Glycemic index, glycemic load, and pancreatic cancer risk (Canada)
Stephanie A.N. Silvera1,*, Thomas E. Rohan1, Meera Jain2, Paul D. Terry3, Geoffrey R. Howe4, Anthony
1Department of Epidemiology and Population Health, Albert Einstein College of Medicine, 1300 Morris Park Avenue,
Bronx, NY 10461, USA;
3Department of Epidemiology, Emory School of Public Health;4Department of Epidemiology, Mailman School of
Public Health, Columbia University, New York, NY
2Department of Public Health Sciences, University of Toronto, Toronto, Canada;
Received 27 August 2004; accepted in revised form 18 October 2004
Key words: glycemic index, glycemic load, pancreatic neoplasms, prospective cohart.
There is some evidence that plasma insulin and postload plasma glucose may be associated with risk of pancreatic
cancer. Glycemic index and glycemic load are measures, which allow the carbohydrate content of individual foods
to be classified according to their postprandial glycemic effects and hence their effects on circulating insulin levels.
Therefore, we examined pancreatic cancer risk in association with glycemic index (GI), glycemic load (GL), and
intake of dietary carbohydrate and sugar in a prospective cohort of 49,613 Canadian women enrolled in the
National Breast Screening Study (NBSS) who completed a self-administered food frequency questionnaire between
1980 and 1985. Linkages to national cancer and mortality databases yielded data on cancer incidence and deaths,
with follow-up ending between 1998 and 2000. During a mean 16.5 years of follow-up, we observed 112 incident
pancreatic cancer cases. There was no association between overall glycemic index, glycemic load, total carbohydrate
and total sugar intake and pancreatic cancer risk. In multivariate adjusted models, the hazard ratio (HR) for the
highest versus lowest quartile levels of overall GI and GL were 1.43 (95% confidence interval [CI]¼0.56–3.65,
Ptrend¼0.58) and 0.80 (95% CI¼0.45–1.41, Ptrend¼0.41), respectively. Our data suggest that overall glycemic
index and glycemic load, as well as total sugar and total carbohydrate intake, are not associated with pancreatic
cancer risk. However, given the limited literature regarding the role of diet in the etiology of pancreatic cancer,
particularly with respect to glycemic index/load, further investigation is warranted.
Relatively little is known about the etiology of pancre-
atic cancer. However, there is some evidence that plasma
insulin levels might be relevant . In an in vitro study,
Fisher et al.  demonstrated that high-affinity insulin
receptors are present in six pancreatic cell lines and also
reported a dose-dependent proliferative response in
these cell lines with increasing insulin concentrations.
This observation is in accord with evidence that insulin
is associated with the activation of mitogenic signals
that stimulate pancreatic cell proliferation . Further-
more, insulin receptor substrate-1 expression has been
found to be increased in human pancreatic cancer .
Additional support for a role for insulin comes from the
observation that postload plasma glucose levels were
positively associated with pancreatic cancer risk in a
cohort of men and women who were non-diabetic at
Glycemic index and glycemic load are measures which
can be used to classify the carbohydrate content of
individual foods according to their postprandial glyce-
mic effects and hence their effects on blood insulin levels
[5–8]. Using glycemic index values, the total glycemic
effect of the diet (glycemic load) can be estimated .
* Address Correspondence to: S.A.N. Silvera, Department of Epide-
miology and Population Health, Albert Einstein College of Medicine,
718-430-3038; Fax:+1-178-430-8653; E-mail address: ssilvera@aecom.
Cancer Causes and Control (2005) 16:431–436
? Springer 2005
Consumption of high GI diets has been associated with
hyperinsulinemia [5, 9, 10], while low GI diets have been
shown to be associated with a lower postprandial rise in
insulin , thus maintaining insulin sensitivity .
