Hormonal Contraception Use and Risk of Lower Urinary Tract Symptoms and Conditions: Findings From the Boston Area Community Health Survey.

Of the 1,228 female BACH participants under 45 years of age, we excluded 95 for clinical reasons (hysterectomy [n=57], bilateral oophorectomy [n=2], natural or treatment-induced menopause [n=23], current or recent pregnancy [n=9], or permanent catheterization [n=4]), two for incomplete information on HC use or LUTS, and 250 for incomplete information on vaginal parity at baseline. This left 881 participants in the analysis of contraception and LUTS prevalence. Approximately 30% of the weighted sample was <35 years of age, 45.0% was 35–39 years of age, and 25.1% was 40–44 years of age ( Table 1 ). Slightly more than half of the weighted population was vaginally parous and 7.2% had ≥4 vaginal deliveries. Slightly less than half had a BMI <25 kg/m 2 , approximately one third had a BMI ≥30 kg/m 2 , and the majority had a waist circumference between 65–89 cm. Approximately one quarter of the weighted population smoked cigarettes currently and a small percentage (1.9%) reported a diagnosis of diabetes. Current HC users were the youngest, followed by past users and then never users. Current users were more likely to self-identify as White, whereas never users were more likely to self-identify as Black. In addition, current users were most likely to be vaginally parous, to have a BMI <25 kg/m 2 , and to smoke in the past. They were also least likely to have diabetes and to smoke currently, consistent with HC use guidelines. Characteristics of ever systemic users were very similar to ever users, as most participants who had used HCs had used systemic forms of HC. No statistically significant associations were observed for ever, current, past, or ever systemic HC use with the prevalence of UI, other bladder storage symptoms, voiding/emptying symptoms, and lower urinary tract pain ( Table 2 ). Similar null results were observed for HC history and type with the prevalence of individual storage and voiding/emptying LUTS, with the exception of suggestive positive associations for ever, past, and systemic use with urinating again after less than two hours and positive associations for ever and past use with post-micturition dribbling ( Appendix Table 2 ). In contrast to findings for most bladder storage and voiding/emptying LUTS, and pain, significant positive associations were observed for rUTIs. Specifically, participants currently using HCs had a 16-fold higher prevalence of rUTIs than never users (adjusted PR [aPR]=16.3, 95% CI: 1.8–152.1) and those who had ever used HCs had a 6-fold higher prevalence (aPR=6.4, 95% CI: 1.0–41.3). Suggestive positive associations were also observed for past (aPR=4.8, 95% CI: 0.9–26.9) and ever systemic use (aPR=4.1, 95% CI: 0.7–24.8). Similar associations were observed in sensitivity analyses limited to participants who had ever or were currently using contraception of any form. Of the 881 participants included in prevalent analyses, 608–867 (depending on the LUTS group) were retained in incident analyses. Similar to findings for prevalent LUTS, no associations were observed for HC history and type with the incidence of UI, other bladder storage LUTS, voiding/emptying LUTS, and lower urinary tract pain, either when evaluated as groups of LUTS ( Table 3 ) or individually ( Appendix Table 3 ). Exceptions were suggestive or significant inverse associations for: 1) past HC use and incident stress incontinence; 2) current use and incident nocturia; 2) ever and ever systemic use with urinating again after less than two hours; and 3) current use with weak urinary stream. Similar to prevalent findings, suggestive, but slightly attenuated, positive associations were observed for ever (aRR=4.4, 95% CI: 0.9–21.2), past (aRR=4.0, 95% CI: 0.9–18.3), current (RR=5.9, 95% CI: 0.7–53.5), and ever systemic HC use (RR=3.7, 95% CI: 0.9–15.9) with rUTI risk, after adjusting for baseline sexual function. Similar inferences were obtained after adjusting for other baseline covariates ( Table 3 ) and for follow-up covariates (data not shown), as well as in sensitivity analyses restricted to participants who had ever used contraception ( Table 3 ). Results attenuated for current HC use in analyses restricted to participants currently using any form of contraception but the sample size for this analysis was small (n=400).

