Trends in colorectal cancer incidence rates in New Zealand, 1981-2004.

Abstract

Incidence rates of colorectal cancer (CRC) in New Zealand rank among the highest worldwide. Internationally, there has been evidence of a shift in colon cancer from left- to right-sided. The objective of this study was to determine trends in left- and right-sided colon and rectal cancers incidence by sex, age and ethnicity.
Using datasets created by linking data from the New Zealand Cancer Registry to the census data, we analysed a total of 47,694 CRCs from 1981 to 2004. Cancers were divided into right-sided colon (cecum to the splenic flexure); left-sided colon (descending and sigmoid colon); and rectal (rectosigmoid junction and rectum).
Left- and right-sided colon, and rectal cancer incidence rates increased by 13-20% among men. In women, colon cancer rates increased by 25% for right-sided cancers, decreased by 8% for left-sided cancers and remained unchanged for rectal cancers. This corresponds with an increase in right-sided cancers from 57% to 65% of total colon cancers in women. The incidence of all CRCs increased at a faster rate among Māori than non-Māori.
We identified a left- to right-sided shift in colon cancer limited to women over the age of 65. While Māori trends in site distribution parallel those of their non-Māori counterparts, the rapid increase in Māori incidence rates is noteworthy. It is unclear why such shifts in CRC site distribution are occurring.

Full text

Page 1

Trends in colorectal cancer incidence rates in New Zealand, 1981–2004
Ankit B. Shah,*† Diana Sarfati,† Tony Blakely,† June Atkinson† and Elizabeth R. Dennett‡
*Department of Public Health and Community Medicine, Tufts University, School of Medicine, Boston, Massachusetts, USA
†Department of Public Health, University of Otago, Wellington, and
‡Department of Surgery and Anesthesia, Wellington School of Medicine and Health Sciences, Wellington, New Zealand
Key words
colorectal neoplasm, epidemiology, incidence, subsite,
trend.
Correspondence
Dr Diana Sarfati, Department of Public Health,
University of Otago, PO Box 7343, Wellington 6242,
New Zealand. Email: diana.sarfati@otago.ac.nz
A. B. Shah MD, MPH; D. Sarfati MBChB, MPH,
FNZCPHM; T. Blakely MBChB, MPH, PhD, FNZCPHM;
J. Atkinson NZC Statistics; E. R. Dennett MBChB,
MMedSci, MAppMgt, FRACS.
Accepted for publication 6 May 2011.
doi: 10.1111/j.1445-2197.2011.05995.x
Abstract
Background: Incidence rates of colorectal cancer (CRC) in New Zealand rank among
the highest worldwide. Internationally, there has been evidence of a shift in colon
cancer from left- to right-sided. The objective of this study was to determine trends in
left- and right-sided colon and rectal cancers incidence by sex, age and ethnicity.
Methods: Using datasets created by linking data from the New Zealand Cancer
Registry to the census data, we analysed a total of 47 694 CRCs from 1981 to 2004.
Cancers were divided into right-sided colon (cecum to the splenic flexure); left-sided
colon (descending and sigmoid colon); and rectal (rectosigmoid junction and rectum).
Results: Left- and right-sided colon, and rectal cancer incidence rates increased by
13–20% among men. In women, colon cancer rates increased by 25% for right-sided
cancers, decreased by 8% for left-sided cancers and remained unchanged for rectal
cancers. This corresponds with an increase in right-sided cancers from 57% to 65% of
total colon cancers in women. The incidence of all CRCs increased at a faster rate
among Ma ¯ori than non-Ma ¯ori.
Conclusion: We identified a left- to right-sided shift in colon cancer limited to women
over the age of 65. While Ma ¯ori trends in site distribution parallel those of their
non-Ma ¯ori counterparts, the rapid increase in Ma ¯ori incidence rates is noteworthy. It is
unclear why such shifts in CRC site distribution are occurring.
Introductionans_59951..7
Worldwide, colorectal cancer (CRC) is one of the most common
cancers.1Furthermore, the incidence and mortality rates of CRC in
New Zealand rank among the highest in the world.1,2Age, personal
history of polyps, family history of CRC and inflammatory bowel
disease are strong, albeit non-modifiable risk factors for the devel-
opment of CRC. Data on modifiable risk factors are inconsistent
but there is good evidence for the role of level of physical inac-
tivity,3,4obesity5and diabetes mellitus (DM)6,7in the development
of CRC.
Exploration into the evolution of CRC has led some,8–15but not
all,16–18to conclude that over time there has been a shift in
CRC incidence to the proximal colon, the so-called left- to right-
sided shift. This finding has kindled research on site-specific aeti-
ologies and risk factors. The proposed hypotheses suggest that
the occurrence of CRC at a particular anatomic site may be associ-
ated with distinctive risk factor profiles within particular geographic
and demographic groups.19–21As a result, studies have attempted to
stratify risk by sex, age, ethnicity and income, and have reported a
right-sided shift among low-income populations,12Caucasians,12
African Americans,19,22males,13females8,12,13and/or the elderly.12,13
Beyond clues from socio-demographic variations in right-sided
shift, studies have suggested that type 2 DM in females,7cigarette
smoking,23prior cholecystectomy24,25and physical inactivity3,4
increase the risk for right-sided colon cancer. Although not compel-
ling, there is evidence that alcohol consumption,26type 2 DM in
males27and dietary habits28are risk factors for left-sided cancers. In
summary, the findings in relation to site-specific risk factors have
been inconsistent, and to date, no risk factor has been identified that
explains the patterns in CRC site distribution.
The aim of this study is to determine whether there has been a
right-sided shift in CRC in New Zealand from 1981 to 2004. Using
cancer registry data linked to the census information in a high CRC
ORIGINAL ARTICLE
ANZJSurg.com
© 2012 The Authors
ANZ Journal of Surgery © 2012 Royal Australasian College of SurgeonsANZ J Surg •• (2012) ••–••

