Body mass index and body size in early adulthood and risk of pancreatic cancer in a central European multicenter case–control study



The relationship between two measures of excess body weight, body mass index (BMI) and body size score, and risk of pancreatic cancer was examined among 574 pancreatic cancer cases and 596 frequency-matched controls from the Czech Republic and Slovakia enrolled between 2004 and 2009. Analyses using multivariable logistic regression showed an increased risk of pancreatic cancer associated with elevated quartiles of BMI at ages 20 [fourth quartile: odds ratio (OR) = 1.79, 95% confidence interval (CI): 1.23, 2.61] and 40 (fourth quartile: OR = 1.57, 95% CI: 1.09, 2.27) compared to the lowest quartile. Consistent results were observed for body size score at ages 20 (high versus low: OR = 1.66, 95% CI: 1.08, 2.57) and 40 (medium versus low: OR = 1.36, 95% CI: 1.00, 1.86), but no association was found for BMI and body size score at 2 years before the interview. Stronger risk estimates for BMI were observed in males than females, particularly at age 20, but the analysis of body size yielded similar estimates by sex. When considering excess body weight at both ages 20 and 40 jointly, the highest risk estimates were observed among subjects with elevated levels at both time periods in the analysis of BMI (OR = 1.86, 95% CI: 1.32, 2.62) and body size (OR = 1.53, 95% CI: 1.09, 2.13). These findings, based on two different measures, provide strong support for an increased risk of pancreatic cancer associated with excess body weight, possibly strongest during early adulthood.

More than 250,000 pancreatic cancer cases are registered worldwide each year with some of the highest rates observed in areas of central Europe including Poland, Czech Republic and Slovakia.1, 2 Disease survival is among the poorest of all neoplasms exhibiting a median of less than 6 months and case fatality reaching close to 90% within 12 months of diagnosis.3, 4 Relatively little is known about the causes of pancreatic cancer with the exception of advanced age, smoking, family history of pancreatic cancer and history of diabetes mellitus and chronic pancreatitis.5

Over the past decade, evidence has accumulated supporting a role for excess body weight in the etiology of pancreatic cancer. The biological plausibility of this association stems from the strong relationships observed between obesity and abnormal glucose metabolism and hyperinsulinemia,6, 7 which are known to have the capacity to increase local blood flow and promote cell proliferation within the pancreas.8, 9 Epidemiologic studies examining the health effects of excess body weight have largely focused on the use of the body mass index (BMI). A recent meta-analysis conducted among 21 prospective cohort studies showed a statistically significant elevated risk of pancreatic cancer associated with each 5 kg/m2 increase in BMI in both males and females.10 Although these findings offer substantial support for a causal relationship, previous studies have been largely inconsistent regarding differential effects by sex and other major risk factors,11–14 and more evidence is beginning to emerge suggesting that onset of overweight or obesity earlier in life may have a larger impact on pancreatic cancer risk.15, 16

Moreover, further clarity and confirmation may be achieved through the evaluation of alternative measures of excess body weight. As a measure based only on height and weight, BMI makes simplistic assumptions about the distribution of muscle and bone mass; thus, it may not reflect an accurate measure of body fat for certain individuals, namely, athletes and the elderly.17, 18 Consistent results among multiple indicators of excess body weight in pancreatic cancer studies may strengthen the evidence in support of this association.

In this current analysis conducted among a large central European sample of pancreatic cancer cases and controls, we examined the association between excess body weight and pancreatic cancer risk using two different measures, BMI and pictogram body size score, obtained for multiple time periods.

