• pancreatic cancer;
  • obesity;
  • height;
  • diabetes;
  • smoking;
  • alcohol;
  • physical activity


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Risk factors for pancreatic cancer, other than smoking and diabetes, are not well-established, especially for women. In a cohort of 1.3 million middle-aged women, followed for 9.2 million person-years for cancer incidence and 11.5 million person-years for mortality, there were 1,338 incident pancreatic cancer cases and 1,710 deaths from the disease. Using proportional hazards models, we calculated adjusted relative risks (RRs) and 95% confidence intervals (CIs) by smoking, height, body mass index (BMI), alcohol consumption, physical activity and history of diabetes. Pancreatic cancer incidence was greater in current than never smokers (RR 2.39, CI 2.10–2.73), the risk increasing with the number of cigarettes smoked. The incidence of pancreatic cancer also increased with increasing BMI (RR 1.34, CI 1.13–1.57 for BMI ≥ 30 vs. 22.5–25 kg/m2), and with a history of diabetes (RR 1.58, CI 1.22–2.03, with vs. without such a history). These factors were also associated with increased mortality from pancreatic cancer. Height, alcohol consumption and physical activity showed little or no association with pancreatic cancer risk. © 2008 Wiley-Liss, Inc.

Pancreatic cancer is the fifth most common cause of death from cancer in women in the United Kingdom.1 Smoking and diabetes are well established risk factors for pancreatic cancer.2–5 Some large prospective studies report that the risk of pancreatic cancer in women increases with body mass index (BMI),6, 7 but others find no association with BMI.8–10 In this report, we use data from a prospective study of more than one million women to examine the separate and joint associations of demographic and lifestyle factors with pancreatic cancer.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

In 1996–2001, 1.3 million middle-aged women who had been invited for screening for breast cancer at breast cancer screening centres throughout England and Scotland joined the study by completing a study questionnaire, which asked about height, weight, smoking, alcohol consumption, strenuous exercise, any exercise, history of diabetes, social and demographic factors, and other personal characteristics. Strenuous physical activity was defined on the questionnaire as exercise “enough to cause sweating or a fast heart rate.” The question on “any exercise” was added to the baseline questionnaire after the first 9% were recruited and so is not available for all women. Full details of the study design and methods are described elsewhere,11 and the questionnaire can be viewed at All participants gave their written consent to take part in the study, and the Oxford and Anglia Multi-Centre Research Ethics Committee approved the study.

All study participants have been flagged on the National Health Service Central Register, so that deaths and cancer registrations are routinely notified to the investigators. For these analyses, the last date of follow-up for cancer incidence was December 31, 2006 in most regions, except in the Thames and West Midlands regions where it was June 30, 2006, North Yorkshire and North West Mersey regions where it was December 31, 2005, and in Scotland where it was December 31, 2002. The last date of follow-up for mortality was December 31, 2007 in all regions.

Women diagnosed before recruitment as having any cancer other than nonmelanoma skin cancer (ICD-10 code C44) were excluded from all analyses. In analyses of incident pancreatic cancer (ICD-10 code C25), eligible women contributed person-years from the date of recruitment to the study until the date of diagnosis of pancreatic cancer, date of death, or the last date of follow-up, whichever was the earliest. In addition, women diagnosed with any other cancer (except nonmelanoma skin cancer) during the follow-up period were censored at the date of diagnosis of that cancer. For analyses of pancreatic cancer mortality, eligible women contributed person-years from recruitment until death from pancreatic cancer, death from some other cause, or end of follow-up, whichever was the earliest.

