Background
Ovarian cancer risk in BRCA1 and BRCA2 mutation carriers has been shown to decrease with longer duration of oral contraceptive use. Although the effects of using oral contraceptives in the general population are well established (approximately 50% risk reduction in ovarian cancer), the estimated risk reduction in mutation carriers is much less precise because of potential bias and small sample sizes. In addition, only a few studies on oral contraceptive use have examined the associations of duration of use, time since last use, starting age, and calendar year of start with risk of ovarian cancer.
Objective
This study aimed to investigate in more detail the associations of various characteristics of oral contraceptive use and risk of ovarian cancer, to provide healthcare providers and carriers with better risk estimates.
Study Design
In this international retrospective study, ovarian cancer risk associations were assessed using oral contraceptives data on 3989 BRCA1 and 2445 BRCA2 mutation carriers. Age-dependent–weighted Cox regression analyses were stratified by study and birth cohort and included breast cancer diagnosis as a covariate. To minimize survival bias, analyses were left truncated at 5 years before baseline questionnaire. Separate analyses were conducted for each aspect of oral contraceptive use and in a multivariate analysis, including all these aspects. In addition, the analysis of duration of oral contraceptive use was stratified by recency of use.
Results
Oral contraceptives were less often used by mutation carriers who were diagnosed with ovarian cancer (ever use: 58.6% for BRCA1 and 53.5% BRCA2 ) than by unaffected carriers (ever use: 88.9% for BRCA1 and 80.7% for BRCA2 ). The median duration of use was 7 years for both BRCA1 and BRCA2 carriers who developed ovarian cancer and 9 and 8 years for unaffected BRCA1 and BRCA2 carriers with ovarian cancer, respectively. For BRCA1 mutation carriers, univariate analyses have shown that both a longer duration of oral contraceptive use and more recent oral contraceptive use were associated with a reduction in the risk of ovarian cancer. However, in multivariate analyses, including duration of use, age at first use, and time since last use, duration of oral contraceptive use proved to be the prominent protective factor (compared with <5 years: 5–9 years [hazard ratio, 0.67; 95% confidence interval, 0.40–1.12]; >10 years [hazard ratio, 0.37; 95% confidence interval, 0.19–0.73]; P trend =.008). The inverse association between duration of use and ovarian cancer risk persisted for more than 15 years (duration of ≥10 years; BRCA1 <15 years since last use [hazard ratio, 0.24; 95% confidence interval, 0.14–0.43]; BRCA1 >15 years since last use [hazard ratio, 0.56; 95% confidence interval, 0.18–0.59]). Univariate results for BRCA2 mutation carriers were similar but were inconclusive because of limited sample size.
Conclusion
For BRCA1 mutation carriers, longer duration of oral contraceptive use is associated with a greater reduction in ovarian cancer risk, and the protection is long term.
Why was this study conducted?
The use of combined oral contraceptives is a strong protective factor for ovarian cancer and has been suggested as chemoprevention for BRCA1 and BRCA2 mutation carriers. Previous studies on oral contraceptive use were limited in sample size, and other than duration of use, only a few studies have examined the other characteristics of oral contraceptive use.
Key findings
For BRCA1 mutation carriers, longer duration of oral contraceptive use is associated with a reduction in ovarian cancer risk, and the protection is long term. Findings for BRCA2 mutation carriers were similar but less definitive given the smaller sample size.
What does this add to what is known?
To date, most studies have examined the association of duration of oral contraceptive use with risk of ovarian cancer, without taking other characteristics of oral contraceptive use into account. For BRCA1 mutation carriers, we have shown that the duration of oral contraceptive use is indeed more important than recency of use or starting age. Moreover, the strong protective effect of long duration of oral contraceptive use persists for a long period. Current results are based on a relatively large sample of BRCA1 and BRCA2 mutation carriers and corrected for potential testing and survival biases.