Material and Methods

Study population

The study had a case–control design and was conducted in five areas of Czech Republic (Prague, Olomouc, Ostrava, Ceske Budejovice and Brno) and five areas of Slovakia (Banska Bystrica, Bratislava, Martin, Zilina and Trencin). In each area, a series of newly diagnosed cases of primary pancreatic cancer was recruited between October 2004 and June 2009 in the Czech Republic and between October 2007 and July 2009 in Slovakia, together with a set of controls that were frequency matched to cases on age and sex. Cases were recruited in hospitals that treated essentially all pancreatic cancers in each area. The patient was asked to participate if he/she presented at the hospital with clinical or imaging suspicion of pancreatic cancer and had lived in the study area for at least 1 year. This ultra rapid case ascertainment method included contact with the patients before confirmation of the diagnosis. Identification of cases occurred through an active search of clinical and pathological departments. In total, 750 patients were approached for inclusion in the study, of whom 652 were diagnosed with pancreatic cancer (excludes known pancreatic endocrine tumors, i.e., islets cell carcinomas); six (0.9%) cases died before the interview (next-of-kin interviews were not used), 53 (8.1%) refused to participate and 11 (1.7%) were too ill to be interviewed. Another five cases were excluded from the study because of missing information for height or weight, and body size score at ages 20 and 40, two cases who were underweight (BMI < 18.5) at both ages 20 and 40 years and one case because of missing date of birth, resulting in 574 cases included in the analyses. Diagnostic methods were not available for cases from Ceske Budejovice (n = 83). In other centers (n = 491), histological or cytological confirmation of the pancreatic cancer (based on pancreatic tissue or on lymph nodes or distant metastases) was available for 68% of the cases; other cases were recruited based on surgical exploration (5%), imaging techniques (11%) or clinical suspicion (16%).

We recruited controls among local residents consulting general practitioners (GPs) that served the same areas for which cases arose. Ten GPs in the Czech Republic and 17 in Slovakia participated in the central recruitment. They were provided with a sampling scheme comprising age and sex, which was used among patients visiting the GP for a minor health condition, routine health examination or preventive purpose. Individuals who agreed to participate were asked to visit a clinic center for the interview, with the exception of controls recruited in Prague and a subset from Slovakia centers in which the interview was conducted at the GP office at the time of the originally scheduled patient's visit. Out of 284 controls approached in Prague, 271 (95%) agreed to participate. In other centers, participation rates were more difficult to assess; estimates from GPs ranged from 60% in Brno to greater than 90% in Olomouc, Ostrava and Ceske Budejovice. At the time of the interview in Slovakia, three (3.5%) individuals refused to participate. A total of 619 interviews were conducted; exclusions included two (0.3%) subjects with missing date of birth or date of interview, 16 (2.6%) subjects with missing information for height or weight and body size score at ages 20 and 40 and five (0.8%) subjects who were underweight (BMI < 18.5) at both ages 20 and 40 years. This resulted in a total of 596 controls included in the analysis.

The study protocol was approved by the institutional review boards of the International Agency for Research on Cancer and all collaborating centers/institutions, and written informed consent was obtained for all participating subjects.

Data collection

Interviewers were trained in each center to perform face-to-face interviews with cases (at the hospital, next-of-kin respondents were not pursued) and controls (at the clinic or GP office) using standard questionnaires that covered tobacco use, alcohol consumption, diet, physical activity, medical history, reproductive history and anthropometric measures over time. Specifically, participants were asked about height and weight at ages 20, 40 and 2 years before the interview. Height and weight were used to calculate a BMI as (weight in kilograms)/(height in meters).2 Using a pictogram, subjects were also asked to choose among nine illustrations of body size ranging from very lean to obese, which they felt represented their body size most accurately at each of the three time points (Fig. 1). We used pictograms resembling those published by Must et al. in 1993.19 Similar pictograms were later validated and used in various epidemiological studies.20–24 Using our data, we estimated the correlation between BMI and body size score using the Spearman rank correlation coefficient in males and females separately. For ages 20, 40 and 2 years before the interview, coefficients were 0.54, 0.63 and 0.74 in males, and 0.62, 0.68 and 0.77 in females, respectively. Correlation coefficients were similar between cases and controls.

Figure 1.

Body image pictogram in males and females given a score of 1–9 ranging from very lean to severely overweight.19