Cox proportional hazards models were applied using the STATA computing package, taking attained age as the underlying time variable, to obtain relative risks (RRs) and corresponding 95% confidence intervals (CIs). Proportional hazards assumptions were verified by the time-varying covariates method (data not shown).12 Analyses were stratified by broad geographical region (10 areas covered by 10 cancer registries) and quintiles of socioeconomic group based on deprivation index,13 and adjusted where appropriate for height (in five categories with cut-points at 155, 160, 165 and 170 cm), BMI (in six categories with cut-points at 22.5, 25.0, 27.5, 30.0 and 32.5 kg/m2) and smoking (never, former or current smoker). We assigned women with missing values for any of the adjustment variables to a separate category for that variable. All variables were treated as categorical variables. Since for many variables (e.g., height) there is no natural reference category, variances are, where appropriate, estimated by treating the RRs as floating absolute risks. Results are therefore presented in the form of plots of RRs and their corresponding floated standard errors (FSEs) and 95% floated confidence intervals (FCIs).14 The position of the square indicates the value of the RR, and its area is inversely proportional to the variance of the logarithm of the RR, providing an indication of the amount of statistical information available for that particular estimate. All results in the text, and results for factors with only two categories (e.g., history of diabetes) and results that refer to a specific comparison of two categories, are presented in the form of conventional RRs and their corresponding 95% CIs. The p values were calculated by likelihood ratio tests. We also examined whether the effects of smoking, BMI and diabetes were modified by other variables, by fitting models with interaction terms.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The characteristics at recruitment of the study population and those developing and dying from pancreatic cancer are shown in Table I. The 1.3 million women were followed for 9.2 million person-years for incident cancer, during which period there were 1,338 incident cases of pancreatic cancer, and followed for 11.5 million person-years for mortality, during which period there were 1,710 deaths from pancreatic cancer. Those with incident and fatal pancreatic cancer were, as expected, somewhat older than the study population on average and were more likely to be current smokers at the time of recruitment.

Table I. Characteristics of the Cohort at Recruitment and of Those Subsequently Developing or Dying from Pancreatic Cancer
 Population at riskIncident pancreatic cancer casesDeaths from pancreatic cancer
  • 1

    Number of pancreatic cancer deaths exceeds number of incident pancreatic cancer cases because of longer follow-up for mortality, as described in the text.

Number of women1.29 million1,3381,7101
Mean years of follow-up7.2 for incidence; 8.9 for mortality  
Age in years [mean (SD)]55.9 (4.5)57.8 (4.5)57.9 (4.4)
Current smokers21% (252,656)33% (421)34% (553)
Height in cm [mean (SD)]162.0 (6.7)162.1 (6.7)162.2 (6.7)
Weight in kg [mean (SD)]68.8 (12.7)70.0 (13.9)69.6 (13.6)
Body mass index in kg/m2 [mean (SD)]26.2 (4.7)26.6 (5.0)26.4 (4.9)
Alcohol: 14 or more units per week6.6% (84,154)6.9% (91)6.7% (114)
Strenuous exercise more than once per week21% (260,899)21% (265)21% (337)
History of diabetes2.7% (34,775)4.9% (65)4.4% (75)
Upper third of socioeconomic status33% (426,808)32% (421)31% (528)

Figure 1 shows the RRs of incident pancreatic cancer by category of smoking, height, BMI, alcohol use and physical activity reported at recruitment. Current smokers at recruitment were at increased risk of incident pancreatic cancer compared with never smokers (RR 2.39, 95% CI 2.10–2.73). We further classified current smokers into those smoking fewer than 15 and 15 or more cigarettes per day. Compared with never smokers, the RR for incident pancreatic cancer was 2.00 (CI 1.69–2.37) in women smoking <15 cigarettes per day and 2.88 (CI 2.46–3.37) in women smoking 15 or more cigarettes per day.

thumbnail image

Figure 1. Relative risks of pancreatic cancer incidence and mortality by categories of smoking, height, body mass index, alcohol consumption and physical activity. *Totals are not always the same because of missing values. †RR, FSE and FCI denote relative risk, floating standard error and floating confidence interval controlling for age, region, socioeconomic status, smoking, body mass index and height as described in the text. ‡This question was added to the recruitment questionnaire after the first 9% were recruited.