Introduction
Mutations in the BRCA1 and BRCA2 genes are associated with a high lifetime risk of ovarian cancer. The average cumulative risk of ovarian cancer up to the age of 70 years has been estimated to be 41% (95% confidence interval [CI], 33–50) for BRCA1 mutation carriers and 15% (95% CI, 10–23) for BRCA2 mutation carriers. The use of oral contraceptives is a strong protective factor (approximately 50%, with a dose-response association observed with duration of use) for ovarian cancer in the general population and has been suggested as chemoprevention for BRCA1 and BRCA2 mutation carriers. Although the effects of using oral contraceptives in the general population are well established, the estimated risk reduction in mutation carriers is much less certain and precise because of potential bias and small sample sizes. All previous studies were retrospective and therefore susceptible to survival bias. Only 1 study, as a sensitivity analysis, minimized potential survival bias by restricting the analyses to person-years within the 3 years before study enrollment. So far, almost all previous BRCA1 and BRCA2 mutation carrier studies restricted analyses to risk of ovarian cancer and duration of use of oral contraceptives.
To provide carriers with better risk estimates, we wanted to investigate in more detail the association between oral contraceptive use and risk of ovarian cancer. We used retrospective data from the International BRCA1/2 Carrier Cohort Study (IBCCS). Here, we were able to mutually adjust for multiple characteristics of oral contraceptive use to better understand their associations with ovarian cancer risk. To minimize the potential for survival bias, we used a left-truncated approach and conducted full-cohort retrospective analyses (ie, without left truncation) for comparison with the literature.
Materials and Methods
Study group
The IBCCS is a collaborative European study of women carrying a pathogenic or likely pathogenic germline mutation in BRCA1 or BRCA2 . Women were eligible if they were between 18 and 80 years of age at recruitment. More than two-thirds of participants were enrolled to 1 of the 3 large ongoing nationwide studies in the United Kingdom and Ireland (Epidemiologic Study of Familial Breast Cancer), France (Gene Etude Prospective Sein Ovaire), and the Netherlands (Hereditary Breast and Ovarian Cancer Research, Netherlands). For the current analyses, women with both BRCA1 and BRCA2 mutations were excluded. In addition, women born before 1920 were excluded, because their reproductive years preceded the availability of oral contraceptives.
Data collection
A baseline questionnaire elicited detailed information on known or suspected risk factors for breast and ovarian cancer. Data on preventive surgeries and cancer occurrence were collected from medical records or linkages to cancer and pathology registries (75%) or questionnaires (25%). Participants provided written informed consent, and each study was approved by the relevant institutional ethical committee.
Statistical analysis
To estimate hazard ratios (HRs), time-dependent Cox proportional hazards regression models with age as the timescale were used, stratified for birth cohort and study. To reduce the possible impact of survival bias, analyses were left truncated, restricting the analyses to person-years within 5 years before study enrollment (age at baseline questionnaire). This implies that we started follow-up 5 years before baseline, with women at risk of developing ovarian cancer (so at least 1 ovary in situ: BRCA1 [n=3989] and BRCA2 [n=2445]). For those, who were diagnosed with ovarian cancer during the next 5 years, the mean survival was 3.2 years for BRCA1 mutation carriers and 2.9 years for BRCA2 mutation carriers, ranging from 0 to 5 years. Person-years were calculated up to the diagnosis of ovarian cancer (event of interest), diagnosis of another cancer (with the exception of breast cancer and nonmelanoma skin cancer), risk-reducing salpingo-oophorectomy (RRSO), mutation testing, or baseline questionnaire completion, whichever came first. Because of the retrospective nature of the study and because only person-years before DNA test was included, women were not aware of their mutation during the ages, the person-years, included for analysis. Breast cancer diagnosis was included as a time-dependent covariate. To correct for the potential testing bias, analyses were performed using the extended weighted regression approach described by Antoniou et al. , Cancer cases are more often genetically tested on unaffected women. Therefore, the cancer incidence in a retrospective cohort of mutation carriers is overestimated, and the estimated HRs are underestimated. To correct for this testing bias, ovarian and breast cancer cases and unaffected women were weighted differentially to ensure that age-specific incidence rates implied by the weighted cohort were consistent with known incidence rates for women with a BRCA1 or BRCA2 mutation. Carriers who developed breast or ovarian cancer were underweighted (weights <1), and the unaffected carriers were overweighted (weights >1). In general, the unweighted HR estimates were closer to the null value than the weighted HRs, and the 95% CIs are narrower. For instance, forever vs never use unweighted HRs were 0.72 (95% CI, 0.58–0.90) for BRCA1 and 0.80 (95% CI, 0.55–1.17) for BRCA2 mutation carriers, compared with the weighted HRs of 51 (95% CI, 0.36–0.71) and 0.65 (95% CI, 0.35–1.19) for BRCA1 and BRCA2 mutation carriers.