Therefore, glycemic index/load might be associated with
pancreatic cancer risk. To date, however, it appears that
the relationship between overall glycemic index and
glycemic load and pancreatic cancer risk has been
examined in only one epidemiological study . In that
investigation, a cohort study in women, there was no
association between overall glycemic index and pancre-
atic cancer risk, although there was a statistically non-
significant trend of increasing risk with increasing
glycemic load intake . Given the current lack of
data regarding these relationships, we examined the
association between overall glycemic index and glycemic
load, as well as total carbohydrate and sugar intake
(included because of their strong association with
postprandial insulin response ), and pancreatic
cancer risk in a cohort of Canadian women.
Materials and methods
The design of our study has been described in detail
elsewhere . Briefly, 89,835 women aged 40–59 years
were recruited into the Canadian National Breast Screen-
ing Study (NBSS) between 1980 and 1985 from the
general Canadian population . At recruitment into
the cohort, participants completed a self-administered
questionnaire which sought information on demographic
characteristics, lifestyle factors (including cigarette smok-
ing), menstrual and reproductive history, and use of oral
contraceptives and replacement estrogens. Starting in
1982, a self-administered food frequency questionnaire
(FFQ)  was distributed to all new attendees at all
screening centers, and to women returning to the
screening centers for re-screening. The FFQ sought
information on usual portion size and frequency of
consumption of 86 food items, and included photographs
of various portion sizes to assist respondents with
quantifying intake. A comparison between the self-
administered questionnaire and a full interviewer-admin-
istered questionnaire, which has been subjected to both
validity and reliability testing  and used in a number
of epidemiologic studies , revealed that the two
methods gave estimates of intake of the major macro-
nutrients and dietary fiber which were moderately to
strongly correlated with each other (reported correlation
coefficients ranged from 0.47 to 0.72) . A total of
49,613 dietary questionnaires were returned and available
Glycemic index values of foods were obtained from
published reports based on studies in North America .
Overall glycemic index was calculated by multiplying the
carbohydrate content of a given food item by the
number of servings per day of that food item and its
glycemic index value, summing over all food items
reported, and dividing by total carbohydrate intake.
Total dietary glycemic load was calculated by multiply-
ing the carbohydrate content of a given food item by the
number of servings consumed per day and its glycemic
index value and summing the values for all food items
reported. When the reported glycemic index values for
foods were observed to vary across studies  we used
the mean of the reported values of glycemic index for
Data from the completed self-administered food
frequency questionnaires were used to estimate daily
intake of nutrients using a database for Canadian foods
that has been described elsewhere . Briefly, nutrient
values, including total carbohydrate, total fiber, and
sugar intakes were calculated by using a data bank
based on Handbook No. 8 of the US Department of
Agriculture , modified and expanded for Canadian
foods. The variable for total sugar was calculated by
summing the nutrient intake values for galactose,
glucose, fructose, lactose, sucrose, and maltose.
Data from the food frequency questionnaire were also
used to estimate overall glycemic index and glycemic
load. Glycemic index values of foods were obtained
from published reports based on studies in North
America . Overall glycemic index was calculated by
multiplying the carbohydrate content (in grams) of a
given food item by the number of servings per day of
that food item and its glycemic index value, summing
over all food items reported, and dividing by the total
carbohydrate in the diet. Total dietary glycemic load
was calculated by multiplying the carbohydrate content
of a given food item by the number of servings
consumed per day and its glycemic index value and
summing the values for all food items reported. Each
unit increase in glycemic load represents the insulin
response to the equivalent of 1 gram of glucose or
carbohydrate from white bread (depending on the
standard used) . When the reported glycemic index
values for foods were observed to vary across studies ,
we used the mean of the reported values of glycemic
index for that food. The main foods contributing to
glycemic load in the cohort are listed in a footnote to
Incident cases of pancreatic cancer and deaths from
all causes were ascertained respectively by means of
computerized record linkages to the Canadian Cancer
Database and to the National Mortality Database, both
432S.A.N. Silvera et al.
of which are maintained by Statistics Canada. The
linkages to the databases yielded data on cancer
incidence and mortality to 31 December 2000 for
women in Ontario, 31 December 1998 for women in
Quebec, and 31 December 1999 for women in other
provinces. There is evidence that the use of record
linkages to ascertain incident cancer cases and deaths in
Canada is both complete and accurate [21, 22].