Download figure to PowerPoint

The incidence of pancreatic cancer increased with increasing BMI, with the RR in obese women (with BMI 30.0 kg/m2 or higher) compared with women with BMI 22.5–25.0 kg/m2 being 1.34 (CI 1.13–1.57). Risk of incident pancreatic cancer did not vary significantly across categories of height (p = 0.6) or alcohol consumption (p = 0.2). Risk of pancreatic cancer was not significantly associated with physical activity whether measured by strenuous exercise or any exercise (p = 0.6 and p = 0.8, respectively). It could be argued that BMI and physical activity may share a causal pathway,7, 15 and so we also considered strenuous exercise and any exercise in an alternative analysis that adjusted for smoking and height but not BMI. In this alternative analysis we found no association of incident pancreatic cancer with either any exercise (p = 0.7) or strenuous exercise (p = 0.6).

Figure 1 shows results for pancreatic cancer mortality, as well as for incidence, and the relationships are mostly similar to those for incidence, including significant associations with smoking (p < 0.0001) and BMI (p = 0.02).

Among 34,775 women reporting a history of diabetes at recruitment, there were 65 incident cases of pancreatic cancer during follow-up. Women with diabetes were at increased risk of incident pancreatic cancer compared with women without diabetes (Fig. 2), with a RR of 1.58 (CI 1.22–2.03) of incident pancreatic cancer in women with vs. without diabetes. Because diabetes and obesity may both be considered markers for insulin resistance, we also calculated a RR for diabetes without adjustment for BMI and the RR of incident pancreatic cancer was 1.70 (CI 1.32–2.18).

thumbnail image

Figure 2. Relative risk of pancreatic cancer associated with smoking, body mass index and diabetes, in subgroups. *Totals are not always the same because of small numbers of missing values. †RR, SE and CI denote relative risk, standard error and confidence interval controlling for age, region, socioeconomic status, smoking, body mass index and height as described in the text.

Download figure to PowerPoint

Figure 2 also shows the associations of pancreatic cancer incidence with smoking (current vs. never smokers), BMI (≥25 vs. <25 k/m2) and diabetes within subgroups of women classified by other factors. The RR for incident pancreatic cancer in current smokers compared with never smokers was 2.62 (CI 2.24–3.07) in women of height <165 cm, and 2.03 (CI 1.64–2.51) in women of height 165 cm or greater (p = 0.04). The RRs associated with smoking did not vary by BMI, alcohol or physical activity. The RR associated with BMI ≥ 25 kg/m2vs. BMI < 25 kg/m2, and the RR associated with diabetes vs. no diabetes, did not vary with smoking, height, alcohol or physical activity.

To address possible reverse causality (risk factor changes induced by undiagnosed pancreatic cancer or precancerous conditions of the pancreas), we also calculated RRs for incident pancreatic cancer excluding the first two and then the first four years of follow-up. When this was done the RRs did not vary materially, but the number of events was reduced (Table II).

Table II. Relative Risks for Pancreatic Cancer Incidence by Smoking, Body Mass Index (BMI) and Diabetes, All Data and Excluding the First 2 or 4 Years of Follow-Up
ComparisonAll dataExcluding first two years of follow-upExcluding first four years of follow-up
Cases: exposed/ unexposedRR (95% CI)1Cases: exposed/ unexposedRR (95% CI)Cases: exposed/ unexposedRR (95% CI)
  • 1

    RR denotes relative risk and CI denotes confidence interval. Analyses control for age, region, socioeconomic status, smoking, body mass index and height as described in the text.

