Abstract
To analyze the potential effect of social inequality on pancreatic cancer risk in Western Europe, by reassessing the association within the European Prospective Investigation into Cancer and Nutrition (EPIC) Study, including a larger number of cases and an extended follow-up.
Data on highest education attained were gathered for 459,170 participants (70% women) from 10 European countries. A relative index of inequality (RII) based on adult education was calculated for comparability across countries and generations. Cox regression models were applied to estimate relative inequality in pancreatic cancer risk, stratifying by age, gender, and center, and adjusting for known pancreatic cancer risk factors.
A total of 1,223 incident pancreatic cancer cases were included after a mean follow-up of 13.9 (±4.0) years. An inverse social trend was found in models adjusted for age, sex, and center for both sexes [HR of RII, 1.27; 95% confidence interval (CI), 1.02–1.59], which was also significant among women (HR, 1.42; 95% CI, 1.05–1.92). Further adjusting by smoking intensity, alcohol consumption, body mass index, prevalent diabetes, and physical activity led to an attenuation of the RII risk and loss of statistical significance.
The present reanalysis does not sustain the existence of an independent social inequality influence on pancreatic cancer risk in Western European women and men, using an index based on adult education, the most relevant social indicator linked to individual lifestyles, in a context of very low pancreatic cancer survival from (quasi) universal public health systems.
The results do not support an association between education and risk of pancreatic cancer.
Introduction
The incidence and mortality of pancreatic cancer have undergone a parallel rise in Europe and North America in the last decades (1). Meanwhile, pancreatic cancer 5-year survival is among the lowest of the common cancers (7%), and treatment advances have been minimal, despite the high-quality and near-universal coverage of health systems in Western Europe.
It is established that most pancreatic cancer are noninherited, although family history also conveys a higher disease risk (2). Nevertheless, little is known on the etiopathogenesis of pancreatic cancer, and effective screening tests are lacking.
Previous literature suggests a causal role for body fatness and, probably, adult height, where an evidence for alcohol, red or processed meat, and other dietary factors is limited or inconclusive (2). The established pancreatic cancer risk factors are tobacco smoking, body mass index (BMI), diabetes, and chronic pancreatitis (3).
Social determinants are linked to lifestyle cancer risk factors. However, a preceding study on the association of pancreatic cancer with socioeconomic status within the European Prospective Investigation into Cancer and Nutrition (EPIC) was inconclusive (4). This is the reason why we reanalyze this association including a larger number of cases and a longer follow-up using updated end-point data from the EPIC cohorts.
Materials and Methods
Details on study methods and sample characteristics can be found elsewhere (4, 5). EPIC recruited volunteers from 10 European countries between 1992 and 2000, who were 35 to 70 years old at baseline. A relative index of inequality (RII) was estimated on the basis of an educational ranking of individuals within each sex, age groups, and center (4). Of the 491,992 participants without prevalent cancer, those without baseline lifestyle or dietary information (n = 6,259), extreme energy reporters (n = 9,573), and individuals with missing data on education (n = 16,931, including 19 pancreatic cancer cases) were excluded. Furthermore, participants who developed a different primary cancer prior to a pancreatic and neuroendocrine cancer (n = 54) or nonmalignant tumors (n = 5) were censored at the date of the event, leaving a final sample of 457,947 noncases and 1,223 pancreatic cancer cases, with a mean follow-up of 13.9 (± 4.0) years and 6,401,413 person-years (Supplementary Table S1).
The RII was estimated through Cox regression with age as the time variable. Effect modification was evaluated by sex, age, BMI, smoking, alcohol, diabetes, and European region. Interactions were assessed using likelihood ratio tests. Sensitivity analyses were conducted to test the robustness of results against potential biases due to reverse causation or residual confounding.
Analyses were conducted using R version 3.3.2, and two-sided P values <0.05 were considered statistically significant.
