Content uploaded by Leo J Schouten
Author content
All content in this area was uploaded by Leo J Schouten
Content may be subject to copyright.
Barrett's esophagus and esophageal and
gastric cancer subtypes:
an epidemiologic perspective
Thesis Jessie Steevens_v04.pdf
© Jessie Steevens, Maastricht 2010
Layout: Tiny Wouters
Cover design: Bram Steevens
Production: Datawyse, Universitaire Pers Maastricht
ISBN: 978-90-5278-994-1
This Ph.D. research was supported by the Dutch Cancer Society (KWF
Kankerbestrijding). The studies presented in this thesis were conducted at the
Department of Epidemiology, GROW - School for Oncology and Developmental
Biology, Maastricht University Medical Centre +.
Financial support for the printing of this thesis was kindly provided by the Department
of Epidemiology and the Dutch Cancer Society.
Thesis Jessie Steevens_v04.pdf
Barrett's esophagus and esophageal and
gastric cancer subtypes:
an epidemiologic perspective
Proefschrift
ter verkrijging van de graad van doctor aan de Universiteit Maastricht
op gezag van de Rector Magnificus, Prof. mr. G.P.M.F Mols,
volgens het besluit van het college van Decanen,
in het openbaar te verdedigen
op woensdag 1 december 2010 om 12:00 uur
door
Jessie Steevens
Thesis Jessie Steevens_v04.pdf
UNIVERSITAIRE
PERS MAASTRICHT
U
P
M
Promotor
Prof. dr. ir. P.A. van den Brandt
Copromotores
Dr. L.J. Schouten
Dr. ir. R.A. Goldbohm, TNO Kwaliteit van Leven, Leiden
Beoordelingscommissie
Prof. dr. F.T. Bosman (voorzitter)
Prof. dr. M.A. van Baak
Prof. dr. L.J. Murray, Queen's University Belfast, Belfast, Northern Ireland
Dr. J. de Nooijer
Prof. dr. P.D. Siersema, Universitair Medisch Centrum Utrecht, Utrecht
Thesis Jessie Steevens_v04.pdf
CONTENTS
Chapter 1 General introduction 7
Chapter 2 A prospective cohort study on overweight, smoking, 23
alcohol consumption, and risk of Barrett’s esophagus
Submitted
Chapter 3 Toenail selenium status and the risk of Barrett’s esophagus: 49
the Netherlands Cohort Study
Cancer Causes Control 2010; in press
Chapter 4 Cancer incidence and cause-specific mortality in a 67
population-based cohort of patients with Barrett’s esophagus
Submitted
Chapter 5 Trends in incidence of esophageal and stomach cancer 83
subtypes in Europe
Eur J Gastroenterol Hepatol 2010;22:669-678
Chapter 6 Alcohol consumption, cigarette smoking and risk of subtypes 105
of esophageal and gastric cancer: a prospective cohort study
Gut 2010;59:39-48
Chapter 7 Selenium status and the risk of esophageal and gastric cancer 131
subtypes: the Netherlands Cohort Study
Gastroenterology 2010;138:1704-1713
Chapter 8 Vegetables and fruits consumption and risk of esophageal and 151
gastric cancer subtypes in the Netherlands Cohort Study
Submitted
Chapter 9 General discussion 179
Summary / Samenvatting 193
Dankwoord 201
Curriculum Vitae / List of publications 205
Thesis Jessie Steevens_v04.pdf
Thesis Jessie Steevens_v04.pdf
General introduction
1
7
Thesis Jessie Steevens_v04.pdf
Chapter 1
8
Thesis Jessie Steevens_v04.pdf
Introduction
9
Cancer has become a disease that affects many people in the world. The International
Agency for Research on Cancer estimated that 12.7 million new cancer cases and 7.6
million cancer deaths occurred in 2008.1 In the Netherlands, cancer has become the
most important cause of death in 2008.2 In that year, over 40,000 persons died of
cancer, which was approximately 30% of all deaths.3
This thesis is concerns two types of cancer: cancer of the esophagus and cancer of
the stomach, also called gastric cancer. Within esophageal and gastric cancer one can
discern subtypes. Besides these cancers, this thesis also examines Barrett’s esophagus
(BE), a precursor of adenocarcinoma of the esophagus.
ESOPHAGEAL AND GASTRIC CANCER
Incidence and mortality
Esophageal cancer was estimated to be the 8th most common cancer worldwide with
regard to new cases, and the 6th most common cause of death from cancer in 2008.
Cancer of the stomach is more common being the 4th cancer with regard to new cases,
and the 2nd leading cause of cancer death in the world.1
Geographically, esophageal and gastric cancer are not distributed evenly (Figures
1.1 and 1.2). Esophageal cancer is specifically common in Asia (Iran, China, Japan,
Korea), Southern and Eastern Africa (Uganda, Zimbabwe), USA (Blacks), and France.
Typically, in these high risk populations the majority of esophageal cancers is of the
squamous cell type, in contrast to low risk populations, where adenocarcinomas are
more common.4 Incidence rates for gastric cancer are highest in Eastern Asia (China,
Japan, Korea), Eastern Europe, Italy, Portugal and Latin America.1,5 In the Netherlands,
1876 and 2020 persons were diagnosed with esophageal and gastric cancer,
respectively, and 1558 and 1424 patients died of these cancers in 2008.6
Course of the diseases
Symptoms that indicate the presence of esophageal cancer are dysphagia, weight loss,
and pain on swallowing foods.8 The most common symptoms that precede a diagnosis
of gastric cancer are weight loss, abdominal pain, nausea, and dysphagia.9 Patients are
often not diagnosed with esophageal or gastric cancer until the cancer has reached an
advanced stage. Partly due to this late diagnosis, treatment options are limited and
curative treatments are often not possible.10 The late diagnosis therefore lies at the
bottom of the very high mortality rate of both cancers. One year after diagnosis, 42%
of the esophageal cancer patients and 45% of the gastric cancer patients are still alive.
Only 14% of esophageal and 21% of gastric cancer patients survive five years.6 These
Dutch rates are comparable with European survival rates.11
Thesis Jessie Steevens_v04.pdf
Chapter 1
10
Figure 1.1 Estimated worldwide incidence rates (per 100,000) of esophageal cancer (both sexes, all ages),
age-standardized to the world population. Source: Globocan 2008.7
Figure 1.2 Estimated worldwide incidence rates (per 100,000) of gastric cancer (both sexes, all ages), age-
standardized to the world population. Source: Globocan 2008.7
Thesis Jessie Steevens_v04.pdf
Introduction
11
Classification of esophageal and gastric cancers
Cancers of the esophagus and stomach can be classified in several ways. In the
International Classification of Diseases for Oncology (ICD-O),12 tumors are classified
according to their topography and histology.
For the esophagus, there are two possible topographic classifications: cervical,
thoracic, and abdominal part of the esophagus, or upper, middle, and lower third of
the esophagus. The latter classification is most commonly used in research, and shown
in Figure 1.3. Another classification is based on the type of tissue from which the
tumor originates: a histological classification. The two most common histological types
of esophageal cancer are esophageal squamous cell carcinoma and esophageal
adenocarcinoma. Esophageal squamous cell carcinoma most often occurs in the
middle third of the esophagus, while the large majority of esophageal adeno-
carcinomas arise in the lower third.13
The ICD-O topographic classification of gastric cancer discerns tumors of the
cardia, fundus, body, pyloric antrum, and pylorus (Figure 1.3). Additionally, tumors can
be classified as located in the lesser or greater curvature of the stomach.12 The large
majority of all gastric cancers are adenocarcinomas.13
Figure 1.3 The esophagus and stomach.
Stomach
Esophagus
Diafragm
Upper third of the
esophagus
Lower third of
the esophagus
Middle third of
the esophagus
Gastric cardia
Fundus
Pyloric antrum
Pylorus
Body
Thesis Jessie Steevens_v04.pdf
Chapter 1
12
Trends in incidence rates
Incidence rates of esophageal and gastric cancer have been carefully monitored for
decades. Before 1940, gastric cancer was the leading cause of cancer death in men in
the USA. In the next decades, however, this cancer became less and less common.
Between 1950 and the 1970s, gastric cancer mortality had started to decline quite
sharply in several countries, including the USA, Scandinavian countries, Western
European countries, and Australia.4,14 Later publications observed, however, that this
decline in gastric cancer did not concern the cardia, the proximal part of the
stomach.15 In fact, some reported that the incidence of gastric cardia cancer had risen
in the 1960s through 1980s in the USA, the UK, Denmark, and Sweden.16-19 Moreover,
these and other studies also observed sharp increases in the incidence of
adenocarcinomas of the esophagus.16-18,20 In contrast, stable or declining incidence
rates have been reported for the other histological type of esophageal cancer:
squamous cell carcinoma.16,17,21-23
The question continually is, however, whether these observed changes in
incidence are true increases in the disease. In the course of time the attention for
subtypes of esophageal and gastric cancer has grown, which may have influenced
clinicians and pathologists. Also, the quality of registration of tumor histology and
subsite has improved greatly.5, 24-28 These developments may have contributed to the
observed changes. In general, it is thought that at least a past of the observed increase
in adenocarcinomas of the esophagus and gastric cardia and decrease in other gastric
adenocarcinomas is real.16,19,20,22,23
As a result of the trends in incidence rates, the ratio of esophageal squamous cell
carcinoma (ESCC) to esophageal adenocarcinoma (EAC) has changed in USA whites
from 83/17 in 1974 to 41/59 in 1995.29 In contrast, in USA blacks the ratio remained
stable at 97/3 during this period.29 In the Netherlands this ratio changed from 55/45 to
34/66 between 1989 and 2003.6 Thus, in the USA and the Netherlands, EAC has
become the most common histological type of esophageal cancer. Conversely, this was
not observed in China, a high risk area of esophageal cancer, where ESCC accounted
for nearly 100% of esophageal cancers, at least in the period 1988-2003.30 The subsite
distribution of gastric cancers has also changed over time as a consequence of the
incidence trends. In the USA, the percentage GCA of all gastric tumors increased from
8% to 15% in black men and from 30% to 47% in white men between 1974 and 1995.29
In contrast, in the Netherlands this percentage remained stable around 25% between
1990 and 2007.31 A relatively high percentage of GCA tumors is seen in China: in 1993
GCA cases accounted for 30% of gastric cancers, and this increased to 37% in 2004.32
Thus, the geographical differences in the percentage of GCA cases are less striking than
the differences in the ratio of ESCC to EAC.
In the future, it remains important to keep monitoring trends in the incidence of
esophageal and gastric cancer subtypes. High quality and detailed data are required for
this monitoring. Interpretation of these trends should be done carefully by taking into
account changes in diagnosis and registration of these cancers. A further aim is to
Thesis Jessie Steevens_v04.pdf
Introduction
13
explain the observed trends and to identify risk factors for these diseases. Primary
prevention is specifically important given the poor prognosis, as well as identification
of high-risk individuals and early detection (secondary prevention) of esophageal and
gastric cancer.
BARRETT’S ESOPHAGUS
Barrett’s esophagus (BE), also referred to as columnar lined esophagus, is a so-called
precancerous lesion for esophageal adenocarcinoma. In persons with BE, the normal
multi-layer squamous epithelium of the esophagus is replaced by a single layer of
columnar epithelium. When a patient is undergoing endoscopy of the esophagus, this
metaplasia is visible. The columnar epithelium has a red color as opposed to the light
pink color of the normal squamous epithelium (Figure 1.4).33 BE is located in the lower
third of the esophagus.
Figure 1.4 Endoscopic picture of Barrett’s esophagus. Arrows indicate columnar epithelium. Source: 39
Definition and diagnosis of the disease
The condition BE was named after Norman Barrett, a surgeon who described it in
1950.34 The definition of BE has been debated and has changed over time.33,35
Currently there are two different accepted case definitions. The American College of
Gastroenterology definition is as follows: “Barrett’s esophagus is a change in the distal
esophageal epithelium of any length that can be recognized as columnar type mucosa
at endoscopy and is confirmed to have intestinal metaplasia by biopsy of the tubular
esophagus.”36 The British Society of Gastroenterology defines Barrett’s esophagus as
“an oesophagus in which any portion of the normal squamous lining has been replaced
Thesis Jessie Steevens_v04.pdf
Chapter 1
14
by a metaplastic columnar epithelium which is visible macroscopically. In order to
make a positive diagnosis of “Barrett’s oesophagus”, a segment of columnar
metaplasia of any length must be visible endoscopically above the oesophago-gastric
junction and confirmed or corroborated histologically.”37,38 Thus, in contrast to the
American definition, the British definition does not require identification of specialized
intestinal metaplasia. The British Society of Gastroenterology reasons that intestinal
metaplasia can always be found, providing a sufficient number of biopsies is taken over
an adequate time-scale.37
Incidence rates
After it was observed that the incidence of esophageal adenocarcinoma was rising in
various countries, researchers hypothesized that this increase was preceded by a
similar rise in the incidence of BE. Substantial increases in diagnosis of BE have indeed
been described in Scotland, the USA, and the Netherlands.40-42
In the Netherlands, the world standardized incidences rates of BE increased from
12.7 to 18.4 per 100,000 men and from 6.5 to 8.6 per 100,000 women between 1992
and 2003. This increase is not explained by an increase in the number of biopsies
taken, or by an increased proportion of BE patients that undergo endoscopy.
Therefore, a real increase in the incidence of BE seems likely.42
Progression and surveillance
As mentioned earlier, BE is a precancerous lesion, or precursor lesion. For three
decades, it has been known that patients diagnosed with BE are at increased risk of
esophageal adenocarcinoma.43 However, the exact size of this risk remains uncertain.
The risk of progression has been estimated by several researchers and the reported
risk estimates are highly variable, depending on the study population, selection of
patients, and follow-up time.44 Large studies with complete follow-up that quantify this
risk are requested. Another question that remains to be answered is whether patients
with BE are maybe also at increased risk of other cancers and other diseases.
The progression of BE can be identified pathologically. The successive steps are BE
without dysplasia, low-grade dysplasia, high-grade dysplasia and finally adeno-
carcinoma (Figure 1.1).45
Studying BE is interesting and important, because it may yield new insights into
the etiology of the disease, which is important for primary prevention. Ultimately,
research should lead to the identification of persons at highest risk of developing
esophageal adenocarcinoma. Through endoscopic surveillance of patients with BE,
esophageal adenocarcinoma may be diagnosed in an earlier stage. This may lead to
improved treatment options and prognosis.
Thesis Jessie Steevens_v04.pdf
Introduction
15
RELEVANCE OF CANCER SUBTYPES AND PRECURSOR LESIONS
So far, we described two indications for a different etiology of esophageal squamous
cell carcinoma (ESCC) vs. esophageal adenocarcinoma (EAC), and gastric cardia
adenocarcinoma (GCA) vs. gastric non-cardia adenocarcinoma (GNCA). These are the
differences in trends in incidence rates and the differences in the geographical
distribution of these cancer subtypes.
A third indication for a different etiology is the observed sex difference in the
occurrence of these cancers. All subtypes of esophageal and gastric cancer occur more
frequently in men than women, at least in most parts of the world.13 But more
importantly, the ratio of the incidence in men to that in women, the male-to-female
ratio, is much higher for EAC (between 6:1 and 8:1),46-48 compared with ESCC (between
2:1 and 3:1).46,49 For BE, the male-to-female ratio is usually around 2:1.50 A different
male-to-female ratio is also observed for the gastric cancer subtypes. The ratio is
between 3:1 and 5:1 for GCA,46,48,51,52 whereas it is around 2:1 for GNCA.18,51
Until recently, relatively little attention was directed towards differences between
the histologic types of esophageal cancer and the majority of existing epidemiologic
studies have not distinguished between the major histologic types of esophageal
cancer. This also applies to the subsite division of gastric cancer.
BE has started to receive more attention since a rise in incidence of EAC has been
observed in several countries. BE is an interesting precursor lesion, because it is very
well visible through endoscopy and can thus be monitored and studied relatively
easily. It thus provides a good model for the investigation of cancer development. One
thing we should keep in mind is that BE is only diagnosed if a person presents with
reflux symptoms or if endoscopy is performed on different medical grounds and BE is
found by chance. Thus, some cases of asymptomatic BE remain undetected. The
estimates of the proportion of asymptomatic individuals that have BE vary widely:
between 1 and 25%.53-55
Precancerous lesions have also been identified for ESCC,56 GCA,57 and GNCA.58
These are, however, not the focus of this thesis.
RISK FACTORS FOR ESOPHAGEAL AND GASTRIC CANCERS AND
BARRETT’S ESOPHAGUS
Demographic factors
As with nearly all cancers, older age is a strong risk factor for esophageal and gastric
cancers,13 as well as for BE.59 As described above, men are at higher risk of all
esophageal and gastric cancer subtypes and BE than women. In the USA, BE and EAC
are more common among Caucasians, while ESCC is more common among blacks.13
Thesis Jessie Steevens_v04.pdf
Chapter 1
16
Medical factors
Several medical conditions and medications have been associated with risk of BE and
esophageal and gastric cancer. The most important risk factor for BE and EAC is gastro-
esophageal reflux disease (GERD),60 which is defined as chronic symptoms or mucosal
damage produced by the abnormal reflux of gastric contents (including acid) into the
esophagus.61 GERD may also increase risk of GCA, although probably less strongly than
for BE and EAC.62 Gastro-esophageal reflux can be increased in the presence of a hiatal
hernia, which means that the upper part of the stomach protrudes into the chest
cavity through a hole in the diafragm.13 Not only reflux of acid can increase risk of
esophageal cancer: ingestion of caustic substances, particularly lye, very strongly
increase the risk of ESCC.63
ESCC risk is also higher in subjects with achalasia and tylosis. Achalasia is an
esophageal motility disorder involving the smooth muscle layer of the esophagus and
the lower esophageal sphincter (LES). Patients with achalasia have difficulty
swallowing and suffer from regurgitation.63 Tylosis is a rare disease associated with
hyperkeratosis of the palms of the hands and soles of the feet. The inherited type of
tylosis (Howell-Evans syndrome) has been most strongly linked to ESCC.63
There is also evidence for a relationship between some medications and risk of BE
or esophageal or gastric cancer. Medications that relax the LES may contribute to
increased risk of BE and EAC via increased chance of gastro-esophageal reflux. These
medications include the commonly used nitroglycerin, aminophyllines, beta-adrenergic
agonists, anticholinergics and benzodiazepines.64, 65 Nonsteroidal anti-inflammatory
drugs (NSAIDs) on the other hand, may be preventive for EAC and maybe also for ESCC
and GNCA,13,66-68 possibly because they stimulate apoptosis and inhibit angiogenesis.66
Infection with Helicobacter (H.) pylori has been identified as the strongest risk
factor for gastric cancer.13 However, this is probably only true for GNCA. In contrast, H.
pylori appears to protect against EAC and GCA.13,47 Infection with Epstein-Barr virus has
also been suggested to play a role in some gastric tumors.13 Other medical conditions
that have been suggested to increase risk of gastric cancer include pernicious anemia,
peptic ulcer disease, and prior gastric surgery.52 Further, persons with a family history
of gastric cancer have been observed to have an increased risk to develop gastric
cancer.13
Lifestyle and dietary habits
Several aspects of lifestyle have been studied in relation to risk of developing
esophageal and gastric cancer. Most studies have not separately studied the cancer
subtypes. Very little information is available on possible relationships between lifestyle
and diet and the risk to develop BE.
Smoking and alcohol consumption have been studied most, and have been found
to increase risk of ESCC. Smoking is also associated with EAC and gastric cancers, but
probably less strongly. The association between alcohol consumption and risk of EAC
Thesis Jessie Steevens_v04.pdf
Introduction
17
and gastric cancers is not entirely clear yet.13 Further, being overweight may increase a
person’s risk of EAC and GCA.13, 69
Higher incidence rates of ESCC and gastric cancer have been observed among the
lower socioeconomic groups. This observation may be explained partly by risk factors
strongly associated with low socioeconomic status (SES), such as infection with H.
pylori and smoking. On the other hand, the association with SES may also reflect yet
unrecognized exposures, as the association between SES and ESCC risk remained after
adjustment for known risk factors.13
Vegetables, fruits, and some vitamins and carotenoids may decrease risk of
esophageal and gastric cancer. Foods containing dietary fiber may decrease risk of
esophageal cancer. On the other hand, there is some indication for a positive
association between processed meat and esophageal and gastric cancer. Esophageal
cancer risk may also be increased by eating red meat and drinking maté and very hot
drinks. Salt, salted foods, chili, smoked foods, and grilled or barbecued animal food
may increase gastric cancer risk.69
RATIONALE AND AIM OF THIS THESIS
As explained in the previous paragraphs, there has been little epidemiologic research
into the risk factors for subtypes of esophageal and gastric cancer. The data on BE is
even more limited. Very few studies have looked into the etiology of this condition.
Besides, most evidence on BE and esophageal and gastric cancer subtypes to date
comes from cross-sectional and case-control studies. These study designs are
susceptible to methodological limitations. The prospective cohort study design is less
susceptible to biases and therefore generally has a higher validity. The research
described in this thesis is carried out within a prospective cohort study: the
Netherlands Cohort Study on diet and cancer (NLCS).
Our study could make an important contribution to the epidemiological body of
evidence on the development of BE and the subtypes of esophageal and gastric cancer.
The cancers of main interest in this thesis are ESCC, EAC, and GCA. GNCA will be
described only for comparison with these cancers. The main focus of this thesis is on
the role of lifestyle and nutritional factors in the etiology of these diseases. These
factors can potentially be modified and can thus have a role in primary prevention.
Additionally, we will use the data of the NLCS to follow-up BE patients for
occurrence of EAC, other cancers and cause-specific mortality, as we indicated that
data are lacking on these topics. The large size and complete follow-up make the NLCS
suitable for these analyses.
Finally, we will also study changes in the incidence of esophageal and gastric
cancers in time, because of the need to keep monitoring trends for above-mentioned
reasons. We will do this with data from European countries, and examine whether
there are differences in trends between countries.
Thesis Jessie Steevens_v04.pdf
Chapter 1
18
STUDY DESIGN
The research described in this thesis is based on a prospective cohort study: the
Netherlands Cohort Study on diet and cancer (NLCS). The NLCS was initiated in
September 1986.70 A total of 120,852 men and women aged 55-69 years participated
in this study. These subjects were randomly selected from Dutch municipal population
registries. At baseline, all subjects filled out a questionnaire on dietary habits, lifestyle,
and other risk factors for cancer. According to the case-cohort approach,71 data are
processed and analyzed for a random sample of the cohort (the subcohort) and cases.
The subcohort consists of 5,000 men and women who were sampled at baseline. This
subcohort is followed-up for vital status and migration and is used to estimate the
person-time at risk for the total cohort. Incident cases of BE and cancer were identified
in the whole cohort. The identification of BE cases took place through linkage with data
from PALGA (the nationwide registry for histopathology and cytopathology in the
Netherlands). The existence of this unique nationwide pathology registry offered us
the opportunity to study BE within a prospective cohort study. A pathologist reviewed
and coded further details and characteristics of the BE cases from the PALGA data. The
cohort was also followed-up for cancer incidence, and this was done through linkage
with PALGA and the Netherlands Cancer Registry.72 Further, the cohort has been
followed-up for vital status through record linkage with the Central Bureau of
Genealogy and automated municipal population registries. For deceased cohort
members we also obtained information on the cause of death from Statistics
Netherlands.
The research described in this thesis was based on 16.3-year follow-up data of the
cohort (September 17, 1986 through December 31, 2002).
OUTLINE OF THE THESIS
This thesis begins with two studies on the etiology of BE. Chapter 2 describes the
associations between the lifestyle factors overweight, smoking, and alcohol
consumption and the risk to develop BE. Our study on selenium status and its
relationship with risk of BE is presented in chapter 3. Chapter 4 focuses on the
question whether cancer incidence and total and cause-specific mortality are increased
in patients with BE. Trends in the incidence of esophageal and gastric cancer subtypes
in European countries are investigated in chapter 5. The next three chapters are about
several possible risk factors of esophageal and gastric cancer subtypes: alcohol
consumption and cigarette smoking (chapter 6); selenium status (chapter 7), and
vegetables and fruits consumption (chapter 8). Finally, chapter 9 concludes the thesis
with a discussion of the main findings and some methodological aspects.
Thesis Jessie Steevens_v04.pdf
Introduction
19
REFERENCES
1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of worldwide burden of cancer in
2008: GLOBOCAN 2008. Int J Cancer 2010.
2. http://www.cbs.nl/en-GB/menu/themas/gezondheid-
welzijn/publicaties/artikelen/archief/2009/2009-2687-wm.htm?Languageswitch=on (accessed on June
22, 2010).
3. http://statline.cbs.nl/StatWeb/publication/?DM=SLEN&PA=7052ENG&D1=7-8&D2=0&D3=a&D4=
l&LA=EN&HDR=G3,G1,G2&STB=T&VW=T (accessed on June 22, 2010).
4. Boyle P, Levin B. World Cancer Report. Lyon, 2008.
5. Curado MP, Edwards B, Shin HR, Storm H, Ferlay J, Heanue M, Boyle P. Cancer incidence in five
continents. Volume IX. Lyon: IARC Scientific publications no. 160, 2007.
6. www.ikcnet.nl (accessed August 08, 2010).
7. http://globocan.iarc.fr (accessed June 23, 2010).
8. Enzinger PC, Mayer RJ. Esophageal cancer. N Engl J Med 2003;349:2241-2252.
9. Wanebo HJ, Kennedy BJ, Chmiel J, Steele G, Jr., Winchester D, Osteen R. Cancer of the stomach. A
patient care study by the American College of Surgeons. Ann Surg 1993;218:583-592.
10. Bird-Lieberman EL, Fitzgerald RC. Early diagnosis of oesophageal cancer. Br J Cancer 2009;101:1-6.
11. Berrino F, De Angelis R, Sant M, Rosso S, Bielska-Lasota M, Coebergh JW, Santaquilani M. Survival for
eight major cancers and all cancers combined for European adults diagnosed in 1995-99: results of the
EUROCARE-4 study. Lancet Oncol 2007;8:773-783.
12. http://apps.who.int/classifications/apps/icd/icd10online/index.htm?gc15.htm+ (accessed July 30,
2010).
13. Schottenfeld D, Fraumeni JF, Jr. Cancer epidemiology and prevention. Oxford University Press, 2006.
14. Howson CP, Hiyama T, Wynder EL. The decline in gastric cancer: epidemiology of an unplanned
triumph. Epidemiol Rev 1986;8:1-27.
15. Levi F, La Vecchia C, Te VC. Descriptive epidemiology of adenocarcinomas of the cardia and distal
stomach in the Swiss Canton of Vaud. Tumori 1990;76:167-171.
16. Blot WJ, Devesa SS, Kneller RW, Fraumeni JF, Jr. Rising incidence of adenocarcinoma of the esophagus
and gastric cardia. Jama 1991;265:1287-1289.
17. Powell J, McConkey CC. The rising trend in oesophageal adenocarcinoma and gastric cardia. Eur J
Cancer Prev 1992;1:265-269.
18. Moller H. Incidence of cancer of oesophagus, cardia and stomach in Denmark. Eur J Cancer Prev
1992;1:159-164.
19. Hansson LE, Sparen P, Nyren O. Increasing incidence of carcinoma of the gastric cardia in Sweden from
1970 to 1985. Br J Surg 1993;80:374-377.
20. Hesketh PJ, Clapp RW, Doos WG, Spechler SJ. The increasing frequency of adenocarcinoma of the
esophagus. Cancer 1989;64:526-530.
21. Blot WJ, Devesa SS, Fraumeni JF, Jr. Continuing climb in rates of esophageal adenocarcinoma: an
update. Jama 1993;270:1320.
22. Zheng T, Mayne ST, Holford TR, Boyle P, Liu W, Chen Y, Mador M, Flannery J. Time trend and age-
period-cohort effects on incidence of esophageal cancer in Connecticut, 1935-89. Cancer Causes
Control 1992;3:481-492.
23. Harrison SL, Goldacre MJ, Seagroatt V. Trends in registered incidence of oesophageal and stomach
cancer in the Oxford region, 1974-88. Eur J Cancer Prev 1992;1:271-274.
24. Waterhouse J, Muir C, Shanmugaratnam K, Powell J. Cancer incidence in five continents. Volume IV.
Lyon: IARC Scientific publication no. 42, 1982.
25. Muir C, Waterhouse J, Mack T, Powell J, Whelan S. Cancer incidence in five continents. Volume V.
Lyon: IARC Scientific publication no. 88, 1987.
26. Ferlay J, Muir CS, Whelan SL, Gao Y-T. Cancer incidence in five continents. Volume VI. Lyon: IARC
Scientific publications no. 120, 1992.
27. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J. Cancer incidence in five continents. Volume VII.
Lyon: IARC Scientific publications no. 143, 1997.
Thesis Jessie Steevens_v04.pdf
Chapter 1
20
28. Parkin DM, Whelan SL, Ferlay J, Teppo L, Thomas DB. Cancer incidence in five continents. Volume VIII.
Lyon: IARC Scientific publications no. 155, 2002.
29. Devesa SS, Blot WJ, Fraumeni JF, Jr. Changing patterns in the incidence of esophageal and gastric
carcinoma in the United States. Cancer 1998;83:2049-2053.
30. He YT, Hou J, Chen ZF, Qiao CY, Song GH, Meng FS, Jin HX, Chen C. Trends in incidence of esophageal
and gastric cardia cancer in high-risk areas in China. Eur J Cancer Prev 2008;17:71-76.
31. Dassen AE, Lemmens VE, van de Poll-Franse LV, Creemers GJ, Brenninkmeijer SJ, Lips DJ, Vd Wurff AA,
Bosscha K, Coebergh JW. Trends in incidence, treatment and survival of gastric adenocarcinoma
between 1990 and 2007: a population-based study in the Netherlands. Eur J Cancer 2010;46:1101-
1110.
32. Zhou Y, Zhang Z, Zhang Z, Wu J, Ren D, Yan X, Wang Q, Wang Y, Wang H, Zhang J, Zhu X, Yang Y, Luo C,
Guo X, Tang C, Qiao L. A rising trend of gastric cardia cancer in Gansu Province of China. Cancer Lett
2008;269:18-25.
33. Spechler SJ. Clinical practice. Barrett's Esophagus. N Engl J Med 2002;346:836-842.
34. Barrett NR. Chronic peptic ulcer of the oesophagus and 'oesophagitis'. Br J Surg 1950;38:175-182.
35. Cameron AJ. The history of Barrett esophagus. Mayo Clin Proc 2001;76:94-96.
36. Wang KK, Sampliner RE. Updated guidelines 2008 for the diagnosis, surveillance and therapy of
Barrett's esophagus. Am J Gastroenterol 2008;103:788-797.
37. Watson A, Shepherd NA. The definition of "Barrett's" columnar-lined esophagus. British Society of
Gastroenterology. Guidelines for the diagnosis and management of Barrett's columnar-lined
esophagus. Loughborough: Q3 print, 2005:4-6.
38. Playford RJ. New British Society of Gastroenterology (BSG) guidelines for the diagnosis and
management of Barrett's oesophagus. Gut 2006;55:442-443.
39. http://www.amc.edu/patient/services/BARRX/barretts_esophagus_definition.html (accessed June 24,
2010).
40. Prach AT, MacDonald TA, Hopwood DA, Johnston DA. Increasing incidence of Barrett's oesophagus:
education, enthusiasm, or epidemiology? Lancet 1997;350:933.
41. Conio M, Cameron AJ, Romero Y, Branch CD, Schleck CD, Burgart LJ, Zinsmeister AR, Melton LJ, 3rd,
Locke GR, 3rd. Secular trends in the epidemiology and outcome of Barrett's oesophagus in Olmsted
County, Minnesota. Gut 2001;48:304-309.
42. Post PN, Siersema PD, van Dekken H. Rising incidence of clinically evident Barrett's oesophagus in The
Netherlands: a nation-wide registry of pathology reports. Scand J Gastroenterol 2007;42:17-22.
43. Naef AP, Savary M, Ozzello L. Columnar-lined lower esophagus: an acquired lesion with malignant
predisposition. Report on 140 cases of Barrett's esophagus with 12 adenocarcinomas. J Thorac
Cardiovasc Surg 1975;70:826-835.
44. Yousef F, Cardwell C, Cantwell MM, Galway K, Johnston BT, Murray L. The incidence of esophageal
cancer and high-grade dysplasia in Barrett's esophagus: a systematic review and meta-analysis. Am J
Epidemiol 2008;168:237-249.
45. Flejou JF. Barrett's oesophagus: from metaplasia to dysplasia and cancer. Gut 2005;54 Suppl 1:i6-12.
46. van Blankenstein M, Looman CW, Hop WC, Bytzer P. The incidence of adenocarcinoma and squamous
cell carcinoma of the esophagus: Barrett's esophagus makes a difference. Am J Gastroenterol
2005;100:766-774.
47. Lagergren J. Adenocarcinoma of oesophagus: what exactly is the size of the problem and who is at
risk? Gut 2005;54 Suppl 1:i1-5.
48. El-Serag HB, Mason AC, Petersen N, Key CR. Epidemiological differences between adenocarcinoma of
the oesophagus and adenocarcinoma of the gastric cardia in the USA. Gut 2002;50:368-372.
49. Cook MB, Chow WH, Devesa SS. Oesophageal cancer incidence in the United States by race, sex, and
histologic type, 1977-2005. Br J Cancer 2009;101:855-859.
50. Cook MB, Wild CP, Forman D. A systematic review and meta-analysis of the sex ratio for Barrett's
esophagus, erosive reflux disease, and nonerosive reflux disease. Am J Epidemiol 2005;162:1050-1061.
51. Wu H, Rusiecki JA, Zhu K, Potter J, Devesa SS. Stomach carcinoma incidence patterns in the United
States by histologic type and anatomic site. Cancer Epidemiol Biomarkers Prev 2009;18:1945-1952.
52. Forman D, Burley VJ. Gastric cancer: global pattern of the disease and an overview of environmental
risk factors. Best Pract Res Clin Gastroenterol 2006;20:633-649.
Thesis Jessie Steevens_v04.pdf
Introduction
21
53. Ward EM, Wolfsen HC, Achem SR, Loeb DS, Krishna M, Hemminger LL, DeVault KR. Barrett's esophagus
is common in older men and women undergoing screening colonoscopy regardless of reflux
symptoms. Am J Gastroenterol 2006;101:12-17.
54. Gerson LB, Shetler K, Triadafilopoulos G. Prevalence of Barrett's esophagus in asymptomatic
individuals. Gastroenterology 2002;123:461-467.
55. Fan X, Snyder N. Prevalence of Barrett's esophagus in patients with or without GERD symptoms: role
of race, age, and gender. Dig Dis Sci 2009;54:572-577.
56. Wang GQ, Abnet CC, Shen Q, Lewin KJ, Sun XD, Roth MJ, Qiao YL, Mark SD, Dong ZW, Taylor PR,
Dawsey SM. Histological precursors of oesophageal squamous cell carcinoma: results from a 13 year
prospective follow up study in a high risk population. Gut 2005;54:187-192.
57. El-Serag HB, Sonnenberg A, Jamal MM, Kunkel D, Crooks L, Feddersen RM. Characteristics of intestinal
metaplasia in the gastric cardia. Am J Gastroenterol 1999;94:622-627.
58. Correa P. A human model of gastric carcinogenesis. Cancer Res 1988;48:3554-3560.
59. Wong A, Fitzgerald RC. Epidemiologic risk factors for Barrett's esophagus and associated
adenocarcinoma. Clin Gastroenterol Hepatol 2005;3:1-10.
60. Shaheen N, Ransohoff DF. Gastroesophageal reflux, barrett esophagus, and esophageal cancer:
scientific review. Jama 2002;287:1972-1981.
61. DeVault KR, Castell DO. Updated guidelines for the diagnosis and treatment of gastroesophageal reflux
disease. Am J Gastroenterol 2005;100:190-200.
62. Lagergren J, Bergstrom R, Lindgren A, Nyren O. Symptomatic gastroesophageal reflux as a risk factor
for esophageal adenocarcinoma. N Engl J Med 1999;340:825-831.
63. http://www.uptodateonline.com/online/content/search.do (accessed June 26, 2010).
64. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
65. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atcddd/
welcome.html (accessed 15 Oct, 2009).
66. Epplein M, Nomura AM, Wilkens LR, Henderson BE, Kolonel LN. Nonsteroidal Antiinflammatory Drugs
and Risk of Gastric Adenocarcinoma: The Multiethnic Cohort Study. Am J Epidemiol 2009.
67. Sadeghi S, Bain CJ, Pandeya N, Webb PM, Green AC, Whiteman DC. Aspirin, nonsteroidal anti-
inflammatory drugs, and the risks of cancers of the esophagus. Cancer Epidemiol Biomarkers Prev
2008;17:1169-1178.
68. Duan L, Wu AH, Sullivan-Halley J, Bernstein L. Nonsteroidal Anti-inflammatory Drugs and Risk of
Esophageal and Gastric Adenocarcinomas in Los Angeles County. Cancer Epidemiol Biomarkers Prev
2008;17:126-134.
69. World Cancer Research Fund, American Institute for Cancer Research. Food, nutrition, physical activity
and the prevention of cancer: a global perspective. AICR, 2007.
70. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
71. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
72. van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol
1990;19:553-558.
Thesis Jessie Steevens_v04.pdf
Chapter 1
22
Thesis Jessie Steevens_v04.pdf
23
Thesis Jessie Steevens_v04.pdf
A prospective cohort study on
overweight, smoking, alcohol
consumption, and risk of Barrett's
esophagus
Jessie Steevens
Leo J Schouten
Ann LC Driessen
Clément JR Huysentruyt
Yolande CA Keulemans
R Alexandra Goldbohm
Piet A van den Brandt
Submied
2
Chapter 2
24
ABSTRACT
Background
Barrett’s esophagus (BE) is a precursor lesion of esophageal adenocarcinoma. Besides
gastroesophageal reflux, possible risk factors for BE include overweight, cigarette
smoking, and alcohol consumption. Our objective was to study these associations
using prospective data.
Methods
The prospective Netherlands Cohort Study, initiated in 1986, consists of 120,852 men
and women, aged 55-69 years at baseline. At baseline, all subjects completed a
questionnaire on dietary habits and lifestyle. After 16.3 years of follow-up, 370 BE
cases with specialized intestinal metaplasia and 3866 subcohort members were
available for case-cohort analysis. Cox proportional hazards models were used to
calculate incidence rate ratios (RR) and 95% confidence intervals (CI).
Results
Body mass index at baseline was associated with risk of BE in women [multivariable
adjusted RR per 1 kg/m2, 1.07 (1.03-1.11)] but not in men [RR per 1 kg/m2, 0.99
(0.93-1.05)]. The association in women was not specifically due to abdominal
overweight. Former cigarette smokers were at increased risk of BE (RR=1.33, 95% CI
1.00-1.77), but current smokers were not. Smoking duration showed a positive
association with BE risk (p trend=0.03). For alcohol consumption, the RR per 10 grams
ethanol/day was 0.95 (0.87-1.03).
Conclusions
Increased body mass index was a risk factor for BE in women, but not in men. Several
aspects of cigarette smoking were positively associated with BE risk. Alcohol
consumption was not associated with an increased risk of BE.
Impact
Future research should focus on risk factors for development as well as for progression
of BE to esophageal adenocarcinoma.
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
25
INTRODUCTION
Barrett’s esophagus (BE) is a condition of the distal esophagus characterized by a
replacement of the normal stratified squamous epithelium with a single layer of
columnar epithelium. Histological confirmation based on biopsies is required for
diagnosis. According to the USA definition of BE the presence of goblet cells, indicating
specialized intestinal metaplasia (SIM), is required.1 The UK definition is less stringent:
the presence of any type of metaplasia is sufficient for a diagnosis of BE.2 Both the USA
and UK definitions may have been used in the Netherlands.
BE is primarily of interest because it is associated with an increased risk of
esophageal adenocarcinoma, a cancer with high mortality. The estimates of this risk of
esophageal adenocarcinoma vary between 4.1 and 6.1 per 1000 person-years.3
The most recognized and strong risk factor for BE is chronic gastroesophageal
reflux.4 Additionally, overweight, cigarette smoking, and alcohol consumption might be
associated with risk of BE. These factors are interesting as they are potentially
modifiable. However, the body of evidence on the role of these factors in the etiology
of BE is sparse. Especially data from prospective cohort studies is lacking.
The possible role of overweight in the etiology of BE has been investigated in
several epidemiological studies. Null associations5,6 as well as positive associations7-9
have been found in case-control studies. The only prospective cohort study, which was
conducted among women, reported a positive association.10 The relationship between
cigarette smoking and risk of BE has been studied in some case-control studies,6,8,11,12
while no cohort studies have reported on this relationship. Some observed a positive
association,6,8 while other studies did not observe an association.5,12 Alcohol
consumption and its possible relationship with BE has been studied in a few case-
control studies. Null-associations with alcohol have been observed,11,13 and one study14
found alcohol to be a risk factor. Inverse associations have been reported for wine
consumption.11,13
The aim of this study was to investigate the associations between overweight,
smoking, and alcohol consumption and risk of BE, within the prospective Netherlands
Cohort Study (NLCS) on diet and cancer.
METHODS
Study design and subjects
The NLCS was started in September 1986. Cohort members were selected at random
from 204 Dutch municipal registries, and 58 279 men and 63 573 women aged 55-69
were enrolled in the study. All members of the cohort completed a self-administered
questionnaire. A detailed study design has previously been published.15
Thesis Jessie Steevens_v04.pdf
Chapter 2
26
For efficiency, we used a case-cohort approach16 for data processing and analysis.
Cases were derived from the entire cohort, while the number of person-years at risk in
the entire cohort was estimated from a subcohort of 5000 subjects. This subcohort
was selected at random from the full cohort at baseline.
The subcohort was followed-up for vital status, first actively by biennially
contacting the subcohort members, and later by linkage to the Dutch municipal
population registries. After 16.3 years (September 1986 to December 2002), only one
male subcohort member was lost to follow-up. We excluded subcohort members who
reported having prevalent BE or cancer (other than skin cancer) (n=230) at baseline
(Figure 2.1).
The Medical Ethics Committee of Maastricht University, the Netherlands, has
approved the study.
Follow-up
Incident BE cases in the total cohort were detected by computerized record linkage to
the nationwide network and registry of histopathology and cytopathology in the
Netherlands (PALGA).17 This network was founded in 1971, and an increasing number
of laboratories joined PALGA such that it covered all 204 municipalities from which the
cohort members were sampled since 1991. At each of the 64 pathology laboratories in
the Netherlands, summaries of all pathology reports are generated automatically. Each
summary contains a so-called PALGA diagnosis that describes topography,
morphology, function, procedure, and disease. These excerpts are transferred to a
central databank.17 The linkage with PALGA was carried out for 16.3 years of follow-up.
Subsequently, one pathologist (A.D.) and one pathologist in training (C.H.), who
were blinded to the exposure status of the cases, reviewed the excerpts of all
pathology records. Excluded were cases with an uncertain diagnosis of BE (n=106) and
cases that had prevalent cancer or BE at baseline (n=76). Additionally, we excluded
cases with a diagnosis of esophageal or gastric cancer before or less than a half year
after the diagnosis of BE (n=58), and cases with a diagnosis specifying the presence of
only non-intestinal type metaplasia (n=108) (Figure 2.1).
Two definitions of BE were used: our primary case definition included only
subjects with esophageal SIM (n=456). The secondary case definition included subjects
(a) fulfilling the primary case definition or (b) with a pathology report stating
‘Barrett’s’, without a description of the type of metaplasia (total n=626).
Exposure data
All cohort members completed a self-administered questionnaire at baseline. This
questionnaire contained a 150-item food frequency questionnaire (FFQ), including
detailed questions on alcohol consumption, and questions on various other cancer risk
factors.
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
27
Overweight was measured by several variables: BMI at baseline, BMI at age 20
years, BMI change, and pant/skirt size (as a proxy for waist circumference18). BMI at
baseline and BMI at age 20 years were calculated using weight at baseline and weight
at age 20 years, respectively, divided by height at baseline squared (kg/m2). Subjects
with missing values for BMI at baseline were excluded from all analyses. Subjects with
a BMI under 18.5 were excluded, because there were no cases in this category. BMI
change since age 20 years was calculated as BMI at baseline minus BMI at age 20
years. Recently, the use of self-reported pant/skirt size as a proxy measure for waist
circumference was validated in the NLCS.18
Questions were asked about the following aspects of cigarette smoking: whether
the subject was a smoker at baseline, age at smoking initiation, age at smoking
cessation, the number of cigarettes smoked daily and the number of smoking years
(excluding stopping periods). Based on these questions, the following variables were
constructed: smoking status (never/former/current), current smoking (yes/no),
frequency (n cigarettes/day), duration (n years), pack-years of cigarette smoking (n)
and time since cessation (years).
The habitual consumption of alcohol during the year preceding the start of the
study was measured by six items in the questionnaire: (a) beer, (b) red wine, (c) white
wine, (d) sherry and other fortified wines, (e) liquor types containing on average 16%
alcohol and (f) (Dutch) gin, brandy and whiskey. Questions were asked about the
frequency of consumption and the number of glasses consumed on each drinking
occasion. For analysis, we combined (b), (c) and (d) into ‘wine’, and (e) and (f) into
‘liquor’. Mean daily alcohol consumption was calculated using the Dutch food-
composition table.19 For ‘beer’ and ‘other alcoholic beverages’, participants could
indicate whether five years ago, they drunk (a) more than, (b) equal amounts of or (c)
less than today. The fourth answering option was (d) ‘I never use this’. Using these
questions, we selected participants with a stable alcohol consumption to perform a
sensitivity analysis.
The FFQ has been validated against a 9-day diet record, and the Spearman
correlation coefficient between the alcohol intake assessed by the questionnaire and
that estimated by the diet record was 0.89 for all subjects and 0.85 for users of
alcoholic beverages.20 The reproducibility of the FFQ was established and the test-
retest correlation was 0.90 for alcohol intake, and this correlation declined only 0.01-
0.02 per year.21 This indicates that the single FFQ measurement was able to rank
subjects according to alcohol intake and this ability dropped only slightly over time.
The single FFQ measurement that is used in our cohort study can characterize dietary
habits for a period of at least 5 years.21
Questionnaire data were key-entered and processed in a standardized manner,
blinded with respect to case/subcohort status in order to minimize observer bias in
coding and data interpretation.
Thesis Jessie Steevens_v04.pdf
Chapter 2
28
Netherlands Cohort Study on diet and cancer (58 279 men and 62 573 women)
↓ ↓
Subcohort randomly drawn
from total cohort Record linkage with PALGA until 31-12-2002
↓ ↓
1954 reports from 974 cases
Review by pathologist: exclusion of uncertain diagnoses
↓
5000 868 cases
Exclusion of prevalent cancer and BE cases at baseline
↓ ↓
792 cases
Exclusion of cases with esophageal or gastric cancer before or <½ year
after Barrett's diagnosis, and cases with only non-intestinal metaplasia
↓ ↓
4770
Subcohort members 456 Barrett's esophagus cases
with SIM
626 Barrett's esophagus cases with
SIM or unknown metaplasia
Exclusion if incomplete or inconsistent dietary data
↓ ↓ ↓
4434 425 582
Exclusion of first two years of follow-up
↓ ↓ ↓
4348 415 565
Exclusion if missing data on confounders
↓ ↓ ↓
3866 370 500
Figure 2.1 Flow diagram of subcohort members and Barrett’s esophagus cases on whom the analyses
were based. PALGA, nationwide network and registry of histo- and cytopathology in the
Netherlands; SIM, specialized intestinal metaplasia.
Statistical analysis
To evaluate the potential influence of prediagnostic BE at baseline on smoking and
alcohol consumption habits and BMI, BE cases were categorized according to the year
of follow-up in which they were diagnosed. A t-test was used to compare the
differences in mean levels of these exposures between early (0-2 y) and late (2-16.3 y)
follow-up. For the t-test, the exposures were ln (natural logarithm) transformed to
normalize the distributions. The average daily alcohol consumption among early cases
(2.8 g ethanol) was statistically significantly lower when compared with late cases (10.2
g ethanol, figures for cases with SIM). Therefore, we decided to exclude these early
cases from all analyses, to prevent bias. Additionally, subcohort members and cases
with inconsistent or incomplete dietary questionnaire data20 and those with missing
data on the confounders were excluded. Complete data were available for 3866
subcohort members, 370 BE cases fulfilling the primary case definition, and 500 BE
cases fulfilling the secondary case definition (Figure 2.1).
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
29
The multivariable regression models included the following variables: age, sex,
BMI (kg/m2), cigarette smoking (current, frequency and duration), and alcohol
consumption (g/day). Exact model specifications can be found in the table footnotes.
The following variables were potential confounders, but were not included in the
models because they did not change the incidence rate ratio (RR) by >5%: highest level
of education, family history of esophageal or gastric cancer, reported long-term
(>0.5 y) use of nonsteroidal anti-inflammatory drugs (NSAIDs) or aspirin, or lower
esophageal sphincter relaxing medication,22,23 non-occupational physical activity, daily
intakes of vegetables and fruit.
Multivariable adjusted incidence rate ratios (RR) and corresponding 95%
confidence intervals (CI) were estimated using Cox proportional hazards models.24 The
Stata 9.2 statistical software package (StataCorp, College Station, Texas, USA) was used
for analysis. Standard errors were estimated using the robust Huber-White sandwich
estimator to account for additional variance introduced by sampling from the cohort.
This method is equivalent to the variance-covariance estimator by Barlow.25 We tested
the proportional hazards assumption using the scaled Schoenfeld residuals.26 Tests for
dose-response trends were assessed by fitting ordinal exposure variables as
continuous terms. Two-sided p values are reported throughout the article. Interaction
with sex was assessed by including a cross-product term in the model. If the
interaction was statistically significant, we only presented results stratified by sex.
RESULTS
Characteristics
With respect to the characteristics mentioned in Table 2.1, there were no important
differences between the BE cases fulfilling the primary and secondary case definitions.
The most marked differences between subcohort members and cases are the
following. Cases were more likely to be men and former smokers than subcohort
members, while cases had a slightly lower ethanol intake. The BMI at baseline and the
change in BMI from age 20 were somewhat higher among cases than among the
subcohort. Furthermore, cases were somewhat less physically active and consumed
somewhat less fruit and vegetables than the subcohort. NSAIDs and aspirin, and lower
esophageal sphincter relaxing medication were more likely used by cases than
subcohort members.
Cox regression results
All results from Cox regression analyses were very similar for BE cases defined by
primary and secondary case definitions. We therefore only showed the results based
on the BE cases that met the primary case definition. Likewise, only multivariable
Thesis Jessie Steevens_v04.pdf
Chapter 2
30
adjusted results are shown, as these were very similar to the age-adjusted results (see
Supplemental Tables 2.1, 2.2, 2.3).
Table 2.1 Characteristics of cases and subcohort members in the Netherlands Cohort Study (1986-2002).
Characteristic Barrett's esophagus cases
Subcohort
(n=3866)a
SIM
(n=370)a
SIM or unknown
metaplasia
(n=500)a
Mean (SD)b Mean (SD)b Mean (SD)b
Age at baseline (years) 61.3 (4.2) 61.1 (4.3) 61.3 (4.2)
Men (%) 49.0 57.3 56.2
Cigarette smoking status
Never smoker (%) 37 33 33
Former smoker (%) 36 46 44
Current smoker (%) 37 21 23
Ever cigarette smokers:
Frequency of cigarette smoking (n/day) 15.3 (10.3) 16.0 (10.8) 16.1 (10.4)
Duration of cigarette smoking (years) 31.5 (12.2) 30.8 (11.8) 30.9 (12.0)
Pack-years of cigarette smoking (n) 22.6 (17.8) 22.6 (16.7) 22.9 (16.8)
Abstainer from alcohol (%) 23 20 21
Alcohol consumers:
Ethanol intake (g/day) 13.5 (15.0) 12.9 (15.0) 13.1 (14.9)
Beer intake (glasses/day) 0.3 (0.8) 0.3 (0.7) 0.3 (0.7)
Wine intake (glasses/day) 0.5 (0.8) 0.5 (0.7) 0.5 (0.8)
Liquor intake (glasses/day) 0.5 (0.8) 0.5 (0.8) 0.5 (0.8)
BMI at baseline (kg/m2) 25.1 (3.0) 25.4 (2.8) 25.5 (2.8)
BMI at age 20 years (kg/m²) 21.5 (2.6) 21.6 (2.5) 21.6 (2.6)
BMI change from age 20 to baseline (kg/m²) 3.6 (3.3) 3.8 (3.0) 3.8 (3.1)
Non-occupational physical activity (min/day) 74 (61) 69 (53) 70 (55)
Fruit consumption (g/day) 178 (120) 171 (118) 172 (116)
Vegetable consumption (g/day) 195 (82) 189 (77) 184 (76)
Highest level of education
Primary (%) 28 28 31
Lower vocational (%) 22 20 20
Secondary and medium vocational (%) 36 36 34
University and higher vocational (%) 14 16 15
Family history of esophageal or gastric cancer (%) 7 7 7
Use of NSAIDs and aspirin (%)c 7 10 10
Use of LES relaxing medication (%)c 14 17 18
a Presented are the number of subcohort members or cases with complete data on age, sex, cigarette
smoking (current yes/no, number of cigarettes smoked daily, number of smoking years), alcohol
consumption and body mass index. Subcohort members and cases with incomplete or inconsistent
questionnaire data are excluded, as well as the first two years of follow-up; b For categorical variables a
percentage is presented; c Self reported use during more than 0.5 year; BMI, body mass index; IQR,
interquartile range; LES, lower esophageal sphincter; NSAIDs, non-steroidal anti-inflammatory drugs; SD,
standard deviation; SIM, specialized intestinal metaplasia.
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
31
Overweight
Women who were overweight (BMI: 25-<30) or obese (BMI: ≥30) at baseline were at
increased risk of BE. Compared with normal weight (BMI 18.5-<25) women, the RRs
were 1.73 (95% CI 1.22-2.44) for overweight and 1.67 (95% CI 0.96-2.90) for obese
women. The results for men were significantly different from those for women (p
interaction=0.01). Men who were overweight or obese at baseline were not at
increased risk of BE: RRs were 0.85 (95% CI 0.63-1.14) and 1.21 (95% CI 0.60-2.47),
respectively (Table 2.2).
An interaction with sex was also found for BMI at age 20 (p interaction=0.01). A
positive association was found for women (RR for BMI ≥25 at age 20 y=1.56, 95% CI
0.79-3.10), p trend=0.05], while an inverse association was found for men (RR for BMI
≥25 at age 20 y=0.37, 95% CI 0.15-0.87, p trend=0.02). A change in BMI during
adulthood was not associated with risk of BE in men, but a positive trend was observed
in women (p trend=0.01), mainly because women who lost weight during adulthood
were at decreased risk (RR=0.33, 95% CI 0.15-0.75) when compared with those who
gained 0-<4 BMI points.
When RRs for BMI at baseline were additionally adjusted for pant/skirt size, the
associations for women were somewhat attenuated, and for men the associations
were similar when compared with analyses without adjustment for pant/skirt size.
Analyses with pant/skirt size as independent variable showed a positive association in
women (p trend=0.01), which disappeared after adjustment for BMI at baseline. For
men, no association was observed for pant/skirt size (p trend=0.73), and adjustment
for BMI at baseline did not change this observation (Table 2.2).
Cigarette smoking
In the multivariable regression analyses on cigarette smoking (Table 2.3), we did not
observe any statistically significant differences between associations for men and
women, which permitted us to report the results for both sexes combined. Nearly all
RRs associated with aspects of cigarette smoking were above unity, although these
were mostly not statistically significant. Former smokers were at highest risk of BE
(RR=1.33, 95% CI 1.00-1.77) when compared with never smokers, but current smokers
were not at higher risk. Smoking frequency was found to be related to a non-
significantly higher risk of BE (p trend=0.46). A longer duration of cigarette smoking
was associated with an increasing risk of BE (p trend=0.03). Results on the association
between pack-years of smoking and BE showed increased RRs, but only for subjects
who had smoked 20-<40 pack-years the RR was statistically significant (RR=1.50, 95%
CI 1.04-2.17, p trend=0.08). The results of analyses on smoking cessation did not
indicate a lower risk of BE with increasing duration of smoking cessation (p
trend=0.82).
Thesis Jessie Steevens_v04.pdf
Chapter 2
32
Thesis Jessie Steevens_v04.pdf
Table 2.2 Association (multivariable-adjusteda) between overweight and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Men Women
categorical
median
person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) p inter-
actionb
BMI at baseline (kg/m2)
18.5- <25 23.3 12267 119 1 (reference) 14327 65 1 (reference)
25- <30 26.6 9801 83 0.85 (0.63-1.14) 9151 75 1.72 (1.22-2.44)
≥ 30 31.5 808 10 1.21 (0.60-2.47) c 2288 18 1.67 (0.96-2.90) 0.01
p trend
d0.73 p trend 0.01
continuous, 1 kg/m2 increments 22876 212 0.99 (0.93-1.05) c 25766 158 1.07 (1.03-1.11) 0.03
BMI at age 20 years (kg/m2)
<20 18.7 3603 32 0.72 (0.45-1.13) 7057 36 1.06 (0.62-1.80)
20-<21.5 20.8 5027 64 1 (reference) 5190 25 1 (reference)
21.5-<23 22.2 4885 37 0.61 (0.40-0.94) 5386 38 1.45 (0.86-2.46)
23-<25 23.8 3578 25 0.56 (0.34-0.91) 4076 31 1.59 (0.92-2.76)
≥25 26.0 1323 6 0.37 (0.15-0.87) 1775 14 1.56 (0.79-3.10) c 0.01
p trend 0.02 p trend 0.05
continuous, 1 kg/m2 increments 18417 164 0.93 (0.87-0.99) 23485 144 1.07 (1.01-1.13) 0.001
Change in BMI after 20 years of age (kg/m2) e
<0 -1.3 1763 8 0.56 (0.26-1.20) 2662 8 0.33 (0.15-0.75)
0-<4 2.2 9908 95 1 (reference) 10499 67 1 (reference)
4-<8 5.5 5799 51 0.77 (0.52-1.15) 7675 51 1.12 (0.76-1.64)
≥8 9.4 947 10 0.83 (0.39-1.78) 2648 18 1.21 (0.69-2.10) 0.95
p trend 0.79 p trend 0.01
continuous, 1 kg/m2 increments 18417 164 0.99 (0.92-1.06) 23485 144 1.06 (1.02-1.10) 0.92
BMI at baseline (kg/m2), additionally adjusted for pant/skirt size f
18.5- <25 23.3 11349 113 1 (reference) 14228 65 1 (reference)
25- <30 26.6 8925 77 0.81 (0.57-1.15) 9037 75 1.56 (1.02-2.39)
≥ 30 31.5 661 10 1.35 (0.59-3.06) c 2205 18 1.38 (0.65-2.89) 0.01
p trend 0.78 p trend 0.23
continuous, 1 kg/m2 increments 20935 200 1.00 (0.92-1.08) c 25469 158 1.06 (0.99-1.13) 0.07
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
33
Thesis Jessie Steevens_v04.pdf
Men Women
categorical
median
person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) p inter-
action b
Pant/skirt size as a proxy for waist circumference (men/women) f
≤48 / ≤40 3085 30 1.20 (0.73-1.98) 4636 12 0.44 (0.22-0.85)
50-51 / 42 4834 40 1 (reference) 6487 39 1 (reference)
52-53 / 44 7188 73 1.22 (0.81-1.84) 7049 55 1.25 (0.82-1.92)
54-55 / 46-48 3961 37 1.11 (0.69-1.78) c 6487 46 1.10 (0.70-1.71)
≥56 / ≥50 1867 20 1.28 (0.72-2.27) c 810 6 1.23 (0.50-3.04)
p trend 0.73 p trend 0.01
continuous 20935 200 1.02 (0.91-1.15) 25469 158 1.16 (1.05-1.28)
Pant/skirt size as a proxy for waist circumference (men/women), additionally
adjusted for BMI at baseline f
≤48 / ≤40 3085 30 1.19 (0.72-1.98) c 4636 12 0.51 (0.26-1.01)
50-51 / 42 4834 40 1 (reference) 6487 39 1 (reference)
52-53 / 44 7188 73 1.23 (0.81-1.87) 7049 55 1.08 (0.70-1.69)
54-55 / 46-48 3961 37 1.12 (0.68-1.85) 6487 46 0.76 (0.44-1.33)
≥56 / ≥50 1867 20 1.26 (0.67-2.38) 810 6 0.52 (0.15-1.83)
p trend 0.76 p trend 0.66
continuous 20935 200 1.03 (0.89-1.18) 25469 158 1.03 (0.87-1.23)
a adjusted for age (years), cigarette smoking (current smoking status (yes/no), frequency (number of cigarettes/day), and duration (years)), alcohol consumption (g/day);
b p-value for interaction between sex and measure of overweight (BMI, BMI at age 20, BMI change or pant/skirt size), based on cross product term in the Cox
proportional hazard model; c proportional hazards assumption was violated for this analysis; d tests for dose-response trends were assessed by fitting ordinal variables as
continuous terms in the Cox proportional hazard model; e additionally adjusted for BMI at age 20 years; f pant size (men) corresponds to the following standard waist
circumferences: ≤50 = ≤88 cm; 50-51 = 88 cm; 52-53 = 93 cm; 54-55 = 98 cm; ≥56 = ≥103 cm, skirt size (women) corresponds to the following standard waist
circumferences: ≤40 = ≤74 cm; 40 = 74 cm; 42 = 78 cm; 44 = 82 cm; 46 = 86 cm; 48 = 91 cm; ≥50 = ≥96 cm. BMI, body mass index; CI, confidence interval; RR, incidence
rate ratio; SIM specialized intestinal metaplasia.
