Lesbian and bisexual women's gynaecological conditions: a systematic review and exploratory meta-analysis

This systematic review was conducted according to a prospective protocol that was lodged with PROSPERO (CRD42015027091) on 6 October 2015. Studies were eligible using the following inclusion criteria: (1) population – women self‐described as LB, women who described themselves as WSW, or having sex with women and men (WSWM); (2) exposure – women with benign (non‐infectious) and malignant gynaecological conditions; (3) comparator – heterosexual women or women self‐describing as only having sex with men; (4) study design – any comparative studies including randomised controlled trials, case–control studies, cohort studies, cross‐sectional analyses, experimental studies, or secondary studies with data of interest. Studies had to contain primary data and be peer‐reviewed. Only studies reporting after the year 2000 were eligible. There were no restrictions on setting or language. Studies were excluded if: the sexual orientation and behaviour of women were not clear; there was no comparison with heterosexual women; there were no outcomes of interest; or if they were opinions, editorials, conference abstracts or case reports. Search terms were developed based on the population and exposures sought. MeSH terms and synonyms were used to widen the search. A total of seven databases were used: Ovid Medline; Ovid Embase; Ovid PsycInfo; Web of Science – Science Citation Index; Cochrane; British Nursing Index; CINAHL. Searches were limited to 1 January 2000–22 October 2015 in view of prior piloting. Reference lists of reviews and primary studies were also searched. The authors checked studies on lesbian health used in other projects. A full table of search terms can be found in Appendix S1. After removing duplicates, the remaining papers were assessed independently for relevance first by title, and then by abstract (KR, KYG, CM). All articles were included for full‐text assessment if any author considered the abstract relevant or there was uncertainty. Full‐text assessment to determine inclusion in the systematic review was carried out by all authors. Any disagreements were resolved by discussion. A standard form was devised prior to data extraction and quality scoring, based on the content of the papers and the aims of the review. Data were independently extracted by two authors (KR, KYG). No authors were contacted about data discrepancies. Studies were appraised for selection, performance, attrition and detection biases, and reported in the categories of risk of bias, study design issues and whether the study would be representative of LB in the general population. No formal quality appraisal was carried out as there was no single validated checklist that would be appropriate for all of the studies due to the diverse study designs. Numbers were converted to n (or reverse n ) and %, using back‐calculation and estimates from figures in the published articles when required. P ‐values were calculated when not provided in the paper using Fishers Exact test or Chi‐squared test (with Yates’ correction) as appropriate. Meta‐analysis was conducted using RevMan version 5.3 on the outcomes of gynaecological conditions in LB women, where three or more studies reported results. Subgroups of LB women in cervical and uterine cancer were conducted.

From 611 records (41 duplicates), 208 abstracts were selected, of which no papers were unavailable and 47 full papers were read. Eleven studies were included [see Figure S1 (PRISMA flow chart) and Table S1 (excluded studies with reasons)]. The 11 studies had a variety of different populations, exposures, study designs, settings and outcomes. All included studies came from high‐income countries. No studies were found in any language apart from English. No primary study was found that directly addressed the question. These are detailed in Table 1 . There were three prospective cohort studies, four retrospective cohort studies, four cross‐sectional surveys, and no case–control studies. Settings varied and included reproduction and infertility clinics, online and telephone questionnaires, and large‐scale health studies. All were from high‐income countries (five USA, two UK, and one each from Canada, Belgium, Denmark and Sweden). Sizes ranged from 31 29 to 91 582 participants. 11 Recruitment methods varied, and included clinic, convenience and population samples. Data collection included telephone and online surveys, self and medical staff completed questionnaires, a national cancer registry, and medical chart review. Funding sources were wide ranging, though six had no details of funding or were unfunded. Characteristics of the included studies IUI, intra‐uterine insemination; PCOS, polycystic ovary syndrome. Table 2 shows that to date there is limited evidence on gynaecological conditions in LB women compared with heterosexuals, and it is particularly sparse for bisexual women. However, there were some interesting findings. Compared with heterosexual women there were significantly more bisexual women with PCOS in one study, 29 and lesbians in one study. 