SEARCH

SEARCH BY CITATION

Keywords:

  • pancreatic cancer;
  • menstruation;
  • reproduction;
  • hormones;
  • estrogens;
  • women;
  • CYP17A1

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Menstrual and reproductive factors and exogenous hormone use have been investigated as pancreatic cancer risk factors in case-control and cohort studies, but results have been inconsistent. We conducted a prospective examination of menstrual and reproductive factors, exogenous hormone use and pancreatic cancer risk (based on 304 cases) in 328,610 women from the EPIC cohort. Then, in a case-control study nested within the EPIC cohort, we examined 12 single nucleotide polymorphisms (SNPs) in CYP17A1 (an essential gene in sex steroid metabolism) for association with pancreatic cancer in women and men (324 cases and 353 controls). Of all factors analyzed, only younger age at menarche (<12 vs. 13 years) was moderately associated with an increased risk of pancreatic cancer in the full cohort; however, this result was marginally significant (HR = 1.44; 95% CI = 0.99–2.10). CYP17A1 rs619824 was associated with HRT use (p value = 0.037) in control women; however, none of the SNPs alone, in combination, or as haplotypes were associated with pancreatic cancer risk. In conclusion, with the possible exception of an early age of menarche, none of the menstrual and reproductive factors, and none of the 12 common genetic variants we evaluated at the CYP17A1 locus makes a substantial contribution to pancreatic cancer susceptibility in the EPIC cohort.

Pancreatic cancer is the 4th leading cause of cancer mortality in both sexes in the US, and the 6th in the European Union.1, 2 Five-year survival is <5%, and incidence is 30–50% higher in men than women. Established risk factors include tobacco smoking, overweight and obesity, history of diabetes, history of pancreatitis, and non-O blood group. Based on the sex difference in pancreatic cancer incidence, and some earlier animal data, it has been hypothesized that hormonal factors related to estrogen exposure may be protective for pancreatic cancer.3, 4 Hormonal, menstrual and reproductive factors have been investigated as pancreatic cancer risk factors in numerous case-control and cohort studies, but with inconsistent results.5–35

CYP17A1 (located on chromosome 10q24.3 and spanning 6569 bp) encodes cytochrome p450c17α, an enzyme with 17α-hydroxylase and 17,20-lyase activities essential in the biosynthesis of glucocorticoids and sex steroids. Genetic variation in CYP17A1 has been associated with steroid hormone levels, menstrual factors and risk of endometrial, breast and prostate cancers, albeit inconsistently.36–38 A recent population-based, case-control study of CYP17A1 variant -34T/C A1/A2 (rs743572) and pancreatic cancer from the San Francisco Bay area reported a statistically significant inverse association between the A2 allele and pancreatic cancer risk.39

We conducted a prospective cohort analysis of mentrual and reproductive factors and exogenous hormone use in relation to pancreatic cancer risk in the EPIC cohort. We also investigated genetic variation at the CYP17A1 locus in relation to risk of pancreatic cancer in a nested case-control study from the EPIC cohort.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

EPIC cohort

The EPIC cohort includes a total of 521,457 participants (368,010 women and 153,447 men) recruited through 23 centers in 10 European countries including Denmark (Aarhus, Copenhagen), France, Germany (Heidelberg, Potsdam), Greece, Italy (Florence, Turin, Varese, Naples, Ragusa), The Netherlands (Bilthoven, Utrecht), Norway, Spain (Asturias, Granada, Murcia, Navarra, Guipuzcoa), Sweden (Malmö, Umeå) and the United Kingdom (Oxford, Cambridge).40 Most of the EPIC participants were enrolled between 1992 and 1998 and between the ages of 35 and 70 years. Participants were recruited from the general population residing in geographic areas including towns and provinces except for the French cohort recruited from a teacher's organization health insurance program, cohorts consisting of women attending breast cancer screening programs (Utrecht and Florence), parts of the Italian and Spanish cohorts in which participants were recruited among blood donors, and most of the Oxford cohort in which participants were recruited among vegetarian volunteers. Eligible participants gave written informed consent and completed questionnaires on diet, lifestyle, and medical history. Ethical review boards from the International Agency for Research on Cancer (IARC) and local centers approved the study. Participants were excluded from analyses because they were prevalent cases of cancer at recruitment (n = 23,633) or because they had no follow-up information (n = 3,448). The following also were excluded from the cohort analyses: participants with missing lifestyle or dietary information or extreme energy intake (n = 15,892); all men (n = 142,601); and women with missing information on all menstrual and reproductive variables (n = 7,249). The final number of EPIC women available for the full cohort analyses of reproductive factors and pancreatic cancer risk was 328,610.

