SEARCH

SEARCH BY CITATION

Keywords:

  • pancreatic neoplasm;
  • risk factors;
  • epidemiology;
  • body weight;
  • physical activity;
  • insulin resistance

Abstract

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

To explore the hypothesis that insulin resistance may be an etiologic factor in pancreatic cancer, we assessed the pancreatic cancer risk associated with anthropometric factors and physical activity, both of which are important determinants of insulin sensitivity in humans. Three hundred and twelve patients with histologically confirmed pancreatic cancer were compared to 2,919 controls in a population-based, case-control study in 7 of the 10 Canadian provinces. Participants were asked to report their exposure status for the period 2 years before interview. Men in the highest quartile of body mass index (BMI, ≥28.3 kg/m2) were at increased risk of pancreatic cancer [adjusted odds ratio (OR) = 1.90, 95% confidence interval (CI) 1.08–3.35]. In addition, men who reported a decrease in weight of at least 2.9% from their lifetime maximum were at reduced risk compared to those reporting a ≤2.9% loss (≥10.2% loss, OR = 0.51, 95% CI 0.30–0.86). BMI 2 years before interview was not associated with pancreatic cancer risk among women, though those reporting a ≥12.5% decrease in weight from their lifetime maximum had substantially lower risk compared to those in the baseline quartile (OR = 0.53, 95% CI 0.29–0.99). After adjustment for age, province of residence, dietary intake and anthropometric factors, men in the highest quartile of the composite moderate and strenuous physical activity index were at reduced risk of pancreatic cancer (OR = 0.53, 95% CI 0.31–0.90). Physical activity did not appear to be associated with pancreatic cancer among women, though a tendency for reduced risk with increasing levels of strenuous activity was suggested (p for trend = 0.06). Our findings support the hypothesis that insulin resistance is an etiologic factor in the development of pancreatic neoplasms among men and possibly women. © 2001 Wiley-Liss, Inc.

Pancreatic cancer is an important public health concern in Western countries, especially considering the extremely grim prognosis and mysterious etiology of this neoplasm.1–5 While there have been consistent positive associations reported between risk of pancreatic cancer and cigarette smoking, this factor has accounted for only a modest proportion of the population-attributable risk for this cancer in most studies conducted to date.1–3

To our knowledge, the roles of adiposity and physical activity have received only limited attention in the epidemiologic literature, and the results have been inconsistent. Positive associations between body mass index (BMI) or obesity and pancreatic cancer have been reported from 3 case-control studies and 2 cohort studies.6–10 In addition, a greater than 10 kg weight gain since age 30 was associated with a 1.8-fold increased risk in a case-control study from Italy.11 A number of earlier investigations, however, reported no association with body weight or BMI.12–18 Maximum lifetime weight was a risk factor for pancreatic cancer in 1 study,6 but Bueno de Mesquita et al.13 reported that there was no association with this factor in a case-control study conducted in the Netherlands. Six previous studies have reported that there was no association between physical activity and risk of pancreatic cancer.19–24 However, care should be taken when interpreting the results of these studies because either they were conducted in nonrepresentative populations19, 21, 23, 24 or the measurement of physical activity was incomplete (missing either occupational or leisure-time activity)22 or subjective.20

Clarifying the importance of anthropometric factors and physical activity in the etiology of pancreatic cancer is of interest in light of the hypothesis suggesting that insulin resistance–related hyperinsulinemia may be important in the pathogenesis of this neoplasm.1, 25 Blood reaches the exocrine pancreas by way of the insulin-producing islets of Langerhans,26, 27 and in vitro studies have indicated that insulin promotes pancreatic tumors.28–30 Additionally, Lui et al.31 demonstrated that N-nitrosobis(2-oxoprophyl)amine–induced peripheral insulin resistance was an early feature of pancreatic cancer in hamsters. In humans, obesity and physical activity are important determinants of variation in insulin sensitivity,32–40 and these variables are the best available surrogate indicators of insulin resistance for use in population-based studies. The Canadian National Enhanced Cancer Surveillance System (NECSS) provided an ideal opportunity to examine these risk factors in Canada in a large, population-based sample of live incident pancreatic cancer patients and population-based controls.