The BACH Survey is a population-based, longitudinal study of Boston residents aged 30–79 at baseline in 2002–5 (BACH I). Participants were sampled randomly using a two-stage cluster design, including stratification by sex, race/ethnicity (equally distributed across Black, Hispanic, and White), and age (30–39, 40–49, 50–59, and 60–79). At baseline, 5,506 participants completed the first study visit, 3,205 (52.4%) of whom identified as female. The first visit included an in-home interview; completion of a self-administered questionnaire; and measurement of height, weight, and waist circumference. Participants repeated these procedures five years later during the follow-up visit from 2008–2010 (BACH II). The BACH Survey was approved by the New England Research Institutes Institutional Review Board and all participants provided written informed consent. 19 We limited the current analysis to subjectively pre-menopausal female participants who completed the baseline interview. Eligible participants were required to: 1) be <45 years of age (to reduce the likelihood of including post-menopausal women); 2) not report a hysterectomy or bilateral oophorectomy; and 3) not report that their menstrual cycles had stopped because of menopause or a therapy (e.g., chemotherapy) other than continuous HC use. We also excluded participants who were currently or recently pregnant or used permanent catheterization at baseline to avoid including participants with transient or chronic LUTS because of well-known factors related to LUTS and not HCs. Lastly, we excluded participants with missing baseline information on current and prior HC use, LUTS, or vaginal parity, a strong risk factor for LUTS in younger women. 1 For analyses of HC use and incident LUTS, we further limited the analyses to participants who completed the follow-up interview, provided information on LUTS at follow-up, and did not have any of the above-described exclusionary conditions at follow-up. We also required participants not to report the particular LUTS or condition of interest at baseline in LUTS/rUTI-specific incident analyses (e.g., no reported UI at baseline in the UI incident analyses). We did not limit incident analyses to participants without any reported LUTS at baseline because of small sample size (n=498). At baseline and follow-up, female participants were asked about their current or past use of contraceptive pills (OCPs), implants, vaginal rings, injections, intrauterine devices (IUDs), tubal sterilization, or barrier methods. We used this information to create the following variables for analysis: never, past, and current HC use (i.e., use of OCPs, implants, vaginal rings, injections, and IUDs) and ever systemic HC use (OCPs, implants, vaginal rings, and injections). Although the survey did not distinguish between combined HC and progestin-only pills or between hormonal and non-hormonal IUDs, we grouped all OCPs and IUDs together as HCs because combined HC pills and hormonal IUDs are much more common in the U.S. than progestin-only pills and non-hormonal IUDs. 21 , 22 We also created a separate variable for only ever local HC (i.e., IUD) use, which we hypothesized would have a null effect on risk of LUTS or rUTIs, but did not use it in the analysis because of low numbers of only ever local users (n=27). At baseline and follow-up, participants completed evaluations of LUTS and rUTIs, using the American Urological Association Symptom Index, 23 Sandvik Incontinence Severity Scale, 24 and Interstitial Cystitis Symptom Index, 25 as well as items written specifically for BACH. Altogether, this yielded 24 items, which we collapsed into 15 unique symptoms or conditions ( Appendix Table 1 ) 26 and five LUTS/condition groups, including: 1) UI (UI in the past year); 2) non-UI bladder storage symptoms (difficulty postponing urination, nocturia [≥2 voids/night], perceived frequency, and urinating again after less than two hours in the past month); 3) bladder voiding and emptying symptoms (dribbling/wet clothes after urination, straining/difficulty to begin voiding, weak stream, intermittency, and incomplete emptying in the past month); 4) pain, burning, or discomfort in the pubic/bladder area in the past month; and 5) rUTIs . We created unique LUTS and groups of LUTS by taking the maximum response of items in each group. Participants were considered to have a particular symptom if they reported it “fairly often” or more (as opposed to never, “rarely,” or “a few times”), and rUTIs if they reported ≥3 UTIs in the past year. At baseline and follow-up, the following additional variables were assessed by self-report: age, race/ethnicity, vaginal parity, cigarette smoking history, and diabetes status. Weight and height were measured and used to calculate body mass index (BMI). Waist circumference was also measured. We included these variables in the analysis because of their known relation with LUTS 1 and possible relation with HC use. For rUTIs, we included an additional variable related to sexual activity in the analyses because rUTI risk is strongly associated with sexual frequency. 2 Specifically, we used three items from the Female Sexual Function Index 27 (ever sexual activity, and level and frequency of sexual desire and interest in the past four weeks) to create a variable with the following categories: 1) no sexual activity in the past four weeks, and some sexual activity accompanied by a: 2) very low/no or almost never/never sexual desire, 3) low/a few times desire, 4) moderate/sometimes desire, 5) high/most times desire, and 6) very high/almost always desire. We used this variable as a marker of sexual frequency because data on frequency of sexual intercourse were not collected in BACH. We also performed sensitivity analyses limited to participants who reported ever using contraception at baseline (i.e., HC, tubal sterilization, or barrier methods) to control for ever engaging in sexual activity. Analyses of current hormonal contraception use were further limited to participants who reported any current method of contraception. To account for the BACH sampling design, we weighted all observations inversely proportional to their probability of selection, with further post-stratification to the Boston population using the 2000 U.S. Census. Confounding was explored by comparing proportions of potential confounders (see covariates section) by baseline contraception history. Associations between HC use and LUTS were investigated by calculating unadjusted and multivariable-adjusted prevalence ratios (PRs), risk ratios (RRs), and their associated 95% confidence intervals (CIs), using Poisson regression with robust variance estimation. Multivariable-adjusted models included terms for age, race/ethnicity, vaginal parity, BMI, waist circumference, cigarette smoking, diabetes, and sexual activity (rUTI analyses only). Baseline covariate values were included in prevalent analyses, whereas two sets of models were run for incident analyses; the first included baseline values and the second included follow-up values.