Results
Table 1 shows baseline participants' characteristics by the educational ranks of RII. An inverse and statistically significant social trend was found in models adjusted for age, sex, and center for both sexes combined [HR of RII, 1.27; 95% confidence interval (CI), 1.02–1.59], which was stronger among women (HR, 1.42; 95% CI, 1.05–1.92; Table 2). Multivariate adjustment attenuated RII estimates causing the loss of statistical significance. Results were similar when considering education as the exposure.
Baseline participants' characteristics in the EPIC Study by the RII
| . | RIIa . | |||
|---|---|---|---|---|
| . | (0–0.25) . | (0.25–0.50) . | (0.50–0.75) . | (0.75–1.00) . |
| N | 112,542 | 99,384 | 149,168 | 98,135 |
| Person-years/cases | 1,544,380/270 | 1,447,272/260 | 2,031,255/337 | 1,378,506/356 |
| Women | 79,201 (70.4) | 66,520 (66.9) | 110,843 (74.3) | 64,274 (65.5) |
| Age (years) | 51.2 (8.5) | 48.7 (11.1) | 51.6 (9.7) | 52.5 (9.5) |
| BMI (kg/m2) | 24.6 (3.8) | 24.9 (4.0) | 26.0 (4.6) | 26.1 (4.3) |
| Normal weight | 67,718 (60.2) | 56,077 (56.4) | 69,011 (46.3) | 43,954 (44.8) |
| Overweight | 35,450 (31.5) | 33,326 (33.5) | 53,340 (35.8) | 38,389 (39.1) |
| Obese | 9,374 (8.3) | 9,981 (10.0) | 26,817 (18.0) | 15,792 (16.1) |
| Smoking | ||||
| Never | 54,101 (48.1) | 47,451 (47.7) | 80,811 (54.2) | 42,639 (43.4) |
| Former | 33,065 (29.4) | 27,929 (28.1) | 33,824 (22.7) | 26,885 (27.4) |
| Current | 22,858 (20.3) | 23,105 (23.2) | 31,396 (21.0) | 27,215 (27.7) |
| Physically inactive | 19,420 (17.3) | 16,633 (16.7) | 38,757 (26.0) | 19,998 (20.4) |
| Diabetes | 2,145 (2.0) | 1,645 (1.8) | 5,492 (3.9) | 2,842 (3.1) |
| Energy intake (kcal/day) | 2,109 (596) | 2,054 (606) | 2,075 (629) | 2,070 (649) |
| Fruit and vegetable intake (g/day) | 455.1 (274.6) | 409.5 (251.7) | 484.6 (278.9) | 390.6 (254.3) |
| Red and processed meat intake (g/day) | 75.8 (48.2) | 68.5 (49.7) | 76.6 (50.9) | 80.3 (54.5) |
| Alcohol consumers | 101,183 (89.9) | 91,415 (92.0) | 118,897 (79.7) | 85,003 (86.6) |
| Alcohol consumption (g/day)b | 14.98542 (17.548) | 12.80948 (15.672) | 13.88367 (18.631) | 12.46815 (17.928) |
| . | RIIa . | |||
|---|---|---|---|---|
| . | (0–0.25) . | (0.25–0.50) . | (0.50–0.75) . | (0.75–1.00) . |
| N | 112,542 | 99,384 | 149,168 | 98,135 |
| Person-years/cases | 1,544,380/270 | 1,447,272/260 | 2,031,255/337 | 1,378,506/356 |
| Women | 79,201 (70.4) | 66,520 (66.9) | 110,843 (74.3) | 64,274 (65.5) |
| Age (years) | 51.2 (8.5) | 48.7 (11.1) | 51.6 (9.7) | 52.5 (9.5) |
| BMI (kg/m2) | 24.6 (3.8) | 24.9 (4.0) | 26.0 (4.6) | 26.1 (4.3) |
| Normal weight | 67,718 (60.2) | 56,077 (56.4) | 69,011 (46.3) | 43,954 (44.8) |
| Overweight | 35,450 (31.5) | 33,326 (33.5) | 53,340 (35.8) | 38,389 (39.1) |
| Obese | 9,374 (8.3) | 9,981 (10.0) | 26,817 (18.0) | 15,792 (16.1) |
| Smoking | ||||
| Never | 54,101 (48.1) | 47,451 (47.7) | 80,811 (54.2) | 42,639 (43.4) |
| Former | 33,065 (29.4) | 27,929 (28.1) | 33,824 (22.7) | 26,885 (27.4) |
| Current | 22,858 (20.3) | 23,105 (23.2) | 31,396 (21.0) | 27,215 (27.7) |
| Physically inactive | 19,420 (17.3) | 16,633 (16.7) | 38,757 (26.0) | 19,998 (20.4) |
| Diabetes | 2,145 (2.0) | 1,645 (1.8) | 5,492 (3.9) | 2,842 (3.1) |