12 However, there was no significantly higher rate amongst lesbians in three other studies. 30 , 31 , 32 There was significantly less chronic pelvic pain for lesbians in one study, 33 but no difference in another. 34 There was significantly more chronic pelvic pain for bisexual women in two studies. 33 , 34 In one study there were significantly higher rates of cervical cancer in both lesbian and bisexual women, and higher rates of uterine cancer in lesbians but lower rates in bisexual women. 35 Another study found lower rates of cervical carcinoma in situ in lesbians. 36 There was no significant difference in rates of endometriosis and fibroids. 12 , 31 , 37 Two studies reported rates of confounders (alcohol, smoking and BMI) by sexual orientation, 11 , 31 and only one study took a confounder into account (no difference in BMI between groups when measuring PCOS). 30 Numerical results from included studies NS, not significant; PCOS, polycystic ovary syndrome; RR, relative risk. Back calculations of n using weighted denominator. Back calculation by authors using weighted prevalence estimates given in paper. Percentages calculated by authors. As given in the paper, comparison with expected rates not given. Statistics calculated with QuickCalcs online calculator ( www.graphpad.com/quickcalcs/contingency1 ). significant differences in bold. Exploratory meta‐analyses in PCOS, endometriosis and fibroids showed no significant differences overall (in both lesbian and bisexual women; Figure S2). There was a higher rate of cervical cancer in bisexual women than heterosexual women [odds ratio (OR) 1.94; 95% CI 1.46–2.59], but no difference overall (OR 0.76; 95% CI 0.15–3.92) or in lesbians (OR 0.30; 95% CI 0.04–2.08). There was a lower rate of uterine cancer in lesbian women than heterosexual women (OR 0.28; 95% CI 0.11–0.73) and overall (OR 0.36; 95% CI 0.13–0.97), but no difference in bisexual women (OR 0.43; 95% CI 0.06–3.07; Figure 1 ). Meta‐analyses of rates of gynaecological cancers in lesbian and bisexual ( LB ) women compared with heterosexual women. (A) Meta‐analysis of rates of cervical cancer. (B) Meta‐analysis of rates of uterine cancer. The ‘events’ heading is the number of patients with that condition; ‘total’ is the number of patients in that group; and the ‘weight’ is the relative impact of each study on the meta‐analysis result. ‘Favours LB ’ means that if the point estimate of the meta‐analysis effect size is in that side of the plot, fewer LB women have the condition than heterosexuals and vice versa.

Too little is known about LB women, and new comparative studies would be valuable to ensure conclusions, for instance regarding cancer incidences, are based on solid evidence rather than a negative hypothesis. A key message is that LB women's possible problems and difficulties have to be handled with care. For the majority of gynaecological conditions, there are no differences between LB and heterosexual women. Clinicians may not know or take into account their patients’ sexuality when treating them for gynaecological disorders, and should be open and non‐stigmatising to obtain this information. Existing notions about PCOS must be undone, and clinicians should not treat sexuality as an association. Clinicians should be aware of higher rates of pelvic pain and cervical cancer in bisexual women. Current sexual identity or behaviour is not a correct indictor of risk for cervical cancer, as women who currently identify as heterosexual may have prior bisexual experience. Similarly, parity does not indicate current sexual identity or behaviour as lesbians may have children. 21 , 23 Furthermore, fertility‐limiting gynaecological diseases, such as endometriosis and fibroids, need to be acknowledged and addressed regardless of sexual orientation and providers’ perceived likelihood of future pregnancies. Clinicians should provide appropriate information to all women, and not make assumptions about their patients’ sexuality in any sexual or reproductive health setting. Heterosexuality should not be assumed in gynaecology as many LB women would prefer to disclose their sexuality but feel unable to, silenced by this assumption. 47 Conversely, they may be reticent to ‘come out’ to their healthcare professional for fear of adverse reactions. It may be that more women do not disclose due to safety issues and the uncertainty about how they will be treated. Past experiences of homophobia, heterosexism and discrimination can directly affect patterns of healthcare seeking, leading to avoidance of routine screening, and reluctance to seek help and advice in future. Healthcare providers should ensure they are aware of potential stigmatisation and issues of cultural competency with sexual minority women, ensuring equitable access and optimal healthcare for patients. Openness and sensitivity allows for an ease in communication and the formation of a better doctor–patient relationship. Half of LB women in a large UK community survey have not disclosed their sexual orientation to their primary healthcare provider. 