Baseline diet and lifestyle questionnaires

Usual diet over the previous 12 months was measured at enrollment using country-specific validated food questionnaires.40 A separate lifestyle questionnaire included questions on lifetime smoking, lifetime alcohol consumption, education, occupation, menstrual and reproductive history, ever-use of exogenous hormones (oral contraceptives, OC and hormone replacement therapy, HRT), physical activity and history of illness including surgical procedures. Body mass index (BMI) was calculated by dividing weight by height squared (kg m−2). Smoking status was defined as follows: never, former time since quit ≥10 years, former time since quit <10 years, current <20 cigarettes per day, and current ≥20 cigarettes per day. Education was defined as none, primary, technical or professional, secondary and university. Diabetes history was based on self-report at cohort enrollment. Average daily alcohol consumption at recruitment was in grams (g) per day [defined as none (0 g day−1), ≤6, ≤18, ≤30, and >30 g day−1]. Physical activity was defined using the Cambridge index.41 We also considered dietary variables for average daily fruit and nut intake, vegetable intake, red meat intake and processed meat intake.

Identification of incident pancreatic adenocarcinoma cases

Follow-up of cohort members for these analyses was performed through 2003–2006, depending upon the study center, and was based on information in population cancer registries except in France, Germany, Greece and Naples where a combination of different methods including health insurance records, hospital-based cancer and pathology registries, and active follow-up were used. Follow-up began on the date of EPIC recruitment and ended on the date of pancreatic cancer diagnosis, the date of death or the date of the last complete follow-up, whichever came first. Adenocarcinoma of the exocrine pancreas (from here on referred to as pancreatic cancer) included diagnoses coded as C25 (25.0–25.3, 25.7–25.9) according to the International Classification of Diseases-Oncology (ICD-O) 2nd edition. After exclusions, a total of 304 incident primary exocrine pancreatic cancers were identified in EPIC women, of which 76% were microscopically confirmed. For pancreatic cancers with missing microscopic confirmation, diagnoses were based upon clinical symptoms, physical examination and imaging.

Nested case-control study of pancreatic adenocarcinoma

A nested case-control study of primary incident pancreatic cancer within the EPIC cohort was conducted to investigate genetic susceptibility. This nested study included EPIC cases and controls (men and women) who were part of a larger genome-wide association study (GWAS) of pancreatic cancer conducted in the pancreatic cancer cohort consortium (PanScan).42 PanScan cases and controls from Swedish centers in EPIC were not included in the IARC database and so were not included in these analyses. For each primary incident pancreatic cancer case, approximately one control who was alive and cancer-free at the time of diagnosis of the index case was matched on age (±6 months), sex, study center, date and time of blood collection, fasting status and exogenous hormone use (women only). The total number of pancreatic cancer cases in the nested study was 324 (161 women and 163 men), and the total number of controls was 353 (174 women and 179 men). The number of cases and controls did not match perfectly due to genotyping quality control (QC) criteria in PanScan.42

Genotyping in PanScan utilized the HumanHap500 chip and included 558,542 SNPs after QC.42 EPIC-PanScan genotype data for all SNPs at the CYP17A1 locus (and within 20 kb upstream and downstream) that passed PanScan QC criteria were included in these analyses. This resulted in 12 SNPs (three in the coding regions of the gene, and nine from 5′ and 3′ noncoding regions). Based on r2 >0.8 and SNPs with MAF ≥5% in the HapMap CEU population, these 12 SNPs cover about 74% of known common genetic variation in the CYP17A1 gene.

Statistical analyses

Cohort study.

Hazards ratios (HR) and 95% confidence interval (CI) for reproductive factors and pancreatic cancer risk in the full EPIC cohort of women were calculated using Cox proportional hazards regression models with age as the time scale and adjusted by EPIC study center and age at recruitment. The proportional hazards assumption was evaluated in all models using graphical methods and likelihood ratio tests. All statistical tests were two-sided, and all cohort analyses were performed using STATA (version 10.0, College Station, TX).