Of the 46,613 women who had returned a FFQ and
had provided information on energy intake, we excluded
88 women with extreme energy intake values (at least
three standard deviations above or below the mean
value for logecaloric intake). Additionally, we excluded
one woman with prevalent pancreatic cancer at baseline.
Analyses were thus based on 49,111 women for whom
complete information on all covariates was available.
Cox proportional hazards models (using age as the
time scale) were used to estimate hazard ratios (HR) and
95% confidence intervals (CI) for the association
between energy-adjusted quartile levels of glycemic
load, overall glycemic index, total carbohydrates,
sucrose, fructose, and total sugar, and pancreatic cancer
risk; energy adjustment was performed using the resid-
ual method . For these analyses, cases contributed
person-time to the study from their date of enrollment
until the date of diagnosis of their pancreatic cancer,
and non-cases contributed person-time from their date
of enrollment until the termination of follow-up (the
date to which cancer incidence data were available for
women in the corresponding province) or death,
whichever occurred earlier. Multivariate models in-
cluded body mass index [defined as weight (kg)/height
(m2); weight and height were measured at baseline ],
self-reported alcohol consumption (frequency of con-
sumption of beer, wine, and liquor) and smoking history
(in pack years, defined as the number of cigarettes per
day multiplied by how many years they reported
smoking), energy intake (kcal/day), study center and
randomization group. To test for trend we fitted the
median value of each quartile as an ordinal variable in
the risk models, and evaluated the statistical significance
of the coefficient using the Wald test . All analyses
were performed using SAS version 8 (SAS Institute
The average duration of follow-up for cohort members
was 16.5 (809,492 person-years), during which 112 cases
of pancreatic cancer were diagnosed. The mean (S.D.)
age at diagnosis for the cases was 61.7 (±7.3) years. For
the cohort as a whole, the means (±SD) of the energy-
adjusted overall glycemic index and glycemic load were
79.4 (±24.5) and 147.2 (±35.1) g/day, respectively.
There was a two-fold variation in mean glycemic load
values between the lowest and highest quintile levels
(Table 1). Compared to those with relatively low glyce-
mic load values, women with high glycemic load values
reported lower alcohol consumption, fewer pack-years
Table 1. Baseline distributions of pancreatic cancer risk factors by quintiles of energy-adjusted glycemic load (GL)
Quartiles of energy-adjusted Glycemic Load (g/day)
<125 (n = 12281)125–147 (n = 12281) 148–169 (n = 12280) >169 (n = 12281)
Mean overall glycemic index
Mean glycemic load (g/day)
Total carbohydrate (g/day)
Total sugar (g/day)
Total fiber (g/day)
Mean energy intake (kcal/day)
Mean age (years)
Mean BMI (kg/m2)
Pack-years of smoking (Mean)
Mean alcohol intake (g/day)
Some vigorous physical activity (%)
HRT use (% ever)
aThe main foods contributing to glycemic load in the cohort include white bread (sliced), rolls, muffins, potatoes (baked, boiled, and mashed),
French fries, cakes, cookies, rice, pasta, pizza, cold breakfast cereals, pies and tarts, cola, other soft drinks, citrus fruits and juices and other fruits,
crisp snacks (such as potato chips or popcorn), candy, chocolate, peas, beans and lentils, hot breakfast cereals, dark and wholegrain breads, corn,
root vegetables other than potatoes, jam, jelly and honey, sugar in tea or coffee, ice cream, and peanut butter.
bNumbers in parentheses represent the standard deviation.