The effect of familial clustering on estimates of precision was accounted for using robust variance estimation. Trend tests were based on modeling the category-specific mean as a continuous variable. We conducted a separate analysis for duration of use, time since last use, and starting age (“oral contraceptive univariate”) and a multivariate analysis, including all of these aspects of oral contraceptive use (“oral contraceptive multivariate”). In addition, the analysis of duration of oral contraceptive use was stratified by recency of use. All characteristics of oral contraceptive use were considered time-dependent covariates, computed for each year of observation.
None of the potential confounders (family history, parity number of pregnancies, and menopausal status) changed the HRs for oral contraceptive use and ovarian cancer risk by more than 10%, and therefore, they were excluded from the final models.
Sensitivity analyses were composed of the following: (1) stratified analyses (birth cohort, study, and attained age), (2) left-truncated analyses censored for breast cancer diagnosis, (3) multiple imputations for missing covariate data, and (4) multiple imputations with a random-effects Cox model approach, where we considered study site as a random term. Covariates were imputed 50 times in 5 iterations. Covariates were imputed with multivariate imputation by changed equations, using conditional multiple imputations that follow an iterative procedure. Furthermore, we conducted “full-cohort” retrospective analyses, where the analysis included person-years from birth instead of being left truncated.
All statistical tests were two-sided and a P value of <.05 was considered statistically significant. Trend tests were based on the P value for the continuous variable based on fitting category-specific means. Analyses were performed using Stata (version 13; StataCorp, College Station, TX), except for the multiple imputations and mixed model sensitivity analyses for which R (version 4.0.2; R Foundation, Vienna, Austria) was used.
Results
In the left-truncated analyses of 3989 BRCA1 mutation carriers, 346 women (8.7%) were diagnosed with ovarian cancer at censoring ( Table 1 ). Ovarian cancer cases completed their questionnaire on average 1.8 years (range, 0–5 years) after their ovarian cancer diagnosis. Of the 3642 BRCA1 mutation carriers (91.3%) who were unaffected by ovarian cancer, 2.4% were censored at age of RRSO. Of the 2445 BRCA2 mutation carriers, 106 women (4.3%) were diagnosed with ovarian cancer at censoring. BRCA2 ovarian cancer cases completed their questionnaire on average 2.1 years (range, 0–5 years) after their ovarian cancer diagnosis. Of the 2339 BRCA2 mutation carriers (95.6%) who were unaffected with ovarian cancer, 1.8% were censored at age of RRSO. For both BRCA1 and BRCA2 mutation carriers, compared with ovarian cancer cases, women unaffected with ovarian cancer were younger ( BRCA1 , 40.5 vs 51.7 years; BRCA2 , 43.4 vs 56.9 years) and thus born more recently (birth year 1952–1980: 81.0% vs 43.4% for BRCA1 and 76.7% vs 29.3% for BRCA2 ). A relatively large proportion was diagnosed with breast cancer before the end of follow-up ( BRCA1 , 37.9% for those affected with ovarian cancer and 37.9% for those unaffected with ovarian cancer; BRCA2 , 33.0% for those affected ovarian cancer and 37.7% for ovarian cancer unaffected).