In our large prospective analysis of subjectively pre-menopausal women, no associations were observed for HC use with the prevalence or risk of LUTS, including UI, other bladder storage symptoms, voiding/emptying symptoms, and lower urinary tract pain. In contrast, significant positive associations were observed for rUTIs. These associations persisted after adjustment for known LUTS risk factors and a marker of sexual activity, as well as in sensitivity analyses restricted to participants who reported ever using any form of contraception. However, we were not able to adjust for frequency of sexual activity, a factor likely associated with HC use and known to be associated with rUTI risk, 2 to investigate whether this association was explained by residual confounding or possibly a hormonal alteration attributable to HC use. We originally investigated the association between HC use and LUTS because of the known influence of sex hormones on urogenital tract development and function across the life course; 3 the consistently observed, but opposing, associations for systemic and local menopausal hormone use with UI in post-menopausal women; 5 , 6 and the known influence of OCPs on SHBG and sex steroid hormone levels. 7 , 8 Based on these observations, we hypothesized that systemic HC use would increase the risks of UI and other bladder storage symptoms in pre-menopausal women, whereas local HC use would have no effect. Contrary to our hypothesis, we observed no association for systemic HC use with risks of UI or other bladder storage symptoms, although we observed a few, likely chance, associations for individual storage LUTS. These generally null findings are consistent with those from several cross-sectional studies, 15 – 17 but conflict with those from several other studies that observed either positive and inverse findings for reasons that are not clear. 10 – 14 Our findings for bladder voiding/emptying symptoms were also generally null in contrast to those from the only study, to our knowledge, to investigate this possible association. That study observed a positive association between OCP use and risk of a sensation of incomplete emptying in adolescent women. 11 Although we did not have strong biologic hypotheses for studying HC use and lower urinary tract pain, we included lower urinary tract pain in our analysis because of previously observed positive, cross-sectional associations between HC use and interstitial cystitis/bladder pain syndrome, 28 which we hypothesized might be explained by reverse causation (i.e., more frequent HC use by individuals with interstitial cystitis/bladder pain syndrome to reduce their overall burden of pelvic pain or to address co-morbid endometriosis). Contrary to findings from previous cross-sectional studies, 28 as well as one longitudinal study that did not assess pain at baseline, 11 we observed no association between HC use and lower urinary pain in both prevalent analyses, which might be more susceptible to reverse causation; and incident analyses, which should be less susceptible to reverse causation. Thus, so far, the literature does not support a role for HC use in contributing to lower urinary tract pain. Although our initial motivation for this analysis was to investigate associations between HC use and risks of UI and other LUTS, we took advantage of the extensive data collected on lower urinary tract symptoms and conditions in BACH to explore additional less well-studied LUTS and conditions, such as rUTIs. This possible association is biologically plausible because, similar to UI and other storage LUTS, rUTI risk is influenced by genitourinary hormone levels and increases with peri-menopause and menopause. 2 Our observed positive association between HC use and risk of rUTIs is consistent with this hypothesis, as well as much of the limited literature on this topic. 11 , 29 However, a lingering concern for this literature is residual confounding by frequency of sexual activity. This behavior is likely correlated with HC use and known to be related to UTI risk. 2 In a previous study that was able to adjust for correlates of sexual frequency (i.e., condom use and number of sexual partners), the magnitude of association between HC use and UTIs decreased after adjustment, 11 suggesting that the original findings were confounded by sexual activity and might attenuate further with adjustment for additional measures of sexual activity. Although our positive associations for rUTIs were only minimally attenuated after adjustment for recent intensity and frequency of sexual desire as a marker of sexual frequency and after restriction to participants who reported ever using any form of contraception, these measures may not have been sufficiently well-correlated with sexual frequency to address potential confounding by sexual activity. Therefore, future studies should investigate these findings further, incorporating data on frequency of sexual activity. Strengths of our study include its: 1) longitudinal design, allowing us to investigate associations between HC use and risk (as opposed to prevalence) of LUTS, thereby reducing concerns for reverse causation; 2) detailed collection of data related to HCs, and lower urinary tract symptoms and conditions, supporting investigation of both history and type of HC use in relation to UI and a wide range of lesser-studied LUTS; 3) detailed collection of covariate data, allowing us to control for a number of variables associated with both HC use and LUTS; and 4) use of probability sampling, high participant response and retention rates, and stratified sampling by race and ethnicity, allowing us to generalize our findings to the U.S. population