Anthropometric Measures, Body Mass Index, and Pancreatic Cancer:  A Pooled Analysis From the Pancreatic Cancer Cohort Consortium (PanScan) FREE

Pancreatic adenocarcinoma is the fourth leading cause of cancer death in the United States¹ and is responsible for about 227 000 deaths per year worldwide.² Because of the lack of effective screening tests for pancreatic cancer, it is often diagnosed at an advanced stage, contributing to a 5-year survival rate that is less than 5%.³ The incidence of pancreatic cancer is higher in men than in women, and in the United States, it is higher in blacks than in whites.³ Smoking, diabetes mellitus, and a family history of pancreatic cancer are known risk factors,^4- 5 but these factors do not account for all the cases of pancreatic cancer.

Obesity and high body mass index (BMI [calculated as weight in kilograms divided by height in meters squared]) have been proposed as additional risk factors for pancreatic cancer. Prospective studies have yielded conflicting results concerning the association between BMI and risk of pancreatic cancer. Most prospective epidemiological studies^6- 15 have found that a high BMI and a lack of physical activity are associated with an increased risk of pancreatic cancer incidence or mortality, independent of a history of diabetes mellitus. However, several prospective studies have not confirmed a significant role of BMI in pancreatic cancer^16- 23 or have found that effect of BMI varied according to smoking status^24- 25 or sex.^26- 28

The objective of the present study was to examine the association between BMI, other anthropometric factors, and pancreatic cancer risk by pooling data from nested case-control studies included in the National Cancer Institute (NCI) Pancreatic Cancer Cohort Consortium (PanScan). With 2170 cases, this is one of the largest analyses to date of BMI and pancreatic cancer.

PanScan is an initiative that was funded jointly by the NCI's Division of Cancer Control and Population Sciences and the Division of Cancer Epidemiology and Genetics in 2006. PanScan includes investigators from 12 prospective epidemiological cohorts and 1 case-control study and was created to identify genetic markers of susceptibility through a genome-wide association scan and to investigate environmental, lifestyle, and genetic causes of pancreatic cancer.

Studies in the pooled analysis included the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study (ATBC)²⁹; CLUE II³⁰; Cancer Prevention Study II (CPS II)³¹; European Prospective Investigation Into Cancer and Nutrition (EPIC)³²; the Health Professionals Follow-up Study³³; the Mayo Clinic study³⁴; the New York University Women's Health Study,³⁵ the Nurses' Health Study³⁶; the Physicians' Health Study³⁷; the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial³⁸; Shanghai Men's and Women's Health Studies (SMWHS)^39- 40; the Women's Health Initiative ⁴¹; and the Women's Health Study.⁴² A total of 2170 cases and 2209 controls were eligible for the present study (Table 1).

Cases included all incident primary pancreatic adenocarcinoma (International Classification of Diseases for Oncology, Third Edition, codes C25.0-C25.3 and C25.7-C25.9). Endocrine pancreatic tumors (International Classification of Diseases for Oncology, Third Edition, code C25.4, histologic types 8150, 8151, 8153, 8155, 8240, and 8246) were excluded because the etiology of these cancers is thought to be different from that of exocrine tumors, which account for most pancreatic tumors. Case ascertainment varied among studies but included linking participants to cancer registries and national death indices and self-report and next-of-kin reports. Most cases were histologically confirmed (ATBC, CLUE II, EPIC, New York University Women's Health Study, SMWHS, and the Women's Health Initiative) or were confirmed through cancer registries (ATBC, CPS II, EPIC, and SMWHS), death certificates (CPS II and EPIC), or review of medical records by medical personnel (ATBC; CPS II; EPIC; Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial; and SMWHS).

Controls were incidence density sampled with a 1:1 control to case ratio and were alive and cancer free on the date of diagnosis of the matched case. At a minimum, controls were matched to cases on calendar year of birth (±5 years), sex, race, and ethnicity. Some cohorts used more stringent matching on age and, in addition, on other relevant factors (for comparisons of blood levels of analytes of interest), such as age at baseline or age at blood collection (±5 years), date and time of day of blood collection, fasting blood collection, and length of follow-up (Table 1).