Chapter 2
34
Thesis Jessie Steevens_v04.pdf
Table 2.3 Association (multivariable-adjusteda) between cigarette smoking and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) p
inter-
action
b
Smoking status a
never smokers 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
former smokers 17390 170 1.33 (1.00-1.77) 12120 127 1.35 (0.87-2.10) 5270 43 1.55 (1.03-2.33)
current smokers 12365 79 0.93 (0.68-1.28) 7376 58 1.04 (0.64-1.69) 4989 21 0.81 (0.49-1.36) 0.45
p trend
c 0.61 p trend 0.71 p trend 0.93
Frequency of cigarette smoking (no./day) d
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 10 18819 149 1.16 (0.75-1.81) 10965 101 0.97 (0.50-1.86) 7853 48 1.69 (0.91-3.13)
≥20 20 10936 100 1.24 (0.74-2.08) 8531 84 1.02 (0.50-2.10) 2405 16 1.89 (0.78-4.59) 0.99
p trend 0.46 p trend 0.79 p trend 0.17
continuous, 10 cigarettes/day increments 48642 370 1.05 (0.93-1.19) 22876 212 1.05 (0.92-1.20) 25766 158 1.09 (0.80-1.50) 0.83
Duration of cigarette smoking (years) e
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 12 5724 41 1.05 (0.68-1.60) 2987 20 0.79 (0.41-1.52) 2737 21 1.55 (0.88-2.73)
20 - <40 30 14353 129 1.37 (0.96-1.94) 9095 95 1.35 (0.81-2.25) 5258 34 1.56 (0.81-3.03)
≥40 43 9677 79 1.51 (0.97-2.34) 7413 70 1.63 (0.90-2.95) 2263 9 1.13 (0.42-3.04) 0.20
p trend 0.03 p trend 0.03 p trend 0.41
continuous, 10 years increments 48642 370 1.07 (0.97-1.19) 22876 212 1.10 (0.96-1.25) 25766 158 1.08 (0.87-1.33) 0.43
Pack-years of cigarette smoking f
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 9 15872 125 1.27 (0.94-1.72) 8787 79 1.24 (0.78-1.97) 7085 46 1.55 (1.01-2.38)
20 - <40 28 9612 87 1.50 (1.04-2.17) 7220 71 1.51 (0.92-2.47) 2393 16 1.80 (0.94-3.45)
≥40 48 4270 37 1.44 (0.91-2.28) 3490 35 1.59 (0.89-2.84) 781 2 0.65 (0.15-2.90) 0.55
p trend 0.08 p trend 0.09 p trend 0.37
continuous, 10 pack-years increments 48642 370 1.05 (0.98-1.12) 22876 212 1.05 (0.98-1.13) 25766 158 1.08 (0.91-1.28) 0.15
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
35
Thesis Jessie Steevens_v04.pdf
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. o
f
cases
RR (95% CI) p
inter-
action
b
Smoking cessation a
never smokers 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
stopped ≥ 20 years ago 25 5413 47 1.16 (0.79-1.71) 3972 37 1.18 (0.70-1.99) 1441 10 1.38 (0.68-2.82)
stopped 10 - <20 years ago 14 5960 64 1.46 (1.03-2.07) 4205 53 1.63 (0.99-2.67) 1755 11 1.21 (0.61-2.38)
stopped >0 - <10 years ago 5 5985 58 1.32 (0.91-1.91) 3943 37 1.23 (0.72-2.10) 2042 21 1.88 (1.12-3.17)
current smokers 0 12365 79 0.93 (0.68-1.28) 7376 58 1.04 (0.64-1.69) 4989 21 0.82 (0.49-1.37) 0.34
p trend 0.82 p trend 0.78 p trend 0.79
a all analyses were adjusted for: age (years), alcohol consumption (g/day), body mass index (kg/m²); b p-value for interaction between sex and cigarette smoking, based on
cross product term in the Cox proportional hazard model; c tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox
proportional hazard model; d additionally adjusted for current smoking status (yes/no) and smoking duration (years); e additionally adjusted for current smoking status
(yes/no) and smoking frequency (number of cigarettes/day); f additionally adjusted for current smoking status (yes/no). CI, confidence interval; RR, incidence rate ratio;
SIM specialized intestinal metaplasia.
Chapter 2
36
Thesis Jessie Steevens_v04.pdf
Table 2.4 Association (multivariable-adjusteda) between alcohol consumption and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) p inter-
action b
Alcohol consumption (g ethanol/day)
abstainer 0 11282 75 1 (reference) 3243 24 1 (reference) 8039 51 1 (reference)
>0-<5 2 14416 122 1.22 (0.90-1.65) 4860 50 1.37 (0.81-2.31) 9556 72 1.19 (0.82-1.73)
5-<15 9 10925 78 0.94 (0.67-1.31) c 6026 58 1.28 (0.77-2.13)c 4899 20 0.66 (0.38-1.16)
15-<30 22 7692 64 1.00 (0.69-1.45) 5335 52 1.26 (0.75-2.13) 3273 15 0.75 (0.40-1.41) d 0.35
>=30 40 4327 31 0.82 (0.51-1.29) 3411 28 1.02 (0.57-1.84)
p trend
e 0.16 p trend 0.54 p trend 0.12
continuous, 10 grams ethanol/day
increments
48642 370 0.95 (0.87-1.03) 22876 212 0.97 (0.89
-
1.07) 25766 158 0.84 (0.65-1.08) 0.20
Alcohol consumption (g ethanol/day)
Stable users f
abstainer 0 9099 58 1 (reference) 2513 17 1 (reference) 6586 41 1 (reference)
>0-<5 2 8447 80 1.43 (1.00-2.06) c 3133 34 1.60 (0.85-3.00) c 5313 46 1.43 (0.92-2.23)
5-<15 9 6694 46 0.96 (0.64-1.45) c 3904 35 1.36 (0.73-2.53) c 2790 11 0.62 (0.30-1.27)
15-<30 22 4239 38 1.19 (0.76-1.88) c 3022 33 1.71 (0.90-3.26) c 1734 6 0.53 (0.21-1.35) d 0.27
>=30 40 2346 17 0.90 (0.50-1.62) 1829 16 1.29 (0.62-2.67)
p trend 0.42 p trend 0.73 p trend 0.05
continuous, 10 g ethanol/day
increments
30825 239 0.97 (0.87-1.09) 14402 135 1.03 (0.91-1.15) 16423 104 0.73 (0.51-1.05) 0.09
Alcoholic beverages (glasses/day) g
Beer
no beer 0 32777 221 1 (reference) 9591 72 1 (reference) h
>0-1 0.2 12552 130 1.31 (1.01-1.69) 10162 121 1.60 (1.16-2.21)
>1-2 1.4 2167 14 0.81 (0.45-1.48) 2043 14 0.95 (0.51-1.77)
>2 3.7 1146 5 0.59 (0.23-1.51) 1080 5 0.67 (0.26-1.75)
p trend 0.10 p trend 0.10
continuous, 1 glass/day
increments
48642 370 0.86 (0.69-1.08) 22876 212 0.86 (0.69-1.07)
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
37
Thesis Jessie Steevens_v04.pdf
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) p inter-
action b
Wine
no wine 0 22196 161 1 (reference) 11012 89 1 (reference) 11184 72 1 (reference)
>0-1 0.2 20214 166 1.19 (0.94-1.50) 9050 97 1.34 (0.98-1.82) 11163 69 1.11 (0.77-1.62)
>1-2 1.4 4097 30 1.15 (0.73-1.79) c 1800 17 1.23 (0.68-2.24) c 3418 17 1.48 (0.72-3.01) d 0.50
>2 2.7 2135 13 1.03 (0.53-2.01) 1014 9 1.20 (0.54-2.69)
p trend 0.91 p trend 0.72 p trend 0.29
continuous, 1 glass/day
increments
48642 370 1.04 (0.86-1.28) c 22876 212 1.06 (0.84-1.33) c 25766 158 1.71 (0.83-3.55) 0.42
Liquor
no liquor 0 24989 169 1 (reference) 8300 70 1 (reference) 16689 99 1 (reference)
>0-1 0.2 17954 149 1.16 (0.91-1.48) c 9790 92 1.15 (0.82-1.62) c 8164 57 1.20 (0.84-1.71)
>1-2 1.9 4068 37 1.35 (0.86-2.11) 3316 36 1.48 (0.89-2.47) 913 2 0.53 (0.13-2.14) d 0.44
>2 2.8 1631 15 1.60 (0.79-3.24) 1470 14 1.50 (0.68-3.28)
p trend 0.17 p trend 0.17 p trend 0.51
continuous, 1 glass/day
increments
48642 370 1.10 (0.89-1.36) c 22876 212 1.13 (0.89-1.42) 25766 158 0.70 (0.35-1.41) 0.14
a Adjusted for age (years), cigarette smoking (current smoking status (yes/no), frequency (number of cigarettes/day), and duration (years)), body mass index (kg/m²); b p-
value for interaction between sex and alcohol consumption, based on cross product term in the Cox proportional hazard model; c Proportional hazards assumption was
violated for this analysis; d For analyses on women, the highest two categories of consumption were combined, because of low case numbers; e Tests for dose-response
trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model; f Subjects who had not changed their alcohol consumption
habits in the 5 years before baseline; g Additionally adjusted for ethanol intake; h It was not possible to perform analyses on beer consumption in women, because too
few women consumed beer. CI, confidence interval; RR, incidence rate ratio; SIM specialized intestinal metaplasia.
Chapter 2
38
Alcohol consumption
As can be seen in Table 2.4, there was no interaction between sex and alcohol
consumption in the analyses of BE risk. Analyses of the association between alcohol
intake and risk of BE indicated no increased risk. Subjects in the highest category of
intake (≥30 g ethanol/day) had an RR of 0.82 (95% CI 0.51-1.29) when compared with
abstainers. A sensitivity analysis was performed based on subjects who had a stable
alcohol consumption in the period starting 5 years before baseline until baseline. This
analysis included approximately 65% of the study population. Broadly speaking, the
results were comparable with those based on the total study population.
Also shown in Table 2.4 are RRs associated with consumption of alcoholic
beverages, adjusted for ethanol intake. The RRs for beer and wine consumption do not
shown clear inverse or positive associations. Increasing liquor consumption was
associated with a non-significantly increased risk of BE (RR for >2 glasses/day=1.60,
95% CI 0.79-3.24, p trend=0.17).
DISCUSSION
In this large prospective cohort study, overweight was found to be a risk factor for BE
in women, but not in men. Several aspects of cigarette smoking were associated with
an increase in risk of BE, while alcohol consumption was not associated with risk of BE.
Overweight
Our observation of a positive association between increased BMI and risk of BE in
women, is comparable with a recent observation in another prospective cohort study
among women. They found obese women to be at increased risk of BE (OR 1.52, 95%
CI 1.02-2.28). However, no association with overweight was found in that study (OR
0.92, 95% CI 0.66-1.27), while we did find a positive association with overweight. Three
case-control studies showed an increased risk of BE for overweight and obese
persons.7-9 These studies were based on predominantly male populations (60-100%
male cases), thus it is likely that overweight and obese men were at increased risk,
even though no separate results by sex were presented. We did not observe an
increased risk in men. Previously, a positive association between overweight and risk
of esophageal adenocarcinoma has also been described, in the NLCS (in women only)27
and a meta-analysis.28
Several possible biological mechanisms may be involved in the association
between overweight and development of BE. Some studies observed that adjustment
for gastroesophageal reflux disease caused attenuation or disappearance of the effect
of increased BMI on BE risk,6,29 whereas in other studies this adjustment did not affect
the RRs.5,10 Therefore, it is probable that the BMI acts through other mechanisms
besides gastroesophageal reflux disease, as was also suggested recently in a
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
39
comprehensive review.30 Another possible mechanism relates to the distribution of
body fat. Abdominal fat is more metabolically active than subcutaneous fat, and
secretes several factors that may be involved in systemic inflammation and cancer
development.30,31 Two case-control studies observed positive associations between
measures of abdominal overweight and BE risk.8,32 In our study, we did not observe an
increased risk of BE for women with abdominal overweight, when total overweight
was accounted for. However, the use of clothing size measures as a proxy for waist
circumference may not have been optimal. Results from the Nurses’ Health Study did
not show an association for abdominal overweight either.10 Thus, there are some
inconsistencies between observations on the role of abdominal overweight in the
development BE.
Cigarette smoking
Our observation that several aspects of cigarette smoking were associated with an
increased risk of BE are in agreement with results from some previous case-control
studies.6,8 In an earlier analysis within the NLCS, we investigated the association of
cigarette smoking with risk of esophageal adenocarcinoma.33 The association of
cigarette smoking with esophageal adenocarcinoma was stronger (RR for ≥40 pack-
years=2.93 (95% CI 1.59-5.40) than the association with its precursor BE. This
difference in strength of the associations with BE and esophageal adenocarcinoma is
consistent with the literature.e.g. 5,8,12,34,35 This observation may indicate that smoking
plays a role in the later stages of esophageal carcinogenesis: in the progression of BE to
esophageal adenocarcinoma. In a population of BE patients, risk of progression was
indeed found to be increased for patients who had smoked, although this was not
statistically significant (RR=1.53 (95% CI 0.68-6.44).36 Further investigation into this
topic may yield interesting insights.
Alcohol consumption
Our finding that there was no association between alcohol intake and risk of BE is
consistent with results from most studies.8,11,13 Also, the findings are analogous to the
null association between alcohol and risk of esophageal adenocarcinoma.33,34,37
Wine consumption has been reported to be inversely associated with BE risk.11,13
However, within the NLCS, we could not confirm this association. The fact that we
adjusted for total ethanol intake while the other two studies adjusted for beer and
liquor intake, cannot explain the different findings, as there were no associations with
ethanol, beer or liquor.
Strengths and limitations
An important strength of this study is its prospective character. The advantage of the
prospective design is the relative insensitivity for selection and information bias
Thesis Jessie Steevens_v04.pdf
Chapter 2
40
compared with a case-control design. Second, this study is one of the largest with
respect to the number of BE cases, specifically the number of female cases. This large
number of cases allowed the investigation of interaction and separate analyses for
men and women. A third strength is the availability of extensive information about the
exposure variables investigated, which allowed a detailed look at these exposures.
There are also some limitations to our study. The first is the lack of information on
the presence of gastroesophageal reflux disease and the use of medications for this
disease. Consequently, we were not able to investigate possible confounding effects of
these factors or their possible intermediate role in the associations investigated.
Second, the use of clothing size as a proxy measure for waist circumference is sub-
optimal compared with measurement of waist circumference. Third, subcohort
members who were not diagnosed with BE were assumed to be disease-free, but we
were unable to verify this assumption. Therefore, there may be some BE cases missing
in our dataset, which may have reduced the power. Finally, the power of the study also
may have been reduced due to the incomplete coverage of the NLCS population by
PALGA before 1991. Due to this incompleteness we may have missed some BE cases or
the registered incidence date may be later than the true incidence date. We estimated
that we may have missed 3% of the BE cases in our cohort at most.
Recommendations for further research
Most studies that investigated overweight, smoking, and alcohol consumption in
relation to the risk of BE, had a case-control design. Reversed causation can be a
problem in studies investigating etiology, specifically in studies with a case-control
design. This is illustrated by our finding that cases diagnosed during the first 2 years of
follow-up reported different alcohol consumption habits at baseline than later cases.
Only one prospective cohort study previously reported on overweight and risk of BE,
while we are the first prospective cohort study to report on alcohol and smoking. We
therefore suggest that additional cohort studies investigate these factors in association
with BE risk. Future studies should, whenever possible, investigate whether
associations differ by sex. Differences in associations between sexes may help to
explain the higher male-to-female ratio in esophageal adenocarcinoma compared with
BE. As mentioned before, studying risk factors for progression of BE to esophageal
adenocarcinoma also deserves further attention.
Thesis Jessie Steevens_v04.pdf
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
41
Thesis Jessie Steevens_v04.pdf
Supplemental Table 2.1 Association (age-adjusted) between overweight and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Men Women
categorical
median
person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) p inter-
action a
BMI at baseline (kg/m2)
18.5- <25 23.3 12267 119 1 (reference) 14327 65 1 (reference)
25- <30 26.6 9801 83 0.88 (0.65-1.18) 9151 75 1.79 (1.27-2.53)
≥ 30 31.5 808 10 1.29 (0.64-2.60) b 2288 18 1.74 (1.01-3.01) 0.01
p trend
c0.90 p trend 0.002
Continuous, 1 kg/m2 increments 22876 212 1.00 (0.94-1.06) b 25766 158 1.07 (1.04-1.11) 0.04
BMI at age 20 years (kg/m2)
<20 18.7 3603 32 0.71 (0.45-1.11) 7057 36 1.06 (0.63-1.80)
20-<21.5 20.8 5027 64 1 (reference) 5190 25 1 (reference)
21.5-<23 22.2 4885 37 0.60 (0.39-0.93) 5386 38 1.46 (0.87-2.47)
23-<25 23.8 3578 25 0.56 (0.34-0.91) 4076 31 1.59 (0.92-2.75)
≥25 26.0 1323 6 0.37 (0.16-0.88) 1775 14 1.64 (0.83-3.24) 0.01
p trend 0.03 p trend 0.04
continuous, 1 kg/m2 increments 18417 164 0.93 (0.87-0.99) 23485 144 1.07 (1.01-1.13) 0.001
Change in BMI after 20 years of age (kg/m2) d
<0 -1.3 1763 8 0.48 (0.23-1.00) 2662 8 0.47 (0.22-1.00)
0-<4 2.2 9908 95 1 (reference) 10499 67 1 (reference)
4-<8 5.5 5799 51 0.90 (0.63-1.29) 7675 51 1.04 (0.71-1.52)
≥8 9.4 947 10 1.10 (0.55-2.20) 2648 18 1.07 (0.62-1.84) 0.96
p trend 0.29 p trend 0.11
continuous, 1 kg/m2 increments 18417 164 1.03 (0.98-1.08) 23485 144 1.03 (0.99-1.08) 0.97
BMI at baseline (kg/m2), additionally adjusted for pant/skirt size e
18.5- <25 23.3 11349 113 1 (reference) 14228 65 1 (reference)
25- <30 26.6 8925 77 0.83 (0.59-1.17) 9037 75 1.61 (1.05-2.46)
≥ 30 31.5 661 10 1.40 (0.62-3.16) b 2205 18 1.40 (0.67-2.95) 0.01
p trend 0.87 p trend 0.20
continuous, 1 kg/m2 increments 20935 200 1.00 (0.93-1.08) b 25469 158 1.06 (1.00-1.14) 0.09
Chapter 2
42
Thesis Jessie Steevens_v04.pdf
Men Women
categorical
median
person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) person time at
risk in
subcohort (y)
no. of
cases
RR (95% CI) p inter-
action a
Pant/skirt size as a proxy for waist circumference (men/women) e
≤48 / ≤40 3085 30 1.18 (0.72-1.95) 4636 12 0.43 (0.22-0.83)
50-51 / 42 4834 40 1 (reference) 6487 39 1 (reference)
52-53 / 44 7188 73 1.23 (0.82-1.86) 7049 55 1.29 (0.84-1.97)
54-55 / 46-48 3961 37 1.13 (0.71- 1.81) b 6487 46 1.16 (0.74-1.81)
≥56 / ≥50 1867 20 1.31 (0.74- 2.33) b 810 6 1.23 (0.50-3.03)
p trend 0.61 p trend 0.002
continuous 20935 200 1.03 (0.92-1.16) 25469 158 1.18 (1.07-1.29)
Pant/skirt size as a proxy for waist circumference (men/women),
additionally adjusted for BMI at baseline e
≤48 / ≤40 3085 30 1.17 (0.71- 1.95) b 4636 12 0.51 (0.26-1.01)
50-51 / 42 4834 40 1 (reference) 6487 39 1 (reference)
52-53 / 44 7188 73 1.23 (0.81-1.88) 7049 55 1.10 (0.71-1.71)
54-55 / 46-48 3961 37 1.13 (0.69-1.86) 6487 46 0.78 (0.45-1.36)
≥56 / ≥50 1867 20 1.28 (0.68-2.42) 810 6 0.49 (0.14-1.70)
p trend 0.70 p trend 0.59
continuous 20935 200 1.03 (0.90-1.19) 25469 158 1.04 (0.87-1.23)
a p-value for interaction between sex and measure of overweight (BMI, BMI at age 20, BMI change or pant/skirt size), based on cross product term in the Cox proportional
hazard model; b Proportional hazards assumption was violated for this analysis; c Tests for dose-response trends were assessed by fitting ordinal variables as continuous
terms in the Cox proportional hazard model; d Additionally adjusted for BMI at age 20 years; e Pant size (men) corresponds to the following standard waist
circumferences: ≤50 = ≤88 cm; 50-51 = 88 cm; 52-53 = 93 cm; 54-55 = 98 cm; ≥56 = ≥103 cm, skirt size (women) corresponds to the following standard waist
circumferences: ≤40 = ≤74 cm; 40 = 74 cm; 42 = 78 cm; 44 = 82 cm; 46 = 86 cm; 48 = 91 cm; ≥50 = ≥96 cm. BMI, body mass index; CI, confidence interval; RR, incidence
rate ratio; SIM specialized intestinal metaplasia.
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
43
Thesis Jessie Steevens_v04.pdf
Supplemental Table 2.2 Association (age-adjusted) between cigarette smoking and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) p
inter-
actiona
Smoking status
never smokers 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
former smokers 17390 170 1.29 (0.97-1.70) 12120 127 1.33 (0.86-2.07) 5270 43 1.37 (0.93-2.01)
current smokers 12365 79 0.89 (0.65-1.21) 7376 58 1.02 (0.63-1.65) 4989 21 0.71 (0.44-1.17) 0.47
p trend b 0.40 p trend 0.66 p trend 0.46
Frequency of cigarette smoking (n/day)
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 10 18819 149 1.08 (0.82-1.42) 10965 101 1.17 (0.75-1.84) 7853 48 1.03 (0.71-1.49)
≥20 20 10936 100 1.19 (0.87-1.63) 8531 84 1.28 (0.81-2.02) 2405 16 1.14 (0.65-1.98) 0.90
p trend 0.29 p trend 0.29 p trend 0.68
continuous, 10 cigarettes/day
increments
48642 370 1.07 (0.96-1.18) 22876 212 1.08 (0.96-1.21) 25766 158 1.02 (0.83-1.27) 0.67
Duration of cigarette smoking (years)
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 12 5724 41 0.99 (0.67-1.47) 2987 20 0.85 (0.46-1.57) 2737 21 1.29 (0.78-2.14)
20 - <40 30 14353 129 1.21 (0.91-1.60) 9095 95 1.34 (0.85-2.10) 5258 34 1.10 (0.72-1.68)
≥40 43 9677 79 1.05 (0.76-1.44) 7413 70 1.22 (0.77-1.96) 2263 9 0.66 (0.33-1.34) 0.20
p trend 0.43 p trend 0.16 p trend 0.66
continuous, 10 years increments 48642 370 1.01 (0.94-1.07) 22876 212 1.03 (0.95-1.13) 25766 158 0.97 (0.88-1.08) 0.40
Chapter 2
44
Thesis Jessie Steevens_v04.pdf
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no. of
cases
RR (95% CI) p
inter-
actiona
Pack-years of cigarette smoking
no cigarette smoker 0 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
>0 - <20 9 15872 125 1.08 (0.82- 1.43) 8787 79 1.14 (0.72-1.81) 7085 46 1.09 (0.75-1.59)
20 - <40 28 9612 87 1.18 (0.85- 1.63) 7220 71 1.27 (0.79-2.03) 2393 16 1.14 (0.66-1.99)
≥40 48 4270 37 1.11 (0.74- 1.67) 3490 35 1.31 (0.77-2.23) 781 2 0.43 (0.10-1.79) 0.56
p trend 0.46 p trend 0.27 p trend 0.66
continuous, 10 pack-years increments 48642 370 1.02 (0.96- 1.08) 22876 212 1.03 (0.96-1.10) 25766 158 0.98 (0.85-1.12) 0.53
Smoking cessation
never smokers 18888 121 1 (reference) 3380 27 1 (reference) 15508 94 1 (reference)
stopped ≥ 20 years 25 5413 47 1.11 (0.76- 1.63) 3972 37 1.17 (0.69-1.98) 1441 10 1.15 (0.58-2.29)
stopped 10 - <20 years 14 5960 64 1.41 (1.00- 1.98) 4205 53 1.61 (0.98-2.63) 1755 11 1.06 (0.55-2.04)
stopped >0 - <10 years 5 5985 58 1.29 (0.90- 1.86) 3943 37 1.20 (0.71-2.03) 2042 21 1.73 (1.04-2.89)
current smokers 0 12365 79 0.88 (0.65- 1.20) 7376 58 1.02 (0.63-1.65) 4989 21 0.72 (0.44-1.17) 0.34
p trend 0.61 p trend 0.72 p trend 0.73
a p-value for interaction between sex and cigarette smoking, based on cross product term in the Cox proportional hazard model; b tests for dose-response trends were
assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model. CI, confidence interval; RR, incidence rate ratio; SIM specialized intestinal
metaplasia.
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
45
Thesis Jessie Steevens_v04.pdf
Supplemental Table 2.3 Association (age-adjusted) between alcohol consumption and risk of Barrett's esophagus with SIM; Netherlands Cohort Study (1986-2002).
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) p
inter-
actiona
Alcohol consumption (g ethanol/day)
abstainer 0 11282 75 1 (reference) 3243 24 1 (reference) 8039 51 1 (reference)
>0-<5 2 14416 122 1.24 (0.92-1.68) 4860 50 1.40 (0.83-2.34) 9556 72 1.18 (0.81-1.72)
5-<15 9 10925 78 0.96 (0.68-1.33) b 6026 58 1.32 (0.80-2.19) b 4899 20 0.64 (0.38-1.09)
15-<30 22 7692 64 1.04 (0.73-1.49) 5335 52 1.33 (0.79-2.21) b 3273 15 0.73 (0.40-1.31) c 0.27
>=30 40 4327 31 0.86 (0.55-1.34) 3411 28 1.11 (0.62-1.97)
p trend d 0.22 p trend 0.75 p trend 0.06
continuous, 10 grams ethanol/day
increments
48642 370 0.96 (0.88-1.04) 22876 212 0.98 (0.90-1.07) 25766 158 0.82 (0.65-1.04) 0.16
Alcohol consumption (g ethanol/day), stable users e
Abstainer 0 9099 58 1 (reference) 2513 17 1 (reference) 6586 41 1 (reference)
>0-<5 2 8447 80 1.43 (1.00-2.04) b 3133 34 1.64 (0.89-3.03) b 5313 46 1.38 (0.89-2.15)
5-<15 9 6694 46 0.96 (0.64-1.44) b 3904 35 1.39 (0.75-2.57) b 2790 11 0.63 (0.32-1.25)
15-<30 22 4239 38 1.20 (0.78-1.86) b 3022 33 1.70 (0.91-3.17) b 1734 6 0.56 (0.23-1.35) c 0.21
>=30 40 2346 17 0.94 (0.53-1.66) 1829 16 1.34 (0.65-2.74)
p trend 0.48 p trend 0.68 p trend 0.04
continuous, 10 g ethanol/day
increments
30825 239 0.98 (0.88-1.09) 14402 135 1.03 (0.92-1.15) 16423 104 0.74 (0.53-1.03) 0.07
Alcoholic beverages (glasses/day) f
Beer
no beer 0 32777 221 1 (reference) 9591 72 1 (reference) g
>0-1 0.2 12552 130 1.33 (1.03-1.72) 10162 121 1.62 (1.17-2.23)
>1-2 1.4 2167 14 0.84 (0.46-1.53) 2043 14 0.96 (0.51-1.80)
>2 3.7 1146 5 0.60 (0.24-1.51) 1080 5 0.67 (0.26-1.72)
p trend 0.10 p trend 0.09
continuous, 1 glass/day increments 48642 370 0.87 (0.70-1.08) 22876 212 0.86 (0.69-1.07)
Chapter 2
46
Thesis Jessie Steevens_v04.pdf
Total Men Women
categorical
median
person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) person time
at risk in
subcohort
(y)
no.
of
cases
RR (95% CI) p
inter-
actiona
Wine
no wine 0 22196 161 1 (reference) 11012 89 1 (reference) 11184 72 1 (reference)
>0-1 0.2 20214 166 1.17 (0.93-1.48) 9050 97 1.33 (0.98-1.81) 11163 69 1.07 (0.74-1.54)
>1-2 1.4 4097 30 1.12 (0.72-1.76) b 1800 17 1.22 (0.68-2.22) b 3418 17 1.41 (0.70-2.82) c 0.46
>2 2.7 2135 13 0.99 (0.51-1.92) 1014 9 1.19 (0.54-2.65)
p trend 0.98 p trend 0.71 p trend 0.33
continuous, 1 glass/day increments 48642 370 1.03 (0.84-1.25) b 22876 212 1.05 (0.84-1.32) b 25766 158 1.61 (0.81-3.22) 0.37
Liquor
no liquor 0 24989 169 1 (reference) 8300 70 1 (reference) 16689 99 1 (reference)
>0-1 0.2 17954 149 1.17 (0.92-1.49) b 9790 92 1.15 (0.82-1.61) b 8164 57 1.21 (0.85-1.71)
>1-2 1.9 4068 37 1.36 (0.87-2.12) 3316 36 1.48 (0.90-2.46) 913 2 0.54 (0.14-2.13) c 0.42
>2 2.8 1631 15 1.65 (0.82-3.32) 1470 14 1.54 (0.70-3.36)
p trend 0.14 p trend 0.15 p trend 0.54
continuous, 1 glass/day increments 48642 370 1.11 (0.90-1.36) b 22876 212 1.13 (0.90-1.43) 25766 158 0.72 (0.37-1.41) 0.12
a p-value for interaction between sex and alcohol consumption based on cross product term in the Cox proportional hazard model; b Proportional hazards assumption was
violated in this analysis; c For analyses on women, the highest two categories of consumption were combined, because of low case numbers; d Tests for dose-response
trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model; e Subjects who had not changed their alcohol consumption
habits in the 5 years before baseline; f Additionally adjusted for ethanol intake; g It was not possible to perform analyses on beer consumption in women, because too few
women consumed beer. CI, confidence interval; RR, incidence rate ratio; SIM specialized intestinal metaplasia.
Overweight, smoking, alcohol consumption, and risk of Barrett’s esophagus
47
REFERENCES
1. Wang KK, Sampliner RE. Updated guidelines 2008 for the diagnosis, surveillance and therapy of
Barrett's esophagus. Am J Gastroenterol 2008;103:788-797.
2. Playford RJ. New British Society of Gastroenterology (BSG) guidelines for the diagnosis and
management of Barrett's oesophagus. Gut 2006;55:442-443.
3. Yousef F, Cardwell C, Cantwell MM, Galway K, Johnston BT, Murray L. The incidence of esophageal
cancer and high-grade dysplasia in Barrett's esophagus: a systematic review and meta-analysis. Am J
Epidemiol 2008;168:237-249.
4. Wild CP, Hardie LJ. Reflux, Barrett's oesophagus and adenocarcinoma: burning questions. Nat Rev
Cancer 2003;3:676-684.
5. Anderson LA, Watson RG, Murphy SJ, Johnston BT, Comber H, Mc Guigan J, Reynolds JV, Murray LJ.
Risk factors for Barrett's oesophagus and oesophageal adenocarcinoma: results from the FINBAR
study. World J Gastroenterol 2007;13:1585-1594.
6. Smith KJ, O'Brien SM, Green AC, Webb PM, Whiteman DC. Current and past smoking significantly
increase risk for Barrett's esophagus. Clin Gastroenterol Hepatol 2009;7:840-848.
7. Bu X, Ma Y, Der R, Demeester T, Bernstein L, Chandrasoma PT. Body mass index is associated with
Barrett esophagus and cardiac mucosal metaplasia. Dig Dis Sci 2006;51:1589-1594.
8. Edelstein ZR, Farrow DC, Bronner MP, Rosen SN, Vaughan TL. Central adiposity and risk of Barrett's
esophagus. Gastroenterology 2007;133:403-411.
9. Stein DJ, El-Serag HB, Kuczynski J, Kramer JR, Sampliner RE. The association of body mass index with
Barrett's oesophagus. Aliment Pharmacol Ther 2005;22:1005-1010.
10. Jacobson BC, Chan AT, Giovannucci EL, Fuchs CS. Body mass index and Barrett's oesophagus in women.
Gut 2009;58:1460-1466.
11. Anderson LA, Cantwell MM, Watson RG, Johnston BT, Murphy SJ, Ferguson HR, McGuigan J, Comber
H, Reynolds JV, Murray LJ. The association between alcohol and reflux esophagitis, Barrett's
esophagus, and esophageal adenocarcinoma. Gastroenterology 2009;136:799-805.
12. Kubo A, Levin TR, Block G, Rumore G, Quesenberry CP, Jr., Buffler P, Corley DA. Cigarette smoking and
the risk of Barrett's esophagus. Cancer Causes Control 2009;20:303-311.
13. Kubo A, Levin TR, Block G, Rumore GJ, Quesenberry CP, Jr., Buffler P, Corley DA. Alcohol types and
sociodemographic characteristics as risk factors for Barrett's esophagus. Gastroenterology
2009;136:806-815.
14. Akiyama T, Inamori M, Iida H, Mawatari H, Endo H, Hosono K, Yoneda K, Fujita K, Yoneda M, Takahashi
H, Goto A, Abe Y, Kobayashi N, Kubota K, Saito S, Nakajima A. Alcohol consumption is associated with
an increased risk of erosive esophagitis and Barrett's epithelium in Japanese men. BMC Gastroenterol
2008;8:58.
15. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
16. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
17. Casparie M, Tiebosch AT, Burger G, Blauwgeers H, van de Pol A, van Krieken JH, Meijer GA. Pathology
databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide
histopathology and cytopathology data network and archive. Cell Oncol 2007;29:19-24.
18. Hughes LA, Schouten LJ, Goldbohm RA, van den Brandt PA, Weijenberg MP. Self-reported clothing size
as a proxy measure for body size. Epidemiology 2009;20:673-676.
19. Nevo table: Dutch food composition table, 1986-1987. (Dutch). Voorlichtingbureau Voor de Voeding,
1986.
20. Goldbohm RA, van den Brandt PA, Brants HA, van't Veer P, Al M, Sturmans F, Hermus RJ. Validation of
a dietary questionnaire used in a large-scale prospective cohort study on diet and cancer. Eur J Clin
Nutr 1994;48:253-265.
21. Goldbohm RA, van 't Veer P, van den Brandt PA, van 't Hof MA, Brants HA, Sturmans F, Hermus RJ.
Reproducibility of a food frequency questionnaire and stability of dietary habits determined from five
annually repeated measurements. Eur J Clin Nutr 1995;49:420-429.
Thesis Jessie Steevens_v04.pdf
Chapter 2
48
22. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
23. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atcddd/welcome.
html (accessed 15 Oct, 2009).
24. Cox DR. Regression models and life-tables (with discussion). J R Stat Soc Ser B 1972;34:187-220.
25. Barlow WE. Robust variance estimation for the case-cohort design. Biometrics 1994;50:1064-1072.
26. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrica 1982;69:
239-241.
27. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:
1503-1511.
28. Kubo A, Corley DA. Body mass index and adenocarcinomas of the esophagus or gastric cardia: a
systematic review and meta-analysis. Cancer Epidemiol Biomarkers Prev 2006;15:872-878.
29. Cook MB, Greenwood DC, Hardie LJ, Wild CP, Forman D. A systematic review and meta-analysis of the
risk of increasing adiposity on Barrett's esophagus. Am J Gastroenterol 2008;103:292-300.
30. Reid BJ, Li X, Galipeau PC, Vaughan TL. Barrett's oesophagus and oesophageal adenocarcinoma: time
for a new synthesis. Nat Rev Cancer 2010;10:87-101.
31. Murray L, Romero Y. Role of obesity in Barrett's esophagus and cancer. Surg Oncol Clin N Am
2009;18:439-452.
32. Corley DA, Kubo A, Levin TR, Block G, Habel L, Zhao W, Leighton P, Quesenberry C, Rumore GJ, Buffler
PA. Abdominal obesity and body mass index as risk factors for Barrett's esophagus. Gastroenterology
2007;133:34-41; quiz 311.
33. Steevens J, Schouten LJ, Goldbohm RA, van den Brandt PA. Alcohol consumption, cigarette smoking
and risk of subtypes of oesophageal and gastric cancer: a prospective cohort study. Gut 2010;59:39-48.
34. Freedman ND, Abnet CC, Leitzmann MF, Mouw T, Subar AF, Hollenbeck AR, Schatzkin A. A prospective
study of tobacco, alcohol, and the risk of esophageal and gastric cancer subtypes. Am J Epidemiol
2007;165:1424-1433.
35. Veugelers PJ, Porter GA, Guernsey DL, Casson AG. Obesity and lifestyle risk factors for
gastroesophageal reflux disease, Barrett esophagus and esophageal adenocarcinoma. Diseases of the
Esophagus 2006;19:321-328.
36. Gatenby PA, Caygill CP, Ramus JR, Charlett A, Watson A. Barrett's columnar-lined oesophagus:
demographic and lifestyle associations and adenocarcinoma risk. Dig Dis Sci 2008;53:1175-1185.
37. Lindblad M, Rodriguez LA, Lagergren J. Body mass, tobacco and alcohol and risk of esophageal, gastric
cardia, and gastric non-cardia adenocarcinoma among men and women in a nested case-control study.
Cancer Causes Control 2005;16:285-294.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of
Barrett's esophagus:
the Netherlands Cohort Study
Jessie Steevens
Leo J Schouten
Ann LC Driessen
Clément JR Huysentruyt
Yolande CA Keulemans
R Alexandra Goldbohm
Piet A van den Brandt
Cancer Causes and Control 2010; in press
3
49
Thesis Jessie Steevens_v04.pdf
Chapter 3
50
ABSTRACT
Objective
To investigate the association between selenium and the risk of Barrett’s esophagus
(BE), the precursor lesion of esophageal adenocarcinoma.
Methods
Data from the prospective Netherlands Cohort Study were used. This cohort study was
initiated in 1986, when 120,852 subjects aged 55-69 years completed a questionnaire
on dietary habits and lifestyle, and provided toenail clippings for the determination of
baseline selenium status. After 16.3 years of follow-up, 253 BE cases (identified
through linkage with the nationwide Dutch pathology registry) and 2039 subcohort
members were available for case-cohort analysis. Cox proportional hazards models
were used to calculate incidence rate ratios (RR).
Results
The multivariable adjusted RR for the highest versus the lowest quartile of toenail
selenium was 1.06 (95% CI 0.71-1.57). No dose-response trend was seen (p-
trend=0.99). No association was found in subgroups defined by sex, smoking status,
body mass index (BMI), or intake of antioxidants. For BE cases that later progressed to
high-grade dysplasia or adenocarcinoma, the RR for a selenium level above the median
vs. below the median was 0.64 (95% CI 0.24-1.76).
Conclusions
In this large prospective cohort study, we found no evidence of an association between
selenium and risk of BE.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
51
INTRODUCTION
Barrett’s esophagus (BE) is a disease of the distal esophagus. Normally, the esophagus
is lined with stratified squamous epithelium, which is replaced by a single layer of
columnar epithelium (metaplasia) in BE patients. Diagnostic criteria for BE differ across
the world. In the USA, the presence of goblet cells (indicating specialized intestinal
metaplasia, SIM) is required for the diagnosis of BE,1 while in the UK any type of
metaplasia is sufficient for the diagnosis of BE.2 In the Netherlands, both definitions
have been used by different pathologists over time.
BE is most common in middle-aged Caucasian men 3 and in the Netherlands, the
incidence is rising among men and women.4 BE has primarily been of interest because
patients are at increased risk to develop adenocarcinoma of the esophagus. The
reported risk of esophageal adenocarcinoma in BE patients is highly variable between
publications and was found in a recent meta-analysis to be between 4.1/1,000 and
6.1/1,000 person-years.5 Patients in whom high-grade dysplasia is present in the BE
tissue are at highest risk of developing esophageal adenocarcinoma.6
Gastroesophageal reflux disease has been identified as a strong risk factor for BE,
and possibly overweight or obesity are risk factors as well.7 Lifestyle factors, including
diet, may also play a role in BE etiology,7-9 but little information is available on this
topic.
One dietary factor of interest is selenium, a trace element, which has been
investigated for its possible role in cancer etiology. Selenium is involved in various
anticarcinogenic processes and may act at a number of stages in cancer
development.10 Selenium is incorporated into some selenoproteins including
glutathione peroxidases (GPx), which are antioxidant enzymes and protect against
oxidative damage. Selenium is further involved in alteration of DNA methylation,
blockage of the cell cycle, induction of apoptosis and inhibition of angiogenesis.10
Selenium has been associated with several cancers: it may protect against prostate,
lung, gastric, and colorectal cancer.11-13 Moreover, we recently found evidence
suggestive of an inverse association with esophageal adenocarcinoma. This association
was found in women, never smokers, and persons with a low intake of antioxidants.14
We identified only one study that looked into the association between selenium and
BE. That study found a lower average serum selenium concentration among patients
with BE than among controls, indicating a possible inverse association. Unfortunately,
no measure of relative risk or multivariable adjusted results were presented. Two
cross-sectional studies among BE cases in the USA found indications for an inverse
association between serum selenium and markers of progression of BE.15,16 Further, a
recent study reported promotor hypermethylation of the gene coding for the
selenoprotein GPx3 in Barrett’s esophagus tissue, thereby epigenetically inactivating
this gene.17 We hypothesize that selenium status is inversely associated with risk of BE,
if selenium is involved early in the process of carcinogenesis.
Thesis Jessie Steevens_v04.pdf
Chapter 3
52
The selenium content of our food may vary considerably between varieties of the
same type of food, depending on the soil where the food was grown.18 For that reason,
selenium intake from diet cannot be estimated reliably using questionnaires. Selenium
status is therefore often measured using biomarkers. In this study, we used toenails as
a biomarker of selenium intake. Toenails are a suitable biomarker, because they reflect
the intake of selenium for a period up to one year.19,20
Within the prospective Netherlands Cohort Study on diet and cancer, we
investigated whether toenail selenium levels were inversely associated with risk of BE.
MATERIALS AND METHODS
Study design and participants
The prospective Netherlands Cohort Study on diet and cancer was started in
September 1986, when 58,279 Dutch men and 62,573 women aged 55-69 were
enrolled. The subjects were selected at random from 204 Dutch municipal registries.
All cohort members completed a self-administered questionnaire and were requested
to provide toenail clippings at baseline. The study was described in detail before.21
A case-cohort approach22 was used for data processing and analysis for efficiency;
case subjects were derived from the entire cohort, and the number of person-years at
risk for the entire cohort was estimated from a subcohort of 3,500 subjects who were
selected at random from the full cohort at baseline.
The subcohort was followed-up for vital status and after 16.3 years (September
1986 to December 2002), only one male subcohort member was lost to follow-up. We
excluded subcohort members who reported having prevalent BE or cancer (other than
skin cancer) at baseline (Figure 3.1).
The Medical Ethics Committee of Maastricht University, the Netherlands, has
approved the study.
Follow-up for BE incidence
Incident BE cases in the total cohort were detected by computerized record linkage to
the nationwide network and registry of histopathology and cytopathology in the
Netherlands (PALGA).23 This network was founded in 1971, and an increasing number
of laboratories joined PALGA such that it has a nationwide coverage since 1991. Due to
this incomplete coverage, we may have missed BE cases diagnosed between baseline
(1986) and 1991. We calculated how many cases we may have missed by multiplying
the mean number of cases per year during the period a lab was connected by the
number of years the lab was not connected to PALGA. This was done for each lab. This
way we calculated that we may have missed approximately 3% of BE cases. At each of
the 64 pathology laboratories in the Netherlands, excerpts of all pathology reports are
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
53
generated automatically. Each excerpt contains a so-called PALGA diagnosis that
describes topography, morphology, function, procedure, and disease. These excerpts
are transferred to a central databank.23 The linkage with PALGA was carried out for
16.3 years of follow-up.24 All links were then manually checked for false-positives.
Thereafter, one pathologist (A.L.C.D.) and one pathologist in training (C.J.R.H.),
who were blinded to the exposure status (i.e. the selenium status) of the cases,
reviewed the excerpts of all pathology records to extract information on the initial
date of diagnosis of BE, the type of metaplasia, and the degree of dysplasia. Excluded
were BE cases with an uncertain diagnosis, or a diagnosis specifying the presence of
only non-intestinal type metaplasia. Also excluded were cases with a diagnosis of
esophageal or gastric cancer before or less than a half year after the diagnosis of BE,
and BE cases that were prevalent at baseline (Figure 3.1).
Netherlands cohort study on diet and cancer (58,279 men and 62,573 women)
↓ ↓
Subcohort randomly
drawn from total cohort Record linkage with PALGA until 31-12-2002
↓ ↓
2,376 reports from 1,185 cases
Exclusion of false-positive links
↓
1,954 reports from 974 cases
Review by pathologist: exclusion of uncertain diagnoses
↓
3,500 868 cases
Exclusion of prevalent cancer and Barrett's esophagus cases at baseline
↓ ↓
792 cases
Exclusion of cases with esophageal or gastric cancer before or <½
year after Barrett's diagnosis, and cases with only non-intestinal
metaplasia
↓ ↓
3,342
subcohort members
456 Barrett's esophagus
cases with specialized
intestinal metaplasia
626 Barrett's esophagus cases with
specialized intestinal metaplasia or
unknown metaplasia
Exclusion if toenail clippings were not provided, <10 mg of toenail clippings were available for
selenium determination, or if problems occurred with selenium determination
↓ ↓ ↓
2,423 285 397
Exclusion if incomplete or inconsistent dietary data a
↓ ↓ ↓
2,294 285 397
Exclusion if missing data on confounders
↓ ↓ ↓
2,039 253 346
Figure 3.1 Flow diagram of subcohort members and Barrett’s esophagus cases on whom the analyses
were based. a For reasons of efficiency, toenail material of Barrett’s esophagus cases were not
sent to laboratory for instrumental neutron activation analysis for determination of selenium
levels, if they had incomplete or inconsistent questionnaire data. Therefore, this exclusion
criterion did not anymore lead to exclusion of cases at this stage.
Thesis Jessie Steevens_v04.pdf
Chapter 3
54
Two definitions of BE were used: our primary case definition included only
subjects with esophageal SIM. The secondary case definition included subjects (a)
fulfilling the primary case definition or (b) with a pathology report stating ‘Barrett’s’,
without a description of the type of metaplasia. Furthermore, BE cases may progress
to high-grade dysplasia or esophageal adenocarcinoma. We followed the BE cases until
December 31, 2002 and we identified the subgroup of BE cases who showed one of
these complications. For analyses of BE cases who progressed to esophageal
adenocarcinoma, we selected only BE cases in whom the adenocarcinoma was
diagnosed more than a half year after the diagnosis of BE. This time lag was chosen to
be more sure that this was a new diagnosis of cancer that followed the BE.
Exposure data
In our cohort, about 75% of the subjects provided toenail clippings. Toenail selenium
determinations were carried out by the Reactor Institute Delft (Delft University of
Technology, Delft, the Netherlands). The determination was based on instrumental
neutron activation analysis of the 77mSe isotope (half-life 17.5 s). Each sample went
through 6 cycles of 17 s irradiation at a thermal neutron flux of 3 x 1016 m
-2s-1, 3 s
decay, and 17 s counting at 1 cm from a 40% germanium detector. This method, its
accuracy and precision, and the use in the Netherlands Cohort Study have been
described in more detail previously.25-28
In 1992, the toenail selenium determinations for the subcohort members were
carried out (for the purpose of analysis of the association between selenium and risk of
several cancers). In 2008, the toenail selenium determinations were carried out for the
BE cases. In 1992, the ‘SBP’ facility was used for instrumental neutron activation
analysis, and since 1996 the ‘CAFIA’ facility has been used. To assess the comparability
of these two methods, the toenail selenium levels for 40 subcohort members were
assessed in 1996 with the ‘CAFIA’ facility in addition to the original assessment with
the ‘SBP’ facility.25 The mean (SD) selenium level assessed by the ‘CAFIA’ facility [0.552
(0.05) µg/g] was comparable with mean selenium levels assessed by the ‘SBP’ facility
[0.551 (0.04) µg/g] for these subjects. The Pearson correlation coefficient between
toenail selenium levels assessed by the ‘CAFIA’ facility and those estimated by the
‘SBP’ facility was 0.95 (p<0.01).25 It was concluded that both methods were
comparable.
In 1992, all toenail clippings provided were sent to the Reactor Institute Delft for
selenium determination. This determination however, yields unreliable results if the
nails weigh <10 mg and these measurements were thus excluded. In 2007, we
discovered that toenail clippings can also be used as a source of DNA.29 We therefore
separated and saved 10-20 mg of toenail clippings from the BE cases. If afterwards,
more than 10 mg of toenail clippings were left, we sent them to the Reactor Institute
Delft for selenium determination. In case there were problems with the determination
of toenail selenium the subject was excluded (Figure 3.1).
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
55
The self-administered questionnaire, which was filled out by all cohort members
at baseline, consisted of a 150-item food frequency questionnaire (FFQ) and other
questions, e.g. on smoking, education, height, weight, and use of medication. The FFQ
asked about habitual consumption in the year before the start of the study. We
calculated mean daily nutrient intakes using the Dutch food-composition table 30. The
questionnaire data were key entered and processed in a standardized manner, blinded
with respect to cases/subcohort status to minimize observer bias in coding and
interpretation of the data.
Statistical analysis
To evaluate the potential influence of prediagnostic BE at baseline on toenail selenium
levels, cases were categorized according to the year of follow-up in which they were
diagnosed. The mean toenail selenium levels of the cases according to the year of
follow-up were compared and differences were tested using a t-test. For the t-test,
selenium levels were ln-transformed to normalize the distribution. For the case-cohort
analysis, toenail selenium levels were categorized into quartiles according to the
distribution in the subcohort. For continuous analysis, a 0.06 µg/g increment in toenail
selenium was chosen. This is equal to the average size of the two central quartiles.
We excluded subjects who had inconsistent or incomplete dietary questionnaire
data, because dietary data were needed as potential confounders and in subgroup
analyses.31 Subjects with missing data on the confounders were also excluded (Figure
3.1).
The following variables were considered confounders and included in the
multivariable regression model: age, sex, cigarette smoking (current yes/no, number of
cigarettes/day, and number of smoking years), alcohol consumption (g/day), and body
mass index (kg/m2). The following variables were considered potential confounders,
but were not included in the models because they did not change the incidence rate
ratio (RR) by >5%: highest level of education, family history of esophageal or gastric
cancer, reported long-term (>0.5 years) use of nonsteroidal anti-inflammatory drugs or
aspirin, or lower esophageal sphincter relaxing medication,32 non-occupational
physical activity, daily intakes of the antioxidants vitamin C, vitamin E, α-carotene, β-
carotene, β-cryptoxanthin, lycopene, and lutein/ zeaxanthin. Figure 3.1 shows that
complete data were available for 2039 subcohort members, 253 BE cases fulfilling the
primary case definition and 346 BE cases fulfilling the secondary case definition.
Multivariable adjusted RR and corresponding 95% confidence intervals (CI) were
estimated using Cox proportional hazards models 33 in Stata 9.2 (StataCorp, College
Station, Texas, USA). Standard errors were estimated using the robust Huber-White
sandwich estimator to account for additional variance introduced by sampling from
the cohort. This method is equivalent to the variance-covariance estimator by
Barlow.34 We tested the proportional hazards assumption using the scaled Schoenfeld
residuals.35 Tests for dose-response trends were assessed by fitting ordinal exposure
Thesis Jessie Steevens_v04.pdf
Chapter 3
56
variables as continuous terms. Two-sided P values are reported throughout the article.
The significance level α was set at 0.05.
We investigated possible interactions between toenail selenium status and sex,
cigarette smoking status, BMI, and intakes of several antioxidants by estimating RR in
strata of these exposures. The p value for interaction was assessed by including a
cross-product term in the model. To evaluate potential differences in the association
during early and late follow-up, we stratified our analysis by follow-up period. Finally,
we analyzed the subgroup of BE cases that progressed to high-grade dysplasia and/or
esophageal adenocarcinoma during follow-up until 2002, as this is the group most
relevant with respect to disease burden. Because of the low number of cases in this
subgroup (n=21), we used two instead of four exposure categories: above and below
the median selenium concentration.
RESULTS
Descriptives
Table 3.1 shows the mean toenail selenium levels of BE cases, stratified by two-year
follow-up periods in which they were diagnosed. The toenail selenium levels in the first
two follow-up years were lower than in later follow-up years, and this difference was
borderline statistically significant (p=0.07) for BE cases with SIM. As this difference may
indicate an effect of prediagnostic disease on toenail selenium levels, we decided to
exclude the first two years of follow-up from the analysis to prevent bias.
Table 3.1 Toenail selenium levels (µg/g) in Barrett's esophagus cases according to sex and time between
baseline and diagnosis; Netherlands Cohort Study (1986-2002, n = 120,852).
Barrett's esophagus
SIM SIM or unknown metaplasia
Casesa Toenail selenium level (µg/g) Toenail selenium level (µg/g)
No. cases mean SD p
b No. cases mean SD p
b
All cases 285 0.563 0.097 397 0.562 0.112
Men 163 0.556 0.100 226 0.558 0.127
Women 122 0.571 0.093 171 0.568 0.087
Follow-up year in which case was diagnosed
0-2 5 0.488 0.047 0.07 10 0.536 0.099 0.39
>2-4 17 0.531 0.086 31 0.566 0.210
>4-6 17 0.540 0.087 28 0.553 0.091
>6-8 38 0.557 0.058 54 0.552 0.068
>8-10 50 0.573 0.113 71 0.565 0.102
>10-12 55 0.555 0.096 74 0.561 0.120
>12-14 54 0.585 0.107 69 0.572 0.103
>14-17 49 0.568 0.099 60 0.565 0.094
a Mean (SD) selenium levels in subcohort members were 0.547 (0.126) µg/g for men (n=1212) and 0.575
(0.109) µg/g for women (n=1244); b T-test of mean toenail selenium levels (ln-transformed) for cases
diagnosed in the first 2 years of follow-up versus levels for cases diagnosed during the remainder of follow-
up years.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
57
Characteristics of the subcohort and BE cases are described in Table 3.2.
Subcohort members had a median toenail selenium level of 0.553 μg/g, with lower
levels in men (median 0.539 μg/g) than in women (0.564 μg/g). The median toenail
selenium levels in the two BE case groups were comparable (0.552 and 0.553 μg/g).
With respect to other characteristics, there were more men and former smokers
among the cases, when compared with the subcohort. Furthermore, cases had
somewhat lower intakes of carotenoids, and were more likely to be long-term users of
non-steroidal anti-inflammatory drugs/aspirin and lower esophageal sphincter relaxing
medication.
Main analysis
Table 3.3 presents multivariable adjusted RR of BE according to toenail selenium levels.
The RR of BE with SIM was 1.06 (95% CI: 0.71, 1.57) for the highest versus the lowest
quartile of toenail selenium, and no dose-response trend was observed (p trend=0.99).
Multivariable analyses stratified by sex showed some differences in associations for
men (RR quartile 4 vs. quartile 1 = 1.31, 95% CI: 0.78, 2.21) and women (RR quartile 4
vs. quartile 1=0.85, 95% CI: 0.46, 1.57), but the interaction between toenail selenium
and sex was not statistically significant (p interaction=0.18). The results were all
comparable to the age- and sex-adjusted results (Table 3.3). When we used the less
stringent secondary case definition of BE, the results were very similar to those based
on the primary case definition (Table 3.3).
Interaction and subgroup analyses
When analyses were performed in strata of smoking status, body mass index, or daily
intake of antioxidants, no statistically significant or important differences in
associations were observed between strata, and the RR were all around unity. Again,
results were similar for the two BE case groups (see Supplemental Table 3.1). Analyses
stratified by follow-up period (early vs. late follow-up) showed no statistically
significant associations or dose-response trends in either period (data not shown).
An inverse association was found between toenail selenium and risk of BE that
progressed to high-grade dysplasia and/or adenocarcinoma (Table 3.4). The
multivariable RR for a toenail selenium level above the median vs. below the median
was 0.64, although the confidence interval was relatively wide (95% CI: 0.24, 1.76), due
to the low case number in this analysis (n=21).
Thesis Jessie Steevens_v04.pdf
Chapter 3
58
Table 3.2 Characteristics of cases and subcohort members in the Netherlands Cohort Study (1986-2002,
n=120,852).
Subcohort Barrett's esophagus cases
Characteristic (n=2039)a SIM (n=253)a SIM or unknown
metaplasia (n=346)a
Median (IQR) Median (IQR) Median (IQR)
Toenail selenium level (µg/g)
total 0.553 (0.498-0.613) 0.552 (0.501-0.615) 0.553 (0.500-0.607)
men 0.539 (0.483-0.602) 0.551 (0.495-0.615) 0.544 (0.492-0.605)
women 0.564 (0.514-0.623) 0.555 (0.508-0.622) 0.565 (0.509-0.613)
Mean (SD) b Mean (SD) b Mean (SD) b
Age at baseline (years) 61.2 (4.2) 60.8 (4.2) 61.1 (4.1)
Men (%) 49 59 58
Cigarette smoking status
never smoker (%) 38 30 31
former smoker (%) 37 51 48
current smoker (%) 25 19 21
Ever cigarette smokers:
Frequency of cigarette smoking (n/day) 15.2 (10.3) 16.6 (11.6) 16.5 (11.1)
Duration of cigarette smoking (years) 31.0 (12.2) 30.4 (11.8) 30.7 (12.0)
Ethanol intake (g/day) 10.2 (14.3) 10.8 (15.0) 10.8 (15.0)
Body mass index (kg/m2) 25.0 (3.1) 25.3 (2.9) 25.3 (2.8)
Non-occupational physical activity
(min/day)
73 (58) 69 (53) 70 (55)
Highest level of education
primary (%) 27 27 30
lower vocational (%) 22 20 20
secondary and medium vocational (%) 37 36 35
university and higher vocational (%) 14 17 15
Vitamin C intake (mg/day) 104 (42) 101 (44) 100 (43)
Vitamin E intake (mg/day) 14 (6) 14 (6) 13 (6)
α-carotene intake (µg/day) 711 (588) 670 (502) 670 (512)
β-carotene intake (µg/day) 3001 (1589) 2821 (1382) 2805 (1398)
β-cryptoxanthin intake (µg/day) 181 (165) 177 (167) 176 (166)
Lycopene intake (µg/day) 1214 (1773) 1203 (2469) 1125 (2202)
Lutein/zeaxanthin intake (µg/day) 2518 (1079) 2353 (904) 2341 (925)
Family history of esophageal or gastric
cancer (%)
8 8 8
Use of non-steroidal anti-inflammatory
drugs and aspirin (%) c
7 11 10
Use of lower esophageal sphincter
relaxing medication (%) c
14 19 20
IQR, interquartile range. a Presented are the number of subcohort members or cases with complete data on
toenail selenium level, age, sex, cigarette smoking (current yes/no, number of cigarettes smoked daily,
number of smoking years), alcohol consumption and body mass index. Subcohort members and cases with
incomplete or inconsistent questionnaire data are excluded; b For categorical variables a percentage is
presented; c Self reported use during more than 0.5 year.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
59
Thesis Jessie Steevens_v04.pdf
Table 3.3 Incidence rate ratios of Barrett's esophagus according to toenail selenium levels; Netherlands Cohort Study (1986-2002, n=120,852).