MATERIAL AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The NECSS was developed to better understand the relationship between environment and cancer. New cases of 18 types of cancer were identified through provincial cancer registries in 8 of the 10 Canadian provinces. All pancreatic cancer cases included in the NECSS were histologically confirmed and defined according to International Classification of Diseases, rubric 157.41 The analyses presented here were based on 312 cases and 2,919 controls from 7 provinces (Newfoundland, Prince Edward Island, Nova Scotia, Manitoba, Saskatchewan, Alberta and British Columbia) who completed questionnaires between 1994 and 1997.

For the 18 cancers studied through the NECSS using the same questionnaire protocol (including pancreatic and several other rapidly fatal cancers), the overall response rates were 70% for men and 68% for women, similar to response rates for controls. However, like most other studies of pancreatic cancer that collected information directly from case subjects, the overall proportion of cases that responded was low. Among women diagnosed with pancreatic cancer, 39% had died before they could be contacted and the attending physician refused consent to approach the subject for an additional 18%. For men, 40% had died and physicians refused consent for an additional 15%. Response among those sent questionnaires was 49.7% for women and 51.2% for men; thus, the overall response rates were 21.4% for women and 28.2% for men. The vast majority of cases were ascertained within 1–3 months of diagnosis, physician consent to send questionnaires to patients was obtained within 1 month and 70% of questionnaires were returned within 2 months.

The NECSS used frequency matching in the selection of population controls to obtain age and sex distributions similar to those of all cancer cases combined. Strategies for control selection varied by province depending on data accessibility. In Prince Edward Island, Nova Scotia, Manitoba, Saskatchewan and British Columbia, provincial health insurance plans were used to obtain a random sample of the provincial population stratified by age group and sex. In each of these provinces, >95% of residents are covered by these public plans. Current military personnel and their families as well as indigenous peoples are excluded because they are covered by other plans. In Newfoundland and Alberta, random-digit telephone dialing was used to obtain a population sample. Response rates for controls in the 7 provinces were 64.6% for women and 59.2% for men.

Mailed questionnaires, with telephone follow-up when necessary for clarification, were used to obtain information on residential and occupational histories and other risk factors for cancer. Questionnaires were not sent to persons known to be deceased or to those for whom physician consent was not granted.

The present analysis focused on variables known to play a role in insulin resistance, including adiposity and physical activity. Age, province of residence, dietary energy and fat intake, smoking, alcohol consumption and, among women, parity and age of menarche were considered as potential confounders. BMI was calculated as weight 2 years before interview divided by current height squared (kg/m2). Maximum lifetime BMI was calculated using maximum lifetime weight (excluding pregnancy) and current height. Physical activity 2 years before interview was assessed based on session frequency, seasons participated and average time per session for each of 12 categories of the most common types of moderate and strenuous leisure-time physical activity in Canada. We created a composite index of physical activity by summing the hours per month of moderate and strenuous activity and applying a weighting factor of 9/5 for strenuous activities.42 Individual activities included walking for exercise, jogging or running, gardening or yard work, home exercise or exercise class, golf, racquet sports, bowling or curling, swimming or water exercises, skiing or skating, bicycling, social dancing and other strenuous exercise.

Women were asked to report age during the first menstrual period as well as the number of pregnancies and live births. Dietary intakes of total energy and fat 2 years before interview were determined using data from a 60-item food-frequency questionnaire. This instrument was modeled after 2 instruments that have been extensively validated: the reduced Block questionnaire43 and the instrument used in the Nurses Health Study cohort.44 Minor modifications to the questionnaire items were made to take into account differences between U.S. and Canadian diets. Estimates of total caloric and dietary fat intake were calculated for each individual by substituting the number of kilojoules and grams of fat for each of the items in the diet questionnaire using the Canadian Nutrient Guide.44a Measurement of alcohol and tobacco consumption has been described previously.45

Logistic regression analysis was used to calculate the odds ratios (ORs) for the risk factors of interest. We decided a priori to present only risk estimates that were adjusted for age and province. This was done because age is strongly associated with both the incidence of pancreatic cancer and several of the risk factors under study and because the mechanisms used to identify cases and controls varied by province. Models were fit separately for males and females as the reproductive characteristics of females have distinctive influences on measures of anthropometry and hormonal pathways of insulin resistance.

