Menopausal hormone therapy and risk of colorectal cancer in the European Prospective Investigation into Cancer and Nutrition

Abstract

Menopausal hormone therapy (HT) may influence colorectal cancer risk. A total of 136,275 postmenopausal women from the European Prospective Investigation into Cancer and Nutrition were followed for an average of 9 years, during which time 1,186 colorectal cancers were diagnosed. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using Cox proportional hazards models stratified by center and age, and adjusted for body mass index, smoking, diabetes, physical activity and alcohol consumption. Compared to never use of HT at study enrolment, current use of estrogen-only (HR, 1.02; 95% CI, 0.79–1.31) or estrogen plus progestin (HR, 0.94; 95% CI, 0.77–1.14) was not significantly associated with the risk of colorectal cancer, and these associations did not vary by recency, duration, route of administration, regimen or specific constituent of HT. Our results show no significant association of estrogen-only or estrogen plus progestin therapy with colorectal cancer risk.

A large number of women have used menopausal hormone therapy (HT), and many will continue its use for treatment of acute menopausal symptoms. Observational studies published in the 1980s and 1990s have shown that HT is associated with a decreased risk of colorectal cancer.1–3 Most of these studies were conducted in North America and have not considered the type of HT formulation but were conducted at a time when estrogen-only therapy was commonly used. The Women's Health Initiative trials found that estrogen plus progestin therapy, but not estrogen-only therapy, was associated with a decreased colorectal cancer risk.4, 5 Recent observational studies have reported inconsistent findings by type of HT formulation, and few studies have examined whether the associations differ by route of administration or type of hormonal constituent.6–20

The aim of this study is to evaluate the association of recency, duration, route of administration, regimen and specific constituent of HT with colorectal cancer risk in the European Prospective Investigation into Cancer and Nutrition (EPIC).

Material and Methods

Study cohort

EPIC is a prospective cohort with 23 centers in ten European countries (Denmark, France, Germany, Greece, Italy, The Netherlands, Norway, Spain, Sweden and the United Kingdom [UK]) with more than half a million participants. It was established in the early 1990s to investigate potential dietary, lifestyle and serologic cancer risk factors. The cohort population and data collection procedures have been described in detail elsewhere.21 Briefly, standardized lifestyle questionnaires, anthropometric data and country-specific food questionnaires were collected between 1992 and 2000 from 521,457 participants, of whom 70% were women. Incident cancer cases were identified through linkage to population cancer registries in Denmark, Italy, The Netherlands, Norway, Spain, Sweden and the UK, or with a combination of methods including linkage to health insurance records, cancer and pathology registries and active follow-up of study participants or their next of kin in France, Germany and Greece. Approval for the study was obtained from the ethical review boards of the participating institutions.

Study population

The study population comprised female participants after excluding women with prevalent cancer at recruitment and those who did not return the baseline lifestyle questionnaire. Women from the Swedish (n = 26,920) and Greek (n = 15,313) subcohorts were excluded because of lack of detailed data on HT use. The study population was further restricted to women who were postmenopausal at recruitment (n = 137,397). Menopausal status was defined according to information on menstruation status and ovariectomy.22 Women were considered postmenopausal if they did not report any menses over the past 12 months or if they reported bilateral ovariectomy. We further excluded 32 women who never menstruated and 1,090 women with missing information on both ever and current HT use. Of the final analytic cohort of 136,275 women, 1,186 developed colorectal cancer (839 colon cancer [International Classification of Diseases 10th revision codes: C18] and 347 rectal cancer [C19, C20]) from recruitment until the end of follow-up (from December 31, 2003 to November 23, 2006, according to center). The colorectal cancer cases comprised 190 cases recruited in Denmark, 282 in France, 75 in Germany, 124 in Italy, 162 in the Netherlands, 54 in Norway, 67 in Spain and 232 in the UK.

Assessment of HT

Information on HT use was collected from the baseline questionnaire. Participants were asked about ever and current use of HT, age at start and total duration of use and among current users the route of administration and the brand name, which was coded according to the World Health Organization's Anatomical Therapeutic Chemical classification system.23 Analyses were performed for ever, current and former use of HT, type of current HT use (estrogen-only, estrogen plus progestin and other formulations [which included tibolone, progestin-only and estrogen plus androgen preparations]) and duration of use. For estrogen-only users, analyses were performed by constituent (estradiol compounds and conjugated equine estrogens) and route of administration (cutaneous and oral), and for estrogen plus progestin users, by progestin constituent (micronized [natural] progesterone, progesterone derivatives and testosterone derivatives)24 and regimen (continuous [progestin used daily throughout the cycle] and sequential [progestin not used daily but usually added for 10–14 days per month]). Route of administration and brand name of former HT use were also available for a subset of participants (mostly in France and Norway); thus, only sensitivity analyses were performed using this information.