| Characteristics | BRCA1 mutation carriers | BRCA2 mutation carriers | ||||||
|---|---|---|---|---|---|---|---|---|
| OvCa+ | OvCa− | OvCa+ | OvCa− | |||||
| N=3989 | N=2445 | |||||||
| n (%) | 346 | (8.7) | 3643 | (91.3) | 106 | (4.3) | 2339 | (95.7) |
| Mean age at start of follow-up (SD), y | 48.4 | (8.9) | 36.4 | (11.5) | 54.0 | (9.1) | 39.0 | (11.8) |
| Mean age at end of follow-up (SD), y | 51.7 | (8.9) | 40.5 | (11.4) | 56.9 | (9.1) | 43.4 | (11.7) |
| Age at end of follow-up, y | ||||||||
| <37 | 14 | (4.1) | 1413 | (38.8) | 4 | (3.8) | 696 | (29.8) |
| 37–46 | 89 | (25.7) | 1202 | (33.0) | 8 | (8.6) | 771 | (33.0) |
| >47 | 243 | (70.2) | 1028 | (28.2) | 94 | (88.7) | 872 | (37.3) |
| Mean person-years (y/person) (SD) | 3.2 | (1.2) | 4.2 | (1.2) | 2.9 | (1.2) | 4.4 | (1.1) |
| Censored for the following: | ||||||||
| Ovarian cancer | 346 | (100.0) | 0 | (0.0) | 106 | (100.0) | 0 | (0.0) |
| DNA test or baseline questionnaire | 0 | (0.0) | 3521 | (96.6) | 0 | (0.0) | 2273 | (97.2) |
| Other cancer | 0 | (0.0) | 41 | (1.1) | 0 | (0.0) | 29 | (1.2) |
| Bilateral RRSO | 0 | (0.0) | 81 | (2.2) | 0 | (0.0) | 37 | (1.6) |
| Year at end of follow-up | ||||||||
| 1990–2000 | 137 | (39.6) | 1122 | (30.8) | 33 | (31.1) | 370 | (15.8) |
| 2001–2005 | 133 | (38.4) | 1354 | (37.2) | 42 | (39.6) | 976 | (41.7) |
| 2006–2012 | 76 | (22.0) | 1167 | (32.0) | 31 | (29.3) | 993 | (42.5) |
| Birth year | ||||||||
| 1920–1944 | 87 | (25.1) | 292 | (8.0) | 55 | (51.9) | 236 | (10.1) |
| 1945–1951 | 109 | (31.5) | 401 | (11.0) | 20 | (18.9) | 308 | (13.2) |
| 1952–1980 | 150 | (43.4) | 2950 | (81.0) | 31 | (29.3) | 1795 | (76.7) |
| Study a | ||||||||
| EMBRACE | 129 | (9.8) | 1183 | (90.2) | 64 | (5.5) | 1095 | (94.5) |
| GENEPSO | 64 | (6.9) | 865 | (93.1) | 15 | (2.7) | 546 | (97.3) |
| Other b | 153 | (8.8) | 1595 | (91.3) | 27 | (3.7) | 698 | (96.3) |
| Breast cancer | ||||||||
| No | 215 | (62.4) | 2255 | (61.9) | 71 | (67.0) | 1458 | (62.3) |
| Yes | 131 | (37.9) | 1388 | (37.9) | 35 | (33.0) | 881 | (37.7) |
| Number of ovarian cancers among first- and second-degree relatives | ||||||||
| 0 | 132 | (51.8) | 1877 | (64.0) | 53 | (68.0) | 1435 | (78.7) |
| 1 | 90 | (35.3) | 752 | (25.6) | 17 | (21.8) | 309 | (17.0) |
| ≥2 | 33 | (12.9) | 305 | (10.4) | 8 | (10.3) | 79 | (4.3) |
| Missing | 91 | 708 | 27 | 516 | ||||
| Cancer type unknown | 0 | 1 | 1 | 0 | ||||
a IBCCS is a collaboration of EMBRACE, GENEPSO, and “other” studies
b Other studies include the following: HEBON, MUV, MODSQUAD, GC-HBOC, Lund-BRCA, OUH, NIO, INHERIT, HCSC, IHCC, CNIO, Stockholm-BRCA, Milan Italy, HSP, DKFZ, and Dusseldorf Germany, Belgium (order is based on number of carriers included in the analyses).
Oral contraceptives were less often used by women who were diagnosed with ovarian cancer (ever use: 58.6% for BRCA1 and 53.5% for BRCA2 ) than by unaffected carriers (ever use:88.9% for BRCA1 and 80.7% for BRCA2 ) ( Table 2 ). The median duration of use was 7 years for both BRCA1 and BRCA2 ovarian cancer cases (interquartile rage (IQR): 3–11 years for BRCA1 and 4–12 years for BRCA2 ) and 9 years (IQR, 5–13 years) and 8 years (IQR, 5–13 years) for BRCA1 and BRCA2 mutation carriers who were unaffected with ovarian cancer, respectively.