Page 2

incidence population, we investigate whether patterns in anatomic
tumour location have varied by gender, age and ethnicity.
Methods
The datasets were created through linking records from the New
Zealand Cancer Registry (NZCR) and records from the five yearly
census of population and dwellings. The NZCR is a population-
based cancer registry that collects information on all malignant
tumours with the exception of non-melanotic skin cancers. Five
closed non-overlapping cohorts were created of the New Zealand
usual resident population (all ages) on census night 1981, 1986,
1991, 1996, 2001, followed up for incident cancer(s) until the sub-
sequent census or in the case of the 2001 cohort, until December 31,
2004 (the most recent data available at the time of the study’s record
linkage). Exact time periods are as follows: March 25, 1981 to
March 4, 1986; March 5, 1986 to March 5, 1991; March 6, 1991 to
March 5, 1996; March 6, 1996 to March 6, 2001; and lastly, March
7, 2001 to December 31, 2004. Cohorts were created using proba-
bilistic record linkage software (QualityStage™; IBM Corp.,
Armonk, NY, USA) to anonymously link census and cancer registry
records within a geographic area (meshblock or census area unit) on
sex, date of birth, ethnicity and country of birth. Further details on
the linkage methods, process and outputs are detailed elsewhere.29
Briefly, 73.2% (1981–1986) to 81.7% (2001–2004) of eligible
cancer registry records were linked to a census record, with 95.2%
(1981–1986) to 96.9% (2001–2004) of these linked census-cancer
records estimated to be true links. To adjust for underestimation of
rates using the linked datasets, and to correct for linkage bias
whereby the percentage of eligible cancer records linked varied by
socio-demographic characteristics, we calculated weights for strata
based on age, sex, ethnicity, small area deprivation and time since
census. All analyses presented in this report use these weights,
and they have been shown to satisfactorily adjust for linkage
bias.30
Approval was granted under the Statistics New Zealand Data
Integration Policy, and the Wellington Ethics Committee granted
ethics approval for CancerTrends (Ref 04/10/093).
Variables
Exposures
Ethnicity was classified as Ma ¯ori and non-Ma ¯ori, which includes
Pacific, Asian and European/Other. By the 2006 census, the New
Zealand population was over 4 million; 14% classified as Ma ¯ori, 7%
Pacific, 9%Asian, 65% New Zealand European and 11% identifying
with other ethnic groups (although largely of European descent).
Outcome
A CRC recorded in the NZCR was required to be included in this
study. CRCs (ICD 10 code C18-20) were categorized into right-
sided (cecum (C18.0), ascending colon (C18.2), hepatic flexure
(C18.3), transverse colon (C18.4) and splenic flexure (C18.5)), left-
sided (descending colon (C18.6) and sigmoid colon (C18.7)) and
rectal (rectosigmoid junction (C19) and rectum (C20)). Those coded
as unspecified (C18.9) or overlapping/appendix (C18.8) were
excluded.19A total of 47 694 CRC are analysed in this paper.
Analyses
Data were restricted to adults (25+ years) who were in their usual
residence on census night.All analyses were conducted in SAS v8.2
(SAS Institute Inc., Cary, NC, USA). Person time was censored after
development of the first CRC with a specific three-digit ICD code.
Standardized rates and 95% confidence intervals (CIs) were calcu-
lated for sex, age and ethnicity using direct standardization to the age
structure of the 2001 World Health Organization World Standard
population.
Regression analysis was used to determine changes in incidence
rates over time, per unit of time of 5 years, by weighted regression
of the standardized rates over time (weights = inverse of the variance
of the standardized rate).
Results
Table 1 shows the number of left- and right-sided colon, and rectal
cancers and person-years of follow-up by time period for each sex.
The overall incidence of CRC in New Zealand continues to increase.
In men, there has been an increase in incidence in all sites: left-sided
(19% from 1981–1986 to 2001–2004, or from our regression analy-
sis, an increased rate of 1.13 per 100 000 per 5 years (95% CI 0.37
to 1.88)); right-sided (22%, or increased rate of 1.32 per 100 000 per
5 years (95% CI 0.77 to 1.87)); and rectal cancer (13%, or an
increased rate of 1.19 per 100 000 per 5 years (95% CI 0.56 to
1.82)). There is no evidence of a shift in colon tumours from left- to
right-sided among men (Fig. 1).
In women, there has been an increase in incidence of right-sided
(25%, or 1.60 per 100 000 per 5 years (95% CI 1.38 to 1.82)), but a
decrease in left-sided cancers (-8%, or -0.38 per 100 000 per 5
years (95% CI -0.70 to -0.05)). Incidence of rectal cancers has
remained unchanged (Fig. 1). The non-overlapping CIs and diver-
gent trends of left- and right-sided cancers indicate statistical sig-
nificance. Accordingly, the percentage of all colon cancers that are
right-sided increased monotonically from 57% to 65% among
women.
Age
Asexpected,age-standardizedincidenceratesamongthoseages65+
were considerably higher than the 25–64 age group at all sites
(Table 2). Risk of CRC was stable or decreased for all sites over time
for the younger age group in both sexes, although the only statisti-
cally significant trend in this group was for left-sided cancers in
females (18% reduction, P = 0.01).
For men aged 65+ years, there was a significant increase at all
sites (52%, P = 0.02; 35%, P = 0.03; and 32%, P = 0.01 for
right-sided colon, left-sided colon and rectal cancers, respectively)
without evidence of a shift to right-sided cancers. Among females
65+, there was a 58% (P < 0.01) increase in right-sided cancers, a
non-significant increase in left-sided cancers and stable rates of
rectal cancers which is consistent with a right-sided shift.
2
Shah et al.
© 2012 The Authors
ANZ Journal of Surgery © 2012 Royal Australasian College of Surgeons