Covariate assessment

Weight, height, waist and hip circumference were measured at recruitment, except for the Oxford cohort, the Norwegian cohort and approximately two-thirds of the French cohort, among whom height and weight were self-reported. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Information on smoking status, menstrual and reproductive factors, diagnosis of diabetes mellitus and physical activity was also obtained from the recruitment questionnaire, along with nutrient intakes estimated from the food-frequency questionnaire, details of which are published elsewhere.25–27

Statistical analysis

The distributions of baseline characteristics were compared by HT use among the entire cohort using chi-square tests for categorical variables, analysis of variance or Kruskal-Wallis tests for continuous variables or logistic and linear regression for age- and center-adjusted means and proportions. Hazard ratios (HR) and their 95% confidence intervals (95% CI) for colorectal cancer were estimated using Cox proportional hazards models with age as the underlying time scale. All models were stratified by EPIC recruitment center and age at enrolment (≤50, 51–53, 54–56, 57–59, 60–62, 63–65 and >65 years). The proportionality of hazards was verified based on the slope of the Schoenfeld residuals over time, which is equivalent to testing that the log HR function is constant over time.28 Linear trends were tested by entering appropriate ordinal variables into the model, the coefficients of which were evaluated by the Wald test. All models were adjusted for known or suspected risk factors for colorectal cancer, including smoking status (never, former and current), diabetes mellitus (no and yes), BMI (<25, ≥25–<30 and ≥30 kg/m²), physical activity (inactive, moderately inactive, moderately active and active), oral contraceptive use (never and ever) and alcohol intake (quartiles: <0.58, ≥0.58–<3.92, ≥3.92–<11.94 and ≥11.94 g/day). Missing values were assigned to separate categories for smoking status, diabetes, physical activity, alcohol intake and oral contraceptive use. An analysis that excluded individuals with missing values for any of these covariates produced almost identical results and is not presented here. Further adjustments for waist and hip circumference, waist:hip ratio, reproductive factors (age at menarche and menopause, number of full-term pregnancies and age at the first full-term pregnancy, ever breastfed and cumulative duration of breastfeeding and reason for menopause) and dietary variables (intake of calcium, energy, fiber, folate, red meat and saturated fat) gave very similar results, and these are not presented. Subset analyses were performed by cancer subsite (colon and rectum).

Analyses were also performed according to potentially modifying factors (age at recruitment [at the median, <57 vs. ≥57 years], BMI [at the median, <24.7 vs. ≥24.7 kg/m²], smoking status [ever vs. never smokers] and oral contraceptive use [ever vs. never]). Tests for interaction were carried out by using the relevant exposure variables, indicator variables for the potentially modifying factors and product terms of the two variables. The statistical significance of the interaction terms was evaluated by the Wald test. Country-specific analyses were also performed. To test for heterogeneity of associations by country, by type of HT use and by cancer subsite, a meta-analytic approach was used, and the Q statistic was calculated.29 A sensitivity analysis was performed to evaluate the assumption that HT use reported at recruitment remained constant throughout follow-up, by repeating the analysis after imposing additional censorship on participants at 4, 6, 8 and 10 years of follow-up. All p-values were two sided, and all analyses were performed using STATA version 10 (College Station, TX).

Results

A total of 1,186 colorectal cancer cases were identified among 136,275 postmenopausal women during 1,225,977 person-years of follow-up over an average of 9 years. Table 1 shows the patterns of HT use across Europe. Overall, the frequency of ever HT use was 46% (67% of these being current users) and ranged from 19% in Spain to 69% in Norway. The mean age at first HT use was 49.7 years, and the median duration of use at recruitment was between 3 and 5 years in most countries. Women in Spain and Italy had considerably shorter durations of use, whereas almost one-fifth of the Danish women had been users for more than 10 years at recruitment. Use of combined estrogen and progestin HT was more common than estrogen-only in most, if not all, countries. The most frequently used progestins were norethisterone acetate, norgestrel, micronized progesterone, dydrogesterone and medroxyprogesterone acetate.