Statistical analysis

To evaluate the risk of pancreatic cancer associated with the two measures of excess body weight, BMI and body size score, odds ratios (ORs) and 95% confidence intervals (CIs) were calculated using logistic regression adjusting for study center (indicator variables for the five Czech Republic centers and the Slovakia centers combined), age at interview (continuous) and sex. Based on prior knowledge of the importance of certain exposures and the results of the univariate analyses, the confounding potential of several other covariates was evaluated using the criteria of whether there was a change in OR of greater than 10% in the variable of interest after the addition of the potential confounder to the model. Potential confounders included education, diabetes mellitus (diagnosed 2 years before the interview), chronic pancreatitis (diagnosed 2 years before the interview), smoking (past versus never and current versus never), alcohol consumption (past versus never and current versus never) and physical activity. The risk associated with BMI for three time points before the interview was evaluated per 5 kg/m2 increase and as categorical variables with cutpoints for each time period based on quartiles of the distribution among controls irrespective of sex. Similarly for body size score, analyses were conducted for each unit increase and as a three-level categorical variable defined as low (score 1–3), medium (score 4) and high (score 5–9). The effect of BMI and body size score was examined stratified by other risk factors for pancreatic cancer including sex, history of diabetes mellitus, smoking status and alcohol consumption. In addition, excess body weight at ages 20 and 40 years was evaluated jointly using dichotomous classifications of BMI (≥24 kg/m2) and body size (≥score 4). Subjects were categorized into one of four possible groups based on their BMI and body size classification at ages 20 and 40 years. All analyses were conducted using SAS version 9 software.25


The characteristics of the 574 cases of pancreatic cancer and 596 controls frequency matched on sex, age and center are presented in Table 1. A large majority of subjects (88% of cases and controls) were recruited from centers located in the Czech Republic. About 56% of cases were males and the age at interview ranged from 20 to 94 years showing a mean of about 64 years. The age distribution of cases by sex showed that, on an average, males tended to be diagnosed at slightly earlier ages than females.

Table 1. Select characteristics of cases and controls, 2004–2009
inline image

The evaluation of BMI and body size at three time points and risk of pancreatic cancer is shown in Table 2. Compared to the lowest quartile, an increase risk of pancreatic cancer was observed for the third (OR = 1.81, 95% CI: 1.24, 2.63) and fourth BMI quartiles (OR = 1.79, 95% CI: 1.23, 2.61) at age 20 years, and a slightly weaker association for the third (OR = 1.40, 95% CI: 0.97, 2.03) and fourth BMI quartiles (OR = 1.57, 95% CI: 1.09, 2.27) at age 40 years. Similarly for body size score, an increased risk of pancreatic cancer was associated with medium (OR = 1.38, 95% CI: 0.99, 1.91) and high (OR = 1.66, 95% CI: 1.08, 2.57) values of body size compared to low at age 20 years, but appeared attenuated at age 40 years (high versus low: OR = 1.23, 95% CI: 0.90, 1.69). No association was observed for BMI or body size assessed 2 years before the interview.

Table 2. Association between body mass index and body size score at ages 20 and 40 years and 2 years before the interview and risk of pancreatic cancer, 2004–2009
inline image

An evaluation of BMI and body size score by sex is presented in Table 3. In the analysis of BMI among males, strong evidence of an association was observed, particularly when comparing the highest quartile to the lowest quartile at ages 20 (OR = 3.01, 95% CI: 1.67, 5.41) and 40 (OR = 1.83, 95% CI: 1.05, 3.20). Although elevated, the effects appeared to be weaker among females, in which a statistical test for interaction showed some evidence of a differential effect by sex for BMI at age 20 (p-interaction = 0.038). In contrast, this difference by sex is not observed in the analysis of body size at both ages 20 and 40. A similarly increased risk per unit increase in body size score at age 20 was found for males (OR = 1.15, 95% CI: 1.00, 1.33) and females (OR = 1.14, 95% CI: 0.97, 1.35). In addition, the effect of BMI and body size score did not significantly differ by smoking and alcohol consumption status or by previous diagnosis of diabetes mellitus (data not shown).

Table 3. Body mass index and body size at ages 20 and 40 years and risk of pancreatic cancer stratified by sex, 2004–2009
inline image

Table 4 shows the results examining the role of excess body weight at ages 20 and 40 years jointly. Subjects were categorized into dichotomous groups based on whether they had a high BMI (≥24 kg/m2) or body size (≥score 4) at ages 20 and 40. Compared to subjects with a low BMI at both ages 20 and 40, having a high BMI at age 40 only was associated with an increased risk (OR = 1.46, 95% CI: 1.05, 2.04), whereas a higher risk was observed for subjects with a high BMI at both ages 20 and 40 (OR = 1.86, 95% CI: 1.32, 2.62). In the analysis of body size, the same pattern was observed with the strongest risk found when having a high body size at both ages 20 and 40 (OR = 1.53, 95% CI: 1.09, 2.13). When stratified by sex, the effect observed in the analysis of BMI appeared to be predominately in males, which was not the case in the analysis of body size. Because of small numbers of subjects experiencing a reduction in BMI and body size between ages 20 and 40, this association could not be evaluated reliably.