The influence of other potential confounding variables was taken into account using stepwise regression procedures. This approach represented an objective method to consider the influence of all identified risk factors, where several are highly correlated with each other. A p value of 0.10 was used for entry into the model, while a value of 0.15 was used for removal of the variable. Both age and province of residence were forced into the model. Risk estimates derived using the fully adjusted model excluded subjects who had missing data for any of the risk factors identified in the stepwise regression method. We reran the minimally adjusted models using the same sample size available for the fully adjusted models, to assess the influence of excluding these subjects. The 95% confidence intervals (CIs) were calculated for all ORs, and where the risk factor could be expressed as a continuous variable, a linear test for trend was conducted using the Wald χ2 test statistic.

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

An increased risk of pancreatic cancer was observed among men who reported having a BMI ≥28.3 kg/m2 2 years before interview compared to those with a reported BMI of <23.7 kg/m2. This was the case in both the minimally and the fully adjusted models (Table I; fully adjusted OR = 1.90, 95% CI 1.08–3.35). A maximum lifetime BMI of ≥30.5 was associated with a 2-fold increased risk of pancreatic cancer in the fully adjusted model (OR = 2.05, 95% CI 1.16–3.64). Additionally, men who reported a ≥2.9% decrease in weight from their lifetime maximum were at significantly reduced risk of pancreatic cancer compared to those reporting a decrease of <2.9% (≥10.2% decrease, OR = 0.51, 95% CI 0.30–0.86). Weekly total caloric intake was associated with pancreatic cancer risk in men after adjustment for anthropometric and lifestyle variables (OR = 1.68, 95% CI 1.01–2.81). Weekly dietary fat consumption, cigarette pack-years and alcohol consumption were not associated with pancreatic cancer risk in men after adjustment for confounding factors.

Table I. Risk of Pancreatic Cancer Among Males for Selected Anthropometric Measures and Potential Confounding Variables, NECSS
Variable1CasesControlsAge- and province-adjusted OR,2 95% CICasesControlsMultivariate OR, 95% CI
  • 1

    Variables were categorized into quartiles based on the frequency distribution observed in the male control population.

  • 2

    ORs adjusted for age and province.

  • 3

    Percent change from maximum lifetime weight.

  • 4

    Adjusted for age, province, percent weight change, caloric intake and composite index of physical activity.

  • 5

    Adjusted for age, province, caloric intake, maximum BMI and composite index of physical activity.

  • 6

    Adjusted for age, province, percent weight change, maximum BMI and composite index of physical activity.