Glycemic index, glycemic load and pancreatic cancer risk433
total carbohydrates, sugar and fiber (Table 1). No
appreciable variation was observed in use of hormone
replacement therapy (HRT), participation in vigorous
physical activity, mean body mass index (BMI), or
parity by quartile levels of glycemic load. The patterns
for overall glycemic index were similar to those for the
glycemic load (data not shown).
there was no association between glycemic load, overall
glycemic index, fructose consumption, or total sugar
smoking,and consumedmore totalcalories,
intake and risk of pancreatic cancer. Although there was
an inverse association of statistical significance between
age- and energy-adjusted total carbohydrate intake and
pancreatic cancer risk (HR¼0.49, 95% CI¼0.25–0.94,
Ptrend¼0.05) and between sucrose consumption and
pancreatic cancer risk (HR¼0.55, 95% CI¼0.31–0.97,
Ptrend¼0.02), additional adjustment for other potential
confounders attenuated the associations and rendered
them statistically non-significant. Additional adjustment
for total fiber consumption had essentially no impact on
the hazard ratios (data not shown).
Table 2. Adjustedahazard ratios and 95% confidence intervals (CI) for the association between quartiles of overall glycemic index, glycemic load,
total carbohydrate, total sugar, and total fiber and risk of pancreatic cancer
Cases/person-years Hazard ratio (95% CI)
Age- & energy adjustedMultivariate adjusted
Overall glycemic index
P for trend
Glycemic Load (g/day)
P for trend
Total Carbohydrate (g/day)
P for Trend
Total Sugar (g/day)
P for Trend
P for Trend
P for Trend
aMultivariable models included age (as time to event variable), BMI in kg/m2(<25, 25–29, >30), alcohol (zero plus four levels of intake),
smoking (pack-years, quartiles), parity (nulliparity plus 3 levels for parous), energy intake (continuous), study center, and randomization group.
434S.A.N. Silvera et al.
High glycemic index diets are associated with increased
insulin secretion [5, 9, 10, 26], which has been shown to
promote pancreatic cancer cell growth in vitro , and
has been hypothesized to affect pancreatic cancer risk by
several mechanisms, including alteration of cell cycle
kinetics (insulin facilitates the transit of cells through the
G1 phase of the cell cycle) , inhibition of apoptosis
, and down-regulation of insulin-like growth factor
binding protein 1 (IGFBP-1) [30, 31].
In the prospective study reported here, there was no
association between either overall glycemic index or
glycemic load and pancreatic cancer risk during a
16-year follow-up period. Although a statistically
significant inverse association was found with total
carbohydrate consumption in the age- and energy-
adjusted model, this association was attenuated in the
multivariate model. By using a prospective study design,
recall bias was avoided, but misclassification of dietary
intake may have attenuated the observed associations. A
limitation of our study is the use of a one-time dietary
assessment that may not have been representative of the
dietary habits of the study participants over the course
of follow-up. In addition, the small number of pancre-
atic cancer cases precluded any subgroup analyses. As
well, the lack of statistical significance in our multivar-
iate models may be due to a lack of power due to the
small number of cases.
Only two prospective studies have previously exam-
ined the association between glycemic index/load, die-
tary sugar, or dietary fiber and pancreatic cancer risk.
Stolzenberg-Solomon et al. , analyzed data from the
Alpha-Tocopherol, Beta-Carotene Cancer Prevention
Study cohort in Finland, in which 163 incident pancre-
atic cancer cases were observed during 13 years of
follow-up, and reported a statistically significant inverse
trend of pancreatic cancer risk with carbohydrate
consumption among male smokers. This study popula-
tion is clearly different from our cohort of Canadian
women and therefore these results are not directly
comparable to ours. In contrast, Michaud et al. ,
(n¼88,802), in which 180 incident pancreatic cancer
cases were found during 18 years of follow-up . Our
overall results are similar to those of Michaud et al. ,
in that there was no association overall with either
glycemic index or glycemic load, but we were unable to
confirm their finding of a statistically significant trend of
increasing risk with increasing glycemic load among
women with a body mass index of at least 25 kg/m2,
because the relatively small number of cases in our study
Michaud et al., we did not collect data on diabetes,
and therefore were unable to adjust for it in the analysis.