Page 3

Ethnicity
Table 3 shows the age-standardized incidence rates of CRC for
males and females by ethnicity. Despite having considerably lower
incidence rates of CRC than non-Ma ¯ori at each site for all time
periods, Ma ¯ori men and women had a higher relative increase in
incidence over the study period. The difference was particularly
marked for left-sided disease among males with relative increases of
11% (P = 0.07) for non-Ma ¯ori and 127% (P = 0.06) for Ma ¯ori males,
and rectal disease with relative increases of 6% (P = 0.10) for
non-Ma ¯ori and 82% (P < 0.01) for Ma ¯ori males.
Significant increases were seen in right-sided cancers in both
non-Ma ¯ori (21%, P < 0.01) and Ma ¯ori women (66%, P = 0.04),
compared with non-significant changes in left-sided cancers (-12%,
P = 0.07 and 16%, P = 0.13 for non-Ma ¯ori and Ma ¯ori, respectively)
and rectal cancers (-1%, P = 0.69 and 39%, P = 0.30, respectively).
Consequently, right-sided tumours increased as a percentage of all
colonic tumours for both non-Ma ¯ori (58% to 66%) and Ma ¯ori
women (45% to 54%), albeit the latter increase being non-
monotonic and erratic. Non-Ma ¯ori men had relative increases in
incidence at all sites, with no change in site distribution. Ma ¯ori men
had larger increases in incidence rates at all sites relative to non-
Ma ¯ori men but lacked statistical precision, thus precluding any con-
clusion of shift in tumour location.
Discussion
From our study of 47 694 CRC from 1981 to 2004 in an unscreened
New Zealand population,31,32we found an increase in CRC for both
sexes, restricted to those 65+ years of age.While there is no evidence
for a left- to right-sided shift in men, we identify a right-sided shift
in both non-Ma ¯ori and Ma ¯ori women aged 65+ years.
We found an increasing incidence of CRC for those aged 65+, but
no change or a reduction in incidence among the younger popula-
tion. This is consistent with a birth cohort analysis of CRC in New
Zealand which concluded that the incidence rates of CRC have been
declining for successive cohorts born from 1943 to 1953.33The high
CRC incidence rates seen in the 65+ age group may be the result of
early-life exposures and risk factors, although how such early-life
risk factors would give rise to a right-sided shift, and differentiate by
sex, is unresolved.
The only other New Zealand study that investigated a shift in CRC
reported a right-sided shift in both sexes that was more pronounced
in the elderly. However, this was a small study that only compared
data from two consecutive 4-year time periods and likely does not
accurately represent the trends over time.10
A number of international population-based studies have
reported a left- to right-sided shift in CRC among both older men
and women.10–12,14,15For example, in Norway, Larsen and Bray
found evidence of a left- to right-sided shift in both sexes seen
among the older population (55–74 years).15These authors also
carried out a birth cohort analysis and concluded that there have
been rapid increases in right-sided colon cancer among generations
born from 1900 with stabilization of colon cancer at both sites in
generations born after 1955. Similarly, Singh et al. analysed 40
years of colon cancer data from Canada, and reported an increase in
rates of right-sided colon cancer in both sexes, largely among the
older population (>70 years).14Saltzstein and Behling investigated
CRC incidence from 1988 to 2003 in California and reported a left-
to right-sided shift in both sexes, but more pronounced among
women.12
Fewer studies have investigated site distribution by ethnicity.
Cucino et al. found increased proportions in right-sided cancers
among black males and whites of both sexes from the Department of
VeteransAffairs database and increased right-sided cancers for both
sexes in blacks and whites from the SEER database of the National
Cancer Institute.11On the other hand, a study by Demers et al.
looking at 22 years of CRC data from Michigan, USA, reported a
right-sided shift only in African Americans (both sexes) and con-
cluded that such shifts may be related to ethnic-specific changes in
environmental exposures.19
It is not clear why there may be variations in CRC subsite trends
over time and between countries or regions. One possibility is that
differential screening uptake may impact site distribution. One study
Table 1 Number of colorectal cancers and person-years of follow-up by site and sex for those aged 25+ years (47 694 total colorectal cancers)
Males
Left colon
Females
Left colonRight colonRectal Right colonRectal
1981–1986†
Total cancers
Person-years
1986–1991
Total cancers
Person-years
1991–1996
Total cancers
Person-years
1996–2001
Total cancers
Person-years
2001–2004
Total cancers
Person-years
1 2151 0741 6861 662 1 1491 242
4 309 024 4 309 5124 307 8814 556 6714 557 967 4 557 666
1 392 1 236 1 8811 8931 272 1 284
4 662 273 4 662 5634 661 070 4 972 9804 974 2094 974 302
1 5481 377 2 0942 1181 248 1 455
4 937 9504 938 3874 936 6515 343 5375 345 6305 345 132
1 7581 6232 3342 4661 4161 569
5 304 0415 304 3275 302 5665 810 3855 812 8885 812 460
1 6471 3261 956 2 3071 122 1 344
4 131 2984 131 7894 130 6554 606 380 4 608 5274 608 177
†Time periods are non-overlapping; please refer to Methods section for exact dates.
Colorectal cancer trends in New Zealand
3
© 2012 The Authors
ANZ Journal of Surgery © 2012 Royal Australasian College of Surgeons