Data on anthropometric measures, demographics, and possible confounders were collected through self-administered written questionnaires or in-person interviews. Detailed descriptions of data collection methods have been published previously by the individual studies.^{29- 30,32- 33,35- 44} From each study, baseline information on BMI, weight, height, waist circumference, waist to hip ratio (WHR), history of cigarette smoking, sex, age, race, family history of pancreatic cancer, alcohol consumption, pancreatitis, and history of diabetes mellitus was requested. Individual data sets were checked for consistency with previously published results. Forty cases and 46 controls had missing data on BMI, resulting in 2130 cases and 2163 controls available for the main analyses.

The Special Studies Institutional Review Board of the NCI approved the pooled PanScan. Each study also was approved by its local institutional review board.

Odds ratios (ORs) and 95% confidence intervals (95% CIs) for pancreatic cancer risk were calculated using unconditional logistic regression, adjusting for cohort, age (categorical), sex, BMI source (self-reported vs measured), and smoking (never, former, or current) (model 1). Several multivariate models were assessed to control the effects of potential confounders. Model 2 was additionally adjusted for diabetes mellitus history (yes or no). In model 3, cases diagnosed in the first 2 years of follow-up were excluded to address the possibility of an effect of early undiagnosed disease. In model 4, current smokers (at baseline) were excluded, and in model 5, cases diagnosed in the first 2 years of follow-up and current smokers were excluded. Furthermore, models including waist circumference and WHR were additionally adjusted for height to remove extraneous variation due to body size. No adjustment was made for family history of pancreatic cancer because few cohorts had this information. Trend tests were conducted using cohort-specific quartiles of BMI, weight, height, waist circumference, and WHR as well as descriptive BMI categories (underweight, <18.5; normal weight, 18.5-24.9; overweight, 25.0-29.9; obese, 30.0-34.9; and severely obese, ≥35.0). To test for heterogeneity, BMI quartile categories were modeled as a continuous variable, and the risk estimates and standard errors from the cohort-specific models were used to generate the Q statistic.

The association between BMI and time of onset for pancreatic cancer was also examined using logistic regression modeling. Differences in time of onset were examined for normal vs overweight vs obese categories of BMI and in a combined category of overweight and obese. The analyses were conducted using a software program (SAS version 9.1.3; SAS Institute Inc, Cary, North Carolina).

The study included 2170 pancreatic cancer cases and 2209 controls aged 37 to 94 years (Table 1). Of the 2170 pancreatic cancer cases, 1059 were men and 1111 were women. Cases and controls were similar for age and racial distribution (Table 2). Most participants were white, and 86% of the study population was 60 years or older. Compared with controls, cases had a higher prevalence of current smoking (18% and 25%), diabetes mellitus (7% and 14%), history of pancreatitis (0.4% and 11%), and family history of pancreatic cancer (4% and 6%) based on data from cohorts with available information. The average age at pancreatic cancer onset in cases was 68.3 years, and the average lag time between cohort enrollment and diagnosis of pancreatic cancer in cases was 6.3 years.

Table 3 describes baseline anthropometric characteristics of cases and controls. Weight, height, and corresponding BMI were self-reported in approximately 50% of participants, measured in 29%, and measured and subsequently adjusted for difference in clothing in approximately 20%. Thirty-six percent of cases and 39% of controls had BMI in the normal range, 41% of cases and 39% of controls were overweight, and 21% of cases and 19% of controls were obese (Table 3). Cases had slightly higher mean weight compared with controls (76.8 and 75.5 kg) and a larger mean waist circumference (86.9 and 85.7 cm). Mean WHR and height were similar.

