Search tips
Search criteria 


Logo of amjepidLink to Publisher's site
Am J Epidemiol. 2010 June 1; 171(11): 1174–1182.
Published online 2010 May 7. doi:  10.1093/aje/kwq061
PMCID: PMC2915491

Available Carbohydrates, Glycemic Load, and Pancreatic Cancer: Is There a Link?


High-carbohydrate diets have been linked to pancreatic cancer risk in case-control studies, but prospective studies have shown mostly null results. The authors investigated the associations of glycemic load, glycemic index, and carbohydrate intake with pancreatic cancer risk in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. Dietary intake was assessed by using a self-administered questionnaire. Between 1998 and 2006 (median follow-up = 6.5 years), 266 incident, confirmed pancreatic cancers were identified among 109,175 participants. Hazards ratios and 95% confidence intervals were adjusted for sex, smoking, body mass index, and total energy. Overall, elevated risks for pancreatic cancer were observed in the 90th versus 10th percentile of glycemic load (hazards ratio (HR) = 1.45, 95% confidence interval (CI): 1.05, 2.00), available carbohydrate (HR = 1.47, 95% CI: 1.05, 2.06), and sucrose (HR = 1.37, 95% CI: 0.99, 1.89) intake. The positive association for available carbohydrate intake was observed during the first 4 years of follow-up (HR<2 years = 2.60, 95% CI: 1.34, 5.06; HR2–<4 years = 1.94, 95% CI: 1.06, 3.55) but not subsequently (HR = 0.86, 95% CI: 0.52, 1.44); the opposite pattern was observed for total fat and saturated fat intake. Rather than being causal, the short-term increase in pancreatic cancer risk associated with high available carbohydrate and low fat intake may be capturing dietary changes associated with subclinical disease.

Keywords: diet, dietary carbohydrates, dietary fats, glycemic index, pancreatic neoplasms, prospective studies

Individuals who are obese or have type 2 diabetes have an elevated risk of pancreatic cancer, which may be attributable to higher circulating concentrations of insulin during the early stages of type 2 diabetes (13). Insulin receptors are expressed in human pancreatic cancers, and insulin is a known pancreatic tumor promoter (3). Recent results from prospective epidemiologic studies provide direct evidence that higher fasting (46) and postprandial (7) glucose levels are associated with greater pancreatic cancer risk, even among nondiabetic persons (6, 7). Consequently, behavioral and dietary factors that influence insulin secretion and susceptibility to diabetes may contribute to the development of pancreatic cancer. A high glycemic load diet, characterized by relatively high intake of carbohydrate-containing foods that are digested and absorbed quickly to induce sharp or prolonged rises in blood glucose (8), has been found to increase the risk of diabetes in some epidemiologic studies (9, 10) and, consequently, may also increase the risk for pancreatic cancer.

However, of the 7 prospective studies examining the association between glycemic load and pancreatic cancer (1117), a positive association between a high dietary glycemic load and pancreatic cancer risk was observed in only 1 study and was statistically significant only in participants who were overweight and inactive (11). On the other hand, several case-control studies (1821), although not all (2225), showed positive associations between dietary carbohydrates and pancreatic cancer risk. In 1 case-control study, positive associations for carbohydrates, including mono- and disaccharides, were attenuated after restricting the analysis to participants whose food habits remained stable during the previous 10 years (18).

As dyspepsia is a common symptom of pancreatic cancer that is exacerbated with the consumption of fatty foods (26), study participants with symptoms of pancreatic cancer, either diagnosed or undiagnosed, may replace fat with carbohydrates in their diet to help alleviate these symptoms, perhaps accounting for some of the inconsistencies in epidemiologic studies of diet and pancreatic cancer. However, the issue of timing between self-reported dietary intake and pancreatic cancer diagnosis has not been evaluated in detail in previous epidemiologic studies. Lagged analyses in cohort studies of diet and pancreatic cancer may be necessary to rule out the short-term influence of subclinical disease on dietary intake.