Barrett's esophagus
Subcohort SIM SIM or unknown metaplasia
Age- and sex-adjusted Multivariable adjusted a Age- and sex-adjusted Multivariable adjusted a
Median
(µg/g)
person time at
risk (years)
no.
cases
RR 95% CI RR 95% CI no.
cases
RR 95% CI RR 95% CI
All subjects
Quartiles of toenail selenium (boundaries in µg/g)
1 (≤0.498) 0.458 6170 59 1 reference 1 reference 83 1 reference 1 reference
2 (≤0.552) 0.525 6548 69 1.15 0.79, 1.67 1.08 0.74, 1.58 90 1.06 0.77, 1.48 1.02 0.73, 1.42
3 (≤0.613) 0.580 6659 59 0.99 0.67, 1.45 0.92 0.62, 1.37 89 1.06 0.76, 1.47 1.01 0.73, 1.42
4 (>0.613) 0.667 6527 66 1.15 0.78, 1.68 1.06 0.71, 1.57 84 1.03 0.74, 1.44 0.98 0.69, 1.39
p trend b 0.69 0.99 0.87 0.90
continuous (0.06 µg/g increment)c25903 253 1.01 0.96, 1.07 1.00 0.94, 1.06 346 1.01 0.96, 1.07 1.01 0.95, 1.07
p interaction
d 0.17 p interaction 0.18 p interaction 0.06 p interaction 0.06
Men
Quartiles of toenail selenium (boundaries in µg/g)
1 (≤0.498) 0.458 3658 39 1 reference 1 reference 55 1 reference 1 reference
2 (≤0.552) 0.525 3078 37 1.10 0.68, 1.79 1.06 0.64, 1.76 54 1.14 0.75, 1.73 1.10 0.72, 1.70
3 (≤0.613) 0.580 2830 34 1.10 0.67, 1.80 1.07 0.64, 1.78 44 1.02 0.66, 1.57 0.99 0.63, 1.55
4 (>0.613) 0.667 2596 38 1.35 0.84, 2.19 1.31 0.78, 2.21 48 1.21 0.79, 1.86 1.18 0.74, 1.87
p trend 0.26 0.33 0.50 0.61
continuous (0.06 µg/g increment) 12162 148 1.03 0.98, 1.09 1.03 0.97, 1.09 201 1.04 0.98, 1.09 1.03 0.98, 1.09
Women
Quartiles of toenail selenium (boundaries in µg/g)
1 (≤0.498) 0.458 2512 20 1 reference 1 reference 28 1 reference 1 reference
2 (≤0.552) 0.525 3469 32 1.16 0.64, 2.09 1.14 0.63, 2.08 36 0.93 0.55, 1.58 0.93 0.54, 1.58
3 (≤0.613) 0.580 3829 25 0.82 0.44, 1.51 0.77 0.41, 1.43 45 1.05 0.63, 1.75 1.01 0.61, 1.68
4 (>0.613) 0.667 3931 28 0.89 0.49, 1.63 0.85 0.46, 1.57 36 0.82 0.48, 1.39 0.81 0.47, 1.37
p trend 0.42 0.33 0.55 0.49
continuous (0.06 µg/g increment) 13742 105 0.95 0.85, 1.06 0.94 0.84, 1.05 145 0.94 0.86, 1.03 0.93 0.85, 1.02
a Adjusted for age (years), sex, cigarette smoking (current smoking status (yes/no), frequency (number of cigarettes/day), and duration (years)), alcohol consumption
(g/day), body mass index (kg/m²); b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model;
c The 0.06 µg/g increment for continuous analyses is equal to the average size of the two central quartiles; d p-value for interaction between sex and toenail selenium
level, based on cross product term in the Cox proportional hazard model.
Chapter 3
60
Table 3.4 Incidence rate ratios of Barrett's esophagus cases who developed high grade dysplasia or
adenocarcinoma according to toenail selenium levels; Netherlands Cohort Study (1986-2002,
n=120,852).
Subcohort Barrett's esophagus with high grade dysplasia
and/or adenocarcinoma
Age- and
sex-adjusted
Multivariable
adjusted a
Median within
category µg/g
person time
at risk
(years)
no.
cases
RR 95% CI RR 95% CI
Categories of toenail selenium, based on a median split (boundary in µg/g)
1 (≤0.552) 0.497 12717 13 1 reference 1 reference
2 (>0.552) 0.615 13185 8 0.63 0.26, 1.56 0.64 0.24, 1.76
Continuous (0.06 μg/g
Increment)b
25903 21 1.10 0.99, 1.22 1.11 0.99, 1.24
a adjusted for age (years), sex, cigarette smoking (current smoking status (yes/no), frequency (number of
cigarettes/day), and duration (years)), alcohol consumption (g/day), body mass index (kg/m²); b The RRs of
the continuous analyses are influenced by one case with a very high toenail selenium level (1.612 μg/g).
When this case was excluded, the age- and sex-adjusted RR was 0.96 (95% CI 0.79-1.17) and the
multivariable adjusted RR was 0.96 (95% CI 0.77-1.20).
DISCUSSION
To our knowledge, this is the second investigation, and first cohort study, into the
possible role of selenium in the etiology of BE, a precursor lesion of esophageal
adenocarcinoma. The results from this prospective cohort study do not support our
hypothesis that an inverse association might exist. Neither did we find evidence of an
association in subgroups defined by sex, smoking status, BMI, intake of antioxidants or
follow-up period.
The Dutch population that was used in this study, has a low to moderate toenail
selenium level (mean: 0.564 µg/g in the subcohort) when compared with mean levels
found in other populations (general population or control subjects): China (Sichuan
province) 0.211 µg/g,36 Finland 0.47 µg/g,37 USA 0.83-0.92 µg/g,38 Colombia
0.945 µg/g,39 and USA (South Dakota/Wyoming) 1.517 µg/g.19
In the Netherlands, both the USA and UK case definitions of BE may have been
used. The use of these definitions may have changed over time and different
pathologists may have used different definitions.
The first strength of our study is its prospective character, which brings the
advantage of measuring the selenium status before the diagnosis of BE. This method
makes it unlikely that the exposure was changed due to the disease or knowledge of
the diagnosis, and therefore it lowers the chance of reversed causation. A second
strength is the size of the study. This study had 80% power to detect an RR of 0.63 for
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
61
all BE cases and an RR of 0.59 for BE cases with SIM.40 Thirdly, we had the possibility to
adjust for confounding by several lifestyle factors. A fourth strength is the use of
toenail clippings for the measurement of selenium exposure, because selenium levels
in clippings from all toes represent an exposure period of up to one year. Measuring
long-term exposure is important as it more likely represents the exposure in the
etiologically relevant time window compared with a measure of short-term exposure.
A limitation of our study is the lack of information about gastroesophageal reflux
disease and medications related to this disease. In the baseline questionnaire one
question concerned the long-term use of medications and the disease for which these
were prescribed. However, we believe that the reported use of reflux medication and
antacids, and presence of gastroesophageal reflux disease were substantially lower
than the actual frequency. This prohibited us to use this information in the analysis.
Consequently, we were not able to evaluate if gastroesophageal reflux disease is a
confounder or intermediate factor in the relation between selenium and BE. We
therefore suggest that this study is replicated in a population for which information on
gastroesophageal reflux is available, to evaluate its possible role as confounder or
intermediate factor. A second limitation is the single measurement of the exposure.
Selenium exposure may have changed during follow-up due to changed selenium
levels in foods or changed dietary habits. Third, we did not have the opportunity to
verify the absence of BE in the subcohort members who were not diagnosed with BE.
BE usually occurs in patients with gastroesophageal reflux disease, but it has also been
described in asymptomatic individuals, who are therefore not diagnosed with the
disease. Three studies investigated the prevalence of BE among asymptomatic
individuals in the USA and reported prevalences ranging from 1% to 25%.41-43 However,
in our cohort study, any undiagnosed BE cases would most likely not have influenced
the RR. This is because the imperfect sensitivity (i.e. false-negatives) is combined with
good specificity (i.e. few false positives), and the chance of diagnosis is likely
independent of selenium status. The misclassification of the disease status then is non-
differential and will not influence the RR.44 Still, a consequence of undiagnosed cases is
a reduced power to detect an existing association. A fourth limitation, related to the
third, is caused by the incomplete coverage of the Netherlands by PALGA before 1991.
There may have been some BE cases that were diagnosed in laboratories when these
had not yet joined PALGA. If these cases were not followed-up, we missed their
diagnosis. If they were followed-up after the laboratory joined PALGA, the incidence
date we registered is more recent than the true incidence date. We assessed that we
may have missed 3% of the BE cases in our cohort at most.
If, in line with our null findings there is truly no association between toenail
selenium and the risk of BE, this does not preclude a possible association between
selenium and progression of BE to high-grade dysplasia or adenocarcinoma. We
observed an RR well below unity for subjects with a toenail selenium level above the
median in those BE cases who progressed to high-grade dysplasia and/or esophageal
adenocarcinoma. This RR, however, was based on few cases and was not statistically
Thesis Jessie Steevens_v04.pdf
Chapter 3
62
significant. In a previous analysis we found indications for an inverse association
between selenium and risk of esophageal adenocarcinoma in some subgroups:
women, never smokers, and low antioxidant consumers.14 These above-mentioned
observations are in agreement with observations by two other studies that found
indications for a role of selenium in later stages of the carcinogenesis of esophageal
adenocarcinoma.15,16 Selenium might be an interesting preventive agent for
progression to esophageal adenocarcinoma in patients with BE, but this requires a
larger body of evidence.
In conclusion, no evidence of an inverse association between toenail selenium
and risk of BE was found in this study. An inverse association might exist for BE cases
that progress to high-grade dysplasia and/or esophageal adenocarcinoma, but results
from other studies are needed before any firm conclusions can be drawn. Preferably,
future studies should have a prospective character and should have the possibility to
investigate the role of gastroesophageal reflux in this association. Also, it would be
informative to study the relation between selenium and risk of BE in a population with
selenium levels that are lower or higher compared with our study population.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
63
Supplemental Table 3.1 Association
a between toenail selenium levels and risk of Barrett's esophagus
stratified by cigarette smoking status, body mass index and intake levels of
antioxidants; Netherlands Cohort Study (1986-2002, n=120,852)
Subcohort Barrett's esophagus
SIM SIM or unknown metaplasia
person time
at risk
(years)
no.
cases
RR b 95% CI no.
cases
RR b 95% CI
Cigarette smoking status
Never smoker 10157 77 0.98 0.87, 1.10 108 0.95 0.86, 1.05
Former smoker 9358 128 1.00 0.92, 1.08 165 1.00 0.93, 1.08
Current smoker 6389 48 1.07 0.88, 1.29 73 1.14 0.91, 1.43
p interaction 0.78, p interaction 0.35,
Body mass index (kg/m²)
< 25 13930 131 1.01 0.96, 1.07 178 1.01 0.95, 1.06
≥ 25 11973 122 0.96 0.83, 1.10 168 0.99 0.86, 1.14
p interaction 0.73, p interaction 0.79,
Antioxidant intake
Vitamin C intake (mg/d)
low (≤ 97) 12700 139 1.01 0.95, 1.09 189 1.01 0.94, 1.07
high (> 97) 13204 114 0.99 0.90, 1.08 157 1.01 0.91, 1.12
p interaction 0.76, p interaction 0.90,
Vitamin E intake (mg/d)
low (≤ 12) 12901 113 1.01 0.90, 1.13 166 1.06 0.95, 1.19
high (> 12) 13002 140 0.99 0.92, 1.07 180 0.96 0.89, 1.04
p interaction 0.78, p interaction 0.16,
α-carotene intake (µg/d)
low (≤ 573) 12708 132 1.04 0.93, 1.16 183 1.03 0.94, 1.14
high (> 573) 13195 121 0.97 0.89, 1.06 163 0.99 0.90, 1.08
p interaction 0.37, p interaction 0.46,
β-carotene intake (µg/d)
low (≤ 2680) 12817 139 1.01 0.95, 1.08 196 1.01 0.95, 1.07
high (> 2680) 13086 114 0.97 0.88, 1.08 150 1.00 0.88, 1.12
p interaction 0.62, p interaction 0.78,
β-cryptoxanthin intake (µg/d)
low (≤ 130) 12669 129 1.01 0.91, 1.13 176 1.03 0.92, 1.15
high (> 130) 13234 124 0.99 0.92, 1.07 170 0.99 0.92, 1.06
p interaction 0.60, p interaction 0.48,
Lycopene intake (µg/d)
low (≤ 823) 12857 141 0.98 0.89, 1.07 203 0.96 0.89, 1.04
high (> 823) 13046 112 1.04 0.94, 1.14 143 1.08 0.97, 1.20
p interaction 0.46, p interaction 0.13,
Lutein/Zeaxanthin intake (µg/d)
low (≤ 2360) 12876 146 1.00 0.92, 1.08 198 0.99 0.92, 1.07
high (> 2360) 13027 107 1.01 0.91, 1.12 148 1.02 0.92, 1.14
p interaction 0.66, p interaction 0.60,
a Adjusted for age (years), sex, cigarette smoking (current smoking status (yes/no), frequency (number of
cigarettes/day), and duration (years)), alcohol consumption (g/day), body mass index (kg/m²), if applicable;
b RR per 0.06 µg/g increment of toenail selenium. This increment is equal to the average size of the two
central quartiles; c p-value for interaction between sex and toenail selenium level, based on cross product
term in the Cox proportional hazard model.
Thesis Jessie Steevens_v04.pdf
Chapter 3
64
REFERENCES
1. Wang KK, Sampliner RE. Updated guidelines 2008 for the diagnosis, surveillance and therapy of
Barrett's esophagus. Am J Gastroenterol 2008;103:788-797.
2. Playford RJ. New British Society of Gastroenterology (BSG) guidelines for the diagnosis and
management of Barrett's oesophagus. Gut 2006;55:442-443.
3. Falk GW. Barrett's esophagus. Gastroenterology 2002;122:1569-1591.
4. Post PN, Siersema PD, van Dekken H. Rising incidence of clinically evident Barrett's oesophagus in The
Netherlands: a nation-wide registry of pathology reports. Scand J Gastroenterol 2007;42:17-22.
5. Yousef F, Cardwell C, Cantwell MM, Galway K, Johnston BT, Murray L. The incidence of esophageal
cancer and high-grade dysplasia in Barrett's esophagus: a systematic review and meta-analysis. Am J
Epidemiol 2008;168:237-249.
6. Rastogi A, Puli S, El-Serag HB, Bansal A, Wani S, Sharma P. Incidence of esophageal adenocarcinoma in
patients with Barrett's esophagus and high-grade dysplasia: a meta-analysis. Gastrointest Endosc
2008;67:394-398.
7. Wild CP, Hardie LJ. Reflux, Barrett's oesophagus and adenocarcinoma: burning questions. Nat Rev
Cancer 2003;3:676-684.
8. Anderson LA, Watson RG, Murphy SJ, Johnston BT, Comber H, Mc Guigan J, Reynolds JV, Murray LJ.
Risk factors for Barrett's oesophagus and oesophageal adenocarcinoma: results from the FINBAR
study. World J Gastroenterol 2007;13:1585-1594.
9. Kubo A, Levin TR, Block G, Rumore GJ, Quesenberry CP, Jr., Buffler P, Corley DA. Dietary patterns and
the risk of Barrett's Esophagus. Am J Epidemiol 2008;167:839-846.
10. Rayman MP. Selenium in cancer prevention: a review of the evidence and mechanism of action. Proc
Nutr Soc 2005;64:527-542.
11. Navarro Silvera SA, Rohan TE. Trace elements and cancer risk: a review of the epidemiologic evidence.
Cancer Causes Control 2007;18:7-27.
12. World Cancer Research Fund, American Institute for Cancer Research. Food, nutrition, physical activity
and the prevention of cancer: a global perspective. AICR, 2007.
13. Clark LC, Combs GF, Jr., Turnbull BW, Slate EH, Chalker DK, Chow J, Davis LS, Glover RA, Graham GF,
Gross EG, Krongrad A, Lesher JL, Jr., Park HK, Sanders BB, Jr., Smith CL, Taylor JR. Effects of selenium
supplementation for cancer prevention in patients with carcinoma of the skin. A randomized
controlled trial. Nutritional Prevention of Cancer Study Group. Jama 1996;276:1957-1963.
14. Steevens J, van den Brandt PA, Goldbohm RA, Schouten LJ. Selenium status and the risk of esophageal
and gastric cancer subtypes: the Netherlands cohort study. Gastroenterology 2010;138:1704-1713.
15. Rudolph RE, Vaughan TL, Kristal AR, Blount PL, Levine DS, Galipeau PC, Prevo LJ, Sanchez CA,
Rabinovitch PS, Reid BJ. Serum selenium levels in relation to markers of neoplastic progression among
persons with Barrett's esophagus. J Natl Cancer Inst 2003;95:750-757.
16. Moe GL, Kristal AR, Levine DS, Vaughan TL, Reid BJ. Waist-to-hip ratio, weight gain, and dietary and
serum selenium are associated with DNA content flow cytometry in Barrett's esophagus. Nutr Cancer
2000;36:7-13.
17. Peng DF, Razvi M, Chen H, Washington K, Roessner A, Schneider-Stock R, El-Rifai W. DNA
hypermethylation regulates the expression of members of the Mu-class glutathione S-transferases and
glutathione peroxidases in Barrett's adenocarcinoma. Gut 2009;58:5-15.
18. Willett WC, Buzzard IM. Foods and nutrients. In: Willett WC, ed. Monographs in Epidemiology and
Biostatistics. Volume 30. Nutritional epidemiology. New York: Oxford Universtiy Press, 1998:18-32.
19. Longnecker MP, Stram DO, Taylor PR, Levander OA, Howe M, Veillon C, McAdam PA, Patterson KY,
Holden JM, Morris JS, Swanson CA, Willett WC. Use of selenium concentration in whole blood, serum,
toenails, or urine as a surrogate measure of selenium intake. Epidemiology 1996;7:384-390.
20. Hunter DJ, Morris JS, Chute CG, Kushner E, Colditz GA, Stampfer MJ, Speizer FE, Willett WC. Predictors
of selenium concentration in human toenails. Am J Epidemiol 1990;132:114-122.
21. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
22. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
Thesis Jessie Steevens_v04.pdf
Toenail selenium status and the risk of Barrett’s esophagus
65
23. Casparie M, Tiebosch AT, Burger G, Blauwgeers H, van de Pol A, van Krieken JH, Meijer GA. Pathology
databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide
histopathology and cytopathology data network and archive. Cell Oncol 2007;29:19-24.
24. van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol
1990;19:553-558.
25. Zeegers MP, Goldbohm RA, Bode P, van den Brandt PA. Prediagnostic toenail selenium and risk of
bladder cancer. Cancer Epidemiol Biomarkers Prev 2002;11:1292-1297.
26. Bode P. Automation and quality assurance in the Neutron Activation Facilities in Delft. J Radioanal Nucl
Chem 2000;245:127-132.
27. van den Brandt PA, Goldbohm RA, van't Veer P, Bode P, Hermus RJ, Sturmans F. Predictors of toenail
selenium levels in men and women. Cancer Epidemiol Biomarkers Prev 1993;2:107-112.
28. van den Brandt PA, Goldbohm RA, van 't Veer P, Bode P, Dorant E, Hermus RJ, Sturmans F. A
prospective cohort study on toenail selenium levels and risk of gastrointestinal cancer. J Natl Cancer
Inst 1993;85:224-229.
29. van Breda SG, Hogervorst JG, Schouten LJ, Knaapen AM, van Delft JH, Goldbohm RA, van Schooten FJ,
van den Brandt PA. Toenails: an easily accessible and long-term stable source of DNA for genetic
analyses in large-scale epidemiological studies. Clin Chem 2007;53:1168-1170.
30. Nevo table: Dutch food composition table, 1986-1987. (Dutch). Voorlichtingbureau Voor de Voeding,
1986.
31. Goldbohm RA, van den Brandt PA, Brants HA, van't Veer P, Al M, Sturmans F, Hermus RJ. Validation of
a dietary questionnaire used in a large-scale prospective cohort study on diet and cancer. Eur J Clin
Nutr 1994;48:253-265.
32. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
33. Cox DR. Regression models and life-tables. J Roy Statistical Society 1972;34:187-220.
34. Barlow WE. Robust variance estimation for the case-cohort design. Biometrics 1994;50:1064-1072.
35. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrika
1982;69:239-241.
36. Gao S, Jin Y, Hall KS, Liang C, Unverzagt FW, Ji R, Murrell JR, Cao J, Shen J, Ma F, Matesan J, Ying B,
Cheng Y, Bian J, Li P, Hendrie HC. Selenium level and cognitive function in rural elderly Chinese. Am J
Epidemiol 2007;165:955-965.
37. Ovaskainen ML, Virtamo J, Alfthan G, Haukka J, Pietinen P, Taylor PR, Huttunen JK. Toenail selenium as
an indicator of selenium intake among middle-aged men in an area with low soil selenium. Am J Clin
Nutr 1993;57:662-665.
38. Garland M, Morris JS, Rosner BA, Stampfer MJ, Spate VL, Baskett CJ, Willett WC, Hunter DJ. Toenail
trace element levels as biomarkers: reproducibility over a 6-year period. Cancer Epidemiol Biomarkers
Prev 1993;2:493-497.
39. Koriyama C, Campos FI, Yamamoto M, Serra M, Carrasquilla G, Carrascal E, Akiba S. Toenail selenium
levels and gastric cancer risk in Cali, Colombia. J Toxicol Sci 2008;33:227-235.
40. Cai J, Zeng D. Sample size/power calculation for case-cohort studies. Biometrics 2004;60:1015-1024.
41. Fan X, Snyder N. Prevalence of Barrett's esophagus in patients with or without GERD symptoms: role
of race, age, and gender. Dig Dis Sci 2009;54:572-577.
42. Gerson LB, Shetler K, Triadafilopoulos G. Prevalence of Barrett's esophagus in asymptomatic
individuals. Gastroenterology 2002;123:461-467.
43. Ward EM, Wolfsen HC, Achem SR, Loeb DS, Krishna M, Hemminger LL, DeVault KR. Barrett's esophagus
is common in older men and women undergoing screening colonoscopy regardless of reflux
symptoms. Am J Gastroenterol 2006;101:12-17.
44. Rothman KJ, Greenland S. Precision and validity in epidemiologic studies. In: Rothman KJ, Greenland S,
eds. Modern epidemiology. Second ed. Philadelphia: Lippincott, 1998:115-134.
Thesis Jessie Steevens_v04.pdf
Chapter 3
66
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific
mortality in a population-based cohort
of patients with Barrett's esophagus
Leo J Schouten
Jessie Steevens
Clément JR Huysentruyt
Ceciel E Coffeng
Yolande CA Keulemans
Flora E van Leeuwen
Ann LC Driessen
Piet A van den Brandt
Submied
4
67
Thesis Jessie Steevens_v04.pdf
Chapter 4
68
ABSTRACT
Objective
Barrett’s esophagus (BE) is associated with an increased risk of esophageal
adenocarcinoma, but evidence with respect to incidence of other cancers and overall
mortality is inconsistent.
Methods
The Netherlands Cohort Study was initiated in 1986, when 120,852 participants aged
55-69 years at baseline were included. Until December 2002, 626 incident BE cases
(excluding non-intestinal metaplasia) were identified by record linkage with the
nationwide Pathology Registry. Cancer and mortality follow-up of this cohort of BE
cases was established by record linkage to the Netherlands Cancer Registry and
Statistics Netherlands. The expected number of cases was calculated using national
cancer incidence and mortality data.
Results
After a median follow-up of 5.7 years of the BE cohort, 13 esophageal and 5 gastric
cancer cases were recorded. The Observed/Expected (O/E) ratios for esophageal and
gastric cancer were 10.0 (95% Confidence Interval (CI), 5.3-17.1) and 1.8 (95% CI,
0.6-4.2), respectively.
Total cancer incidence (excluding esophageal and gastric cancer) in the BE cohort
was nearly statistically significantly increased: O/E-ratio 1.3 (95% CI, 1.0-1.6). Of the
subtypes, small intestinal and pancreatic cancer incidence were increased, but not
statistically significantly after exclusion of the first six months of follow-up.
During follow-up 225 BE cases died. Mortality of all causes (excluding esophageal
and gastric cancer) was not increased: O/E-ratio 1.0 (95% CI, 0.9-1.2), neither was
mortality according to main causes of death.
Conclusion
In this population-based cohort of BE cases esophageal cancer incidence was
increased. Total cancer incidence and overall mortality were not increased.
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
69
INTRODUCTION
Barrett’s esophagus (BE) is an abnormal appearing distal esophageal lining with
replacement of the normal squamous epithelium by specialized or intestinal columnar
epithelium.1 BE is thought to be a pre-cancerous condition and patients with BE have
an increased risk to develop esophageal adenocarcinoma.2 Although many studies
have investigated the risk of esophageal adenocarcinoma in patients with BE,3,4 most
follow-up studies are small and have been conducted in selected patient series. Larger
studies with complete follow-up and a strict definition of BE have found lower risks of
esophageal adenocarcinoma compared to studies with a smaller study-size.4,5
Incidence of gastric cancer was also increased in some studies, possibly due to
misclassification of junctional tumors.3 The number of studies examining incidence
rates of other cancers and mortality is limited. There are indications that colorectal
cancer risk is increased in patients with BE, but this may be due to selected patient
series and detection bias.6,7
Several studies have observed increased mortality rates among BE cases.8-11
However, the only population-based cohort that studied mortality did not observe an
increased risk.12
We therefore decided to study cancer incidence and overall mortality in the
population-based Netherlands Cohort Study on Diet and Cancer. By record-linkage to
the Netherlands nationwide pathology registry, cases of BE in the cohort were
identified. By follow-up of these cases we were able to study whether cancer incidence
and cause-specific mortality were increased in this population-based cohort of BE
cases.
PATIENTS AND METHODS
Study design and subjects
The Netherlands Cohort Study was started in September 1986, when 58,279 Dutch
men and 62,573 women aged 55-69 years were enrolled. The subjects were selected at
random from 204 Dutch municipal population registries. All cohort members
completed a self-administered questionnaire. The study design has been described in
detail before.13
Incident BE cases in the total cohort were identified by computerized record-
linkage to the nationwide network and registry of histopathology and cytopathology in
the Netherlands (PALGA) using the search term “Barrett” or the search terms
“metaplasia” combined with “esophagus”.5 The linkage was carried out for the period
until 31 December 2002, using the linkage protocol that we described before.14 All links
were checked for false-positives (see Figure 4.1). Thereafter, one pathologist (A.L.C.D.)
and one pathologist in training (C.J.R.H.) reviewed the available summary and
Thesis Jessie Steevens_v04.pdf
Chapter 4
70
conclusion of the pathology records from PALGA to extract information on the date of
diagnosis of BE, the type of metaplasia, and the presence and degree of dysplasia.
Excluded were BE cases with an uncertain diagnosis or with a diagnosis specifying
the presence of only non-intestinal metaplasia. Also excluded were BE cases with
prevalent cancer (self-reported) or prevalent BE (if known) in September 1986. To
eliminate the possibility that the diagnosis of BE was related to (preclinical) symptoms
of esophageal or gastric cancer, we excluded BE cases with a diagnosis of esophageal
or gastric cancer before or less than a half year after the diagnosis of BE (leaving 626
cases). This constituted a population-based cohort of BE cases. The Medical Ethics
Committee of Maastricht University, the Netherlands, has approved the study.
Netherlands Cohort Study on Diet and Cancer
(58,279 men and 62,573 women)
↓
Record linkage with PALGA until 31-12-2002
2376 reports from 1185 BE cases
↓
Exclusion of false-positive linkages
1954 reports from 974 BE cases
↓
Review by pathologist: exclusion of uncertain diagnoses
868 BE cases
↓
Review by pathologist: exclusion of non-intestinal metaplasia
742 BE cases
↓
Exclusion of prevalent cancer and BE cases at baseline
680 BE cases
↓
Exclusion of esophageal and gastric cancer cases
before or <½ year after the BE diagnosis
626 BE cases
↓ ↓ ↓
Exclusion of first half
year of follow-up
Exclusion of cancer and
BE cases
↓ ↓
Analysis of esophageal
and gastric cancer
incidence
Analysis of cancer
incidence
Analysis of mortality
605 BE cases 561 BE cases 626 BE cases
Figure 4.1 Flow diagram of Barrett’s esophagus (BE) cases on whom the analyses were based.
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
71
Follow-up for cancer incidence and cause-specific mortality
For the analyses with respect to esophageal and gastric cancer incidence, total cancer
incidence, and overall mortality, three different datasets were constructed from the BE
cohort. Because we had excluded all esophageal or gastric cancer in the first half year
of follow-up after BE diagnosis, the follow-up for esophageal and gastric cancer started
a half year after BE diagnosis (leaving 605 patients at risk). For the follow-up of cancer
incidence, BE cases with cancer (other than non-melanoma skin cancer) diagnosed
before the BE diagnosis were excluded (leaving 561 patients). No additional exclusion
criterion was necessary for the dataset for the mortality follow-up (626 patients).
Follow-up for vital status was established by record linkage with the Central
Bureau of Genealogy and the automated municipal population registries. Follow-up for
vital status was complete for all 626 members of the Barrett’s cohort at 1 May 2005.
Follow-up for cancer incidence in BE cases was performed by record linkage to the
Netherlands Cancer Registry and PALGA.14 For the cancer incidence analyses, follow-up
data were available until 31 December 2003. In this analysis only the first occurrence
of a primary cancer was counted.
We were able to obtain the cause of death from Statistics Netherlands (until 1
May 2005). Causes of death were coded in the International Classification of Diseases
(ICD)-9 until December 1995 and ICD-10 from January 1996. Only the primary cause of
death was used in the analysis.
Statistical analysis
Cancer incidence in the Barrett’s cohort was compared with the incidence in the Dutch
population. Time at risk began at diagnosis of BE (or a half year after BE diagnosis in
the esophageal and gastric cancer analysis) and ended on the date of cancer diagnosis,
the date of death, or the last date of the follow-up for cancer incidence (31 December
2003), whichever occurred first. The expected numbers of cancer, taking into account
the person-years of observation in the Barrett’s cohort, were computed with the use
of sex-, age-, and calendar year – specific cancer incidence rates from the Eindhoven
Cancer Registry up to 1990 or,15 for the period after 1990, from the Netherlands
Cancer Registry.16,17
Mortality in the Barrett’s cohort was compared with mortality in the Dutch
population. Time at risk began at diagnosis of BE and ended on the date of death or
the last date of the follow-up for mortality (1 May 2005), whichever occurred first. For
the mortality analysis sex-, age- and calendar year -specific mortality rates were
obtained from Statistics Netherlands.
The Observed/Expected (O/E) ratios for cancer incidence and cause-specific
mortality were calculated as the ratio of the observed number of cancer cases or
deaths to the expected number, and the 95% confidence intervals (CI) were calculated
based on the Poisson distribution.18 The absolute excess risk of cancer was calculated
by subtracting the expected number of cases from the observed number, divided by
Thesis Jessie Steevens_v04.pdf
Chapter 4
72
the person-years at risk. All analyses were conducted using a statistical program
written in SPSS developed by the Netherlands Cancer Institute.19
RESULTS
The dataset for esophageal and gastric cancer incidence included 605 BE cases, the
dataset for incidence of other cancers included 561 persons, while the dataset for
mortality follow-up counted 626 persons (see Table 4.1). The majority of the BE cases
was male and the mean age at diagnosis of Barrett’s esophagus was slightly higher
than 71 years. In 73-74% of the BE cases, the histology was specified as intestinal
metaplasia, while in the remaining cases the histology was unspecified. Dysplasia was
reported not to be present in 81% of the cases, while high-grade dysplasia was rare
(1%). The median follow-up duration was 5.7 years in the esophageal and gastric
cancer analysis, 5.4 years in the cancer incidence analysis and 6.8 years in the mortality
analysis.
Table 4.1 Baseline characteristics of cohort of patients with Barrett's esophagus for cancer incidence and
mortality analysis. Netherlands Cohort Study.
Esophageal and gastric
cancer analysisa
Cancer incidence
analysis b
Mortality
analysis c
Characteristic n (%) n (%) n (%)
Total 605 (100.0) 561 (100.0) 626 (100.0)
Sex
Males 325 (53.7) 292 (52.0) 341 (54.5)
Females 280 (46.3) 269 (48.0) 285 (45.5)
Age at diagnosis of Barrett's esophagus, mean
(SD)d
71.2 (5.3) 71.1 (5.3) 71.2 (5.3)
Incidence year
1986-1990 72 (11.9) 68 (12.1) 75 (12.0)
1991-1994 145 (24.0) 138 (24.6) 149 (23.8)
1995-1998 214 (35.4) 193 (34.4) 222 (35.5)
1999-2002 174 (28.8) 162 (28.9) 180 (28.8)
Histology
Intestinal metaplasia 446 (73.7) 415 (74.0) 456 (72.8)
Barrett’s esophagus NOS 159 (26.3) 146 (26.0) 170 (27.2)
Dysplasia
No dysplasia 493 (81.5) 457 (81.5) 508 (81.2)
Low-grade dysplasia 92 (15.2) 88 (15.7) 97 (15.5)
High-grade dysplasia 8 (1.3) 7 (1.2) 9 (1.4)
Unspecified dysplasia 12 (2.0) 9 (1.6) 12 (1.9)
Follow-up yearse, mean (SD) 6.4 (3.9) 6.0 (3.9) 7.2 (4.2)
a Cohort used for esophageal and gastric cancer analysis, start of follow-up a half year after diagnosis of
Barrett’s esophagus; b Cohort used for cancer incidence analysis, only patients who were cancer-free at
diagnosis of the Barrett’s esophagus; c Cohort used for mortality analysis; d SD= Standard Deviation; NOS=
Not Otherwise Specified; e Calculated from diagnosis of Barrett’s esophagus.
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
73
In the Barrett’s cohort, 13 cases of esophageal cancer and 5 cases of gastric
cancer were reported during follow-up (Table 4.2). The 13 esophageal cancer cases
were all located in the lower third of the esophagus and the histological type was
adenocarcinoma. Based on national incidence rates only 1.3 cases of esophageal
cancer were expected, and the O/E-ratio was statistically significantly increased: 10.0
(95% CI: 5.3-17.1). The O/E-ratios were higher in males than in females and in cases
with a BE diagnosis at age 68 to 72 years (Table 4.2). The O/E ratios were higher in
cases with intestinal metaplasia, and higher in cases with unspecified dysplasia or high-
grade dysplasia, although numbers were small. O/E-ratios were lower after five years
of follow-up.
The O/E-ratio was 1.8 (95% CI: 0.6-4.2) for gastric cancer (Table 4.2). Four of the
five observed gastric cancer cases were located in the gastric cardia; the subsite of the
remaining case was unspecified. All cases of gastric cancer in the follow-up were
detected in males and in cases with intestinal metaplasia.
The incidence rate of esophageal cancer was 3.7 per 1000 person-years and the
absolute excess risk was 3.3 per 1000 person-years (95% CI: 1.6-5.9 per 1000 person-
years). The incidence rate of gastric cancer was 1.4 per 1000 person-years and the
absolute excess risk was 0.6 per 1000 person-years (95% CI: -0.3 to 2.5 per 1000
person-years) (Table 4.2).
In the dataset for cancer incidence follow-up 92 cases of invasive cancer were
diagnosed during the follow-up (Table 4.3). The O/E-ratio was calculated to be 1.5
(95% CI: 1.2-1.8). Without the esophageal and gastric cancer cases the O/E-ratio was
still increased but not statistically significantly anymore: O/E-ratio: 1.3 (95% CI: 1.0-
1.6). The cancer incidence rate was statistically significantly increased for small
intestinal cancer (O/E-ratio: 11.8; 95% CI: 1.4-42.5) and pancreatic cancer (O/E-ratio:
3.1; 95% CI: 1.0-7.2). When cases diagnosed in the first six months of follow-up were
excluded, the increased risks were no longer statistically significant: O/E-ratio 6.3 (95%
CI: 0.2-35.3) for small intestinal cancer and O/E-ratio 2.7 (95% CI: 0.7-6.8) for
pancreatic cancer. Colorectal cancer incidence was not increased. The incidence rates
of lung and prostate cancer were increased, but not statistically significantly.
The total mortality in the cohort was not statistically significantly increased
(n=626 at risk) (Table 4.4). During follow-up, 225 cases died (n=213.2 expected) and
the O/E-ratio was 1.1 (95% CI: 0.9-1.2). After exclusion of esophageal and gastric
cancer deaths, there were 214 cases who died (n=208.5 expected). The O/E-ratio for
this comparison was 1.0 (95% CI: 0.9-1.2). Mortality was not statistically significantly
increased either for any of the specific causes of death (Table 4.4).
Thesis Jessie Steevens_v04.pdf
Chapter 4
74
Thesis Jessie Steevens_v04.pdf
Table 4.2 Incidence of esophageal and stomach cancer in Barrett’s esophagus cohorta according to sex, age at diagnosis, histology, dysplasia and duration of follow-up
(n=605). Netherlands Cohort Study, 1986-2003.
Incidence rate (per 1000 person-years)
Category n at risk n person-years Observed Expected O/E ratiob 95% CI of O/E ratio Incidence rate AERc 95% CI of AER
Esophageal cancer
Total 605 3560.0 13 1.30 10.0 (5.3-17.1) 3.7 3.3 (1.6- 5.9)
Sex
Females 280 1760.5 2 0.33 6.0 (0.7-21.7) 1.1 0.9 (-0.1- 3.9)
Males 325 1799.6 11 0.97 11.3 (5.7-20.3) 6.1 5.5 (2.5- 10.4)
Age at diagnosis BE
56-67 years 132 1200.0 4 0.38 10.6 (2.9-27.1) 3.3 3.0 (0.6- 8.2)
68-72 years 196 1215.3 7 0.45 15.7 (6.3-32.4) 5.8 5.4 (2.0- 11.5)
73+ years 277 1144.8 2 0.48 4.2 (0.5-15.1) 1.7 1.3 (-0.2- 5.9)
Histological Classification
Intestinal Metaplasia 446 2581.4 11 0.95 11.6 (5.8-20.7) 4.3 3.9 (1.8- 7.3)
Barrett’s NOS 159 978.7 2 0.35 5.7 (0.7-20.4) 2.0 1.7 (-0.1- 7.0)
Dysplasia
No dysplasia 493 2958.1 9 1.06 8.5 (3.9-16.2) 3.0 2.7 (1.0- 5.4)
Low-grade dysplasia 92 487.6 2 0.19 10.3 (1.2-37.1) 4.1 3.7 (0.1- 14.4)
High-grade dysplasia 8 44.5 1 0.02 50.5 (1.3-281.1) 22.5 22.0 (0.1- 124.8)
Unspecified dysplasia 12 69.9 1 0.03 29.9 (0.8-166.3) 14.3 13.8 (-0.1- 79.2)
Follow-up duration d
0-<1 years 605 578.9 2 0.19 10.4 (1.3-37.6) 3.5 3.1 (0.1- 12.1)
1-4 years 549 1745.2 9 0.62 14.5 (6.6-27.6) 5.2 4.8 (2.0- 9.4)
5-9 years 318 1006.6 2 0.39 5.1 (0.6-18.5) 2.0 1.6 (-0.2- 6.8)
10+ years 110 229.4 - 0.10 - -- - - -
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
75
Thesis Jessie Steevens_v04.pdf
Incidence rate (per 1000 person-years)
Category n at risk n person-years Observed Expected O/E ratiob 95% CI of O/E ratio Incidence rate AERc 95% CI of AER
Stomach cancer
Total 605 3560.0 5 2.78 1.8 (0.6-4.2) 1.4 0.6 (-0.3- 2.5)
Sex
Females 280 1760.5 0 0.78 - - - - -
Males 325 1799.6 5 2.00 2.5 (0.8-5.8) 2.8 1.7 (-0.2- 5.3)
Age at diagnosis BE
56-67 years 132 1200.0 2 0.77 2.6 (0.3-9.4) 1.7 1.0 (-0.4- 5.4)
68-72 years 196 1215.3 2 0.97 2.1 (0.3-7.5) 1.6 0.8 (-0.6- 5.2)
73+ years 277 1144.8 1 1.04 1.0 (0.0-5.4) 0.9 0.0 (-0.9- 4.0)
Histological Classification
Intestinal Metaplasia 446 2581.4 5 2.01 2.5 (0.8-12.6) 1.9 1.2 (-0.2- 3.7)
Barrett’s NOS 159 978.7 0 0.77 - - - - -
Dysplasia
No dysplasia 493 2958.1 4 2.24 1.8 (0.5-4.6) 1.4 0.6 (-0.4- 2.7)
Low-grade dysplasia 92 487.6 0 0.42 - - - - -
High-grade dysplasia 8 44.5 1 0.04 24.8 (0.6-137.9) 22.5 21.6 (-0.4- 124.3)
Unspecified dysplasia 12 69.9 0 0.07 - - - - -
Follow-up duration
0-<1 years 605 578.9 1 0.43 2.3 (0.1-12.8) 1.7 1.0 (-0.7- 8.9)
1-4 years 549 1745.2 3 1.34 2.2 (0.5-6.5) 1.7 1.0 (-0.4- 4.3)
5-9 years 318 1006.6 . 0.80 . .. - - -
10+ years 110 229.4 1 0.20 5.1 (0.1-28.5) 4.4 3.5 (-0.7- 23.4)
a Barrett's cases with esophageal and gastric cancer before diagnosis or within half year after diagnosis of Barrett’s esophagus are excluded (leaving 605 cases at risk and
3560.0 person-years at risk); b O/E ratio = Observed/expected ratio; 95% CI = 95% Confidence Intervals; NOS = Not Otherwise Specified; BE= Barrett's esophagus; AER=
Absolute excess risk; c Calculated as ((Observed-Expected)*1000)/person-years; d The first six months of follow-up were excluded. T=0 starts therefore at six months.
Chapter 4
76
Thesis Jessie Steevens_v04.pdf
Table 4.3 Cancer incidence in Barrett’s esophagus cohorta for selected sites (n=561). Netherlands Cohort Study, 1986-2003.
ICD-10 Primary site
b Observed (n) Expected (n) O/E-ratio c 95% CI
C00-C43, C45-C96 All Malignant, skin cancer excluded 92 63.3 1.5 (1.2-1.8)
C00-C14, C16-C43, C45-C96 All Malignant, skin, stomach & esophagus cancer excluded 74 59.5 1.2 (1.0-1.6)
C00-C14, C30-C32, C65 Head & Neck 3 2.0 1.5 (0.3-4.4)
C15-C26 Digestive Organs 34 16.6 2.0 (1.4-2.9) d
C15 Esophagus 13 1.2 11.0 (5.9-18.8) d
C16 Stomach 5 2.5 2.0 (0.6-4.6)
C17 Small Intestine 2 0.2 11.8 (1.4-42.5) d
C18-C20 Colon & Rectum 9 10.0 0.9 (0.4-1.7)
C25 Pancreas 5 1.6 3.1 (1.0-7.2) d
C30-C39 Respiratory & Intrathoracic Organs 15 11.6 1.3 (0.7-2.1)
C34 Lung & Bronchus 14 10.7 1.3 (0.7-2.2)
C40-C41 Bone & Joint 0 0.1 - -
C43 Skin, Melanoma 2 1.1 1.9 (0.2-6.8)
C45-C49 Mesothelial & Soft Tissue 1 0.8 1.2 (0.0-6.8)
C50 Breast 7 6.2 1.1 (0.5-2.3)
C51-C58 Female Genital Organs 2 2.6 0.8 (0.1-2.8)
C60-C63 Male Genital Organs 15 10.0 1.5 (0.8-2.5)
C61 Prostate 15 9.9 1.5 (0.9-2.5)
C64-C68 Urinary Tract 6 4.8 1.2 (0.5-2.7)
C69-C72 Eye, Meninges, Central Nervous System 4 1.5 2.7 (0.7-7.0)
C73-C75 Endocrine Glands 1 0.7 1.5 (0.0-8.4)
C76-C80 Primary Unknown or Ill Defined 0 0.2 - -
C81-C96 Blood, Bone Marrow & Lymphatic organs 4 4.4 0.9 (0.3-2.4)
a No prevalent cancer (skin cancer excluded) at diagnosis of Barrett’s esophagus (561 cases at risk and 3382.8 person-years); b All main categories are presented. Subsites
are presented if >5 observed cases or if O/E-ratio statistically significant decreased or increased; c O/E-ratio = Observed/expected ratio; 95% CI = 95% Confidence Interval.
d Statistically significantly increased.
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
77
Thesis Jessie Steevens_v04.pdf
Table 4.4 Total and cause-specific mortality in Barrett’s esophagus cohort according to cause of death (n=626)a. Netherlands Cohort Study, 1986-2005.
ICD-10 codes Cause of Death Observed (n) Expected (n) O/E-ratio b 95% CI
A00-Y89 All Causes
c 225 213.2 1.1 (0.9-1.2)
A00-B99 Infectious & Parasitic diseases 1 2.3 0.4 (0.0-2.4)
D50-D89 Blood(-forming) & Immune system 0 0.6 - -
E00-E90 Endocrine, Nutritional & Metabolic 7 6.6 1.1 (0.4-2.2)
F00-F90 Mental & Behavioral disorders 4 7.1 0.6 (0.2-1.4)
G00-H95 Central & Peripheral nervous system 9 4.8 1.9 (0.9-3.6)
I00-I99 Circulatory system 84 80.0 1.1 (0.8-1.3)
I21-I22 Myocardial infarction 23 21.9 1.1 (0.7-1.6)
I20, I23-I25 Other Ischemic heart disease 13 8.8 1.5 (0.8-2.5)
I30-I33, I39-I52 Other Heart disease 17 18.1 0.9 (0.6-1.5)
I60-I69 Cerebrovascular disease 17 18.8 0.9 (0.5-1.4)
I00-I15, I26-I28, I34-I38, I70-I99 Other Circulatory diseases 31 12.4 1.1 (0.6-1.9)
J00-J99 Respiratory system 22 23.8 0.9 (0.6-1.4)
J10-J11 Influenza 0 0.3 - -
J12-J18 Pneumonia 6 8.1 0.7 (0.3-1.6)
J40-J47 COPD & Asthma 14 13.1 1.1 (0.6-1.8)
J00-J06, J20-J39, J60-J99 Other Respiratory organs 2 2.3 0.9 (0.1-3.1)
K00-K93 Digestive system 7 7.5 0.9 (0.4-1.9)
L00-L99 Skin & Subcutaneous tissue 0 0.5 - -
M00-M99 Musculoskeletal & Connective tissue 0 1.3 - -
N00-N99 Urogenital system 4 4.2 1.0 (0.3-2.4)
P00-P96 Perinatal period 0 0.0 - -
Q00-Q99 Congenital malformations 0 0.1 - -
R00-R99 Symptoms & signs, NOS 6 8.1 0.7 (0.3-1.6)
V01-Y89 External causes 5 4.1 1.2 (0.4-2.9)
V01-V99 Vehicle (transport) accidents 2 0.7 2.9 (0.4-10.6)
WOO-X59 Other Accidents 3 2.5 1.2 (0.3-3.5)
X60-X84 Suicide (intentional self-harm) 0 0.7 - -
X85-Y09 Kill, Murder & Homicide 0 0.0 - -
Y10-Y89 Other External events 0 0.2 - -
a 626 persons at risk and 4490.5 person-years; b O/E ratio = Observed/expected ratio; 95% CI = 95% Confidence Interval; COPD = Chronic Obstructive Pulmonary Disease;
NOS= Not Otherwise Specified; c Including mortality because of cancer.
Chapter 4
78
DISCUSSION
In this large population-based cohort of cases with Barrett’s esophagus, only the O/E-
ratios of esophageal cancer were statistically significantly increased. Presence of
intestinal metaplasia and unspecified or high-grade dysplasia were associated with the
highest risks of esophageal cancer, although numbers were small. Cancer incidence
without esophageal and gastric cancer incidence was slightly increased, but not
statistically significantly. Incidence of small intestinal and pancreatic cancer was
increased, but was based on small numbers and not statistically significant after
excluding the first half year of follow-up. Cause-specific and overall mortality were not
increased.
Barrett’s esophagus is considered to be a precursor lesion of esophageal
adenocarcinoma.20 In agreement with this assumption, we observed a strongly
increased risk of esophageal cancer during follow-up. The incidence rate in the cohort
was 3.7 per 1000 person-years after excluding the first half year of follow-up. The
pooled estimate in a recent meta-analysis was estimated to be 5.3 per 1000 person-
years.4 However, when the pooled analysis was restricted to studies with a large study
size (>500 person-years), without possible selection bias (>70% follow-up) and a well-
defined definition of Barrett’s esophagus, the incidence was considerably lower and
very close to the results in the current study: 3.9 per 1000 person-years.4 In a recent
follow-up study, an even lower incidence of esophageal cancer was observed: only 1.6
per 1000 person-years.3 The largest follow-up study, using the same Pathology
database as the current study, found an incidence rate of 4.3 per 1000 person-years.5
The relatively low incidence in these studies confirms the opinion that the incidence of
esophageal cancer in Barrett’s esophagus has been overestimated in the literature
because of publication bias.20
In the current study, the incidence rate of esophageal cancer in males was much
higher than that in females, which is in agreement with the meta-analysis.4 Cases with
unspecified or high-grade dysplasia at baseline had a strongly increased risk of
esophageal cancer, although the O/E-ratios were based on one case in each category.
The increased risk is in accordance with the literature.21,22 The increased risk in cases
with unspecified dysplasia is possibly explained by classification problems. In the
summaries of the pathology reports of these cases, dysplasia was sometimes reported
without specification as low- or high-grade, and we classified these cases as
unspecified dysplasia. The elevated risk may therefore be explained by this
misclassification. In contrast to findings from a recent large study,5 the O/E-ratios were
highest during the first five years of follow-up. It is conceivable, that the incidence
after five years of follow-up was underestimated in our study, because the number of
BE cases and cancer events is small, especially regarding cases with a longer follow-up.
The incidence rate of gastric cancer was also slightly increased. In a recent cohort
study from Cook et al the O/E-ratio of gastric cancer was 2.0 (7 cases of gastric cancer
observed and 3.43 expected),3 which is comparable with the result in the current study
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
79
(O/E-ratio: 1.8). Because most gastric cancers in the current study were located in the
gastric cardia, the increased risk might be caused by misclassification of esophageal
cancer as gastric cardia cancer. However, the Netherlands Cancer Registry has good
access to clinical files and its data have been shown to be of good quality,23 making
misclassification less likely.
Incidence rates of small intestinal and pancreatic cancer were also increased, but
not statistically significantly after exclusion of the cases in the first six months after BE
diagnosis. It is therefore conceivable that preclinical complaints of these cancers
warranted endoscopy, in which the Barrett’s esophagus was discovered. The number
of observed cases in our study is also limited. However, in patients with Multiple
Endocrine Neoplasia type 1 (MEN1) and the Zollinger Ellison Syndrome (ZES) the risks
of both duodenal and endocrine pancreatic malignancies and of Barrett’s esophagus
were increased.24,25 In many patients with ZES, excess gastrin is produced by a
gastrinoma in e.g. the duodenum. Gastrin increases gastric acid production and might
therefore increase the risk of Barrett’s esophagus. This is in agreement with the
observation in the current study that one of the detected duodenal cancers was
reported to be a neuroendocrine carcinoma (carcinoid). This finding should be
replicated in other longitudinal studies, before it can be concluded whether a causal
association is likely.
Barrett’s esophagus has also been associated with an increased incidence of
colorectal cancer.26 However, results since this first publication in 1985 have been
inconsistent.3,7,27 The possible association between Barrett’s esophagus and colorectal
cancer has been attributed to common environmental exposures such as bile acids,
vegetables and fruit intake, and shared genetic aberrations.26 A recent study, using
data from the Netherlands Pathology Registry, showed a slightly increased risk of
colorectal cancer, but incidence was especially increased in the first year of follow-up.6
The authors concluded that the greater part of the excess risk is most likely explained
by diagnostic bias. We did not observe an increased risk at all, making it likely that
there is no or only a small increased risk of colorectal cancer in patients with Barrett’s
esophagus.
The O/E-ratio for total mortality was slightly increased, but not statistically
significantly. None of the specific causes of death (excluding cancer mortality) was
increased. Some, but not all, recent studies observed moderately increased O/E-ratios
for mortality. In a British cohort study with 502 patients, an increased mortality was
observed (O/E ratio 1.21; 95% CI: 1.06-1.37).3 Several other studies also reported
increased mortality rates.8-11 Risks in these studies remained elevated after exclusion
of deaths that were related to esophageal cancer. Only one study did not observe an
increased risk.12 In this study, data were used from the Northern Irish Registry of
Barrett’s esophagus, and the O/E ratio was 0.96 (95% CI, 0.84-1.07). The current study
and the Irish study are the only studies that are population-based and not hospital-
based. A possible explanation for the observed discrepancies might therefore be
selection bias. The hospital-based cohorts were assembled at hospitals and may
Thesis Jessie Steevens_v04.pdf
Chapter 4
80
include patients with more complicated disease or with more co-morbidities and were
therefore at increased risk to die. Lack of power is not a likely explanation, as the
current study has 225 deaths, which is comparable with the other published
studies.3,11,12
This study suffered from some limitations. Case definitions of Barrett’s esophagus
differ between the United States (USA) and the United Kingdom (UK). In the USA, the
presence of goblet cells (indicating intestinal metaplasia) is required for the diagnosis
of BE,28 while in the UK any type of metaplasia is sufficient for the diagnosis of BE. 29
Both the USA and UK definitions may have been used in the Netherlands. The use of
definitions may have changed over time and different pathologists may have used
different definitions. Cases were excluded from the analysis if the pathological
summary only reported non-intestinal metaplasia (gastric or junctional metaplasia). In
the pathology reports of approximately 25% of the cases the type of Barrett’s
esophagus was not specified. Some misclassification may therefore have occurred and
this may have attenuated the risks slightly. A second limitation is that we only had
access to summaries of the pathology reports as recorded in the pathology registry
PALGA. However, we were able to compare the conclusions of 60 reports from 29
patients with the full reports from one hospital. In this substudy, the agreement on
patient level was 29/29 (100%) with respect to the eligibility of the cases, 26/29 (90%)
with respect to histology and 27/29 (93%) with respect to dysplasia.
Our study also has several strengths. The study is population-based and has a long
follow-up. Follow-up for vital status was 100% complete and we were able to retrieve
all causes of death from the death certificates. Follow-up for cancer is also presumed
to be >95% complete.30
In conclusion, in this study we observed that patients with Barrett’s esophagus
had an increased risk of esophageal cancer and possibly also for stomach, small
intestinal and pancreatic cancer. The incidence of colorectal cancer was not increased.
Neither overall nor cause-specific mortality were increased.
Thesis Jessie Steevens_v04.pdf
Cancer incidence and cause-specific mortality of patients with Barrett’s esophagus
81
REFERENCES
1. Shaheen NJ, Richter JE. Barrett's oesophagus. Lancet 2009;373:850-861.
2. Wild CP, Hardie LJ. Reflux, Barrett's oesophagus and adenocarcinoma: burning questions. Nat Rev
Cancer 2003;3:676-684.
3. Cook MB, Wild CP, Everett SM, Hardie LJ, Bani-Hani KE, Martin IG, Forman D. Risk of Mortality and
Cancer Incidence in Barrett's Esophagus. Cancer Epidemiol Biomarkers Prev 2007.
4. Yousef F, Cardwell C, Cantwell MM, Galway K, Johnston BT, Murray L. The incidence of esophageal
cancer and high-grade dysplasia in Barrett's esophagus: a systematic review and meta-analysis. Am J
Epidemiol 2008;168:237-249.
5. De Jonge PJ, Van Blankenstein M, Looman CW, Casparie MK, Meijer GA, Kuipers EJ. Risk of Malignant
Progression in Patients with Barrett's oesophagus: a Dutch nationwide cohort study. Gut
2010;59:1030-1036.
6. de Jonge PJ, van Blankenstein M, Looman CW, Casparie MK, Meijer GA, Kuipers EJ. Risk of colorectal
cancer in patients with Barrett's esophagus: A Dutch population-based study. Am J Gastroenterol
2010;105:77-83.
7. Howden CW, Hornung CA. A systematic review of the association between Barrett's esophagus and
colon neoplasms. Am J Gastroenterol 1995;90:1814-1819.
8. van der Burgh A, Dees J, Hop WC, van Blankenstein M. Oesophageal cancer is an uncommon cause of
death in patients with Barrett's oesophagus. Gut 1996;39:5-8.
9. Solaymani-Dodaran M, Logan RF, West J, Card T. Mortality associated with Barrett's esophagus and
gastroesophageal reflux disease diagnoses-a population-based cohort study. Am J Gastroenterol
2005;100:2616-2621.
10. Conio M, Cameron AJ, Romero Y, Branch CD, Schleck CD, Burgart LJ, Zinsmeister AR, Melton LJ, 3rd,
Locke GR, 3rd. Secular trends in the epidemiology and outcome of Barrett's oesophagus in Olmsted
County, Minnesota. Gut 2001;48:304-309.
11. Moayyedi P, Burch N, Akhtar-Danesh N, Enaganti SK, Harrison R, Talley NJ, Jankowski J. Mortality rates
in patients with Barrett's oesophagus. Aliment Pharmacol Ther 2008;27:316-320.
12. Anderson LA, Murray LJ, Murphy SJ, Fitzpatrick DA, Johnston BT, Watson RG, McCarron P, Gavin AT.
Mortality in Barrett's oesophagus: results from a population based study. Gut 2003;52:1081-1084.
13. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
14. van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol
1990;19:553-558.
15. Muir C, Waterhouse J, Mack T, Powell J, Whelan S. Cancer incidence in five continents. Volume V.
Lyon: IARC Scientific publication no. 88, 1987.
16. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J. Cancer incidence in five continents. Volume VII.
Lyon: IARC Scientific publications no. 143, 1997.
17. Parkin DM, Whelan SL, Ferlay J, Teppo L, Thomas DB. Cancer incidence in five continents. Volume VIII.
Lyon: IARC Scientific publications no. 155, 2002.
18. Pearson ES, Hartley HO, (eds). Biometrika Tables for Statisticians. Biometrika Trust, 1976.
19. van Leeuwen FE, Klokman WJ, Hagenbeek A, Noyon R, van den Belt-Dusebout AW, van Kerkhoff EH,
van Heerde P, Somers R. Second cancer risk following Hodgkin's disease: a 20-year follow-up study. J
Clin Oncol 1994;12:312-325.
20. Shaheen NJ, Crosby MA, Bozymski EM, Sandler RS. Is there publication bias in the reporting of cancer
risk in Barrett's esophagus? Gastroenterology 2000;119:333-338.
21. Schnell TG, Sontag SJ, Chejfec G, Aranha G, Metz A, O'Connell S, Seidel UJ, Sonnenberg A. Long-term
nonsurgical management of Barrett's esophagus with high-grade dysplasia. Gastroenterology
2001;120:1607-1619.
22. Weston AP, Sharma P, Topalovski M, Richards R, Cherian R, Dixon A. Long-term follow-up of Barrett's
high-grade dysplasia. Am J Gastroenterol 2000;95:1888-1893.
23. Schouten LJ, Jager JJ, van den Brandt PA. Quality of cancer registry data: a comparison of data
provided by clinicians with those of registration personnel. Br J Cancer 1993;68:974-977.
Thesis Jessie Steevens_v04.pdf
Chapter 4
82
24. Hoffmann KM, Gibril F, Entsuah LK, Serrano J, Jensen RT. Patients with multiple endocrine neoplasia
type 1 with gastrinomas have an increased risk of severe esophageal disease including stricture and
the premalignant condition, Barrett's esophagus. J Clin Endocrinol Metab 2006;91:204-212.