Current smoker vs. never smokers421/5162.39 (2.10, 2.73)346/4082.51 (2.17–2.91)249/3052.45 (2.06–2.91)
BMI ≥ 25 vs. < 25 kg/m2719/5571.14 (1.02–1.28)582/4411.17 (1.04–1.33)412/3301.12 (0.96–1.29)
Diabetes vs. no diabetes65/12731.58 (1.22–2.03)49/10211.51 (1.13–2.03)36/7441.57 (1.12–2.20)


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

In this large prospective study of middle-aged UK women, we found that pancreatic cancer incidence and mortality were increased with smoking, with increasing BMI and with a history of diabetes.

Smoking was associated with a greater RR of pancreatic cancer than any of the other factors examined. Current smokers were at two-fold or higher risk than never smokers across all categories of height, BMI, alcohol and physical activity. Cigarette smoking is a well-established risk factor for pancreatic cancer.2, 3 The adjusted RR of 2.4 that we observed in current smokers compared with never smokers is similar to the adjusted RRs between 1.8 and 2.5 observed in other cohort studies of pancreatic cancer in women.6, 16, 17 We observed greater risk in those smoking 15 or more than in those smoking <15 cigarettes per day, consistent with previous reports.6, 16

Prospective studies that have reported on the relationship between obesity and pancreatic cancer in women have reached varying conclusions. An increased risk with increasing BMI has been observed in some,6, 7 but not all,9, 18 cohort studies of middle-aged women in the United States. Prospective studies of women in Japan10 and Korea8 found no association between BMI and pancreatic cancer. A recent meta-analysis found a significant but small association between increased BMI and increased pancreatic cancer risk in women,19 which is consistent with our results.

The increased risk of pancreatic cancer with a history of diabetes is well established. Our adjusted RR of 1.6 is consistent with the RR of 1.7 reported in a recent meta-analysis of cohort studies.5 It is possible that the association is partly due to diabetes induced by undiagnosed pancreatic cancer or precancerous conditions of the pancreas.20, 21 However, it seems unlikely that this wholly explains the association, since we find an association even when diabetes precedes diagnosis of cancer by four years or more (Table II). Meta-analyses have found increased risk of pancreatic cancer even 10 years or more after diagnosis of diabetes.4, 5, 22 We found no strong evidence to suggest that the effect of diabetes varies with other factors, but because diabetes and pancreatic cancer are both relatively rare conditions, even this large study has little power for these tests.

The existing evidence for an association between pancreatic cancer and physical activity is mixed. Some studies have found inverse associations for some but not other measures of physical activity, in analyses restricted to specific groups23 or in analyses not adjusting for BMI.7 Other prospective studies have found no association of physical activity with pancreatic cancer, the largest of these having 402 pancreatic cancers.24 We found no association of pancreatic cancer incidence or mortality with physical activity in this study of 1,338 cases and 1,710 deaths. Further examination is warranted, especially in studies with more detailed measurement of physical activity.

We found little evidence of an association between height and pancreatic cancer and no association between alcohol consumption and risk of pancreatic cancer, although our study has relatively few heavy drinkers and we cannot examine the hypothesis of increased risk with the heaviest levels of alcohol consumption.25

The Million Women Study includes one in four UK women who were aged 50–64 during the period of recruitment, and is the largest study to date of women's health. To our knowledge, the results presented here include more women with incident pancreatic cancer than in any previous cohort study. Previous reports found that women participating in the Million Women Study differ little from the general UK population of women of a similar age.11, 26 As with most very large epidemiological studies, our exposure measurements were based on self-report. A validation study of 2,500 UK women of similar age to our cohort found a close numerical agreement between self-reported and measured BMI.27 It has been shown in many studies, including a substudy of our cohort,28 that diabetes is self-reported with high accuracy,29–31 as is smoking.32

In summary, this study shows that the risk of incident and fatal pancreatic cancer is affected by modifiable lifestyle factors including smoking and obesity. Current smokers have more than twice the risk of pancreatic cancer than never smokers. Other factors such as increasing BMI and diabetes are also associated with increased risk of pancreatic cancer, but the magnitudes of the RRs associated with diabetes, and even the most extreme differences in BMI, are materially lower than that of smoking.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