We investigated the overall and time-dependent associations of dietary glycemic load, glycemic index, and total and specific carbohydrate with the risk of pancreatic cancer in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial.


Study population

The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial is an ongoing clinical trial designed to investigate whether screening tests reduce mortality from these types of cancer. The design and objectives of the trial have been described in detail elsewhere (27). Briefly, between 1993 and 2001, 149,939 men and women aged 55–74 years without a prior history of prostate, lung, colon, rectal, or ovarian cancer were recruited from 1 of 10 US centers (Birmingham, Alabama; Denver, Colorado; Detroit, Michigan; Honolulu, Hawaii; Marshfield, Wisconsin; Minneapolis, Minnesota; Pittsburgh, Pennsylvania; Salt Lake City, Utah; St. Louis, Missouri; Washington, DC) and then randomized to either the intervention (regular screening of the 4 main cancer sites by prostate-specific antigen tests and digital rectal examination (prostate cancer), chest radiograph (lung cancer), flexible sigmoidoscopy (colorectal cancer), and cancer antigen 125 screening and transvaginal ultrasound (ovarian cancer)) or control (usual care) arm of the trial. All eligible participants provided written, informed consent. The study was approved by all participating institutional review boards and the National Cancer Institute.

At baseline, a total of 111,916 participants completed a self-administered demographic and lifestyle questionnaire and a food frequency questionnaire. This study population included subjects who had fewer than 8 missing responses on the food frequency questionnaire, had no history of pancreatic cancer at baseline, and accrued follow-up time. We then excluded participants who had missing (n = 1,246) or extreme (greater than twice the interquartile range above the 75th or below the 25th percentile after transforming to a normal distribution) values for body mass index (n = 735) or total energy (n = 760). After these exclusions, 109,175 participants comprised the analytical study population.

Data collection


At the time of enrollment (October 1993–September 2001), all participants were asked to complete a baseline questionnaire that included questions on general demographics, occupational history, personal and family medical history, lifestyle habits, and history of screening for prostate, lung, colorectal, and ovarian cancer. Dietary data were collected by using a self-administered, validated food frequency questionnaire, the Diet History Questionnaire, version 1 (DHQ) (28), which was distributed to the intervention and control arms of the trial between 1998 and 2005 and solicited information on the frequency and portion size of 124 foods and supplements during the past year (29). The median time between completion of the baseline questionnaire and the DHQ was 3.0 years (interquartile range = 2.9–4.0). The DHQ has been validated against a reference instrument (four 24-hour recalls) (30). In a calibration study, correlations between the DHQ and dietary recalls for total energy, energy-adjusted protein, carbohydrates, and fat were approximately 0.5, 0.6, 0.65, and 0.65, respectively (30).

Glycemic measures.

The glycemic index measures the percent incremental area under the 2-hour postprandial glucose response curve for consumption of a given carbohydrate-containing food relative to the corresponding area under the curve for an equivalent amount of carbohydrate from a reference food, such as glucose or white bread. Glycemic index values were added to the DHQ database by linking 1,300 glycemic index values compiled from published and unpublished laboratory reports through 2001 to individual foods within the 225 US Department of Agriculture and Continuing Survey of Food Intakes by Individuals (CSFII) food groups that form the basis of the DHQ nutrient database (31). An overall glycemic load value was calculated for each participant by multiplying the glycemic index (divided by 100) and average daily gram intake of available carbohydrates (carbohydrates available for digestion and absorption or total carbohydrates minus fiber) for each food included in the DHQ and summing this product across all foods. An overall glycemic index value was assigned to each participant by dividing his/her overall glycemic load value by that person's average daily gram intake of available carbohydrates and multiplying this number by 100 (31).