Table 4. Body mass index and body size score at ages 20 and 40 years jointly and risk of pancreatic cancer, 2004–2009
inline image


The findings from this large case–control study conducted in central Europe are consistent with previous associations observed for elevated BMI and additionally support the role for excess body weight through an analysis of pictogram body size scores. Recent evidence from a large U.S. case–control study for the first time provided strong support for a potentially stronger role of excess body weight during adolescence and early adulthood.15 Our collection of anthropometric measures at different time periods allowed us to confirm these results with a conclusion that early life excess body weight, based on two measures, is associated with an increased risk of pancreatic cancer. However, we were not able to perform a more detailed evaluation of excess body weight after age 40 years.

Two recent meta-analyses of prospective cohort studies10, 26 showed an overall consistency in the association between elevated BMI and an increased pancreatic cancer risk but differed slightly with respect to conclusions about sex-specific effects. Both meta-analyses reported estimates in females as being associated with an ∼ 10% increased risk for each 5 kg/m2 increase in BMI, while only one observed a statistically significant association in males.10 Among 13 of the more recent reports on this topic published subsequent to these meta-analyses,10, 26, 27 all observed an effect for elevated BMI on pancreatic cancer risk,13, 15, 28–36 with the exception of one study that assessed BMI at interview37 and another presenting only a sex-stratified analysis with relatively small numbers of cases.38 In this current analysis, strong evidence of increased risks was observed for elevated levels of BMI in males and females combined, which appeared to be driven by a stronger association in males, particularly for BMI at age 20 years. On average, males tended to have higher BMI than females at both ages 20 and 40 years. Other studies have observed similar distributions by sex and have also shown the effects on risk to be predominately in males.30, 31

However, a differential effect by sex was not supported by the results observed for body size score. To date, there is no overall consensus for a potentially differential role of excess body weight by sex on pancreatic cancer risk. As suggested by previous meta-analyses10, 27 and several of the more recent studies including our study, excess body weight does appear to contribute to risk in both males and females possibly differentially, but the associations may be complicated by sex-specific differences in dietary habits, physical activity and body fat distribution.7 One plausible explanation for a potentially larger effect in males may be related to different tendencies in fat deposition. Compared to women, there is a greater tendency in men for fat to accumulate in the abdominal area versus other areas of the body.7 Abdominal adiposity has been associated with metabolic abnormalities and hyperinsulinemia,6 conditions which are also thought be the biological link between excess body weight and pancreatic cancer.

Furthermore, we are not able to conclude whether the absence of an association at 2 years before the interview is due to it being etiologically unimportant or if the findings are influenced by the severe weight loss commonly experience by pancreatic cancer patients prediagnosis.39 Because of the tendency for pancreatic cancer to be diagnosed at advanced stages, it is conceivable that at 2 years before the interview (prediagnosis), the subjects may have already started experiencing disease-related symptoms (i.e., weight loss and lifestyle changes).

The evaluation of excess body weight on risk of pancreatic cancer has predominately been through an analysis of BMI with additional support provided by a few studies examining waist and hip circumference and waist-to-hip ratio.11, 14, 40 Although recognized to be highly correlated with body fat, the BMI measure has its limitations with potentially failing to discriminate between body fat percentage and lean mass.17, 41 Results from the current analysis of pictogram body size score, as a complementary measure of excess body weight, corroborate findings observed for BMI and pancreatic cancer risk. In studies requiring long-term recall by the subjects, it has been shown that the use of pictograms, rather than asking subjects to recall specific body weights, may aid in the effort to obtain valid measures of excess body weight.19–21 This measure has been effectively used in previous studies of other cancers,22, 24, 42 but to our knowledge this is the first report among pancreatic cancer studies.

