Hormone Contraception before the First Birth and Endometrial Cancer Risk

Endometrial cancer is the most common gynecologic malignancy, and its occurrence is related to both endogenous and exogenous hormone exposures (1–4). Combination oral contraceptives (COC) and menopausal hormone therapy are the most common forms of exogenous hormones. COCs are the dominant form of female contraception in most developed countries, and more than 300 million women have used them worldwide since their introduction in the 1960s (5). Apart from preventing unwanted pregnancy, COC use may have a long-lasting influence on carcinogenesis.

A number of combined studies and systematic reviews have found that longer durations of COC use, and more recent COC use, strongly reduce endometrial cancer risk, regardless of progestin dose (1, 2, 6). In addition, studies report that after cessation of COC use, a reduction in risk persists for 20 years or more (1). Given that COCs were approved for use in North America in the early 1960s, it is only among endometrial cancer cases diagnosed in more recent years that we have the ability to evaluate endometrial cancer risk with use in the distant past (more than 30 years ago) among a high percentage of women. In addition, studies of breast cancer have reported an intriguing long-term effect with the use of COCs before the first full-term pregnancy (7–9), highlighting a specific type of use in the distant past that has not been investigated in endometrial cancer. Therefore, using a case–control study design, we investigated the risks for endometrial cancer associated with COC use, specifically focusing on COC use before the first full-term pregnancy.

The methods used have been reported previously (10). This study received ethical approval from the University of Calgary and the Alberta Cancer Board ethics review boards, and all women provided written informed consent. Briefly, women with first primary endometrial cancer were identified through the Alberta Cancer Registry. Eligible cases were <80 years of age, diagnosed between January 2002 and February 2006, and residents of central/southern Alberta (n = 900). Physicians provided permission to contact 808 cases and 549 (68%) of these completed the interview. Seven cases were excluded because of questionable interviews, resulting in 542 cases. Controls with intact uteri, from the same geographic region, were identified by random digit dialing (RDD) and were frequency-matched to cases in 5-year age groups (11). We screened 60.9% of households and identified 1,984 eligible women of whom 1,036 (52.2%) were interviewed. Four controls were excluded because of questionable interviews, resulting in 1,032 controls.

In the structured, in-person interviews, information was recorded only for exposures that occurred before the reference date among cases (hysterectomy date) and the reference date for controls (an assigned date that preceded the interview date by the average time between hysterectomy and date of interview for the cases). Women provided a variety of information and, to facilitate recall, a hormone calendar was completed during the interview that recorded major life events and reproductive events. Weight and height were measured at interview; body mass index (BMI; kg/m²) was estimated from the average of the three measurements of weight and height, except for five cases and five controls with reported weight and height 1 year before the reference date.

Women reported all premenopausal hormone contraception in the form of birth control pills, shots, or implants. To aid reporting, women were shown colored, photographic displays of common hormone contraception and packaging, as well as a written list. In addition, women were queried about blank periods in the calendar (no pregnancy, no lactation, and no hormone contraception use already indicated) for complete reporting. Users were further questioned about the start and end dates of each episode of use. The specific names of hormone contraception used were not recorded because of interview-time constraints; thus, the exact type/potency of the hormone contraception could not be evaluated.

All variables were coded as in Table 1. Among parous women, we also estimated the duration of COC use before the first full-term pregnancy categorized as never (never use or <0.5 years use), 0.5–1.9 years, 2.0–4.9 years, ≥5 years, and unknown. Menopausal status was coded as premenopausal (menstruating, pregnant, lactating ≤6 months before reference date), peri/postmenopausal [last menstrual period (LMP) ≥12 months before reference date or LMP 6–12 months previously and using hormone therapy], and unknown. Peri-/postmenopausal women were further subcategorized by hormone therapy: estrogen only, estrogen plus progesterone only, other combinations, and unknown.

Unconditional logistic regression was used to estimate ORs and 95% confidence intervals (12). Trends for variables with ≥3 levels of exposure were evaluated with the likelihood ratio test, and P values are two-sided. Final ORs were adjusted for age, residence, parity, BMI, and menopausal status/hormone therapy as categorized in Table 1. Further adjustment for education level, marital status, age at menarche, duration of intrauterine device (IUD) use, smoking, first/second degree family history of uterine or colorectal cancer and number of first/second degree relatives, history of fibroids/endometriosis, history of hypertension, and history of diabetes did not alter the ORs. Most (85%) of our cases were type I and the small percentage of type II cancers precluded a stratified analysis. We performed all analyses using SAS version 9.1.3 and STATA version 10.1.

Overall, cases were more likely than control subjects to be nulliparous, to have a high BMI, to have a family history of uterine or colorectal cancer, and to have a history of fibroids/endometriosis, hypertension, or diabetes (Table 1). Cases were less likely to be married, to have an older age at menarche, and to have used IUDs.

Use of premenopausal hormone contraception was common with a predominance of COC use (Table 2). Any COC use was associated with a reduction in risk (adjusted OR, 0.64; 95% CI, 0.50–0.82), which was stronger with increasing duration of COC use (for each additional year of use, the adjusted OR, 0.95; 95% CI, 0.92–0.98; P_trend = 0.0003) and decreasing time since last use (for each decreasing year since last use, adjusted OR, 0.97; 95% CI, 0.95–0.99; P_trend = 0.004). We also found a clear pattern of decreased risk with more recent use and longer durations of use (P_trend < 0.001), with the greatest reduction in risk among women who had a shorter time since last use (<25 years) and longer duration of use (≥3 years; adjusted OR, 0.31; 95% CI, 0.21–0.46).