Page 4

suggests that polypectomies have led to a disproportionate decrease
in incidence of left-sided cancers and therefore an increased propor-
tion of right-sided cancers.12There is also some evidence that
colonoscopy differentially reduces mortality from cancers on the left
compared with right.34Such theories do not help explain our find-
ings, as there is no significant utilization of screening,31,32and cer-
tainly no clear increased utilization by women in New Zealand.
Another source for inconsistent findings is the variation in cat-
egorization of left and right colon. For example, Vobecky et al. used
a classification that included five groups – caecum, ascending colon,
transverse and descending colon (as one group), sigmoid colon, and
rectum – while Demers et al. used three groups – right colon from
caecum to splenic flexure, left colon from descending to sigmoid
colon, and rectum.8,19Such variation in categorization may result in
contradicting conclusions even if identical data are used. To allow
for reproducibility and comparison between studies, it would be
useful to standardize the categorization of left and right colon.
However, such classification errors are presumably non-differential
by sex, age and ethnicity, and therefore would not affect patterns of
varying site distribution by these demographic groups.
Strengths and limitations
This study allowed us to investigate CRC incidence for the entire
New Zealand population over a 24-year period. The resulting com-
prehensive database enabled us to identify trends and populations at
risk for CRC. By using data from the entire New Zealand (census
night) population as well as data from NZCR, it was possible to
overcome undercounting of Ma ¯ori on cancer registry data.35
Fig. 1. Trends in age-standardized inci-
dence rates of colorectal cancer per
100 000 person-years (95% CI) by site
for adults 25+ years.
4
Shah et al.
© 2012 The Authors
ANZ Journal of Surgery © 2012 Royal Australasian College of Surgeons