25. Lindor NM, Lindor CJ, Greene MH. Hereditary Neoplastic Syndromes. In: Schottenfeld D, Fraumeni JF,
Jr., eds. Cancer Epidemiology and Prevention. Third Edition. Oxford: Oxford University Press, 2006:
562-576.
26. Sontag SJ, Schnell TG, Chejfec G, O'Connell S, Stanley MM, Best W, Chintam R, Nemchausky B, Wanner
J, Moroni B. Barrett's oesophagus and colonic tumours. Lancet 1985;1:946-949.
27. Bollschweiler E, Schloesser T, Leers J, Vallbohmer D, Schafer H, Holscher AH. High prevalence of
colonic polyps in white males with esophageal adenocarcinoma. Dis Colon Rectum 2009;52:299-304.
28. Wang KK, Sampliner RE. Updated guidelines 2008 for the diagnosis, surveillance and therapy of
Barrett's esophagus. Am J Gastroenterol 2008;103:788-797.
29. Playford RJ. New British Society of Gastroenterology (BSG) guidelines for the diagnosis and
management of Barrett's oesophagus. Gut 2006;55:442-443.
30. Goldbohm R, van Den Brandt PA, Dorant E. Estimation of the coverage of Dutch municipalities by
cancer registries and PALGA based on hospital discharge data. Tijdschr Soc Gezondheidsz 1994;72:
80-84.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and
stomach cancer subtypes in Europe
Jessie Steevens
Anita AM Boerweck
Miranda JM Dirx
Piet A van den Brandt
Leo J Schouten
European Journal of Gastroenterology and Hepatology 2010;22(6):669-678
5
83
Thesis Jessie Steevens_v04.pdf
Chapter 5
84
ABSTRACT
Objective
Time trend studies in the USA have shown that the incidences of adenocarcinomas of
the esophagus and gastric cardia have risen strongly since the 1970s, whereas the
incidence of squamous cell carcinomas of the esophagus has declined. Previously, we
found that the incidence of these adenocarcinomas also rose in some European
countries until the early 1990s. The main goal of this study was to investigate more
recent trends in the incidence of esophageal and stomach cancer subtypes in European
countries.
Methods
Eurocim cancer incidence data of 23 cancer registries from 13 European countries
were used to investigate the incidence trends in esophageal and stomach cancer
subtypes in the 1983-1997 period. We calculated estimated annual percent changes
(EAPCs) in European age- standardized incidence rates and 95% confidence intervals.
Results
The incidence of adenocarcinomas of the esophagus and gastric cardia rose in most,
but not all, registration areas (EAPCs were usually 1 to 7%), the strongest in the UK and
Ireland. Esophageal squamous cell carcinoma incidence rose mostly in Northern
European and Slovakian men (EAPCs: 1 to 5%) and in women (EAPCs: 1 to 8%), but
declined mostly in Southern and Western European men (EAPCs: -1 to -5%).
Conclusions
Our results are partly in line with earlier findings on adenocarcinomas of the
esophagus and gastric cardia. There was, however, substantial heterogeneity in trends
of subtypes of these cancers within Europe. There may be different risk factors for
these cancers, and the prevalence of these risk factors may differ among countries.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
85
INTRODUCTION
These days, esophageal cancer and stomach cancer are often being considered to
comprise more than two types of cancer (e.g. in epidemiology, pathology). Esophageal
cancer is usually divided into histological subtypes, the two main types being
adenocarcinoma and squamous cell carcinoma. Subtypes of stomach cancer are
usually based on topography (i.e. the location of the tumor in the stomach),
distinguishing cardia tumors from tumors in other parts of the stomach. The incidences
of both esophageal and stomach cancers are relatively high1 and the 5-year survival
rates are very low: 12% for esophageal cancer and 24% for stomach cancer in Europe.2
Survival rates of gastric cardia adenocarcinoma are worse than those of other stomach
cancers.3,4
Time trend studies in the United Stated have shown that the incidences of
adenocarcinomas of the esophagus and gastric cardia (AEGC) have risen strongly from
the 1970s, whereas the incidence of squamous cell carcinomas of the esophagus has
declined.5,6 In an earlier publication, we investigated whether these changes also took
place in Europe. We showed that this rise in the incidence of AEGC did indeed occur in
Denmark, part of Italy, Slovakia, England and Wales and Scotland, in the 1980s and
early 1990s.7
Such time trends should always be interpreted cautiously, as an observed trend
may reflect changes in diagnostics, in classification of tumors or in the quality of cancer
registries. The quality and level of detail of data on esophageal and stomach cancer
incidences have increased over time.8-10 More esophageal tumors are being
histologically verified and the exact topography of stomach tumors is more often being
registered.8-10 In contrast, observed trends can of course also indicate a true change in
the occurrence of the disease, which may follow a change in the prevalence of one or
more risk factors. There are indications that the subtypes of esophageal and stomach
cancer have different risk factors. e.g.11
Two studies have investigated these trend in Europe in more recent years,12,13 but
one did not include any Southern or Eastern European countries12 and the other
studied only esophageal cancers.13
The main goals of the present study were therefore: (a) to study trends in the
incidence of histological subtypes of esophageal cancer and topographical subtypes of
stomach cancer, and (b) to include data from all European regions. This allowed a
comparison between cancers and between regions as well. We used Eurocim14 data
and we presented the results in a conveniently arranged way.
Thesis Jessie Steevens_v04.pdf
Chapter 5
86
METHODS
Eurocim database
Cancer registries that are members of the European Network of Cancer Registries are
asked to regularly submit incidence, mortality and population data to a central
databank (Eurocim14) held at the International Agency for Research on Cancer (IARC).
We obtained permission from the steering committee of the European Network of
Cancer Registries to use the Eurocim database, which contains incidence data from 118
population-based cancer registries in 25 countries. Available registration years range
from 1953 to 1998. In addition, the National Cancer Registry of Ireland has extracted
data for the area covered by the Southern Tumors Registry for the period 1983-1997,
because these data were not available in the present version of the Eurocim database.
We wanted to include this region, because it was also included in our earlier study.7
Selection of cancer registries
Figure 5.1 gives an overview of the registry selection, and Table 5.1 lists all registries
included. A cancer registry had to be situated in Europe and had to have registration
data available in Eurocim of esophageal and stomach cancer, for a period of at least 15
years. To ensure that all Europe was covered in this study, at least one registry in every
country was selected. If several registries were available in one country, we selected
those with best data quality. Data quality of cancer registries is defined by three
indicators: (a) the percentage of histologically or microscopically verified cases (%HV),
(b) the percentage of cases derived from death certificates only (%DCO) and (c) the
ratio of mortality to incidence (M/I ratio). We looked at the indicator values from
Cancer Incidence in Five Continents (CI-5) volumes IV-VIII.8-10,15,16 Not all quality indices
were available for all registries; mostly the %DCO was missing and sometimes the M/I
ratio.
From 27 of 29 selected registries, we obtained permission to use their Eurocim
data. The registries were asked some questions regarding registration practice, as this
could be of importance when interpreting the results. Four registries were excluded
from the study, because they did not have data that were detailed enough in terms of
esophageal cancer histology or stomach cancer topography. Data of 23 cancer
registries from 13 countries were available for analysis and all included the 1983-1997
period. When compared with the earlier analyses,7 this study covered more countries
and five more years.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
87
Figure 5.1 Selection of cancer registries and reasons for exclusion.
Selection of tumors
We included all primary tumors of the esophagus and stomach (International
Classification of Diseases for Oncology version 2 (ICD-O-2)17 codes C15 and C16,
respectively). Esophageal cancers were divided in three histological subgroups: (a)
squamous cell carcinomas (ESCC), (b) adenocarcinomas and (c) other esophageal
cancers. Stomach cancers were divided into three topographical subgroups: (a) cardia,
(b) other specified sites (OSS) and (c) not otherwise specified (NOS) (Appendix).
We then combined AEGC to form one group (Appendix). Although it is subject of
discussion whether or not these diseases are the same,4,18,19 it was necessary to
combine them, because some cancer registries used to classify adenocarcinomas of
the lower esophagus as cardia tumors, as was also advised by the International
Classification of Diseases version 9 (ICD-9). In the analysis, we distinguished stomach
tumors OSS from NOS, so that we could examine whether trends observed in cardia
118 Cancer registries (25 countries)
in Eurocim database
-3outside Europe
-1esophageal & stomach cancers were
not registered
-77<15 years of continuous registration
in Eurocim
-8excluded because better quality
registries were available in the same
country
-2 no permission to use Eurocim data
-4necessary details in histology and
topography were not registered or only
for a period < 15 years
29 registries were sent a consent
form and a questionnaire
23 registries (13 countries)
available for data analysis
115
114
27
37
118 Cancer registries (25 countries)
in Eurocim database
-3outside Europe
-1esophageal & stomach cancers were
not registered
-77<15 years of continuous registration
in Eurocim
-8excluded because better quality
registries were available in the same
country
-2 no permission to use Eurocim data
-4necessary details in histology and
topography were not registered or only
for a period < 15 years
29 registries were sent a consent
form and a questionnaire
23 registries (13 countries)
available for data analysis
115
114
27
37
Thesis Jessie Steevens_v04.pdf
Chapter 5
88
tumors could be partly explained by increasing registry quality or specification of
tumor sublocalisation.
Table 5.1 Characteristics of selected cancer registries, by region.
Population size in
1997 (x 1 million)
Esophageal
cancer cases (n/year)a
Stomach
cancer cases (n/year) a
male female male female
Northern Europe
Denmark 5.3 167 75 470 310
Estonia 1.5 47 11 290 245
Iceland 0.3 7 5 44 23
Ireland, Southernb 0.6 23 18 46 26
Sweden 8.8c 225 96 1034 664
UK England, East Angliad 2.2 126 82 266 140
UK England, Merseyside and Cheshiree2.4 152 117 358 249
UK England, South Thames Region 6.9 306 235 868 599
UK England, Yorkshiref 3.7 196 138 477 304
UK Scotland 5.1 333 274 684 509
Central & Eastern Europe
Slovakia 5.4 138 18 903 555
Southern Europe
Italy, Parma 0.4 14 4 146 103
Italy, Lombardy (Varese) 0.8 37 7 168 130
Slovenia 2.0 77 13 294 202
Spain, Navarra 0.5 23 4 102 60
Spain, Catalonia, Tarragona 0.6 21 2 64 37
Western Europe
France, Bas-Rhin 1.0 88 8 82 55
France, Calvados 0.6 86 8 55 37
France, Côte d’Or 0.5 37 4 43 25
France, Doubs 0.5c 34 4 39 23
the Netherlands, Eindhoven 1.0 16 7 99 59
Switzerland, Geneva 0.4 18 5 34 27
Switzerland, St Gallen Appenzell 0.5 17 3 44 32
a Average number of cases in the 1983-1996 period. This period is available in Eurocim for all selected
registries; b Data were obtained from the National Cancer Registry of Ireland. Southern Ireland represents
the area formerly covered by the Southern Tumour Registry; c Population size of 1998 (Sweden), population
size of 1996 (France, Doubs). These years were the latest available in Eurocim; d The registry’s current name
is Eastern Cancer Registration & Information Service (ECRIC); e This registry is now part of the North West
Cancer Intelligence Service (NWCIS); f This registry is now part of the Northern and Yorkshire Cancer Registry
and Information Service (NYCRIS).
Statistical analyses
Data analyses were performed using Eurocim version 4.0 and Stata 9.2 (StataCorp,
College Station, Texas, USA). We calculated age-standardized incidence rates (ESR)
according to the European Standard Population.14 A graph of these rates against time
was plotted for each registry and sex, showing trends in the five tumor groups
mentioned above. These graphs were inspected to see whether the trends were linear
and whether a change in trend (a joinpoint) seemed to be present. In case a joinpoint
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
89
seemed to be present, we used the Joinpoint regression program (Version 3.0,
Statistical Research and Applications Branch, National Cancer Institute, Bethesda, MD,
USA) to test whether the trend change was statistically significant.20 The minimal
number of joinpoints was set to zero, and the maximum to three; which are the
default values of the program.
Estimated annual percentage changes (EAPC) and corresponding 95% confidence
intervals (95% CI) were calculated for each registry, sex and cancer subtype, using the
following formulae: ln(rate) = a x period + b and EAPC = 100 (ea -1), where a is
regression coefficient and b is intercept. The EAPCs were calculated for the same
15-year period (1983-1997) for all registries to make these comparable with one
another. Data of different cancer registries within one country were not pooled before
analysis for two reasons: (a) this would make differences within in a country
indiscernible and (b) regional registries included do not always cover the whole
country (e.g. France, UK). In addition, French regional cancer registries are not
representative for the national population.21 The EAPCs were visualized in forest plots,
to facilitate comparisons of the trends in Europe. Results are presented by regions of
Europe, according to the United Nations classification.22 For each region, a summary
estimate is presented, which was obtained by a fixed effects model using the method
of Mantel and Haenszel. The I-squared measure is shown to quantify the heterogeneity
within the regions.23
RESULTS
Descriptives
Table 5.1 lists some characteristics of the 23 cancer registries included. The population
size covered by the cancer registries ranged from 0.3 to 8.8 million. The average
numbers of cancer cases per year are listed in Table 5.1. For esophageal cancer, these
ranges were 7-333 cases per year in men and 2-274 cases per year in women and for
stomach cancer, these ranges were 34-1034 and 23-664 cases per year in men and
women, respectively. ESRs, averaged over three-year periods, were higher among men
than among women (for details, see Tables 5.2A and 5.2B). Figure 5.2 shows two
examples of ESR graphs.
The male-to-female ratios of ESRs varied greatly between countries and between
the cancer subtypes. These were highest for ESCC: about six to nine, averaged for all
registries. Very high ratios were seen in France, Spain, Switzerland (St Gallen
Appenzell) and Slovakia. In contrast, the ratios were just above one in Great Britain
and Southern Ireland. Much less variation was observed in the male-to-female ratios of
AEGC, with an average ratio of approximately five. Lower average ratios
(approximately two) were found for stomach cancers OSS and NOS.
Thesis Jessie Steevens_v04.pdf
Chapter 5
90
Thesis Jessie Steevens_v04.pdf
Table 5.2A Annual incidence rates according to the European Standard Population (ESRa) of esophageal and stomach cancer subtypes, men.
Esophageal squamous cell
carcinoma
Adenocarcinomas of
esophagus and gastric cardia
Stomach
other specified sites
Stomach
not otherwise specified
1982-1984b 1995-1997 1982-1984 1995-1997 1982-1984 1995-1997 1982-1984 1995-1997
Northern Europe
Denmark 3.0 3.4 6.6 7.7 5.0 2.3 9.7 5.3
Estonia 3.4 5.8 5.8 3.6 30.3 29.0 20.4 13.7
Iceland 5.2 5.2 7.0 7.7 10.2 6.7 27.4 12.8
Ireland, Southernc 3.9 5.4 5.4 10.9 3.1 8.9 15.3 3.0
Sweden 3.5 2.6 3.3 4.4 0 0 18.9 9.8
UK England, East Angliad 2.7 3.3 7.0 14.9 5.7 6.0 14.6 7.8
UK England, Merseyside and Cheshiree 3.7 3.6 8.7 14.6 18.0 6.0 11.5 10.3
UK England, South Thames Region 2.7 2.9 7.0 10.1 5.7 3.0 15.0 10.1
UK England, Yorkshiref 3.1 3.4 8.3 13.0 7.6 4.7 15.1 11.2
UK Scotland 4.9 5.7 8.7 15.1 6.8 4.3 18.7 12.3
Central & Eastern Europe
Slovakia 3.5 8.1 3.2 4.0 23.5 22.7 17.5 5.4
Southern Europe
Italy, Parma 3.0 4.4 3.5 5.3 28.8 22.9 32.2 11.6
Italy, Lombardy (Varese) 6.8 6.3 4.1 5.0 32.3 24.9 18.1 3.7
Slovenia 7.0 5.9 4.8 4.6 9.7 13.1 25.2 17.3
Spain, Navarra 8.5 7.7 3.0 5.1 16.0 23.1 22.4 5.4
Spain, Catalonia, Tarragona 5.3 5.4 2.4 2.3 3.1 5.4 19.2 13.7
Western Europe
France, Bas-Rhin 25.1 14.9 5.0 6.3 4.1 4.7 16.6 9.5
France, Calvados 32.8 18.3 6.0 7.0 20.5 13.6 1.0 2.0
France, Côte d’Or 15.7 8.5 4.7 7.0 18.9 9.6 0.4 0
France, Doubs 14.5 12.4 7.6 3.4 16.9 7.8 1.8 4.5
the Netherlands, Eindhoven 3.1 2.8 7.0 8.7 19.4 15.7 4.6 1.2
Switzerland, Geneva 8.6 7.5 5.2 5.8 12.3 8.7 1.5 0.7
Switzerland, St Gallen Appenzell 9.6 3.5 6.4 7.0 14.8 9.7 0.5 2.4
a ESR, incidence rate according to the European Standard Population. Number of cases per 100 000 person-years; b Average incidences rates over 3-year periods were
calculated to create more stable rates. When only 2 years of the 3-year period were available, a 2-year average was calculated; c Data were obtained from National
Cancer Registry of Ireland; d The registry’s current name is Eastern Cancer Registration & Information Service (ECRIC); e This registry is now part of the North West Cancer
Intelligence Service (NWCIS); f This registry is now part of the Northern and Yorkshire Cancer Registry and Information Service (NYCRIS).
Trends in incidence of esophageal and stomach cancer subtypes in Europe
91
Thesis Jessie Steevens_v04.pdf
Table 5.2B Annual incidence rates according to the European Standard Population (ESRa) of esophageal and stomach cancer subtypes, women
Esophageal squamous cell
carcinoma
Adenocarcinomas of
esophagus and gastric cardia
Stomach
other specified sites
Stomach
not otherwise specified
1982-1984b 1995-1997 1982-1984 1995-1997 1982-1984 1995-1997 1982-1984 1995-1997
Northern Europe
Denmark 1.1 1.5 1.6 1.4 2.4 1.7 5.7 2.8
Estonia 0.5 0.7 1.5 1.1 17.6 13.9 8.8 5.9
Iceland 3.3 2.4 1.8 1.6 7.1 3.2 11.1 5.0
Ireland, Southernc 4.5 3.3 0.9 2.9 1.0 5.0 7.5 2.0
Sweden 1.1 0.9 0.7 1.0 0 0 10.3 5.7
UK England, East Angliad 2.4 3.2 1.5 2.9 2.1 2.1 6.6 2.9
UK England, Merseyside and Cheshiree 3.1 3.5 2.4 3.7 7.7 2.8 5.5 5.3
UK England, South Thames Region 2.3 2.4 1.4 2.2 2.5 1.3 6.3 4.1
UK England, Yorkshiref 2.6 2.6 1.6 3.3 3.4 2.4 6.9 5.0
UK Scotland 3.5 4.5 2.6 4.2 3.6 2.3 10.1 6.5
Central & Eastern Europe
Slovakia 0.2 0.4 0.8 0.9 10.5 10.1 9.1 2.5
Southern Europe
Italy, Parma 0.5 1.2 0.7 0.8 14.4 13.8 14.4 6.7
Italy, Lombardy (Varese) 0.7 0.9 0.9 1.6 15.7 13.8 9.3 2.9
Slovenia 0.8 0.6 1.2 1.3 3.5 6.1 12.4 7.8
Spain, Navarra 0.7 0.3 0.7 0.8 6.8 9.9 8.0 1.7
Spain, Catalonia, Tarragona 0.4 0.4 0.5 0.4 1.9 2.3 9.5 5.6
Western Europe
France, Bas-Rhin 0.9 1.2 0.8 0.8 2.4 2.6 6.5 3.4
France, Calvados 1.7 2.3 1.6 1.4 8.4 6.2 0.5 1.0
France, Côte d’Or 0.9 1.5 1.1 0.5 7.7 4.5 0.4 0.2
France, Doubs 1.4 2.1 1.4 1.2 8.9 4.1 0.7 2.2
the Netherlands, Eindhoven 1.1 1.8 2.3 1.9 9.8 7.2 1.4 0.5
Switzerland, Geneva 1.8 2.7 1.1 1.2 5.0 5.4 1.6 0.4
Switzerland, St Gallen Appenzell 0.3 0.9 1.1 1.2 12.3 4.8 0.4 1.1
a ESR, incidence rate according to the European Standard Population. Number of cases per 100 000 person-years; b Average incidences rates over 3-year periods were
calculated to create more stable rates. When only 2 years of the 3-year period were available, a 2-year average was calculated; c Data were obtained from National
Cancer Registry of Ireland; d The registry’s current name is Eastern Cancer Registration & Information Service (ECRIC); e This registry is now part of the North West Cancer
Intelligence Service (NWCIS); f This registry is now part of the Northern and Yorkshire Cancer Registry and Information Service (NYCRIS).
Chapter 5
92
Figure 5.2 Trends in incidence rates (according to the European Standard Population, ESR) of esophageal
and stomach cancer subtypes for men and women in Slovenia (A) and UK Scotland (B), 1983-
1997
Estimated annual percentage changes
Figures 5.3A to 5.3H show EAPCs and 95% CIs for ESCC (Figures 5.3A and 5.3B), AEGC
(5.3C and 5.3D), stomach tumors OSS (5.3E and 5.3F) and stomach tumors NOS (5.3G
and 5.3H), for men and women separately, during the 1983-1997 period. The EAPCs
were calculated for all groups, but the incidence rates for women were often very low,
which resulted in wide confidence intervals for these EAPCs.
ESCC incidence rose slightly; approximately 1 to 2% per year in most Northern
European countries and Slovakia in men and women, but none of the EAPCs was
010 20 30
1985 1990 1995 2000 1985 1990 1995 2000
men women
Slovenia Slovenia
Age-standardiz ed incidence rates p er 100,000
Year
A
010 20 30
1985 1990 1995 2000 1985 1990 1995 2000
men women
Slovenia Slovenia
Age-standardiz ed incidence rates p er 100,000
Year
A
0 5 10 15 20
1985 1990 1995 2000 1985 1990 1995 2000
men women
UK Scotland UK Scotland
ESCC
Stomach N OS
AEGC
Stomach OSS
Esophagu s other
Age-standardized incidence rates per 100,000
Year
B
0 5 10 15 20
1985 1990 1995 2000 1985 1990 1995 2000
men women
UK Scotland UK Scotland
ESCC
Stomach N OS
AEGC
Stomach OSS
Esophagu s other
Age-standardized incidence rates per 100,000
Year
B
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
93
statistically significant. In Southern and Western Europe the incidence declined
approximately 1 to 5% per year in men, and mostly rose in women. The latter EAPCs
were based on less than 10 cases per year, which is reflected by the wide 95% CIs.
AEGC incidence rose in men and women in most (but not all) European regions
covered by the cancer registries. The increases were strongest in Northern Europe (1
to 7% per year), specifically in Great Britain and Southern Ireland (approximately 3 to
7%). Smaller increases (1 to 3% per year) were seen in the other European regions.
Individual cancer registration areas where the incidence of AEGC did not rise were:
Estonia, Iceland, Slovenia, Spain Navarra (women), Spain Tarragona, France Doubs,
Netherlands Eindhoven (women) and Switzerland Geneva (men). The above trends
were not always statistically significant; for details see Figures 5.3A-H.
In Northern European men and Southern European women, AEGC was more
common than ESCC. This difference had become larger during the 1983-1997 period. In
Northern European women, AEGC also became more frequent relative to ESCC, and by
the end of the 1990s both types had become equally common. A similar tendency was
seen in men from Southern and Western Europe, although AEGC had not, or not yet,
outnumbered ESCC. In contrast, the ratio of ESCC to AEGC increased in Western
European women. The incidence of other esophageal tumors was considerably low in
most countries, when compared with AEGC and ESCC (data not shown).
Stomach tumors OSS showed a declining incidence in nearly all registration areas,
in men as well as in women. The greatest declines were seen in Western and Northern
Europe, but the picture was less clear for Southern Europe and Slovakia. A remarkable
exception was Southern Ireland, where a strong and statistically significant increase in
this type of cancer was found (EAPCs: 16% in men and 18% in women).
Incidence of stomach tumors NOS declined (approximately 2 to 10% per year) in
both sexes in practically all registration areas in Northern, Central & Eastern, and
Southern Europe, and the decline was often statistically significant. In Western Europe,
the incidence of these tumors also declined in some areas but, in contrast, increased in
some parts of France and Switzerland. In the latter regions, however, this relative
increase represents a small absolute increase, as the percentage of stomach tumors
NOS was already low in the early 1980s, because of a good topographic specification.
Joinpoint analyses
Results of the Joinpoint analyses showed that when a change in trend seemed present,
it was mostly also statistically significant (data not shown). The majority of trend
changes were seen in stomach tumors NOS and some in stomach tumors OSS. A strong
decline in the incidence of stomach tumors NOS was mostly followed by a weaker
decline. We did not observe any trend changes in the incidence of AEGC. In only one
registration area (France Bas-Rhin) a trend change was observed in ESCC incidence in
men: it rose until 1984 and declined thereafter.
Thesis Jessie Steevens_v04.pdf
Chapter 5
94
A
B
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
95
C
D
Thesis Jessie Steevens_v04.pdf
Chapter 5
96
E
F
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
97
G
H
Figure 5.3 Estimated annual percent changes (EAPCs) and 95% confidence intervals of standardized
incidence rates (according to the European Standard Population, ESR) of esophageal and
stomach cancer subtypes in European countries, 1983-1997.
A summary estimate of the EAPCs is depicted for each European region, which was obtained
by a fixed effects model using the method of Mantel and Haenszel. The I-squared measure
quantifies the heterogeneity within the regions.
Thesis Jessie Steevens_v04.pdf
Chapter 5
98
DISCUSSION
Summary of results
This study showed that the incidence of adenocarcinomas of esophagus and gastric
cardia (AEGC) rose for both sexes in many, but not all, European countries during the
1983-1997 period. The magnitude of this increase differed among countries. Incidence
rates of ESCC rose in some European regions, but declined in others, and there was a
difference between men and women. In nearly all regions, the ratio of AEGC to ESCC
incidence was increasing.
Interpretation of trends in incidence
When interpreting results of studies on trends in incidence rates, one always has to be
cautious. A careful look should be taken at possible explanations, other than a true
increase of the disease. The growing use of endoscopies may have caused more
tumors of the esophagus to be histologically verified, but we do not expect this
growing use of endoscopies to have caused much overdiagnosis of esophageal cancers.
The attention for esophageal and stomach cancer increased and might have influenced
pathologists, and the histological description of the tumors became more detailed.
This may have led to improved detection of these tumors. In some countries the
incidence of not specified esophageal tumors indeed decreased, but mainly before the
period studied and the incidence of these tumors is generally low.
A better specification and registration of the topography of tumors could in part
explain the decreasing trend in stomach cancers NOS, which goes together with the
increase of cardia tumors. However, the decline of stomach cancers NOS is often much
stronger than the increase in cardia tumors. Furthermore, one would then expect
stomach cancers OSS to increase as well, but in most countries these showed a
decreasing incidence instead. In Southern Ireland, the trends (rising AEGC and stomach
cancers OSS and declining stomach cancers NOS) may in part have been caused by a
better topography registration. However, we do not know the quality indices of the
mid 1990s to support this.
In this study we did not separate adenocarcinomas of the esophagus from tumors
of the gastric cardia. Some studies have made this separation and have found different
trends in incidence for these tumors.5,12,18,24 There are studies indicating that these two
cancers have a different etiology,e.g.25 but other studies point to similar risk factors.e.g.11
It is difficult to distinguish tumors of the lower esophagus from those of the gastric
cardia. This applies in the clinic and in cancer registration, because registration differs
among registries and over time. Therefore, combining the two tumors was the best
solution.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
99
Comparison with other studies
In the literature, rising incidences in AEGC were found in the USA, Australia and New
Zealand,5,12,18,24 but no such trend was found in Hong Kong and Taiwan.26,27 The results
of our study show that beside heterogeneity in trends between continents,
heterogeneity can also be found within Europe, which is consistent with results from
our earlier study.7
Eurocim database only contains data up to 1997, but it is a valuable source of data
for trend studies in which sublocalisation and histological subgroups are needed, and
should therefore be updated regularly. A comparison with more recent data from
individual countries is limited, because, to our knowledge there are only very few
studies available. In Sweden, the trends in the incidences of ESCC, esophageal
adenocarcinoma, and gastric cardia adenocarcinoma, continued in the same direction
until 2004.28 In the Netherlands, trends were described for AEGC until 2003, showing a
further increase in incidence in men, and in women as well.29 In France, the incidences
of esophageal cancer and stomach cancer were estimated until 2000 and were
expected to decrease, except for stomach cancer in women, which was thought to
rise.30 These figures, however, are estimated for the whole country, whereas individual
regions may have different trends, as we also found in our study. In Italy, the incidence
of stomach cancer was estimated until 2010 and was thought to further decrease.31
Possible explanations for the trends observed
If the observed trends represent real changes in incidence, what then underlies these
changes? Ecological studies compare trends in the prevalence of possible risk factors
with the observed trends in cancer incidence. Trend data of these factors however, are
mostly not available until the 1980s, whereas earlier data are probably more relevant,
because one expects a certain latency time before changes in risk factors have their
effect on cancer incidence. Therefore, the following must be interpreted cautiously.
Results from the MONICA study showed that the prevalence of obesity increased
between the mid 1980s and late 1990s,32 most strongly in the UK. This may explain
part of the trends in incidence of AEGC as overweight and obesity are strong risk
factors for these tumors.11 The rising prevalence described in that study can affect
AEGC incidence later on. The decreasing prevalence of Helicobacter pylori infection in
developed countries33 could also be important, because this infection is related to
increased risk of noncardia stomach cancer, but to decreased risk of AEGC.34,35
Smoking and alcohol consumption are differently related to the subtypes of
esophageal and stomach cancer36,37 and Bosetti et al. concluded that these factors
might explain trends in incidence.13 In addition, other factors (e.g. diet and nutrition)
could explain the observed trends.
Thesis Jessie Steevens_v04.pdf
Chapter 5
100
Suggestions for the future
Eurocim database is a valuable source of information for all kinds of studies, and
should therefore be updated regularly. If the Eurocim database would be updated,
data from more countries could be used because they will then have data available for
a longer period. This is especially important with respect to Eastern Europe, which was
not well represented in this study. Thus there is a need for better standards in
timeliness and comparability in cancer registry datasets.
Etiological epidemiological research can identify environmental risk factors for
esophageal and stomach cancer and its subtypes and should be applied in primary
prevention, as survival is still very low.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
101
APPENDIX
Classification of tumors of the esophagus and stomach
Description of tumor Topography codes
ICD-O-2 17
Morphology codes
ICD-O-2 17
Eurocim entities14
Esophagus, squamous
cell carcinoma (ESCC)
C15.0-C15.9 8050-8082 134, 141, 148
Esophagus,
adenocarcinoma a
C15.0-C15.9 8140-8473, 8480-8490,
8500-8550
135-136, 142-143,
149-150
Esophagus, other
subtypes
C15.0-C15.9 All other morphologies 137-140, 144-147,
151-154
Stomach, cardia a C16.0 All morphologies 155-161
Stomach, other
specified sites (OSS)
C16.1-C16.8 All morphologies 162-175
Stomach, not otherwise
specified (NOS)
C16.9 All morphologies 176-182
a These tumors were combined in analyses into adenocarcinomas of the esophagus and gastric cardia
(AEGC).
Thesis Jessie Steevens_v04.pdf
Chapter 5
102
REFERENCES
1. Ferlay J, Autier P, Boniol M, Heanue M, Colombet M, Boyle P. Estimates of the cancer incidence and
mortality in Europe in 2006. Ann Oncol 2007;18:581-592.
2. Berrino F, De Angelis R, Sant M, Rosso S, Bielska-Lasota M, Coebergh JW, Santaquilani M. Survival for
eight major cancers and all cancers combined for European adults diagnosed in 1995-99: results of the
EUROCARE-4 study. Lancet Oncol 2007;8:773-783.
3. Verdecchia A, Corazziari I, Gatta G, Lisi D, Faivre J, Forman D. Explaining gastric cancer survival
differences among European countries. Int J Cancer 2004;109:737-741.
4. Dolan K, Sutton R, Walker SJ, Morris AI, Campbell F, Williams EM. New classification of oesophageal
and gastric carcinomas derived from changing patterns in epidemiology. Br J Cancer 1999;80:834-842.
5. Devesa SS, Blot WJ, Fraumeni JF, Jr. Changing patterns in the incidence of esophageal and gastric
carcinoma in the United States. Cancer 1998;83:2049-2053.
6. Trivers KF, Sabatino SA, Stewart SL. Trends in esophageal cancer incidence by histology, United States,
1998-2003. Int J Cancer 2008;123:1422-1428.
7. Botterweck AA, Schouten LJ, Volovics A, Dorant E, van Den Brandt PA. Trends in incidence of
adenocarcinoma of the oesophagus and gastric cardia in ten European countries. Int J Epidemiol
2000;29:645-654.
8. Ferlay J, Muir CS, Whelan SL, Gao Y-T. Cancer incidence in five continents. Volume VI. Lyon: IARC
Scientific publications no. 120, 1992.
9. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J. Cancer incidence in five continents. Volume VII.
Lyon: IARC Scientific publications no. 143, 1997.
10. Parkin DM, Whelan SL, Ferlay J, Teppo L, Thomas DB. Cancer incidence in five continents. Volume VIII.
Lyon: IARC Scientific publications no. 155, 2002.
11. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:1503-
1511.
12. Vizcaino AP, Moreno V, Lambert R, Parkin DM. Time trends incidence of both major histologic types of
esophageal carcinomas in selected countries, 1973-1995. Int J Cancer 2002;99:860-868.
13. Bosetti C, Levi F, Ferlay J, Garavello W, Lucchini F, Bertuccio P, Negri E, La Vecchia C. Trends in
oesophageal cancer incidence and mortality in Europe. Int J Cancer 2008;122:1118-1129.
14. European Network of Cancer Registries. European incidence database V2.3. Eurocim version 4. Lyon:
IARC, 2001.
15. Waterhouse J, Muir C, Shanmugaratnam K, Powell J. Cancer incidence in five continents. Volume IV.
Lyon: IARC Scientific publication no. 42, 1982.
16. Muir C, Waterhouse J, Mack T, Powell J, Whelan S. Cancer incidence in five continents. Volume V.
Lyon: IARC Scientific publication no. 88, 1987.
17. Percy C, van Holten V, Muir C. International classification of diseases for oncology, Second edition.
Geneva: World Health Organization, 1990.
18. Crane SJ, Richard Locke G, 3rd, Harmsen WS, Diehl NN, Zinsmeister AR, Joseph Melton L, 3rd, Romero
Y, Talley NJ. The changing incidence of oesophageal and gastric adenocarcinoma by anatomic sub-site.
Aliment Pharmacol Ther 2007;25:447-453.
19. Rusch VW. Are cancers of the esophagus, gastroesophageal junction, and cardia one disease, two, or
several? Semin Oncol 2004;31:444-449.
20. Kim HJ, Fay MP, Feuer EJ, Midthune DN. Permutation tests for joinpoint regression with applications to
cancer rates. Stat Med 2000;19:335-351.
21. Remontet L, Esteve J, Bouvier AM, Grosclaude P, Launoy G, Menegoz F, Exbrayat C, Tretare B, Carli PM,
Guizard AV, Troussard X, Bercelli P, Colonna M, Halna JM, Hedelin G, Mace-Lesec'h J, Peng J, Buemi A,
Velten M, Jougla E, Arveux P, Le Bodic L, Michel E, Sauvage M, Schvartz C, Faivre J. Cancer incidence
and mortality in France over the period 1978-2000. Rev Epidemiol Sante Publique 2003;51:3-30.
22. United Nations. http://unstats.un.org/unsd/methods/m49/m49regin.htm#europe, date accessed: 30
September 2008.
23. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. Bmj
2003;327:557-560.
Thesis Jessie Steevens_v04.pdf
Trends in incidence of esophageal and stomach cancer subtypes in Europe
103
24. Armstrong RW, Borman B. Trends in incidence rates of adenocarcinoma of the oesophagus and gastric
cardia in New Zealand, 1978-1992. Int J Epidemiol 1996;25:941-947.
25. Brewster DH, Fraser LA, McKinney PA, Black RJ. Socioeconomic status and risk of adenocarcinoma of
the oesophagus and cancer of the gastric cardia in Scotland. Br J Cancer 2000;83:387-390.
26. Chang SS, Lu CL, Chao JY, Chao Y, Yen SH, Wang SS, Chang FY, Lee SD. Unchanging trend of
adenocarcinoma of the esophagus and gastric cardia in Taiwan: a 15-year experience in a single
center. Dig Dis Sci 2002;47:735-740.
27. Yee YK, Cheung T, Chan AO, Yuen M, Wong BC. Decreasing trend of esophageal adenocarcinoma in
Hong Kong. Cancer Epidemiol Biomarkers Prev 2007;16:2637-2640.
28. Falk J, Carstens H, Lundell L, Albertsson M. Incidence of carcinoma of the oesophagus and gastric
cardia. Changes over time and geographical differences. Acta Oncologica 2007;46:1070-1074.
29. van Blankenstein M, Looman CW, Siersema PD, Kuipers EJ, Coebergh JW. Trends in the incidence of
adenocarcinoma of the oesophagus and cardia in the Netherlands 1989-2003. Br J Cancer
2007;96:1767-1771.
30. Bouvier AM, Remontet L, Jougla E, Launoy G, Grosclaude P, Buemi A, Tretarre B, Velten M, Dancourt V,
Menegoz F, Guizard AV, Mace Lesec'h J, Peng J, Bercelli P, Arveux P, Esteve J, Faivre J. Incidence of
gastrointestinal cancers in France. Gastroenterol Clin Biol 2004;28:877-881.
31. Inghelmann R, Grande E, Francisci S, Verdecchia A, Micheli A, Baili P, Capocaccia R, De Angelis R.
Regional estimates of stomach cancer burden in Italy. Tumori 2007;93:367-373.
32. Seidell JC. Prevalence and time trends of obesity in Europe. J Endocrinol Invest 2002;25:816-822.
33. Everhart JE. Recent developments in the epidemiology of Helicobacter pylori. Gastroenterol Clin North
Am 2000;29:559-578.
34. Helicobacter and Cancer Collaborative Group. Gastric cancer and Helicobacter pylori: a combined
analysis of 12 case control studies nested within prospective cohorts. Gut 2001;49:347-353.
35. Islami F, Kamangar F. Helicobacter pylori and esophageal cancer risk: a meta-analysis. Cancer Prev Res
(Phila Pa) 2008;1:329-338.
36. Lindblad M, Rodriguez LA, Lagergren J. Body mass, tobacco and alcohol and risk of esophageal, gastric
cardia, and gastric non-cardia adenocarcinoma among men and women in a nested case-control study.
Cancer Causes Control 2005;16:285-294.
37. Freedman ND, Abnet CC, Leitzmann MF, Mouw T, Subar AF, Hollenbeck AR, Schatzkin A. A prospective
study of tobacco, alcohol, and the risk of esophageal and gastric cancer subtypes. Am J Epidemiol
2007;165:1424-1433.
Thesis Jessie Steevens_v04.pdf
Chapter 5
104
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking
and risk of subtypes of
esophageal and gastric cancer:
a prospective cohort study
Jessie Steevens
Leo J Schouten
R Alexandra Goldbohm
Piet A van den Brandt
Gut 2010;59:39-48
6
105
Thesis Jessie Steevens_v04.pdf
Chapter 6
106
ABSTRACT
Objective
Alcohol consumption and cigarette smoking may be differentially associated with
esophageal squamous cell carcinoma (ESCC), esophageal adenocarcinoma (EAC),
gastric cardia adenocarcinoma (GCA), and gastric non-cardia adenocarcinoma (GNCA).
However, because this was based on retrospective studies, these hypotheses were
examined in a prospective cohort.
Methods
The prospective Netherlands Cohort Study consists of 120 852 participants who
completed a baseline questionnaire on diet and other cancer risk factors in 1986. After
16.3 years of follow-up, 107 ESCC, 145 EAC, 164 GCA, and 491 GNCA cases were
available for analysis using Cox proportional hazards models and the case-cohort
approach.
Results
The multivariable adjusted incidence rate ratio (RR) for ESCC was 4.61 (95% confidence
interval (CI) 2.24-9.50) for ≥30 g ethanol/day compared with abstainers (p
trend<0.001), while no associations with alcohol were found for EAC, GCA or GNCA.
Compared with never smokers, current smokers had RRs varying from 1.60 for GCA to
2.63 for ESCC, and were statistically significant or borderline statistically significant.
Frequency, duration, and pack-years of smoking were independently associated with
risk of all four cancers. A positive interaction was found between alcohol consumption
and smoking status regarding ESCC risk. The RR for current smokers who consumed
>15 g/day of ethanol was 8.05 (95% CI 3.89-16.60; p interaction=0.65), when
compared with never smokers who consumed <5 g/day of ethanol.
Conclusions
This prospective study found alcohol consumption to be associated with increased risk
of only ESCC. Cigarette smoking was associated with risk of all four cancers.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
107
INTRODUCTION
In the past, esophageal and gastric cancer have been regarded as two disease entities,
but nowadays it is becoming clearer that probably each represents several diseases.
The two main subtypes of esophageal cancer are squamous cell carcinoma (ESCC) and
adenocarcinoma (EAC). Gastric cancer can be subdivided into gastric cardia
adenocarcinomas (GCA) and gastric non-cardia adenocarcinomas (GNCA).
The incidence rates of EAC and GCA have risen greatly in the USA and Europe
during the past decades.1,2 For ESCC, conversely, rates declined or remained stable,
and rates of GNCA declined greatly.1,2 These differences in trends indicate that these
cancer subtypes may have a different etiology and thus differential associations with
lifestyle factors such as alcohol consumption and cigarette smoking.
On this topic, mainly case-control studies have been performed, but only few
have compared subtypes of esophageal and gastric cancer.3-6 This study design is
sensitive for misclassification,7 which might be of importance in studies investigating
smoking and alcohol consumption, because of participants’ awareness of the
established health risks of these exposures.8 Prospective cohort studies are less
susceptible to this bias. However, only two cohort studies9,10 have reported on alcohol
consumption and risk of ESCC, EAC, GCA and GNCA, and three cohort studies have
reported on cigarette smoking.9-11 These studies often lacked sufficient information on
confounders.9,11 We therefore investigated this association within a large-scale
prospective cohort: the Netherlands Cohort Study on diet and cancer (NLCS).12
We hypothesized that: (1) alcohol consumption and cigarette smoking are
strongly positively associated with ESCC risk, with multiplicative interaction; (2) alcohol
consumption is not associated with EAC, GCA, and GNCA; and (3) cigarette smoking is
positively associated with EAC, GCA, and GNCA, but less strongly than with ESCC.
METHODS
Study design and participants
This study was conducted within the NLCS, which started in September 1986 with the
enrolment of Dutch men (n=58 279) and women (n=62 573) women aged 55-70
years.12
For reasons of efficiency, the case-cohort approach was used for data processing
and analysis.13 Cases were derived from the entire cohort, and the number of person-
years at risk for the entire cohort was estimated from a subcohort of 5000, who were
randomly sampled from the total cohort at baseline. Person-years at risk were
calculated from the start of the study until esophageal or gastric cancer diagnosis,
death, emigration, loss to follow-up or end of follow-up, whichever occurred first.
Thesis Jessie Steevens_v04.pdf
Chapter 6
108
Follow-up for cancer incidence was performed by record linkage to the
Netherlands Cancer Registry (NCR) and the nationwide network and registry of
histopathology and cytopathology in the Netherlands (PALGA).14 The completeness of
cancer follow-up is ≥96%,15 and follow-up of the subcohort was nearly 100% complete
(only one male subcohort member was lost to follow-up) after 16.3 years (from 17
September 1986 until 31 December 2002). The following numbers of incident,
microscopically confirmed, primary carcinomas were identified: 130 ESCC16 (ICD-O-3
C15), 181 EAC16 (C15), 206 GCA (C16.0) and 594 GNCA (C16.1-C16.9) (Figure 6.1).
Excluded were non-microscopically confirmed cancers, which formed 1% of both
esophageal and gastric cancers. The group of GNCA included cancers with a lesion with
overlapping subsites of the stomach (C16.8, n=160) and some gastric not otherwise
specified cancers (C16.9, n=75), raising the possibility that some cardia cancers might
be included in de non-cardia category. However, as we found risk estimates to be
similar in separate analyses of gastric cancers of specified sites (C16.1-C16.5) and other
gastric cancers (C16.6-C16.9) (data not shown), we combined the groups in the
analysis.
We excluded subcohort members who reported having prevalent cancer other
than skin cancer at baseline. Also excluded were cases and subcohort members with
incomplete or inconsistent dietary data17 or missing data on exposure or confounding
variables. Figure 6.1 shows that 3962 subcohort members and 107 ESCC, 145 EAC, 164
GCA, and 491 GNCA cases were available for analysis.
Exposure information
All cohort members completed a self-administered questionnaire at baseline. This
questionnaire included a 150-item food frequency questionnaire (FFQ), with questions
on alcohol consumption, and questions on other cancer risk factors, such as smoking
habits, level of education, body mass index (BMI), physical activity, family history of
cancer and use of medication.
We asked about the habitual consumption of alcohol during the year preceding
the start of the study, and this was measured by six items: (1) beer, (2) red wine, (3)
white wine, (4) sherry and other fortified wines, (5) liquor types containing on average
16 alcohol and (6) (Dutch) gin, brandy and whisky. Questions were asked about the
frequency of consumption and the number of glasses consumed on each drinking
occasion. For analysis, we combined (2), (3) and (4) into “wine”, and (5) and (6) into
“liquor”. Mean daily alcohol consumption was calculated using the Dutch food-
composition table.18 Based on a pilot study, a glass of beer, wine and liquor was
assumed to contain 200, 105 and 45 ml of the beverage, respectively (unpublished
findings). For “beer” and “other alcoholic beverages”, participants could indicate
whether five years ago, they drunk (1) more than, (2) equal amounts of or (3) less than
today. The fourth answering option was (4) “I never use this”. Using these questions,
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
109
we selected those participants with stable alcohol consumption to perform a
sensitivity analysis to investigate the robustness of our results.
Netherlands cohort study on diet and cancer (120,852)
↓ ↓
Subcohort
randomly drawn
from total cohort
Record linkage with Netherlands Cancer Registry and PALGA*
↓ ↓ ↓ ↓ ↓
5000
Esophageal
squamous cell
carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
Exclusion of participants with prevalent cancer at baseline
↓ ↓ ↓ ↓ ↓
4774 130 181 206 594
Exclusion of participants with incomplete or inconsistent dietary data
↓ ↓ ↓ ↓ ↓
4438 120 168 187 551
Exclusion of participants with missing data on alcohol consumption or cigarette smoking
↓ ↓ ↓ ↓ ↓
4111 111 153 168 512
Exclusion of participants with missing data on confounders
↓ ↓ ↓ ↓ ↓
3962 107 145 164 491
Figure 6.1 Flow diagram of subcohort members and cases on whom the analyses were based. *PALGA:
nationwide network and registry of histopathology and cytopathology in The Netherlands.
Regarding cigarette smoking, questions were asked about the following: whether
the subject was a smoker at baseline, age at which they started smoking, age at
smoking cessation, the number of cigarettes smoked daily and the number of smoking
years (excluding stopping periods). Based on these questions, the following variables
were constructed: smoking status (never, former, current), current smoking (yes/no),
frequency (number of cigarettes/day), duration (number of years), pack-years of
cigarette smoking (number), and time since cessation (years).
The FFQ has been validated against a 9-day diet record, and the Spearman
correlation coefficient between the alcohol intake assessed by the questionnaire and
that estimated by the diet record was 0.89 for all subjects and 0.85 for users of
alcoholic beverages.17 The reproducibility of the FFQ was established and the test-
retest correlation was 0.90 for alcohol intake, and this correlation declined only 0.01-
0.02 per year.19 This indicates that the single FFQ measurement was able to rank
Thesis Jessie Steevens_v04.pdf
Chapter 6
110
subjects according to alcohol intake and this ability dropped only slightly over time.
The single FFQ measurement that is used in our cohort study can characterize dietary
habits for a period of at least 5 years.19
Questionnaire data were key-entered and processed in a standardized manner,
blinded with respect to case/subcohort status in order to minimize observer bias in
coding and data interpretation.
Data analysis
The confounders considered were:20-22 age (years), sex, level of education (primary,
lower vocational, secondary and medium vocational, and university and higher
vocational), BMI (kg/m2), non-occupational physical activity (<30, 30 to <60, 60 to <90,
≥90 min/day), energy intake (kJ), tea consumption (cups/day), intakes of fruit,
vegetables, legumes, fish, red meat and meat products (all g/day), family history of
esophageal or gastric cancer (yes/no in first-degree relatives), reported long-term
(>0.5 years) use of non-steroidal anti-inflammatory drugs (NSAIDs) or aspirin (ATC
codes M01A, N02B) (yes/no), and reported long-term use (yes/no) of lower esophageal
sphincter (LES)-relaxing medication.23-25 Alcohol consumption and cigarette smoking
were mutually adjusted in the statistical models. A variable was considered a
confounder if including it in the model changed the rate ratio (RR) for any of the cancer
types by >5%. For all four cancers, the same confounders were used in analyses.
Incidence RRs and corresponding 95% CIs for alcohol consumption and cigarette
smoking were estimated in age- and sex-adjusted and multivariable adjusted case-
cohort analyses using Cox proportional hazards models.26 Analyses were done using
Stata 9.2 statistical software package (StataCorp, College Station, Texas, USA).
Standard errors were estimated using the robust Huber-White sandwich estimator to
account for additional variance introduced by sampling from the cohort. This method
is equivalent to the variance-covariance estimator by Barlow.27 The proportional
hazards assumption was tested using the scaled Schoenfeld residuals.28 If the
assumption was violated for a confounder, a time-varying covariate was added to the
model. In case the assumption was violated for the exposure variable, it was checked
whether the time-varying covariate for this variable was statistically significant in the
model. Tests for dose-response trends were assessed by fitting ordinal exposure
variables as continuous terms. Two-sided p values are reported throughout the article.
To evaluate whether substances in alcoholic beverages, other than ethanol, have
an effect on the risk of ESCC, EAC, GCA, or GNCA, we also analyzed beer, wine, and
liquor consumption, adjusted for ethanol intake.
In smoking analyses, the different aspects of smoking were mutually adjusted for,
to investigate which aspect is most important in terms of esophageal or gastric cancer
risk.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
111
Interaction
To evaluate possible interaction between alcohol and cigarette smoking status, we
estimated RRs of all four cancers for combinations of these exposures. The p value for
interaction was assessed by including a cross-product term in the model. In the
interaction analyses it was necessary to change the alcohol categories to include
subjects with low alcohol consumption in the reference category, besides abstainers,
because of too low numbers of cases per stratum.
RESULTS
Descriptives
Table 6.1 presents the characteristics of the subcohort members and cases, including
all variables that were considered potential confounders. Esophageal and gastric
cancer cases were more often men, especially EAC and GCA cases. Alcohol
consumption was higher in cases, especially ESCC cases, than in subcohort members,
while the proportion of alcohol abstainers was higher in the subcohort. About 60% of
subcohort members and cases were stable alcohol consumers. Cases were exposed
more to cigarette smoke in terms of smoking status, frequency, duration, and pack-
years compared with the subcohort.
Cases also differed from subcohort members with respect to some other
characteristics: EAC and GCA cases had a higher BMI than subcohort members. All
cases consumed less fruit and slightly less vegetables than the subcohort, but cases
more often reported a family history of esophageal or gastric cancer and use of LES-
relaxing medication (Table 6.1).
The proportional hazards assumption was violated only in a few analyses (see
footnotes in Table 6.2 and Supplemental Table 6.1 available online).
Alcohol consumption
A daily alcohol consumption of ≥30 g, when compared with abstaining, was associated
with a significantly increased risk of ESCC (multivariable adjusted RR=4.61, 95% CI 2.24
to 9.50, p trend<0.001) (Table 6.2). Women were at somewhat higher risk than men,
and the interaction with sex was statistically significant in continuous analyses
(p=0.04), but not in categorical analyses (p=0.68). No association was observed
between alcohol consumption and EAC (RR for ≥30 g/day=1.04, 95% CI 0.54 to 2.02),
GCA (RR=0.90, 95% CI 0.50 to 1.64) or GNCA (RR=1.00, 95% CI 0.68 to 1.47).
Multivariable RRs for alcohol consumption and ESCC were slightly attenuated
when compared with age- and sex-adjusted RRs (see Supplemental Table 6.1). For EAC,
GCA, and GNCA, age- and sex-adjusted RRs were very similar to multivariable adjusted
RRs.
Thesis Jessie Steevens_v04.pdf
Chapter 6
112
Table 6.1 Characteristics of cases and subcohort members in the Netherlands Cohort Study on diet and
cancer (NLCS), 1986-2002.
Subcohort Cases
Exposure variables and potential
confounders
(n=3962) a
ESCC
(n=107) a
EAC
(n=145) a
GCA
(n=164) a
GNCA
(n=491) a
age at baseline (years) 61.3 (4.2) 62.5 (4.2) 61.6 (4.1) 61.5 (4.1) 62.5 (4.1)
sex
men (n (%)) 1944 (49.1) 59 (55.1) 114 (78.6) 140 (85.4) 331 (67.4)
women (n (%)) 2018 (50.9) 48 (44.9) 31 (21.4) 24 (14.6) 160 (32.6)
abstainer from alcohol (n (%)) 943 (23.8) 16 (15.0) 25 (17.2) 29 (17.7) 107 (21.8)
alcohol consumers:
ethanol intake (g/day) 13.6 (15.1) 27.6 (27.9) 17.3 (18.3) 16.6 (15.5) 15.9 (17.2)
beer intake (glasses/day) 0.3 (0.8) 0.8 (1.8) 0.5 (1.1) 0.5 (1.0) 0.5 (1.2)
wine intake (glasses/day) 0.5 (0.8) 0.6 (1.1) 0.4 (0.7) 0.4 (0.7) 0.4 (0.8)
liquor intake (glasses/day) 0.5 (0.8) 1.1 (1.7) 0.7 (1.0) 0.7 (0.9) 0.6 (0.9)
cigarette smoking status
never smoker (n (%)) 1460 (36.9) 22 (20.6) 28 (19.3) 28 (17.1) 113 (23.0)
former smoker (n (%)) 1425 (36.0) 31 (29.0) 70 (48.3) 80 (48.8) 201 (40.9)
current smoker (n (%)) 1077 (27.2) 54 (50.5) 47 (32.4) 56 (34.2) 177 (36.1)
ever cigarette smokers:
frequency of cigarette smoking (n/day) 15.3 (10.3) 18.3 (10.3) 20.1 (12.6) 17.5 (11.1) 16.5 (10.4)
duration of cigarette smoking (years) 31.6 (12.2) 36.9 (10.5) 34.0 (11.0) 34.9 (11.4) 34.7 (11.8)
pack-years of cigarette smoking (n) 22.7 (17.7) 30.3 (19.0) 30.9 (20.4) 27.2 (18.0) 26.3 (19.2)
level of education
primary (n (%)) 1104 (27.9) 34 (31.8) 37 (25.5) 45 (27.4) 185 (37.7)
lower vocational (n (%)) 868 (21.9) 22 (20.6) 37 (25.5) 42 (25.6) 117 (23.8)
secondary and medium vocational (n (%)) 1420 (35.8) 38 (35.5) 49 (33.8) 45 (27.4) 142 (28.9)
university and higher vocational (n (%)) 570 (14.4) 13 (12.2) 22 (15.2) 32 (19.5) 47 (9.6)
body mass index (kg/m2) 25.0 (3.1) 24.2 (3.6) 26.3 (3.4) 25.7 (3.0) 24.9 (3.2)
non-occupational physical activity
(min/day)
73 (60) 68 (65) 74 (58) 86 (78) 80 (71)
energy intake (kJ/day)
men 9090 (2130) 8703 (2033) 9054 (2025) 8834 (2160) 9166 (2112)
women 7063 (1651) 7406 (1806) 6981 (1807) 7445 (1510) 6995 (1510)
tea consumption (cups/day) 2.8 (2.0) 2.7 (2.4) 2.6 (2.3) 2.3 (1.8) 2.8 (1.9)
fruit consumption (g/day) 177 (120) 130 (108) 169 (135) 154 (116) 164 (125)
vegetable consumption (g/day) 194 (83) 189 (76) 187 (83) 188 (84) 188 (81)
legumes consumption (g/day) 8 (12) 7 (10) 10 (12) 9 (13) 8 (11)
fish consumption (g/day) 13 (15) 14 (13) 15 (19) 15 (20) 14 (19)
red meat consumption (g/day) 87 (40) 93 (41) 90 (37) 92 (45) 91 (41)
meat products consumption (g/day) 13 (15) 14 (14) 14 (15) 15 (16) 15 (16)
family history of esophageal or gastric
cancer (n (%))
296 (7.5) 10 (9.4) 18 (12.4) 14 (8.5) 63 (12.8)
reported long-term use of NSAIDs (n (%)) 262 (6.6) 7 (6.5) 11 (7.6) 9 (5.5) 23 (4.7)
reported long-term use of LES relaxing
medication (n (%))
564 (14.2) 17 (15.9) 28 (19.3) 25 (15.2) 77 (15.7)
a The number of subcohort members or cases used in age- and sex-adjusted analyses and multivariable
analyses of alcohol consumption and cigarette smoking. Values are given as mean (SD); for categorical
variable n (%) is presented. EAC, esophageal adenocarcinoma; ESCC, esophageal squamous cell carcinoma;
GCA, gastric cardia adenocarcinoma; GNCA, gastric non-cardia adenocarcinoma; LES, lower esophageal
sphincter; NSAIDs, non-steroidal anti-inflammatory drugs.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
113
When the analyses were restricted to stable alcohol consumers, the association
with ESCC became somewhat stronger. The results for EAC, GCA, and GNCA barely
changed (Table 6.2).
Results regarding alcoholic beverages are shown in table 2. After adjustment for
total alcohol intake, beer consumption was associated with an increased risk of ESCC
(p trend=0.23) and GNCA (p trend = 0.06). Wine consumption was inversely associated
with risk of ESCC (RR per 1 glass/day increment=0.67, 95% CI 0.50 to 0.90), but was not
associated with EAC, GCA or GNCA. Consumption of liquor was not significantly
associated with risks of ESCC, EAC, and GCA, and with a reduced risk of GNCA.
Cigarette smoking
Current and former cigarette smoking were associated with an increased risk of all four
cancer types, when compared with never smokers (Table 6.3). The strongest
associations were found for ESCC and GNCA.
Frequency of cigarette smoking was found to be associated with increased risks of
all four cancers, independent of other smoking aspects. The association was
statistically significant for ESCC and EAC. The RRs for the gastric cancers were lower
than for the esophageal cancers (Table 6.3). Duration of cigarette smoking was
independently associated with increased risks of ESCC, GCA, and GNCA, but the RR was
only statistically significant for GNCA. Pack-years of smoking were associated with
increased risks (mostly statistically significant) of all four cancers. The highest RRs were
found in those who smoked ≥40 pack-years.
We also looked into the effect of cigarette smoking cessation on the risk of
esophageal and gastric cancers. The risks of all four cancer types declined for smokers
who had stopped since <10, 10 to <20 or ≥20 years (p trend <0.05 for all four cancers),
when compared with current smokers. However, when compared with never smokers,
the risks were still elevated for ESCC, EAC, and GNCA. For ESCC, EAC, and GCA, the RRs
of time since smoking cessation were attenuated when we adjusted for smoking
duration (data not shown).
Multivariable RRs for cigarette smoking were attenuated when compared with
age- and sex-adjusted RRs (Supplemental Table 6.2). This difference could largely be
explained by the mutual adjustment of cigarette smoking aspects. For ESCC,
confounding by alcohol consumption also caused attenuation of the associations.
Thesis Jessie Steevens_v04.pdf
Chapter 6
114
Thesis Jessie Steevens_v04.pdf
Table 6.2 Association (multivariablea adjusted incidence RRs) between alcohol consumption and risk of esophageal and gastric cancer subtypes; Netherlands Cohort
Study on diet and cancer (NLCS) 1986-2002.
