Combined oral contraceptive utilization and uterine fibroid incidence: A prospective study in a cohort of African-American women.

More than 1 billion U.S. dollars are spent each year to treat uterine leiomyoma (fibroids) (non-malignant smooth muscle uterine tumors) [ 1 ]. Fibroid symptoms include bulk symptoms (pelvic pressure, urinary symptoms), pain, and heavy menstrual bleeding which can diminish quality of life and impact work productivity [ 2 ]. Fibroids can contribute to infertility and may be associated with preterm delivery [ 2 , 3 ]. Fibroids are the leading indication for hysterectomy [ 4 ]. Estimates based on ultrasound screening, irrespective of prior diagnosis or symptoms, suggest that more than 70% of women develop fibroids by menopause [ 5 ]. Though the prevalence of symptomatic fibroids is unknown, most women who are diagnosed have had symptoms for years [ 4 ]. Estimated annual costs of fibroids vary by assumptions about the prevalence of symptomatic fibroids, ranging from $5.89 to $34 billion [ 6 ]. The estimates include one aspect of quality of life, loss of productivity which was estimated to range from $5 –$17.2 billion. There are important racial disparities in the health burden associated with fibroids. In the U.S., African Americans develop fibroids an estimated 10 years earlier than White women, have a higher estimated incidence by the age of menopause, have larger fibroids at the time of diagnosis, and are 2–3 times more likely to require major medical or surgical procedures to treat fibroids [ 5 , 7 – 9 ]. While there has been advancement in surgical and medical treatment of symptomatic fibroids, considerably less research has focused on fibroid prevention. This is surprising given the high prevalence and public health burden of the disease. It is currently thought that estrogen and progesterone have complex, interrelated roles in fibroid tumor development and growth [ 10 , 11 ]. Progesterone is an important driver [ 12 ] of fibroid growth, and estrogen increases availability of progesterone receptors [ 11 ]. Injectable progestin (synthetic progesterone) contraceptives may offer lasting protection against uterine fibroids [ 13 ], and other forms of hormonal contraceptives are commonly used as a first-line treatment for fibroid symptoms [ 14 , 15 ]. Oral contraceptives (OCs) are generally composed of both estrogen and progestin [ 16 ]. Exposure to OCs is widespread in the U.S. [ 17 – 19 ]. Among sexually experienced, African-American women aged 15–44, 80% have used OCs in their lifetime [ 17 ]. Despite the availability of longer acting methods in the U.S., OCs remain the most commonly used form of hormonal contraception [ 20 ]. To date, published literature on the association between OC use and uterine fibroid development [ 21 – 32 ] have yielded mixed findings. Existing studies identified prevalent fibroid cases with cross-sectional ultrasound data, or relied on self-reported diagnoses of fibroid status which may be influenced by symptom severity and access to health care [ 33 ]. In addition, self-reported new diagnoses reflect detection of fibroids that likely developed years earlier. Thus, the current literature is largely unable to establish temporality in the reported relationships between OCs and fibroids. Further, existing studies have only compared any history of oral contraceptive use to never use. Comparing the incidence of fibroids associated with different durations of use and other utilization characteristics would be valuable since most women use oral contraceptives in their lifetime [ 17 – 19 ]. Finally, few studies [ 21 , 25 – 27 ] included ≥10% African-American participants, the population with the largest fibroid burden both in terms of occurrence and severity [ 34 – 37 ]. We sought to address these limitations by examining the longstanding question of how OC use influences uterine fibroid development, using data from the Study of Environment, Lifestyle, and Fibroids (SELF) [ 8 ]. SELF is the first large prospective, ultrasound-based study of risk factors for uterine fibroid incidence, and exclusively enrolled young, African-American participants [ 8 ]. Specifically, we examined the associations between different characteristics of oral contraceptive use and incident fibroids at ~40 months of follow-up among participants who were fibroid-free at enrollment.

Among the 1,308 SELF participants who were fibroid-free at enrollment, median age at enrollment was 28 years [IQR: 25, 31] ( Table 1 ). Very few baseline characteristics differed between ever and never users with the exception that ever users of COCs more frequently reported higher annual household income, higher educational attainment, never smoking, heavy menstrual bleeding, and childbirth, and less frequently reported Long/Recent (>24 months, within 8 years) use of Depo-Provera ( Table 1 ). *Education was missing for n = 1 participants in the never used COCs group. Abbreviations: SELF, Study of Environment, Lifestyle, and Fibroids; COC, combined oral contraceptives; IQR, interquartile range; hrs, hours; wk, week; PhD, Doctor of Philosophy; kg/m 2 , kilograms per square meter; H-IUD, hormonal intrauterine device. Among the 913 (70%) SELF participants who had ever used COCs, 320 (35%) reported initiating COCs prior to age 17 years, and most (73%) used COCs for less than 5 years in total ( Table 2 ). Relatively few (24%) were currently using COCs at study enrollment. Nearly half (48%) had last used COCs 5 or more years ago ( Table 2 ). Unweighted and weighted risks for developing fibroids during the follow-up period are provided in the “Incident Fibroid Cases” columns of Table 2 . Approximately 17% (n = 221) developed fibroids by the 40-month follow-up. *Weighted (IPW/SMR) or Adjusted (MVR) for age in years (continuous), age at menarche (<11 years), Depo-Provera duration of and years since last use (never use, short/past, long/past, short/recent, long/recent), total implant and H-IUD duration of use (≥24 months), years since last birth (<5 years, 5–9.99 years, ≥10 years and no birth), parity (nulliparous or never pregnant, 1 birth, 2 births, ≥3 births), BMI (≥30, remaining values according to tertile), education (Bachelor’s degree or higher). Weighted risk ratios, risk differences, and incident fibroid case counts are shown in the columns corresponding to this footnote. A total of n = 198 women were censored, including n = 136 COC users. Inverse probability of censoring weights were applied to upweight individuals most likely to have been censored who remained in the study. The model for probability of censoring included the exposure of interest (e.g., duration of use), all covariates used in the IPW/SMR models, with the addition of annual household income, baseline employment status, smoking history, and history of heavy gushing type bleeding. † The weighted estimate for Ever-use was weighted according to the covariate distribution of COC users; i.e., standardized morbidity ratio [SMR] weighting was employed for this exposure. ‡ Excludes n = 27 participants who used both pill and mini-pill, for whom years since last COC use could not be distinguished. § Duration of use was characterized as short (< 2 years) or long (≥ 2 years). Years since last use was characterized as recent (< 5 years) or past (≥ 5 years). Abbreviations: SELF, Study of Environment, Lifestyle, and Fibroids; CI, confidence interval; MVR, multivariable logistic regression; IPW, inverse probability weighting; SMR, standardized morbidity ratio; H-IUD, hormonal intrauterine device; BMI, body mass index. At the 40-months’ follow-up, we observed a possible inverse association between ever use of COCs and fibroid cumulative incidence (wRR 0.78, 95% CI: 0.60, 1.00; wRD -0.05, 95% CI: -0.11, 0; Table 2 , Fig 1 , S3 and S4 Appendices in S1 File ). Among ever users of COCs, age at first use ≥17 years was associated with an elevated risk of fibroid incidence (wRR: 1.25; 95% CI: 0.89, 1.76; wRD: 0.04, 95% CI: -0.02, 0.10) compared to those initiating COCs prior to age 17, although the estimate was imprecise. Comparing each level of duration of COC use to <1 year of use, weighted risk ratios varied from 1.10 to 1.49, with elevated risk in every category other than the shortest-term use ( Table 2 , Fig 1 , S3 and S4 Appendices in S1 File ). Weighted risk ratios comparing each level of years since last use to current use ranged from 0.87 to 1.23, declining with increased years since last use ( Table 2 , Fig 1 , S3 and S4 Appendices in S1 File ). Compared to short-term use in the past (<2 years of COC use, ≥5 years age), other cross-classified patterns of use demonstrated no clear trend. *Weighted (IPW/SMR) for age in years (continuous), age at menarche (<11 years), Depo-Provera duration of and years since last use (never use, short/past, long/past, short/recent, long/recent), total implant and H-IUD duration of use (≥24 months), years since last birth (<5 years, 5–9.9 years, ≥10 years and no birth), parity (nulliparous or never pregnant, 1 birth, 2 births, ≥3 births), BMI (≥30, remaining values according to tertile), education (Bachelor’s degree or higher). A total of n = 198 women were censored, including n = 136 COC users. Inverse probability of censoring weights were applied to upweight individuals most likely to have been censored who remained in the study. The model for probability of censoring included the exposure of interest (e.g., duration of use), all covariates used in the IPW/SMR models, with the addition of annual household income, baseline employment status, smoking history, and history of heavy gushing type bleeding. † The weighted estimate for Ever-use was weighted according to the covariate distribution of COC users; i.e., standardized morbidity ratio [SMR] weighting was employed for this exposure. ‡ Excludes n = 27 participants who used both pill and mini-pill, for whom years since last COC use could not be distinguished. § Duration of use was characterized as short (< 2 years) or long (≥ 2 years). Years since last use was characterized as recent (< 5 years) or past (≥ 5 years). Abbreviations: IPW, inverse probability weighting; SMR, standardized morbidity ratio; SELF, Study of Environment, Lifestyle, and Fibroids; CI, confidence interval; MVR, multivariable logistic regression; H-IUD, hormonal intrauterine device; BMI, body mass index.