48 It is important to remember that not all LB women will want to disclose. 49 The paucity of primary studies may relate to a lack of interest, lack of funding, or stigmatisation. These exploratory results need confirmation with high‐quality large‐scale studies into LB women's gynaecological health. Whilst identity and behaviour are overlapping categories, they must be distinguished in future research. An important implication is that sexual orientation should be routinely recorded as part of data collection in cohort studies, alongside medical records, to allow more large‐scale interpretation of disease patterns (and potential confounders) as previous authors have also concluded. 28 , 50 More work is required in developing countries, although matters such as routine recording may be problematic due to stigma or illegality of homosexuality. Research has suggested that a relatively large proportion of GPs have difficulties with discussing sexual identity with patients. 51 , 52 , 53 If this is also true with secondary care and gynaecology staff, work is required to help practitioners be more confident and comfortable with their LB patients. How best to achieve this is unclear, and studies exploring methods of training health staff should be developed. None declared. Completed disclosure of interests form available to view online as supporting information. The authors were involved as follows: KR, SB, CM conception and design; KR, KYG, SB, CM execution, analysis, drafting manuscript and critical discussion; all were responsible for revision and final approval of the manuscript. All authors had full access to all of the data (including statistical reports and tables) in the study, and can take responsibility for the integrity of the data and the accuracy of the data analysis. Not required. The study arose from an initial project undertaken during a part‐funded BSc at King's College London (KR) but was otherwise unfunded. Open access for this article was funded by King's College London.

Despite a paucity of existing primary research, key findings were: higher rates of chronic pelvic pain in bisexual compared with heterosexual women; no statistically significant differences in PCOS, endometriosis and fibroids; a higher rate of cervical cancer in bisexual than heterosexual women but no difference overall; a lower rate of uterine cancer in lesbian than heterosexual women and overall, but no difference in bisexual women. Strengths include protocol preregistration in the PROSPERO database, no language restriction, careful presentation of numerical results and inclusion of global data from a variety of sources. Rates of gynaecological conditions in LB women were not the primary focus of the majority of included papers, therefore other publications may exist that might have yielded further useful information. There was no formal quality review and we were unable to interrogate confounders (i.e. country, age, ethnicity, class, education, BMI, actual sexual behaviours). This makes it difficult to ascertain whether differences in reported rates are truly due to sexual orientation or confounding factors. Furthermore, the lack of consistent definitions of sexual orientation (Table 1 ) poses a challenge in discerning the differences in rates of gynaecological disorders between LB and heterosexual women. Different aspects of sexual orientation, including behaviour, identity and attractions, may confound each other, making this a difficult research area. The meta‐analyses should be approached with caution due to the heterogeneity of studies included. Nevertheless, there is sufficient consistent information to draw some clinical conclusions with generalisability. There have been no previous systematic reviews of gynaecological disorders in LB women. The lack of a significant difference in rates of PCOS between heterosexual and LB women contrasts with Agrawal et al. 12 who found significantly raised prevalence of PCOS in lesbians (although the researchers were not blind to sexual orientation), and information published through public health websites for patients, advising lesbians they could have higher rates of PCOS than heterosexuals. 20 Chronic pelvic pain is typically associated with endometriosis and infection. 38 There were no differences in endometriosis rates in lesbians, and no studies examining endometriosis in bisexual women were found. Sexually transmitted diseases were out of scope of this review, and reported differences in bacterial vaginosis and the vaginal microbiome deserve further exploration. 