The authors thank the women who participated in the Million Women Study.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • 1
    Office for National Statistics. Mortality statistics: cause. Review of the Registrar General on deaths by cause, sex and age, in England and Wales, 2005.
  • 2
    IARC Working Group on the Evaluation of Carcinogenic Risks to Humans.Tobacco smoke and involuntary smoking IARC monographs on the evaluation of carcinogenic risks to humans, vol. 83. International Agency for Research on Cancer: Lyon, France, 2004.
  • 3
    Iodice S,Gandini S,Maisonneuve P,Lowenfels AB. Tobacco and the risk of pancreatic cancer: a review and meta-analysis. Langenbecks Arch Surg 2008; 393: 53545.
  • 4
    Everhart J,Wright D. Diabetes mellitus as a risk factor for pancreatic cancer: a meta-analysis. J Am Med Assoc 1995; 273: 16059.
  • 5
    Huxley R,Ansary-Moghaddam A,Berrington de Gonzalez A,Barzi F,Woodward M. Type-II diabetes and pancreatic cancer: a meta-analysis of 36 studies. Br J Cancer 2005; 92: 207683.
  • 6
    Coughlin SS,Calle EE,Patel AV,Thun MJ. Predictors of pancreatic cancer mortality among a large cohort of United States adults. Cancer Causes Control 2000; 11: 91523.
  • 7
    Michaud DS,Giovannucci E,Willett WC,Colditz GA,Stampfer MJ,Fuchs CS. Physical activity, obesity, height, and the risk of pancreatic cancer. J Am Med Assoc 2001; 286: 9219.
  • 8
    Berrington de Gonzalez A,Yun JE,Lee S-Y,Klein AP,Jee SH. Pancreatic cancer and factors associated with the insulin resistance syndrome in the Korean cancer prevention study. Cancer Epidemiol Biomarkers Prev 2008; 17: 35964.
  • 9
    Sinner PJ,Schmitz KH,Anderson KE,Folsom AR. Lack of association of physical activity and obesity with incident pancreatic cancer in elderly women. Cancer Epidemiol Biomarkers Prev 2005; 14: 15713.
  • 10
    Luo J,Iwasaki M,Inoue M,Sasazuki S,Otani T,Ye W,Tsugane S. Body mass index, physical activity and the risk of pancreatic cancer in relation to smoking status and history of diabetes: a large-scale population-based cohort study in Japan—the JPHC study. Cancer Causes Control 2007; 18: 60312.
  • 11
    Million Women Study Collaborative Group. The Million Women Study: design and characteristics of the study population. Breast Cancer Res 1999; 1: 7380.
  • 12
    Collett D. Model checking in the Cox regression. In: ChatfieldC,TannerM,ZidekJ, eds.Modelling survival data in medical researched, Vol. 2. USA: Chapman & Hall/CRC Press, 2003: 1468.
  • 13
    Townsend P,Phillimore P,Beattie A. Health and deprivation: inequality and the North. London: Croom Helm, 1988.
  • 14
    Easton DF,Peto J,Babiker AGAG. Floating absolute risk: an alternative to relative risk in survival and case-control analysis avoiding an arbitrary reference group. Stat Med 1991; 10: 102535.
  • 15
    Stolzenberg-Solomon RZ,Graubard BI,Chari S,Limburg P,Taylor PR,Virtamo J,Albanes D. Insulin, glucose, insulin resistance, and pancreatic cancer in male smokers. J Am Med Assoc 2005; 294: 28728.
  • 16
    Fuchs CS,Colditz GA,Stampfer MJ,Giovannucci EL,Hunter DJ,Rimm EB,Willett WC,Speizer FE. A prospective study of cigarette smoking and the risk of pancreatic cancer. Arch Intern Med 1996; 156: 225560.
  • 17