and multiple racial/ethnic groups. Despite its prospective design, limitations of our study include collection of data at only two times, precluding an incident user design 30 (i.e., studying participants from the time they initiate HC use). Therefore, if HC users who experienced LUTS or other correlated side effects discontinued use before baseline or follow-up and if the impact of HCs is transient, then we may have potentially missed early, positive and transient associations. Additional limitations of our analysis include its: 1) susceptibility to residual confounding, as described earlier for sexual frequency and rUTIs; 2) low prevalence of IUD use, precluding a well-powered investigation of this form of contraception; and 3) exposure and participant misclassification. Specifically, we were also not able to distinguish between combined HC and progestin-only pills and between hormonal and non-hormonal IUD use. However, use of progestin-only pills and non-hormonal IUDs is relatively uncommon in the U.S. 21 , 22 In addition, although we limited the analysis to women under 45 years of age who reported recent menstrual cycles to avoid including post-menopausal women, it is possible that some women at the upper end of our age range may have been peri-menopausal. Finally, our study population was limited to women who self-identified as Black, Hispanic, or White. However, we have no reason to believe that our findings would not generalize to women of other races or ethnicities, such as Asian women. In summary, findings from our large prospective analysis of Boston area residents do not support associations between HC use and risk of most LUTS, including UI, other bladder storage LUTS, voiding LUTS, and lower urinary tract pain. Our positive findings for rUTIs should be investigated further to determine whether they are explained by residual confounding by sexual frequency or by the possible influence of HCs on the hormonal genitourinary tract milieu. Ultimately, a causal association between HC use and rUTIs could have implications for HC choice (e.g., systemic versus local), and for counseling about rUTI risk and additional prevention strategies.

Lower urinary tract symptoms (LUTS), including dysfunction with bladder storage and voiding/emptying, and recurrent urinary tract infections (rUTIs) are common and associated with significant direct and indirect economic costs in women, as well as negative impact on quality of life. The etiology of these common and bothersome symptoms is incompletely understood and likely multifaceted. 1 , 2 One potential, but understudied, risk factor is use of exogenous sex steroid hormones, including hormonal contraception (HC). Sex steroids impact lower urinary tract function through several physiological pathways in women. The female genital and lower urinary tracts share an embryological origin, with estrogen and progesterone receptors distributed throughout the lower urinary tract and pelvic floor. Estrogen enhances periurethral tissue integrity, epithelial thickness, blood flow, and neuronal control. 3 Furthermore, changes in systemic estrogen and progesterone levels are associated with increased risks of LUTS (mainly urinary incontinence [UI] and other bladder storage symptoms) and rUTIs in the postpartum period, 4 as well as during and following menopause. 1 , 2 Local estrogen applied vaginally has been shown to reduce these symptoms and conditions in peri- and post-menopausal women. 5 , 6 However, other observations suggest a more complex relationship between hormones and LUTS. For instance, systemic menopausal hormone therapy has been shown to increase risks of UI onset and progression in post-menopausal women. 5 , 6 The role of local and systemic hormone use in younger, pre-menopausal women (i.e., HC use) is much less well-studied and understood. Combined HC use is known to raise levels of sex-hormone binding globulin (SHBG), 7 leading to a corresponding decrease in levels of circulating sex steroid hormones, including testosterone, 7 endogenous estrogen, 8 and potentially exogenous estrogen. Therefore, it is possible that combined HC use may contribute to hormone-mediated genitourinary symptoms, such as vaginal dryness and decreased libido―two reported or hypothesized combined HC side effects 9 ―and other menopausal symptoms, such as UI, in pre-menopausal women. Consistent with this hypothesis, one analysis of the Nurses’ Health Study II cohort found that having ever used oral contraceptive pills (OCPs) was associated with a greater risk of UI among pre-menopausal women, with increasing risk as the duration of OCP use increased. 10 Similar positive findings were observed for several other LUTS in a prospective analysis of adolescents in the Avon Longitudinal Study of Parents and Children 11 and in two cross-sectional analyses of pre- and post-menopausal women. 12 , 13 However, null and even inverse associations were observed in other studies, 14 – 18 drawing into question a possible influence of exogenous hormone use on lower urinary tract function in younger women. To help resolve this question, we used data from the Boston Area Community Health (BACH) Survey 19 to perform a comprehensive examination of HC use and lower urinary tract dysfunction, including separate analyses for current, past, and systemic use with risks of UI, a wide range of other LUTS, and rUTIs in pre-menopausal women. This investigation extends a previous analysis of reproductive factors and LUTS progression in BACH that explored use of any contraception in relation to progression of a composite measure of LUTS (i.e., the American Urological Association Symptom Index) in pre- and post-menopausal women combined. 20