BMI (kg/m2)
 <23.7313731.0222771.04
 23.7 to <25.8443711.41, 0.87–2.30362701.79, 1.01–3.19
 25.8 to <28.3403761.34, 0.82–2.20282681.36, 0.74–2.49
 ≥28.3573751.91, 1.20–3.05412591.90, 1.08–3.35
 Unknown110
 p for trend0.00010.03
Maximum lifetime BMI (kg/m2)
 <25.2333681.0232721.04
 25.2 to <27.8503761.55, 0.93–2.47382761.75, 1.00–3.08
 27.8 to <30.5363781.18, 0.72–1.94262751.34, 0.72–2.47
 ≥30.5543661.81, 1.14–2.88402752.05, 1.16–3.64
 Unknown017
 p for trend0.040.02
Percent change in weight3
 <2.9673711.0492661.05
 2.9 to <5.7333750.45, 0.29–0.70212880.35, 0.20–0.61
 5.7 to <10.2343710.50, 0.32–0.78282530.55, 0.33–0.93
 ≥10.2383750.55, 0.36–0.84292670.51, 0.30–0.86
 Unknown113
 p for trend0.140.25
Weekly dietary fat consumption (g/week)
 <261.7303371.0272651.05
 261.7 to <353.6353371.15, 0.69–1.93262711.00, 0.53–1.89
 353.6 to <467.6413371.42, 0.86–2.34342731.20, 0.60–2.39
 ≥467.6463371.66, 1.02–2.71402651.19, 0.55–2.60
 Unknown21157
 p for trend0.00120.02
Caloric intake (kj/week)
 <40,868373371.0312631.06
 40,900 to <50,700253370.67, 0.39–1.15202770.62, 0.34–1.13
 50,700 to <63,952393371.11, 0.68–1.79312641.10, 0.64–1.90
 ≥63,952513371.44, 0.91–2.27452701.68, 1.01–2.81
 Unknown21157
 p for trend0.00600.0019
Cigarette pack-years
 <5464901.0373491.05
 5 to <15172270.76, 0.42–1.36131570.71, 0.36–1.40
 15 to <25162070.79, 0.44–1.44111600.63, 0.28–1.16
 ≥25885571.83, 1.23–2.72643931.50, 0.81–2.11
 Unknown624
 p for trend0.00590.49
Alcohol consumption (drinks/week)
 0464411.0333291.05
 >0 to <3323050.98, 0.61–1.59252191.07, 0.60–1.89
 3 to <7293110.90, 0.55–1.48232311.02, 0.57–1.82
 7 to <14261891.28, 0.76–2.14221301.62, 0.88–2.99
 ≥14332311.39, 0.86–2.24241651.22, 0.67–2.23
 Unknown728
 p for trend0.030.59

Women in the highest BMI quartile (≥27.4 kg/m2) had a 1.5-fold increased pancreatic cancer risk compared to those in the referent category (<22.1 kg/m2) after adjustment for age and province of residence, though this estimate was attenuated in the fully adjusted model (Table II; OR = 1.21, 95% CI 0.70–2.06). In addition, women reporting a ≥12.5% decrease from their maximum lifetime weight experienced a significant reduction after adjustment for age, province of residence and other confounders (OR = 0.53, 95% CI 0.29–0.99). A reduction in risk was suggested among women who reported 4 or more pregnancies compared to those who reported none (OR = 0.51, 95% CI 0.24–1.01), as well as among women reporting an older age at first menstruation (p for trend = 0.0009). Heavy smokers were at significantly increased pancreatic cancer risk, though this was not the case for heavy drinkers.

Table II. Risk of Pancreatic Cancer Among Females for Selected Anthropometric Measures and Potential Confounding Variables, NECSS
Variable1CasesControlsAge- and province-adjusted OR,2 95% CICasesControlsMultivariate OR, 95% CI
  • 1

    Variables were categorized into quartiles based on the frequency distribution observed in female controls.

  • 2

    ORs adjusted for age and province.

  • 3

    Percent change from maximum lifetime weight.

  • 4

    Adjusted for age, province, caloric intake, age at first menstruation and cigarette pack-years.

  • 5

    Adjusted for age, province, caloric intake, age at first menstruation, cigarette pack-years and maximum BMI.

  • 6

    Adjusted for age, province, caloric intake, age at first menstruation, cigarette pack-years and BMI.

  • 7

    Adjusted for age, province, age at first menstruation, cigarette pack-years and BMI.

  • 8

    Adjusted for age, province and age at first menstruation.