Outcome assessment and follow-up

Incident pancreatic cancers were ascertained through the use of annually mailed, self-administered study questionnaires or death certification information and were subsequently confirmed by pathology reports or medical record abstraction. The definition of a pancreatic cancer case was limited to primary pancreatic adenocarcinomas (International Classification of Diseases for Oncology, Third Edition, codes C250–C259) and excluded endocrine pancreatic tumors (histology types 8150, 8151, 8153, 8155, 8240, 8246, 8502, and 8520). Follow-up began at the date of randomization, completion of the baseline questionnaire, or completion of the DHQ, whichever came last, and ended at pancreatic cancer diagnosis, death, or most recent National Death Index search date (December 31, 2006). Although they were not considered to be a case in this analysis, participants who reported having a diagnosis of pancreatic cancer that was not subsequently confirmed ended follow-up on their self-reported date of cancer diagnosis.

Statistical analysis

The nutrient density method was used to adjust for total energy by dividing gram intakes of glycemic load and nutrients by total energy and multiplying this value by 1,000. The glycemic index was not energy adjusted because it was not correlated with total energy in this study (r = 0.04). Spearman correlation coefficients were used to assess the correlation among glycemic index, energy-adjusted glycemic load, and energy-adjusted carbohydrate intakes.

Cox proportional hazards models were used to calculate hazards ratios and 95% confidence intervals for glycemic index, glycemic load, available carbohydrates, and specific carbohydrate intakes in relation to pancreatic cancer risk. Attained age was used as the underlying time metric and, in this way, all models were adjusted for age. Dietary intakes were first included in the models as categorical variables, with quintile cutpoints based on the distribution of the entire cohort. The dietary variables were also included in models as continuous values. First, Box-Cox transformations were used to normalize the distributions of dietary exposures. These normally distributed values were then divided by the difference between the 90th and 10th percentiles. This scaling produced estimated glycemic loads comparable to the estimated glycemic load between the first and fifth quintiles in analyses using dietary exposure categories. Modeling the exposures on the continuous scale allowed for greater statistical precision compared with the categorical analysis. We considered using higher-order (up to cubic) effects of the transformed nutrient density variables, but tests of these higher-order effects showed no significant improvement in the fit of the models. Multivariable models included sex, body mass index (<18.5, 18.5–24.9, 25–29.9, ≥30 kg/m2), and cigarette smoking status (never smoked cigarettes, quit smoking ≥2 years ago, recent/current smoking <20 cigarettes per day, currently smoking ≥20 cigarettes per day). Additionally, we adjusted for normally transformed total energy, except in models in which the main exposure was glycemic index. Because information on physical activity was missing for the majority of participants (53.0%), we did not include physical activity as a covariate in the models. Alcohol intake was not included in the models as it was not associated with pancreatic cancer risk in this cohort. In categorical models, tests for trend were conducted by treating the quintile of intake (or glycemic index) as a continuous variable with values 1–5; the Wald test was used to evaluate statistical significance. Self-reported diabetics were excluded in sensitivity analyses because of the potential for dietary modification after diagnosis of this disease.

For variables with missing values for race (<0.1%), education (<0.1%), diabetes (4.1%), and cigarette smoking status (<0.1%), we imputed the mean or mode (for categorical variables) based on the distribution in the cohort. Participants who were missing values for number of cigarettes smoked (<0.2%) were assigned the mean conditioned on smoking status. Use of a missing exposure category in adjusted models yielded nearly identical results.

To evaluate effect modification, we fit continuous exposure models and examined the results within strata of sex, body mass index (<30 and ≥30 kg/m2), and smoking status (noncurrent or current smokers). We tested the statistical significance of interactions between each transformed exposure of interest and the stratification variable in multivariable models using the Wald test associated with each cross-product term.

Schoenfeld residuals and graphical methods were used to evaluate the assumption of proportional hazards. There was some evidence that the proportional hazards assumption was violated. To account for this, we conducted a time-varying analysis examining the associations of continuous, transformed intakes of total and specific dietary carbohydrates with pancreatic cancer stratified by follow-up period. We tested the time interaction by including follow-up period as a stratification variable and adding a cross-product term for nutrient value and follow-up period to the model; the Wald test was used to evaluate statistical significance. Because we expected that pancreatic cancer cases diagnosed soon after baseline may have changed the composition of their diet, whereby carbohydrates were substituted in place of fat in the diet, we also examined the time-dependent associations for fat intake and pancreatic cancer risk.