| Variable | BRCA1 mutation carriers | BRCA2 mutation carriers | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OvCa+, n (%) c | OvCa−, n (%) c | Weighted, a , b HR (95% CI) d | OvCa+, n (%) c | OvCa−, n (%) c | Weighted, a , b HR (95% CI) e | |||||
| Oral contraceptive use | ||||||||||
| Never (<6 mo) | 133 | (41.4) | 659 | (19.1) | 1.00 | 46 | (46.5) | 426 | (19.3) | 1.00 |
| Ever | 188 | (58.6) | 2788 | (88.9) | 0.51 (0.36–0.71) | 53 | (53.5) | 1782 | (80.7) | 0.65 (0.35–1.19) |
| Ever, starting age unknown | 11 | 125 | 4 | 87 | ||||||
| Missing | 14 | 71 | 3 | 44 | ||||||
| Calendar year at start | ||||||||||
| Never (<6 mo) | 133 | (41.4) | 659 | (19.1) | 1.00 | 46 | (46.5) | 426 | (19.3) | 1.00 |
| ≤1975 | 114 | (35.5) | 684 | (19.8) | 0.45 (0.33–0.62) | 39 | (39.4) | 492 | (22.3) | 0.73 (0.38–1.39) |
| >1975 | 74 | (23.1) | 2104 | (61.0) | 0.56 (0.35–0.88) | 14 | (14.1) | 1290 | (58.4) | 0.49 (0.23–1.05) |
| Ever, starting year unknown | 11 | 125 | 4 | 87 | ||||||
| Missing | 14 | 71 | 3 | 44 | ||||||
| Total duration of use | ||||||||||
| Never (<6 mo) | 133 | (43.0) | 659 | (20.2) | 1.00 | 46 | (47.4) | 426 | (20.4) | 1.00 |
| <5 y | 67 | (21.7) | 616 | (18.9) | 0.79 (0.53–1.19) | 17 | (17.5) | 401 | (19.2) | 0.87 (0.42–1.80) |
| 5–9 y | 53 | (17.2) | 867 | (26.6) | 0.54 (0.35–0.85) | 13 | (13.4) | 557 | (26.6) | 0.51 (0.23–1.12) |
| ≥10 y | 56 | (18.2) | 1121 | (34.4) | 0.32 (0.21–0.50) | 21 | (21.7) | 707 | (33.8) | 0.60 (0.28–1.27) |
| Ever, no period specific data | 23 | 309 | 6 | 204 | ||||||
| Missing | 14 | 71 | 3 | 44 | ||||||
| Trend f | P =2.0E-04 | P =.449 | ||||||||
| Time since last use | ||||||||||
| Never (<6 mo) | 133 | (43.0) | 659 | (20.2) | 1.00 | 46 | (47.4) | 426 | (20.4) | 1.00 |
| <10 y | 29 | (9.4) | 1478 | (45.3) | 0.40 (0.22–0.71) | 7 | (7.2) | 825 | (39.5) | 0.36 (0.14–0.92) |
| 10–19 y | 60 | (19.4) | 648 | (19.9) | 0.54 (0.36–0.82) | 12 | (12.4) | 415 | (19.9) | 0.58 (0.24–1.42) |
| ≥20 y | 87 | (28.2) | 478 | (14.7) | 0.61 (0.43–0.87) | 32 | (33.0) | 425 | (20.3) | 0.78 (0.40–1.52) |
| Ever, no period specific data | 23 | 309 | 6 | 204 | ||||||
| Missing | 14 | 71 | 3 | 44 | ||||||
| Trend f | P =.025 | P =.258 | ||||||||
| Starting age | ||||||||||
| Never (<6 mo) | 133 | (41.4) | 659 | (19.1) | 1.00 | 46 | (46.5) | 426 | (19.3) | 1.00 |
| ≤19 y | 62 | (19.3) | 1753 | (50.9) | 0.43 (0.28–0.65) | 20 | (20.2) | 1030 | (46.7) | 1.02 (0.43–2.42) |
| 20–23 y | 58 | (18.1) | 615 | (17.8) | 0.51 (0.33–0.78) | 12 | (12.1) | 461 | (20.9) | 0.49 (0.20–1.23) |
| >23 y | 68 | (21.2) | 420 | (12.2) | 0.63 (0.41–0.97) | 21 | (21.2) | 291 | (13.2) | 0.60 (0.29–1.21) |
| Ever, starting age unknown | 11 | 125 | 4 | 87 | ||||||
| Missing | 14 | 71 | 3 | 44 | ||||||
| Trend f | P =.154 | P =.114 | ||||||||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