We estimated HR in the full cohort of women for the following menstrual and reproductive variables: age at menarche, age at menopause (both categorical, in years), duration of any oral contraceptive (OC) use (categorical, in years), ever use of hormone replacement therapy (HRT), number of full-term pregnancies, age at first full-term pregnancy (categorical, in years), ever having breast fed, miscarriage, induced abortion, ovariectomy, hysterectomy and cumulative duration of menstrual cycling (in years). Cumulative duration of menstrual cycling in years was calculated for each woman in the EPIC cohort as follows: for postmenopausal women, it was the difference between the age at menopause and the age at menarche minus the total amount of time the participant was pregnant (number of full-term pregnancies × 9 months). For premenopausal and perimenopausal women, total menstrual cycling was the difference between the age at recruitment and age at menarche minus the total amount of time that the participant was pregnant. We also estimated cumulative duration of menstrual cycling as described above, but subtracting the total time spent taking OC. Total years of menstrual cycling was categorized based upon quintiles in the entire cohort of women. Self-reported menopausal status was defined as follows: menopausal (menses had ceased in the last 12 months or surgical menopause due to bilateral ovariectomy), perimenopausal (no longer menstruating at the time of recruitment or less than nine menstrual cycles in the past 12 months) and premenopausal (regular menses or more than nine cycles in the past 12 months). The menopausal status of women with missing or incomplete questionnaire data was imputed based on cut-points for age at recruitment (postmenopausal if 55 years or older; perimenopausal if between 46 and 55 years, and premenopausal if 46 years or younger).

In the cohort analyses of pancreatic cancer risk, final models were adjusted for age, center and known or suspected pancreatic cancer risk factors including smoking status (as described above), BMI (continuous in kg m−2), self-reported type 2 diabetes history at baseline, education (as described above), average daily alcohol consumption (grams per day) at recruitment, and average daily energy-adjusted red meat intake (grams per day). Dietary intakes were energy-adjusted using the residual method.43 Physical activity and intakes of fruits/nuts, vegetables and processed meats were evaluated but not included in final models because they were not risk factors for pancreatic cancer or did not change effect estimates by >10%. Potential interactions between smoking status and reproductive and menstrual factors in relation to pancreatic cancer risk were assessed by comparing models with and without product terms using a likelihood ratio test at alpha-level 0.05.

Nested case-control study

In the nested case-control study of CYP17A1 SNPs, odds ratios (OR) and 95% CI were estimated using unconditional logistic regression models. Final models were adjusted for age, sex, country, smoking status, BMI and diabetes history. Potential associations between CYP17A1 SNPs and menstrual and reproductive factors and exogenous hormone use in control women were assessed by means of a chi-square test for categorical variables (such as ever use of postmenopausal HRT) and a nonparametric Wilcoxon rank sum test for continuous variables (such as age at menarche and cumulative duration of menstrual cycling).

Number of risk alleles in CYP17A1

A variable for the total number of risk alleles in CYP17A1 was created by adding at-risk alleles for all independent SNPs (coded as 0, 1 or 2). For each pair of CYP17A1 PanScan SNPs in LD based on HapMap (r2 > 0.8), one was removed to avoid over-counting. For this variable, the following three SNPs were removed from the analysis: rs4290163 removed due to perfect LD with rs6163; and rs3824754 and rs4409766 removed due to perfect LD with rs17115100. Coding for rs6163, rs17115100, rs2486758 and rs284859 was reversed to add only at-risk alleles (OR at or above the null of 1.0) in relation to pancreatic cancer.

Haplotype analysis

Haplotypes frequencies and haplotype-specific OR and 95% CI were estimated using SNPassoc in R. Associations between haplotypes with frequency >5% and pancreatic cancer risk (haplotype-specific ORs) were estimated using the most common haplotype as a reference category. Models were adjusted for age, sex, country, smoking, BMI and diabetes history. Sex-specific models were also run.

In control women, the mean age at menopause and mean total years of menstrual cycling were estimated for each CYP17A1 haplotype. Box-Cox transformation with λ = 2 was applied to satisfy normality assumptions. Associations between CYP17A1 haplotypes and HRT use also were assessed by estimating haplotype-specific OR and adjusting for age and country.

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Menstrual and reproductive factors

The contribution of person-years and pancreatic cancer cases for each country in EPIC shows that the majority of cases for these analyses came from France (n = 65) while the fewest came from Greece (n = 9) (Supporting Information Appendix 1). Baseline use of exogenous hormones and total years of menstrual cycling (lower and upper quintiles) varied somewhat by country (Supporting Information Appendix 1). Analysis of baseline characteristics in EPIC women by ever use of OC and HRT and total years of menstrual cycling (Table 1) show that ever OC users tend to be younger, have less body mass, smoke more, attain more education and have a lower prevalence of diabetes than women in the entire cohort. Ever HRT users tended to be older at baseline, have more body mass, consume more red meat, smoke less and to have attained less education. Women with longer total cumulative years of menstrual cycling tended to be older, have slightly more body mass, consume more alcohol and red meat and have a higher prevalence of diabetes (Table 1). Pancreas cancer cases tended to be older, use less OC and have more total years of menstrual cycling than women in the overall cohort (Table 1).