However, although there is evidence from epidemiologic
studies to support a causal association between diabetes
mellitus and pancreatic cancer , it is not immediately
obvious that failure to adjust for diabetes would have
confounded the associations observed here. Specifically,
in the 1980’s dietary recommendations for diabetics
stressed high intake of complex carbohydrates and fiber
and low fat consumption [34, 35]. These recommenda-
tions were similar to those commonly promoted at that
time for the population as a whole . Hence, it is
unlikely that there were any substantial differences in
dietary patterns for diabetics and non-diabetics during
the period in which dietary data were collected for the
present study. Nevertheless, if those diagnosed with
diabetes had altered their diet to (for example) include
more foods with low glycemic index values, then it is
possible that our inability to adjust for diabetes could
have obscured a positive association between glycemic
index/load and pancreatic cancer risk.
Evidence regarding the role of diet in the etiology of
pancreatic cancer is limited , particularly with
respect to the role of glycemic index/load. However,
given that diet is a potentially modifiable exposure, the
possibility that high glycemic load diets might be
associated with increased pancreatic cancer risk war-
rants further investigation.
We thank Statistics Canada, the provincial and territo-
rial Registrars of Vital Statistics, and the Cancer
Registry directors for their assistance in making the
cancer incidence and mortality data available.
1. Ekbom A, Hunter D (2002) Pancreatic cancer. In: Adami H,
Hunter D, and Trichopoulos D, eds. Textbook of Cancer Epide-
miology. New York, NY: Oxford University Press.
2. Fisher WE, Laszlo GB, Schirmer WJ (1996) Insulin promotes
pancreatic cancer: evidence for endocrine influence on exocrine
pancreatic tumors. J Surg Res 63: 310–313.
3. Gupta K, Krishnaswamy G, Karnad A, Peiris A (2002) Insulin: a
novel factor in carcinogenesis. Am J Med Sci 232: 140–145.
4. Gapstur SM, Gann PH, Lowe W, et al. (2000) Abnormal glucose
metabolism and pancreatic cancer mortality. J Am Med Assoc 283:
5. Holt SH, Miller JC, Petocz P (1997) An insulin index of foods: the
insulin demand generated by 1000-kJ portions of common foods.
Am J Clin Nutr 66: 1264–1276.
Glycemic index, glycemic load and pancreatic cancer risk 435
6. Ionescu-Tirgoviste C, Popa E, Sintu E, et al. (1983) Blood glucose Download full-text
and plasma insulin responses to various carbohydrates in type 2
(non-insulin-dependent diabetes). Diabetologia 24: 80–84.
7. Jenkins DJ, Kendall CW, Augustin LS, et al. (2002) Glycemic
index: overview of implications in health and disease. Am J Clin
Nutr 76: 266S–273S.
8. Ludwig DS (2002) The glycemic index: physiological mechanisms
relating to obesity, diabetes, and cardiovascular disease. J Am Med
Assoc 287: 2414–2423.
9. Foster-Powell K, Holt SH, Brand-Miller JC (2002) International
table of glycemic index and glycemic load values: 2002. Am J Clin
Nutr 76: 5–56.
10. Byrnes SE, Miller JCB, Denyer GS (1995) Amylopectin starch
promotes the development of insulin resistance in rats. J Nutr 125:
11. Ludwig DS (2000) Dietary glycemic index and obesity. J Nutr 130:
12. Plymate SR, Jones RE, Matej LA, Friedl KA (1988) Regulation of
sex hormone binding globulin (SHBG) production in Hep G2 cells
by insulin. Steroids 52: 339–340.
13. Michaud DS, Liu S, Giovannucci E, et al. (2002) Dietary sugar,
glycemic load, and pancreatic cancer risk in a prospective study.
J Natl Cancer Inst 94: 1293–1300.
14. Jenkins D, Wolever TM, Buckley G, et al. (1988) Low-glycemic-
index starchy foods in the diabetic diet. Am J Clin Nutr 48:
15. Terry P, Jain M, Miller AB, Howe GR, Rohan TE (2002) Dietary
carotenoid intake and colorectal cancer risk. Nutr Cancer 42:
16. Miller AB, Baines CJ, To T, Wall C (1992) Canadian National
Breast Screening Study. I. Breast cancer detection and death rates
among women aged 40–49 years. II. Breast cancer detection and
death rates among women aged 50–59 years. Can Med Assoc J 147:
17. Jain MG, Harrison L, Howe GR, Miller AB (1982) Evaluation of a
self-administered dietary questionnaire for use in a cohort study.