All analyses were conducted using STATA, version 9.2, software (StataCorp LP, College Station, Texas). All tests were 2 sided and considered statistically significant at α < 0.05.


A total of 667,734 person-years of follow-up (median, 6.5 and up to 9 years) were contributed by 109,175 participants between January 1998 and December 2006. During follow-up, 266 participants (162 men and 104 women) reported a diagnosis of pancreatic cancer that was subsequently confirmed by medical record abstraction.

Energy-adjusted glycemic load was more strongly correlated with energy-adjusted available carbohydrate than with glycemic index. Spearman correlation coefficients for glycemic load and glycemic index were 0.36 for men and 0.28 for women, while these values for glycemic load and available carbohydrates were 0.95 and 0.93 for men and women, respectively. The glycemic index was not strongly correlated with available carbohydrate (for men, r = 0.08; for women, r = −0.05). The top food sources of glycemic load were desserts (13.1%), milk (9.2%), fruits (8.2%), pasta/rice/cooked cereal (6.8%), candy (6.6%), regular soft drinks (5.9%), and fruit juices/drinks (5.5%). The ranges of body mass index and total energy values were 15.7–45.6 kg/m2 and 338–5,545 kcal/day, respectively.

Compared with participants in the lowest quintile of available carbohydrate intake, participants in the highest quintile had a higher proportion of females, non-Hispanic blacks, and Asian/Pacific Islanders, as well as a lower proportion of non-Hispanic whites and Hispanics. Participants in the highest quintile were also generally older at study entry, more likely to be never smokers, had lower body mass indexes, spent more time doing vigorous physical activity, and consumed lower intakes of total energy, total and saturated fat, protein, red meat, and alcohol (Table 1). A higher intake of available carbohydrates was also associated with higher glycemic load and intake of specific carbohydrates (starch, sucrose, and fructose), fiber, and folate (Table 1).

Table 1.
Characteristics of the Participants According to Quintiles of Dietary Available Carbohydrate, the Prostate, Lung, Colorectal, and Ovarian Cancer Trial (n = 109,175), 1998–2006

Glycemic loads for pancreatic cancer according to quintiles of glycemic load, glycemic index, available carbohydrate, and specific carbohydrate intake are presented in Table 2. Sex did not modify any of the associations (P > 0.05); therefore, we present our results using the total study population. Higher glycemic load (for high vs. low quintile, hazards ratio (HR) = 1.41, 95% confidence interval (CI): 0.97, 2.07; Ptrend = 0.03) and available carbohydrate intake (HR = 1.56, 95% CI: 1.06, 2.30; Ptrend = 0.004) were associated with an increased risk of pancreatic cancer. Of the specific carbohydrates that we examined, a similar positive association was observed for sucrose, with a hazards ratio of 1.55 (95% CI: 1.06, 2.27) comparing the highest with lowest quintile of intake, although the test for trend was not statistically significant (Ptrend = 0.12). Glycemic index and intakes of starch and fructose were not associated with pancreatic cancer. The results were similar after excluding self-reported diabetic persons (Table 2).

Table 2.
Hazard Ratios and 95% Confidence Intervals for Dietary Carbohydrate and Pancreatic Cancer, the Prostate, Lung, Colorectal, and Ovarian Cancer Trial (n = 109, 175), 1998–2006

We did not observe statistically significant interactions of the dietary exposures and pancreatic cancer risk by smoking status, body mass index, or diabetes.

