Even though pancreatic cancer is the 5th-leading cause of cancer deaths in the Western world,1 little is known about the etiology of the disease. Lifestyle and other risk factors for pancreatic cancer have been examined in many epidemiologic studies, most of which were case-control studies. The conclusions of many of these are dubious because of small sample size, low participation and use of proxy respondents. Only cigarette smoking stands out as a likely causal agent in epidemiologic studies. However, cigarette smoking can explain only about 25% of the incidence.2 Studies have also suggested that coffee, ethanol, obesity, high energy intake and high consumption of fat, carbohydrates and animal protein may be risk factors.3–5 Some studies have shown decreased risks associated with frequent consumption of fruits and vegetables, but this has not been observed consistently.3 It is possible that factors other than those that have been investigated may affect the development of pancreatic cancer.
Our study examined the effects of diet, coffee, ethanol, tobacco use, BMI, physical activity and change in body weight on pancreatic cancer incidence in a cohort study of twins established in 1958 and followed by the Swedish Twin Registry. The cohort, which provided prospective information concerning exposure, included male and female same-sexed twin pairs who were born from 1886–1925 and were both living in Sweden in 1961. At the 1961 baseline, self-administered questionnaires regarding lifestyle factors were mailed to 25,778 registrees. However, 1,288 of these were found to have already died. Additional questionnaires were administered in 1963 and 1967. Our current study included 12,204 females and 9,680 males who responded to these questionnaires with information on at least some of the potential risk factors. The overall response rate to the questionnaires was 85%.
Subjects with pancreatic cancer were included in the analysis only if the diagnosis was made from 1969–97. Ninety percent of the tumors were histologically confirmed. The median age at enrollment in the cohort was 56 years. Subjects were followed from exposure assessment to 1 of 3 study endpoints: diagnosis of pancreatic cancer, death or the end of the study on December 31, 1997. Cancer incidence was ascertained by record linkage to the Swedish Cancer Registry (documented to be 98% complete).6 Death was ascertained by linkage to the Swedish Cause of Death Registry. During the follow-up period (median 16 years), there were 176 incident cases of pancreatic cancer diagnosed at a median age of 73 years. The total number of pancreatic cancer cases included in the risk analysis of each lifestyle factor is less than 176 incident cases because some subjects did not answer all the questions on the questionnaires. The internal dropout rate varied considerably between the different questions.
In the questionnaires, smoking status was assessed as nonsmoker, former smoker or current smoker. Current smokers reported the number of cigarettes smoked per day and were coded as light (1–10 cigarettes/day) and regular (11+/day) based on the combination of responses to the questionnaires in 1961 and 1967. Ethanol consumption was self-reported in 1967 as amount and frequency of beer, wine and spirits intake. Ethanol consumption was coded as grams pure ethanol per month (Table I). Coffee consumption was assessed as number of cups per day (Table I). The 1967 questionnaires also included questions about the frequency of consumption of 10 food groups: fruit/vegetables, eggs, pork, beef, sausages, fish, potatoes, flour/grain products, pastry and sweets. The response alternatives were no part, a small part, a moderate part and a large part of the diet (for pastry and sweets: less than daily, daily and several times a day). The alternatives no part and a small part were combined as low or no part for analysis. Physical activity during leisure hours was assessed as low (responses were hardly any physical exercise or light physical exercise, e.g., regular walks, light gardening) and high (responses were regular exercise or hard physical training). Physical activity at work was assessed as sedentary or physical (responses were physical or hard physical). Weight was self-reported in kilograms and height in centimeters. BMI (in kg/m2) was used as a measure of relative body weight. BMI scores were divided into 4 groups according to WHO criteria for thinness and overweight.7 BMI 18.5–24.99 was selected as the reference category. Weight was self-reported for the actual age when the subject answered the questionnaire and for ages 25 and 40. Adult weight gain was assessed by subtracting weight at age 25 years from weight at enrollment.
|Former cigarette smokers||15||0.75||0.42–1.43|
|Light (1–10 cigarettes/day)||48||1.37||0.94–2.00|
|Regular (11+ cigarettes/day)||21||1.25||0.75–2.08|
|Cigars or pipe||9||0.58||0.28–1.19|
The relative risk of pancreatic cancer was estimated through Cox proportional hazards modeling using the SAS program (SAS Institute, Cary, NC). Relative risk estimates are presented with 95% confidence intervals, adjusted for smoking (current, former and nonsmoker), age and sex. Age was used as a continuous variable. In addition, the estimates for food items were adjusted for BMI. We did not adjust for nonindependence of observations (the “twinness” of the sample). Doing so would have resulted in wider confidence intervals but would not change the point estimates or the conclusions of our study. Since genetic effects are of little or no importance for pancreatic cancer,8 a co-twin control design using cancer discordant pairs would provide little additional evidence. Because the sample size is relatively small, estimates from such a study would have little power.
Cigarette smoking was associated with a small increase in pancreatic cancer risk, with an age- and sex-adjusted relative risk of 1.4 (Table I). Ex-smokers were not at increased risk. There was no difference in risk between light and regular smokers. Smoking of cigars or pipes was not associated with an increased risk. If anything, the risk was reduced (relative risk 0.58).