| Energy intake (kcal/day) | 2,109 (596) | 2,054 (606) | 2,075 (629) | 2,070 (649) |
| Fruit and vegetable intake (g/day) | 455.1 (274.6) | 409.5 (251.7) | 484.6 (278.9) | 390.6 (254.3) |
| Red and processed meat intake (g/day) | 75.8 (48.2) | 68.5 (49.7) | 76.6 (50.9) | 80.3 (54.5) |
| Alcohol consumers | 101,183 (89.9) | 91,415 (92.0) | 118,897 (79.7) | 85,003 (86.6) |
| Alcohol consumption (g/day)b | 14.98542 (17.548) | 12.80948 (15.672) | 13.88367 (18.631) | 12.46815 (17.928) |
NOTE: Values are mean and SD or numbers and percentages.
aEducational rank from most educated (0) to least educated (1), corresponding to the mean proportion of the population with a higher education within the corresponding group of sex, age category, and center.
bMean intake estimated among alcohol consumers only.
Association between education and the RII with pancreatic cancer in the EPIC Cohorts Study
| . | . | . | . | Model 1 . | Model 2 . |
|---|---|---|---|---|---|
| . | . | Person-years . | Cases . | HR (95% CI) . | HR (95% CI) . |
| All | University | 1,570,473 | 233 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 2,842,462 | 479 | 1.08 (0.92–1.28) | 1.05 (0.89–1.23) | |
| Primary or less | 1,988,478 | 511 | 1.20 (1.01–1.42) | 1.12 (0.94–1.33) | |
| Plinear trend | 0.036 | 0.201 | |||
| RII | 6,401,413 | 1,223 | 1.27 (1.02–1.59) | 1.17 (0.93–1.46) | |
| Plinear trend | 0.030 | 0.173 | |||
| Pnonlinear trend | 0.113 | 0.464 | |||
| Women | University | 1,050,615 | 104 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 2,091,455 | 291 | 1.06 (0.84–1.35) | 1.05 (0.83–1.34) | |
| Primary or less | 1,332,521 | 290 | 1.27 (0.98–1.64) | 1.18 (0.91–1.54) | |
| Plinear trend | 0.046 | 0.164 | |||
| RII | 4,474,592 | 685 | 1.42 (1.05–1.92) | 1.29 (0.95–1.75) | |
| Plinear trend | 0.022 | 0.103 | |||
| Pnonlinear trend | 0.113 | 0.331 | |||
| Men | University | 519,858 | 129 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 751,007 | 188 | 1.12 (0.89–1.41) | 1.04 (0.83–1.31) | |
| Primary or less | 655,957 | 221 | 1.13 (0.89–1.44) | 1.03 (0.81–1.31) | |
| Plinear trend | 0.330 | 0.735 | |||
| RII | 1,926,821 | 538 | 1.12 (0.81–1.54) | 1.00 (0.72–1.38) | |
| Plinear trend | 0.488 | 0.981 | |||
| Pnonlinear trend | 0.729 | 0.923 | |||
| Psex interaction | 0.310 | 0.249 |
| . | . | . | . | Model 1 . | Model 2 . |
|---|---|---|---|---|---|
| . | . | Person-years . | Cases . | HR (95% CI) . | HR (95% CI) . |
| All | University | 1,570,473 | 233 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 2,842,462 | 479 | 1.08 (0.92–1.28) | 1.05 (0.89–1.23) | |
| Primary or less | 1,988,478 | 511 | 1.20 (1.01–1.42) | 1.12 (0.94–1.33) | |
| Plinear trend | 0.036 | 0.201 | |||
| RII | 6,401,413 | 1,223 | 1.27 (1.02–1.59) | 1.17 (0.93–1.46) | |
| Plinear trend | 0.030 | 0.173 | |||
| Pnonlinear trend | 0.113 | 0.464 | |||
| Women | University | 1,050,615 | 104 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 2,091,455 | 291 | 1.06 (0.84–1.35) | 1.05 (0.83–1.34) | |
| Primary or less | 1,332,521 | 290 | 1.27 (0.98–1.64) | 1.18 (0.91–1.54) | |
| Plinear trend | 0.046 | 0.164 | |||
| RII | 4,474,592 | 685 | 1.42 (1.05–1.92) | 1.29 (0.95–1.75) | |