BMI (kg/m2)
 <22.1323411.0282861.04
 22.1 to <24.5223580.64, 0.36–1.12203070.64, 0.35–1.18
 24.5 to <27.4343450.97, 0.58–1.63252960.78, 0.44–1.40
 ≥27.4513631.51, 0.94–2.44393021.21, 0.70–2.06
 Unknown0803
 p for trend0.010.39
Maximum lifetime BMI (kg/m2)
 <23.8263461.0233011.04
 23.8 to <26.6373561.29, 0.76–2.19312881.22, 0.68–2.18
 26.6 to <30.4303511.06, 0.61–1.85222990.78, 0.42–1.46
 ≥30.4463501.85, 1.11–3.09363001.43, 0.81–2.53
 Unknown011
 p for trend0.090.79
Percent change in weight3
 <3.5483521.0332941.05
 3.5 to <7.3353500.73, 0.46–1.16312970.94, 0.55–1.61
 7.3 to <12.5333400.75, 0.47–1.21292930.91, 0.53–1.58
 ≥12.5233620.52, 0.31–0.88193040.53, 0.29–0.99
 Unknown010
 p for trend0.130.17
Weekly dietary fat consumption (g/week)
 <236.0253271.0252871.06
 236.0 to <323.0323281.37, 0.79–2.39302951.19, 0.63–2.24
 323.0 to <419.8303271.22, 0.70–2.14283041.19, 0.57–2.50
 ≥419.8323261.41, 0.81–2.45293051.52, 0.68–3.40
 Unknown20106
 p for trend0.610.34
Caloric intake (kj/week)
 <38,178283271.0272801.07
 38,178 to <47,674303281.06, 0.62–1.83292961.01, 0.57–1.79
 47,674 to <57,941383271.40, 0.83–2.35363111.24, 0.72–2.15
 ≥57,941233270.86, 0.86–1.53203040.69, 0.37–1.28
 Unknown20105
 Linear test for trend0.700.71
Number of pregnancies
 0181371.011981.05
 19980.65, 0.28–1.526820.61, 0.20–1.81
 2293140.71, 0.37–1.33292760.94, 0.43–2.08
 3302650.80, 0.43–1.51252360.89, 0.40–2.00
 ≥4525910.55, 0.31–1.00414980.51, 0.24–1.01
 Unknown1901
 p for trend0.080.07
Age at first menstruation (years)
 <12322311.0232141.06
 12 to <15808600.61, 0.39–0.94737870.79, 0.47–1.32
 ≥15182120.54, 0.29–1.01161900.66, 0.33–1.32
 Unknown911598
 p for trend<0.0010.0009
Cigarette pack-years
 >0 to <5628441.0587791.08
 5 to <15171891.36, 0.77–2.40151751.17, 0.64–2.16
 15 to <25191401.88, 1.08–3.28181291.95, 1.09–3.47
 ≥25392192.42, 1.56–3.74372012.38, 1.50–3.76
 Unknown222219
 p for trend<0.0001<0.0001
Alcohol consumption (drinks/week)
 0706581.0675901.04
 >0 to <3263960.59, 0.36–0.94233640.52, 0.31–0.86
 3 to <7181790.92, 0.53–1.61161650.70, 0.38–1.28
 7 to <1412831.14, 0.58–2.2212771.11, 0.56–2.22
 ≥146690.75, 0.31–1.825630.42, 0.15–1.14
 Unknown727520
 p for trend0.670.17

After adjustment for age and province of residence, men who reported participation in ≥8.2 hr/month of strenuous physical activity had a significantly reduced risk of pancreatic cancer compared to those in the baseline category (Table III; OR = 0.59, 95% CI 0.29–0.87). The magnitude of this estimate was reduced slightly in the fully adjusted model. In addition, men in the highest quartile of the composite moderate and strenuous physical activity index were at significantly reduced pancreatic cancer risk in both the minimally and the fully adjusted models (fully adjusted OR = 0.53, 95% CI 0.31–0.90). Physical activity did not appear to be associated with risk of pancreatic cancer among women after adjustment for anthropometric and lifestyle factors, though a tendency for reduced risk with increasing strenuous activity was suggested (Table IV, p for trend = 0.06).