Table 1. Use of menopausal hormone therapy (HT) at recruitment among postmenopausal women in the EPIC cohort by country

Compared to never HT users, current users were younger (mean age: 59.0 vs. 55.0 years, respectively), leaner (median BMI: 25.3 vs. 23.8 kg/m²), more physically active (% moderately active or active: 32.5 vs. 40.4%), drank more alcohol (mean: 8.1 vs. 9.8 g/day), were more likely to be current smokers (16.5 vs. 19.2%), have ever used oral contraceptives (36.2 vs. 60.8%) and less likely to be diabetic (3.8 vs. 1.8%; all p < 0.001). Compared to estrogen plus progestin users, women who used estrogen-only therapy were on average older (mean age: 54.5 vs. 56.4 years, respectively), heavier (median BMI: 23.4 vs. 24.6 kg/m²) and less likely to have ever used oral contraceptives (62.4 vs. 56.2%; all p < 0.001). Adjustment for age and center did not alter these associations (data not shown).

Table 2 shows the association of recency, duration, route of administration, regimen and specific constituent of HT with risk of colorectal cancer. Ever and current use of any HT were not statistically significantly associated with colorectal cancer risk. Current use of combined estrogen plus progestin therapy and estrogen-only therapy was also not significantly associated with colorectal cancer risk. No statistically significant associations were found for total or current HT duration, constituent, route of administration or regimen for either current estrogen-only or estrogen plus progestin use. None of these associations varied significantly by country (data not shown).

Table 2. Hazard ratio (HR) and 95% confidence interval (CI) for menopausal hormone therapy (HT) and colorectal cancer among postmenopausal women in the EPIC cohort

When analyzed according to colorectal cancer subsite, there were no significant associations for HT use and risk of colon cancer, although ever HT use was inversely associated with the risk of rectal cancer (Table 3). Current HT use among women using estrogen plus progestin preparations that either contained a testosterone-derived progestin or had a sequential regimen was also inversely associated with rectal cancer risk, but these analyses were based on small numbers and were of marginal statistical significance.

Table 3. Hazard ratio (HR) and 95% confidence interval (CI) for menopausal hormone therapy (HT) patterns of use vs. never use by colorectal cancer site among postmenopausal women in the EPIC cohort

The association of HT use and colorectal cancer was not modified by age, BMI, smoking or ever use of oral contraceptives (data not shown). Excluding women with prevalent diabetes also did not appreciably change the risk estimates (data not shown). There was, however, a significant interaction for former HT use, but not for current use, and colorectal cancer risk by smoking status (p_interaction, 0.03) with an inverse association among ever smokers (HR, 0.73; 95% CI, 0.55–0.98 [57 exposed cases]), and no association among never smokers (HR, 1.11; 95% CI, 0.89–1.38 [103 exposed cases]). Finally, after repeating the analysis imposing additional censorship on participants at 4 (the median duration of HT use in most countries), 6, 8 and 10 years of follow-up, the findings remained the same (data not shown).

Discussion

In this prospective European study, with the largest number of colorectal cancer cases taking estrogen plus progestin therapy reported to date, current use of estrogen-only or estrogen plus progestin therapy was not significantly associated with colorectal cancer risk. This absence of association did not vary by recency, duration, route of administration, regimen or specific constituent of HT.

Results from two clinical trials have shown a lower risk of colorectal cancer with estrogen plus progestin therapy, but no association with estrogen-only HT, although these findings were based on few cases after a relatively short period of follow-up4, 5, 30 (Figs. 1 and 2). Data from observational studies are more inconsistent. Our null findings for estrogen-only HT agree with the WHI trial findings; however, the two largest prospective studies, with over 200 and 450 exposed cases, respectively, both reported significant inverse associations,7, 12 as shown in Figure 1. Several mechanisms have been proposed through which estrogen may exert a protective effect on colorectal carcinogenesis, including its effect on reducing secondary bile acid production,31 circulating insulin-like growth factor-132 or estrogen receptor gene methylation.33

Figure 1. Published results on the association between ever (unless otherwise specified below) estrogen-only therapy vs. never use of any HT and colorectal cancer. †Prospective study simulating the WHI trial comparing conjugated equine estrogens to no HT use. ††Estimates are for current estrogen-only use. ‡Estimated relative risk for colorectal cancer after summarizing, using random effects models, the colon and rectal cancer estimates that were reported in the paper. §Estimates are for current estrogen-only use for 5+ years. *The estimated relative risk is matched for age, sex, clinic and ethnicity but unadjusted for further covariates. ‡‡Estimates are for current estrogen-only therapy per 5 years of use for colon cancer. Abbreviations: HT: menopausal hormone therapy; WHI: Women's Health Initiative; GPRD: General Practice Research Database; CPS-II: Cancer Prevention Study II; BCDDP: Breast Cancer Detection Demonstration Project; CTS: California Teachers Study; EPIC: European Prospective Investigation into Cancer and Nutrition; NR: not reported.