Table 1. Baseline characteristics of EPIC cohort and pancreas cancer cases by hormonal factors in women
inline image

With the possible exception of age at menarche, none of the menstrual and reproductive factors we analyzed in this study were statistically significantly associated with the risk of pancreatic cancer (Table 2). Having an early age at menarche (<12 years) was associated with an increased risk of pancreatic cancer, but this statistical association was only marginally significant (Table 2). With the exception of quintile 3, total cumulative years of menstrual cycling was not associated with pancreatic cancer risk in this study (Table 2). Total cumulative years of menstrual cycling accounting for OC use was not associated with pancreatic cancer risk and neither variable for years of menstrual cycling showed a dose–response relationship (Table 2).

Table 2. Hazard ratios for reproductive and menstrual risk factors and pancreatic cancer risk in women, EPIC cohort
inline image

Exogenous hormone use

Women who reported any past or current use of exogenous hormones (OC and/or HRT) were not at altered risk of pancreatic cancer (Table 2). There were no statistically significant interactions between exogenous hormone use (OC use, HRT use) and smoking status in relation to pancreatic cancer risk (all LRT p values >0.7, data not shown).

CYP17A1 (nested case-control study)

Based upon HapMap, CYP17A1 rs6163 (in Exon 1) is in perfect LD (r2 = 1) with SNP rs743572 (the 5′-34T/C A1/A2 polymorphism most studied to date, and not included in the Illumina GWAS platform). Thus, we analyzed rs6163 as proxy for rs743572 in these analyses. The following LD relationships were obtained from the HapMap CEU population: rs4290163 (r2 = 1 with rs6163), rs4919683 (r2 = 0.96 with rs6163), and rs619824 (r2 = 0.70 with rs6163).

When we analyzed the association between CYP17A1 SNP genotypes and reproductive factors in control women (n = 164), only the rare allele of rs619824 was associated with never use of hormone replacement for menopause (Mantel Haenszel Chi-square p value = 0.037) (for 1 vs. 0 variant alleles and ever-use of HRT, OR = 0.56, 95%CI = 0.27–1.19; for 2 vs. 0 variant alleles and ever-use of HRT use, OR = 0.35, 95%CI = 0.12–1.05). None of the remaining SNPs in CYP17A1, including rs6163, were associated with the other factors we analyzed including age at menarche, age at menopause, age started using menopausal hormones, age at first full-term pregnancy, number of full-term pregnancies, use of OC (ever/never), cumulative duration of menstrual cycles (with or without considering OC use), ovariectomy (ever/never) or hysterectomy (ever/never) (data not shown).

None of the CYP17A1 SNP genotypes that we evaluated in this study were statistically significantly associated with pancreatic cancer risk in women or men, combined or separately (Table 3). Our adjusted models in the nested case-control study of CYP17A1 SNPs and pancreatic cancer included variables for country and age at recruitment, smoking status, BMI (kg m−2) and diabetes history (known risk factors for pancreatic cancer). Education, dietary intake variables (including red and processed meat), total energy, alcohol consumption and physical activity were not confounders and not included in the models for CYP17A1 SNPs and pancreatic cancer risk. SNP rs10883782 gave the lowest p value for association with pancreatic cancer in these analyses (dominant model, p value = 0.17).

Table 3. Odds ratios for CYP17A1 SNPs and pancreatic cancer in women and men, EPIC nested case-control study
inline image

We examined total number of independent risk alleles in CYP17A1 (4–7, 8–9, 10–11, ≥12) in relation to pancreatic cancer risk and found no evidence of statistically significant associations or trends in men and women combined, or separately (data not shown). We also evaluated pancreatic cancer risk and total number of risk alleles in CYP17A1 as a continuous variable and found no associations (women and men: per allele OR = 1.02, 95%CI = 0.96–1.09; women: per allele OR = 1.04, 95%CI = 0.95–1.15; men: per allele OR = 0.99, 95% CI = 0.90–1.09).

CYP17A1 haplotypes

There were numerous rare (<5%) haplotypes (n = 23) across the CYP17A1 locus, indicating low LD in and around the locus. None of the more common haplotype (≥5%) was associated with pancreatic cancer risk in women and men combined (Table 4), or in women only (data not shown). We also investigated associations between CYP17A1 haplotypes and menstrual factors (age at menarche, total years of menstrual cycling, parity and HRT/OC use) in control women, and found no evidence for associations with any of these factors (data not shown).