We found no clear association between ethanol consumption and pancreatic cancer (Table I). Consumption of coffee was associated with a reduced risk of pancreatic cancer. Among those who consumed 7 or more cups per day, the relative risk of pancreatic cancer was 0.4 compared to those who reported a consumption of 2 cups or less (Table I).
There were no clear associations for the food items, with the exceptions of pork and sausages, which were associated with reduced risks. The relative risks for high or moderate pork consumption were 0.25 (95% CI 0.08–0.81) and 0.66 (95% CI 0.46–0.93), respectively, compared to low or no consumption. High consumption of sausages was also associated with a reduced risk of pancreatic cancer (relative risk 0.37; 95% CI 0.13–1.03).
Table II shows associations of BMI, physical activity and pancreatic cancer. Mild obesity (BMI 25–30) at registration was associated with an increased risk of pancreatic cancer (relative risk 1.36) compared to subjects with a BMI of 18.5–25. Subjects with a BMI below 18.5 had an even greater increase in risk for pancreatic cancer with a relative risk of 2. Adult weight gain was associated with a trend toward increased pancreatic cancer risk. Subjects with a weight gain of 12 kg or more had a relative risk of 1.46. Sedentary work was associated with an increased risk of pancreatic cancer compared to physically active work (relative risk 1.58). Low physical activity outside work was also associated with an increased risk of pancreatic cancer (relative risk 1.53).
|BMI at enrollment|
|Adult weight gain|
|First quartile (≤1 kg)||25||1.02||0.60–1.73|
|Second quartile (2–5 kg)||31||1.0||Referent|
|Third quartile (6–11 kg)||35||1.28||0.77–2.14|
|Fourth quartile (≥12 kg)||36||1.46||0.87–2.45|
|Physical activity at work|
|Physical activity during leisure time|
Our present follow-up study confirmed previous findings that smoking of cigarettes is associated with increased risk of pancreatic cancer.9 Previous cohort and case-control studies have usually found a relative risk of 2–4 for moderate or heavy smokers, in contrast to the modest 1.4-fold increased risk for current smokers in our study. This may be due to the lack of repeated measures of exposure in our study, since some subjects may have stopped or started smoking during the study period. However, the responses to the 3 different questionnaires (1961–67) showed that only 2% of the subjects changed their smoking habits during that interval. In agreement with some previous studies,10–12 we found no change in risk across categories of current daily cigarette dose and did not see an increased risk for former cigarette smokers or for individuals smoking pipes or cigars.
In our study, coffee consumption was associated with a decreased risk of pancreatic cancer. A protective effect of coffee consumption has been reported previously,13 although most studies have either shown no association9–14 or an increased risk.15, 16 The lack of association of ethanol consumption and increased pancreatic cancer risks is consistent with most studies,9–14 although some have shown increased risks.12–17
The protective effect of fruit and vegetables that has been seen in a number of previous studies3–9 was not confirmed by our data. However, the dietary instrument used in our study only looked at the relative contribution of the fruits/vegetables group to the total diet and may be too crude an instrument to detect a small effect. In addition, changes in dietary habits could have occurred during the follow-up period, although we know of no data suggesting that Swedish dietary habits changed radically during that time. We found that high and moderate consumption of pork and high consumption of sausages were protective against pancreatic cancer. This inverse association has been reported in a previous study,18 but the mechanisms that might be involved are unclear.
In our study, mild obesity was associated with a modest increased risk. Adult weight gain was also associated with an increased risk of pancreatic cancer, suggesting that excess of energy intake over expenditure could be a risk factor. In some previous studies, BMI was associated with modest increased risks, with ORs ranging from 1.2–1.7,4, 19 but others have not found an association.20 Energy restriction in rats has been found to inhibit pancreatic carcinogenesis,21 and obesity is associated with increased circulating levels of insulin, which has been suggested to act as a promoter of pancreatic tumor growth.22 The finding of an increased risk for subjects with a BMI less than 18.5 should be viewed with caution, since it is based on only 5 cases. These 5 subjects all had a weight gain of less than 2 kg between age 25 and enrollment, so the increased risk was not associated with substantial adult weight gain. We found that low physical activity, both during leisure hours and at work, was associated with an increased risk of pancreatic cancer. Some animal experimental data support the hypothesis that physical activity suppresses pancreatic tumorigenesis.23
Our study used prospective assessment of exposure, thereby avoiding the potential for differential recall that is a concern in many previous studies. The nearly complete end point in our study reduces the potential for bias from differential follow-up. Data from our study suggest that smoking, obesity and low physical activity increase the risk of pancreatic cancer and that the consumption of pork and sausage may have a protective effect. Our findings do not support the hypothesis that coffee or ethanol increases pancreatic cancer risk.
Bengt ISAKSSON*, Fredrik JONSSON, Nancy L. PEDERSEN, Jörgen LARSSON, Maria FEYCHTING and Johan PERMERT