Page 5

However, the study has limitations. Statistical imprecision
becomes a problem when stratifying by ethnicity. It was not possible
to link all cancer records back to the census, and so weighting for
linkage bias was undertaken; we are confident that this will have
adjusted for bias due to misclassification of the cancer outcome. The
Cancer RegistryAct, which came into effect in 1994, required man-
datory registration of all cancers, and there was an increase in CRC
registrations following its introduction, and probably reflects
improved notification or may reflect a ‘holding over’of registrations
from one year to the next but this probably impacted ethnic groups,
Table 2 Trends in age-standardized right- and left-sided colon, and rectal cancer incidence rates per 100 000 person-years (95% CI) by sex and age
MalesFemales
25–6465+
25–6465+
Right colon
1981–1986†
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
Left colon
1981–1986
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
Rectal
1981–1986
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
12.5 (11.3–13.7)
13.6 (12.4–14.9)
12.8 (11.6–13.9)
13.2 (12.1–14.3)
11.3 (10.2–12.3)
-10
0.36
86.9 (79.8–94.0)
86.9 (78.9–94.9)
96.2 (89.4–103.0)
110.0 (103.0–117.0)
132.0 (123.0–142.0)
52
0.02*
14.8 (13.5–16.1)
15.1 (13.7–16.5)
15.5 (14.3–16.8)
14.8 (13.6–16.0)
12.9 (11.7–14.1)
-13
0.22
90.9 (83.3–98.5)
97.9 (91.1–105.0)
106.0 (98.0–113.0)
123.0 (116.0–130.0)
144.0 (135.0–152.0)
58
<0.01
11.6 (10.4–12.7)
13.9 (12.7–15.2)
13.2 (12.1–14.4)
13.0 (12.0–14.1)
11.8 (10.7–12.8)
2
0.92
72.8 (65.5–80.0)
73.9 (67.4–80.4)
83.0 (75.7–90.2)
102.0 (95.0–109.0)
98.1 (90.9–105.0)
35
0.03
13.1 (11.7–14.5)
13.1 (11.9–14.4)
12.2 (11.1–13.4)
12.0 (11.0–13.1)
10.8 (9.8–11.8)
-18
0.01
59.7 (53.7–65.6)
58.8 (53.4–64.2)
55.9 (50.3–61.6)
62.9 (57.9–67.8)
64.9 (58.4–71.3)
9
0.25
19.3 (17.6–20.9)
21.9 (20.3–23.5)
20.9 (19.4–22.5)
22.5 (21.0–24.0)
18.9 (17.6–20.3)
-2
0.83
112.0 (103.0–121.0)
114.0 (105.0–123.0)
134.0 (125.0–144.0)
132.0 (124.0–141.0)
148.0 (138.0–158.0)
32
0.01
13.0 (11.8–14.2)
12.1 (10.9–13.3)