The potentially adverse effects of excess body weight beginning in childhood and early adulthood are widely demonstrated in studies of cardiovascular diseases43, 44 and diabetes,45, 46 but little is known about the effects on cancer risk.47 In a recent large case–control study that examined BMI at 10-year intervals across a lifespan, the strongest risks associated with each 5-unit increase were observed for BMI at ages 14–19, 20–29 and 30–39 years.15 Risk estimates appeared to drop between ages 40 and 69 years, particularly among women. Their analysis examining age at onset of overweight also showed the strongest associations among subjects who became overweight at ages 14–19 and 20–29 years compared to never being overweight.15 Consistent with this previous report and others that have evaluated the early adulthood period,30, 31 our analysis showed the strongest risks among subjects having an elevated measure of excess body weight at both ages 20 and 40. Although the underlying mechanisms are not well understood, an association describing the importance of excess body weight in early adulthood in the risk of a disease typically diagnosed after age 50 years suggests that there may be a cumulative effect over an extended duration that is etiologically important.15, 47 Because of limitations in sample size, we were not able to adequately evaluate the potential for a reduction in risk of pancreatic cancer following a loss of excess body weight between ages 20 and 40 years.

A major strength of this study is the ultra rapid case ascertainment protocol implemented at each of the study areas, which allowed us to achieve a population-based ascertainment of incident pancreatic cancer cases with a low refusal rate and administer all nonproxy interviews. Because of the short survival period and high case-fatality rate of this disease, many previous studies have used pancreatic cancer mortality as the endpoint rather than incidence contributing to misclassification of the outcome and potentially limited generalizability of study results. Also, case–control studies often were left with using anthropometric information from a mixture of proxy and nonproxy respondents. Subgroup analyses from a previous meta-analysis showed no association with BMI among four studies that used proxy interview, but found a statistically significant increased risk among three studies that did not use proxies.27

Although our case definition excluded pancreatic endocrine cancer (i.e., islets cell carcinoma) based on topography, it is possible that a small number of these cases were included among those who were not histologically or cytologically confirmed. However, the proportion of pancreatic endocrine tumor cases is expected to be very low (∼ 1%), and analyses restricted to pathology confirmed pancreatic exocrine tumors (331 cases and 596 controls) resulted in the same overall conclusions (results not shown).

The degree of misclassification in past recall of height, weight and body size is difficult to quantify. However, previous studies have shown very high correlations between self-reported long-term recall and measured weight, even over several decades.48 Notably though, these studies showed a tendency for underreporting weights among overweight subjects which tend to be more pronounced in females.49 The misclassification nondifferentially between cases and controls would have the effect of biasing results toward one of no association. This may be one explanation contributing to the weaker association observed in females, which was evident only in the analysis of BMI, but not body size. In terms of differential misclassification, it is difficult to predict the direction of the bias. It can be envisioned that as many cases experience significant cancer-related weight loss, there may be a tendency for cases to underestimate weights for time periods earlier in life, producing risk estimates that would underestimate the true effect of excess body weight. One intended use of the body size pictograms in this study was to provide a second measure of excess body weight that may have recall advantages over specific measures of height and weight.

Importantly, the influence of uncontrolled or residual confounding on risk estimates is always a possibility. We presented results for two logistic regression models, one of which included known major risk factors for pancreatic cancer. In most analyses, additionally adjusting for these risk factors did not appear to change the risk estimates substantially suggesting that the confounding potential for these variables was likely small. Because of limitations in our data, we did not present results adjusted for other potential confounders including physical activity and dietary factors. However, among a subset of subjects for which data were available (missing for 49 subjects), adjusting for leisure physical activity did not remarkably change the risk estimates.

Overall, these results support the conclusion that excess body weight is a significant risk factor for pancreatic cancer. Through an analysis examining different points in time, we provide evidence that excess body weight beginning from early adulthood may have a larger impact on risk than later onset excess weight gain. More research is necessary to evaluate whether a reduction in risk can be achieved by a loss in excess body weight at various intervals within a lifespan. Furthermore, current evidence suggests a specific role for central adiposity in the pathogenesis of pancreatic cancer. The additional use of waist circumference and/or waist-hip ratio in future epidemiologic studies may help to clarify the role of body fat distribution in the risk of pancreatic cancer.17 The implications of these findings on public health are important, particularly in the western world where obesity rates have reached epidemic proportions even among children.50