Table III. Risk of Pancreatic Cancer Among Males According to Levels of Moderate and Strenuous Physical Activity Performed 2 Years Prior to Interview, NECCS
Physical activity (hr/month)CasesControlsAge- and province-adjusted OR,1 95% CICasesControlsMultivariate OR, 95% CI
  • 1

    Adjusted for age and province of residence.

  • 2

    Adjusted for age, province, caloric intake, percent change in weight, maximum BMI and composite index of physical activity.

  • 3

    Adjusted for age, province, caloric intake, percent change in weight and maximum BMI.

Moderate activity
 <4.49463241.0412591.02
 4.49 to <12.24323230.67, 0.41–1.09292670.70, 0.36–1.34
 12.24 to <25.25283260.53, 0.53–0.88252840.73, 0.33–1.61
 ≥25.25403250.72, 0.45–1.14322641.21, 0.50–2.91
 Unknown27207
 p for trend0.060.69
Strenuous activity
 0795471.0634571.02
 >0 to <1.6312350.82, 0.53–1.30271930.94, 0.56–1.57
 1.6 to <8.2182340.48, 0.28–0.84182130.68, 0.36–1.26
 ≥8.2202360.59, 0.29–0.87192110.61, 0.28–1.30
 Unknown25253
 p for trend0.210.94
Composite moderate and strenuous physical activity index
 <7.118432911.0412561.03
 7.118 to <19.036362910.81, 0.50–1.31342690.76, 0.46–1.27
 19.036 to <36.66272900.55, 0.33–0.92222740.42, 0.24–0.75
 ≥36.66302910.59, 0.36–0.98302750.53, 0.31–0.90
 Unknown3734225264
 p for trend0.080.04
Table IV. Risk of Pancreatic Cancer Among Females According to Levels of Moderate and Strenuous Physical Activity Performed 2 Years Prior to Interview, NECCS
Physical activity (hr/month)CasesControlsAge- and province-adjusted OR,1 95% CICasesControlsMultivariate OR, 95% CI
  • 1

    Adjusted for age and province of residence.

  • 2

    Adjusted for age, province of residence, cigarette pack-years and age at menarche.

Moderate activity
 <5.2303091.0222231.02
 5.2 to <11.97253090.73, 0.41–1.28192370.76, 0.39–1.50
 11.97 to <22.84273100.75, 0.43–1.32242520.97, 0.51–1.87
 ≥22.84333090.88, 0.51–1.50282491.00, 0.53–1.91
 Unknown24177
 p for trend0.630.98
Strenuous activity
 0605651.0424421.02
 >0 to <0.78242081.19, 0.71–2.00221691.40, 0.81–2.43
 0.78 to <3.88172080.79, 0.44–1.40161800.90, 0.61–1.64
 ≥3.88132110.58, 0.31–1.10121770.68, 0.35–1.32
 Unknown25222
 p for trend0.030.06
Composite moderate and strenuous physical activity index
 <6.11262791.0222501.02
 6.11 to <14.88252780.84, 0.46–1.51242560.98, 0.52–1.83
 14.88 to <28.83342791.06, 0.61–1.86322651.23, 0.68–2.23
 ≥28.83202790.63, 0.34–1.18202570.80, 0.41–1.54
 Unknown3429930256
 p for trend0.150.35

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We have demonstrated associations between obesity and physical activity and risk of pancreatic cancer in a large case-control study conducted in Canada. Our findings are of interest in light of the results of previous research, particularly with regard to the hypothesis that insulin resistance plays a role in pancreatic cancer etiology.1, 25