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric non-cardia
adenocarcinoma
categorical
median
person
time at risk
(years)
no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI)
Alcohol consumption (g ethanol/day)
abstainer 0 13336 16 1 (reference) 25 1 (reference) 29 1 (reference) 107 1 (reference)
>0-<5 2 16740 15 0.85 (0.42-1.73) 38 1.17 (0.69-1.98) 30 0.75 (0.44-1.30) 121 0.92 (0.69-1.23)
5-<15 9 12714 23 1.65 (0.85-3.17) 30 0.91 (0.51-1.60) 47 1.11 (0.68-1.82) 126 1.07 (0.80-1.43)
15-<30 22 8979 24 2.11 (1.08-4.14) 31 1.01 (0.56-1.82) 36 0.92 (0.54-1.57) 78 0.77 (0.55-1.08)
>=30 40 5037 29 4.61 (2.24-9.50) 21 1.04 (0.54-2.02) 22 0.90 (0.50-1.64) 59 1.00 (0.68-1.47)
p trend
b <0.001 p trend 0.93 p trend 0.99 p trend 0.68
continuous, 10 grams ethanol/day
increments
overall 56806 107 1.32 (1.19-1.45) 145 1.01 (0.90-1.14) 164 0.98 (0.88-1.08) 491 1.02 (0.95-1.09)
men 26679 59 1.28 (1.15-1.43) 114 0.99 (0.88-1.12) 140 0.98 (0.89-1.09) 331 1.03 (0.95-1.11)
women 30128 48 1.62 (1.31-2.00) 31 1.23 (0.93-1.64) 24 0.84 (0.41-1.75) 160 0.96 (0.78-1.18)
p interaction
c 0.04 p interaction 0.41 p interaction 0.65 p interaction 0.47
Alcohol consumption (g ethanol/day)
Stable users d
abstainer 0 10763 13 1 (reference) 21 1 (reference) 23 1 (reference e) 88 1 (reference)
>0-<5 2 9780 8 0.74 (0.30-1.80) 19 0.91 (0.48-1.73) 12 0.46 (0.22-0.97) 75 0.91 (0.65-1.28)
5-<15 9 7831 13 1.59 (0.72-3.49) 19 0.88 (0.46-1.68) 22 0.78 (0.41-1.47) 75 0.98 (0.69-1.39)
15-<30 22 4932 14 2.44 (1.06-5.60) 18 1.07 (0.53-2.16) 19 0.83 (0.41-1.69) 39 0.65 (0.42-1.01)
>=30 40 2735 15 5.34 (2.16-13.18) 15 1.43 (0.66-3.11) 10 0.76 (0.34-1.73) 33 0.98 (0.60-1.61)
p trend <0.001 p trend 0.28 p trend 0.79 p trend 0.52
continuous, 10 g ethanol/day
increments
36041 63 1.42 (1.24-1.63) 92 1.08 (0.95-1.24) 86 0.99 (0.85- 1.14) 310 0.99 (0.89-1.09)
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
115
Thesis Jessie Steevens_v04.pdf
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric non-cardia
adenocarcinoma
categorical
median
person
time at risk
(years)
no. of
cases
RR (95% CI) no. of
cases
RR (95% CI) no. of
cases
RR (95% CI) no. of
cases
RR (95% CI)
Alcoholic beverages (glasses/day) f
Beer
no beer 0 38324 60 1 (reference) 79 1 (reference) 86 1 (reference g) 294 1 (reference)
>0-1 0.2 14606 28 0.98 (0.57-1.67) 48 0.84 (0.55-1.28) 59 0.86 (0.59-1.25) 140 0.90 (0.70-1.15)
>1-2 1.4 2543 8 1.36 (0.59-3.16) 10 0.84 (0.41-1.71) 12 0.86 (0.44-1.68) 30 1.02 (0.65-1.60)
>2 3.7 1333 11 1.62 (0.64-4.09) 8 1.07 (0.44-2.62) 7 0.90 (0.39-2.07) 27 1.58 (0.95-2.63)
p trend 0.23 p trend 0.80 p trend 0.87 p trend 0.06
continuous, 1 glass/day
increments
56806 107 1.10 (0.92-1.32) 145 0.98 (0.82- 1.17) 164 1.03 (0.86- 1.24) 491 1.15 (1.03-1.29)
Wine
no wine 0 26060 50 1 (reference e) 74 1 (reference) 84 1 (reference) 268 1 (reference)
>0-1 0.2 23467 42 0.93 (0.60-1.44) 55 0.97 (0.64-1.47) 64 1.01 (0.69-1.49) 169 0.88 (0.70-1.10)
>1-2 1.4 4806 8 0.51 (0.22-1.17) 11 0.92 (0.45-1.87) 9 0.69 (0.31-1.54) 36 0.91 (0.59-1.39)
>2 2.7 2473 7 0.30 (0.07-1.23) 5 0.79 (0.28-2.20) 7 1.04 (0.40-2.70) 18 0.88 (0.48-1.63)
p trend 0.05 p trend 0.64 p trend 0.70 p trend 0.73
continuous, 1 glass/day
increments
56806 107 0.67 (0.50-0.90) 145 0.89 (0.67- 1.19) 164 0.87 (0.62-1.21) 491 0.93 (0.77-1.13)
Liquor
no liquor 0 29359 41 1 (reference) 60 1 (reference) 60 1 (reference) 230 1 (reference)
>0-1 0.2 20833 31 1.13 (0.69-1.84) 53 0.96 (0.65-1.41) 69 1.14 (0.79-1.65) 183 0.93 (0.74-1.15)
>1-2 1.9 4714 21 1.82 (0.97-3.41) 20 1.11 (0.61-2.03) 29 1.43 (0.81-2.53) 59 0.84 (0.58-1.21)
>2 2.8 1900 14 1.55 (0.64-3.78) 12 1.53 (0.68-3.48) 6 0.72 (0.24-2.18) 19 0.58 (0.33-1.03)
p trend 0.11 p trend 0.36 p trend 0.68 p trend 0.12
continuous, 1 glass/day
increments
56806 107 1.21 (0.92-1.60) 145 1.12 (0.87-1.43) 164 1.07 (0.79-1.45)e491 0.90 (0.76-1.07)
a Adjusted for age, sex, cigarette smoking (current smoking status (yes/no), frequency, and duration), body mass index, level of education, energy intake, consumption of
fruits, vegetables, fish; b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model; c p value for
interaction between sex and alcohol consumption, based on cross product term in the Cox proportional hazard model; d Subjects who had not changed their alcohol
consumption habits in the five years before baseline; e Proportional hazards assumption was violated for the exposure variable, but there was no statistically significant
interaction with time; f Additionally adjusted for ethanol intake; g Proportional hazards assumption was violated for the exposure variable, and there was a statistically
significant interaction with time. CI, confidence interval; RR, rate ratio.
Chapter 6
116
Thesis Jessie Steevens_v04.pdf
Table 6.3 Association (multivariable adjusteda incidence rate ratios) between cigarette smoking and risk of esophageal and gastric cancer subtypes; Netherlands
Cohort Study on diet and cancer (NLCS) 1986-2002.
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric non-cardia
adenocarcinoma
categorical
median
person time
at risk (years)
no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI) no.
of
cases
RR (95% CI)
Smoking status
never smokers 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
former smokers 20271 31 1.38 (0.72-2.63) 70 1.47 (0.92-2.35) 80 1.40 (0.86-2.27) 201 1.47 (1.11-1.96)
current smokers 14577 54 2.63 (1.47-4.69) 47 1.67 (1.01-2.77) 56 1.60 (0.97-2.66) 177 1.86 (1.40-2.48)
p trend
b <0.001 p trend 0.05 p trend 0.07 p trend <0.001
Smoking status, additional adjustment for smoking
frequency and duration
never smokers 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
former smokers 20271 31 0.68 (0.27-1.74) 70 0.81 (0.37-1.77) 80 0.80 (0.37-1.76) 201 1.09 (0.70-1.70)
current smokers 14577 54 1.07 (0.38-3.04) 47 0.87 (0.31-2.40) 56 0.74 (0.27-2.04) 177 1.26 (0.71-2.25)
p trend 0.27 p trend 0.98 p trend 0.63 p trend 0.34
Frequency of cigarette smoking (n/day)
no cigarette smoker 0 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
>0 - <20 10 21933 44 0.79 (0.32-1.97) 50 0.98 (0.46-2.06) 74 0.84 (0.39-1.81) 222 1.16 (0.76-1.77)
≥20 20 12916 41 1.15 (0.44-2.99) 67 1.77 (0.80-3.88) 62 0.92 (0.40-2.14) 156 1.31 (0.81-2.11)
p trend 0.24 p trend 0.01 p trend 0.82 p trend 0.22
continuous, 10 cigarettes/day
increments
56806 107 1.18 (1.00-1.40) 145 1.26 (1.08-1.46) 164 1.04 (0.89-1.22) 491 1.09 (0.98-1.20)
Duration of cigarette smoking (years)
no cigarette smoker 0 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
>0 - <20 12 6623 4 0.48 (0.15-1.57) 13 0.78 (0.37-1.66) 16 1.04 (0.51-2.11) 52 1.34 (0.90-1.98)
20 - <40 30 16850 33 1.26 (0.60-2.67) 61 1.11 (0.64-1.92) 59 1.28 (0.72-2.27) 153 1.27 (0.91-1.77)
≥40 43 11376 48 1.72 (0.73-4.03) 43 0.89 (0.45-1.73) 61 1.62 (0.82-3.20) 173 1.51 (1.00-2.29)
p trend 0.08 p trend 0.88 p trend 0.14 p trend 0.10
continuous, 10 years increments 56806 107 1.11 (0.93-1.33) 145 1.04 (0.90-1.20) 164 1.14 (0.98-1.33) 491 1.10 (1.00-1.20)
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
117
Thesis Jessie Steevens_v04.pdf
Subcohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric non-cardia
adenocarcinoma
categorical
median
person time
at risk (years)
no.
of
cases RR (95% CI)
no.
of
cases RR (95% CI)
no.
of
cases RR (95% CI)
no.
of
cases RR (95% CI)
Pack-years of cigarette smoking
no cigarette smoker 0 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
>0 - <20 9 18505 25 1.07 (0.53-2.17) 38 1.15 (0.68-1.94) 52 1.23 (0.73-2.07) 154 1.32 (0.98-1.78)
20 - <40 28 11247 39 2.14 (1.02-4.46) 42 1.72 (0.97-3.05) 55 1.67 (0.96-2.90) 153 1.88 (1.34-2.63)
≥40 48 5096 21 2.22 (1.01-4.86) 37 2.93 (1.59-5.40) 29 1.74 (0.93-3.25) 71 1.68 (1.12-2.52)
p trend 0.01 p trend <0.001 p trend 0.04 p trend 0.004
Continuous, 10 pack-years
increments
56806 107 1.14 (1.03-1.25) 145 1.16 (1.07-1.26) 164 1.07 (0.99-1.16) 491 1.08 (1.02-1.15)
Smoking cessation
never smokers 21958 22 1 (reference) 28 1 (reference) 28 1 (reference) 113 1 (reference)
stopped ≥ 20 years 25 6280 10 1.46 (0.63-3.42) 19 1.32 (0.70-2.47) 18 1.00 (0.53-1.91) 49 1.13 (0.77- 1.67)
stopped 10 - <20 years 14 6953 10 1.28 (0.58-2.84) 23 1.42 (0.80-2.51) 28 1.43 (0.81-2.52) 66 1.41 (0.98- 2.02)
stopped >0 - <10 years 5 6984 11 1.42 (0.62-3.23) 28 1.66 (0.95-2.91) 34 1.72 (0.97-3.05) 85 1.81 (1.30- 2.52)
current smokers 0 14577 54 2.63 (1.47-4.69) 47 1.68 (1.01-2.78) 56 1.61 (0.97-2.66) 177 1.86 (1.39- 2.47)
p trend 0.001 p trend 0.03 p trend 0.02 p trend <0.001
a all analyses were adjusted for age, sex, alcohol consumption, body mass index, level of education, energy intake, consumption of fruits, vegetables, fish. Analyses of
smoking frequency, duration and pack-years of smoking were adjusted for current smoking status (yes/no). Additionally, smoking frequency and duration were mutually
adjusted for in analyses; b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard model. CI,
confidence interval; RR, rate ratio.
Chapter 6
118
Interaction
Table 6.4 shows RRs for several exposure combinations of alcohol consumption and
cigarette smoking, using never smokers and low alcohol consumers as a reference. A
graphical presentation of these results can be found in Figure 6.2A-D.
The RR of ESCC was 8.05 (95% CI 3.89 to 16.60) in the category of current smokers
who consumed >15 g/day of ethanol. This RR is compatible with multiplicative
interaction, but was not statistically significant (p interaction=0.65). We found no
evidence for interaction between alcohol consumption and cigarette smoking on risk
of EAC, GCA, or GNCA.
Figure 6.2 Combined exposure to alcohol and cigarette smoking and risk of esophageal and gastric
cancer subtypes; Netherlands Cohort Study on diet and cancer, 1986-2002. (A) Esophageal
squamous cell carcinoma, (B) Esophageal adenocarcinoma, (C) Gastric cardia
adenocarcinoma, (D) Gastric non-cardia adenocarcinoma. Multivariable RRs: adjusted for
age, sex, body mass index, level of education, energy intake, consumption of fruits,
vegetables, fish.
never for me r current
0-5
>5-15
>15
1
2
3
4
5
6
7
8
RR (95% CI
)
Cigarette smoking status
Alcohol
consumption
(grams/day)
A
never for me r current
0-5
>5-15
>15
1
2
3
4
5
6
7
8
RR (95% CI
)
Cigarette smoking status
Alcohol
consumption
(grams/day)
B
never for me r current
0-5
>5-15
>15
1
2
3
4
5
6
7
8
RR (95% CI
)
Cigarette smoking status
Alcohol
consumption
(grams/day)
C
never for me r current
0-5
>5-15
>15
1
2
3
4
5
6
7
8
RR (95% CI
)
Cigarette smoking status
Alcohol
consumption
(grams/day)
D
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
119
Thesis Jessie Steevens_v04.pdf
Table 6.4 Combinations of alcohol consumption and cigarette smoking and risk (multivariable adjusted a incidence RRs) of esophageal and gastric cancer subtypes;
Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002.
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric non-cardia
adenocarcinoma
Alcohol consumption
(g ethanol/ day)
Alcohol consumption
(g ethanol/ day)
Alcohol consumption
(g ethanol/ day)
Alcohol consumption
(g ethanol/ day)
Cigarette smoking status 0-5 >5-15 >15 0-5 >5-15 >15 0-5 >5-15 >15 0-5 >5-15 >15
never smoker
cases/person time at risk
(years) 14/16351 3/3750 5/1857 18/16351 5/3750 5/1857 18/16351 6/3750 4/1857 81/16351 18/3750 14/1857
RR 1 (ref) 1.01 3.74 1 (ref) 0.92 1.76 1 (ref) 0.92 1.15 1 (ref) 0.89 1.40
95% CI 0.28-3.59 1.25-11.20 0.32-2.67 0.61-5.06 0.35-2.41 0.37-3.54 0.52-1.52 0.76-2.58
former smoker
cases/person time at risk
(years) 8/7830 8/5600 15/6842 28/7830 14/5600 28/6842 23/7830 31/5600 26/6842 73/7830 63/5600 65/6842
RR 1.33 2.04 3.16 1.79 1.24 1.72 1.15 2.08 1.18 1.41 1.78 1.40
95% CI 0.53-3.35 0.81-5.15 1.36-7.31 0.93-3.44 0.58-2.64 0.88-3.39 0.57-2.31 1.10-3.91 0.61-2.32 0.98-2.02 1.21-2.62 0.96-2.04
current smoker
cases/person time at risk
(years) 9/5896 12/3364 33/5317 17/5896 11/3364 19/5317 18/5896 10/3364 28/5317 74/5896 45/3364 58/5317
RR 1.70 4.48 8.05 1.87 1.87 1.81 1.51 1.27 1.89 2.04 2.18 1.62
95% CI 0.72-4.05 1.97-10.20 3.89-16.60 0.92-3.82 0.83-4.21 0.87-3.79 0.75-3.01 0.56-2.90 0.97-3.68 1.43-2.91 1.45-3.27 1.09-2.41
p interactionb0.65 p interaction 0.79 p interaction 0.29 p interaction 0.42
a Adjusted for age, sex, body mass index, level of education, energy intake, consumption of fruits, vegetables, fish; b p value for interaction between cigarette smoking
status and alcohol consumption was assessed by including a cross-product term in the Cox proportional hazard model. CI, confidence interval; ref, reference category; RR,
rate ratio.
Chapter 6
120
DISCUSSION
Main findings
In this study, which is one of the first prospective cohort studies on the topic, we found
the following results. Alcohol consumption was related to a strongly increased risk of
ESCC, whereas it was unrelated to EAC, GCA, or GNCA. Cigarette smoking status,
frequency, duration, and pack-years were associated with increased risks of all four
cancer types. A multiplicative interaction was found between alcohol consumption and
cigarette smoking on the risk of ESCC, but no such interaction existed for EAC, GCA, or
GNCA.
Strengths and limitations
A strength of our study includes the prospective character, which makes selection and
information bias unlikely. Moreover, the division we made into histological subtypes of
esophageal cancer and localizations of gastric cancer is important, as this allows the
evaluation of different risk factors for the cancer subtypes and more precise estimates
of the strength of the associations. Our study is also one of the largest prospective
cohort studies investigating this topic.
The ranges of the participants’ alcohol consumption and cigarette smoking habits
were wide, which gave good contrasts between high and low exposures. Also, alcohol
consumption and cigarette smoking habits were addressed extensively in the
questionnaire. We lack data on smoking and alcohol consumption after baseline, but
we did ask participants about their habits during a long period before baseline. As the
development of a tumor probably takes several decades, we believe we asked about
the exposure in a relevant time window.
Participants could have changed their alcohol consumption and smoking habits
due to preclinical cancer, which may influence results. This is unlikely though, because
the smoking data reflect the lifetime exposure of the participants. As for alcohol,
results were robust for restriction of analyses to stable users.
In this epidemiological study, we analyze the possibility of residual confounding,
by, for example, Helicobacter pylori infection. This infection is associated with
increased risk of GNCA,29 and possibly with decreased risk of EAC,30,31 while it is not
associated with GCA29 or ESCC risk.31 Unfortunately, we lack data on H pylori infection,
but we estimate the prevalence to have been ~50%.32 The infection might have
confounded the associations if it is correlated with alcohol consumption or smoking.
However, studies investigating this correlation have been inconsistent.33-35 Residual
confounding by other variables cannot be completely ruled out, but the associations
found were strong and residual confounding probably cannot explain these
associations entirely. Moreover, the true associations may be stronger than observed,
because we probably underestimated their strength due to random measurement
error.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
121
Comparison with previous research: alcohol
The strong positive association between alcohol consumption and ESCC risk we found
is in agreement with associations found in the other prospective cohorts.9,10,36 The
association we found showed a statistically significant trend (p<0.001), although low
consumers of alcohol were found to be at slightly decreased ESCC risk, and this
association remained when we restricted our analyses to the group of stable alcohol
consumers. These findings are consistent with findings from two other cohorts,10,36 and
an explanation was suggested by Freedman et al.10 Possibly, some of the subjects who
were abstainers at baseline used to be heavy drinkers long >5 years) before baseline.
These drinkers remain at increased risk for a decade after cessation of alcohol
consumption.37
Our findings that alcohol was not associated with EAC, GCA and GNCA confirm the
null results found in two other Western cohorts.9,10
To investigate whether substances in alcoholic beverages other than ethanol are
relevant for cancer risk, we adjusted analyses of these beverages for total ethanol
intake. This approach has not been used for esophageal or gastric cancers before and
gave some new insights. For the beverages, ESCC risk was lower compared with the
association with ethanol. This suggests that ethanol is probably the key substance in
these beverages, and this suggestion is strengthened by the fact that associations have
been found across the world, for many different alcoholic beverages.10,38,39 Still, other
substances in alcoholic beverages might be relevant. For instance, flavonoids40 in wine
might explain the inverse association we found with ESCC. This inverse association has
also been found by a case-control study,41 but this association was unadjusted for
ethanol intake. Regarding beer, N-nitroso compounds present in this beverage may be
partly responsible for the positive associations with ESCC and GNCA.42
Comparison with previous research: cigarette smoking
We found that cigarette smoking was associated with an increased risk of all
esophageal and gastric cancer subtypes. For these cancer subtypes, we are the first
cohort study to analyze the association with all kinds of smoking aspects. Our findings
are consistent with previous reviews.43-45 It appeared that smoking frequency as well
as duration were important in terms of increasing a person’s risk. Therefore, the
number pack-years smoked may be a good indicator of a person’s total exposure.
The relationships we found between smoking and esophageal cancer types were
weaker compared with a previous case-control45 and two cohort10,11 studies, but
similar to the results of a third cohort study.9 There may be several reasons why other
studies found stronger associations: under-reporting of smoking habits in case-control
studies, no mutual adjustment for smoking aspects, or no or insufficient adjustment
for confounding (e.g., by alcohol consumption). Our RRs might be attenuated because
some smokers may have stopped smoking after baseline, but this is unlikely because
after smoking cessation people stay at risk for a long period.
Thesis Jessie Steevens_v04.pdf
Chapter 6
122
Our results showed that smoking cessation decreased risks of all four cancers, but
only after one to two decades and these results are in agreement with two
reviews.43,46
Interaction
This is the first prospective study to investigate the interaction between alcohol
consumption and cigarette smoking and the risk of ESCC. An interaction for ESCC has
been established by previous case-control studies (see, for example,4,41,47,48). We
confirm this interaction, although this was not statistically significant, probably due to
low power. Yet, for EAC, GCA and GNCA it was uncertain whether a similar interaction
is present, as only one study examined this.10 According to our findings, no interaction
was present.
The interaction between alcohol consumption and cigarette smoking can be
explained biologically. Cigarette smoke contains many carcinogens, and the ethanol
from alcoholic beverages is metabolized to acetaldehyde, which was classified as
human carcinogen by the International Agency for Research on Cancer (IARC).49
Acetaldehyde can circulate in the blood after formation in the liver, but can also be
formed locally by oral bacteria.50 Ethanol itself can cause local irritation of the upper
gastrointestinal tract51 and may facilitate the uptake of the carcinogens present in
cigarette smoke.50
In conclusion, alcohol and cigarette smoke each have an individual effect on ESCC
risk, but, when combined, they act synergistically. We found no interaction for EAC,
GCA, or GNCA.
Conclusions and recommendations
In summary, we found alcohol to be positively associated with ESCC risk, but not with
EAC, GCA or GNCA. Cigarette smoking was positively associated with risk of all four
cancer types. Alcohol consumption and smoking interacted in a multiplicative way on
the risk of ESCC.
Alcohol consumption cannot explain the previously mentioned rising incidences of
EAC and GCA, because it was not associated with risk of either. Cigarette smoking
habits cannot explain the increasing trends either, because these habits have not
increased over time in Western countries. Therefore, we suggest that further research
should focus on other risk factors for esophageal and gastric cancer subtypes, to
search for explanations for these increases.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
123
Thesis Jessie Steevens_v04.pdf
Supplemental Table 6.1 Association (age- and sex-adjusted incidence rate ratios) between alcohol consumption and risk of esophageal and gastric cancer subtypes;
Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002.
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
person time
at risk (years)
no.
cases
RR a (95% CI)
a no.
cases
RR (95% CI) no.
cases
RR (95% CI) no.
cases
RR (95% CI)
Alcohol consumption (grams ethanol/day)
Abstainer 0 13336 16 1 reference 25 1 reference 29 1 reference 107 1 reference
>0-<5 2 16740 15 0.78 0.38-1.60 38 1.13 0.68-1.89 30 0.76 0.45-1.28 121 0.90 0.68-1.19
5-<15 9 12714 23 1.66 0.86-3.20 30 0.90 0.52-1.57 47 1.13 0.70-1.82 126 1.05 0.79-1.39
15-<30 22 8979 24 2.51 1.29-4.87 31 1.13 0.65-1.98 36 1.02 0.62-1.70 78 0.82 0.60-1.13
>=30 40 5037 29 5.52 2.77-11.00 21 1.24 0.67-2.27 22 0.99 0.56-1.76 59 1.03 0.72-1.48
p-trend
b <0.001 p-trend 0.50 p-trend 0.63 p-trend 0.98
continuous, 10 g ethanol/day increments
overall 56806 107 1.34 1.22-1.46 145 1.05 0.95-1.16 164 1.00 0.91-1.10 491 1.02 0.96-1.09 c
men 26679 59 1.30 1.18-1.42 114 1.04 0.94-1.16 140 1.01 0.92-1.10 331 1.03 0.96-1.10
women 30128 48 1.69 1.42-2.02 31 1.16 0.89-1.52 24 0.88 0.44-1.75 160 0.98 0.81-1.18
p-interaction d 0.01 p-interaction 0.51 p-interaction 0.67 p-interaction 0.63
Alcohol consumption (grams ethanol/day)
Stable users e
abstainer 0 10763 13 1 reference 21 1 reference 23 1 reference 88 1 reference
>0-<5 2 9780 8 0.71 0.29-1.73 19 0.86 0.46-1.62 12 0.48 0.23-0.98 75 0.89 0.64-1.24
5-<15 9 7831 13 1.53 0.69-3.37 19 0.83 0.44-1.58 22 0.80 0.43-1.47 75 0.95 0.67-1.33
15-<30 22 4932 14 2.77 1.25-6.17 18 1.10 0.57-2.14 19 0.93 0.49-1.77 39 0.72 0.48-1.09
>=30 40 2735 15 5.47 2.42-12.35 15 1.55 0.77-3.12 10 0.81 0.38-1.75 33 1.03 0.66-1.62
p-trend <0.001 p-trend 0.11 p-trend 0.62 p-trend 0.81
continuous, 10 grams ethanol/day
increments
36041 63 1.40 1.27-1.55 92 1.12 0.99-1.26 86 1.00 0.88-1.13 310 1.01 0.92-1.10
Chapter 6
124
Thesis Jessie Steevens_v04.pdf
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
person time
at risk (years)
no.
cases
RR a (95% CI)
a no.
cases
RR (95% CI) no.
cases
RR (95% CI) no.
cases
RR (95% CI)
Alcoholic beverages (glasses/day) f
Beer
no beer 0 38324 60 1 reference 79 1 reference 86 1 reference 294 1 reference
>0-1 0.2 14606 28 0.96 0.55-1.69 48 0.84 0.55-1.28 59 0.86 0.60-1.25 140 0.89 0.69-1.13
>1-2 1.4 2543 8 1.35 0.57-3.16 10 0.90 0.44-1.82 12 0.91 0.47-1.77 30 1.10 0.70-1.72
>2 3.7 1333 11 2.12 0.85-5.30 8 1.28 0.55-2.97 7 1.01 0.45-2.28 27 1.91 1.15-3.15
p-trend 0.06 p-trend 0.48 p-trend 0.89 p-trend 0.01
continuous, 1 glass/day increments 56806 107 1.15 0.98-1.34 145 1.04 0.87-1.23 164 1.06 0.89-1.26 491 1.20 1.07-1.34
Wine
no wine 0 26060 50 1 reference g 74 1 reference 84 1 reference 268 1 reference
>0-1 0.2 23467 42 0.78 0.51-1.20 55 0.84 0.58-1.21 64 0.89 0.63-1.26 169 0.71 0.58-0.88
>1-2 1.4 4806 8 0.45 0.20-1.01 11 0.75 0.38-1.45 9 0.62 0.29-1.33 36 0.72 0.48-1.09
>2 2.7 2473 7 0.27 0.07-1.06 5 0.54 0.19-1.52 7 0.83 0.34-2.01 18 0.62 0.34-1.11
p-trend 0.03 p-trend 0.21 p-trend 0.42 p-trend 0.12
continuous, 1 glass/day increments 56806 107 0.67 0.50-0.89c 145 0.81 0.60-1.08 164 0.82 0.60-1.13 491 0.82 0.68-1.00
Liquor
no liquor 0 29359 41 1 reference c 60 1 reference 60 1 reference 230 1 reference
>0-1 0.2 20833 31 1.01 0.62-1.64 53 0.94 0.63-1. 69 1.14 0.79-1.65 183 0.91 0.73-1.13
>1-2 1.9 4714 21 1.96 1.03-3.72 20 1.19 0.65-2.18 29 1.54 0.87-2.73 59 0.92 0.64-1.34
>2 2.8 1900 14 1.53 0.59-3.95 12 1.72 0.74-4.03 6 0.76 0.25-2.34 19 0.65 0.36-1.17
p-trend 0.08 p-trend 0.21 p-trend 0.52 p-trend 0.39
continuous, 1 glass/day increments 56806 107 1.20 0.89-1.63 145 1.15 0.89-1.50 164 1.10 0.81-1.50 c 149 0.95 0.80-1.13
a CI, confidence interval; RR, rate ratio; b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard
model; c Proportional hazards assumption was violated for the exposure variable, and there was a statistically significant interaction with time; d p-value for interaction
between sex and alcohol consumption, based on cross product term in the Cox proportional hazard model; e subjects who had not changed their alcohol consumption
habits in the five years before baseline; f additionally adjusted for ethanol intake; g proportional hazards assumption was violated for the exposure variable, but there was
no statistically significant interaction with time.
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
125
Thesis Jessie Steevens_v04.pdf
Supplemental Table 6.2 Association (age- and sex-adjusted incidence rate ratios) between cigarette smoking and risk of esophageal and gastric cancer subtypes;
Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002.
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
person time
at risk
(years)
no.
cases
RR a (95% CI)
a no.
cases
RR (95% CI) no.
cases
RR (95% CI) no.
cases
RR (95% CI)
Smoking status
never smokers 21958 22 1 reference 28 1 reference 28 1 reference 113 1 reference
former smokers 20271 31 1.58 0.85-2.95 70 1.45 0.92-2.29 80 1.35 0.84-2.17 201 1.41 1.08-1.85
current smokers 14577 54 4.09 2.34-7.14 47 1.57 0.96-2.55 56 1.54 0.94-2.51 177 1.95 1.49-2.55
p-trend
b <0.001 p-trend 0.09 p-trend 0.09 p-trend <0.001
Frequency of cigarette smoking (no./day)
no cigarette smoker 0 21958 22 1 reference 28 1 reference 28 1 reference 113 1 reference
>0-<20 10 21933 44 2.27 1.27-4.04 50 1.12 0.69-1.81 74 1.33 0.83-2.14 222 1.58 1.22-2.04
>=20 20 12916 41 3.88 2.09-7.21 67 2.25 1.39-3.63 62 1.59 0.97-2.61 156 1.79 1.34-2.40
p-trend <0.001 p-trend <0.001 p-trend 0.07 p-trend <0.001
continuous, 10 cigarettes/day increments
overall 56806 107 1.43 1.26-1.63 145 1.32 1.17-1.48 164 1.13 1.00-1.28 491 1.16 1.07-1.25
men 26679 59 1.25 1.07-1.47 114 1.34 1.18-1.51 140 1.13 0.99-1.28 331 1.14 1.05-1.24
women 30128 48 1.96 1.57-2.44 31 1.13 0.73-1.76 24 1.17 0.78-1.78 160 1.24 1.04-1.48
p-interactionc 0.001 p-interact ion 0.40 p-interaction 1.00 p-interaction 0.42
Duration of cigarette smoking (years)
no cigarette smoker 0 21958 22 1 reference 28 1 reference 28 1 reference 113 1 reference
>0-<20 12 6623 4 0.73 0.24-2.21 13 1.03 0.52-2.01 16 1.07 0.55-2.08 52 1.39 0.96-2.00
20-<40 30 16850 33 2.41 1.31-4.44 61 1.68 1.06-2.67 59 1.34 0.83-2.18 153 1.47 1.11-1.95
>=40 43 11376 48 4.82 2.64-8.80 43 1.51 0.90-2.53 61 1.72 1.05-2.83 173 2.06 1.56-2.72
p-trend <0.001 p-trend 0.04 p-trend 0.02 p-trend <0.001
continuous, 10 years increments 56806 107 1.41 1.24-1.61 145 1.13 1.02-1.26 164 1.15 1.03-1.27 491 1.17 1.11-1.25
Chapter 6
126
Thesis Jessie Steevens_v04.pdf
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
person time
at risk
(years)
no.
cases
RR a (95% CI)
a no.
cases
RR (95% CI) no.
cases
RR (95% CI) no.
cases
RR (95% CI)
Pack-years of cigarette smoking
no cigarette smoker 0 21958 22 1 reference 28 1 reference 28 1 reference 113 1 reference
>0-<20 9 18505 25 1.62 0.86-3.03 38 1.05 0.63-1.75 52 1.17 0.71-1.94 154 1.37 1.04-1.81
20-<40 28 11247 39 4.53 2.41-8.48 42 1.65 0.99-2.76 55 1.68 1.01-2.78 153 2.10 1.57-2.81
>=40 48 5096 21 4.99 2.47-10.05 37 2.98 1.74-5.11 29 1.78 1.01-3.13 71 1.90 1.34-2.69
p-trend <0.001 p-trend <0.001 p-trend 0.01 p-trend <0.001
continuous, 10 pack-years increments
overall 56806 107 1.27 1.18-1.37 145 1.18 1.10-1.27 164 1.09 1.02-1.17 491 1.11 1.06-1.17
men 26679 59 1.20 1.10-1.31 114 1.19 1.10-1.28 140 1.09 1.02-1.17 331 1.11 1.05-1.16
women 30128 48 1.50 1.32-1.69 31 1.14 0.88-1.46 24 1.08 0.83-1.40 160 1.15 1.03-1.28
p-interaction 0.004 p-interaction 0.66 p-interaction 0.81 p-interaction 0.64
Smoking cessation
never smokers 21958 22 1 reference 28 1 reference 28 1 reference 113 1 reference
stopped >= 20 years 25 6280 10 1.60 0.71-3.62 19 1.21 0.66-2.22 18 0.93 0.49-1.74 49 1.06 0.73- 1.55
stopped 10-<20 years 14 6953 10 1.50 0.69-3.27 23 1.38 0.78-2.43 28 1.36 0.78-2.37 66 1.34 0.95- 1.89
stopped >0-<10 years 5 6984 11 1.67 0.74-3.75 28 1.77 1.03-3.03 34 1.75 1.01-3.03 85 1.79 1.30- 2.47
current smokers 0 14577 54 4.10 2.35-7.15 47 1.56 0.96-2.55 56 1.54 0.94-2.51 177 1.94 1.48- 2.54
p-trend <0.001 p-trend 0.04 p-trend 0.02 p-trend <0.001
a CI, confidence interval; RR, rate ratio; b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the Cox proportional hazard
model; c p-value for interaction between sex and cigarette smoking frequency, based on cross product term in the Cox proportional hazard model.
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
127
REFERENCES
1. Botterweck AA, Schouten LJ, Volovics A, Dorant E, van Den Brandt PA. Trends in incidence of
adenocarcinoma of the oesophagus and gastric cardia in ten European countries. Int J Epidemiol
2000;29:645-654.
2. Vizcaino AP, Moreno V, Lambert R, Parkin DM. Time trends incidence of both major histologic types of
esophageal carcinomas in selected countries, 1973-1995. Int J Cancer 2002;99:860-868.
3. Gammon MD, Schoenberg JB, Ahsan H, Risch HA, Vaughan TL, Chow WH, Rotterdam H, West AB,
Dubrow R, Stanford JL, Mayne ST, Farrow DC, Niwa S, Blot WJ, Fraumeni JF, Jr. Tobacco, alcohol, and
socioeconomic status and adenocarcinomas of the esophagus and gastric cardia. J Natl Cancer Inst
1997;89:1277-1284.
4. Lagergren J, Bergstrom R, Lindgren A, Nyren O. The role of tobacco, snuff and alcohol use in the
aetiology of cancer of the oesophagus and gastric cardia. Int J Cancer 2000;85:340-346.
5. Pandeya N, Williams GM, Sadhegi S, Green AC, Webb PM, Whiteman DC. Associations of duration,
intensity, and quantity of smoking with adenocarcinoma and squamous cell carcinoma of the
esophagus. Am J Epidemiol 2008;168:105-114.
6. Wu AH, Wan P, Bernstein L. A multiethnic population-based study of smoking, alcohol and body size
and risk of adenocarcinomas of the stomach and esophagus (United States). Cancer Causes Control
2001;12:721-732.
7. Rothman KJ, Greenland S. Modern epidemiology. Lippincott, 1998.
8. Pell JP, Haw SJ, Cobbe SM, Newby DE, Pell AC, Oldroyd KG, Murdoch DL, Pringle SD, Dunn FG, Macintyre
PD, Gilbert TJ, Fischbacher CM, Borland W. Validity of self-reported smoking status: Comparison of
patients admitted to hospital with acute coronary syndrome and the general population. Nicotine Tob
Res 2008;10:861-866.
9. Lindblad M, Rodriguez LA, Lagergren J. Body mass, tobacco and alcohol and risk of esophageal, gastric
cardia, and gastric non-cardia adenocarcinoma among men and women in a nested case-control study.
Cancer Causes Control 2005;16:285-294.
10. Freedman ND, Abnet CC, Leitzmann MF, Mouw T, Subar AF, Hollenbeck AR, Schatzkin A. A prospective
study of tobacco, alcohol, and the risk of esophageal and gastric cancer subtypes. Am J Epidemiol
2007;165:1424-1433.
11. Zendehdel K, Nyren O, Luo J, Dickman PW, Boffetta P, Englund A, Ye W. Risk of gastroesophageal cancer
among smokers and users of Scandinavian moist snuff. Int J Cancer 2008;122:1095-1099.
12. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
13. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
14. Van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol 1990;19:
553-558.
15. Goldbohm RA, van den Brandt PA, Dorant E. Estimation of the coverage of Dutch municipalities by
cancer registries and PALGA based on hospital discharge data. Tijdschr Soc Gezondheidsz 1994;72:
80-84.
16. Parkin DM, Shanmugaratnam K, Sobin L, Ferlay J, Whelan SL. Histological Groups for comparative
studies. IARC Technical reports. Volume 31. Lyon: International Agency for Research on Cancer, 1998.
17. Goldbohm RA, van den Brandt PA, Brants HA, van't Veer P, Al M, Sturmans F, Hermus RJ. Validation of a
dietary questionnaire used in a large-scale prospective cohort study on diet and cancer. Eur J Clin Nutr
1994;48:253-265.
18. Nevo table: Dutch food composition table, 1986-1987. (Dutch). Voorlichtingbureau Voor de Voeding,
1986.
19. Goldbohm RA, van 't Veer P, van den Brandt PA, van 't Hof MA, Brants HA, Sturmans F, Hermus RJ.
Reproducibility of a food frequency questionnaire and stability of dietary habits determined from five
annually repeated measurements. Eur J Clin Nutr 1995;49:420-429.
20. World Cancer Research Fund, American Institute for Cancer Research. Food, nutrition, physical activity
and the prevention of cancer: a global perspective. AICR, 2007.
Thesis Jessie Steevens_v04.pdf
Chapter 6
128
21. Schottenfeld D, Fraumeni JF, Jr. Cancer epidemiology and prevention. Oxford University Press, 2006.
22. Zeegers MP, Schouten LJ, Goldbohm RA, van den Brandt PA. A compendium of familial relative risks of
cancer among first degree relatives: a population-based study. Int J Cancer 2008;123:1664-1673.
23. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:
1503-1511.
24. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
25. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atcddd/
welcome.html (accessed 15 Oct, 2009).
26. Cox DR. Regression models and life-tables (with discussion). J R Stat Soc Ser B 1972;34:187-220.
27. Barlow WE. Robust variance estimation for the case-cohort design. Biometrics 1994;50:1064-1072.
28. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrica 1982;69:
239-241.
29. Helicobacter and Cancer Collaborative Group. Gastric cancer and Helicobacter pylori: a combined
analysis of 12 case control studies nested within prospective cohorts. Gut 2001;49:347-353.
30. Islami F, Kamangar F. Helicobacter pylori and esophageal cancer risk: a meta-analysis. Cancer Prev Res
(Phila Pa) 2008;1:329-338.
31. Rokkas T, Pistiolas D, Sechopoulos P, Robotis I, Margantinis G. Relationship between Helicobacter pylori
infection and esophageal neoplasia: a meta-analysis. Clin Gastroenterol Hepatol 2007;5:1413-1417,
1417 e1411-1412.
32. Taylor DN, Blaser MJ. The epidemiology of Helicobacter pylori infection. Epidemiol Rev 1991;13:42-59.
33. Tsugane S, Tei Y, Takahashi T, Watanabe S, Sugano K. Salty food intake and risk of Helicobacter pylori
infection. Jpn J Cancer Res 1994;85:474-478.
34. Siman JH, Forsgren A, Berglund G, Floren CH. Tobacco smoking increases the risk for gastric
adenocarcinoma among Helicobacter pylori-infected individuals. Scand J Gastroenterol 2001;36:
208-213.
35. Brenner H, Rothenbacher D, Bode G, Adler G. Relation of smoking and alcohol and coffee consumption
to active Helicobacter pylori infection: cross sectional study. Bmj 1997;315:1489-1492.
36. Ishiguro S, Sasazuki S, Inoue M, Kurahashi N, Iwasaki M, Tsugane S. Effect of alcohol consumption,
cigarette smoking and flushing response on esophageal cancer risk: a population-based cohort study
(JPHC study). Cancer Lett 2009;275:240-246.
37. Castellsague X, Munoz N, De Stefani E, Victora CG, Castelletto R, Rolon PA, Quintana MJ. Independent
and joint effects of tobacco smoking and alcohol drinking on the risk of esophageal cancer in men and
women. Int J Cancer 1999;82:657-664.
38. Zambon P, Talamini R, La Vecchia C, Dal Maso L, Negri E, Tognazzo S, Simonato L, Franceschi S. Smoking,
type of alcoholic beverage and squamous-cell oesophageal cancer in northern Italy. Int J Cancer
2000;86:144-149.
39. Znaor A, Brennan P, Gajalakshmi V, Mathew A, Shanta V, Varghese C, Boffetta P. Independent and
combined effects of tobacco smoking, chewing and alcohol drinking on the risk of oral, pharyngeal and
esophageal cancers in Indian men. Int J Cancer 2003;105:681-686.
40. Neuhouser ML. Dietary flavonoids and cancer risk: evidence from human population studies. Nutr
Cancer 2004;50:1-7.
41. Pandeya N, Williams G, Green AC, Webb PM, Whiteman DC. Alcohol Consumption and the Risks of
Adenocarcinoma and Squamous Cell Carcinoma of the Esophagus. Gastroenterology 2009;136:
1215-1224.
42. Liu C, Russell RM. Nutrition and gastric cancer risk: an update. Nutr Rev 2008;66:237-249.
43. IARC. Monographs on the evaluation of carcinogenic risks to humans: Tobacco smoke and involuntary
smoking. Volume 83. Lyon: International Agency for Research on Cancer, 2004.
44. Ladeiras-Lopes R, Pereira AK, Nogueira A, Pinheiro-Torres T, Pinto I, Santos-Pereira R, Lunet N. Smoking
and gastric cancer: systematic review and meta-analysis of cohort studies. Cancer Causes Control
2008;19:689-701.
45. Pelucchi C, Gallus S, Garavello W, Bosetti C, La Vecchia C. Alcohol and tobacco use, and cancer risk for
upper aerodigestive tract and liver. Eur J Cancer Prev 2008;17:340-344.
Thesis Jessie Steevens_v04.pdf
Alcohol consumption, cigarette smoking and risk of subtypes of esophageal and gastric cancer
129
46. Bosetti C, Gallus S, Garavello W, La Vecchia C. Smoking cessation and the risk of oesophageal cancer: An
overview of published studies. Oral Oncol 2006;42:957-964.
47. Gallus S, Bosetti C, Franceschi S, Levi F, Simonato L, Negri E, La Vecchia C. Oesophageal cancer in
women: tobacco, alcohol, nutritional and hormonal factors. Br J Cancer 2001;85:341-345.
48. Lee CH, Wu DC, Lee JM, Wu IC, Goan YG, Kao EL, Huang HL, Chan TF, Chou SH, Chou YP, Lee CY, Chen
PS, Ho CK, He J, Wu MT. Carcinogenetic impact of alcohol intake on squamous cell carcinoma risk of the
oesophagus in relation to tobacco smoking. Eur J Cancer 2007;43:1188-1199.
49. IARC. Monographs on the evaluation of carcinogenic risks to humans: re-evaluation ofsome organic
chemicals, hydrazine and hydrogen peroxide. Volume 71. Lyon: International Agency for Research on
Cancer, 1999.
50. Poschl G, Seitz HK. Alcohol and cancer. Alcohol Alcohol 2004;39:155-165.
51. IARC. Monographs on the evaluation of carcinogenic risks to humans: alcohol drinking. Volume 44.
Lyon: International Agency for Research on Cancer, 1988.
Thesis Jessie Steevens_v04.pdf
Chapter 6
130
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of
esophageal and gastric cancer subtypes:
the Netherlands Cohort Study
Jessie Steevens
Piet A van den Brandt
R Alexandra Goldbohm
Leo J Schouten
Gastroenterology 2010;138:1704-1713
7
131
Thesis Jessie Steevens_v04.pdf
Chapter 7
132
ABSTRACT
Background & Aims
Selenium may protect against the development of esophageal squamous cell
carcinoma (ESCC), esophageal adenocarcinoma (EAC) and gastric cardia
adenocarcinoma (GCA). Only in very few studies have the associations with ESCC and
GCA been investigated, and no epidemiologic studies exist on EAC.
Methods
We studied the association between selenium and risk of ESCC, EAC, and GCA within
the prospective Netherlands Cohort Study, conducted among 120,852 men and
women aged 55-69 years at baseline. In September 1986, the cohort members
completed a questionnaire on risk factors for cancer and provided toenail clippings for
determination of baseline selenium status. After 16.3 years of follow-up, 64 ESCC, 112
EAC, and 114 GCA cases and 2072 subcohort members were available for case-cohort
analysis. Incidence rate ratios (RR) were calculated using Cox proportional hazards
models.
Results
In multivariable analyses of selenium status, we found an inverse association with
ESCC (RRper standard unit increment 0.80, 95% confidence interval (CI) 0.67-0.96), and a
borderline significant inverse association with GCA (RR 0.91, 95% CI 0.80-1.02). No
overall association was observed for EAC (RR 1.05, 95% CI 0.95-1.15), but for women
and never smokers, significant inverse associations were found (RRper standard unit increment
0.72, 95% CI 0.61-0.84 and 0.74, 95% CI 0.64-0.86, respectively).
Conclusions
This prospective study supports an inverse association between toenail selenium and
risk of ESCC and GCA and suggests an inverse association with risk of EAC in subgroups
(women, never smokers, and low antioxidant consumers). These associations need
confirmation.
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
133
INTRODUCTION
Selenium has been studied in relation to cancer risk, and most epidemiologic studies
have been performed on breast, colorectal, lung, gastric, bladder, and prostate
cancer.1 Evidence from observational studies so far suggests that selenium is inversely
associated with prostate cancer, and possibly with lung and gastric cancer.1 The reason
for the interest in selenium in relation to cancer etiology is the antioxidant capacity of
the selenium-dependent glutathione peroxidase enzymes.2,3 Other possible
mechanisms through which selenium could be associated with lower cancer risk
include reduction of inflammation, induction of detoxifying phase II enzymes, increase
of p53, alteration of DNA methylation, blockage of the cell cycle, induction of
apoptosis of cancer cells and inhibition of angiogenesis.3
The mineral is an essential trace element present in food. The selenium content of
foods depends on the selenium content of the soil on which the food is grown. Regions
with very low soil selenium levels (<0.05 ppm) are New Zealand, parts of China, and
Finland, while high levels (>5 ppm) are found in Canada, parts of the western United
States, parts of China, Ireland, France, and Germany.4 The recommended daily
allowance for selenium is 55 µg/day in the United States5, and 50-150 µg/day in The
Netherlands.6 For the Dutch population, the most important sources of selenium are
meat, bread, milk, and milk product, fish, eggs, and cheese.7 Generally high serum
selenium levels are reported in the United States, whereas in The Netherlands, blood
and toenail selenium levels are intermediate between those reported from New
Zealand and the United States.8 In epidemiologic research, the measurement of
selenium status through biomarkers is preferred over the measurement of selenium
intake through questionnaires because of the high variability of food selenium
content.9 Toenails are a suitable biomarker because they reflect the intake of selenium
for a period up to 1 year.10,11
Data on possible associations between selenium and esophageal and gastric
cancer are sparse, and even less is known about the subtypes of these cancers. A few
epidemiologic studies12-16 of different design found indications for an inverse
association between selenium and total esophageal cancer. Limited evidence
suggestive of an inverse association with esophageal squamous cell carcinoma (ESCC)
arises from studies in Chinese populations.17-20 We could not find any case-control or
cohort studies on selenium and esophageal adenocarcinoma (EAC) risk, but 1 US cross-
sectional study found an inverse association between serum selenium and markers of
neoplastic progression in Barrett’s esophagus cases (a precursor lesion of EAC).21
Selenium and its association with gastric cancer risk has been described in a review,22
but most studies, including ours,23 did not study the cardia specifically. In our case, this
was due to the limited number of cases after 3.3 years of follow-up of our cohort. A
cohort study17 and trial24 in Linxian, China, studying gastric cardia adenocarcinoma
(GCA) specifically, were able to identify an inverse association.
Thesis Jessie Steevens_v04.pdf
Chapter 7
134
It is important to study subtypes of esophageal and gastric cancer separately in
analysis because these probably differ with regard to their etiology.25,26 EAC and GCA
are specifically of interest because the incidences of these cancers have risen strongly
in the United States and Europe during the past decades.27,28
The purpose of this study was to prospectively investigate the association
between prediagnostic toenail selenium levels and risk of ESCC, EAC, and GCA. For this
investigation, we used 16.3-year follow-up data from The Netherlands Cohort Study on
diet and cancer (NLCS).
Our hypothesis was that toenail selenium levels are inversely associated with risk
of ESCC, EAC, and GCA. The association with ESCC was expected to be the strongest
because ESCC is strongly associated with smoking, which causes oxidative stress. This
oxidative stress may be counteracted by the antioxidant capacity of selenium.
We studied the potential interaction between selenium and cigarette smoking,
intake of vitamins C, E, and several carotenoids. These factors were chosen because
they may all be involved in an antioxidant mechanism,3,29 and evidence suggestive of
these interactions was previously found for total and prostate cancer.30,31 Additionally,
potential interaction with body mass index (BMI) was investigated because of its
association with esophageal cancer.
PATIENTS AND METHODS
Study design and participants
In September 1986, the prospective NLCS was initiated when 58,279 Dutch men and
62,573 women aged 55-69 years were enrolled. The subjects were randomly selected
from 204 Dutch municipal registries. At baseline, all cohort members completed a self-
administered questionnaire and were asked to provide toenail clippings. A detailed
description of the study was previously published.32
The case-cohort approach is used for data processing and analysis, for reasons of
efficiency.33 Cases are derived from the entire cohort. A subcohort of 3500 subjects
was selected at random from the total cohort at baseline. These 3500 subjects were
used for the estimation of person-years at risk in the total cohort. Person-years at risk
were calculated from the start of the study until esophageal or gastric cancer
diagnosis, death, emigration, loss to follow-up or end of follow-up (December 31,
2002), whichever occurred first. Follow-up of the subcohort was established by linkage
to Dutch municipal population registries and after 16.3 years (September 1986 to
December 2002), only 1 male subcohort member was lost to follow-up. We excluded
subcohort members (n=154) who reported at baseline to have prevalent cancer (other
than skin cancer), leaving 3346 subcohort members.
The entire cohort was linked to The Netherlands Cancer Registry and the
nationwide network and registry of histopathology and cytopathology in The
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
135
Netherlands34 for cancer incidence.35 Follow-up for cancer incidence was ≥96%
complete,36 and after 16.3 years, the following numbers of incident, microscopically
confirmed, primary carcinomas were identified: 130 ESCC37 [International Classification
of Diseases (ICD) for Oncology (O)-3, C15], 181 EAC37 (C15), and 206 GCA (C16.0). The
Medical Ethics Committee of Maastricht University, The Netherlands, has approved the
study.
Exposure data
At baseline, all cohort members completed a self-administered questionnaire. The
questionnaire consisted of a 150-item food frequency questionnaire and other
questions on cancer risk factors, eg, smoking habits, alcohol consumption, height, and
weight. The food frequency questionnaire asked about habitual consumption in the
year preceding the start of the study. Mean daily nutrient intakes were calculated
using the Dutch food-composition table.38 The questionnaire data were key entered
and processed in a standardized manner, blinded with respect to cases/subcohort
status to minimize observer bias in coding and interpretation of the data.
In our cohort, about 75% of the subjects provided toenail clippings. Toenail
selenium determinations were carried out by the Reactor Institute Delft (Delft
University of Technology, Delft, The Netherlands). The determination was based on
instrumental neutron activation analysis of the 77mSe isotope (metastable selenium-77
isotope, half-life 17.5 seconds). Each sample went through 6 cycles of 17-second
irradiation at a thermal neutron flux of 3*1016 m-2s-1, 3-second decay, and 17-second
counting at 1 cm from a 40% germanium detector. This method and the use in the
NLCS have been described in more detail previously.23,39-41
In 1992, the toenail selenium determinations for the subcohort members and
gastric cancer cases diagnosed until 3.3 years of follow-up (until December 31, 1990)
were carried out. In 2008, the selenium determinations were carried out for all
esophageal cancer cases and the gastric cardia cancer cases occurring from 3.3-16.3
years of follow-up (December 31, 1990, until December 31, 2002). In 1992, the “SBP”
(abbreviation of Snelle Buizen Post, Dutch for fast pneumatic system) facility was used
for instrumental neutron activation analysis, and, since 1996, the “CAFIA” (Carbonfiber
Autonomous Facility for Irradiation and Analysis) facility has been used. To assess the
validity and comparability of these 2 methods, the toenail selenium levels for 40
subcohort members were assessed in 1996 with the “CAFIA” facility in addition to the
original assessment with the “SBP” facility.39 The mean selenium level ± standard
deviation (SD) assessed by the “CAFIA” facility (0.552±0.05 µg/g) was comparable with
mean selenium levels assessed by the “SBP” facility (0.551±0.04 µg/g) for these
subjects. The Pearson correlation coefficient between toenail selenium levels assessed
by the “CAFIA” facility and those estimated by the “SBP” facility was 0.95 (p<0.01).39 It
was concluded that both methods were valid and comparable.
Thesis Jessie Steevens_v04.pdf
Chapter 7
136
In 1992, all toenail clippings provided by cases and subcohort were sent to the
Reactor Institute Delft for selenium determination. This determination, however,
yields unreliable results if the nail samples weigh <10 mg, and these measurements
were thus excluded. In 2007, we discovered that toenail clippings can also be used as a
source of DNA42. We therefore separated and saved 10-20 mg of toenail clippings from
the cancer cases for future genetic research. If, afterwards, more than 10 mg of toenail
clippings was left, we sent them to the Reactor Institute Delft. In case there were
problems with the determination of toenail selenium due to very high calcium
contents, the subject was excluded. Complete selenium data were available for 2426
subcohort members, and 71 ESCC, 129 EAC, and 127 GCA cases.
Statistical analysis
To evaluate the potential influence of prediagnostic cancer at baseline on toenail
selenium levels, cases were categorized according to the year of follow-up in which
they were diagnosed. Mean toenail selenium levels of ESCC, EAC and GCA cancer cases
were compared according to year of follow-up, and differences were tested using a
t-test. For the t-test, selenium levels were ln (natural logarithm)-transformed to
normalize the distribution. For the case-cohort analyses, toenail selenium levels were
categorized into quartiles according to the distribution in the subcohort. For
continuous analyses, the selenium levels were standardized to the average size of the
2 central quartiles in the subcohort. One standardized selenium unit is equal to
0.06 µg/g.
Excluded were subcohort members and cases with inconsistent or incomplete
dietary questionnaire data43 and those with missing data on the confounders.
Complete data for statistical analysis was available for 2072 subcohort members, and
64 ESCC, 112 EAC, and 114 GCA cases.
The following variables were considered confounders because of their established
association with esophageal or gastric cancer:44 age, sex, cigarette smoking (current
yes/no, number of cigarettes smoked daily, and number of smoking years), alcohol
consumption (g/day), and BMI (kg/m2). The following variables were considered
potential confounders but were eventually not included in the models because they
did not change the incidence rate ratio (RR) by >5%: nonoccupational physical activity;
highest level of education; daily intakes of vitamin C, vitamin E, α-carotene,
β-carotene, β-cryptoxanthin, lycopene, and lutein/zeaxanthin; family history of
esophageal or gastric cancer; reported long-term (>0.5 years) use of nonsteroidal anti-
inflammatory drugs or aspirin (Anatomical therapeutic chemical (ATC) codes M01A,
N02B); or lower esophageal sphincter relaxing medication.45-47
Cox proportional hazards models were used to estimate age- and sex-adjusted
and multivariable adjusted incidence RRs and corresponding 95% confidence intervals
(CI).48 Analyses were done using Stata 9.2 statistical software package (StataCorp,
College Station, Texas, USA). Standard errors were estimated using the robust Huber-
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
137
White sandwich estimator to account for additional variance introduced by sampling
from the cohort. This method is equivalent to the variance-covariance estimator by
Barlow.49 The proportional hazards assumption was tested using the scaled Schoenfeld
residuals,50 and -log(-log) plots. Tests for dose-response trends were assessed by fitting
ordinal exposure variables as continuous terms. Two-sided p-values are reported
throughout the article.
Because of the limited numbers of cases, all analyses were carried out for both
sexes combined. Nevertheless, tests for interaction between toenail selenium and sex
were performed using the continuous selenium variable. We only report results
stratified by sex for the continuous selenium variable.
We investigated possible interactions between toenail selenium status and
cigarette smoking status (current, former, never), BMI, and intakes of several vitamins
and carotenoids by estimating RRs in strata of these exposures. The p-value for
interaction was assessed by including a cross-product term in the model. In the
interaction analyses, we used a continuous selenium variable, because of the limited
numbers of cases. To evaluate potential bias introduced by prediagnostic cancer at
baseline, which may influence toenail selenium levels, we performed sensitivity
analyses, excluding cases from the first 2 years of follow-up.
RESULTS
Table 7.1 shows that, when the cases were categorized according to 2-year follow-up
periods in which they were diagnosed, there was no clear trend towards lower toenail
selenium levels in cases occurring during early follow-up, indicating no effect of
preclinical disease on toenail selenium levels. The t-tests were also not statistically
significant.
The median toenail selenium level (in micrograms/grams) of all subcohort
members was 0.552. The levels were lower in men (0.538) than in women (0.564). For
the cases, the following median toenail selenium levels (in µg/g) were observed: ESCC,
0.493; EAC, 0.532; and GCA, 0.529 (Table 7.2).
Table 7.2 presents some characteristics of the subcohort members and ESCC, EAC,
and GCA cases. The most salient differences are described hereafter. There were far
more men than women among the EAC and GCA cases. Among the subcohort
members, there were relatively more never cigarette smokers, and subcohort
members had, on average, higher intakes of vitamins C and E, and carotenoids than
the cases. With respect to separate cancer groups, we found that ESCC cases
consumed much more alcohol and more often had a low level of education compared
with the subcohort. EAC cases had a higher BMI, and more often reported a family
history of esophageal or gastric cancer and use of lower esophageal sphincter relaxing
medication. GCA cases spent more time on nonoccupational physical activity than the
subcohort.
Thesis Jessie Steevens_v04.pdf
Chapter 7
138
Table 7.1 Toenail selenium levels (µg/g) in esophageal and gastric cancer cases according to sex and time
between baseline and esophageal or gastric cancer diagnosis; Netherlands Cohort Study (1986-
2002).