Table 4 displays ORs and 95% CIs for pancreatic cancer according to baseline anthropometric factors for all individuals in the study. In all of the participants, a positive association between increasing BMI and risk of pancreatic cancer was observed (adjusted OR for the highest vs lowest BMI quartile, 1.33; 95% CI, 1.12-1.58; P_trend < .001 in model 1). Statistically significant trends of increasing risk of pancreatic cancer with increasing BMI (both quartiles and clinical categories) were observed in all 5 models analyzed.

The figures demonstrate the individual study results (model 1) and pooled risk estimates for overweight (Figure 1), obese (Figure 2), and severely obese individuals (Figure 3). Further adjustment for diabetes mellitus history (model 2) resulted in attenuation of risk estimates compared with model 1, but P values for trend were statistically significant for BMI quartiles and categories (Table 4). In addition, waist circumference and WHR were positively associated with risk of pancreatic cancer in all participants with top vs bottom quartile ORs of 1.23 (95% CI, 0.94-1.62) and 1.71 (95% CI, 1.27-2.30), respectively (Table 4). Stratification by BMI source (self-reported vs measured) resulted in similar risk estimates: ORs (95% CIs) for obese vs normal-weight BMI were 1.24 (0.92-1.68) for measured BMI and 1.21 (0.95-1.53) for self-reported BMI. The OR per BMI increase of 5 was 1.13 (95% CI, 1.11-1.14).

The risk estimates did not change significantly in the sensitivity analysis excluding the Mayo Clinic case-control study (data not shown); therefore, we decided to include the Mayo participants in the final analyses. There was no evidence of significant heterogeneity between different cohorts for the BMI–pancreatic cancer results (P_{heterogeneity} = .36).

Table 5 and Table 6 provide ORs and 95% CIs for pancreatic cancer in men and women, respectively. In men, the adjusted risk estimate (model 1) for the top vs bottom quartile of BMI was 1.33 (95% CI, 1.04-1.69). Higher risk estimates were observed after the exclusion of current smokers (model 4). In men who never smoked, there was a significant trend of increasing risk with increasing BMI (P_trend = .007) with the top vs bottom quartile (OR, 1.51; 95% CI, 1.13-2.03). Height, waist circumference, and WHR ratio were not significantly associated with pancreatic cancer in men (Table 5).

In women, statistically significant trends of increasing risk of pancreatic cancer with increasing BMI were observed overall (model 1) and after the exclusion of cases diagnosed in the first 2 years of follow-up (model 3) or current and former smokers (model 4) (Table 6). Compared with normal-weight BMI (model 1), the ORs for pancreatic cancer were 1.31 (95% CI = 1.07-1.60) for overweight women and 1.61 (95% CI, 1.12-2.33; P_trend = .003) for severely obese women. Increasing waist circumference and WHR were significantly associated with pancreatic cancer risk in women. Compared with the reference group, women in the highest quartile of WHR had an OR of 1.87 (95% CI, 1.31-2.69) after adjustment for cohort, age, BMI source, and smoking status. Inclusion of BMI (categorical) and WHR (quartiles) in the same model suggested that the effect of increasing WHR is stronger (P = .006) compared with that of BMI categories (P = .44) after adjustment for cohort, age, sex, BMI source, smoking, and diabetes mellitus history.

We did not observe clinically meaningful differences in time of onset for pancreatic cancer between normal-weight and overweight/obese individuals. Overweight and obese individuals together were diagnosed approximately 4 months earlier than were normal-weight individuals (data not shown). When comparing obese individuals only with normal-weight individuals, obese participants were diagnosed on average approximately 1 year earlier than were normal-weight individuals, and the difference was significant (P = .03).

The results of this large pooled set of studies support the hypothesis that obesity is associated with an increased risk of pancreatic cancer. The present findings are consistent with most previous epidemiological studies that found a positive association between BMI and pancreatic cancer risk⁴⁵ and support the conclusion from a recent review panel from the World Cancer Research Fund⁴⁵ that the strength of the evidence supporting an association between obesity and pancreatic cancer is convincing.