Our finding of a positive relationship between BMI and pancreatic cancer risk is in agreement with the results of 5 previous studies.6–10 Additional evidence supporting this finding has come from a report which documented weight gain since age 30 as a significant pancreatic cancer risk factor.11 A number of studies, however, did not find congruous relationships between BMI and pancreatic cancer risk.12–18 Explanations for these inconsistencies are not readily apparent, though 2 issues require consideration. First, pancreatic cancer risk appears to be especially pronounced in the upper ranges of adiposity. In 2 studies that did not report positive associations,14, 18 the population prevalence of overweight and obesity may have been relatively low, though this information was not discernable from many reports.6, 9, 13, 15, 17 Second, both weight and BMI are relatively inaccurate surrogate measures of total body adiposity and body fat distribution, the specific parameters of interest with regard to increased insulin resistance. The relationship between BMI and percent body fat has been shown to differ among different ethnic, gender and age groups.46, 47 Further, the distribution of body fat is physiologically important, with insulin resistance being especially pronounced in subjects with high levels of intraabdominal adipose tissue.35 Future studies of anthropometry and pancreatic cancer risk would benefit from the use of superior measures of total body and regional adiposity, such as waist circumference48 and percent body fat assessed by bioelectric impedance analysis.49 Unfortunately, these techniques usually are applicable only in prospective cohort or nested case-control studies and not in a study design such as this, where we relied primarily on retrospective data collected from mailed questionnaires. Our finding of a protective effect of weight loss on risk of pancreatic cancer is of interest given the documentation of improvements in insulin sensitivity after weight loss in obesity-intervention studies.50–52

We have identified a possible role of physical activity in the etiology of pancreatic cancer. While our findings were not universally significant across genders and categories of activity, there are patterns apparent in the results that suggest a protective relationship between physical activity and pancreatic cancer risk. Of particular interest is the consistent reduced risk related to participation in strenuous as well as combined moderate and strenuous activity among men. The lack of association among women may be due to a larger magnitude of nondifferential misclassification in the measurement of physical activity in this gender.53 Six previous studies reported no association between physical activity and risk of pancreatic cancer.19, 20, 21, 22, 23, 24 However, for a number of reasons, care should be taken in interpreting the results of these studies. First, subjects in several studies were members of groups that are not representative of the general population (e.g., longshore workers, university alumni, professional baseball players).19, 21, 23, 24 Although the use of specialized populations in these studies does not invalidate the findings, it does hinder their extrapolation to the general population. Second, only occupational activity was measured in the majority of investigations,19, 21–24 so a possibly important contribution of leisure-time activity may have been missed. Third, the measurement of exercise in the Cancer Prevention Study II did not take into account frequency, duration, intensity or type of activity.20

It was noted above that physical activity has been demonstrated to be inversely related to insulin resistance and hyperinsulinemia in both experimental and population-based studies.33, 36–40 It appears reasonable to suggest that low levels of physical activity over the long term might result in insulin resistance and the consequent chronic exposure of the exocrine pancreas to high levels of insulin,26, 27 which has been shown to have a promoting effect on pancreatic cancer in vitro.28, 29, 30 Further, protective effects of physical activity have been demonstrated in studies of cardiovascular disease,54 type II diabetes,55 and cancers of the breast and colon,56 conditions that are also suspected to be related to insulin resistance.25, 32, 57–59

In the present study, there appeared to be a trend of decreasing risk of pancreatic cancer with increasing age at first menstruation. This finding is consistent with 3 of 4 previous studies that have examined early age at menarche and pancreatic cancer risk.13, 16, 60 No association was found among Chinese women living in Shanghai,7 though the youngest age group cutoff in this study was higher than in the 3 other reports (≤147vs. ≤1113 or 1316, 60). It is possible that this association operates through obesity. Insulin resistance in children and adolescents is strongly related to total body and regional obesity,61 and a number of studies have reported positive associations between these adiposity variables and age at menarche.62–65

It is unlikely that the present findings can be explained by reverse causality (i.e., that high BMI and low physical activity are the result of early-stage disease) for 2 reasons. First, in 1 cohort study9 and 2 nested case-control studies,6, 11 indices of obesity measured at 4.8, 7.5 and 12.5 years before diagnosis were significantly elevated in those who eventually developed pancreatic cancer compared to those who did not. These obesity measures included weight, weight gain, maximum lifetime weight and clinically significant obesity. Second, in another study,66 only 35.4% of pancreatic cancer patients reported prior health disturbances 6 months or less before diagnosis, while an even lower percentage (14.1%) reported such disturbances more than 6 months before diagnosis. Taken together, these findings suggest that it is highly unlikely that symptoms of pancreatic cancer would have influenced physical activity patterns 2 years before diagnosis, the time frame in which exposures were measured in our study.