Esophageal squamous
cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Cases a Toenail selenium level
(µg/g)
Toenail selenium level
(µg/g)
Toenail selenium level
(µg/g)
No.
cases
mean sd p
b
No.
cases
mean sd p b
No.
cases
mean sd p b
All cases 71 0.514 0.088 129 0.564 0.163 127 0.535 0.080
Men 44 0.505 0.092 106 0.571 0.177 109 0.536 0.082
Women 27 0.527 0.081 23 0.533 0.049 18 0.532 0.071
Year of follow-up
0-2 4 0.462 0.103 0.20 8 0.514 0.065 0.30 12 0.537 0.085 0.96
>2-4 9 0.568 0.144 12 0.549 0.070 17 0.573 0.087
>4-6 8 0.446 0.051 15 0.617 0.296 11 0.534 0.093
>6-8 11 0.509 0.072 20 0.590 0.162 17 0.534 0.082
>8-10 9 0.535 0.090 14 0.537 0.081 18 0.509 0.069
>10-12 11 0.531 0.077 20 0.612 0.244 24 0.550 0.077
>12-14 7 0.523 0.058 19 0.515 0.063 12 0.538 0.083
>14-17 12 0.503 0.066 21 0.547 0.064 16 0.502 0.061
a Mean ± sd selenium levels in subcohort members were 0.547 ± 0.126 µg/g for men (n=1211) and 0.575 ±
0.109 µg/g for women (n=1248); b T-test of mean toenail selenium level (ln-transformed) in first 2 years of
follow-up versus rest of follow-up years.
Table 7.3 reports on the association between quartiles of toenail selenium levels
and risk of ESCC, EAC, and GCA. For ESCC, an inverse association was found with
toenail selenium (multivariable RR for the highest vs the lowest quartile 0.37, 95% CI
0.16-0.86), and the test for trend was statistically significant (p-trend=0.02). The
association was nearly the same in age- and sex-adjusted analyses, and the results
were also robust to the exclusion of the first 2 years of follow-up.
No clear association was found between toenail selenium levels and EAC risk
(multivariable RR for the highest vs the lowest quartile 0.76, 95% CI 0.41-1.40). The
results from age- and sex-adjusted analyses and analyses excluding the first 2 years of
follow-up showed no association either. However, a statistically significant (p<0.001)
interaction was found between toenail selenium and sex: an inverse association was
observed in women, but not in men (Table 7.3). As mentioned before, we analyzed
men and women together in further analyses because of the small number of cases.
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
139
Table 7.2 Characteristics of cases and subcohort members in the Netherlands Cohort Study (1986-2002).
Subcohort Cases
Exposure variables and potential confounders
(n=2072) a
ESCC b
(n=64) a
EAC b
(n=112) a
GCA b
(n=114) a
median median median median
(IQR) (IQR) (IQR) (IQR)
Toenail selenium level (µg/g)
total 0.552 0.493 0.532 0.529
(0.498-0.613) (0.453-0.583) (0.488-0.584) (0.477-0.582)
men 0.538 0.490 0.536 0.529
(0.483-0.602) (0.448-0.558) (0.485-0.588) (0.477-0.581)
women 0.564 0.524 0.516 0.518
(0.514-0.623) (0.473-0.597) (0.501-0.564) (0.485-0.583)
mean (sd) c mean (sd) mean (sd) mean (sd)
Age at baseline (y) 61.2 (4.2) 62.2 (4.0) 61.5 (4.1) 61.3 (4.2)
Sex
Men (n (%)) 1020 (49) 40 (63) 93 (83) 97 (85)
Women (n (%)) 1052 (51) 24 (37) 19 (17) 17 (15)
Cigarette smoking status
Never smoker (n (%)) 782 (38) 11 (17) 19 (17) 15 (13)
Former smoker (n (%)) 760 (37) 21 (33) 61 (54) 57 (50)
Current smoker (n (%)) 530 (25) 32 (50) 32 (29) 42 (37)
Ever cigarette smokers
Frequency of cigarette smoking (no./d) 15 (10) 17 (10) 21 (14) 17 (11)
Duration of cigarette smoking (y) 31 (12) 37 (10) 34 (11) 35 (11)
Ethanol intake (g/d) 10 (14) 24 (29) 15 (18) 13 (14)
Body mass index (BMI) (kg/m2) 25.0 (3.1) 23.9 (3.0) 26.4 (3.3) 25.5 (2.6)
Nonoccupational physical activity (min/d) 73 (58) 74 (62) 77 (58) 89 (82)
Highest level of education d
Primary (n (%)) 565 (27) 23 (36) 25 (23) 26 (23)
Lower vocational (n (%)) 457 (22) 12 (19) 26 (24) 34 (30)
Secondary and medium vocational (n (%)) 754 (37) 21 (33) 40 (36) 32 (28)
University and higher vocational (n (%)) 283 (14) 8 (12) 19 (17) 22 (19)
Vitamin C intake (mg/d) 104 (42) 100 (47) 98 (50) 90 (40)
Vitamin E intake (mg/d) 14 (6) 13 (5) 14 (6) 14 (7)
α-carotene intake (µg/d) 711 (587) 680 (513) 701 (539) 659 (555)
β-carotene intake (µg/d) 3001 (1600) 2869 (1359) 2970 (1525) 2924 (1597)
β-cryptoxanthin intake (µg/d) 181 (165) 174 (194) 150 (174) 131 (146)
Lycopene intake (µg/d) 1210 (1762) 1056 (1297) 985 (1722) 1067 (1118)
Lutein/zeaxanthin intake (µg/d) 2520 (1096) 2411 (925) 2567 (1140) 2555 (1317)
Family history of esophageal or gastric cancer
(n (%))
173 (8) 7 (11) 15 (13) 10 (9)
Reported long-term use of NSAIDsb (n (%)) 152 (7) 5 (8) 10 (9) 5 (4)
Reported long-term use of LESb relaxing
medication (n (%))
288 (14) 9 (14) 27 (24) 18 (16)
a Presented are the number of subcohort members or cases with complete data on toenail selenium level,
age, sex, cigarette smoking (current yes/no, number of cigarettes smoked daily, number of smoking years),
alcohol consumption and BMI. Subcohort members and cases with incomplete or inconsistent questionnaire
data are excluded; b ESCC: esophageal squamous cell carcinoma, EAC: esophageal adenocarcinoma, GCA:
gastric cardia adenocarcinoma, LES: lower esophageal sphincter, NSAIDs: nonsteroidal anti-inflammatory
drugs; c For categorical variables n (%) is presented; d Numbers do not always add up to the total, because of
missing values on this variable.
Thesis Jessie Steevens_v04.pdf
Chapter 7
140
Thesis Jessie Steevens_v04.pdf
Table 7.3 Incidence rate ratios of esophageal and gastric cancer subtypes according to toenail selenium levels; Netherlands Cohort Study (1986-2002).
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
person time
at risk
(years)
no.
cases
RR a(95% CI) a no.
cases
RR a(95% CI) a no.
cases
RR a(95% CI) a
All years of follow-up
Age- and sex-adjusted analyses
Quartiles of toenail selenium (boundaries in µg/g)
1 ( ≤ 0.498) 7207 33 1 reference 33 1 reference 41 1 reference
2 ( ≤ 0.552) 7614 10 0.30 0.14- 0.61 35 1.15 0.70-1.89 27 0.73 0.44-1.21
3 ( ≤ 0.613) 7705 12 0.37 0.19- 0.73 24 0.85 0.49-1.48 31 0.90 0.55-1.49
4 (> 0.613) 7584 9 0.28 0.13- 0.59 20 0.74 0.41-1.32 15 0.46 0.25-0.85
p-value for linear trend b 0.001 0.20 0.04
increment in standard units c 30110 64 0.72 0.61- 0.85 112 1.03 0.94-1.12 114 0.88 0.79-0.98
men 14207 40 0.74 0.58- 0.93 93 1.05 0.97-1.12 97 0.91 0.81-1.02
women 15903 24 0.71 0.57- 0.89 19 0.74 0.64-0.85 17 0.73 0.57-0.94
p-interaction
d 0.81 p-interaction
d <0.001 p-interaction
d 0.12
Multivariable adjusted analyses e
Quartiles of toenail selenium (boundaries in µg/g)
1 ( ≤ 0.498) 7207 33 1 Reference 33 1 reference 41 1 reference
2 ( ≤ 0.552) 7614 10 0.36 0.17-0.74 35 1.13 0.67-1.91 27 0.75 0.44-1.27
3 ( ≤ 0.613) 7705 12 0.45 0.21-0.98 24 0.84 0.48-1.49 31 0.97 0.57-1.65
4 ( > 0.613) 7584 9 0.37 0.16-0.86 20 0.76 0.41-1.40 15 0.52 0.27-1.02
p-value for linear trend b 0.02 0.25 0.14
increment in standard units c 30110 64 0.80 0.67-0.96 112 1.05 0.95-1.15 114 0.91 0.80-1.02
men 14207 40 0.81 0.64-1.04 93 1.07 0.99-1.15 97 0.94 0.84-1.06
women 15903 24 0.79 0.63-0.99 19 0.72 0.61-0.84 17 0.73 0.56-0.95
p-interaction
d 0.84 p-interaction
d <0.001 p-interaction
d 0.07
Selenium status and the risk of esophageal and gastric cancer subtypes
141
Thesis Jessie Steevens_v04.pdf
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
person time
at risk
(years)
no.
cases
RR a(95% CI) a no.
cases
RR a(95% CI) a no.
cases
RR * (95% CI) a
First 2 years of follow-up excluded
Multivariable adjusted analyses e
Quartiles of toenail selenium (boundaries in µg/g)
1 ( ≤ 0.498) 6190 30 1 reference 30 1 reference 38 1 reference
2 ( ≤ 0.552) 6579 10 0.40 0.19-0.82 32 1.13 0.66-1.94 24 0.72 0.42-1.25 f
3 ( ≤ 0.613) 6666 11 0.46 0.20-1.01 23 0.88 0.49-1.58 30 1.01 0.59-1.74 f
4 ( > 0.613) 6560 9 0.42 0.18-0.96 f 19 0.79 0.42-1.49 14 0.53 0.27-1.07 f
p-value for linear trend b 0.04 0.35 0.21
increment in standard units c 25996 60 0.82 0.68-0.98 104 1.06 0.97-1.15 106 0.92 0.81-1.04 f
men 12188 38 0.82 0.64-1.05 86 1.08 1.00-1.15 90 0.95 0.85-1.08
women 13808 27 0.82 0.66-1.02 18 0.74 0.63-0.87 16 0.73 0.55-0.97
p-interaction
d 0.96 p-interaction
d <0.001 p-interaction
d 0.08
a RR, Incidence Rate Ratio; 95% CI, 95 percent confidence interval; b Tests for dose-response trends were assessed by fitting ordinal variables as continuous terms in the
Cox proportional hazard model; c For continuous analyses, the selenium levels were standardized to the average size of the two central quartiles. One standardized
selenium unit is equal to 0.06 µg/g; d p-value for interaction between sex and toenail selenium level, based on cross product term in the Cox proportional hazard model.
e adjusted for age (years), sex, cigarette smoking (current smoking status (yes/no), frequency (number of cigarettes/day), and duration (years)), alcohol consumption
(g/day), body mass index (kg/m²); f proportional hazards assumption was violated for selenium.
Chapter 7
142
Thesis Jessie Steevens_v04.pdf
Table 7.4 Association a between toenail selenium b levels and risk of esophageal and gastric cancer subtypes stratified by cigarette smoking status, body mass index and
intake levels of antioxidants; Netherlands Cohort Study (1986-2002)
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
person time at
risk (years)
No. cases RRc (95% CI) c No. cases RRc (95% CI) c No. cases RRc (95% CI) c
Cigarette smoking status
Never smoker 11745 11 0.73 (0.53-0.99) 19 0.74 (0.64-0.86) 15 0.69 (0.52-0.90)
Former smoker 10906 21 0.72 (0.53-0.97) 61 1.07 (0.99-1.15) 57 0.88 (0.73-1.06)
Current smoker 7459 32 0.92 (0.71-1.19) 32 1.18 (0.84-1.65) 42 1.10 (0.91-1.33)
p-interaction 0.50 p-interaction <0.001 p-interaction 0.01
Body mass index (kg/m²)
<25 16181 42 0.89 (0.74-1.07) 40 1.03 (0.91-1.17) 49 0.94 (0.81-1.09)
≥25 13929 22 0.59 (0.43-0.82) 72 1.10 (0.81-1.49)
d 65 0.89 (0.74-1.06)
p-interaction 0.16 p-interaction 0.64 p-interaction 0.81
Antioxidant intake
Vitamin C intake (mg/d)
low (≤97) 14755 38 0.77 (0.58-1.02) 67 0.86 (0.73-1.01) 71 0.85 (0.72-1.12)
high (>97) 15355 26 0.81 (0.64-1.01) 45 1.21 (1.04-1.41) 43 0.97 (0.81-1.15)
p-interaction 0.37 p-interaction <0.001 p-interaction 0.12
Vitamin E intake (mg/d)
low (≤12) 14992 29 0.84 (0.62-1.12) 54 1.24 (1.01-1.53) 49 0.94 (0.78-1.13)
high (>12) 15117 35 0.77 (0.61-0.96) 58 0.90 (0.77-1.06) 65 0.88 (0.75-1.04)
p-interaction 0.50 p-interaction 0.02 p-interaction 0.94
α-carotene intake (µg/d)
low (≤573) 14779 33 0.78 (0.61-1.00) 59 0.83 (0.70-0.99) 58 0.88 (0.71-1.09)
high (>573) 15330 31 0.76 (0.58-0.99) 53 1.08 (1.01-1.16) 56 0.93 (0.81-1.07)
p-interaction 0.87 p-interaction 0.01 p-interaction 0.38
β-carotene intake (µg/d)
low (≤2680) 14903 36 0.76 (0.58-1.01) 58 0.83 (0.69-0.99) 63 0.87 (0.73-1.05)
high (>2680) 15207 28 0.82 (0.65-1.02) 54 1.24 (1.07-1.43) 51 0.94 (0.79-1.11)
p-interaction 0.91 p-interaction 0.004 p-interaction 0.53
β-cryptoxanthin intake (µg/d)
low (≤130) 14705 35 0.71 (0.57-0.89) 66 1.00 (0.76-1.31) 71 0.93 (0.81-1.08)
high (>130) 15404 29 0.89 (0.69-1.14) 46 1.05 (0.97-1.14) 43 0.86 (0.69-1.07)
p-interaction 0.05 p-interaction 0.46 p-interaction 0.63
Selenium status and the risk of esophageal and gastric cancer subtypes
143
Thesis Jessie Steevens_v04.pdf
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
person time at
risk (years)
No. cases RRc (95% CI) c No. cases RRc (95% CI) c No. cases RRc (95% CI) c
Lycopene intake (µg/d)
low (≤823) 14943 33 0.91 (0.74-1.12) 71 0.82 (0.70-0.96) 58 0.83 (0.69-0.99)
high (>823) 15167 31 0.64 (0.48-0.87) 41 1.25 (1.09-1.44) 56 0.98 (0.84-1.14)
p-interaction 0.21 p-interaction <0.001 p-interaction 0.21
Lutein/Zeaxanthin intake (µg/d)
low (≤2360) 14956 35 0.86 (0.66-1.11) 60 0.94 (0.68-1.29) 63 0.93 (0.78-1.11)
high (>2360) 15154 29 0.72 (0.57-0.90) 52 1.16 (0.99-1.36) 51 0.88 (0.74-1.05)
p-interaction 0.33 p-interaction 0.20 p-interaction 0.86
a adjusted for age (years), sex, cigarette smoking (current smoking status (yes/no), frequency (number of cigarettes/day), and duration (years)), alcohol consumption
(g/day), body mass index (kg/m²); b for continuous analyses, the selenium levels were standardized to the average size of the two central quartiles. One standardized
selenium unit is equal to 0.06 µg/g; c RR, Incidence Rate Ratio; 95% CI, 95 percent confidence interval; d proportional hazards assumption was violated for selenium.
Chapter 7
144
We found an inverse association between toenail selenium levels and risk of GCA
(multivariable RR for the highest vs the lowest quartile 0.52, 95% CI 0.27-1.02),
although the test for trend was not statistically significant (p=0.14). As for EAC, the RRs
in women were lower (and statistically significant) than in men, and the interaction
between selenium and sex was borderline significant (p=0.07). The age- and sex-
adjusted results were nearly the same as the multivariable results, but statistical
significance was only reached in the former. Exclusion of the first 2 years of follow-up
did not alter the results (Table 7.3). Because the t-tests (Table 7.1) and the exclusion of
the first 2 years of follow-up indicated that selenium levels were not affected by
prediagnostic disease, all following analyses are based on the complete 16.3-year
follow-up data.
Presented in Table 7.4 are the results of the interaction analyses. For EAC and
GCA, a statistically significant interaction was found between selenium and smoking
status: only for never smokers was an inverse association with selenium present. The
RRs for 1 standardized selenium unit increment were 0.74 (95% CI 0.64-0.86) for EAC
and 0.69 (95% CI 0.52-0.90) for GCA for never smokers. No interaction was found for
ESCC (p-interaction=0.50), although an inverse association was more apparent for
never and former smokers.
With regard to BMI, no interaction with selenium was observed for ESCC, EAC, or
GCA (Table 7.4). From Table 7.4, which presents results stratified by intake of several
antioxidants, it can be seen that the inverse association we observed between toenail
selenium levels and risk of ESCC did not differ between strata, defined by low or high
intake of these antioxidants. However, for EAC we found several statistically significant
interactions between selenium and antioxidant intake. For subjects with a low intake
of vitamin C, α-carotene, β-carotene, or lycopene, (borderline) significant inverse
associations with selenium were present. For GCA, no appreciable differences in
associations were found between strata of antioxidant intake.
DISCUSSION
This study is the first prospective cohort study into selenium and risk of subtypes of
esophageal and gastric cancer in a Western population, and the first to separately
investigate ESCC, EAC, and GCA. Our study provides evidence for an inverse association
between toenail selenium levels and risk of ESCC and GCA. The association with GCA
was more apparent in women, and only in women was an inverse association found
with EAC risk, but this was based on small numbers. We found these results in a Dutch
population with a median toenail selenium level of 0.522 µg/g (subcohort). This is a
low to moderate toenail selenium level when compared to mean levels found in other
populations (general population or control subjects): China (Sichuan province),
0.211 µg/g;51 Finland, 0.47 µg/g;52 United States, 0.83-0.92 µg/g;53 Colombia,
0.945 µg/g;54 and USA (South Dakota/Wyoming), 1.517 µg/g.11
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
145
Only 1 case-control study, in a Chinese population with low selenium intake,
investigated selenium intake, measured by means of a food frequency questionnaire,
in relation to ESCC risk.20 Intake of selenium, measured by questionnaires, correlates
poorly with actual selenium intake.55 However, the Chinese study was able to find a
statistically significant inverse association (p-trend<0.001).20 In the Chinese Linxian trial
cohort,17, 18 which is a population with low selenium levels, a significant inverse
association between serum selenium and ESCC was also found, and the range of
selenium levels in that population was comparable to the range in our cohort.8 Two
other case-control studies investigated esophageal cancer, but did not make a
distinction between ESCC and EAC.12,16 However, these esophageal cancer cases were
presumably mostly ESCC cases. One of these studies was performed in a Japanese
population with a moderate selenium status12 and the other in an Indian population
with a high selenium status.16 These studies reported significantly lower blood
selenium levels in esophageal cancer cases than in controls, but did not report rate
ratios or odds ratios. Also, the US Nutritional Prevention of Cancer trial found lower
esophageal cancer risk in subjects who received daily selenium supplements.15 The
results of the studies mentioned above suggest inverse associations with ESCC over the
range from low to high selenium status, whereas too few studies have been performed
to make any inferences on EAC.
For GCA, the Linxian trial cohort reported a statistically significant inverse
association with serum selenium.17 Two other studies, in a Chinese and Finnish
population with a relatively low selenium status, found an inverse association between
serum selenium levels and total gastric cancer.12,30 Within the NLCS, we previously
found evidence suggestive of an inverse association with total gastric cancer.23 In
contrast, no inverse association was found between serum or toenail selenium levels
and gastric cancer in 3 other studies in Japanese, Hawaiian (with Japanese ancestry)
and Colombian populations with a higher selenium status.54,56,57 This suggests that a
ceiling effect may exist, such that selenium is only associated with gastric cancer risk in
populations with relatively low selenium levels.
If selenium is associated with lower esophageal and gastric cancer risk, the
question arises by which mechanism selenium exerts its effect and in which stage of
carcinogenesis. Two studies, investigating subjects with Barrett’s esophagus21 and with
dysplasia of the squamous epithelium,19 can give clues about the timing of the effect of
selenium. The former study found less signs of progression in subjects with the highest
serum selenium levels, whereas the latter found that selenium supplements were
associated with less progression and more regression of the dysplasia. These findings
may mean that selenium can have an effect after the epithelium of the esophagus has
already changed. No effect, however, of selenium containing supplements was found
in a third study among subjects with dysplasia.58
The possibility for a selenium effect earlier in the carcinogenesis also remains
open. Any antioxidant effects2,3 of selenium, for instance, might occur early in
carcinogenesis. In our interaction analyses of EAC, we found inverse associations
Thesis Jessie Steevens_v04.pdf
Chapter 7
146
specifically for those subjects with a low intake of antioxidants. This finding could
indicate an antioxidant mechanism of selenium for EAC. Other findings, however, did
not match an antioxidant mechanism. We found an inverse association with EAC and
GCA only in never smokers, whereas, in case of an antioxidant mechanism, one would
expect these associations with selenium in the current smokers because smokers are
exposed to oxidative stress,29 and this oxidative stress could be counteracted by
antioxidants.
The associations we observed between selenium and esophageal and gastric
cancers may also be a result of other mechanisms. In fact, it has been suggested that
selenium may act at a number of stages in carcinogenesis and by a number of different
mechanisms (see introduction) that may operate simultaneously, or consecutively.3
The few in vitro and vivo studies that have been performed on selenium and
esophageal carcinogenesis, show effects on different processes in early and late stages
(eg, mutagenesis, development of precancerous lesions and cancer, cancer cell
proliferation, and tumor growth).59-62
To learn more about the possible role of selenium in the etiology of ESCC, EAC,
and GCA, the observations reported here need confirmation or refutation from other
epidemiologic studies. Additionally, it would be interesting to study selenium in
association with risk of Barrett’s esophagus. A comparison of both associations may
give more insight into the mechanism of action of selenium and the stage in the
development of normal esophageal epithelium into EAC at which selenium exerts its
effect.
Two strengths of our study are the size and the division of esophageal cancer into
2 main histologic subtypes: ESCC and EAC. This division enables the identification of
possible differences in the etiology of these cancers. Thirdly, we are the first to
investigate selenium in relation to risk of ESCC, EAC, and GCA in a Western population.
Most previous studies were performed in Asia. A fourth strength is the prospective
character of our study, which prevents bias by reverse causation, a concern in case-
control studies. We investigated whether undiagnosed cancer at baseline may have
influenced our results, by excluding cases diagnosed in the first 2 years of follow-up,
but the results proved robust to this exclusion. Finally, the observation that our results
hardly changed after adjustment for confounders, makes residual confounding
unlikely.
Our study has some limitations that should be described. Toenail selenium levels
were measured only once in our cohort. In the Nurses’ Health Study, a correlation
coefficient of 0.48 was found between toenail selenium levels for specimens taken 6
years apart.53 This moderate reliability may have resulted in some misclassification of
exposure. This misclassification is nondifferential and thus may have attenuated the
association. Another limitation to our study was the relatively low number of female
EAC and GCA cases, which prevented a more detailed investigation of the association
with selenium by sex. As a final point, many statistical tests were performed, which
increases the chance of false positive results.
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
147
In conclusion, this prospective study supports an inverse association between
toenail selenium levels and risk of ESCC and GCA and suggests an inverse association
with risk of EAC in subgroups (women, never smokers, and low antioxidant
consumers).
Thesis Jessie Steevens_v04.pdf
Chapter 7
148
REFERENCES
1. Navarro Silvera SA, Rohan TE. Trace elements and cancer risk: a review of the epidemiologic evidence.
Cancer Causes Control 2007;18:7-27.
2. Jackson MI, Combs GF, Jr. Selenium and anticarcinogenesis: underlying mechanisms. Curr Opin Clin
Nutr Metab Care 2008;11:718-726.
3. Rayman MP. Selenium in cancer prevention: a review of the evidence and mechanism of action. Proc
Nutr Soc 2005;64:527-542.
4. Navarro-Alarcon M, Cabrera-Vique C. Selenium in food and the human body: a review. Sci Total
Environ 2008;400:115-141.
5. Institute of Medicine. Dietary reference intakes for vitamin C, vitamin E, selenium, and carotenoids.
Washington, D.C., 2000.
6. Netherlands Nutrition Centre. http://www.voedingscentrum.nl/nl/eten-gezondheid/voedingstoffen/
vitamines-en-mineralen/hoeveel-heb-je-nodig.aspx#c (accessed 13 August 2009)
7. Hulshof KFAM, Kistemaker C, Kruizinga AG. [The contribution of food groups to the intake of energy
and nutrients. Dutch National Food Consumption Survey 1997-1998]. Zeist: TNO, 1998.
8. van 't Veer P, van der Wielen RP, Kok FJ, Hermus RJ, Sturmans F. Selenium in diet, blood, and toenails
in relation to breast cancer: a case-control study. Am J Epidemiol 1990;131:987-994.
9. Willett WC, Buzzard IM. Foods and nutrients. In: Willett WC, ed. Monographs in Epidemiology and
Biostatistics. Volume 30. Nutritional epidemiology. New York: Oxford Universtiy Press, 1998:18-32.
10. Hunter DJ, Morris JS, Chute CG, Kushner E, Colditz GA, Stampfer MJ, Speizer FE, Willett WC. Predictors
of selenium concentration in human toenails. Am J Epidemiol 1990;132:114-122.
11. Longnecker MP, Stram DO, Taylor PR, Levander OA, Howe M, Veillon C, McAdam PA, Patterson KY,
Holden JM, Morris JS, Swanson CA, Willett WC. Use of selenium concentration in whole blood, serum,
toenails, or urine as a surrogate measure of selenium intake. Epidemiology 1996;7:384-390.
12. Ujiie S, Kikuchi H. The relation between serum selenium value and cancer in Miyagi, Japan: 5-year
follow up study. Tohoku J Exp Med 2002;196:99-109.
13. Jaskiewicz K, Marasas WF, Rossouw JE, van Niekerk FE, Heine Tech EW. Selenium and other mineral
elements in populations at risk for esophageal cancer. Cancer 1988;62:2635-2639.
14. Knekt P, Aromaa A, Maatela J, Alfthan G, Aaran RK, Nikkari T, Hakama M, Hakulinen T, Teppo L. Serum
micronutrients and risk of cancers of low incidence in Finland. Am J Epidemiol 1991;134:356-361.
15. Clark LC, Combs GF, Jr., Turnbull BW, Slate EH, Chalker DK, Chow J, Davis LS, Glover RA, Graham GF,
Gross EG, Krongrad A, Lesher JL, Jr., Park HK, Sanders BB, Jr., Smith CL, Taylor JR. Effects of selenium
supplementation for cancer prevention in patients with carcinoma of the skin. A randomized
controlled trial. Nutritional Prevention of Cancer Study Group. Jama 1996;276:1957-1963.
16. Krishnaswamy K, Prasad MP, Krishna TP, Pasricha S. A case control study of selenium in cancer. Indian J
Med Res 1993;98:124-128.
17. Mark SD, Qiao YL, Dawsey SM, Wu YP, Katki H, Gunter EW, Fraumeni JF, Jr., Blot WJ, Dong ZW, Taylor
PR. Prospective study of serum selenium levels and incident esophageal and gastric cancers. J Natl
Cancer Inst 2000;92:1753-1763.
18. Wei WQ, Abnet CC, Qiao YL, Dawsey SM, Dong ZW, Sun XD, Fan JH, Gunter EW, Taylor PR, Mark SD.
Prospective study of serum selenium concentrations and esophageal and gastric cardia cancer, heart
disease, stroke, and total death. Am J Clin Nutr 2004;79:80-85.
19. Limburg PJ, Wei W, Ahnen DJ, Qiao Y, Hawk ET, Wang G, Giffen CA, Wang G, Roth MJ, Lu N, Korn EL,
Ma Y, Caldwell KL, Dong Z, Taylor PR, Dawsey SM. Randomized, placebo-controlled, esophageal
squamous cell cancer chemoprevention trial of selenomethionine and celecoxib. Gastroenterology
2005;129:863-873.
20. Lu H, Cai L, Mu LN, Lu QY, Zhao J, Cui Y, Sul JH, Zhou XF, Ding BG, Elashoff RM, Marshall J, Yu SZ, Jiang
QW, Zhang ZF. Dietary mineral and trace element intake and squamous cell carcinoma of the
esophagus in a Chinese population. Nutr Cancer 2006;55:63-70.
21. Rudolph RE, Vaughan TL, Kristal AR, Blount PL, Levine DS, Galipeau PC, Prevo LJ, Sanchez CA,
Rabinovitch PS, Reid BJ. Serum selenium levels in relation to markers of neoplastic progression among
persons with Barrett's esophagus. J Natl Cancer Inst 2003;95:750-757.
22. Liu C, Russell RM. Nutrition and gastric cancer risk: an update. Nutr Rev 2008;66:237-249.
Thesis Jessie Steevens_v04.pdf
Selenium status and the risk of esophageal and gastric cancer subtypes
149
23. van den Brandt PA, Goldbohm RA, van 't Veer P, Bode P, Dorant E, Hermus RJ, Sturmans F. A
prospective cohort study on toenail selenium levels and risk of gastrointestinal cancer. J Natl Cancer
Inst 1993;85:224-229.
24. Blot WJ, Li JY, Taylor PR, Guo W, Dawsey S, Wang GQ, Yang CS, Zheng SF, Gail M, Li GY, et al. Nutrition
intervention trials in Linxian, China: supplementation with specific vitamin/mineral combinations,
cancer incidence, and disease-specific mortality in the general population. J Natl Cancer Inst
1993;85:1483-1492.
25. Holmes RS, Vaughan TL. Epidemiology and pathogenesis of esophageal cancer. Semin Radiat Oncol
2006;17:2-9.
26. Kelley JR, Duggan JM. Gastric cancer epidemiology and risk factors. J Clin Epidemiol 2003;56:1-9.
27. Vizcaino AP, Moreno V, Lambert R, Parkin DM. Time trends incidence of both major histologic types of
esophageal carcinomas in selected countries, 1973-1995. Int J Cancer 2002;99:860-868.
28. Steevens J, Botterweck AA, Dirx MJ, van den Brandt PA, Schouten LJ. Trends in incidence of
oesophageal and stomach cancer subtypes in Europe. Eur J Gastroenterol Hepatol 2010;22:669-678.
29. Mena S, Ortega A, Estrela JM. Oxidative stress in environmental-induced carcinogenesis. Mutat Res
2009;674:36-44.
30. Knekt P, Aromaa A, Maatela J, Alfthan G, Aaran RK, Hakama M, Hakulinen T, Peto R, Teppo L. Serum
selenium and subsequent risk of cancer among Finnish men and women. J Natl Cancer Inst
1990;82:864-868.
31. van den Brandt PA, Zeegers MP, Bode P, Goldbohm RA. Toenail selenium levels and the subsequent
risk of prostate cancer: a prospective cohort study. Cancer Epidemiol Biomarkers Prev 2003;12:
866-871.
32. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
33. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
34. Casparie M, Tiebosch AT, Burger G, Blauwgeers H, van de Pol A, van Krieken JH, Meijer GA. Pathology
databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide
histopathology and cytopathology data network and archive. Cell Oncol 2007;29:19-24.
35. van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol
1990;19:553-558.
36. Goldbohm R, van Den Brandt PA, Dorant E. Estimation of the coverage of Dutch municipalities by
cancer registries and PALGA based on hospital discharge data. Tijdschr Soc Gezondheidsz 1994;72:
80-84.
37. Parkin DM, Shanmugaratnam K, Sobin L, Ferlay J, Whelan SL. Histological Groups for comparative
studies. IARC Technical reports. Volume 31. Lyon: International Agency for Research on Cancer, 1998.
38. Nevo table: Dutch food composition table, 1986-1987. (Dutch). Voorlichtingbureau Voor de Voeding,
1986.
39. Zeegers MP, Goldbohm RA, Bode P, van den Brandt PA. Prediagnostic toenail selenium and risk of
bladder cancer. Cancer Epidemiol Biomarkers Prev 2002;11:1292-1297.
40. Bode P. Automation and quality assurance in the Neutron Activation Facilities in Delft. J Radioanal Nucl
Chem 2000;245:127-132.
41. van den Brandt PA, Goldbohm RA, van't Veer P, Bode P, Hermus RJ, Sturmans F. Predictors of toenail
selenium levels in men and women. Cancer Epidemiol Biomarkers Prev 1993;2:107-112.
42. van Breda SG, Hogervorst JG, Schouten LJ, Knaapen AM, van Delft JH, Goldbohm RA, van Schooten FJ,
van den Brandt PA. Toenails: an easily accessible and long-term stable source of DNA for genetic
analyses in large-scale epidemiological studies. Clin Chem 2007;53:1168-1170.
43. Goldbohm RA, van den Brandt PA, Brants HA, van't Veer P, Al M, Sturmans F, Hermus RJ. Validation of
a dietary questionnaire used in a large-scale prospective cohort study on diet and cancer. Eur J Clin
Nutr 1994;48:253-265.
44. Schottenfeld D, Fraumeni JF, Jr. Cancer epidemiology and prevention. Oxford University Press, 2006.
45. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:
1503-1511.
Thesis Jessie Steevens_v04.pdf
Chapter 7
150
46. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
47. WHO Collaborating Centre for Drug Statistics Methodology. http://www.whocc.no/atcddd/
welcome.html (accessed 15 Oct, 2009).
48. Cox DR. Regression models and life-tables. J Roy Statistical Society 1972;34:187-220.
49. Barlow WE. Robust variance estimation for the case-cohort design. Biometrics 1994;50:1064-1072.
50. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrica 1982;69:
239-241.
51. Gao S, Jin Y, Hall KS, Liang C, Unverzagt FW, Ji R, Murrell JR, Cao J, Shen J, Ma F, Matesan J, Ying B,
Cheng Y, Bian J, Li P, Hendrie HC. Selenium level and cognitive function in rural elderly Chinese. Am J
Epidemiol 2007;165:955-965.
52. Ovaskainen ML, Virtamo J, Alfthan G, Haukka J, Pietinen P, Taylor PR, Huttunen JK. Toenail selenium as
an indicator of selenium intake among middle-aged men in an area with low soil selenium. Am J Clin
Nutr 1993;57:662-665.
53. Garland M, Morris JS, Rosner BA, Stampfer MJ, Spate VL, Baskett CJ, Willett WC, Hunter DJ. Toenail
trace element levels as biomarkers: reproducibility over a 6-year period. Cancer Epidemiol Biomarkers
Prev 1993;2:493-497.
54. Koriyama C, Campos FI, Yamamoto M, Serra M, Carrasquilla G, Carrascal E, Akiba S. Toenail selenium
levels and gastric cancer risk in Cali, Colombia. J Toxicol Sci 2008;33:227-235.
55. Zhang ZW, Shimbo S, Qu JB, Watanabe T, Nakatsuka H, Matsuda-Inoguchi N, Higashikawa K, Ikeda M.
Dietary selenium intake of Chinese adult women in the 1990s. Biol Trace Elem Res 2001;80:125-138.
56. Kabuto M, Imai H, Yonezawa C, Neriishi K, Akiba S, Kato H, Suzuki T, Land CE, Blot WJ. Prediagnostic
serum selenium and zinc levels and subsequent risk of lung and stomach cancer in Japan. Cancer
Epidemiol Biomarkers Prev 1994;3:465-469.
57. Nomura A, Heilbrun LK, Morris JS, Stemmermann GN. Serum selenium and the risk of cancer, by
specific sites: case-control analysis of prospective data. J Natl Cancer Inst 1987;79:103-108.
58. Li JY, Taylor PR, Li B, Dawsey S, Wang GQ, Ershow AG, Guo W, Liu SF, Yang CS, Shen Q, et al. Nutrition
intervention trials in Linxian, China: multiple vitamin/mineral supplementation, cancer incidence, and
disease-specific mortality among adults with esophageal dysplasia. J Natl Cancer Inst 1993;85:
1492-1498.
59. Xiao HJ, Huang CY, Wang HY, Li M. [Effect of selenium and zinc on the proliferation of human
esophageal cancer Eca109 cell line in vitro]. Nan Fang Yi Ke Da Xue Xue Bao 2008;28:2117-2120.
60. Guttenplan JB, Kosinska W, von Pressentin MM, Rosa J, El-Bayoumy K. Effects of 1,4-
phenylenebis(methylene)selenocyanate (p-XSC) and vitamin E on 4-nitroquinoline-N-oxide (4-NQO)-
induced mutagenesis in lacZ mouse upper aerodigestive tissue. Mutat Res 2002;518:85-93.
61. Guttenplan JB, Spratt TE, Khmelnitsky M, Kosinska W, Desai D, El-Bayoumy K. Effects of 3H-1,2-
dithiole-3-thione, 1,4-phenylenebis(methylene)selenocyanate, and selenium-enriched yeast
individually and in combination on benzo[a]pyrene-induced mutagenesis in oral tissue and esophagus
in lacZ mice. Mutat Res 2004;559:199-210.
62. Bogden JD, Chung HR, Kemp FW, Holding K, Bruening KS, Naveh Y. Effect of selenium and molybdenum
on methylbenzylnitrosamine-induced esophageal lesions and tissue trace metals in the rat. J Nutr
1986;116:2432-2442.
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and
risk of esophageal and gastric cancer
subtypes in the Netherlands Cohort Study
Jessie Steevens
Leo J Schouten
R Alexandra Goldbohm
Piet A van den Brandt
Submied
8
151
Thesis Jessie Steevens_v04.pdf
Chapter 8
152
ABSTRACT
Prospective epidemiologic data on vegetables and fruits consumption and risk of
subtypes of esophageal and gastric cancer are sparse. We studied the association
between vegetables and fruits consumption and risk of esophageal squamous cell
carcinoma (ESCC), esophageal adenocarcinoma (EAC), gastric cardia adenocarcinoma
(GCA), and gastric noncardia adenocarcinoma (GNCA) in the Netherlands Cohort Study.
In 1986, 120852 Dutch men and women aged 55-69 filled out a questionnaire on diet
and other cancer risk factors. After 16.3 years of follow-up, 101 ESCC, 144 EAC, 156
GCA, 460 GNCA cases and 4035 subcohort members were available for case-cohort
analysis using Cox proportional hazards models. Multivariable adjusted incidence rate
ratios (RR) were generally below unity. Significant inverse associations were observed
for raw vegetables and EAC risk [RR per 25g/day: 0.81, 95% confidence interval (CI)
0.68-0.98], and Brassica vegetables and GCA risk (RR per 25g/day: 0.72, 95% CI
0.54-0.95). Citrus fruits were inversely associated with EAC and GCA risk (RRs for
highest vs. lowest intake: 0.55, 95% CI 0.31-0.98 and 0.38, 95% CI 0.21-0.69,
respectively). Specifically for current smokers, vegetables and fruits intake was
inversely associated with ESCC and EAC risk. Consumption of vegetables and fruits may
protect against subtypes of esophageal and gastric cancer.
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
153
INTRODUCTION
Vegetables and fruits have been of interest for their possible favorable influence on
risk of cancer and other diseases for some decades. These foods contain many
substances that are potentially anticarcinogenic, such as vitamins, carotenoids, and
flavonoids. Possible mechanisms of action include modulation of DNA methylation,
protection from and repair of DNA damage, induction of detoxifying phase II enzymes,
and promotion of apoptosis.1,2. A recent expert report concluded that consumption of
vegetables and fruits probably protects against development of esophageal and gastric
cancer.2 Unfortunately, in this report, no distinction was made between the main
subtypes of these cancers: esophageal squamous cell carcinoma (ESCC), esophageal
adenocarcinoma (EAC), gastric cardia adenocarcinoma (GCA), and gastric noncardia
adenocarcinoma (GNCA). It is interesting and important to look into subtypes of these
cancers, as the evidence is growing that there are differences in risk factors between
these cancers.3-5 Case-control studies on vegetables or fruits consumption and ESCC,6-9
EAC,7,8,10,11 GCA,7,8,12 or GNCA7,12 mostly reported inverse associations. However, case-
control studies are vulnerable for biases because of their retrospective nature. This can
specifically be a problem when investigating diet and gastrointestinal cancer.13 Cohort
studies are less vulnerable for biases. Cohort studies that reported on ESCC have
mostly found inverse associations,14-16 but some cohorts found no association with
vegetables16,17 or fruits17 consumption. For EAC,14,18 GCA,16,18-21 and GNCA16,18-21 mainly
null associations have been observed, except for two studies that found inverse
associations for fruits consumption and GCA,16 and GNCA20.
Using 3.322 and 6.313 year follow-up data of the Netherlands Cohort Study on diet
and cancer (NLCS), we have previously reported on the associations between
vegetables and fruits consumption and risk of total gastric cancer.13 Only for Allium
vegetables, separate results for GCA and GNCA were presented.22
The aim of this study was to prospectively investigate the role of vegetables and
fruits consumption in the development of ESCC, EAC, GCA, and GNCA. Also, we aimed
to study several groups of vegetables and fruits and individual vegetable and fruit
items separately, as few studies have performed such comprehensive analyses. For
this study, we used 16.3-year follow-up data from the NLCS.
MATERIALS AND METHODS
Study design and participants
This study was conducted within the prospective Netherlands Cohort Study, which
started in September 1986 with the inclusion of 58279 men and 62573 women aged
55-69 years, who were randomly sampled from Dutch municipal registries.23 At
Thesis Jessie Steevens_v04.pdf
Chapter 8
154
baseline, all subjects completed a self-administered questionnaire on dietary and
lifestyle habits and other cancer risk factors.
For reasons of efficiency, the case-cohort approach was used for data processing
and analysis.24,25 Cases were derived from the entire cohort, and the number of
person-years at risk for the entire cohort was estimated from a subcohort of 5000,
who were randomly sampled from the total cohort at baseline. The subcohort has
been followed up by contacting the participants and using data from municipal
population registries.23 Person-years at risk were calculated from the start of the study
until esophageal or gastric cancer diagnosis, death, emigration, loss to follow-up or
end of follow-up, whichever occurred first. Only one male subcohort member was lost
to follow-up after 16.3 years.
Follow-up for cancer incidence was performed by record linkage26 to the
Netherlands Cancer Registry (NCR) and PALGA, the nationwide network and registry of
histopathology and cytopathology in the Netherlands.27 Data of 16.3 years of follow-up
(baseline to December 31, 2002) were used. The completeness of cancer follow-up is
≥96%.28 The following numbers of incident, microscopically confirmed, primary
carcinomas were identified: 130 ESCC29 (ICD-O-3 C15), 181 EAC29 (C15), 206 GCA
(C16.0) and 594 GNCA (C16.1-C16.9) (Figure 8.1). The group of GNCA included cancers
with a lesion with overlapping subsites of the stomach (C16.8, n=160) and some
gastric, not otherwise specified cancers (C16.9, n=75), raising the possibility that some
cardia cancers might be included in the non-cardia category. However, as we found
risk estimates to be similar in separate analyses of gastric cancers of specified sites
(C16.1-C16.5) and other gastric cancers (C16.6-C16.9) (data not shown), we combined
the groups in the analysis.
We excluded subjects who reported having prevalent cancer other than skin
cancer at baseline. Also excluded were cases and subcohort members with incomplete
or inconsistent dietary data.30 Further details on the design of the study and methods
of follow-up have been published previously.23,26 The Medical Ethics Committee of
Maastricht University has approved the study.
Questionnaire data
The self-administered questionnaire included a 150-item food frequency questionnaire
(FFQ). The FFQ concentrated on the habitual consumption of foods and beverages
during the year preceding the start of the study.
With regard to vegetables consumption, subjects were asked to report their
frequency of consumption of a number of vegetables (see Table 8.2), both in summer
and in winter. They could choose one of six categories, ranging from “never or less
than 1x per month” to “3-7x per week”. Subjects were asked about usual serving sizes
only for string beans and cooked endive; the mean of these values served as an
indicator to derive the serving sizes of all cooked vegetables, according to a vegetable-
specific algorithm based on results of a pilot study. For onions and tomatoes,
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
155
participants were asked to report their consumption in number per week; for sweet
peppers per month, and for mushrooms the number of 250 g boxes per month.
Subjects with inconsistent vegetable data were excluded from analysis.
With regard to fruits consumption, subjects were asked to report the frequency
and amount of consumption of a number of fruits (listed in Table 8.2). Categories
ranged from “never or less than 1x per month” to “6-7x per week”. Using household
units, these frequencies and amounts have been converted to consumption in
grams/day.
The choice of items in the questionnaire was such that it covered almost all
vegetables and fruits eaten regularly, with the exception of chicory, red cabbage, and
cucumber. Furthermore, in an open-ended question, participants could fill in other
foods they consumed regularly as well as the frequency and amount consumed on
each occasion.
Questionnaire data were key-entered and processed in a standardized manner,
blinded with respect to case/subcohort status to minimize observer bias in the coding
and interpretation of the data.
Netherlands cohort study on diet and cancer (120,852)
↓ ↓
Subcohort randomly drawn
from total cohort
Record linkage with Netherlands Cancer
Registry and PALGA*
↓ ↓ ↓ ↓ ↓
5000 ESCC* EAC* GCA* GNCA*
Exclusion of participants with prevalent cancer at baseline
↓ ↓ ↓ ↓ ↓
4774 130 181 206 594
Exclusion of participants with incomplete or inconsistent dietary data
↓ ↓ ↓ ↓ ↓
4438 120 168 187 551
Exclusion of participants with missing data on confounders
↓ ↓ ↓ ↓ ↓
4111 111 153 168 512
Exclusion of first 2 years of follow-up
↓ ↓ ↓ ↓ ↓
fruit analyses: 4035 101 144 156 460
Exclusion of participants with inconsistent vegetable data
↓ ↓ ↓ ↓ ↓
vegetable analyses: 3827 96 137 148 443
Figure 8.1 Flow diagram of subcohort members and cases on whom the analyses were based. *PALGA:
nationwide network and registry of histo- and cytopathology in the Netherlands, ESCC:
esophageal squamous cell carcinoma, EAC: esophageal adenocarcinoma, GCA: gastric cardia
adenocarcinoma, GNCA: gastric noncardia adenocarcinoma.
Thesis Jessie Steevens_v04.pdf
Chapter 8
156
The FFQ was validated against a 9-day diet record.30 The Spearman correlation
coefficients were 0.38 for total vegetables consumption and 0.60 for total fruits
consumption. On average, vegetables consumption appeared to be slightly
overestimated and fruits consumption to be underestimated by the FFQ as compared
with the diet records.
Statistical analysis
We checked for possible influence of preclinical cancer at baseline on the consumption
of vegetables and fruits.13 This was done by comparing the mean consumption of
vegetables and fruits for cases that were diagnosed during the first 2 years of follow-
up with the mean consumption for cases diagnosed later in follow-up. An independent
samples t-test was used to test for statistical significance of differences, after a square-
root transformation was applied to normalize the distribution of the vegetables and
fruits variables. Vegetables consumption was statistically significantly lower among
early vs. late GCA (difference 43 g/day, P=0.04) and GNCA (difference 22 g/day,
P=0.02) cases, while fruits consumption was lower among early vs. late ESCC cases
(difference 74 g/day, P=0.01). Based on these observations, we decided to exclude the
first 2 years of follow-up from all analyses, to prevent bias caused by reversed
causation due to preclinical cancer.
Analyses were performed for total vegetables consumption, cooked and raw
vegetables, several vegetable subgroups, and the most frequently consumed individual
vegetables. Furthermore, we analyzed total fruits consumption, citrus fruits and the
most frequently consumed individual fruits. The composition of each vegetable and
fruit group can be found in the Appendix. For categorical analyses, vegetables and
fruits consumption levels were categorized into quintiles according to the distribution
in the subcohort. For continuous analyses, an increment of 25 g/day was chosen.
The following confounders were included in all models because they changed the
incidence rate ratios (RR) by more than 5%: age (years), sex, cigarette smoking (current
yes/no, number of cigarettes smoked daily, and number of smoking years), and
consumption of alcohol (g/day), red meat (g/day), and fish (g/day). The following
variables were considered potential confounders but were not included in the models
because they did not change the RRs by more than 5%: consumption of tea,
consumption of meat products, non-occupational physical activity, body mass index
(BMI), highest educational level, family history of esophageal or gastric cancer, long-
term use of non-steroidal anti-inflammatory drugs or aspirin, or lower esophageal
sphincter (LES) relaxing medication.31,32 Analyses of vegetables were adjusted for fruits
consumption and vice versa. For exact model specifications, see the table footnotes.
Cox proportional hazards models were used to estimate age- and sex-adjusted
and multivariable adjusted RRs and corresponding 95% confidence intervals (CI).33
Standard errors were estimated using a robust covariance matrix estimator to account
for increased variance due to sampling from the cohort.24 The proportional hazards
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
157
assumption was tested using the scaled Schoenfeld residuals.34 Tests for dose-
response trends were assessed by fitting ordinal exposure variables as continuous
terms.
Because of the limited numbers of cases, specifically female cases, all analyses
were carried out for both sexes combined. Nevertheless, analyses stratified by sex
were performed for total vegetables and fruits consumption using continuous
variables. We also investigated possible interaction between vegetables and fruits
consumption and cigarette smoking status (current, former, never) by performing
stratified analyses. P-values for interaction were assessed by including a cross-product
term in the models.
All analyses were done using Stata 9.2 statistical software package (StataCorp,
College Station, Texas, USA). The significance level α was set at 0.05. All reported p-
values are two-sided.
RESULTS
Characteristics of the study population
Baseline characteristics of the subcohort and case groups are presented in Table 8.1.
The most striking difference is the high percentage of men among EAC (78%) and GCA
cases (85%), compared with the subcohort (49% men). Furthermore, cases were more
likely current cigarette smokers than subcohort members and among ever smokers,
cases smoked longer and more cigarettes than did subcohort members. Ethanol intake
was also higher in all case groups than in the subcohort, particularly among ESCC
cases. EAC cases had a higher BMI than the subcohort.
Baseline vegetables and fruits consumption
Table 8.2 presents baseline intakes of the individual vegetables and fruits that were
listed in the questionnaire. Because the vegetables and fruits had a right-skewed
distribution, we present medians and interquartile ranges. The vegetables and fruits
are ranked by increasing percentage of nonusers in the subcohort. The median daily
consumption of the subcohort was 179 grams of vegetables and 157 grams of fruits.
The most frequently eaten vegetables in our population were string beans, cauliflower,
and lettuce. When we look at the median daily consumption among users, the
vegetables eaten in the largest amounts were tomatoes, onions, and string beans. The
fruits eaten in the largest amounts were apples and pears, and oranges.
Thesis Jessie Steevens_v04.pdf
Chapter 8
158
Thesis Jessie Steevens_v04.pdf
Table 8.1 Characteristics of cases and subcohort members in the Netherlands cohort study on diet and cancer (NLCS), 1986-2002.
Characteristic Subcohort Cases
(n=4035)a
ESCC
(n=101)a
EAC
(n=144)a
GCA
(n=156)a
GNCA
(n=460)a
Age at baseline (years) 61.3 (4.2) 62.6 (4.1) 61.4 (4.1) 61.4 (4.0) 62.5 (4.1)
Men 49% 54% 78% 85% 66%
Level of education:
primary 28% 34% 26% 29% 38%
lower vocational 22% 20% 25% 24% 24%
secondary and medium vocational 36% 34% 32% 28% 29%
university and higher vocational 14% 12% 16% 19% 9%
Cigarette smoking status:
never smoker 37% 23% 19% 17% 24%
former smoker 36% 30% 49% 47% 40%
current smoker 27% 48% 32% 35% 36%
Ever cigarette smokers:
frequency of cigarette smoking (n/day) 15.2 (10.3) 18.1 (9.9) 20.0 (12.8) 17.2 (10.8) 16.2 (10.3)
duration of cigarette smoking (years) 31.5 (12.3) 36.1 (11.0) 33.8 (10.8) 35.0 (11.4) 34.5 (12.0)
Ethanol intake (g/day) 10.2 (14.2) 23.1 (28.4) 15.1 (18.0) 13.7 (16.0) 11.9 (15.8)
Body mass index (kg/m2) 25.0 (3.1) 24.3 (3.6) 26.3 (3.4) 25.7 (3.0) 25.0 (3.2)
Non-occupational physical activity (min/day) 73.1 (60.6) 69.4 (65.7) 75.1 (57.1) 88.1 (80.0) 78.5 (69.6)
Family history of esophageal or gastric cancer 7% 10% 12% 8% 13%
Use of NSAIDs b 6% 7% 8% 5% 5%
Use of LES relaxing medication b 14% 16% 19% 15% 15%
EAC, esophageal adenocarcinoma; ESCC, esophageal squamous cell carcinoma; GCA, gastric cardia adenocarcinoma; GNCA, gastric non-cardia adenocarcinoma; LES,
lower esophageal sphincter; NSAIDs, non-steroidal anti-inflammatory drugs. Values are given as mean (SD); for categorical variables percentages are presented.
a The number of subcohort members or cases used after exclusion of prevalent cancer cases, first two years of follow-up, subjects with inconsistent or incomplete
questionnaire data, and subjects with missing data on alcohol consumption or cigarette smoking (status, frequency and duration); b longer than six months.
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
159
Table 8.2 Vegetables and fruit consumption of the subcohort members in the Netherlands Cohort Study
on diet and cancer (NLCS), 1986-2002.
Food group or food item Nonusers Users: daily intake (g)
% median (P25-P75)
Total vegetables consumption (n=3827)a 0 179 (137-229)
String/French beans 2 17 (10-25)
Cauliflower 7 13 (8-20)
Lettuce 9 7 (4-11)
Carrots, cooked 12 8 (4-13)
Endive, cooked 14 12 (7-19)
Brussels sprouts 15 8 (5-12)
Sauerkraut 16 6 (3-9)
Tomatoes 18 24 (14-33)
Onion 20 22 (11-33)
Spinach 20 10 (6-16)
Beetroot 25 9 (5-14)
Kale 25 4 (2-5)
Cabbage 29 8 (4-13)
Leek 32 10 (5-17)
Dried pulses b 39 11 (4-17)
Mushrooms 39 4 (4-9)
Broad beans 43 6 (3-11)
Sweet peppers 47 4 (3-6)
Endive, raw 56 4 (2-7)
Rhubarb 59 3 (1-6)
Carrots, raw 67 5 (2-10)
Gherkins 70 3 (1-6)
Vegetable juices b 90 12 (5-25)
Total fruit consumption (n=4035)c 1 157 (96-236)
Apples and pears 13 80 (45-116)
Strawberries 14 7 (4-11)
Oranges and fresh orange juice 16 42 (15-83)
Grapes 37 3 (1-8)
Mandarins 41 4 (2-8)
Bananas 47 11 (4-19)
Processed fruit juices b 51 23 (12-63)
Grapefruits and fresh grapefruit juice 70 16 (7-40)
Raisins and other dried fruits 75 1 (1-3)
a The number of subcohort members used after exclusion of prevalent cancer cases, first two years of follow-
up, subjects with inconsistent or incomplete questionnaire data, or inconsistent vegetable data, and subjects
with missing data on alcohol consumption or cigarette smoking (status, frequency and duration); b Dried
pulses and vegetable juices are not included in total vegetables consumption. Processed fruit juices are not
included in total fruits consumption; c The number of subcohort members used after exclusion of prevalent
cancer cases, first two years of follow-up, subjects with inconsistent or incomplete questionnaire data, and
subjects with missing data on alcohol consumption or cigarette smoking (status, frequency and duration).
Main analyses
Results of the multivariable Cox regression analyses on total vegetables and fruits and
subgroups of vegetables and fruits are shown in Table 8.3. The results from the age-
and sex-adjusted analyses are presented in Supplementary Table 8.1, due to limited
space. These results were comparable with the multivariable adjusted results.
Thesis Jessie Steevens_v04.pdf
Chapter 8
160
For total vegetables consumption we observed that subjects in quintiles 2 to 5
had lower risks of esophageal and gastric cancers, when compared with quintile 1. The
clearest inverse association was observed for EAC (RR for Q5 vs. Q1: 0.59, 95% CI 0.33-
1.06), while the association was weakest for GNCA (RR for Q5 vs. Q1: 0.90, 95% CI
0.64-1.26). RRs or trends were not statistically significant for any tumor type.
For specific groups of vegetables, we can see that the majority of the RRs
observed were below unity. These inverse associations, however, mostly did not reach
statistical significance. No clear associations were observed for cooked vegetables,
while for raw vegetables we found a significant inverse trend (P=0.05) for EAC risk (RR
per 25 g/day increment: 0.80, 95% CI 0.68-0.98). Decreased risks associated with raw
vegetables were also observed for ESCC and GCA, but these were not statistically
significant. A quite strong, but not statistically significant, inverse association was
observed between consumption of legumes and pulses and ESCC risk, while for GNCA
this association was weaker, and no apparent associations were present for EAC and
GCA. Consumption of Brassica vegetables was associated with a statistically
significantly decreased risk of GCA (RR per 25 g/day: 0.72, 95% CI 0.54-0.95, p-
trend=0.01), and slightly decreased risk of EAC (RR per 25 g/day: 0.85, 95% CI 0.49-
1.48), while Brassicas were not clearly associated with risk of ESCC or GNCA. No clear
associations were seen for consumption of Allium vegetables or cooked leafy
vegetables and any tumor type, while for raw leafy vegetables inverse associations
were seen for ESCC and EAC.
Also shown in Table 8.3 are RRs for consumption of total fruits and citrus fruits.
Total fruits consumption was nonsignificantly inversely associated with ESCC risk (RR
for Q5 vs. Q1: 0.62, 95% CI 0.32-1.22), while RRs only slightly below unity were seen
for EAC, GCA, and GNCA. Citrus fruits were quite consistently associated with a
decreased risk of all tumors: for ESCC, the RR for Q5 vs. Q1 was 0.54 (95% CI 0.27-
1.07), for EAC this RR was 0.55 (95% CI 0.31-0.98), for GCA 0.38 (95% CI 0.21-0.69), and
finally for GNCA this RR was 0.80 (95% CI, 0.56-1.15).
In Table 8.4, we present multivariable adjusted results for the individual vegetable
and fruit items that were most frequently consumed in our cohort. Again, the age- and
sex-adjusted results can be found in a supplement (i.e. Supplementary Table 8.2).
None of the individual vegetable or fruit items was statistically significantly associated
with risk of ESCC or EAC. For GCA, we observed inverse associations for cauliflower (RR
per 25 g/day increment: 0.58, 95% CI 0.35-0.96), and oranges and fresh orange juices
(RR per 25 g/day: 0.86, 95% CI 0.77-0.95). Positive associations were observed for
tomato consumption and GNCA risk (RR per 25 g/day: 1.13, 95% CI 1.00-1.28), for
spinach consumption and GCA risk (RR per 25 g/day: 1.77, 95% CI 1.04-3.02), and for
apples and pears consumption and GCA risk (RR per 25 g/day: 1.05, 95% CI 1.00-1.09).