| Plinear trend | 0.022 | 0.103 | |||
| Pnonlinear trend | 0.113 | 0.331 | |||
| Men | University | 519,858 | 129 | 1 (ref.) | 1 (ref.) |
| Secondary or vocational | 751,007 | 188 | 1.12 (0.89–1.41) | 1.04 (0.83–1.31) | |
| Primary or less | 655,957 | 221 | 1.13 (0.89–1.44) | 1.03 (0.81–1.31) | |
| Plinear trend | 0.330 | 0.735 | |||
| RII | 1,926,821 | 538 | 1.12 (0.81–1.54) | 1.00 (0.72–1.38) | |
| Plinear trend | 0.488 | 0.981 | |||
| Pnonlinear trend | 0.729 | 0.923 | |||
| Psex interaction | 0.310 | 0.249 |
NOTE: Model 1, adjusted by sex and stratified by center and baseline age categories; model 2, as model 1, plus further adjustment by smoking intensity, alcohol consumption, BMI, prevalent diabetes, and physical activity. The RII expresses the ratio of the expected pancreatic cancer risk between the most educated (reference) and the least educated participants in the cohorts.
There was no effect modification in stratified analysis (Supplementary table S2). Sensitivity analyses adding new variables or excluding participants caused minor attenuations, which remained not significant (Supplementary table S2). Country-wise exclusion of participants resulted in a significant RII when excluding the Netherlands (RII, 1.29; 95% CI, 1.02–1.63; Supplementary table S3).
Discussion
Education is the most common individual measure of social position because it allows classifying all individuals from young adulthood. Our results do not endorse a social stratification of pancreatic cancer risk in Western Europe, after accounting for major potential confounders.
We cannot discard plausible generation effects and misclassification due to the differences across educational systems. Furthermore, the assumption that all educational categories are hierarchically ordered is not always straightforward, as for vocational and secondary education. However, the alternative use of education as the exposure and the sensitivity analyses conducted exhibited similar associations, supporting the robustness of results. Grouping secondary and vocational education did not result in higher pancreatic cancer risk (Table 2), and the comparison of extreme levels (university versus primary or lower) was not significant either.
Our results are in agreement with an earlier study evaluating the occupational status of United Kingdom's government employees, which did not obtain a significant risk of pancreatic cancer among the least affluent (6). On the contrary, a cohort study performed in Norway found higher risk of pancreatic cancer in farmers versus low occupational groups, which did not change after lifestyle adjustments (7).
Among the limitations, we had no data on developmental factors affecting linear growth (2). Nevertheless, a previous case–control study evaluating serum insulin-like growth factor I (IGF-I) and IGFBP-3 concentrations was unable to support a role for the IGF signaling axis on pancreatic cancer risk (8). Finally, we did not have information on family history (pancreatic cancer is more frequent among family members). However, it is established that over 90% of incident pancreatic cancers are sporadic (mainly attributable to genetic mutations or epigenetic dysregulation), and not inherited.