Table 3 shows the hazards ratios comparing the 90th and 10th percentiles of normally distributed glycemic load, glycemic index, available carbohydrate, and specific carbohydrate intakes by follow-up period (<2, 2–<4, and ≥4 years after baseline) after excluding self-reported diabetic persons. The magnitudes of the positive associations for continuous glycemic load (90th percentile vs. 10th percentile, HR = 1.45, 95% CI: 1.05, 2.00) and available carbohydrates (HR = 1.47, 95% CI: 1.05, 2.06) were similar to results in which we compared the fifth versus first quintile (Table 2). The association for available carbohydrates was strongest in the first 2 years of follow-up (HR = 2.60, 95% CI: 1.34, 5.06), slightly attenuated in the 2–<4-year follow-up period (HR = 1.94, 95% CI: 1.06, 3.55), and null after excluding the first 4 years of follow-up (HR = 0.86, 95% CI: 0.52, 1.44); these differences were statistically significant (Pinteraction = 0.002). We observed similar trends in the associations for glycemic load, sucrose, fructose, saturated fat, and total fat, in which the associations were strongest in the first 2 years and attenuated with subsequent follow-up. The ratio of available carbohydrates to saturated fat intake was also strongly positively associated with pancreatic cancer risk in the first 2 years (HR = 3.09, 95% CI: 1.58, 6.02) and attenuated after excluding the first 4 years of follow-up (HR = 0.88, 95% CI: 0.53, 1.46; Pinteraction = 0.001). The magnitudes of the associations were similar, although slightly more attenuated in early follow-up, before excluding self-reported diabetic persons (Web Table 1). (This information is described in the first of 2 supplementary tables; each is referred to as “Web Table” in the text and is posted on the Journal's website ( Results using 1-, 2-, 3-, 4-, and 5-year lags showed a similar trend toward stronger relative risks in earlier versus later follow-up (Web Table 2).

Table 3.
Hazards Ratiosa and 95% Confidence Intervals for Dietary Carbohydrate and Fat and Pancreatic Cancer Excluding Self-reported Diabetics With Results Stratified by Follow-up Period, the Prostate, Lung, Colorectal, and Ovarian Cancer Trial (n = 101,690), ...


Higher dietary intakes of glycemic load, available carbohydrates, and sucrose were associated with greater risk of pancreatic cancer, while dietary glycemic index, starch, and fructose intakes were not associated with risk in this cohort. However, the positive associations we observed for glycemic load, available carbohydrates, and sucrose were limited to the first 4 years of follow-up. No association was found for carbohydrate intake in the later follow-up period. We also observed the reverse for total fat intake, particularly saturated fat, whereby an inverse association was observed in the earlier, but not later, follow-up period.

Although high glycemic load diets have a biologically plausible role in the development of pancreatic cancer, epidemiologic studies have shown consistent null results. In the Nurses’ Health Study, a significant dose-response association for dietary glycemic load was observed only among women who were physically inactive and overweight (RR = 2.67, 95% CI: 1.02, 6.99) (11). No statistically significant associations between glycemic load and pancreatic cancer were observed in subsequent prospective studies, including analyses restricted to physically inactive and/or overweight participants (1217). In the Cancer Prevention Study-II cohort, a nonsignificant increased risk among the high-body mass index/high-sedentary behavior group compared with the low-risk group was observed for dietary carbohydrates (RR = 2.25, 95% CI: 0.93, 5.45) but not glycemic load (RR = 1.12, 95% CI: 0.46, 2.70), and no associations were observed for glycemic load or carbohydrates in the overall cohort (15). In the Multi-Ethnic Cohort Study, sucrose intake, but not glycemic load, increased the risk of pancreatic cancer among overweight and obese participants (14). No clear patterns have emerged from prospective studies examining the relation between sugary foods and pancreatic cancer risk (32, 33).