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
161
Thesis Jessie Steevens_v04.pdf
Table 8.3 Multivariable adjusted* associations between vegetables and fruit consumption and risk of esophageal and gastric cancer subtypes; Netherlands Cohort Study
on diet and cancer (NLCS) 1986-2002a
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI)
Total vegetables c
Q1 104 9073 31 1 reference 35 1 reference 43 1 reference 111 1 reference
Q2 146 9829 22 0.74 (0.38-1.43) 30 0.71 (0.42-1.19) 26 0.63 (0.37-1.06) 86 0.79 (0.58-1.08)
Q3 181 9662 17 0.56 (0.28-1.13) 25 0.61 (0.36-1.06) 27 0.64 (0.37-1.09) 91 0.88 (0.65-1.20)
Q4 222 9835 12 1.36 (0.76-2.44) 21 0.83 (0.50-1.38) 27 0.87 (0.53-1.45) 79 0.80 (0.58-1.10)
Q5 297 9823 14 0.61 (0.29-1.32) 26 0.59 (0.33-1.06) 25 0.87 (0.50-1.52) 76 0.90 (0.64-1.26)
p-trend
d 0.67 p-trend 0.18 p-trend 0.93 p-trend 0.65
continuous, per 25 g/day increment 48223 96 0.96 (0.89-1.04) 137 0.95 (0.89-1.02) 148 1.00 (0.94-1.07) 443 0.98 (0.94-1.02)
Cooked vegetables e
Q1 80 9495 22 1 reference 36 1 reference 33 1 reference 99 1 reference
Q2 115 9841 14 0.69 (0.34-1.38) 28 0.78 (0.47-1.30) 25 0.74 (0.43-1.26) 90 0.90 (0.66-1.22)
Q3 144 10269 17 0.75 (0.38-1.45) 21 0.56 (0.32-0.97) 25 0.71 (0.41-1.22) 88 0.83 (0.61-1.13)
Q4 177 10012 24 1.26 (0.69-2.29) 21 0.57 (0.32-1.02) 36 1.04 (0.64-1.70) 93 0.93 (0.68-1.27)
Q5 239 8606 19 0.95 (0.45-2.02) 31 1.01 (0.59-1.73) 29 0.93 (0.53-1.63) 73 0.79 (0.56-1.13)
p-trend 0.62 p-trend
0.93 p-trend
0.76 p-trend
0.28
continuous, per 25 g/day increment 48223 96 0.97 (0.89-1.06) 137
1.00 (0.91-1.09) 148 1.01 (0.93-1.09) 443 0.96 (0.92-1.01)
Raw vegetables e
Q1 8 9426 26 1 reference 31 1 reference 38 1 reference 95 1 reference
Q2 22 9550 17 0.72 (0.37-1.39) 37 1.22 (0.74-2.00) 29 0.79 (0.48-1.31) 79 0.87 (0.63-1.19)
Q3 34 9805 19 0.84 (0.45-1.56) 24 0.84 (0.48-1.47) 24 0.74 (0.43-1.26) 97 1.15 (0.84-1.57)
Q4 48 9968 17 0.70 (0.37-1.33) 30 1.00 (0.57-1.74) 29 0.86 (0.52-1.44) 88 1.05 (0.76-1.45)
Q5 77 9473 17 0.70 (0.35-1.41) 15 0.53 (0.27-1.05) 28 0.93 (0.53-1.61) 84 1.16 (0.83-1.62)
p-trend 0.35 p-trend 0.05 p-trend 0.94 p-trend 0.22
continuous, per 25 g/day increment 48223 96 0.94 (0.73-1.19) 137
0.81 (0.68- 0.98) 148
0.99 (0.85-1.16) 443 1.04 (0.95-1.14)
Legumes and pulses e
Q1 11 9613 26 1 reference 26 1 reference 34 1 reference 98 1 reference
Q2 19 9595 13 0.56 (0.28-1.11) 23 0.89 (0.49- 1.61) 28 0.80 (0.47-1.35) 85 0.92 (0.67-1.26)
Q3 28 10408 25 0.88 (0.49-1.58) 29 0.98 (0.55- 1.74) 22 0.51 (0.29-0.90) 92 0.83 (0.61-1.14) f
Q4 39 9903 17 0.65 (0.33-1.31) 23 0.83 (0.45- 1.54) 40 0.99 (0.60-1.62) 89 0.87 (0.63-1.20)
Q5 62 8704 15 0.57 (0.26-1.27) 36 1.42 (0.81- 2.51) 24 0.60 (0.34-1.06) 79 0.83 (0.59-1.18)
p-trend 0.26 p-trend 0.15 p-trend 0.28 p-trend 0.36
continuous, per 25 g/day increment 48223 96 0.81 (0.58-1.13)
137
1.12 (0.93- 1.34) 148 0.96 (0.76-1.20) 443 0.90 (0.78-1.03)
Chapter 8
162
Thesis Jessie Steevens_v04.pdf
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI)
Brassica vegetables e
Q1 1 9569 19 1 reference 35 1 reference 38 1 reference 103 1 reference
Q2 21 10064 22 1.20 (0.63-2.28) 26 0.77 (0.45-1.31) 35 0.87 (0.54-1.40) 74 0.71 (0.52-0.98)
Q3 29 9804 19 1.16 (0.60-2.24) 27 0.78 (0.46-1.33) 25 0.56 (0.34-0.95) 83 0.81 (0.59-1.11)
Q4 39 9879 20 1.26 (0.63-2.52) 22 0.62 (0.34-1.12) 24 0.50 (0.29-0.86) 100 1.02 (0.75-1.40)
Q5 59 8906 16 1.08 (0.47-2.49) 27 0.85 (0.49-1.48) 26 0.51 (0.28-0.92) 83 0.95 (0.66-1.35)
p-trend 0.86 p-trend 0.54 p-trend 0.01 p-trend 0.62
continuous, per 25 g/day increment 48223 96 0.99 (0.70-1.39) 137 1.06 (0.77-1.46) g 148 0.72 (0.54-0.95) 443 1.05 (0.90-1.21)
Allium vegetables e
Q1 5 13046 25 1 reference 49 1 reference 37 1 reference 118 1 reference
Q2 16 6359 11 1.00 (0.48-2.10) 16 0.67 (0.37-1.20) 17 0.92 (0.50-1.68) 76 1.32 (0.96-1.83)
Q3 25 9686 19 1.07 (0.56-2.02) 22 0.64 (0.38-1.10) 35 1.35 (0.83-2.19) 72 0.85 (0.62-1.18)
Q4 38 9920 14 0.80 (0.41-1.58) 23 0.65 (0.38-1.11) 24 0.95 (0.55-1.65) 104 1.25 (0.93-1.68)
Q5 62 9213 27 1.64 (0.88-3.08) 27 0.84 (0.50-1.43) 35 1.55 (0.94-2.56) 73 0.97 (0.69-1.35)
p-trend 0.18 p-trend 0.54 p-trend 0.12 p-trend 0.95
continuous, per 25 g/day increment 48223 96 1.15 (0.90-1.47) g 137 0.96 (0.77-1.20) 148 1.14 (0.97-1.35) 443 0.98 (0.87-1.11)
Leafy vegetables, cooked e
Q1 4 9505 21 1 reference 30 1 reference 29 1 reference 94 1 reference
Q2 12 10208 22 0.87 (0.45-1.68) 31 0.95 (0.56-1.60) 32 1.00 (0.59-1.69) 101 0.95 (0.70-1.29)
Q3 19 9557 19 1.00 (0.53-1.89) 28 0.96 (0.56-1.65) 32 1.13 (0.66-1.94) 79 0.83 (0.60-1.14) g
Q4 27 9803 21 1.07 (0.55-2.08) 24 0.76 (0.42-1.36) 26 0.93 (0.53-1.65) 89 0.92 (0.67-1.27) f
Q5 42 9150 13 0.75 (0.35-1.60) 24 0.83 (0.47-1.46) 29 1.18 (0.66-2.09) 80 0.87 (0.61-1.23)
p-trend 0.66 p-trend 0.40 p-trend 0.65 p-trend 0.45
continuous, per 25 g/day increment 48223 96 0.94 (0.66-1.33)
137 0.89 (0.65-1.22) 148 1.29 (0.96-1.74) 443 0.92 (0.76-1.10)
Leafy vegetables, raw e
Q1 1 9254 24 1 reference 34 1 reference 32 1 reference 101 1 reference
Q2 4 10167 17 0.70 (0.36-1.35) 32 0.89 (0.54-1.47) 27 0.79 (0.46-1.35) 94 0.92 (0.67-1.25)
Q3 7 9573 21 0.87 (0.46-1.67) 29 0.83 (0.49-1.41) 28 0.82 (0.48-1.39) 90 0.93 (0.68-1.28)
Q4 12 9649 22 1.03 (0.56-1.92) 21 0.62 (0.35-1.09) 33 0.97 (0.57-1.63) 76 0.81 (0.58-1.13)
Q5 22 9581 12 0.50 (0.23-1.10) 21 0.63 (0.35-1.13) 28 0.86 (0.48-1.54) 82 0.92 (0.65-1.30)
p-trend 0.17 p-trend 0.08 p-trend 0.92 p-trend 0.59
continuous, per 25 g/day increment 48223 96 0.83 (0.36-1.89) g 137 0.62 (0.33-1.15) 148 0.89 (0.54-1.47) 443 0.86(0.62- 1.18) f
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
163
Thesis Jessie Steevens_v04.pdf
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI)
Total fruit h
Q1 43 9508 33 1 reference 39 1 reference 46 1 reference 113 1 reference
Q2 107 10213 22 0.77 (0.43-1.38) 31 0.89 (0.54-1.47) 26 0.63 (0.38-1.04) 90 0.84 (0.62-1.13)
Q3 155 10123 18 0.72 (0.39-1.33) 26 0.81 (0.48-1.37) 28 0.75 (0.46-1.21) 94 0.95 (0.70-1.28)
Q4 215 10368 13 0.53 (0.27-1.04) 22 0.73 (0.41-1.29) 30 0.86 (0.53-1.40) 83 0.83 (0.61-1.13)
Q5 326 10573 15 0.62 (0.32-1.22) 26 0.97 (0.57-1.67) 26 0.85 (0.50-1.42) 80 0.86 (0.62-1.18)
p-trend 0.11 p-trend 0.77 p-trend 0.82 p-trend 0.39
continuous, per 25 g/day increment 50785 101 0.95 (0.90-1.01)
144
1.00 (0.96-1.05) 156 0.99 (0.95-1.03) 460 1.00 (0.97-1.02)
Citrus fruits i,j
C1 0 5048 18 1 reference 30 1 reference 32 1 reference 61 1 reference
C2 8 11169 19 0.47 (0.24-0.95) 35 0.59 (0.35- 1.00) 47 0.76 (0.47-1.22) 104 0.86 (0.61-1.21)
C3 33 11494 27 0.72 (0.37-1.43) 24 0.46 (0.26- 0.81) 30 0.54 (0.32-0.92) 100 0.89 (0.62-1.27)
C4 77 11490 18 0.51 (0.25-1.04) 29 0.60 (0.35- 1.05) 28 0.55 (0.32-0.94) 109 0.99 (0.70-1.40)
C5 156 11583 19 0.54 (0.27-1.07) 26 0.55 (0.31- 0.98) 19 0.38 (0.21-0.69) 86 0.80 (0.56-1.15)
p-trend 0.38 p-trend 0.37 p-trend 0.003 p-trend 0.46
continuous, per 25 g/day increment 50785 101 1.01 (0.92-1.10)
144
0.97 (0.90-1.04) 156 0.88 (0.81-0.97) 460 0.99 (0.95-1.03)
C, category; CI, confidence interval; RR, Incidence Rate Ratio; Q, quintile. * Adjusted for age (years), sex, cigarette smoking [(current smoking (yes/no), frequency (number
of cigarettes per day), duration (number of years)], alcohol consumption (g ethanol/day), consumption of red meat (g/day), consumption of fish (g/day); a The first two
years of the follow-up were excluded from all analyses, see text; b The number of cases in the vegetable analyses is lower than in the fruit analyses, due to the exclusion
of subjects with inconsistent vegetable data; c Additionally adjusted for total fruit intake; d Tests for dose-response trends were assessed by fitting ordinal exposure
variables as continuous terms in the Cox proportional hazards model; e Additionally adjusted for total fruit intake and all other vegetables; f The proportional hazards
assumption was violated for the exposure variable in this analysis. The interaction between the exposure variable and time was statistically significant; g The proportional
hazards assumption was violated for the exposure variable in this analysis. The interaction between the exposure variable and time was not statistically significant; h
Additionally adjusted for total vegetable intake; i Additionally adjusted for total vegetable intake and all other fruits; j Due to a large number of nonusers of citrus fruits,
categories instead of quintiles were used in analyses: C1 are non-users, C2 to C5 are users, divided into quartiles.
Chapter 8
164
Thesis Jessie Steevens_v04.pdf
Table 8.4 Multivariable adjusted* associations between consumption of individual vegetable and fruit items and risk of esophageal and gastric cancer subtypes;
Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002a
Subcohort Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
person
years at risk
cases
(n) b
RR c(95% CI) cases
(n) b
RR c(95% CI) cases
(n) b
RR c(95% CI) cases
(n) b
RR c(95% CI)
Vegetable item d
String/French beans 48223 96 0.80 (0.50-1.28) 137 1.13 (0.85-1.50) 148 0.91 (0.64-1.31) e 443 0.86 (0.68-1.08)
Cauliflower 48223 96 0.94 (0.52-1.70) 137 0.88 (0.51-1.52) 148 0.58 (0.35-0.96) 443 1.04 (0.77-1.39)
Lettuce 48223 96 0.72 (0.28-1.85) 137 0.51 (0.22-1.17) 148 1.12 (0.58-2.17) 443 0.87 (0.58-1.31) f
Carrots, cooked 48223 96 1.12 (0.57-2.22) 137 0.85 (0.44-1.64) 148 0.72 (0.40-1.31) f 443 0.79 (0.55-1.13)
Endive, cooked 48223 96 0.90 (0.52-1.57) 137 0.92 (0.55-1.54) 148 1.03 (0.70-1.51) 443 0.80 (0.60-1.07)
Brussels sprouts 48223 96 1.29 (0.71-2.34) 137 1.37 (0.69-2.71) 148 1.14 (0.55-2.36) 443 1.24 (0.85-1.80)
Sauerkraut 48223 96 0.56 (0.16-2.00) 137 1.45 (0.58-3.68) 148 0.86 (0.36-2.07) 443 1.03 (0.63-1.69)
Tomatoes 48223 96 0.87 (0.62-1.22) 137 0.79 (0.61-1.01) 148 0.93 (0.71-1.23) 443 1.13 (1.00-1.28)
Onion 48223 96 1.10 (0.79-1.53) 137 0.84 (0.65-1.09) 148 1.21 (0.95-1.53) 443 0.91 (0.78-1.06)
Spinach 48223 96 1.02 (0.51-2.05)e 137 0.80 (0.44-1.45 ) 148 1.77 (1.04-3.02 ) 443 1. 02 (0.73-1.44) e
Beetroot 48223 96 0.90 (0.41-1.96) 137 1.09 (0.60-2.00) 148 1.49 (0.84-2.63) 443 1.13 (0.74-1.75)
Kale 48223 96 0.55 (0.09-3.45) 137 1.12 (0.32-3.92)e 148 1.17 (0.34-4.02) 443 1.69 (0.76-3.75)
Fruit item g
Apples and pears 50785 101 0.94 (0.87-1.03) 144 1.04 (0.99-1.10) 156 1.05 (1.00-1.09) 460 1.00 (0.96-1.03)
Strawberries 50785 101 0.83 (0.40-1.73) 144 0.86 (0.47-1.57) 156 1.06 (0.60-1.85) f 460 0.96 (0.69-1.32)
Oranges and fresh orange juice 50785 101 1.01 (0.91-1.13) 144 0.95 (0.87-1.04) 156 0.86 (0.77-0.95) 460 0.99 (0.95-1.04)
CI, confidence interval; RR, Incidence Rate Ratio. * Adjusted for age (years), sex, cigarette smoking [(current smoking (yes/no), frequency (number of cigarettes/day),
duration (number of years)], alcohol consumption (g ethanol/day), consumption of red meat (g/day), consumption of fish (g/day); a The first two years of the follow-up
were excluded from all analyses, see text; b The number of cases in the vegetable analyses is lower than in the fruit analyses, due to the exclusion of subjects with
inconsistent vegetable data; c Continuous variables, RR per 25 g/day increments; d Additionally adjusted for total fruit intake. The individual vegetable items were
adjusted for all other vegetables consumed; e The proportional hazards assumption was violated for the exposure variable in this analysis. The interaction between the
exposure variable and time was statistically significant; f The proportional hazards assumption was violated for the exposure variable in this analysis. The interaction
between the exposure variable and time was not statistically significant; g Additionally adjusted for total vegetable intake. The individual fruit items were adjusted for all
other fruits consumed.
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
165
Thesis Jessie Steevens_v04.pdf
Table 8.5 Multivariable adjusted* associations between consumption of individual vegetable and fruit items and risk of esophageal and gastric cancer subtypes,
stratified by sex and cigarette smoking status; Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002a
Subcohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
person
years at risk
cases
(n) bRR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
Total vegetables c, RRs per 25 g/day increments
Sex
Men 22621 54 0.90 (0.80-1.00) 106 0.99 (0.91-1.06) 125 0.99 (0.92-1.07) 294 0.96 (0.92-1.01)
Women 25601 42 1.03 (0.95-1.12) 31 0.86 (0.75-0.97) 23 1.04 (0.93-1.16) 149 1.01 (0.95-1.07)
p-interaction d0.04 p-interaction 0.06 p-interaction0.53 p-interaction0.24
Cigarette smoking status
Never smoker 18728 22 1.08 (0.98-1.19) 26 0.97 (0.84-1.13) 24 1.11 (0.99-1.25) 105 1.01 (0.94-1.09)
Former smoker 17369 28 0.96 (0.83-1.11) 68 1.02 (0.93-1.11) 72 0.95 (0.86-1.05) 180 0.98 (0.92-1.04)
Current smoker 12126 46 0.90 (0.81-0.99) 43 0.85 (0.75-0.97) 52 0.99 (0.90-1.10) 158 0.96 (0.90-1.01)
p-interaction 0.03 p-interaction 0.08 p-interaction 0.12 p-interaction 0.51
Total fruit e, RRs per 25 g/day increments
Sex
Men 23722 55 0.91 (0.83-1.00) 112 1.00 (0.96-1.05) 132 0.96 (0.92-1.01) 304 0.99 (0.96-1.02)
Women 27062 46 0.98 (0.91-1.05) 32 0.98 (0.91-1.06) 24 1.08 (1.02-1.14) 156 1.00 (0.97-1.04)
p-interaction 0.24 p-interaction 0.65 p-interaction 0.003 p-interaction 0.47
Cigarette smoking status
Never smoker 19774 23 1.01 (0.94-1.08) 28 0.99 (0.92-1.07) 27 1.02 (0.95-1.10) 110 0.99 (0.95-1.03)
Former smoker 18099 30 0.94 (0.85-1.03) 70 1.03 (0.97-1.08) 74 0.99 (0.95-1.04) 186 1.01 (0.97-1.04)
Current smoker 12912 48 0.91 (0.82-1.01) 46 0.93 (0.86-1.01) 55 0.95 (0.87-1.04) 164 0.98 (0.93-1.03)
p-interaction 0.25 p-interaction 0.15 p-interaction 0.49 p-interaction 0.63
CI, confidence interval; RR, Incidence Rate Ratio. * Adjusted for age (years), sex, cigarette smoking [(current smoking (yes/no), frequency (number of cigarettes per day),
duration (number of years)], alcohol consumption (g ethanol/day), consumption of red meat (g/day), consumption of fish (g/day); a The first two years of the follow-up
were excluded from all analyses, see text; b The number of cases in the vegetable analyses is lower than in the fruit analyses, due to the exclusion of subjects with
inconsistent vegetable data; c Additionally adjusted for total fruit intake; d p-values for interaction were calculated by including a cross-product term in the Cox
proportional hazards model; e Additionally adjusted for total vegetable intake.
Chapter 8
166
Interaction analyses
Results of analyses stratified by sex and smoking status are presented in Table 8.5
(multivariable adjusted) and Supplemental Table 8.3 (age- and sex-adjusted).
Adjustment for confounding variables did not change the RRs.
Total vegetables consumption was associated with risk of ESCC only in men (RR
per 25 g/day increment: 0.90, 95% CI 0.80-1.00, p-interaction=0.04), whereas for EAC
an inverse association was observed in women only (RR per 25 g/day: 0.86, 95% CI
0.75-0.97, p-interaction=0.06). For GCA and GNCA, no clear sex differences were
observed. In current cigarette smokers only, we found significant inverse associations
between vegetables consumption and risk of ESCC (RR per 25 g/day: 0.90, 95% CI
0.81-0.99, p-interaction=0.03) and EAC (RR per 25 g/day: 0.85, 95% CI 0.75-0.97,
p-interaction=0.08). We found no interaction between vegetables consumption and
smoking for GCA and GNCA.
An inverse association between consumption of fruits and ESCC was present in
women (RR per 25 g/day: 0.91, 95% CI 0.83-1.00), but the interaction with sex was not
statistically significant (p-interaction=0.24). For women, a positive association between
fruits consumption and GCA risk was observed (RR per 25 g/day: 1.08, 95% CI
1.02-1.14, p-interaction=0.003). No significant interactions between smoking status
and fruits consumption were observed, although for ESCC, EAC, and GCA the risk
estimates were further below one for current smokers than for never and former
smokers.
DISCUSSION
In the Netherlands Cohort Study, we found mostly inverse associations between
vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes.
Most RRs were not statistically significant. However, our extensive questions on
specific vegetables and fruits allowed us to study these foods in more detail. We
identified significant inverse associations for raw vegetables and EAC risk, Brassicas
and GCA risk, and citrus fruits and EAC and GCA risk. Associations were generally
stronger inverse in current smokers, but the interaction tests were not always
statistically significant.
The current evidence from case-control and cohort studies does not clearly point
to any specific group of vegetables that is responsible for the observed inverse
associations. For ESCC, significant inverse associations have been observed with
Cruciferous vegetables in one cohort study,15 and with cooked vegetables,35 raw
vegetables,7,36 tomatoes37 and spinach37 in case-control studies. The few studies on
EAC found significant inverse associations with raw and cooked spinach14, dark green
vegetables,7 dark yellow vegetables,7,38 and raw vegetables.7 Only one study found
significant inverse associations with GCA; for dark green vegetables and beetroot.39
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
167
Beetroot, carrots, Allium vegetables,39 dark green, light green, and yellow vegetables40
have shown significant inverse associations with GNCA in two studies. This diversity of
vegetables for which significant inverse associations have been observed may have to
do with the fact that vegetable consumption habits differ between continents,
countries, and populations with respect to the amount and types eaten.
For fruit groups, the evidence is more consistent than for vegetable groups. Citrus
fruits have been shown to be inversely associated with risk of ESCC,14,35,36 EAC,38 GCA,18
and GNCA18,39 by a considerable number of studies. Some other fruits that showed
significantly lower risks were Rosaceae for ESCC,14 fruit juices for GCA and GNCA,39 and
apples and pears for GNCA.39
Our data suggest that for current smokers, decreased risks of EAC and ESCC
associated with vegetables and fruits consumption were more apparent than for never
and former smokers. This finding is consistent with three case-control studies,6,8,36
which found inverse associations between vegetables consumption and ESCC only6 or
most strongly8,36 in smokers. Cohort studies found no interactions for ESCC,14,15 EAC,14
GCA,19 or GNCA.19 Smoking causes oxidative stress and DNA damage in the body.41
Substances in vegetables and fruits that have anticarcinogenic properties (e.g. vitamin
C, carotenoids, flavonoids, folate) could counteract these effects. Smokers could
therefore benefit most from consuming vegetables and fruits.
Case-control studies have generally found stronger and more often statistically
significant associations compared with cohort studies. An explanation may be reversed
causation. Individuals who are diagnosed with gastrointestinal cancer may already
have experienced intestinal complaints quite some time before the diagnosis. This may
have caused them to change their dietary habits, including consumption of vegetables
and fruits, leading to biased measurement of the diet as it was before the onset of the
disease.13 Asking about dietary habits long before the diagnosis will not solve this
problem, as reporting past diet is influenced by current diet.42 Cohort studies have
fewer problems with reversed causation thanks to their prospective nature.
Additionally, we minimized the influence of reversed causation on our results by
excluding early cases. On the other hand, measurement of vegetables and fruits in
cohort studies is not perfect either. It is very difficult for persons to accurately report
their intake, specifically of vegetables. This is reflected in the relatively low correlations
in validation studies.43 A consequence of these low correlations is attenuation of risk
estimates. Thus, the true associations may have been stronger than we observed,
strengthening our conclusion of protective effects of some vegetables and fruits on
esophageal and gastric cancers. Besides the attenuation of RRs, another explanation
for the nonsignificance of the results is the limited power. Even though one strength of
the NLCS is its large size, the number of ESCC, EAC, and GCA cases is limited. The large
size of the study, combined with a long follow-up did allow for separate analysis of
ESCC vs. EAC and GCA vs. GNCA. A strength of our study was the ability to check for
confounding by several important risk factors (see methods), and to make adjustments
if necessary. One factor that we did not measure and may have confounded the
Thesis Jessie Steevens_v04.pdf
Chapter 8
168
associations with GNCA in particular is infection with Helicobacter pylori, which is a risk
factor for GNCA and may protect against GCA. We estimated that half of our study
population might have been infected.44,45 It is not clear whether adjustment for
Helicobacter pylori infection would have changed our results. One study that made this
adjustment,46 found no effect on the estimates. Two studies that looked into
interaction with Helicobacter pylori found no evidence of this.18,46
In conclusion, we observed significant inverse associations between some specific
vegetables and fruits and risk of ESCC, EAC, GCA, and GNCA. Most other observed
associations were inverse, but not statistically significant. Consumption of fruits and
vegetables may specifically help protect smokers from developing ESCC, EAC, and GCA.
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
169
APPENDIX
Composition of vegetable and fruit groups
Food group Composition
Total vegetables Cooked vegetables plus raw vegetables
Cooked vegetables Beetroot, broad beans, Brussels sprouts, cauliflower, cabbage
(white/green), cooked carrots, cooked endive, kale, leek, mushrooms,
onions, rhubarb, sauerkraut, spinach, string beans, sweet peppers, and
other cooked vegetables originating from an open-ended question on
frequently consumed items not listed in the questionnaire.
Raw vegetables Gherkins, lettuce, raw carrots, raw endive, tomatoes, and other raw
vegetables originating from an open ended question on frequently
consumed items not listed in the questionnaire
Brassica vegetables Brussels sprouts, cabbage (white/green), cauliflower, kale
Leafy vegetables, cooked Cooked endive, spinach
Leafy vegetables, raw Lettuce, raw endive
Legumes and pulses Broad beans, dried pulses, string beans
Allium vegetables Leek, onions
Total fruit Apples/pears, bananas, grapefruits and fresh grapefruit juice, grapes,
mandarins, oranges and fresh orange juice, raisins/other dried fruit,
strawberries, and other fruits originating from an open ended question on
frequently consumed items not listed in the questionnaire.
Citrus fruits Fresh lemon juice, grapefruits and fresh grapefruit juice, mandarins,
oranges and fresh orange juice
Thesis Jessie Steevens_v04.pdf
Chapter 8
170
Thesis Jessie Steevens_v04.pdf
Supplemental Table 8.1 Age- and sex-adjusted associations between vegetables and fruit consumption and risk of esophageal and gastric cancer subtypes;
Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002 a.
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI) cases
(n) b
RR (95% CI)
Total vegetables
Q1 104 9073 31 1 reference 35 1 reference 43 1 reference 111 1 reference
Q2 146 9829 22 0.71 (0.38- 1.35) 30 0.71 (0.42-1.19) 26 0.62 (0.37-1.05) 86 0.78 (0.57-1.06)
Q3 181 9662 17 0.54 (0.27- 1.07) 25 0.62 (0.36-1.07) 27 0.62 (0.37-1.06) 91 0.87 (0.64-1.17)
Q4 222 9835 12 1.24 (0.71- 2.17) 21 0.85 (0.52-1.41) 27 0.87 (0.54-1.42) 79 0.78 (0.57-1.07)
Q5 297 9823 14 0.69 (0.35- 1.36) 26 0.66 (0.38-1.16) 25 0.88 (0.53-1.45) 76 0.88 (0.64-1.21)
p-trend
c 0.83 p-trend 0.31 p-trend 0.91 p-trend 0.53
continuous, per 25 g/day increment 48223 96 0.99 (0.92- 1.06) 137 0.96 (0.90-1.03) 148 1.00 (0.94-1.07) 443 0.98 (0.94-1.01)
Cooked vegetables
Q1 80 9495 22 1 reference 36 1 reference 33 1 reference 99 1 reference
Q2 115 9841 14 0.62 (0.31- 1.23) 28 0.75 (0.45-1.25) 25 0.73 (0.43-1.25) 90 0.89 (0.65-1.20)
Q3 144 10269 17 0.73 (0.38- 1.38) 21 0.54 (0.31-0.94) 25 0.71 (0.41-1.21) 88 0.84 (0.61-1.14)
Q4 177 10012 24 1.08 (0.60- 1.95) 21 0.55 (0.32-0.95) 36 1.03 (0.63-1.68) 93 0.92 (0.68-1.25)
Q5 239 8606 19 0.97 (0.52- 1.81) 31 0.93 (0.57-1.53) 29 0.95 (0.57-1.59) 73 0.82 (0.59-1.13)
p-trend 0.56 p-trend 0.73 p-trend 0.69 p-trend 0.33
continuous, per 25 g/day increment 48223 96 1.00 (0.92- 1.08) 137 0.99 (0.91-1.07) 148 1.01 (0.94-1.09) 443 0.97 (0.93-1.01)
Raw vegetables
Q1 8 9426 26 1 reference 31 1 reference 38 1 reference 95 1 reference
Q2 22 9550 17 0.64 (0.34- 1.18) 37 1.17 (0.72-1.92) 29 0.76 (0.46-1.25) 79 0.82 (0.59-1.12)
Q3 34 9805 19 0.72 (0.40- 1.30) 24 0.82 (0.47-1.42) 24 0.70 (0.41-1.18) 97 1.05 (0.78-1.43)
Q4 48 9968 17 0.63 (0.34- 1.17) 30 1.00 (0.60-1.69) 29 0.82 (0.50-1.35) 88 0.93 (0.68-1.27)
Q5 77 9473 17 0.69 (0.37- 1.29) 15 0.55 (0.29-1.05) 28 0.88 (0.54-1.46) 84 0.99 (0.72-1.35)
p-trend 0.33 p-trend 0.04 p-trend 0.82 p-trend 0.80
continuous, per 25 g/day increment 48223 96 0.93 (0.74- 1.16) 137 0.83 (0.70-0.98) 148 0.98 (0.85-1.14) 443 1.00 (0.91-1.09)
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
171
Thesis Jessie Steevens_v04.pdf
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) bRR (95% CI)
Legumes and pulses
Q1 11 9613 26 1 reference 26 1 reference 34 1 reference 98 1 reference
Q2 19 9595 13 0.53 (0.27- 1.04) 23 0.86 (0.48-1.54) 28 0.79 (0.47-1.33) 85 0.90 (0.66-1.23)
Q3 28 10408 25 0.88 (0.50- 1.56) 29 0.91 (0.53-1.58) 22 0.51 (0.30-0.90) 92 0.82 (0.61-1.12) d
Q4 39 9903 17 0.64 (0.34- 1.21) 23 0.78 (0.43-1.38) 40 1.00 (0.62-1.61) 89 0.85 (0.63-1.16)
Q5 62 8704 15 0.63 (0.33- 1.22) 36 1.32 (0.78-2.23) 24 0.65 (0.38-1.11) 79 0.84 (0.61-1.16)
p-trend 0.30 p-trend 0.22 p-trend 0.35 p-trend 0.34
continuous, per 25 g/day increment 48223 96 0.90 (0.69- 1.18) 137 1.09 (0.92-1.31) 148 0.97 (0.78-1.20) 443 0.91 (0.80-1.03)
Brassica vegetables
Q1 11 9569 19 1 reference 35 1 reference 38 1 reference 103 1 reference
Q2 21 10064 22 1.11 (0.60- 2.07) 26 0.72 (0.43-1.22) 35 0.90 (0.56-1.46) 74 0.69 (0.51-0.96)
Q3 29 9804 19 0.98 (0.51- 1.86) 27 0.74 (0.44-1.24) 25 0.63 (0.37-1.05) 83 0.78 (0.57-1.06)
Q4 39 9879 20 1.04 (0.55- 1.97) 22 0.59 (0.34-1.03) 24 0.59 (0.35-1.01) 100 0.94 (0.70-1.27)
Q5 59 8906 16 0.91 (0.46- 1.79) 27 0.79 (0.47-1.32) 26 0.69 (0.41-1.16) 83 0.85 (0.62-1.15)
p-trend 0.71 p-trend 0.40 p-trend 0.10 p-trend 0.88
continuous, per 25 g/day increment 48223 96 0.93 (0.71- 1.23) 137 1.00 (0.75-1.33) § 148 0.84 (0.65-1.08) 443 1.00 (0.88-1.13)
Allium vegetables
Q1 5 13046 25 1 reference 49 1 reference 37 1 reference 118 1 reference
Q2 16 6359 11 0.92 (0.45- 1.89) 16 0.66 (0.37-1.17) 17 0.92 (0.51-1.66) 76 1.34 (0.98-1.83)
Q3 25 9686 19 1.04 (0.57- 1.91) 22 0.61 (0.37-1.03) 35 1.30 (0.81-2.09) 72 0.84 (0.62-1.15)
Q4 38 9920 14 0.78 (0.40- 1.51) 23 0.66 (0.40-1.10) 24 0.92 (0.54-1.56) 104 1.25 (0.94-1.66)
Q5 62 9213 27 1.61 (0.93- 2.79) 27 0.81 (0.50-1.32) 35 1.42 (0.88-2.28) 73 0.94 (0.69-1.28)
p-trend 0.13 p-trend 0.44 p-trend 0.20 p-trend 0.75
continuous, per 25 g/day increment 48223 96 1.20 (0.98- 1.46) 137 0.94 (0.77-1.16) 148 1.10 (0.95-1.29) 443 0.97 (0.87-1.08)
Leafy vegetables, cooked
Q1 4 9505 21 1 reference 30 1 reference 29 1 reference 94 1 reference
Q2 12 10208 22 0.93 (0.51- 1.71) 31 0.90 (0.54- 1.50) 32 0.95 (0.56-1.59) 101 0.93 (0.69-1.26)
Q3 19 9557 19 0.89 (0.47- 1.66) 28 0.93 (0.55- 1.57) 32 1.09 (0.65-1.83) 79 0.81 (0.59-1.12) d
Q4 27 9803 21 0.96 (0.52- 1.78) 24 0.75 (0.43- 1.31) 26 0.84 (0.49-1.45) 89 0.90 (0.66-1.23) e
Q5 42 9150 13 0.63 (0.31- 1.27) 24 0.81 (0.47- 1.40) 29 1.01 (0.59-1.71) 80 0.86 (0.62-1.18)
p-trend 0.23 p-trend 0.38 p-trend 0.93 p-trend 0.37
continuous, per 25 g/day increment 48223 96 0.82 (0.60- 1.13) 137 0.90 (0.68- 1.19) 148 1.13 (0.85-1.50) 443 0.91 (0.77-1.07) d
Chapter 8
172
Thesis Jessie Steevens_v04.pdf
Sub-
cohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
categorical
median
(g/day)
person
years at
risk
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) bRR (95% CI)
Leafy vegetables, raw
Q1 1 9254 24 1 reference 34 1 reference 32 1 reference 101 1 reference
Q2 4 10167 17 0.68 (0.36- 1.29) 32 0.87 (0.53-1.43) 27 0.79 (0.46-1.34) 94 0.91 (0.67-1.23)
Q3 7 9573 21 0.88 (0.49- 1.60) 29 0.80 (0.48-1.34) 28 0.82 (0.49-1.39) 90 0.89 (0.66-1.21)
Q4 12 9649 22 0.92 (0.51- 1.66) 21 0.59 (0.34-1.02) 33 0.99 (0.59-1.63) 76 0.75 (0.55-1.04)
Q5 22 9581 12 0.51 (0.25- 1.04) 21 0.60 (0.34-1.05) 28 0.86 (0.50-1.63) 82 0.84 (0.62-1.15)
p-trend 0.13 p-trend 0.05 p-trend 0.88 p-trend 0.24
continuous, per 25 g/day increment 48223 96 0.77 (0.36- 1.63) 137 0.57 (0.32-1.03) 148 0.90 (0.57-1.43) 443 0.79 (0.59-1.05)
Total fruit
Q1 43 9508 33 1 reference 39 1 reference 46 1 reference 113 1 reference
Q2 107 10213 22 0.62 (0.36- 1.08) 31 0.84 (0.51-1.36) 26 0.61 (0.37-1.01) 90 0.80 (0.59-1.07)
Q3 155 10123 18 0.51 (0.29- 0.92) 26 0.72 (0.44-1.21) 28 0.69 (0.42-1.12) 94 0.85 (0.63-1.14)
Q4 215 10368 13 0.35 (0.19- 0.67) 22 0.65 (0.38-1.11) 30 0.79 (0.49-1.27) 83 0.74 (0.55-1.00)
Q5 326 10573 15 0.40 (0.21- 0.74) 26 0.82 (0.49-1.38) 26 0.76 (0.46-1.25) 80 0.73 (0.54-1.00)
p-trend 0.002 p-trend 0.36 p-trend 0.50 p-trend 0.06
continuous, per 25 g/day increment 50785 101 0.92 (0.86- 0.97) 144 0.99 (0.95-1.03) 156 0.98 (0.94-1.02) 460 0.98 (0.96-1.01)
Citrus fruits f
C1 0 5048 18 1 reference 30 1 reference 32 1 reference 61 1 reference
C2 8 11169 19 0.48 (0.25- 0.93) 35 0.57 (0.34-0.95) 47 0.74 (0.46-1.17) 104 0.81 (0.58-1.14)
C3 33 11494 27 0.67 (0.36- 1.24) 24 0.43 (0.24-0.75) 30 0.53 (0.31-0.88) 100 0.81 (0.57-1.14)
C4 77 11490 18 0.43 (0.22- 0.85) 29 0.55 (0.32-0.95) 28 0.53 (0.31-0.90) 109 0.89 (0.63-1.25)
C5 156 11583 19 0.46 (0.23- 0.89) 26 0.52 (0.30-0.91) 19 0.39 (0.22-0.69) 86 0.72 (0.51-1.03)
p-trend 0.13 p-trend 0.29 p-trend 0.004 p-trend 0.23
continuous, per 25 g/day increment 50785 101 0.98 (0.89- 1.07) 144 0.96 (0.89-1.04) 156 0.89 (0.81-0.97) 460 0.98 (0.95-1.02)
CI, confidence interval; RR, Incidence Rate Ratio; Q, quintile; T, tertile. a The first two years of the follow-up were excluded from all analyses, see text; b The number of
cases in the vegetable analyses is lower than in the fruit analyses, due to the exclusion of subjects with inconsistent vegetable data; c Tests for dose-response trends were
assessed by fitting ordinal exposure variables as continuous terms in the Cox proportional hazards model; d The proportional hazards assumption was violated for the
exposure variable in this analysis. The interaction between the exposure variable and time was statistically significant; e The proportional hazards assumption was
violated for the exposure variable in this analysis. The interaction between the exposure variable and time was not statistically significant; f Due to a large number of
nonusers of citrus fruits, categories instead of quintiles were used in analyses: C1 are non-users, C2 to C5 are users, divided into quartiles.
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
173
Thesis Jessie Steevens_v04.pdf
Supplemental Table 8.2 Age- and sex-adjusted associations between consumption of individual vegetable and fruit items and risk of esophageal and gastric cancer
subtypes; Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002 a.
Subcohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
person years
at risk
cases
(n) b RR c (95% CI)
cases
(n) bRR c (95% CI)
cases
(n) bRR c (95% CI)
cases
(n) bRR c (95% CI)
Vegetable item
String/French beans 48223 96 0.91 (0.62-1.32) 137 1.14 (0.87-1.49) 148 0.84 (0.58-1.21)d 443 0.87 (0.71-1.06)
Cauliflower 48223 96 0.87 (0.53-1.44) 137 0.95 (0.56-1.61) 148 0.65 (0.42-1.02) 443 1.01 (0.80-1.27)
Lettuce 48223 96 0.74 (0.31-1.78) 137 0.46 (0.21-1.02) 148 0.95 (0.52-1.71) 443 0.78 (0.54-1.14)
Carrots, cooked 48223 96 0.63 (0.30-1.33) 137 0.85 (0.45-1.60) 148 0.69 (0.36-1.34) e 443 0.72 (0.53-0.99)
Endive, cooked 48223 96 0.75 (0.48-1.17) 137 0.92 (0.61-1.38) 148 1.05 (0.72-1.54) 443 0.82 (0.64-1.06)
Brussels sprouts 48223 96 1.24 (0.71-2.17) 137 1.29 (0.63-2.66) 148 0.95 (0.49-1.83) 443 1.17 (0.81-1.68)
Sauerkraut 48223 96 0.61 (0.21-1.79) 137 1.31 (0.58-2.97) 148 0.87 (0.39-1.96) 443 0.99 (0.64-1.52)
Tomatoes 48223 96 0.93 (0.69-1.25) 137 0.76 (0.60-0.96) 148 0.96 (0.75-1.24) 443 1.07 (0.95-1.21)
Onion 48223 96 1.24 (0.95-1.61) 137 0.86 (0.67-1.11) 148 1.19 (0.97-1.47) 443 0.92 (0.80-1.06)
Spinach 48223 96 0.81 (0.45-1.45) d 137 0.80 (0.45-1.40) 148 1.37 (0.84-2.24) 443 0.98 (0.73-1.30) d
Beetroot 48223 96 0.50 (0.22-1.13) 137 1.03 (0.58-1.81) 148 1.25 (0.79-1.98) 443 0.91 (0.62-1.33)
Kale 48223 96 0.25 (0.04-1.62) 137 0.97 (0.28-3.38) d 148 1.12 (0.32-3.92) 443 1.28 (0.60-2.72) e
Fruit item
Apples and pears 50785 101 0.88 (0.80-0.96) 144 1.02 (0.96-1.08) 156 1.02 (0.98-1.07) 460 0.98 (0.94-1.01)
Strawberries 50785 101 0.66 (0.30-1.49) 144 0.77 (0.42-1.42) 156 0.95 (0.54-1.67) e 460 0.90 (0.65-1.24)
Oranges and fresh orange juice 50785 101 0.97 (0.87-1.09) 144 0.94 (0.86-1.03) 156 0.86 (0.77-0.95 460 0.98 (0.94-1.03
CI, confidence interval; RR, Incidence Rate Ratio. a The first two years of the follow-up were excluded from all analyses, see text; b The number of cases in the vegetable
analyses is lower than in the fruit analyses, due to the exclusion of subjects with inconsistent vegetable data; c Continuous variables, RR per 25 g/day increments; d The
proportional hazards assumption was violated for the exposure variable in this analysis. The interaction between the exposure variable and time was statistically
significant; e The proportional hazards assumption was violated for the exposure variable in this analysis. The interaction between the exposure variable and time was not
statistically significant.
Chapter 8
174
Thesis Jessie Steevens_v04.pdf
Supplemental Table 8.3 Age- and sex-adjusted associations between consumption of vegetables and fruits and risk of esophageal and gastric cancer subtypes,
stratified by sex and cigarette smoking status; Netherlands Cohort Study on diet and cancer (NLCS) 1986-2002a.
Subcohort
Esophageal
squamous cell carcinoma
Esophageal
adenocarcinoma
Gastric cardia
adenocarcinoma
Gastric noncardia
adenocarcinoma
person years
at risk
cases
(n) bRR (95% CI)
cases
(n) b RR (95% CI)
cases
(n) bRR (95% CI)
cases
(n) bRR (95% CI)
Total vegetables, RRs per 25 g/day increments
Sex
Men 22621 54 0.94 (0.84-1.05) 106 0.99 (0.92-1.07) 125 0.99 (0.93-1.07) 294 0.96 (0.92- 1.01)
Women 25601 42 1.04 (0.96-1.13) 31 0.86 (0.76-0.98) 23 1.04 (0.93-1.17) 149 1.01 (0.95- 1.08)
p-interaction c 0.15 p-interaction 0.06 p-interaction 0.48 p-interaction 0,21
Cigarette smoking status
Never smoker 18728 22 1.08 (0.98-1.19) 26 0.98 (0.85-1.13) 24 1.12 (1.00-1.25) 105 1.01 (0.94- 1.09)
Former smoker 17369 28 0.99 (0.86-1.13) 68 1.03 (0.94-1.12) 72 0.96 (0.87-1.05) 180 0.98 (0.93- 1.04)
Current smoker 12126 46 0.95 (0.85-1.06) 43 0.85 (0.74-0.97) 52 1.00 (0.90-1.10) 158 0.96 (0.91- 1.01)
p-interaction 0.18 p-interaction 0.06 p-interaction 0.12 p-interaction 0.54
Total fruit, RRs per 25 g/day increments
Sex
Men 23722 55 0.87 (0.79-0.96) 112 1.00 (0.95-1.05) 132 0.96 (0.91-1.00) 304 0.98 (0.95- 1.01)
Women 27062 46 0.96 (0.89-1.03) 32 0.98 (0.90-1.06) 24 1.07 (1.01-1.14) 156 1.00 (0.96- 1.04)
p-interaction 0.11 p-interaction 0.67 p-interaction 0.003 p-interaction 0,45
Cigarette smoking status
Never smoker 19774 23 1.01 (0.94-1.08) 28 1.00 (0.92-1.07) 27 1.02 (0.95-1.10) 110 0.99 (0.95- 1.03)
Former smoker 18099 30 0.92 (0.83-1.02) 70 1.02 (0.97-1.08) 74 0.99 (0.94-1.04) 186 1.00 (0.97- 1.04)
Current smoker 12912 48 0.89 (0.80-0.99) 46 0.93 (0.86-1.01) 55 0.95 (0.87-1.04) 164 0.98 (0.93- 1.03)
p-interaction 0.12 p-interaction 0.16 p-interaction 0.46 p-interaction 0.66
CI, confidence interval; RR, Incidence Rate Ratio. a The first two years of the follow-up were excluded from all analyses, see text; b The number of cases in the vegetable
analyses is lower than in the fruit analyses, due to the exclusion of subjects with inconsistent vegetable data; c p-values for interaction were calculated by including a
cross-product term in the Cox proportional hazards model.
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
175
REFERENCES
1. McCullough ML, Giovannucci EL. Diet and cancer prevention. Oncogene 2004;23:6349-6364.
2. World Cancer Research Fund, American Institute for Cancer Research. Food, nutrition, physical activity
and the prevention of cancer: a global perspective. AICR, 2007.
3. Steevens J, Botterweck AA, Dirx MJ, van den Brandt PA, Schouten LJ. Trends in incidence of
oesophageal and stomach cancer subtypes in Europe. Eur J Gastroenterol Hepatol 2010;22:669-678.
4. Kamangar F, Chow WH, Abnet CC, Dawsey SM. Environmental causes of esophageal cancer.
Gastroenterol Clin North Am 2009;38:27-57, vii.
5. Forman D, Burley VJ. Gastric cancer: global pattern of the disease and an overview of environmental
risk factors. Best Pract Res Clin Gastroenterol 2006;20:633-649.
6. Yang CX, Wang HY, Wang ZM, Du HZ, Tao DM, Mu XY, Chen HG, Lei Y, Matsuo K, Tajima K. Risk factors
for esophageal cancer: a case-control study in South-western China. Asian Pac J Cancer Prev
2005;6:48-53.
7. Navarro Silvera SA, Mayne ST, Risch H, Gammon MD, Vaughan TL, Chow WH, Dubrow R, Schoenberg
JB, Stanford JL, West AB, Rotterdam H, Blot WJ, Fraumeni JF, Jr. Food group intake and risk of subtypes
of esophageal and gastric cancer. Int J Cancer 2008;123:852-860.
8. Terry P, Lagergren J, Hansen H, Wolk A, Nyren O. Fruit and vegetable consumption in the prevention of
oesophageal and cardia cancers. Eur J Cancer Prev 2001;10:365-369.
9. Hung HC, Huang MC, Lee JM, Wu DC, Hsu HK, Wu MT. Association between diet and esophageal
cancer in Taiwan. J Gastroenterol Hepatol 2004;19:632-637.
10. Anderson LA, Watson RG, Murphy SJ, Johnston BT, Comber H, Mc Guigan J, Reynolds JV, Murray LJ.
Risk factors for Barrett's oesophagus and oesophageal adenocarcinoma: results from the FINBAR
study. World J Gastroenterol 2007;13:1585-1594.
11. Cheng KK, Sharp L, McKinney PA, Logan RF, Chilvers CE, Cook-Mozaffari P, Ahmed A, Day NE. A case-
control study of oesophageal adenocarcinoma in women: a preventable disease. Br J Cancer
2000;83:127-132.
12. Lunet N, Valbuena C, Vieira AL, Lopes C, Lopes C, David L, Carneiro F, Barros H. Fruit and vegetable
consumption and gastric cancer by location and histological type: case-control and meta-analysis.
European Journal of Cancer Prevention 2007;16:312-327.
13. Botterweck AA, van den Brandt PA, Goldbohm RA. A prospective cohort study on vegetable and fruit
consumption and stomach cancer risk in The Netherlands. Am J Epidemiol 1998;148:842-853.
14. Freedman ND, Park Y, Subar AF, Hollenbeck AR, Leitzmann MF, Schatzkin A, Abnet CC. Fruit and
vegetable intake and esophageal cancer in a large prospective cohort study. Int J Cancer
2007;121:2753-2760.
15. Yamaji T, Inoue M, Sasazuki S, Iwasaki M, Kurahashi N, Shimazu T, Tsugane S. Fruit and vegetable
consumption and squamous cell carcinoma of the esophagus in Japan: the JPHC study. Int J Cancer
2008;123:1935-1940.
16. Tran GD, Sun XD, Abnet CC, Fan JH, Dawsey SM, Dong ZW, Mark SD, Qiao YL, Taylor PR. Prospective
study of risk factors for esophageal and gastric cancers in the Linxian general population trial cohort in
China. Int J Cancer 2005;113:456-463.
17. Guo W, Blot WJ, Li JY, Taylor PR, Liu BQ, Wang W, Wu YP, Zheng W, Dawsey SM, Li B, et al. A nested
case-control study of oesophageal and stomach cancers in the Linxian nutrition intervention trial. Int J
Epidemiol 1994;23:444-450.
18. Gonzalez CA, Pera G, Agudo A, Bueno-De-Mesquita HB, Ceroti M, Boeing H, Schulz M, Del Giudice G,
Plebani M, Carneiro F, Berrino F, Sacerdotde C, Tumino R, Panico S, Berglund G, Siman H, Hallmans G,
Stenling R, Martinez C, Dorronsoro M, Barricarte A, Navarro C, Quiros JR, Allen N, Key TJ, Bingham S,
Day NE, Linseisen J, Nagel G, Overvad K, Jensen MK, Olsen A, Tjonneland A, Buchner FL, Peeters PH,
Numans ME, Clavel-Chapelon F, Boutron-Ruault MC, Roukos D, Trichopolou A, Psaltopoulou T, Lund E,
Casagrande C, Slimani N, Jenab M, Riboli E. Fruit and vegetable intake and the risk of stomach and
oesophagus adenocarcinoma in the European Prospective Investigation into Cancer and Nutrition
(EPIC-EURGAST). International Journal of Cancer 2006;118:2559-2566.
Thesis Jessie Steevens_v04.pdf
Chapter 8
176
19. Freedman ND, Subar AF, Hollenbeck AR, Leitzmann MF, Schatzkin A, Abnet CC. Fruit and vegetable
intake and gastric cancer risk in a large United States prospective cohort study. Cancer Causes Control
2008;19:459-467.
20. Nouraie M, Pietinen P, Kamangar F, Dawsey SM, Abnet CC, Albanes D, Virtamo J, Taylor PR. Fruits,
vegetables, and antioxidants and risk of gastric cancer among male smokers. Cancer Epidemiol
Biomarkers Prev 2005;14:2087-2092.
21. Kobayashi M, Tsubono Y, Sasazuki S, Sasaki S, Tsugane S. Vegetables, fruit and risk of gastric cancer in
Japan: a 10-year follow-up of the JPHC Study Cohort I. Int J Cancer 2002;102:39-44.
22. Dorant E, van den Brandt PA, Goldbohm RA, Sturmans F. Consumption of onions and a reduced risk of
stomach carcinoma. Gastroenterology 1996;110:12-20.
23. van den Brandt PA, Goldbohm RA, van 't Veer P, Volovics A, Hermus RJ, Sturmans F. A large-scale
prospective cohort study on diet and cancer in The Netherlands. J Clin Epidemiol 1990;43:285-295.
24. Barlow WE. Robust variance estimation for the case-cohort design. Biometrics 1994;50:1064-1072.
25. Barlow WE, Ichikawa L, Rosner D, Izumi S. Analysis of case-cohort designs. J Clin Epidemiol
1999;52:1165-1172.
26. van den Brandt PA, Schouten LJ, Goldbohm RA, Dorant E, Hunen PM. Development of a record linkage
protocol for use in the Dutch Cancer Registry for Epidemiological Research. Int J Epidemiol
1990;19:553-558.
27. Casparie M, Tiebosch AT, Burger G, Blauwgeers H, van de Pol A, van Krieken JH, Meijer GA. Pathology
databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide
histopathology and cytopathology data network and archive. Cell Oncol 2007;29:19-24.
28. Goldbohm RA, van den Brandt PA, Dorant E. Estimation of the coverage of Dutch municipalities by
cancer registries and PALGA based on hospital discharge data. Tijdschr Soc Gezondheidsz 1994;72:80-
84.
29. Parkin DM, Shanmugaratnam K, Sobin L, Ferlay J, Whelan SL. Histological Groups for comparative
studies. IARC Technical reports. Volume 31. Lyon: International Agency for Research on Cancer, 1998.
30. Goldbohm RA, van den Brandt PA, Brants HA, van't Veer P, Al M, Sturmans F, Hermus RJ. Validation of
a dietary questionnaire used in a large-scale prospective cohort study on diet and cancer. Eur J Clin
Nutr 1994;48:253-265.
31. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:1503-
1511.
32. Lagergren J, Bergstrom R, Adami HO, Nyren O. Association between medications that relax the lower
esophageal sphincter and risk for esophageal adenocarcinoma. Ann Intern Med 2000;133:165-175.
33. Cox DR. Regression models and life-tables. J Roy Statistical Society 1972;34:187-220.
34. Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrica 1982;69:239-
241.
35. De Stefani E, Boffetta P, Deneo-Pellegrini H, Ronco AL, Correa P, Mendilaharsu M. The role of
vegetable and fruit consumption in the aetiology of squamous cell carcinoma of the oesophagus: a
case-control study in Uruguay. Int J Cancer 2005;116:130-135.
36. Bosetti C, La Vecchia C, Talamini R, Simonato L, Zambon P, Negri E, Trichopoulos D, Lagiou P, Bardini R,
Franceschi S. Food groups and risk of squamous cell esophageal cancer in northern Italy. Int J Cancer
2000;87:289-294.
37. Sapkota A, Hsu CC, Zaridze D, Shangina O, Szeszenia-Dabrowska N, Mates D, Fabianova E, Rudnai P,
Janout V, Holcatova I, Brennan P, Boffetta P, Hashibe M. Dietary risk factors for squamous cell
carcinoma of the upper aerodigestive tract in central and eastern Europe. Cancer Causes Control
2008;19:1161-1170.
38. Chen H, Ward MH, Graubard BI, Heineman EF, Markin RM, Potischman NA, Russell RM, Weisenburger
DD, Tucker KL. Dietary patterns and adenocarcinoma of the esophagus and distal stomach. Am J Clin
Nutr 2002;75:137-144.
39. Ekstrom AM, Serafini M, Nyren O, Hansson LE, Ye W, Wolk A. Dietary antioxidant intake and the risk of
cardia cancer and noncardia cancer of the intestinal and diffuse types: a population-based case-control
study in Sweden. Int J Cancer 2000;87:133-140.
Thesis Jessie Steevens_v04.pdf
Vegetables and fruits consumption and risk of esophageal and gastric cancer subtypes
177
40. Nomura AM, Hankin JH, Kolonel LN, Wilkens LR, Goodman MT, Stemmermann GN. Case-control study
of diet and other risk factors for gastric cancer in Hawaii (United States). Cancer Causes Control
2003;14:547-558.
41. Mena S, Ortega A, Estrela JM. Oxidative stress in environmental-induced carcinogenesis. Mutat Res
2009;674:36-44.
42. Willett WC. Recall of remote diet. In: Willett WC, ed. Monographs in Epidemiology and Biostatistics.
Volume 30. Nutritional epidemiology. New York: Oxford Universtiy Press, 1998:148-156.
43. Vainio H, Bianchini F. Fruit and vegetables. IARC Handbooks of Cancer Prevention. Vol. 8. Lyon: IARC
Press, 2003.
44. Taylor DN, Blaser MJ. The epidemiology of Helicobacter pylori infection. Epidemiol Rev 1991;13:42-59.
45. Loffeld RJ, Stobberingh E, van Spreeuwel JP, Flendrig JA, Arends JW. The prevalence of anti-
Helicobacter (Campylobacter) pylori antibodies in patients and healthy blood donors. J Med Microbiol
1990;32:105-109.
46. Machida-Montani A, Sasazuki S, Inoue M, Natsukawa S, Shaura K, Koizumi Y, Kasuga Y, Hanaoka T,
Tsugane S. Association of Helicobacter pylori infection and environmental factors in non-cardia gastric
cancer in Japan. Gastric Cancer 2004;7:46-53.
Thesis Jessie Steevens_v04.pdf
Chapter 8
178
Thesis Jessie Steevens_v04.pdf
179
Thesis Jessie Steevens_v04.pdf
General discussion
9
Chapter 9
180
This chapter contains a discussion of the most important findings described in this
thesis in relation to other published research. Methodological points related to the
thesis topic are also discussed. In addition, strengths and weaknesses of the study
design and analyses are considered. Finally, the implications of our findings, and
suggestions for future research will be addressed.
MAIN FINDINGS
In this thesis, we described two studies into lifestyle factors and their association with
risk of Barrett’s esophagus (BE), using data from the Netherlands Cohort Study on Diet
and Cancer (NLCS). We observed that being overweight or obese was a risk factor for
BE in women, but not in men. Cigarette smoking moderately increased the risk of BE in
both sexes. No association was observed between consumption of alcohol and risk of
BE (Chapter 2). In Chapter 3 we described that toenail selenium status was not
associated with the development of BE. However, in one subgroup, namely BE patients
that progressed to high-grade dysplasia or esophageal adenocarcinoma (EAC), we did
observe an inverse association with toenail selenium status. These findings are
summarized in Table 9.1.
When we followed patients who were diagnosed with BE, we found that these
patients were at increased risk of not only esophageal cancer, but possibly also other
cancers, specifically small intestinal and pancreatic cancer. Overall mortality of BE
patients was not increased (Chapter 4).
In Chapter 5, we used data from European cancer registries to study trends in
incidence of esophageal and gastric cancers during the 1983-1997 period.
Adenocarcinomas of the esophagus and gastric cardia rose in incidence in most
countries, the strongest in the UK and Ireland. Incidence of esophageal squamous cell
carcinoma (ESCC) rose in women from all countries, and in men from Northern Europe
and Slovakia, but declined mostly in men from Southern and Western Europe. In nearly
all countries, the incidence of gastric non-cardia adenocarcinomas (GNCA) declined.
In the NLCS, we found that alcohol consumption was related to a strong increase
in risk of ESCC, while it was not related to risk of EAC, gastric cardia adenocarcinoma
(GCA), and GNCA (Table 9.1). Cigarette smoking clearly increased the risk of all four
cancers. Furthermore, combined exposure to alcohol and cigarette smoking was
associated with a very strong increase in ESCC risk, compatible with a multiplicative
interaction model (Chapter 6). By contrast, a strong inverse association was described
for higher toenail selenium status and risk of ESCC. A less strong inverse association
was observed with GCA risk, while toenail selenium status was associated with EAC risk
only in some groups: women, never smokers, and people with low antioxidant
consumption (Chapter 7). Finally, Chapter 8 described the association between
consumption of vegetables and fruits and risk of esophageal and gastric cancer
subtypes. Associations were generally inverse, and some statistically significant inverse
Thesis Jessie Steevens_v04.pdf
General discussion
181
associations were found with risk of ESCC, EAC, and GCA for specific vegetables and
fruits (Table 9.1). Specifically in smokers, vegetables and fruits consumption was
inversely associated with ESCC and EAC risk.
Table 9.1 Main findings of the etiological studies described in this thesis.
Direction and strength of observed association
Risk factor BE ESCC EAC GCA GNCA Chapter
Overweight Men: 0
Women: +
– ++ ++ – 2, Merry et al.1
Cigarette smoking + ++ ++ + + 2, 6
Alcohol consumption 0 ++ 0 0 0 2, 6
Toenail selenium status 0 – – – – Not studied 3, 7
Vegetables consumption Not studied – – – 0 8
Fruits consumption Not studied – – – 0 8
++ strong positive association (RRs ≥ 2.0), + positive association, 0 no association (RRs very close to 1 and no
dose-response trends), – inverse association, – – strong inverse association (RRs ≤ 0.5).
STRENGTHS OF THE NETHERLANDS COHORT STUDY
The Netherlands Cohort Study has a prospective cohort study design, which brings
several advantages. Prospective cohort studies are less susceptible to information and
selection bias compared with case-control studies. The validity of the former study
design is therefore usually higher. A second major strength of the prospective
character is the possibility to exclude cancer cases who were diagnosed early during
follow-up. This exclusion can diminish or eliminate bias caused by reverse causation.