12.9 (11.7–14.1)
13.3 (12.2–14.4)
13.1 (11.9–14.2)
1
0.39
73.1 (63.8–82.4)
62.7 (56.6–68.8)
71.7 (65.3–78.1)
73.5 (67.7–79.3)
74.5 (68.3–80.6)
2
0.25
*Significant values (P < 0.05) indicated in bold font. †Time periods are non-overlapping; please refer to Methods section for exact dates.
Table 3 Trends in age-standardized right- and left-sided colon, and rectal cancer incidence rates per 100 000 person-years (95% CI) by ethnicity (25+ years)
Non-Ma ¯ori Ma ¯ori
MaleFemale MaleFemale
Right colon
1981–1986†
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
Left colon
1981–1986
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
Rectal
1981–1986
1986–1991
1991–1996
1996–2001
2001–2004
% change
P (trend)
26.4 (24.9–27.9)
27.7 (26.2–29.2)
28.2 (26.8–29.6)
29.3 (27.9–30.6)
31.4 (30.0–32.9)
19
<0.01*
29.8 (28.3–31.3)
30.6 (29.1–32.0)
31.9 (30.4–33.3)
33.9 (32.5–35.3)
36.2 (34.7–37.8)
21
<0.01
12.3 (7.9–16.7)
10.7 (5.7–15.8)
10.7 (6.5–14.8)
18.0 (13.7–22.3)
19.5 (14.6–24.4)
59
0.11
8.7 (5.0–12.4)
12.9 (8.5–17.3)
13.2 (8.7–17.7)
14.2 (10.3–18.1)
14.4 (10.7–18.2)
66
0.04
23.1 (21.7–24.5)
24.7 (23.3–26.0)
25.2 (23.9–26.6)
26.8 (25.5–28.1)
25.7 (24.4–27.0)
11
0.07
21.5 (20.2–22.8)
22.1 (20.7–23.4)
20.1 (18.9–21.3)
20.6 (19.5–21.7)
19.0 (17.8–20.1)
-12
0.07
8.1 (3.8–12.5)
12.3 (7.4–17.2)
14.5 (8.9–20.1)
23.6 (18.1–29.1)
18.4 (13.9–22.9)
127
0.06
10.7 (6.7–14.8)
9.0 (5.3–12.7)
10.8 (6.5–15.0)
11.5 (8.7–14.3)
12.4 (8.7–16.1)
16
0.13
36.2 (34.5–37.9)
37.5 (35.8–39.3)
38.1 (36.4–39.8)
39.3 (37.7–40.9)
38.2 (36.5–39.8)
6
0.10
23.0 (21.7–24.4)
21.4 (20.1–22.6)
22.0 (20.8–23.3)
22.6 (21.4–23.8)
22.8 (21.6–24.1)
-1
0.69
17.6 (11.2–24.0)
20.1 (14.7–25.4)
27.8 (21.3–34.3)
31.1 (23.7–38.5)
32.0 (26.0–38.0)
82
<0.01
10.5 (5.4–15.5)
11.3 (7.3–15.3)
19.4 (13.4–25.4)
19.2 (14.9–23.4)
14.6 (10.9–18.2)
39
0.30
*Significant values (P < 0.05) indicated in bold font. †Time periods are non-overlapping; please refer to Methods section for exact dates.
Colorectal cancer trends in New Zealand
5
© 2012 The Authors
ANZ Journal of Surgery © 2012 Royal Australasian College of Surgeons