We examined COC use before the first full-term pregnancy in an analysis restricted to parous women to see if COCs had a lasting effect beyond the significant hormonal and endometrial changes caused by pregnancy. A strong reduction in risk for endometrial cancer was observed for women who used COCs for ≥5 years before the first full-term pregnancy (adjusted OR, 0.42; 95% CI, 0.25–0.72). A similar reduction was seen with ≥5 years of use before the first full-term pregnancy, even if this was a woman's only use of COCs (adjusted OR, 0.35; 95% CI, 0.18–0.68). Of the 279 postmenopausal women who had any COC use before their first birth, 87 (31%) also used combined continuous estrogen plus progesterone menopausal therapy (29 cases and 58 controls). When we restricted the analysis to postmenopausal women without any hormone therapy use, we still found a reduction in risk with any COC use and with exclusive COC use before the first pregnancy (for ≥5 years any COC use, OR, 0.27; 95% CI, 0.12–0.57; and for exclusive COC use, OR, 0.24; 95% CI, 0.10–0.59; data not shown).

In our study, the use of premenopausal COCs was associated with a reduced risk for endometrial cancer, and risk reductions were greater with increased duration of use and decreased time since last use. COC use is known to be associated with a reduced risk for endometrial cancer, and the reduced risk persists for more than 20 years after cessation (1, 2, 13). In addition, our results provide the first evidence that, among parous women, ≥5 years of COC use before the first full-term pregnancy is associated with a significant reduction in endometrial cancer risk, even if this was a woman's only use of COCs. This risk reduction has not been previously reported in endometrial cancer to our knowledge. A long-term effect of COCs before a first full-term pregnancy has been reported in breast cancer, but in this case it is associated with a modest increase in risk, particularly among women diagnosed under the age of 45 years (7–9). It is hypothesized that the undifferentiated cells of the nulliparous breast may be more susceptible to carcinogenic transformation following exposure to exogenous hormones than the fully developed breast of parous women (9).

Although the more immediate effects of COC use on biologic endpoints such as hormone levels, gene expression, and endometrial proliferation are well documented, surprisingly little is known about the molecular mechanisms underlying the long-term effects. The progestin-related reduction in estrogen-stimulated cellular proliferation with COC use has been hypothesized as one of the main mechanisms underlying endometrial cancer risk reduction (14). Results from a recent randomized controlled trial of COC and progestin-only contraception in high-risk women provide additional support for this theory (15). Progestins promote cell-cycle arrest, induce apoptosis, and regulate expression of multiple signaling pathways involved in oncogenesis, but how these relatively immediate effects relate to a long-term reduction in endometrial cancer risk is still unclear (16–19). Some studies suggest that COC use has a lasting effect on hormone levels: estradiol, estrone, androstenedione, testosterone, and sex hormone-binding globulin levels were lower in past versus never users of COCs (20), and age at menopause was older in users versus nonusers of COCs (21). Although the mechanism behind such long-term effects is unclear, we can speculate that enduring epigenetic changes, progestin-mediated decrease in estrogen receptor expression, or long-term effects on hormone production/metabolism may be involved (20, 22–24). Regardless of the nature of these long-term effects, our results suggest that they endure and last beyond any reduction in risk conferred by one or more full-term pregnancies. Understanding the long-term effects of COC use may guide the timing of chemoprevention efforts that are already being initiated in women at high risk for endometrial cancer (15).

Our study had a number of strengths, including the population-based design, large sample size, detailed information on a wide range of risk factors, and the high prevalence of COC use. Limitations include the relatively low RDD response percentage; however, our controls were similar to the Canadian Community Health Survey population, a survey with a 77.4% response, and our overall results are very consistent with the reported literature (25). We also did not collect COC name/dosage and, thus, could not address the potency of the hormones. Although our interviews were structured with memory aids, cases may have recalled past COC use more fully than controls, but this would result in conservative risk reduction estimates.

In summary, the significant reduction in endometrial cancer risk we observed with ≥5 years of COC use before the first full-term pregnancy among parous women is a novel finding that requires replication. Despite the consistently reported reduction in endometrial cancer risk with COCs, questions remain about the timing of use and the underlying biologic mechanisms of long-term effects. Further research in these areas will help guide future chemoprevention strategies for endometrial cancer.

No potential conflicts of interest were disclosed.

Conception and design: L.S. Cook, A.M. Magliocco, C.M. Friedenreich

Development of methodology: L.S. Cook, Y. Dong, A.M. Magliocco, C.M. Friedenreich

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.S. Cook, C.M. Friedenreich

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.S. Cook, Y. Dong, X. Huang, A.M. Magliocco

Writing, review, and/or revision of the manuscript: L.S. Cook, Y. Dong, P. Round, A.M. Magliocco, C.M. Friedenreich

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.S. Cook, Y. Dong, P. Round, X. Huang

Study supervision: L.S. Cook, C.M. Friedenreich

The authors thank Rita Biel, Lisa Strosher, and Aleata Ryhorchuk for completing study logistics and coordination; Colleen Lachance, Maryann Lester, Lisa Miller, Catherine Munro, and Polly Pratt for RDD recruitment; and Tamara Bellmont, Kay Christie, Pearl Cooke, Linda Davison, Carolyn Henderson, Tacey Lawrence, Rosalie Merkosky, Jodi Parrotta, Brenda Platzer, Cyndi Rasa, Nicole Slot, Keely Winnitoy, and Carol-Anne Zawalykut for conducting subject interviews.

This study was supported by grants from the National Cancer Institute of Canada with funding from the Canadian Cancer Society (#012018). L.S. Cook held a Canada Research Chair (#202105) and a career award from the Alberta Heritage Foundation for Medical Research (AHFMR; #20051033). C.M. Friedenreich held career awards from the Canadian Institutes of Health Research (#87852) and AHFMR (#200100352 and #200700840).