Reverse causation can specifically be a problem in case-control studies on esophageal
and gastric cancer, as we argued before.2,3 Third, in the NLCS there is much
information available on potentially confounding variables. If necessary, we used this
information to adjust the associations between risk factors and disease. This
adjustment further increased the validity of our results.
DIFFERENCES BETWEEN ESOPHAGEAL AND GASTRIC CANCER
TYPES
Based on the observed differences in time trends and risk factors, we understand more
and more that ESCC and EAC are two different diseases and should be investigated as
such. The same is true for gastric cancer, which is also more and more often being
viewed as two separate diseases: GCA and GNCA. Below we will discuss the differences
Thesis Jessie Steevens_v04.pdf
Chapter 9
182
we observed in the light of findings by others. Also, some methodological issues
related to our research into ESCC, EAC, GCA, and GNCA will be dealt with.
ESCC versus EAC
In our trend study, we observed differences in incidence trends for ESCC and EAC in
many European countries. Incidence of EAC rose in most countries, while the incidence
trends were mixed for ESCC. As mentioned in Chapter 1, this may indicate a different
etiology of these two histologic types of esophageal cancer. ESCC and EAC also differ
with respect to pathology, tumor biology, and prognosis. This is acknowledged in the
newest TNM staging system, that provides separate stage groupings for ESCC and
adenocarcinomas of the esophagus and esophagogastric junction.4 ESCC patients are
on average somewhat younger than EAC patients.5, 6 Also, some studies have reported
a better prognosis of EAC than of ESCC, particularly in early stage disease.5,6
In our studies on risk factors for ESCC and EAC, summarized in Table 9.1, we
observed differences for overweight, alcohol consumption, and toenail selenium
status. These differences are in agreement with observations in other studies. Positive
associations are found between overweight and EAC, and inverse associations with
ESCC.7-9 Alcohol consumption is only associated with increased risk of ESCC, not with
EAC.8,10,11 Because our study is the first to investigate selenium and EAC risk, we cannot
compare our findings with the literature yet. We consequently recommend others to
study selenium and EAC risk.
ESCC and EAC were similar with respect to their associations with cigarette
smoking, vegetables consumption, and fruits consumption. Other cohort studies also
found increased risks of ESCC and EAC for smokers,8,10,12 although in two out of three
studies the risks were higher for ESCC than for EAC.10,12 In the literature, inverse
associations with vegetables and fruits consumption are reported somewhat more
frequently for ESCC than for EAC.7,13-16
From the above, we can conclude that ESCC and EAC are different with regard to
various aspects: risk factors, incidence trends, pathology, tumor biology, and
prognosis. We therefore believe that it is not informative to study esophageal cancers
as one entity. ESCC and EAC should be regarded as two distinct diseases in future
research.
EAC versus GCA
When making a comparison between EAC and GCA, one has to consider the
classification of these cancers. EAC most commonly arises in the distal part of the
esophagus, which is close to the gastric cardia. If an esophageal tumor is located near
the gastroesophageal junction, it can grow into the cardia. The opposite is also
possible: a tumor that originates from the gastric cardia can grow into the distal
esophagus. At the time of diagnosis, however, it is sometimes impossible to determine
whether a cancer has arisen from the distal esophagus or from the proximal
Thesis Jessie Steevens_v04.pdf
General discussion
183
stomach.17,18 Sometimes, a tumor is then classified as gastroesophageal junction
cancer.17
The cancer incidence data used in our study come from the Netherlands Cancer
Registry. The Netherlands Cancer Registry is a high quality cancer registry. Specially
trained registration clerks register directly from the medical records in the hospitals.
Clinical information as well as pathology information are collected.19-22 The
classification of EAC and GCA is based on these cancer registry data.
When our findings on EAC and GCA are compared, we can see that these cancers
share all risk factors that we investigated, i.e., overweight, smoking, alcohol
consumption, toenail selenium status, and vegetables and fruits consumption (Table
9.1). These findings are in agreement with other studies, which also found similarities
in risk factors for EAC and GCA.e.g. 8-10,14, 23
Data on trends in incidence are not separately presented for EAC and GCA in the
majority of trend studies. This has to do with the fact that in the past,
adenocarcinomas of the lower esophagus had to be coded as cardia tumors, according
to version 9 of the International Classification of Diseases. Also, researchers may not
want to separate these cancers, because of possible misclassification.
To sum up, it may be said that in our studies based on high quality cancer registry
data, we observed that EAC and GCA are etiologically very similar diseases. Further,
clinically it is very difficult to distinguish these diseases. In view of these facts, we
believe it may not be useful to keep trying to separate EAC and GCA, and therefore
recommend treating these diseases as one entity: gastroesophageal junction
adenocarcinomas.
GCA versus GNCA
In this thesis, we also found some differences between GCA and GNCA. Trends in
incidence rates were strikingly different for adenocarcinomas of the esophagus and
gastric cardia than for GNCA (the latter separately analyzed as gastric cancers of other
specified sites and not otherwise specified gastric cancers) (Chapter 5). This
observation is in agreement with studies on trends in gastric cancer subtypes in other
countries.e.g. 24-28
The associations we observed in the etiological analyses were indeed different for
GCA and GNCA for some risk factors (Table 9.1), although the contrast is less sharp
than for ESCC versus EAC. The largest difference was present for overweight, which
was a strong risk factor for GCA but inversely associated with GNCA. Recent data from
the prospective AARP Diet and Health study confirm our findings: increased BMI was
found to be a risk factor for GCA, but not for GNCA.29 We are unsure whether toenail
selenium status is differently associated with risk of GCA than with risk of GNCA. In a
previous study, we found that selenium was inversely associated (not statistically
significant) with risk of total gastric cancer, but we did not report on GNCA
separately.30 Possibly, selenium is only inversely associated with GCA, and not with
Thesis Jessie Steevens_v04.pdf
Chapter 9
184
GNCA. This was observed by a prospective study in Linxian, China.23 Further, the
associations between consumption of vegetables and fruits were inverse for GCA,
while this was not the case for GNCA. Other prospective studies have generally found
few associations between vegetables and fruits consumption and risk of GCA as well as
GNCA.7,14,31-33
As mentioned in several chapters, we lack information on infection with
Helicobacter (H.) pylori. On this subject, case-control studies have an advantage over
cohort studies. Medical data, including data on H. pylori infection, are more readily
available in case-control studies. One reason is that these studies are often carried out
in a medical setting, in contrast to the general population setting of cohort studies.
Also, the number of participants is usually much lower in case-control studies than in
cohort studies.
The lack of data on H. pylori is particularly important for GNCA, as H. pylori is the
main risk factor for this cancer. A recent review of epidemiologic studies that
investigated interaction between H. pylori infection and other risk factors for gastric
cancer concluded that there is a suggestion of a positive interaction between H. pylori
infection and smoking on gastric cancer risk. Further, the potential protective effect of
dietary antioxidants seemed to be observed only in subjects infected with H. pylori,
although the results were inconsistent.34 Analyzing the interaction between H. pylori
and other risk factors for GNCA within prospective studies would be of interest.
In accordance with our belief that it is not informative to study esophageal cancer
as one entity, we neither believe it is informative to study gastric cancer as one entity.
This thesis shows that GCA and GNCA are quite different diseases with respect to their
etiology, and future research should therefore make a distinction between these two
diseases.
BARRETT’S ESOPHAGUS VERSUS ESOPHAGEAL
ADENOCARCINOMA
Besides making comparisons between subtypes of esophageal and gastric cancer, an
important focus of this thesis is also to make a comparison between BE and EAC. We
believe this thesis contributes significantly to the knowledge of these two diseases,
also because the evidence is based on prospective data. The existence of the
nationwide network and registry of histopathology and cytopathology in the
Netherlands (PALGA)35 offered us the unique possibility to research BE in a prospective
cohort study.
The results of the etiologic analyses on BE and EAC clearly show some differences
between these diseases. In our cohort, overweight, cigarette smoking, and toenail
selenium status all have a different association with BE than with EAC (Table 9.1). It
seems that the associations found for EAC are less strong or not present for BE, at least
for the risk factors described in this thesis. In the literature, obese persons are
Thesis Jessie Steevens_v04.pdf
General discussion
185
generally found to have 3 times more risk of BE as well as EAC, and overweight persons
are found to have 1.5 times the risk of BE and EAC compared with normal weight
persons.36-41 These risk estimates for BE come from case-control studies. The only
other cohort study found no association between BMI and BE risk.42 Therefore, the
inconsistency between our results and most other studies may be due to the study
design. Smoking is probably more strongly associated with EAC compared with BE,
although the exact risk estimates vary between studies.8,10,37,43-45 For selenium and risk
of BE and EAC, there are no other studies available for comparison. It would thus be
interesting if others examined these associations.
The association between vegetables and fruits consumption and risk of BE is not
described in this thesis, but is certainly worthwhile investigating given the inverse
associations we observed with EAC risk. More in general, epidemiological studies on
(lifestyle) risk factors for BE are requested as the evidence is quite limited to date. It
also remains important to compare BE and EAC, to be able to find clues about the
etiologically relevant time window for exposures.
In our analyses of risk factors for BE and EAC, we were able to check for possible
confounding effects of various dietary and lifestyle habits of our cohort participants.
No information, however, was available on the presence of gastroesophageal reflux
disease (GERD). Use of medications for the treatment of reflux, such as antacids,
histamine 2 receptor antagonists (H2RA), and proton pump inhibitors (PPI), were not
measured well in our cohort. Here, case-control studies may have an advantage over
cohort studies, because this medical information is generally more readily available in
these studies. GERD or reflux symptoms may be confounders, effect-modifiers, or
intermediates in the associations we investigated. Therefore, future studies are
recommended to evaluate the role of GERD and reflux. This can best be done by
comparing the results from different statistical models, with and without adjustment
for these factors. This approach has already been applied by some researchers.43,46
Another point of interest is that there are some foods that have been associated
with temporary symptoms of reflux. These include dietary fat, chocolate, mints, coffee,
onions, citrus fruits, and tomatoes.47 Persons already suffering from reflux symptoms
may therefore try to avoid eating these foods in order to reduce their complaints. This
may influence not only our study, but all studies investigating diet and risk of BE,
because the dietary habits of BE cases (who often have a history of reflux) will then be
different from the habits of the non-cases. Specifically, cases may consume less reflux-
inducing foods than non-cases. In this way, reverse causation may play a role in the
observed associations between diet and risk of BE. We attempted to remove the
effects of reverse causation from our results by excluding BE cases that were
diagnosed in the first 2 years of the follow-up. These patients may be the patients that
already had considerable complaints at the time they filled out our baseline food
frequency questionnaire. These complaints may have influenced their diet and might
also have led to their diagnosis of BE. Nevertheless, excluding these early cases may
Thesis Jessie Steevens_v04.pdf
Chapter 9
186
not have solved the problem entirely, as reflux symptoms can already be present a
long time before BE diagnosis.48,49
The paradox of Barrett’s esophagus
It has recently been proposed that BE is a successful adaptation to the harsh intra-
esophageal environment of chronic gastroesophageal reflux disease. The BE tissue has
several functions not present in the normal squamous cell epithelium, that protect
against further reflux injury.50 Besides being a successful adaptation, BE is also
associated with increased risk of EAC, which is the reason that clinicians and
researchers are interested in BE. This increased risk of EAC is also the reason for
regular endoscopic surveillance of BE patients, to try to detect progression to EAC in an
early stage.51, 52 As mentioned before, the incidence of this highly lethal cancer is rising
in the Western world. Making a valid estimate of the risk of EAC in BE patients is
therefore essential for these patients, the cost-efficiency of surveillance, and the
relevance of BE as a precancerous lesion. Quite some studies have tried to estimate
the risk, but there are quite large differences between the estimates from different
studies.53 A recent meta-analysis estimated the risk of esophageal adenocarcinoma in
BE patients to be between 4.1 and 6.1% per 10 year.53 In our study described in
Chapter 4, we have therefore tried to make a valid estimate of, among other things,
EAC risk in BE using a population-based design. Our estimate of the incidence rate was
3.7% per 10 year, which is very close to the estimate from the meta-analysis.
Even though BE patients have a risk of EAC that is equal to or larger than 10 times
the risk of the general population (Chapter 4), their absolute risk of EAC is still quite
low. The majority of BE patients never develop EAC.53 Thus, one could say there is
some overdiagnosis of BE. This overdiagnosis contrasts sharply to the underdiagnosis
of early stage EAC. A large proportion of persons diagnosed with EAC has not had any
previous complaints like reflux (50%), has not been diagnosed with GERD (80%), and
did not have a prior diagnosis of BE (95%).54 Because of the absence of early
complaints and diagnosis of precursor lesions, these persons are not under
surveillance and therefore the EAC is detected very late. By that time, the cancer is
already advanced and the associated mortality high.
Reid et al. have recently nicely described this contrast between overdiagnosis of
BE and underdiagnosis of early EAC, which they named the paradox of Barrett’s
esophagus.54 Changing this paradox of BE is necessary to improve patient care and
lower esophageal cancer mortality. One possibility to achieve this, is to identify those
patients with BE that are at highest risk of progression, thus making a risk
stratification. In our follow-up study in Chapter 4, we observed that male BE patients
have a higher risk of progression to EAC than women. This was also reported in a meta-
analysis on risk of progression in BE.53 Also, BE patients with high-grade dysplasia
(HGD) were at greatest risk of progression. However, the assessment of the grade of
dysplasia (negative, indefinite, low grade, or high grade)55 is subject to a large inter-
Thesis Jessie Steevens_v04.pdf
General discussion
187
observer variation.54 This assessment may thus be improved. Several researchers have
studied biomarkers that may be valuable for risk stratification in BE. These biomarkers
include measures of genomic instability, chromosomal alterations, chromosomal
instability, microsatellite instability, mutations, and disruption of regulatory pathways
in BE tissue.54 Future use of these biomarkers in clinical practice may be promising.
Besides these biomarkers, other characteristics of the BE patient, such as
demographic, lifestyle and medical characteristics might also be useful for risk
stratification. However, the associations between these factors and risk of progression
of BE remain to be well assessed. Valuable information on this topic could be obtained
by setting up a prospective cohort of BE patients. At the start of that cohort study,
information on all kinds of possibly relevant factors for progression should be
collected.
Ultimately, the results of all these investigations could be used to develop a risk
score that clinicians can use to predict the individual risk of progression in their BE
patient, communicate this risk to the patient, and apply personalized patient care.56
Patient care may include endoscopic surveillance, endoscopic treatment,
chemoprevention with NSAIDs/aspirin, and advice to make lifestyle changes. For
patients at low risk of progression, no intervention at all may be the best care.
However, these options remain to be demonstrated as effective in trials.54
IMPLICATIONS FOR PUBLIC HEALTH
Naturally, the lifestyle factors described in this thesis are not only related to risk of
Barrett’s esophagus and esophageal and gastric cancer subtypes. Risk of many other
diseases is influenced (increased or decreased) by smoking, overweight and obesity,
alcohol consumption and dietary habits. The directions of the associations we
observed with BE and esophageal and gastric cancers generally correspond to the
direction of the associations with other diseases (Table 2). As a consequence of these
diseases, individuals live some years of their lives in poor health. Further, a number of
life years are lost due to early mortality. The National Institute for Public Health and
the Environment (RIVM) in the Netherlands has quantified the contribution of several
risk factors to the burden of disease (Table 2).57, 58 For example, compared with never
smokers, smokers lose on average 4.1 life years and 4.6 healthy life years. Overweight
and obesity also cause significant loss of life years and healthy life years, while the
effects of excess alcohol consumption are more limited. No figures were available for
the effect of selenium status and vegetables and fruit consumption.
These figures stress the importance of recommendations and guidelines for
healthy living, as described by e.g. the World Cancer Research Fund 59, the National
Health Council of the Netherlands,60 and the Dutch Cancer Society.61 Of course, only
formulating these recommendations and guidelines is not sufficient. It is not always
self-evident or easy for individuals to follow these recommendations and therefore
Thesis Jessie Steevens_v04.pdf
Chapter 9
188
they should be supported.58 The findings of the research described in this thesis
underline the guidelines for a healthy lifestyle and the importance of motivating and
helping the population to comply with these guidelines.
Table 9.2 Effects of risk factors on life years and healthy life years (with 95% confidence intervals)
(Chronic Diseases Model). a
Risk factor Life years lost Healthy life years lost
(HALE) b
Individuals in the risk groups:
Smokers (including former smokers) 4.1 (3.7-4.6) 4.6 (4.1-5.4)
Overweight subjects c 1.2 (1.0-1.5) 2.1 (1.8-2.3)
Obese subjects c 3.0 (2.3-3.6) 5.1 (4.5-5.6)
Users of excess alcohold 0.6 (0.5-0.7) 0.9 (0.8-1.0)
a Adapted from reference 57; b HALE, health-adjusted life expectancy: a measure for the number of healthy
life years someone can expect to live from a certain age. This is calculated based on the prevalence and the
seriousness of diseases; c Overweight is defined as BMI ≥25 kg/m2. Obesity is defined as BMI ≥30 kg/m2. The
reference group is normal weight (BMI 18.5-25 kg/m2); d Defined as ≥3 glasses of alcohol per day for men,
and ≥2 glasses of alcohol per day for women.
SUGGESTIONS FOR FUTURE RESEARCH
Besides the suggestions for future research described in the texts above, below we
formulated a few additional recommendations.
It is important to keep monitoring the incidence of histologic types of esophageal
cancer and subsites of gastric cancer. Also in countries in which no changes have been
observed yet, these can still take place later. Preferably, figures of recent date should
be used in trends studies. Observed trends can further be extrapolated to make
predictions about future trends. Results from trends studies can give clues about yet
unknown risk factors, and moreover they can be used for e.g. future health care
planning.
An exposure of interest that could be investigated further is physical activity,
which might reduce risk of esophageal and gastric cancers independently of BMI.62, 63
Additionally, reproductive and hormonal factors may be related to risk of esophageal
and gastric cancer subtypes and may also explain the high male predominance of EAC
and GCA. Only a few studies have been conducted on this topic and this requires
further attention.64, 65
Finally, besides studying BE, studying precursor lesions of other esophageal and
gastric cancer subtypes (ESCC, GCA, and GNCA) may be valuable. Issues that deserve
particular attention are the diagnosis and definition of the precursor lesion, risk factors
for the development and malignant progression of the lesion, and the magnitude of
the risk of progression to cancer.
Thesis Jessie Steevens_v04.pdf
General discussion
189
REFERENCES
1. Merry AH, Schouten LJ, Goldbohm RA, van den Brandt PA. Body mass index, height and risk of
adenocarcinoma of the oesophagus and gastric cardia: a prospective cohort study. Gut 2007;56:
1503-1511.
2. Botterweck AA, van den Brandt PA, Goldbohm RA. A prospective cohort study on vegetable and fruit
consumption and stomach cancer risk in The Netherlands. Am J Epidemiol 1998;148:842-853.
3. Botterweck AAM. Diet and risk of stomach cancer [dissertation]. Maastricht University, 2000.
4. Edge SB, Byrd DR, Compton CC, Fritz AG, Green FL, Trotti A. AJCC (American Joint Committee on
Cancer) Cancer Staging Manual. New York: Springer, 2010.
5. Siewert JR, Ott K. Are squamous and adenocarcinomas of the esophagus the same disease? Semin
Radiat Oncol 2007;17:38-44.
6. Mariette C, Finzi L, Piessen G, Van Seuningen I, Triboulet JP. Esophageal carcinoma: prognostic
differences between squamous cell carcinoma and adenocarcinoma. World J Surg 2005;29:39-45.
7. Tran GD, Sun XD, Abnet CC, Fan JH, Dawsey SM, Dong ZW, Mark SD, Qiao YL, Taylor PR. Prospective
study of risk factors for esophageal and gastric cancers in the Linxian general population trial cohort in
China. Int J Cancer 2005;113:456-463.
8. Lindblad M, Rodriguez LA, Lagergren J. Body mass, tobacco and alcohol and risk of esophageal, gastric
cardia, and gastric non-cardia adenocarcinoma among men and women in a nested case-control study.
Cancer Causes Control 2005;16:285-294.
9. Kubo A, Corley DA. Body mass index and adenocarcinomas of the esophagus or gastric cardia: a
systematic review and meta-analysis. Cancer Epidemiol Biomarkers Prev 2006;15:872-878.
10. Freedman ND, Abnet CC, Leitzmann MF, Mouw T, Subar AF, Hollenbeck AR, Schatzkin A. A prospective
study of tobacco, alcohol, and the risk of esophageal and gastric cancer subtypes. Am J Epidemiol
2007;165:1424-1433.
11. Ishiguro S, Sasazuki S, Inoue M, Kurahashi N, Iwasaki M, Tsugane S. Effect of alcohol consumption,
cigarette smoking and flushing response on esophageal cancer risk: a population-based cohort study
(JPHC study). Cancer Lett 2009;275:240-246.
12. Zendehdel K, Nyren O, Luo J, Dickman PW, Boffetta P, Englund A, Ye W. Risk of gastroesophageal
cancer among smokers and users of Scandinavian moist snuff. Int J Cancer 2008;122:1095-1099.
13. Freedman ND, Park Y, Subar AF, Hollenbeck AR, Leitzmann MF, Schatzkin A, Abnet CC. Fruit and
vegetable intake and esophageal cancer in a large prospective cohort study. Int J Cancer
2007;121:2753-2760.
14. Gonzalez CA, Pera G, Agudo A, Bueno-De-Mesquita HB, Ceroti M, Boeing H, Schulz M, Del Giudice G,
Plebani M, Carneiro F, Berrino F, Sacerdotde C, Tumino R, Panico S, Berglund G, Siman H, Hallmans G,
Stenling R, Martinez C, Dorronsoro M, Barricarte A, Navarro C, Quiros JR, Allen N, Key TJ, Bingham S,
Day NE, Linseisen J, Nagel G, Overvad K, Jensen MK, Olsen A, Tjonneland A, Buchner FL, Peeters PH,
Numans ME, Clavel-Chapelon F, Boutron-Ruault MC, Roukos D, Trichopolou A, Psaltopoulou T, Lund E,
Casagrande C, Slimani N, Jenab M, Riboli E. Fruit and vegetable intake and the risk of stomach and
oesophagus adenocarcinoma in the European Prospective Investigation into Cancer and Nutrition
(EPIC-EURGAST). International Journal of Cancer 2006;118:2559-2566.
15. Yamaji T, Inoue M, Sasazuki S, Iwasaki M, Kurahashi N, Shimazu T, Tsugane S. Fruit and vegetable
consumption and squamous cell carcinoma of the esophagus in Japan: the JPHC study. Int J Cancer
2008;123:1935-1940.
16. Guo W, Blot WJ, Li JY, Taylor PR, Liu BQ, Wang W, Wu YP, Zheng W, Dawsey SM, Li B, et al. A nested
case-control study of oesophageal and stomach cancers in the Linxian nutrition intervention trial. Int J
Epidemiol 1994;23:444-450.
17. McColl KE, Going JJ. Aetiology and classification of adenocarcinoma of the gastro-oesophageal
junction/cardia. Gut 2010;59:282-284.
18. Rusch VW. Are cancers of the esophagus, gastroesophageal junction, and cardia one disease, two, or
several? Semin Oncol 2004;31:444-449.
19. Parkin DM, Whelan SL, Ferlay J, Raymond L, Young J. Cancer incidence in five continents. Volume VII.
Lyon: IARC Scientific publications no. 143, 1997.
Thesis Jessie Steevens_v04.pdf
Chapter 9
190
20. Parkin DM, Whelan SL, Ferlay J, Teppo L, Thomas DB. Cancer incidence in five continents. Volume VIII.
Lyon: IARC Scientific publications no. 155, 2002.
21. Curado MP, Edwards B, Shin HR, Storm H, Ferlay J, Heanue M, Boyle P. Cancer incidence in five
continents. Volume IX. Lyon: IARC Scientific publications no. 160, 2007.
22. Schouten LJ, Jager JJ, van den Brandt PA. Quality of cancer registry data: a comparison of data
provided by clinicians with those of registration personnel. Br J Cancer 1993;68:974-977.
23. Mark SD, Qiao YL, Dawsey SM, Wu YP, Katki H, Gunter EW, Fraumeni JF, Jr., Blot WJ, Dong ZW, Taylor
PR. Prospective study of serum selenium levels and incident esophageal and gastric cancers. J Natl
Cancer Inst 2000;92:1753-1763.
24. Brown LM, Devesa SS. Epidemiologic trends in esophageal and gastric cancer in the United States. Surg
Oncol Clin N Am 2002;11:235-256.
25. Hansen S, Wiig JN, Giercksky KE, Tretli S. Esophageal and gastric carcinoma in Norway 1958-1992:
incidence time trend variability according to morphological subtypes and organ subsites. Int J Cancer
1997;71:340-344.
26. Levi F, La Vecchia C, Te VC. Descriptive epidemiology of adenocarcinomas of the cardia and distal
stomach in the Swiss Canton of Vaud. Tumori 1990;76:167-171.
27. Harrison SL, Goldacre MJ, Seagroatt V. Trends in registered incidence of oesophageal and stomach
cancer in the Oxford region, 1974-88. Eur J Cancer Prev 1992;1:271-274.
28. Moller H. Incidence of cancer of oesophagus, cardia and stomach in Denmark. Eur J Cancer Prev
1992;1:159-164.
29. Abnet CC, Freedman ND, Hollenbeck AR, Fraumeni JF, Jr., Leitzmann M, Schatzkin A. A prospective
study of BMI and risk of oesophageal and gastric adenocarcinoma. Eur J Cancer 2008;44:465-471.
30. van den Brandt PA, Goldbohm RA, van 't Veer P, Bode P, Dorant E, Hermus RJ, Sturmans F. A
prospective cohort study on toenail selenium levels and risk of gastrointestinal cancer. J Natl Cancer
Inst 1993;85:224-229.
31. Nouraie M, Pietinen P, Kamangar F, Dawsey SM, Abnet CC, Albanes D, Virtamo J, Taylor PR. Fruits,
vegetables, and antioxidants and risk of gastric cancer among male smokers. Cancer Epidemiol
Biomarkers Prev 2005;14:2087-2092.
32. Kobayashi M, Tsubono Y, Sasazuki S, Sasaki S, Tsugane S. Vegetables, fruit and risk of gastric cancer in
Japan: a 10-year follow-up of the JPHC Study Cohort I. Int J Cancer 2002;102:39-44.
33. Freedman ND, Subar AF, Hollenbeck AR, Leitzmann MF, Schatzkin A, Abnet CC. Fruit and vegetable
intake and gastric cancer risk in a large United States prospective cohort study. Cancer Causes Control
2008;19:459-467.
34. Gonzalez CA, Lopez-Carrillo L. Helicobacter pylori, nutrition and smoking interactions: their impact in
gastric carcinogenesis. Scand J Gastroenterol 2010;45:6-14.
35. Casparie M, Tiebosch AT, Burger G, Blauwgeers H, van de Pol A, van Krieken JH, Meijer GA. Pathology
databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide
histopathology and cytopathology data network and archive. Cell Oncol 2007;29:19-24.
36. Bu X, Ma Y, Der R, Demeester T, Bernstein L, Chandrasoma PT. Body mass index is associated with
Barrett esophagus and cardiac mucosal metaplasia. Dig Dis Sci 2006;51:1589-1594.
37. Edelstein ZR, Farrow DC, Bronner MP, Rosen SN, Vaughan TL. Central adiposity and risk of Barrett's
esophagus. Gastroenterology 2007;133:403-411.
38. Stein DJ, El-Serag HB, Kuczynski J, Kramer JR, Sampliner RE. The association of body mass index with
Barrett's oesophagus. Aliment Pharmacol Ther 2005;22:1005-1010.
39. Hampel H, Abraham NS, El-Serag HB. Meta-analysis: obesity and the risk for gastroesophageal reflux
disease and its complications. Ann Intern Med 2005;143:199-211.
40. Corley DA, Kubo A, Zhao W. Abdominal obesity and the risk of esophageal and gastric cardia
carcinomas. Cancer Epidemiol Biomarkers Prev 2008;17:352-358.
41. Reeves GK, Pirie K, Beral V, Green J, Spencer E, Bull D. Cancer incidence and mortality in relation to
body mass index in the Million Women Study: cohort study. Bmj 2007;335:1134.
42. Corley DA, Kubo A, Levin TR, Block G, Habel L, Zhao W, Leighton P, Quesenberry C, Rumore GJ, Buffler
PA. Abdominal obesity and body mass index as risk factors for Barrett's esophagus. Gastroenterology
2007;133:34-41; quiz 311.
Thesis Jessie Steevens_v04.pdf
General discussion
191
43. Anderson LA, Watson RG, Murphy SJ, Johnston BT, Comber H, Mc Guigan J, Reynolds JV, Murray LJ.
Risk factors for Barrett's oesophagus and oesophageal adenocarcinoma: results from the FINBAR
study. World J Gastroenterol 2007;13:1585-1594.
44. Smith KJ, O'Brien SM, Green AC, Webb PM, Whiteman DC. Current and past smoking significantly
increase risk for Barrett's esophagus. Clin Gastroenterol Hepatol 2009;7:840-848.
45. Kubo A, Levin TR, Block G, Rumore G, Quesenberry CP, Jr., Buffler P, Corley DA. Cigarette smoking and
the risk of Barrett's esophagus. Cancer Causes Control 2009;20:303-311.
46. Whiteman DC, Sadeghi S, Pandeya N, Smithers BM, Gotley DC, Bain CJ, Webb PM, Green AC.
Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the
oesophagus. Gut 2008;57:173-180.
47. Terry P, Lagergren J, Wolk A, Nyren O. Reflux-inducing dietary factors and risk of adenocarcinoma of
the esophagus and gastric cardia. Nutr Cancer 2000;38:186-191.
48. Eisen GM, Sandler RS, Murray S, Gottfried M. The relationship between gastroesophageal reflux
disease and its complications with Barrett's esophagus. Am J Gastroenterol 1997;92:27-31.
49. Conio M, Filiberti R, Blanchi S, Ferraris R, Marchi S, Ravelli P, Lapertosa G, Iaquinto G, Sablich R,
Gusmaroli R, Aste H, Giacosa A. Risk factors for Barrett's esophagus: a case-control study. Int J Cancer
2002;97:225-229.
50. Orlando RC. Mucosal defense in Barrett's esophagus. In: Sharma P, Sampliner R, eds. Barrett's
esophagus and esophageal adenocarcinoma. Massachusetts: Blackwell, 2006.
51. Wang KK, Sampliner RE. Updated guidelines 2008 for the diagnosis, surveillance and therapy of
Barrett's esophagus. Am J Gastroenterol 2008;103:788-797.
52. Playford RJ. New British Society of Gastroenterology (BSG) guidelines for the diagnosis and
management of Barrett's oesophagus. Gut 2006;55:442-443.
53. Yousef F, Cardwell C, Cantwell MM, Galway K, Johnston BT, Murray L. The incidence of esophageal
cancer and high-grade dysplasia in Barrett's esophagus: a systematic review and meta-analysis. Am J
Epidemiol 2008;168:237-249.
54. Reid BJ, Li X, Galipeau PC, Vaughan TL. Barrett's oesophagus and oesophageal adenocarcinoma: time
for a new synthesis. Nat Rev Cancer 2010;10:87-101.
55. Schlemper RJ, Riddell RH, Kato Y, Borchard F, Cooper HS, Dawsey SM, Dixon MF, Fenoglio-Preiser CM,
Flejou JF, Geboes K, Hattori T, Hirota T, Itabashi M, Iwafuchi M, Iwashita A, Kim YI, Kirchner T,
Klimpfinger M, Koike M, Lauwers GY, Lewin KJ, Oberhuber G, Offner F, Price AB, Rubio CA, Shimizu M,
Shimoda T, Sipponen P, Solcia E, Stolte M, Watanabe H, Yamabe H. The Vienna classification of
gastrointestinal epithelial neoplasia. Gut 2000;47:251-255.
56. Prasad GA, Bansal A, Sharma P, Wang KK. Predictors of progression in Barrett's esophagus: current
knowledge and future directions. Am J Gastroenterol 2010;105:1490-1502.
57. http://www.nationaalkompas.nl/gezondheid-en-ziekte/sterfte-levensverwachting-en-daly-
s/ziektelast-in-daly-s/wat-is-de-bijdrage-van-risicofactoren/ (accessed July 23, 2010).
58. van der Lucht F, Polder JJ. Rijksinstituut voor Volksgezondheid en Milieu (RIVM). Van gezond naar
beter. Kernrapport Volksgezondheid Toekomst Verkenning 2010. Houten: Bohn Stafleu van Loghum,
2010.
59. World Cancer Research Fund, American Institute for Cancer Research. Food, nutrition, physical activity
and the prevention of cancer: a global perspective. AICR, 2007.
60. Health Council of the Netherlands. [Guidelines for a healthy diet 2006]. Volume publication no.
2006/21. The Hague: Health Council of the Netherlands, 2006.
61. http://www.kwfkankerbestrijding.nl/index.jsp?objectid=16194 (accessed July 23, 2010).
62. Balbuena L, Casson AG. Physical activity, obesity and risk for esophageal adenocarcinoma. Future
Oncol 2009;5:1051-1063.
Thesis Jessie Steevens_v04.pdf
Chapter 9
192
63. Huerta JM, Navarro C, Chirlaque MD, Tormo MJ, Steindorf K, Buckland G, Carneiro F, Johnsen NF,
Overvad K, Stegger J, Tjonneland A, Boutron-Ruault MC, Clavel-Chapelon F, Morois S, Boeing H, Kaaks
R, Rohrmann S, Vigl M, Lagiou P, Trichopoulos D, Trichopoulou A, Bas Bueno-de-Mesquita H,
Monninkhof EM, Numans ME, Peeters PH, Mattiello A, Pala V, Palli D, Tumino R, Vineis P, Agudo A,
Ardanaz E, Arriola L, Molina-Montes E, Rodriguez L, Lindkvist B, Manjer J, Stenling R, Lund E, Crowe FL,
Key TJ, Khaw KT, Wareham NJ, Jenab M, Norat T, Romaguera D, Riboli E, Gonzalez CA. Prospective
study of physical activity and risk of primary adenocarcinomas of the oesophagus and stomach in the
EPIC (European Prospective Investigation into Cancer and nutrition) cohort. Cancer Causes Control
2010;21:657-669.
64. Freedman ND, Lacey JV, Jr., Hollenbeck AR, Leitzmann MF, Schatzkin A, Abnet CC. The association of
menstrual and reproductive factors with upper gastrointestinal tract cancers in the NIH-AARP cohort.
Cancer 2010;116:1572-1581.
65. Chandanos E, Lagergren J. Oestrogen and the enigmatic male predominance of gastric cancer. Eur J
Cancer 2008;44:2397-2403.
Thesis Jessie Steevens_v04.pdf
Summary / Samenvatting
193
Thesis Jessie Steevens_v04.pdf
194
SUMMARY
Cancer affects millions of persons in the world. In the Netherlands, cancer has become
the most important cause of death in 2008. This thesis concerns two types of cancer:
cancer of the esophagus, and cancer of the stomach. Esophageal and gastric cancer are
diseases with a poor prognosis: the 5-year survival rate is only 14% for esophageal
cancer and 21% for gastric cancer. Besides these cancers, this thesis also concerns
Barrett’s esophagus (BE). BE is a condition of the lower esophagus, in which the normal
squamous epithelium is replaced by columnar epithelium. Patients with BE are at
increased risk of developing esophageal adenocarcinoma (Chapter 1).
There are some indications that the two main subtypes of esophageal cancer,
esophageal squamous cell carcinoma (ESCC) and esophageal adenocarcinoma (EAC),
have a different etiology. This is also true for the two subtypes of gastric cancer: gastric
cardia adenocarcinoma (GCA) and gastric non-cardia adenocarcinoma (GNCA). First,
esophageal and gastric cancer subtypes are geographically differently distributed.
Second, the trends in incidence rates of these cancer subtypes differ. Third, the male-
to-female ratio of the incidence of these cancers differs by subtype.
In this thesis, we investigated whether the risk factors for ESCC, EAC, GCA, and
GNCA are indeed different. We studied cigarette smoking, alcohol consumption,
overweight, toenail selenium status, and consumption of vegetables and fruits.
Further, we also looked into the association of these factors with the risk of BE. These
lifestyle factors can potentially be modified and can thus have a role in primary
prevention.
The research described in this thesis is based on a prospective cohort study: the
Netherlands Cohort Study on diet and cancer (NLCS). The NLCS was initiated in
September 1986, with the enrollment of 120,852 men and women aged 55-69 years.
These subjects were randomly selected from Dutch municipal population registries. At
baseline, all subjects filled out a questionnaire on dietary habits, lifestyle, and other
risk factors for cancer. According to the case-cohort approach, data are processed and
analyzed for a random sample of the cohort (the subcohort) and cases. The subcohort
consists of 5,000 men and women who were sampled at baseline. This subcohort is
followed-up for vital status and migration and is used to estimate the person-time at
risk for the total cohort. Incident cases of BE and cancer were identified in the whole
cohort. BE cases were identified through record linkage with data from PALGA (the
nationwide registry for histopathology and cytopathology in the Netherlands). The
existence of this unique nationwide pathology registry offered us the opportunity to
study BE within a prospective cohort study. A pathologist reviewed and coded further
characteristics of the BE cases from the PALGA data. Incident cancer cases were
identified through record linkages with PALGA and the Netherlands Cancer Registry.
The cohort has been followed-up for vital status through record linkage with the
Central Bureau of Genealogy and automated municipal population registries. For
deceased cohort members we also obtained information on the cause of death from
Thesis Jessie Steevens_v04.pdf
Summary
195
Statistics Netherlands. The research described in this thesis was based on 16.3-year
follow-up data of the cohort (September 17, 1986 through December 31, 2002).
Chapter 2 describes the association between overweight, smoking, and alcohol
consumption and the risk of BE. Overweight and obesity were associated with an
increased risk of BE in women, but not in men. The association in women was not
specifically due to abdominal overweight. Former cigarette smokers were at 30%
increased risk of BE, but current smokers were not. Smoking duration showed a
positive association with BE risk, while alcohol consumption was not associated with
an increased risk of BE.
A study on toenail selenium status and its relationship with risk of BE is presented
in Chapter 3. We did not find an association between selenium and risk of BE. No
association was found either in subgroups defined by sex, smoking status, body mass
index, or intake of antioxidants. For BE cases that later progressed to high-grade
dysplasia or adenocarcinoma, we observed an inverse association with selenium.
In Chapter 4 we used the data of the NLCS to follow BE patients for incidence of
esophageal and gastric cancer, other cancers, and cause-specific mortality. We found
that esophageal and gastric cancer occurred 10 and nearly 2 times more frequently,
respectively, in BE patients than in the general population. Total incidence of all other
cancers was 30% increased in BE patients. Small intestinal and pancreatic cancer
incidence were increased in particular. All-cause mortality and cause-specific mortality
were not increased.
For the study in Chapter 5, we used data from the European Network of Cancer
Registries. We described the incidence trends of esophageal and gastric cancers in 13
European countries in the period 1983-1997. The incidence of adenocarcinomas of the
esophagus and gastric cardia rose in most, but not all, countries, mostly 1-7% per year.
Incidence of ESCC rose in women from all countries and in men from Northern Europe
and Slovakia, but declined mostly in men from Southern and Western Europe. In nearly
all countries, the incidence of GNCA declined.
The results described in Chapter 6 indicate that consumption of 3 or more glasses
alcoholic beverages per day was related to a 4- to 5-fold increase in risk of ESCC.
Alcohol consumption was not related to risk of EAC, GCA, and GNCA, while cigarette
smoking increased the risk of all four cancers 1.5 to 2 times. Furthermore, combined
exposure to alcohol and cigarette smoking was associated with a very strong increase
in ESCC risk: up to 8 times compared with people who did not drink or smoke.
In Chapter 7, the results are presented for the association between toenail
selenium status and risk of ESCC, EAC, and GCA. These results indicate that a high
selenium status is associated with a strong decrease in risk of ESCC. An inverse
association was also found for GCA. For EAC, inverse associations with selenium were
found in subgroups: women, never smokers, and persons with a low antioxidant
intake.
The relation between consumption of vegetables and fruits and risk of ESCC, EAC,
GCA, and GNCA is described in Chapter 8. The results generally indicated inverse
Thesis Jessie Steevens_v04.pdf
196
associations. Specifically, raw vegetables were associated with a lower EAC risk and
Brassica vegetables were associated inversely with GCA risk. Citrus fruits were
inversely associated with EAC and GCA risk. Specifically for current smokers,
vegetables and fruits intake was inversely associated with ESCC and EAC risk.
Consumption of vegetables and fruits may therefore protect against development of
esophageal and gastric cancer subtypes.
This thesis concludes with a discussion of the findings described in this thesis
(Chapter 9). We make comparisons between ESCC and EAC, and conclude that these
cancers are different with regard to various aspects, among others risk factors and
incidence trends. GCA and GNCA are also compared, and here we also draw the
conclusion that these cancers are different in these respects. Because of these
observed differences, we believe it is neither informative to study esophageal cancer
as one entity nor to study gastric cancer as one entity in etiologic research. Therefore,
we recommend separate analyses of ESCC, EAC, GCA, and GNCA in future studies.
Further, it appears that EAC and GCA are quite similar diseases that are very difficult to
distinguish. It may therefore not be useful to keep trying to separate these, and we
therefore suggest treating these diseases and one entity: gastroesophageal junction
adenocarcinomas. Finally, we compare BE and EAC. It appears that the associations
found for EAC are less strong or not present for BE, for the risk factors described in this
thesis. A comparison between BE and EAC in future studies is valuable, as it may give
clues about the etiologically relevant time window for exposures. More - preferably
high quality - research into the relation between lifestyle factors and risk of
development and progression of BE is needed, as data are still very limited.
Thesis Jessie Steevens_v04.pdf
Samenvatting
197
SAMENVATTING
Kanker treft miljoenen mensen in de wereld. In Nederland is kanker in 2008 de
belangrijkste doodsoorzaak geworden. Dit proefschrift gaat over twee vormen van
kanker: slokdarmkanker en maagkanker. Slokdarm- en maagkanker zijn ziekten met
een slechte prognose: de 5-jaarsoverleving is slechts 14% voor slokdarm- en 21% voor
maagkanker. Naast deze kankers gaat dit proefschrift ook over Barrett’s slokdarm.
Barrett’s slokdarm is een aandoening van het onderste deel van de slokdarm, waarbij
het normale plaveiselepitheel is vervangen door cilinderepitheel. Patiënten met
Barrett’s slokdarm hebben een verhoogd risico op het ontwikkelen van
adenocarcinoom van de slokdarm (Hoofdstuk 1).
Er zijn aanwijzingen dat de twee belangrijkste vormen van slokdarmkanker,
plaveiselcelcarcinoom (PC) en adenocarcinoom (AC), een verschillende
ontstaansgeschiedenis hebben. Dit geldt ook voor de twee vormen van maagkanker:
kanker van de cardia van de maag (CC) en kanker van andere delen van de maag (NCC).
Ten eerste zijn de vier vormen van slokdarm- en maagkanker anders verdeeld over de
wereld. Ten tweede zijn er verschillen in de tijdstrends in het vóórkomen van deze
vormen van kanker. Ten derde is de man-vrouw verhouding van het vóórkomen van
deze vormen van kanker verschillend.
In dit proefschrift onderzochten we of de risicofactoren voor PC, AC, CC en NCC
inderdaad verschillend zijn. We bestudeerden het roken van sigaretten, alcohol
consumptie, overgewicht, seleniumgehalte in teennagels, en groente- en
fruitconsumptie. Verder bekeken we ook het verband tussen deze risicofactoren en de
kans op het krijgen van Barrett’s slokdarm. Deze leefstijlfactoren kunnen mogelijk
worden veranderd en kunnen daarom een rol spelen bij het voorkómen van ziekten.
Het onderzoek dat is beschreven in dit proefschrift is voornamelijk gebaseerd op
een prospectief cohortonderzoek: de Nederlandse Cohort Studie naar voeding en
kanker (NLCS). De NLCS is in 1986 gestart met 120.852 mannen en vrouwen van 55-69
jaar oud. Deze deelnemers zijn willekeurig geselecteerd uit Nederlandse
bevolkingsregisters. Aan het begin van de studie hebben alle deelnemers een
vragenlijst ingevuld die ging over voedingsgewoonten, leefstijl en andere risicofactoren
voor kanker. De verwerking en analyse van de onderzoeksgegevens gebeurde volgens
de ‘case-cohort’ benadering. Daarbij worden de kankergevallen uit het gehele cohort
geanalyseerd samen met het subcohort. Dit subcohort bestaat uit een willekeurige
steekproef van 5.000 mannen en vrouwen, die aan het begin van de studie getrokken
is uit het cohort. Het subcohort wordt opgevolgd voor informatie over overlijden en
emigratie en het wordt gebruikt om de persoonstijd ‘at risk’ in het gehele cohort te
schatten. Nieuwe gevallen van Barrett’s slokdarm en kanker worden opgespoord in het
gehele cohort. De identificatie van patiënten met Barrett’s slokdarm vond plaats door
koppeling met gegevens van PALGA (Pathologisch-Anatomisch Landelijk
Geautomatiseerd Archief). Het bestaan van dit unieke landelijke archief gaf ons de
mogelijkheid om Barrett’s slokdarm te bestuderen in een prospectief
Thesis Jessie Steevens_v04.pdf
198
cohortonderzoek. Een patholoog bekeek de PALGA gegevens en codeerde verdere
kenmerken van de patiënten met Barrett’s slokdarm. Het cohort is ook opgevolgd voor
informatie over nieuwe kankergevallen, die verkregen is via PALGA en de Nederlandse
Kanker Registratie. Via het Centraal Bureau voor de Genealogie en het
bevolkingsregister (GBA) is informatie verkregen over welke deelnemers zijn
overleden. Met betrekking tot overleden cohortleden hebben we van het Centraal
Bureau voor de Statistiek informatie verkregen over de doodsoorzaak. Voor het in dit
proefschrift beschreven onderzoek is het cohort gedurende 16,3 jaar gevolgd (17
september 1986 t/m 31 december 2002).
Hoofdstuk 2 beschrijft het verband tussen overgewicht, roken en alcohol
consumptie en het risico op Barrett’s slokdarm. Overgewicht en obesitas waren
geassocieerd met een verhoogd risico op Barrett’s slokdarm bij vrouwen, maar niet bij
mannen. Het verband bij vrouwen werd niet specifiek veroorzaakt door overgewicht in
de buik. Ex-rokers hadden 30% meer risico op Barrett’s slokdarm, maar er was geen
hoger risico voor huidig rokers. Rookduur verhoogde het risico op Barrett’s slokdarm,
terwijl alcohol consumptie niet geassocieerd was met een verhoogd risico op Barrett’s
slokdarm.
Onze studie naar het seleniumgehalte van teennagels en de relatie met risico op
Barrett’s slokdarm wordt gepresenteerd in Hoofdstuk 3. We vonden geen verband
tussen selenium en risico op Barrett’s slokdarm. Ook vonden we geen verband in
subgroepen gedefinieerd op basis van geslacht, rookstatus, overgewicht, of inname
van antioxidanten. Voor patiënten met Barrett’s slokdarm waarbij later voortgang
werd gezien naar hooggradige dysplasie of AC vonden we een beschermend verband
met selenium.
In Hoofdstuk 4 gebruikten we de NLCS gegevens om patiënten met Barrett’s
slokdarm te volgen voor informatie over het optreden van slokdarm- en maagkanker,
andere kankers, en oorzaakspecifieke sterfte. We vonden dat slokdarm- en
maagkanker respectievelijk 10 en bijna 2 keer zoveel voorkwamen bij patiënten met
Barrett’s slokdarm als bij de algemene bevolking. Het vóórkomen van totaal kanker
(exclusief slokdarm- en maagkanker) was 30% verhoogd onder patiënten met Barrett’s
slokdarm. Dunne darm- en alvleesklierkanker kwamen vaker voor. Totale sterfte en
oorzaakspecifieke sterfte waren niet verhoogd.
Voor de studie beschreven in Hoofdstuk 5 gebruikten we gegevens van het
Europese Netwerk van Kanker Registraties. We beschreven de trends in het
vóórkomen van slokdarm- en maagkanker in 13 Europese landen in de periode 1983-
1997. Het vóórkomen van adenocarcinomen van de slokdarm en cardia van de maag
steeg in de meeste, maar niet alle landen, meestal 1-7% per jaar. Het vóórkomen van
PC steeg bij vrouwen uit alle landen en bij mannen uit Noord-Europa en Slowakije,
maar daalde meestal bij mannen uit Zuid- en West-Europa. In bijna alle landen nam
het vóórkomen van NCC af.
De resultaten die beschreven zijn in Hoofdstuk 6 geven aan dat het drinken van 3
of meer glazen alcoholische drank per dag was gerelateerd aan een 4- tot 5-maal
Thesis Jessie Steevens_v04.pdf
Samenvatting
199
verhoogd risico op PC. Alcoholconsumptie was niet gerelateerd aan het risico op AC,
CC en NCC. Het roken van sigaretten verhoogde het risico op alle vier vormen van
kanker 1.5 tot 2 maal. Bovendien was een gecombineerde bloostelling aan alcohol en
sigaretten geassocieerd met een zeer sterke verhoging van het risico op PC: tot 8 maal
vergeleken met mensen die niet dronken noch rookten.
In Hoofdstuk 7 worden de resultaten gepresenteerd voor het verband tussen het
seleniumgehalte van teennagels en risico op PC, AC en CC. Deze resultaten geven aan
dat een hoog seleniumgehalte het risico op PC sterk verlaagt. Een beschermend effect
werd ook gevonden voor CC. Voor AC werden er specifiek bij vrouwen, nooit rokers en
mensen met een lage inname van antioxidanten beschermende effecten van selenium
gevonden.
De relatie tussen groente- en fruitconsumptie en risico op PC, AC, CC en NCC is
beschreven in Hoofdstuk 8. De resultaten geven over het algemeen beschermende
effecten aan. Specifiek werden verbanden gevonden tussen rauwe groenten en een
lager AC risico en koolsoorten en een lager CC risico. Citrusfruit verlaagde het risico op
AC en CC. Specifiek bij huidig rokers waren groente- en fruitconsumptie geassocieerd
met een lager risico op PC en AC. Consumptie van groente en fruit zou kunnen
beschermen tegen het krijgen van vormen van slokdarm- en maagkanker.
Dit proefschrift eindigt met een discussie van de beschreven bevindingen
(Hoofdstuk 9). We maken een vergelijking tussen PC en AC en concluderen dat deze
kankers verschillend zijn op diverse aspecten, onder andere risicofactoren en
tijdstrends in het vóórkomen. CC en NCC worden ook vergeleken en ook hier trekken
we de conclusie dat deze kankers verschillend zijn op deze aspecten. Vanwege de
verschillen die wij hebben gevonden, geloven wij dat het noch informatief is om
slokdarmkanker als één ziekte te bestuderen, noch om maagkanker als één ziekte te
bestuderen, wanneer men de risicofactoren onderzoekt. Daarom stellen we voor om in
toekomstige studies aparte analyses te doen voor PC, AC, CC en NCC. Verder lijkt het
dat AC en CC zeer vergelijkbare ziekten zijn, die moeilijk te onderscheiden zijn. Daarom
is het misschien niet zinvol om te blijven proberen deze te scheiden. Vandaar bevelen
we aan om deze ziekten als één te beschouwen: adenocarcinomen van de slokdarm-
maag overgang. Als laatste vergelijken we Barrett’s slokdarm met AC. Het blijkt dat de
verbanden die we vonden voor AC, minder sterk of afwezig waren voor Barrett’s
slokdarm, voor de in dit proefschrift beschreven risicofactoren. Het vergelijken van
Barrett’s slokdarm en AC in toekomstige studies is waardevol, omdat het aanwijzingen
kan geven over de periode die relevant is voor blootstellingen. Meer - bij voorkeur
kwalitatief hoogstaand - onderzoek naar de relatie tussen leefstijlfactoren en het risico
op het ontwikkelen en de voortgang van Barrett’s slokdarm is nodig, want de gegevens
zijn nu erg beperkt.
Thesis Jessie Steevens_v04.pdf
200
Thesis Jessie Steevens_v04.pdf
Dankwoord
201
Thesis Jessie Steevens_v04.pdf
202
DANKWOORD
Mijn promotietraject, vier jaar onderzoek doen aan de universiteit. De tijd is voorbij
gevlogen. De artikelen zijn geschreven, presentaties gegevens, het boekje is af. In de
afgelopen vier jaar hebben vele mensen hieraan bijgedragen. Als ik je onverhoopt
vergeet te noemen hieronder, bedankt voor jouw bijdrage!
Als eerste wil ik mijn promotor en copromotoren bedanken.
Piet, jij overzag altijd het geheel en lette goed op de planning. Je maakte steeds
genoeg tijd vrij voor een overleg en ik kon laagdrempelig binnenlopen om een vraagje
te stellen. Bedankt.
Leo, elke week een uur om te overleggen over belangrijke dingen en allerlei
details. Wat een luxe, dat kan niet iedere promovendus zeggen. Je hield me goed in de
gaten: ‘lukt het nog allemaal?’. Dankjewel voor alle hulp.
Sandra, het meeste contact hadden we door de afstand via de mail. Zeker tegen
het einde van het project bleef ik je maar stukken sturen. Bedankt voor je kritische blik
op de artikelen.
Janneke, vier jaar lang zaten we bij elkaar op de kamer. Je bent een super
kamergenoot: genoeg tijd om te kletsen en te discussiëren, maar ook om rustig door te
werken. En het raam kon altijd open voor frisse lucht, ook in de winter. Dankjewel voor
je plezierige gezelschap! Ik hoop dat we elkaar nog spreken.
Een bedankje ook voor alle NLCS leden. Allemaal hebben jullie bijgedragen aan
mijn onderzoek, in de vorm van het invoeren van duizenden vragenlijsten,
beantwoorden van mijn vraagjes, zoeken van nagels of op welke manier ook.
Epidemiologie AIO’s en oud-AIO’s Audrey, Mirjam, Laura, Colinda, Brenda, Stefan,
Esther, Monique, Karolien, Karolina, Sander, Paul, Anne, Nadine en Milan: bedankt
voor alle gezellige lunches en gesprekken bij de koffiehoek of op de gang. Succes met
jullie promotieonderzoek - voor zover dat niet al lang af is :-).
Angela, José en Marionne, ik hoop dat jullie de yogalessen tijdens de lunchpauze
net zo relaxed vonden als ik.
Andere collega’s van epidemiologie: bedankt voor jullie collegialiteit!
Anita, Miranda, Ann, Clément, Yolande en Menno: heel fijn dat ik gebruik kon maken
van jullie expertise! Dank voor de vruchtbare samenwerking. Professor Murray, thank
you for the interesting and helpful discussion during your visit in Maastricht.
Vrienden zijn heel belangrijk voor de nodige ontspanning in de vrije tijd. Gezellige
etentjes, bezoekjes en feestjes genoeg.
Saskia, Ingrid, Susan, Ingrid, Irene en Argonde: door de vaste jaarclub
dinsdagavond hoefde ik het studentenleven nog niet helemaal los te laten terwijl ik
Thesis Jessie Steevens_v04.pdf
Dankwoord
203
wende aan het werkende leven. Hopelijk zien we elkaar nog vaak, waar dan ook in
Nederland, met of zonder de mannen!
Karin, Renske en Jeroen, Kim en Hans, met jullie is het altijd gezellig. Jullie zijn
altijd te vinden om iets te leuks te gaan doen. Bedankt ook voor de interesse in mijn
onderzoek. Op een vriendschap die nog lang mag duren!
Caroline en Ilona, superleuk dat jullie mijn paranimfen zijn! De lunches in het azM
waren altijd te kort om goed bij te kletsen... Ik verwacht het komende jaar trouwens
wel twee cocktailavonden ter gelegenheid van ‘het boekje af’ :-).
Papa en mama, vier jaar geleden wisten jullie niks over promoveren. Nu ietsje meer
denk ik. Leuk dat jullie iedereen zo trots vertellen over mijn onderzoek. Ik ben heel blij
dat ik jullie als ouders heb!
Bram en Anne, het is telkens lekker vertrouwd en gezellig met jullie, of we elkaar
nu in Breda, Nijmegen of Maastricht zien.
Yvonne en Rico, Bram en Vianna, Tessa, het is fijn om een schoonfamilie in de
buurt te hebben als je eigen familie ver weg woont.
Bram, supercool dat je deze mooie kaft voor me wilde ontwerpen!
Lieve Joep, wat ben ik toch gelukkig met jou! Je weet wel hoe belangrijk jij voor me
bent geweest tijdens mijn promotieonderzoek. Ik hou van je, monster van me! X
Jessie
Thesis Jessie Steevens_v04.pdf
204
Thesis Jessie Steevens_v04.pdf
Curriculum Vitae / List of publications
205
Thesis Jessie Steevens_v04.pdf
206
Thesis Jessie Steevens_v04.pdf
Curriculum Vitae
207
CURRICULUM VITAE
Jessie Steevens was born on December 14, 1981 in Breda, the Netherlands. She
completed secondary school (Gymnasium) at the Mencia de Mendoza lyceum in Breda
in 2000. That year, she started studying Health Sciences at Maastricht University. As a
part of this study, she fulfilled an internship at the Municipal Health Service (GGD) in
Rotterdam in 2004. In 2005, she obtained her Master’s degree in Health Sciences,
specialization Environmental Health Science. From January through May 2005, she
worked as employee environmental health at the Regional Health Service in Arnhem
(Hulpverlening Gelderland Midden). In September 2005, she started studying for a
Master’s degree in Epidemiology at Maastricht University. At the Department of
Epidemiology, Jessie fulfilled an internship on the topic of coffee and tea consumption
in relation to ovarian cancer risk, and in September 2006, she graduated (cum laude).
From August 2006 until August 2010, she worked on the PhD project entitled “Risk
factors for Barrett’s esophagus, adenocarcinoma of the esophagus and gastric cardia: a
prospective cohort study“, described in this thesis at the Department of Epidemiology,
GROW - School for Oncology and Developmental Biology, Maastricht University
Medical Centre +. Since October 2010 Jessie has worked at the Comprehensive Cancer
Centre Limburg (IKL) as an epidemiologist. Jessie is living with her husband Joep
Urlings.
Thesis Jessie Steevens_v04.pdf
208
LIST OF PUBLICATIONS
Submitted manuscripts
Steevens J, Schouten LJ, Goldbohm RA, Van den Brandt PA. Vegetables and fruits
consumption and risk of esophageal and gastric cancer subtypes in the Netherlands
Cohort Study.
Schouten LJ, Steevens J, Huysentruyt CJR, Coffeng C, Keulemans YCA, van
Leeuwen FE, Driessen ALC, van den Brandt PA. Cancer incidence and total and cause-
specific mortality in patients with Barrett’s esophagus.
Steevens J, Schouten LJ, Driessen ALC, Huysentruyt CJR, Keulemans YCA,
Goldbohm RA, Van den Brandt PA. A prospective cohort study on overweight, smoking,
alcohol consumption, and risk of Barrett's esophagus.
Peer reviewed publications
Steevens J, Schouten LJ, Driessen ALC, Huysentruyt CJR, Keulemans YCA,
Goldbohm RA, Van den Brandt PA. Toenail selenium status and the risk of Barrett’s
esophagus: the Netherlands Cohort Study. Cancer Causes Control 2010; in press
Steevens J, van den Brandt PA, Goldbohm RA, Schouten LJ. Selenium status and
the risk of esophageal and gastric cancer subtypes: the Netherlands cohort study.
Gastroenterology 2010;138:1704-13.
Steevens J, Schouten LJ, Goldbohm RA, van den Brandt PA. Alcohol consumption,
cigarette smoking and risk of subtypes of oesophageal and gastric cancer: a
prospective cohort study. Gut 2010;59:39-48.
Steevens J, Botterweck AA, Dirx MJ, van den Brandt PA, Schouten LJ. Trends in
incidence of oesophageal and stomach cancer subtypes in Europe. Eur J Gastroenterol
Hepatol 2010;22:669-78.
Steevens J, Schouten LJ, Verhage BA, Goldbohm RA, van den Brandt PA. Tea and
coffee drinking and ovarian cancer risk: results from the Netherlands Cohort Study and
a meta-analysis. Br J Cancer 2007;97:1291-4.
Thesis Jessie Steevens_v04.pdf