There are a number of possible limitations to the present study. As mentioned, the measurement of physical activity was probably characterized by nondifferential exposure misclassification, with the possibility of a larger degree of misclassification among women.53 It is likely that this type of information bias reduced some of the risk estimates toward the null value. Further, our physical activity questionnaire did not capture occupational activity, a possibly important segment of the physical activity spectrum. However, in a parallel study of Ontario subjects, very few pancreatic cancer cases and controls were employed in occupations that involved strenuous levels of activity; furthermore, the observed proportions were equivalent (7%) in both cases and controls (data not shown). Recall bias (differential exposure misclassification) is also a possibility, though we do not suspect this to be the case given the lack of well-established risk factors for pancreatic cancer. The NECSS did not collect information on diabetes status, which has been implicated as a risk factor for pancreatic cancer in a number of previous studies.1, 2, 3 However, we do not believe that diabetes confounded the association between obesity and pancreatic cancer because of the likelihood that obesity is the initial exposure in both disease pathways, which subsequently leads to insulin resistance, followed by diabetes32 and/or pancreatic cancer.

The substantial loss of eligible cases due to death, physician refusal and poor response rates to mailed questionnaires was not entirely surprising considering the extremely grim prognosis of pancreatic cancer. Population-based data from Western countries indicate that 1- and 5-year survival rates from this condition can be as low as 11% and 2%, respectively.67 The level of response in the present study was generally similar to that which has characterized other epidemiologic studies of pancreatic cancer not using proxies. For example, Silverman et al.8 were able to directly interview only 46% of eligible cases, while Ghadirian et al.14 interviewed 25% of their cases directly and 75% through proxies. Bias would have been introduced to the present study if rapidly fatal cases had a different etiology from cases with longer survival or if the etiologic factors under study impacted survival, and our results should be interpreted with caution in this regard. However, the majority of cases of pancreatic cancer have either locally advanced or metastatic disease at presentation.67a Furthermore, unlike several other cancers that have highly variable survival rates depending on the histologic stage at diagnosis (e.g., prostate, breast, leukemia, lymphoma, colon), the overwhelming majority of patients diagnosed with pancreatic cancer die rapidly after diagnosis. These points indicate that cancer of the pancreas after clinical presentation is more homogeneous than cancers at other sites and that the impact of excluding those who die early from the analysis would have been small. We chose not to use proxy interviews since this approach results in the introduction of substantial nondifferential misclassification bias to lifestyle measures such as anthropometry, physical activity and diet.8 In addition to diluting the magnitude of associations with the exposures of primary interest (BMI and physical activity), this type of bias would have serious implications for our ability to control for potential confounding influences of dietary components.

We previously reported that cigarette smoking is a significant risk factor for pancreatic cancer in the full NECSS population (including Ontario subjects),45 a finding that is consistent with previous studies.1–3 It has been suggested by some investigators that smoking induces insulin resistance,68–70 though contrary results have also been reported.71, 72

In conclusion, the results of the present study support previous findings suggesting that obesity is associated with risk of pancreatic cancer. Further studies are required, however, to elucidate the specific roles of total and regional adiposity. Additionally, we have presented evidence indicating a protective role of physical activity in the etiology of pancreatic cancer. Taken together with previous in vivo and in vitro findings, these results lend further support to the hypothesis that insulin resistance is important in the etiology of pancreatic cancer. More work is required to demonstrate a causative role for this factor, including cohort studies using more rigorous measures of insulin sensitivity.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

AJGH was supported by a Canadian Institutes for Health Research Postdoctoral fellowship. AJGH and PJV were supported by the Surveillance and Risk Assessment Division, Health, Canada.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIAL AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES