Using UK Biobank for sexual minority health research

UK Biobank recruited 503,325 participants aged between 40 and 69 years from 2006 to 2010 in 22 assessment centers across the UK. All people registered with a National Health Service family doctor and living within approximately 25 miles of an assessment center were invited. From 9.2 million invitations, the recruitment response rate was 5.5%. Participants provided detailed information at baseline and follow-up assessments, and provided samples and consent for linkage. Further methodological details are available at www.ukbiobank.ac.uk. The data used in this study were accessed through UK Biobank (application number 42861). Sexual history information was collected using a self-completion touch-screen tool.

NATSAL-3 is a nationally representative survey of women and men living in private households in the UK. Data collection was between September 2010 and August 2012 [21]. History of same- and opposite-sex sexual behavior was collected through a self-completion questionnaire. Access to these data was obtained through UK Data Archive (project ID: 176077) [25]; respondents aged 40–69 were included for comparability with the UK Biobank sample.

In the UK Biobank baseline assessment, participants were asked a series of questions about their sexual history.

The initial question asked participants whether they were willing to continue to the sexual history questions: (‘The next section contains questions about your sexual history. If you feel that a question is too sensitive, you can skip the questions or skip the entire section if you prefer’).

Respondents were then asked: ‘What was your age when you first had sexual intercourse? (Sexual intercourse includes vaginal, oral or anal intercourse),’ with an additional response option allowing respondents to identify that they had never had sex.

Respondents were additionally asked: ‘Have you ever had sexual intercourse with someone of the same sex?,’ ‘How many sexual partners of the same-sex have you had in your lifetime?’ and ‘About how many sexual partners have you had in your lifetime?’ There were additional responses to this question, ‘Do not know’ and ‘Prefer not to say.’

Based on responses to these questions about numbers of lifetime sexual partners and numbers of same-sex sexual partners, we identified women and men who reported sex exclusively with men and women, respectively (WSEM and MSEW). We additionally identified women and men who reported exclusively same-sex sex (WSEW and MSEM) and women and men who reported sex with both women and men (WSWM and MSWM).

A subgroup of UK Biobank participants were asked the sexual history questions again at a repeat assessment visit in 2012/3 and a further subgroup were asked again at a further assessment occurring from 2014 onward.

Using NATSAL-3, we considered responses to the question: ‘Sexual experience is any kind of contact with another person that you felt was sexual (it could be just kissing or touching, or intercourse or any other form of sex). I have had some sexual experience…,’ with response options paired for sex with both men and women ‘never’/‘only,’ ‘at least once’/‘more often’ and ‘about equally often.’ In addition, respondents in NATSAL-3 were asked ‘Which of the options on this card best describes how you think of yourself?’ with response options ‘Heterosexual/straight,’ ‘gay/lesbian,’ ‘bisexual’ and ‘other.’

In both UK Biobank and NATSAL-3, age and sex at baseline assessment were recorded, and the Townsend deprivation index (a small area measure, which is scaled to a mean of 0 in the population, with lower or negative values indicating less deprived areas) was used as a measure of material deprivation for respondents from UK Biobank. The index is constructed from the following four census variables, households without a car, overcrowded households, households not owner-occupied, persons unemployed, and in the UK it is widely used in identifying areas of the greatest social need in resource allocation decisions [26]. The Townsend deprivation index was calculated immediately prior to participant joining UK Biobank based on the preceding national census output areas; each participant is assigned a score corresponding to the output area in which their postcode is located [27].

In our first set of analyses, we calculated the numbers of UK Biobank participants for whom exclusively same-sex, exclusively opposite-sex and both same- and opposite-sex sexual behavior could be described.

Among women and men who reported sexual history at both the baseline UK Biobank assessment and then again at either the first repeat assessment visit or, if these data were not missing, the second, we calculated longitudinal consistency for exclusively same-sex, exclusively opposite-sex and both same- and opposite-sex sexual behavior.

We additionally described the age, deprivation and lifetime number of sexual partners of UK Biobank participants, stratified by sex and sexual history.

In our second set of analyses, we compared reported sexual behavior in UK Biobank with estimates from NATSAL-3, using the survey weights from NATSAL-3 to give nationally representative estimates in the UK population. We used NATSAL-3 to estimate the positive predictive value (PPV), and sensitivity of sexual behavior as a measure of sexual identity. These PPV are the percentage of women and men in each group defined by sexual behavior who identify with the most closely related measured sexual orientation and these measures of sensitivity present the percentage of women and men who identify with each sexual orientation who report the most closely related sexual behavior history.

Finally, we produced a set of weights that will allow researchers to explore health disparities based on sexual behavior using UK Biobank and use these weighted measures to estimate of disparities experienced by gay, lesbian, bisexual and other sexual minority women and men.

The lay abstract for this work was revised and improved following comments from members of an LGBTQ+ research participation panel [28].

UK Biobank includes a large population-based sample of almost 3000 participants with a history of exclusively same-sex sexual behavior and over 9000 people with a history of sex with both men and women, with acceptable longitudinal consistency over repeated assessments. Reported history of same-sex sexual behavior overall was lower among UK Biobank participants than among respondents to NATSAL-3. Based on data from NATSAL-3, the appropriateness of using sexual behavior alone in UK Biobank (or other resources where sexual behavior, rather than sexual identity is measured) as a measure of sexual orientation is mixed; sexual behavior has high PPV and low sensitivity for sexual identity among women and men who exclusively have same-sex sex; a history of both same- and opposite-sex sex has low PPV and high sensitivity for bisexual sexual identity.

Where previous research exists, our research findings are consistent. Our finding that WSWM have more sexual partners overall is consistent with previous analyses of a prior NATSAL survey [29]; as is our finding that people with a history of same-sex sexual behavior are younger, on average [24,30]. Our finding that WSW and MSM live in less affluent areas compared with WSEM and MSEW is consistent with previous research [1]. The link between material deprivation and poorer health is also well known [31], and so this finding additionally highlights the importance of addressing these health disparities. Although based on slightly different measures, our analyses of the intersection between sexual behavior and identity are also consistent with earlier work from the USA [23] and the UK [24].

Regarding the lower reported same-sex sexual history in UK Biobank compared with NATSAL-3, changes in sexual orientation and sexual behavior over time, with increased reporting of same-sex behavior and minority sexual orientations [22,30] are one possible explanation. A second is that people from less deprived areas are over-represented among UK Biobank respondents [17], while estimates from NATSAL-3 are nationally representative, and WSW and MSM on average live in more deprived areas [1]. A third explanation might be differences in the question wording across the two surveys, with a potentially broader definition of sexual experience in the NATSAL-3 survey question used in this analysis. Nonetheless the patterns of sexual behavior reported are consistent across the two datasets, and with prior external comparisons [18] and we argue that this is a reasonable validation of the UK Biobank responses. The availability of an independent data source (NATSAL-3) is a real strength for this work exploring the validity of UK Biobank as a tool to study sexual minority health disparities. A limitation of this study is that UK Biobank contains no self-reported measure of gender identity, although gender identity disorders in 90 cohort participants are captured through health record linkages [32]. We also note that many people have a sexual identity that is something other than heterosexual, gay/lesbian or bisexual [33] and finally that sexual orientation may represent a continuum rather than a set of mutually exclusive terms [34].

The issues of the representativeness of UK Biobank for the whole UK population are well known [17], however, the value of the resource is also well established. Over 15,000 people report a history of ever having sex with someone of the same sex and over 12,000 participants can be identified as exclusively having same-sex sex or having sex with both women and men, on the basis of their responses to questions about lifetime numbers of sexual partners. This makes UK Biobank an incredibly important resource for sexual minority health research, as the depth of health, risk factor and other information is unique. The fact that respondents were not sampled on the basis of their sexual orientation and that there is a comparable group of participants who have never had same-sex sex is a second strength.

However, there are limitations to an approach that considers researching health disparities on the basis of sexual behavior rather than sexual orientation. First, this ignores the fact that in the UK it is sexual orientation rather than sexual behavior that is the protected characteristic under the 2010 Equality Act (and for which, there is therefore a statutory duty for public bodies to identify and address any inequalities) [11]. Second, there is the more fundamental challenge that research to address health inequalities should be person-centered, and focus on the identity rather than the behavior of the groups under study.

A second important limitation is that only approximately 10% of the participants identified in UK Biobank with a history of sex with both women and men will identify with a bisexual sexual orientation. Bisexual women and men have particularly poor mental health [1], however, this is a group that cannot be well identified in UK Biobank through the measure of sexual behavior available.

We cautiously suggest that UK Biobank is a valuable research resource for sexual minority health, although with limitations. We recommend that future waves of data collection with the UK Biobank cohort consider asking new questions about sexual orientation and gender identity to address in part these data gaps, but for researchers using data collected at baseline we present here three possible approaches for how best to use this resource.

Although the causes of health disparities experienced by sexual minority women and men are complex, for the long-term health conditions that UK Biobank is designed to study, it is plausible that some are associated with sexual behavior rather than sexual orientation, for example, the higher risks of human papillomavirus-associated oropharyngeal and anal and penile cancers [35]. Furthermore, although bisexual women and men are not well identified in UK Biobank, among women history of sex with both women and men is none the less a marker of adverse health outcomes, and this may well be an important group to study [29].

Although bisexual women and men are not well identified in UK Biobank, women and men who exclusively have same-sex sex has much higher PPV for gay/lesbian sexual orientation. For some research projects, this may be a valuable approach to take.

The weights that we present in Table 6 provide a third possible approach for researchers. By creating new variables for each sexual orientation with these values based on sexual history, and using these new variables as the predictors in a model, adjusted estimates of disparities by sexual orientation can be estimated [36,37].