Though ever use of COCs showed an inverse association with uterine fibroid development, estimates for the COC-use factors that might therefore have been expected to be associated with reduced fibroid risk were less precise and directionally inconsistent (e.g., wRR >1 for longer durations of use and more recent use; Table 2 , Fig 1 , S3 and S4 Appendices in S1 File ). Contrary to prior reports [ 21 , 34 ] which relied on never users as the referent group and showed increased self-reported fibroid incidence among young (age <17 years) COC initiators, we observed a higher incidence of fibroids in COC users who initiated COC use on or after age 17 years, although the estimate was imprecise. Estimates for duration of use or years since last COC use were limited in precision due to study size. Further studies should be conducted in larger data sets that would lead to estimates that are more precise and that build upon the methodological improvements represented by this study. Additional tissue-based studies using the new formulations of COCs are also needed to clarify the biological mechanisms at play, particularly among early COC initiators.

This study was completed as part of a PhD dissertation at the University of North Carolina at Chapel Hill [ 38 ], which is publicly available at the Carolina Digital Repository ( https://cdr.lib.unc.edu/ ). The Study of Environment, Lifestyle & Fibroids (SELF) is a prospective cohort study of 1,693 young (23–35 years), African-American women living in the Detroit, Michigan area. SELF was designed to investigate risk factors for uterine fibroid incidence and growth [ 8 , 39 – 41 ]. Recruitment and baseline data collection were completed in 2010–2012 [ 8 ]. Participants were recruited from the Detroit area via local media commercials and advertisements, brochures at healthcare clinics, information booths at community events, and via Henry Ford Health (HFH) [ 8 ]. The primary eligibility requirements were age (23–35 years), self-identified African American/Black, and having no prior clinical diagnosis of uterine fibroids. Additional eligibility criteria included U.S. residence, a willingness to return for clinic visits over 5 years and provide contact information. Exclusion criteria included a prior diagnosis of fibroids, receipt of radiation or chemotherapy for treatment of cancer, and prior diagnosis of autoimmune disorders requiring medication [ 8 ]. Participants had ultrasounds and questionnaire data collected at enrollment and at each follow-up, scheduled at approximately 20-month intervals. This analysis includes data collected through the 40-month follow-up. Study retention rate was high at >85% for the 40-month follow-up. All participants received a study orientation in-person or over the phone describing all aspects of the study including a detailed walk-through of the consent form with an opportunity to ask questions. Following this orientation all participants provided written informed consent witnessed by trained study staff. SELF was approved by the institutional review boards of the National Institute of Environmental Health Sciences and Henry Ford Health. All exposure and covariate data were self-reported using telephone or computer-assisted questionnaires, with the exception of BMI, which was calculated using height and weight as measured at the enrollment clinic visit. History of hormonal contraceptive (HC) use was collected via telephone interview as part of an enrollment questionnaire. Participants were asked if they had ever used each of the following types of HC: “birth control pills” (OCs), “mini-pill” (progestin-only OCs), hormonal implant, hormonal patch, vaginal ring, “hormone shots like Depo-Provera ® ”, and hormonal intrauterine devices (H-IUD). Brief descriptions and examples of common brand names were provided for hormonal implants and shots. For each HC type (and separately for each H-IUD), participants were asked about their age at first use, whether or not they were currently using and how old they were when they stopped using. For age at first use, participants were asked, “How old were you when you started using birth control pills, whether or not it was to prevent pregnancy?” Participants who had used OCs were asked whether or not they had ever stopped using for a month or longer, between initial start and last stop (or study enrollment if currently using), and if so, how much of that time was spent on OCs (“very little of that time,” “less than half of that time,” “about half,” “more than half,” “most of that time”). Participants who used the “mini-pill,” implant, patch, ring, or shot were asked to state the total number of months and years that they had used each method prior to study enrollment. Fibroid identification was performed using transvaginal ultrasonography–the current standard of care for uterine fibroid assessment [ 42 , 43 ]. Transvaginal ultrasonography has 99% sensitivity and 91% specificity for detecting uterine fibroids [ 44 ]. If any fibroids with at least one diameter of 0.5 cm or greater were detected, the largest six fibroids were measured in three separate passes through the uterus. Ultrasound examinations were conducted by a consistent group of registered diagnostic sonographers with at least 3 years of experience in gynecologic sonography. The study sonographers were given initial and refresher trainings including care in distinguishing fibroids from other pathologic changes in the uterus including adenomyosis and polyps, protocol for conducting the exam, and recording the data. Video and still images were archived and an 8% sample for each sonographer, oversampled for fibroid cases, was reviewed every month by the lead sonographer [ 8 ]. Among