Am J Clin Nutr 36: 931–935.
18. Morgan AB, Jain M, Miller AB, et al. (1978) A comparison of
dietary methods in epidemiologic studies. Am J Epidemiol 107:
19. Watt BK, Merrill AL. (1963, expansion March 1972) Composition
of foods: raw, processed, prepared. USDA Handbook No. 8.
Washington, DC: Agriculture Research Service.
20. Liu S, Manson JE, Stampfer MJ, et al. (2001) Dietary glycemic
load assessed by food-frequency questionnaire in relation to
plasma high-density-lipoprotein cholesterol and fasting plasma
triacylglycerols in postmenopausal women. Am J Clin Nutr 73:
21. Robles SC, Marrett LD, Clarke EA, Risch HA (1988) An
application of capture-recapture methods to the estimation of
completeness of cancer registration. J Clin Epidemiol 41: 495–501.
22. Shannon HS, Jamieson E, Walsh C, et al. (1989) Comparison
of individual follow-up and computerized record linkage using
the Canadian Mortality Data Base. Can J Public Health 80:
23. Willett WC, Stampfer MJ (1986) Total energy intake: implications
for epidemiologic analyses. Am J Epidemiol 124: 17–27.
24. Jain M, Miller AB (1997) Tumor characteristics and survival of
breast cancer patients in relation to premorbid diet and body size.
Breast Cancer Res Treat 42: 43–55.
25. Rothman K, Greenland S (1998) Modern Epidemiology 2nd edn.
Philadelphia (PA): Lippincott-Raven; p. 319.
26. Kaaks R, Lukanova A (2002) Effects of weight control and
physical activity in cancer prevention: role of endogenous hormone
metabolism. Ann N Y Acad Sci 963: 268–281.
27. Fisher WE, Boros LG, Schirmer WJ (1996) Insulin promotes
pancreatic cancer: evidence for autocrine and paracrine roles.
J Surg Res 63: 310–313.
28. Gross G, Boldt DH, Osborne CK (1984) Perturbation by insulin
of human breast cancer cell kinetics. Cancer Res 44: 3570–
29. Geier A, Beery R, Haimshon M, Hemi R, Lunenfeld B (1992)
Serum and insulin inhibit cell death induced by cycloheximide in
the human breast cancer cell line MCF-7. In Vitro Cell Dev Biol
30. Musey VC, Goldstein S, Farmer PK, Moore PB, Phillips LS
(1993) Differential regulation of IGF-1 and IGF-binding protein-1
by dietary composition in humans. Am J Med Sci 305: 131–
31. Suikkari AM, Koivisto VA, Rutanen EM, et al. (1988) Insulin
regulates the serum levels of low molecular weight insulin-like
growth factor-finding protein. J Clin Endocrinol Metab 66:
32. Stolzenberg-Solomon RZ, Pietinen P, Taylor PR, Virtamo J,
Albanes D (2002) Prospective study of diet and pancreatic cancer
in male smokers. Am J Epidemiol 155: 783–192.
33. Michaud DS (2004) Epidemiology of pancreatic cancer. Minerva
Chir 59: 99–111.
34. American Diabetes Association (1979) Principles of nutrition and
dietary recommendations for individuals with diabetes mellitus.
Diabetes 28: 1027.
35. (1982) Dietary recommendations for diabetics for the 1980s–a
policy statement by the British Diabetic Association: prepared by
the Nutrition Sub-Committee of the British Diabetic Association’s
Medical Advisory Committee. Hum Nutr Appl Nutr 36: 378, 380,
36. (1986) Statement on nutrition and your health: dietary guidelines
for Americans. J Am Diet Assoc 86: 107–108.
37. World Cancer Research Fund (1997) Food, Nutrition and the
Prevention of Cancer: A Global Perspective. Washington DC:
American Institute for Cancer Research.
436S.A.N. Silvera et al.