Conclusions
These results do not support an association between education and risk of pancreatic cancer.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: L. Cirera, J.M. Huerta, M.-D. Chirlaque, K. Overvad, A. Tjonneland, M.-C. Boutron-Ruault, H. Boeing, R. Tumino, E. Weiderpass, N. Larrañaga, K.-T. Khaw, M. Jenab, B. Bueno-de-Mesquita
Development of methodology: L. Cirera, J.M. Huerta, M.-D. Chirlaque, E. Weiderpass, N. Larrañaga, M. Jenab, B. Bueno-de-Mesquita
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.-D. Chirlaque, K. Overvad, M. Lindstrom, S. Regner, A. Tjonneland, M.-C. Boutron-Ruault, V. Rebours, G. Fagherazzi, H. Boeing, A. Trichopoulou, D. Palli, S. Panico, R. Tumino, F. Ricceri, R. Vermeulen, G. Skeie, E. Weiderpass, S. Merino, M.J. Sánchez, N. Larrañaga, M. Sund, K.-T. Khaw, T.J. Key, B. Bueno-de-Mesquita, C. Navarro
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Cirera, J.M. Huerta, M.-D. Chirlaque, E. Weiderpass, N. Larrañaga, B. Bueno-de-Mesquita
Writing, review, and/or revision of the manuscript: L. Cirera, J.M. Huerta, M.-D. Chirlaque, K. Overvad, M. Lindstrom, S. Regner, A. Tjonneland, M.-C. Boutron-Ruault, G. Fagherazzi, V.A. Katzke, H. Boeing, E. Peppa, A. Trichopoulou, D. Palli, S. Grioni, S. Panico, R. Tumino, F. Ricceri, C. van Gils, R. Vermeulen, G. Skeie, T. Braaten, E. Weiderpass, M.J. Sánchez, N. Larrañaga, M. Sund, K.-T. Khaw, T.J. Key, M. Jenab, S. Naudin, N. Murphy, D. Aune, H.A. Ward, E. Riboli, B. Bueno-de-Mesquita, C. Navarro, E.J. Duell
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Regner, H. Boeing, R. Tumino, E. Weiderpass, N. Larrañaga, K.-T. Khaw
Study supervision: L. Cirera, M.-D. Chirlaque, R. Tumino, E. Weiderpass, N. Larrañaga, B. Bueno-de-Mesquita, C. Navarro
Acknowledgments
The authors would like to thank all EPIC cohort-participants, logistic staff, and scientists for their contribution to the study. The coordination of EPIC is financially supported by the European Commission (DG-SANCO) and the International Agency for Research on Cancer (IARC). The national cohorts are supported by Danish Cancer Society (Denmark); Ligue Contre le Cancer, Institut Gustave Roussy, Mutuelle Générale de l’Education Nationale, Institut National de la Santé et de la Recherche Médicale (INSERM; France); German Cancer Aid, German Cancer Research Center (DKFZ), Federal Ministry of Education and Research (BMBF), Deutsche Krebshilfe, Deutsches Krebsforschungszentrum, and Federal Ministry of Education and Research (Germany); the Hellenic Health Foundation (Greece); Associazione Italiana per la Ricerca sul Cancro-AIRC-Italy and National Research Council (Italy); Dutch Ministry of Public Health, Welfare and Sports (VWS), Netherlands Cancer Registry (NKR), LK Research Funds, Dutch Prevention Funds, Dutch ZON (Zorg Onderzoek Nederland), World Cancer Research Fund (WCRF), and Statistics Netherlands (the Netherlands); ERC-2009-AdG 232997 and Nordforsk, Nordic Centre of Excellence programme on Food, Nutrition and Health (Norway); Health Research Fund (FIS), PI13/00061 to Granada; PI13/01162 to EPIC-Murcia, Regional Governments of Andalucía, Asturias, Basque Country, Murcia and Navarra, ISCIII RETIC (RD06/0020; Spain); Swedish Cancer Society, Swedish Research Council, and County Councils of Skåne and Västerbotten (Sweden); Cancer Research UK (14136 to EPIC-Norfolk; C570/A16491 and C8221/A19170 to EPIC-Oxford), Medical Research Council (1000143 to EPIC-Norfolk, MR/M012190/1 to EPIC-Oxford; United Kingdom).