In our study, we observed a positive association for glycemic load and available carbohydrates with pancreatic cancer risk, but this association was restricted to the first 4 years of follow-up. Although our findings for glycemic load, available carbohydrates, and specific carbohydrates could be explained by a promotional effect of high carbohydrate consumption on the development of pancreatic cancer in the few years prior to diagnosis, our findings of an inverse association between fat and pancreatic cancer that became null with longer follow-up would suggest an alternative explanation. Similar to our study when a 4-year lag was applied to the Nurses’ Health Study, no associations were observed for glycemic load, glycemic index, or carbohydrates (11). As most food frequency questionnaires query intake during the past 12 months, assessment of dietary intake in epidemiologic studies of pancreatic cancer may capture disease-related changes in diet, rather than intake prior to cancer symptoms which may be more germane to pancreatic cancer development. Because there are no accurate methods for diagnosing early stage disease, pancreatic cancer patients typically present in late disease stages with severe, though fairly nonspecific symptoms, including diabetes, fatigue, jaundice, abdominal pain, weight loss, early satiety, nausea, and vomiting (34); however, dyspepsia associated with hiccups, flatulence, and regurgitation may precede more severe symptoms by several months (34). Dyspepsia in general may be attributed to the delayed gastric emptying that often accompanies pancreatic cancer (35). Because high-fat diets may exacerbate symptoms of dyspepsia (26), disease-related gastrointestinal symptoms may have been a source of bias in previous case-control studies that found an increased risk of pancreatic cancer with high intakes of carbohydrate and/or low intakes of fat (1821, 25, 36). In our study, total fat, particularly saturated fat, was inversely associated with pancreatic cancer in early but not later follow-up, suggesting that short-term decreases in fat intake may be a consequence, rather than a cause, of pancreatic cancer. Although prospective studies are less prone to this type of temporal bias sometimes termed “reverse causation,” our results suggest that such bias can occur in prospective studies of pancreatic cancer, given the cancer's unknown latency, and that differences in associations by follow-up time should be considered in studies assessing dietary intake within the period directly preceding cancer diagnosis.

The major strengths of this study include its prospective design, which eliminates differential recall between pancreatic cancer cases and noncases, and the large number of male and female participants. Relying on self-reported pancreatic cancer alone may result in substantially attenuated relative risks because of misclassification of the outcome (37), but limiting the definition of a pancreatic cancer case to ductal adenocarcinoma confirmed through medical record abstraction most likely reduced the potential for this bias. There are some limitations as well. As with most other prospective studies, by assessing dietary intake during the previous 12 months our study was designed to evaluate the association of recent, rather than lifelong, diet on pancreatic cancer risk and, therefore, may not capture the most etiologically relevant time period. Much of the information collected from self-administered questionnaires may be reported with error, and changes in diet, smoking, and body mass index over follow-up were not captured, which would attenuate the relative risk and does not explain the positive or inverse associations that were observed.

In conclusion, in this prospective study, an elevated risk of pancreatic cancer was associated with a high glycemic load and available carbohydrate intake only among participants with less than 4 years of follow-up. The association of total and saturated fat, in turn, was inversely associated with pancreatic cancer in the earlier, but not later, follow-up period. This finding suggests that participants who were diagnosed with incident pancreatic cancer earlier during follow-up may have been symptomatic at the time the food frequency questionnaire was completed, altering their diets to include more easily digestible carbohydrates and less fat. Future prospective studies examining recent diet and pancreatic cancer risk should consider the potential influence of preclinical disease on self-reported intake.


Author affiliations: Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health, Department of Health and Human Services, Rockville, Maryland (Cari L. Meinhold, Li Jiao, Richard B. Hayes, Rachael Z. Stolzenberg-Solomon); Surveillance Research Program, Division of Cancer Control and Population Sciences, National Cancer Institute, Rockville, Maryland (Kevin W. Dodd); Division of Epidemiology and Community Health and Masonic Cancer Center, University of Minnesota, Minneapolis, Minnesota (Andrew Flood); Division of Preventive Medicine, School of Medicine, University of Alabama at Birmingham, Birmingham, Alabama (James M. Shikany); and Department of Epidemiology, Mailman School of Public Health, New York, New York (Jeanine M. Genkinger).