39 , 40 , 41 Both studies investigating chronic pelvic pain examined pain not attributed to a medical cause, i.e. functional. The higher rate of functional pelvic pain among bisexuals could be linked to lifestyle factors. Sexual minorities are exposed to more adverse childhood experiences than heterosexuals, including child abuse, housing adversity and intimate partner violence. 42 Roberts et al. 34 suggested that dysregulation of the hypothalamic–pituitary–adrenal system and related inflammatory processes resulting from abuse or violence victimisation may predispose individuals to experience functional pain. The higher rate of cervical cancer in bisexuals could plausibly be related to higher exposure to risk factors, including smoking and unprotected sex with men (especially in adolescence), and lower attendance for screening. 43 The lower rate of uterine cancer in lesbian women is surprising due to the lower parity and higher BMI when compared with heterosexual women, 4 , 23 which are both risk factors for uterine cancer. 44 These risks may be offset by the higher rates of smoking among lesbian women, 3 which is a protective factor. 45 Oestrogen excess is also a risk factor for uterine cancer. There is little evidence on the comparison of oestrogen levels between LB and heterosexual women. One small study found no difference in oestrogen levels between lesbians and bisexual women. 46 Another study investigating PCOS found no difference in oestrogen levels between lesbians and heterosexuals with normal ovaries. 12 An old review also found no difference in oestrogens in lesbians compared with heterosexual women. 13 It is unclear whether the lower rate of uterine cancer is due to a combination of differing risk factors and protective factors, but further large‐scale research is necessary to confirm this correlation.

Health research in sexual minority women indicates that lesbian and bisexual (LB) women differ from heterosexual women in health risks, health behaviours, and how they experience healthcare. 1 The health risks and health behaviours of LB women are thought to negatively impact their gynaecological health. These include increased tobacco and alcohol use, 2 , 3 higher body mass index (BMI), 4 , 5 , 6 and reduced uptake of cervical cytology screening. 7 , 8 , 9 Lesbians are also said to have fewer reproductive behaviours that are associated with protection from various reproductive system cancers than heterosexual women, including use of oral contraceptives, childbirth and breastfeeding. 10 , 11 Furthermore, there is a theory that lesbians have higher levels of testosterone, which could contribute to higher rates of polycystic ovary syndrome (PCOS). 12 , 13 , 14 Excessive alcohol use and smoking are risk factors for a range of gynaecological disorders, including uterine, breast and ovarian cancers. 15 Increased BMI is a risk factor for conditions such as ovarian cancer and PCOS. 16 , 17 Accordingly, journal articles and public health sources have suggested the combination of increased risk factors may lead to LB women having increased rates of gynaecological conditions. 18 , 19 , 20 Nevertheless, a paucity of research actually measures gynaecological disorder incidence and prevalence in this population. Conversely, conflicting results regarding reproductive behaviours have been reported. Previous pregnancy and hormonal contraceptive use are common among women who have sex with women (WSW), whether they self‐identify as lesbian or not. 21 Teenage hormonal contraceptive use has been found to be lower in lesbian women, but higher in other sexual minority groups when compared with exclusively heterosexual women. 22 Both lesbian and bisexual adolescents have been found to have higher rates of pregnancy. 23 Inconsistent evidence makes it difficult to justify the claim that LB women have fewer reproductive behaviours than heterosexual women. Previous systematic reviews on the health of LB women have considered smoking cessation, 24 mental health, 25 substance abuse, 26 weight, 4 breast cancer 27 and cervical screening. 7 One narrative review concluded there was little to no published literature on incidence and prevalence of endometrial and cervical cancer in LB women, and that further research was necessary to fill the knowledge gap. 28 The aim of this study was to examine the association between LB women's identity or behaviour and their gynaecological health, focusing on the null hypothesis that LB women have the same rates of gynaecological conditions as heterosexual women.

Figure S1. PRISMA flow diagram. Click here for additional data file. Figure S2. Meta‐analyses of rates of gynaecological conditions in lesbian/bisexual women compared to heterosexual women. Click here for additional data file. Table S1. Full Text Papers Excluded. Click here for additional data file. Appendix S1 . List of Search Terms. Click here for additional data file. Video S1. Author Insights. Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file. Click here for additional data file.