those fibroid-free at the enrollment ultrasound, incident fibroid by the 40-month follow-up was “yes” if fibroids were seen at either the 20- or 40-month follow-up. It was “no” if no fibroids were seen at the 20- or 40-month follow-up. Participants with no follow-up ultrasounds as well as those with no fibroids at 20-months but no 40-month ultrasound were excluded during analysis because their cumulative incidence at 40 months could not be assessed. Of those fibroid-free at enrollment (n = 1,308), 198 were missing the outcome variable. Reasons for missing data for the outcome included missing the study visit (n = 190), non-fibroid related hysterectomy (n = 7), and poor ultrasound quality (n = 1). Excluded individuals were included in our propensity score calculations, and inverse probability of censoring weights were applied in order to upweight similar individuals with complete data in the calculations of our estimates (see S1 Appendix in S1 File ). Ever users of combined oral contraceptives (COCs, containing both estrogen and progestin) were participants who answered “yes” to any use of oral contraceptives, excluding those who used only the mini-pill [ 45 ]. Participants who reported using “the pill” were considered to have used COCs if they answered “No” or “I don’t know” to “Have you ever used a progesterone-only birth control pill, or “mini-pill”, such as Micronor, Nora-BE, or Ovrette?” Age at first COC use was dichotomized into <17 and ≥17 based on findings in previous literature [ 21 , 29 ]. Duration of use estimates were calculated in three steps: (1) Subtract age at first OC use from age at last OC use, or age at enrollment for current users, (2) Subtract years and months on the mini-pill, (3) Multiply by the self-reported proportion of time spent using OCs. Weights were applied as follows: 10% for “very little of that time,” 25% for “less than half of that time,” 50% for “about half,” 75% for “more than half,” 90% for “most of that time,” and 100% for those who had no interruption in their use. When age at first use and age at last use were identical, a duration of six months was assigned. For former COC users, years since last COC use was calculated by subtracting self-reported age at last pill use from age at enrollment. Current users were assigned a “years since last use” value of 0. For n = 64 participants who used both combined and progestin-only OCs, years since last COC use specifically could be determined for 31 subjects. Years since last use of COCs could not be determined for the remaining 33 participants due to limitations in the enrollment questionnaire. Joint categories of duration of use and years since last use were created. Duration of use was characterized as short (<2 years) or long (≥2 years) and years since last use was characterized as recent (<5 years) or past (≥5 years), based on available data, creating four joint categories. The covariate set for all weight calculations included the following variables categorized as follows: age in years (continuous), age at menarche (≤10 years, yes/no), Depo-Provera use (4 categories of duration and years since last use), implant and H-IUD use (use ≥ 24 months, yes/no), years since last birth (0–4, 5–9, 10+ years or no birth), parity (nulliparous or never pregnant, 1 birth, 2 births, ≥ 3 births), BMI (in four categories: >30, and remaining values according to tertile), and education (Bachelor’s degree or higher, yes/no). In addition to these covariates, censoring weights also included the exposure of interest, annual household income (<$20,000, $20-$50,000, ≥$50,000), baseline employment status (not employed, employed <30 hrs/wk, employed ≥30 hrs/wk), smoking history (never, former, current <10/day, current ≥10/day), and history of heavy “gushing” type menstrual bleeding (yes/no). All data management and analyses were performed in SAS 9.4 (SAS Institute, Cary, NC, USA). We used inverse probability weights methods to account for differences in fibroid risk factors and possible selection bias between exposure groups, and we report this propensity score-based methods in accordance with the recommendations provided by Ali et al. (2014) [ 46 ] in the Journal of Clinical Epidemiology in Supplemental Appendices 1–3. In brief, inverse probability weights were constructed for all COC exposures, and for censoring. Standardized morbidity ratio weights were constructed for ever-COC use. Similarly, we used weights as described in S1 Appendix in S1 File to estimate associations among all COC users. Absolute standardized differences were used to assess covariate balance with the conventional a priori threshold of 0.1. Weighted log-binomial regression models were used to estimate risk ratios and risk differences for uterine fibroids. Confidence intervals for weighted models were generated using robust variance (“sandwich”) estimator by use of the SAS REPEAT statement. Unweighted multivariable log-binomial regression models, with the same set of covariates used in the weighted models, were run for comparison. As recommended by the American Statistical Association leadership, associations were interpreted based on the strength of the estimate, width of the confidence interval (precision) and consistency across models, not on statistical significance at p≤0.05 [ 47 ]. We first compared ever to never users of COCs, and then compared different levels of COC use (e.g., age at first use, duration of use, etc.) among the COC users. This is in contrast to existing studies, which compared different levels of COC use to never-use–a comparison that is unlikely to be useful given that distinctions by duration of use and time since last use are relevant primarily to ever COC users.

(DOCX) (DOCX)