Conflict of interest: none declared.



confidence interval
Diet History Questionnaire
hazards ratio


1. Berrington de Gonzalez A, Sweetland S, Spencer E. A meta-analysis of obesity and the risk of pancreatic cancer. Br J Cancer. 2003;89(3):519–523. [PMC free article] [PubMed]
2. Huxley R, Ansary-Moghaddam A, Berrington de González A, et al. Type-II diabetes and pancreatic cancer: a meta-analysis of 36 studies. Br J Cancer. 2005;92(11):2076–2083. [PMC free article] [PubMed]
3. Fisher WE, Boros LG, Schirmer WJ. Insulin promotes pancreatic cancer: evidence for endocrine influence on exocrine pancreatic tumors. J Surg Res. 1996;63(1):310–313. [PubMed]
4. Stolzenberg-Solomon RZ, Graubard BI, Chari S, et al. Insulin, glucose, insulin resistance, and pancreatic cancer in male smokers. JAMA. 2005;294(22):2872–2878. [PubMed]
5. Jee SH, Ohrr H, Sull JW, et al. Fasting serum glucose level and cancer risk in Korean men and women. JAMA. 2005;293(2):194–202. [PubMed]
6. Stattin P, Björ O, Ferrari P, et al. Prospective study of hyperglycemia and cancer risk. Diabetes Care. 2007;30(3):561–567. [PubMed]
7. Gapstur SM, Gann PH, Lowe W, et al. Abnormal glucose metabolism and pancreatic cancer mortality. JAMA. 2000;283(19):2552–2558. [PubMed]
8. Jenkins DJ, Wolever TM, Taylor RH, et al. Glycemic index of foods: a physiological basis for carbohydrate exchange. Am J Clin Nutr. 1981;34(3):362–366. [PubMed]
9. Salmerón J, Manson JE, Stampfer MJ, et al. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA. 1997;277(6):472–477. [PubMed]
10. Salmerón J, Ascherio A, Rimm EB, et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care. 1997;20(4):545–550. [PubMed]
11. Michaud DS, Liu S, Giovannucci E, et al. Dietary sugar, glycemic load, and pancreatic cancer risk in a prospective study. J Natl Cancer Inst. 2002;94(17):1293–1300. [PubMed]
12. Johnson KJ, Anderson KE, Harnack L, et al. No association between dietary glycemic index or load and pancreatic cancer incidence in postmenopausal women. Cancer Epidemiol Biomarkers Prev. 2005;14(6):1574–1575. [PubMed]
13. Silvera SA, Rohan TE, Jain M, et al. Glycemic index, glycemic load, and pancreatic cancer risk (Canada) Cancer Causes Control. 2005;16(4):431–436. [PubMed]
14. Nöthlings U, Murphy SP, Wilkens LR, et al. Dietary glycemic load, added sugars, and carbohydrates as risk factors for pancreatic cancer: the Multiethnic Cohort Study. Am J Clin Nutr. 2007;86(5):1495–1501. [PubMed]
15. Patel AV, McCullough ML, Pavluck AL, et al. Glycemic load, glycemic index, and carbohydrate intake in relation to pancreatic cancer risk in a large US cohort. Cancer Causes Control. 2007;18(3):287–294. [PubMed]
16. Henen MM, Verhage BA, Lumey L, et al. Glycemic load, glycemic index, and pancreatic cancer risk in the Netherlands Cohort Study. Am J Clin Nutr. 2008;87(4):970–977. [PubMed]
17. Jiao L, Flood A, Subar AF, et al. Glycemic index, carbohydrates, glycemic load, and the risk of pancreatic cancer in a prospective cohort study. Cancer Epidemiol Biomarkers Prev. 2009;18(4):1144–1151. [PMC free article] [PubMed]
18. Bueno de Mesquita HB, Moerman CJ, Runia S, et al. Are energy and energy-providing nutrients related to exocrine carcinoma of the pancreas? Int J Cancer. 1990;46(3):435–444. [PubMed]
19. Howe GR, Jain M, Miller AB. Dietary factors and risk of pancreatic cancer: results of a Canadian population-based case-control study. Int J Cancer. 1990;45(4):604–608. [PubMed]