Previous studies that did not observe a positive association between BMI and pancreatic cancer were often limited by the use of proxy respondents^46- 49 or by inadequate statistical power to examine associations at BMI levels that correspond with obesity (<10 cases with BMI >30.0).^7,10,17,50 Controversy regarding the role of smoking in the BMI and pancreatic cancer relationship still remains. Many previous studies^7,47- 48,50 that did not observe an association with obesity did not properly control for smoking history. It is possible that residual confounding due to improper adjustment for smoking history may have biased the association between BMI and pancreatic cancer toward the null. When stratifying on smoking status, some previous studies found that the relationship between BMI and pancreatic cancer was strongest in never smokers.^24- 25 The present findings are consistent with previous reports that the association between BMI and pancreatic cancer is stronger in nonsmokers (adjusted OR for BMI ≥30, 1.37; 95% CI, 1.06-1.78) than in smokers (adjusted OR for BMI ≥30, 1.14; 95% CI, 0.91-1.78).

Unlike a recent study⁵¹ in which the authors reported that individuals who were overweight or obese from age 20 to 49 years had earlier onset of pancreatic cancer compared with those with normal body weight, we did not find a substantial difference in age at diagnosis between normal-weight and the combined group of overweight and obese individuals.

In this study, BMI was assessed between ages 37 and 94 years, and overweight or obese individuals were diagnosed a mean of 4 months earlier than were normal-weight individuals. The study by Li et al,⁵¹ using a hospital-based case-control study design, found that overweight or obese patients aged 20 to 49 years had a median age at pancreatic cancer onset 2 to 6 years earlier than that of normal-weight patients. However, these differences were based on BMI as recalled from early adulthood and may have been subject to recall bias. Nevertheless, as suggested by Li et al,⁵¹ obesity at younger ages might have a more profound effect on risk and age at onset of pancreatic cancer than would obesity at older ages.

There are established biologic pathways to support a relationship between excess body weight and the development of pancreatic cancer. Body fatness has a direct linear relationship with insulin production and is related to the development of insulin resistance.⁵² Furthermore, insulin resistance and abnormal glucose metabolism, even in the absence of diabetes mellitus, is associated with pancreatic cancer risk.^8,53- 54 In vitro studies have also shown that insulin has growth-promoting effects in the pancreas.⁵⁵A hyperinsulinemic state allows increased insulin to pass through the pancreas and trigger mitotic activity.^8,53,56 In addition, excess insulin can also downregulate insulinlike growth factor I–binding proteins, which would result in more bioavailable insulinlike growth factor I, which has been associated with cell proliferation and pancreatic cancer risk.^54,57

The results of this study also support a specific role of central adiposity in pancreatic cancer risk, especially in women. In addition to general body fatness, there is a direct linear relationship between intra-abdominal fat deposits, insulin production, and the development of insulin resistance.⁵²

The present study has several strengths, including the very large sample size, the wide range of BMI, and the ability to control for most known or suspected pancreatic cancer risk factors. In addition, the study population was largely a nested sample from various prospective cohort studies so that BMI was measured before pancreatic cancer diagnosis, thus reducing differential reporting of past exposure information. Limitations include the use of some self-reported exposure information; however, adjusting for source of exposure information (self-reported or measured) did not alter the association. Another potential limitation is the wide range of lag periods between BMI measurement (collected at baseline for each cohort) and the date of diagnosis; however, sensitivity analyses examining this lag time by excluding the first 2 years of follow-up did not change the point estimates appreciably, thus arguing against an effect of prediagnostic disease-related changes in anthropometric measures (reverse causation). Participating cohorts had different coding systems for physical activity that were not readily comparable; therefore, we could not assess whether the association between BMI and pancreatic cancer varies by level of physical activity. To address potential residual confounding by smoking, we performed the analyses in never smokers and found a slightly stronger association between BMI and pancreatic risk. Last, only a few cohorts had data available on waist and hip circumference so that there was limited statistical power to examine these relationships.