Figure 2. Published results on the association between ever (unless otherwise specified below) estrogen plus progestin therapy vs. never use of any HT and colorectal cancer. §Total relative risk for HERS and HERS II; estimates provided for colon cancer only. *Prospective study simulating the WHI trial comparing conjugated equine estrogens plus norgestrel to no HT use. ††Estimates are for current estrogen plus progestin use. ‡Estimated relative risk for colorectal cancer after summarizing, using random effects models, the colon and rectal cancer estimates that were reported in the article. †Estimates are for current estrogen plus progestin use for 5+ years. **The estimated relative risk is matched for age, sex, clinic and ethnicity but unadjusted for further covariates. ‡‡Estimates are for current estrogen plus progestin therapy per 5 years of use for colon cancer. Abbreviations: HT: menopausal hormone therapy; HERS: Heart and Estrogen/progestin Replacement Study; WHI: Women's Health Initiative; GPRD: General Practice Research Database; CPS-II: Cancer Prevention Study II; BCDDP: Breast Cancer Detection Demonstration Project; CTS: California Teachers Study; EPIC: European Prospective Investigation into Cancer and Nutrition; NR: not reported.

Our finding that estrogen plus progestin HT was not associated with colorectal cancer risk is consistent with some prospective studies,10, 15 although others have reported a reduction in risk,7, 12, 19 as shown in Figure 2. Although our findings differ from those of the WHI trial, it should be noted that the colorectal cancer cases in the estrogen plus progestin arm of the WHI trial were at a more advanced stage than the cases in the placebo arm,4 and a protective effect on colorectal cancer mortality was not seen.15

Some of the inconsistencies in findings for estrogen-only and estrogen plus progestin HT on risk between different studies may be due to the small number of cases in some HT subgroups, resulting in a lack of statistical power to detect a small effect, and to varying duration of use and type of constituent between study populations. For example, the lack of association for estrogen-only HT found in our study is consistent with others in which HT was only used for a short duration of time5, 18; studies that have included women on HT for a longer duration have tended to show significant inverse associations with risk.7, 8, 10, 12 In addition, medroxyprogesterone acetate (a progesterone-derived progestin) is the main progestin prescribed in the US, whereas a wide variety of progestins are used in Europe, including micronized progesterone and testosterone derivatives. Testosterone derivatives have been shown to be, on average, more progestagenic than micronized progesterone and progesterone derivatives.34 However, the type of progestin constituent could not explain our null findings for estrogen plus progestin and colorectal cancer risk in EPIC.

Our finding that use of combined estrogen plus progestin HT and, in particular, use of testosterone derivatives was associated with a reduction in risk of rectal cancer is of potential interest, but is based on small numbers and should be interpreted with caution because of the number of subgroup analyses performed. The studies that have examined the association of HT with colorectal cancer risk by subsite have generally reported similar associations for both colon and rectal cancer,4–7, 12, 14 although one study found a stronger inverse association for the use of estrogen-only therapy with rectal cancer compared to colon cancer, but no differences by subsite for estrogen plus progestin therapy.10 Additional large cohort studies are needed to clarify the associations for estrogen plus progestin HT according to different progestin constituents.

The main strengths of this study lie in its prospective design and the ability to investigate specific types of HT on colorectal cancer risk. However, several factors should be considered in the interpretation of our observational findings. Women who receive menopausal HT may differ from women who do not in other ways that influence colorectal cancer risk. First, women taking hormones are likely to visit their physicians and may undergo examinations more frequently, including screening to remove premalignant adenomas. Although we do not have information about screening attendance among participants in the EPIC study, adjustment for screening sigmoidoscopy did not change the findings in several other large prospective studies.1 Second, although hormone users were younger, leaner and more likely to smoke, drink alcohol and to have previously taken oral contraceptives than nonusers, these variables were included in the final model and had a negligible effect on the risk estimates. However, hormone users may also be more likely to use other medications, such as nonsteroidal anti-inflammatory drugs, which have been shown to reduce the risk of colorectal cancer in some studies.35 Although this might lead to an overestimation of an inverse association found between HT use and colorectal cancer risk, it is unlikely to explain our overall null findings. Another limitation of this study is that information of HT use was only available at recruitment and some women will have ceased use during the follow-up period, especially after the Women's Health Initiative trial was terminated in 2002.36 However, an analysis that additionally censored participants at 4, 6, 8 and 10 years of follow-up produced similar risk estimates to that of the main findings, although it is possible that misclassification may have diluted the associations toward the null.

In conclusion, estrogen-only and estrogen plus progestin therapy were not significantly associated with colorectal cancer risk in EPIC, and risks did not differ by recency, duration, route of administration, regimen and specific constituent of HT.