Table 4. Frequencies and odds ratios for common (≥5%) haplotypes in CYP17A1 and pancreatic cancer risk in women and men, EPIC nested case-control study
inline image

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

In this prospective cohort analysis of menstrual and reproductive factors, exogenous hormone use and genetic variation in CYP17A1 (based on 12 SNPs), we found no evidence that any of these factors is associated with pancreatic cancer except for the possibility of an early age at menarche (<12 years vs. 13 years) moderately increasing risk. This result was of marginal statistical significance and similar to what was observed in the SEARCH study for menarche at 11 years or younger compared with 12–13 years (OR = 1.8; 95% CI = 1.1–3.1).11 Six other studies12, 18, 20, 31, 32, 35 that evaluated age at menarche and pancreatic cancer reported slight or moderate increased risk with earlier ages, but none was statistically significant; and six studies reported slight inverse or no associations.10, 16, 17, 24, 28, 29 If an earlier onset of menses increases lifetime exposure to estrogen, we would expect a lowering of risk (according the hypothesis that estrogens are protective) which is not what we observed. An alternative explanation, other than a direct effect of estrogen, is an effect of the constellation of early-life and maternal characteristics that have been associated with an earlier onset of menses such as birth weight, maternal smoking, diet, obesity and socioeconomic status.44–47 These characteristics (in particular, smoking and overweight) are known pancreatic cancer risk factors later in life and collectively, could be important in altering risk independent of any putative hormonal effect of an early onset of menses. In support of this possibility, a recent prospective analysis of maternal smoking observed a statistically significant elevated risk of pancreatic cancer among offspring later in life.48

We evaluated 12 CYP17A1 SNPs in men and women from a nested case-control study in the EPIC cohort. We analyzed the data by sex, unadjusted and fully adjusted, and found no evidence for a main effect of these genetic variants on pancreatic cancer risk. Further, in control women from the nested case-control study within EPIC, we assessed each CYP17A1 SNP for association with reproductive and menstrual factors, and the rare allele of rs619824 (r2 = 0.70 with rs6163) was associated with never use of HRT for menopause (p value = 0.037). None of the other variants we tested were associated with menstrual or reproductive factors. Previous case-control studies of breast cancer have reported associations between the A2 allele (of the 5′-34T/C A1/A2 SNP) of CYP17A1 and HRT use and menstrual cycle length, and age at menarche.38 A recent study of CYP17A1 variation using multiple variants and association with age at menarche did not identify an association for any of the seven variants they analyzed (including rs6163 in LD with A1/A2).49 We observed an association between HRT use and rs619824, but not rs6163. Variant rs619824 is located in a noncoding region 9170 bp 3′ of the STP codon of CYP17A1, thus, this variant may be related with post-transcriptional regulation, but the functional relevance of this and most CYP17A1 SNPs is unknown. Despite previous reports of associations between CYP17A1 variants and menstrual factors, we cannot rule out the role of chance in this result as well as the borderline result between age at menarche and pancreatic cancer risk.

Since a recent population-based case-control study identified a main effect for the CYP17A1 A1/A2 SNP and pancreas cancer risk,39 we investigated a SNP is perfect LD with A1/A2 and additional CYP17A1 variants. In the EPIC study, we found no evidence for associations with individual SNPs, haplotypes or total number of at-risk alleles in CYP17A1 and pancreas cancer risk.

Two recent GWAS of cases and controls from the Pancreatic Cancer Cohort Consortium (PanScan) and the International Pancreatic Cancer Case-Control Consortium (PanC4) identified four susceptibility loci for pancreatic cancer.42, 50 Neither scan identified the CYP17A1 locus. Based on these two GWA scans and on our more detailed analyses of the EPIC data that was a part of these two GWA scans, it appears that CYP17A1 is not a susceptibility locus for pancreatic cancer.

Our data should be considered in light of some weaknesses. The precision of our effect estimates was affected by the modest number of cases available for analysis in the nested case-control study (women and men: 324 cases and 353 controls; women: 161 cases and 174 controls). Furthermore, all data on menstrual and reproductive factors and exogenous hormone use were based on self-reported information that could be subject to misclassification error which could bias estimates of effect to the null value. Our study has a number of strengths including its prospective design allowing us to avoid biases present in many retrospective study designs and the availability of DNA and genotype data on a nested sample of pancreatic cancer cases and controls from the EPIC cohort.

In conclusion, with the possible exception of an early age at menarche, no other menstrual or reproductive factors, including allelic variation in the CYP17A1 gene, were associated with pancreatic cancer risk in the EPIC cohort. Further analyses of menstrual factors and pancreas cancer risk in large consortia with thousands of cases are warranted.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. References
  7. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_27875_sm_SuppAppendix1.doc51KSupporting Information

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.