    Nordlund LA,Carstensen JM,Pershagen G. Cancer incidence in female smokers: a 26-year follow-up. Int J Cancer 1997; 73: 6258.
  • 18
    Stolzenberg-Solomon RZ,Adams K,Leitzmann M,Schairer C,Michaud DS,Hollenbeck A,Schatzkin A,Silverman DT. Adiposity, physical activity, and pancreatic cancer in the National Institutes of Health-AARP diet and health cohort. Am J Epidemiol 2008; 167: 58697.
  • 19
    Renehan AG,Tyson M,Egger M,Heller RF,Zwahlen M. Body-mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008; 371: 56978.
  • 20
    Wang CY,Hsu L,Feng D,Prentice RL. Regression calibration in failure time regression. Biometrics 1997; 53: 13145.
  • 21
    World Congress of Gastroenterology Symposium Participants. Diabetes and pancreatic cancer: the egg or the hen? Abstracts of the symposium. A meeting of the World Congress of Gastroenterology. October 5, 1994, Los Angeles, CA. Int J Pancreatol 1994; 16: 8198.
  • 22
    Stevens RJ,Roddam AW,Beral V. Pancreatic cancer in type 1 and young-onset diabetes: systematic review and meta-analysis. Br J Cancer 2007; 96: 5079.
  • 23
    Stolzenberg-Solomon RZ,Pietinen P,Taylor PR,Virtamo J,Albanes D. A prospective study of medical conditions, anthropometry, physical activity, and pancreatic cancer in male smokers (Finland). Cancer Causes Control 2002; 13: 41726.
  • 24
    Lin Y,Kikuchi S,Tamakoshi A,Yagyu K,Obata Y,Inaba Y,Kurosawa M,Kawamura T,Motohashi Y,Ishibashi T. Obesity, physical activity and the risk of pancreatic cancer in a large Japanese cohort. Int J Cancer 2007; 120: 266571.
  • 25
    Boffetta P,Hashibe M. Alcohol and cancer. Lancet Oncol 2006; 7: 14956.
  • 26
    Banks E,Beral V,Cameron R,Hogg A,Langley N,Barnes I,Bull D,Reeves G,English R,Taylor S,Elliman J,Lole Harris C. Comparison of various characteristics of women who do and do not attend for breast cancer screening. Breast Cancer Res 2002; 4: R1.
  • 27
    Spencer EA,Appleby PN,Davey GK,Key TJ. Validity of self-reported height and weight in 4808 EPIC-Oxford participants. Public Health Nutr 2002; 5: 5615.
  • 28
    Banks E,Beral V,Cameron R,Hogg A,Langley N,Barnes I,Bull D,Elliman J,Harris CL. Agreement between general practice prescription data and self-reported use of hormone replacement therapy and treatment for various illnesses. J Epidemiol Biostat 2001; 6: 35763.
  • 29
    Amaducci L,Baldereschi M,Di Carlo A,Maggi S,Scarlato G,Candelise L,Scarpini E,Grigoletto F,Minicuci N,Volonnino G,Zucchetto M,Bressan M, et al. Prevalence of chronic diseases in older Italians: comparing self-reported and clinical diagnoses. Int J Epidemiol 1997; 26: 9951002.
  • 30
    Bowlin SJ,Morrill BD,Nafziger AN,Lewis C,Pearson TA. Reliability and changes in validity of self-reported cardiovascular disease risk factors using dual response: the behavioral risk factor survey. J Clin Epidemiol 1996; 49: 51117.
  • 31
    Heliovaara M,Aromaa A,Klaukka T,Knekt P,Joukamaa M,Impivaara O. Reliability and validity of interview data on chronic disease. The Mini-Finland Health Survey. J Clin Epidemiol 1993; 46: 18191.
  • 32
    Patrick DL,Cheadle A,Thompson DC,Diehr P,Koepsell T,Kinne S. The validity of self-reported smoking: a review and meta-analysis. Am J Public Health 1994; 84: 108693.