In summary, the results of this study provide additional evidence that obesity is associated with increased risk of pancreatic cancer. In addition, the association between waist circumference and pancreatic cancer risk, especially in women, suggests a possible association with the distribution of body fat. These findings, along with those from previous studies, strongly support the role of obesity in pancreatic cancer development.

Correspondence: Alan A. Arslan, MD, Department of Obstetrics and Gynecology, New York University School of Medicine, 550 First Ave, TH-528, New York, NY 10016 (alan.arslan@nyumc.org).

Accepted for Publication: October 31, 2009.

Author Contributions: Dr Arslan had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analyses. Study concept and design: Kooperberg, Shu, Bueno-de-Mesquita, Fuchs, Stolzenberg-Solomon, Zheng, Amundadottir, Barricarte, Bingham, Chanock, Hartge, Hoover, K. B. Jacobs, Manson, McTiernan, Mendelsohn, Palli, Slimani, Tjønneland, and Tobias. Acquisition of data: Helzlsouer, Kooperberg, Shu, Bueno-de-Mesquita, Fuchs, E. J. Jacobs, LaCroix, Petersen, Stolzenberg-Solomon, Albanes, Bamlet, Barricarte, Boeing, Buring, Clipp, Gaziano, Giovannucci, Hankinson, Hartge, Hoover, Hunter, Hutchinson, K. B. Jacobs, Lynch, Manson, McWilliams, Michaud, Palli, Thomas, Tjønneland, Trichopoulos, Virtamo, Wolpin, and Zeleniuch-Jacquotte. Analysis and interpretation of data: Arslan, Helzlsouer, Kooperberg, Shu, Steplowski, Fuchs, Gross, E. J. Jacobs, Boutron-Ruault, Giovannucci, Hankinson, Hartge, K. B. Jacobs, Kraft, Manjer, Rohan, Trichopoulos, Yu, and Patel. Drafting of the manuscript: Arslan, Helzlsouer, Kooperberg, and Patel. Critical revision of the manuscript for important intellectual content: Arslan, Helzlsouer, Kooperberg, Shu, Steplowski, Bueno-de-Mesquita, Fuchs, Gross, E. J. Jacobs, LaCroix, Petersen, Stolzenberg-Solomon, Zheng, Albanes, Amundadottir, Bamlet, Barricarte, Bingham, Boeing, Boutron-Ruault, Buring, Chanock, Clipp, Gaziano, Giovannucci, Hankinson, Hartge, Hunter, Hutchinson, K. B. Jacobs, Kraft, Lynch, Manson, McTiernan, McWilliams, Mendelsohn, Michaud, Palli, Rohan, Slimani, Thomas, Tjønneland, Tobias, Trichopoulos, Virtamo, Wolpin, Yu, and Zeleniuch-Jacquotte. Statistical analysis: Kooperberg, Shu, E. J. Jacobs, Bamlet, Hunter, K. B. Jacobs, Manjer, and Yu. Obtained funding: Shu, Bueno-de-Mesquita, Fuchs, Petersen, Stolzenberg-Solomon, Albanes, Boeing, Gaziano, Giovannucci, Hoover, Kraft, Lynch, Palli, Tjønneland, Trichopoulos, and Zeleniuch-Jacquotte. Administrative, technical, and material support: Arslan, Kooperberg, Steplowski, Fuchs, Gross, Petersen, Zheng, Albanes, Amundadottir, Bingham, Boeing, Chanock, Clipp, Gaziano, Hartge, Hunter, K. B. Jacobs, Lynch, Manjer, Manson, Mendelsohn, Slimani, Thomas, Tobias, Trichopoulos, and Virtamo. Study supervision: Arslan, Helzlsouer, Kooperberg, Bueno-de-Mesquita, Hankinson, Hartge, and Hoover.

Funding/Support: This project was funded in whole or in part by contract HHSN261200800001E from the NCI, National Institutes of Health. The ATBC was supported by contracts N01-CN-45165, N01-RC-45035, and N01-RC-37004 from the Intramural Research Program of the NCI and the US Public Health Service. CLUE II was supported by grant 5U01AG018033 from the National Institute on Aging and by grants CA105069 and CA73790 from the NCI. The EPIC was supported by the European Commission: Public Health and Consumer Protection Directorate 1993-2004; Research Directorate-General 2005; Ligue contre le Cancer; Societé 3M; Mutuelle Générale de l’Education Nationale; Institut National de la Santé et de la Recherche Médicale (France); German Cancer Aid, German Cancer Research Center, Federal Ministry of Education and Research (Germany); Danish Cancer Society (Denmark); Health Research Fund of the Spanish Ministry of Health, participating regional governments and institutions (Spain); Cancer Research UK, Medical Research Council, Stroke Association, British Heart Foundation, Department of Health, Food Standards Agency, the Wellcome Trust (United Kingdom); Greek Ministry of Health and Social Solidarity, Hellenic Health Foundation and Stavros Niarchos Foundation (Greece); Italian Association for Research on Cancer (Italy); Dutch Ministry of Public Health, Welfare and Sports, Dutch Prevention Funds, LK Research Funds, Dutch ZON (Zorg Onderzoek Nederland) (the Netherlands); Swedish Cancer Society, Swedish Scientific Council, Regional Government of Skane and Västerbotten (Sweden); and World Cancer Research Fund. The New York University Women's Health Study is supported by research grants R01CA034588, R01CA098661, and P30CA016087 from the NCI and by grant ES000260 from the National Institute of Environmental Health Sciences. The Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial was supported by contracts from the NCI (University of Colorado: NO1-CN-25514; Georgetown University: NO1-CN-25522; Pacific Health Research Institute: NO1-CN-25515; Henry Ford Health System: NO1-CN-25512; University of Minnesota: NO1-CN-25513; Washington University: NO1-CN-25516; University of Pittsburgh: NO1-CN-25511; University of Utah: NO1-CN-25524; Marshfield Clinic Research Foundation: NO1-CN-25518; University of Alabama at Birmingham: NO1-CN-75022; Westat Inc: NO1-CN-25476; and University of California, Los Angeles: NO1-CN-25404). The SMWHS were supported by extramural research grants R01 CA82729 and R01 CA70867 from the NCI and by the Intramural Research Program of the NCI (Division of Cancer Epidemiology and Genetics). The Women's Health Initiative is funded by contracts N01WH22110, 24152, 32100-2, 32105-6, 32108-9, 32111-13, 32115, 32118-32119, 32122, 42107-26, 42129-32, and 44221 from the National Heart, Lung, and Blood Institute.

Role of the Sponsors: The funding agencies had no role in the conduct of the study, the interpretation of the data, or the decision to submit the manuscript for publication.

PanScan Writing Committee Members: Alan A. Arslan, MD; Kathy J. Helzlsouer, MD, MHS; Charles Kooperberg, PhD; Xiao-Ou Shu, MD, PhD; Emily Steplowski, BS; and Alpa V. Patel, PhD.

Steering Committee Members: Alan A. Arslan, MD; Kathy J. Helzlsouer, MD, MHS; H. Bas Bueno-de-Mesquita, MD, PhD, MPH; Charles S. Fuchs, MD, MPH; Myron D. Gross, PhD; Eric J. Jacobs, PhD; Andrea Z. LaCroix, PhD; Gloria M. Petersen, PhD; Rachael Z. Stolzenberg-Solomon, PhD, MPH; and Wei Zheng, MD, PhD.

Disclaimer: The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, and neither does the mention of trade names, commercial products, or organizations imply endorsement by the US government.

Additional Contributions: We thank the investigators from the PanScan centers and the study participants, without whom this study would not have been possible.