20. Zatonski W, Przewozniak K, Howe GR, et al. Nutritional factors and pancreatic cancer: a case-control study from south-west Poland. Int J Cancer. 1991;48(3):390–394. [PubMed]
21. Silverman DT, Swanson CA, Gridley G, et al. Dietary and nutritional factors and pancreatic cancer: a case-control study based on direct interviews. J Natl Cancer Inst. 1998;90(22):1710–1719. [PubMed]
22. Baghurst PA, McMichael AJ, Slavotinek AH, et al. A case-control study of diet and cancer of the pancreas. Am J Epidemiol. 1991;134(2):167–179. [PubMed]
23. Kalapothaki V, Tzonou A, Hsieh CC, et al. Nutrient intake and cancer of the pancreas: a case-control study in Athens, Greece. Cancer Causes Control. 1993;4(4):383–389. [PubMed]
24. Ghadirian P, Simard A, Baillargeon J, et al. Nutritional factors and pancreatic cancer in the francophone community in Montréal, Canada. Int J Cancer. 1991;47(1):1–6. [PubMed]
25. Ji BT, Chow WH, Gridley G, et al. Dietary factors and the risk of pancreatic cancer: a case-control study in Shanghai China. Cancer Epidemiol Biomarkers Prev. 1995;4(8):885–893. [PubMed]
26. Pilichiewicz AN, Feltrin KL, Horowitz M, et al. Functional dyspepsia is associated with a greater symptomatic response to fat but not carbohydrate, increased fasting and postprandial CCK, and diminished PYY. Am J Gastroenterol. 2008;103(10):2613–2623. [PubMed]
27. Prorok PC, Andriole GL, Bresalier RS, et al. Design of the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial. Control Clin Trials. 2000;21 (6 suppl):273S–309S. [PubMed]
28. Diet History Questionnaire. Risk factor monitoring and methods. Bethesda, MD: National Cancer Institute; 2009. ( (Accessed May 10, 2009)
29. Subar AF, Midthune D, Kulldorff M, et al. Evaluation of alternative approaches to assign nutrient values to food groups in food frequency questionnaires. Am J Epidemiol. 2000;152(3):279–286. [PubMed]
30. Subar AF, Thompson FE, Kipnis V, et al. Comparative validation of the Block, Willett, and National Cancer Institute food frequency questionnaires: the Eating at America's Table Study. Am J Epidemiol. 2001;154(12):1089–1099. [PubMed]
31. Flood A, Subar AF, Hull SG, et al. Methodology for adding glycemic load values to the National Cancer Institute Diet History Questionnaire database. J Am Diet Assoc. 2006;106(3):393–402. [PubMed]
32. Larsson SC, Bergkvist L, Wolk A. Consumption of sugar and sugar-sweetened foods and the risk of pancreatic cancer in a prospective study. Am J Clin Nutr. 2006;84(5):1171–1176. [PubMed]
33. Wang F, Herrington M, Larsson J, et al. The relationship between diabetes and pancreatic cancer [electronic article] Mol Cancer. 2003;2(1):4. [PMC free article] [PubMed]
34. Holly EA, Chaliha I, Bracci PM, et al. Signs and symptoms of pancreatic cancer: a population-based case-control study in the San Francisco Bay area. Clin Gastroenterol Hepatol. 2004;2(6):510–517. [PubMed]
35. Barkin JS, Goldstein JA. Diagnostic and therapeutic approach to pancreatic cancer. Biomed Pharmacother. 2000;54(7):400–409. [PubMed]
36. Nkondjock A, Krewski D, Johnson KC, et al. Specific fatty acid intake and the risk of pancreatic cancer in Canada. Br J Cancer. 2005;92(5):971–977. [PMC free article] [PubMed]
37. Silverman DT, Schiffman M, Devesa S. Diagnostic certainty in pancreatic cancer. J Clin Epidemiol. 1996;49(5):601–603. [PubMed]

Articles from American Journal of Epidemiology are provided here courtesy of Oxford University Press