Objective To review the bisexual-specific prevalence and likelihood of cigarette smoking relative to lesbian/gay and heterosexual individuals.
Data sources We searched MEDLINE, PsycInfo, CINAHL, Scopus and LGBT Life databases (from 1995 to September 2019) for studies reporting cigarette smoking among bisexuals versus their comparators.
Study selection Observational, quantitative, peer-reviewed studies providing estimates for lifetime, past 30 days or current cigarette smoking among bisexuals and any of the two comparators were selected.
Data extraction Data on sexual orientation groups, cigarette smoking, sample type and mechanism, data collection mode, country and median year, as well as gender and age groups were extracted.
Data synthesis Random-effects meta-analysis was used to estimate the pooled OR (95% CIs) of cigarette smoking. Meta-regression was used to examine the difference in the prevalence of cigarette smoking by study and sample characteristics. Of 4663 unduplicated records, 47 unique studies were included (14, 23 and 22 studies on lifetime, past 30 days and current cigarette smoking, respectively). Compared with lesbians/gays and heterosexuals, bisexuals were 1.25 (1.15 to 1.37) and 2.18 (1.84 to 2.59) times more likely to report lifetime smoking, 1.17 (1.08 to 1.27) and 2.49 (2.20 to 2.83) times more likely to report past 30 days smoking and 1.19 (1.00 to 1.43) and 2.26 (1.97 to 2.59) times more likely to report current smoking. Gender was a significant covariate in the meta-regression models.
Conclusions Cigarette smoking was more prevalent among bisexuals than lesbians/gays and heterosexuals, with the estimates showing a greater magnitude among bisexual women relative to all other sexual orientation/gender subgroups.
- smoking caused disease
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Cigarette smoking is an important public health concern and one of the leading causes of preventable morbidity and mortality worldwide. The 2017 Global Burden of Disease Study found smoking as the second most significant contributor to death (average: 7.1 million) and disability-adjusted life years (average: 182 million).1 In the USA, cigarette smoking is responsible for an estimated 480 000 deaths per year.2 Disparities in the prevalence of cigarette smoking have been observed across sociodemographic groups, in particular, sexual orientation groups.3 4 For example, the National Health Interview Survey reported a prevalence of 20.5% for cigarette smoking among sexual minority individuals (inclusive of gay, lesbian, bisexual (LGB) persons) compared with 15.3% among heterosexual persons in the population of US adults in 2016.4 Ryan et al 5 in the first systematic review and Lee et al 6 in an updated systematic review of smoking among sexual minorities reported some evidence for greater prevalence of smoking among sexual minorities versus heterosexuals; however, neither study conducted meta-analysis in order to quantify the magnitude of this finding.
Over the past decade, research has indicated that bisexual people have the poorest health outcomes and risk behaviours when compared with their heterosexual and gay/lesbian counterparts. However, most studies tend to combine data from sexual minority populations, despite literature that consistently highlights intragroup variations across a wide variety of health outcomes and behaviours.7–10 For example, bisexual people are more likely to report current depression,7 current anxiety7 and suicide attempts8 than lesbian/gay and heterosexual participants. Disparities in tobacco use have also been reported among sexual minorities, with the emerging evidence showing a higher prevalence of cigarette smoking among bisexuals than lesbian/gay and heterosexual individuals9–12 and underscoring the need for more specific public health interventions within this population.9 13 For example, Matthews et al discussed the importance of the disaggregation of the study outcomes when examining health disparities by sexual minorities, highlighting the greater risk of bisexual adolescents for health risk behaviours.13 However, fewer studies have systematically documented important health behaviours such as cigarette smoking among bisexuals relative to their lesbian/gay and heterosexual peers. Furthermore, existing data suggest that there may be gender differences in smoking, with the limited available data reporting that bisexual women may be at a higher risk9–14; however, the magnitude of this difference has yet to be quantified. Analyses by sexual orientation and gender can help better understand health disparities13 and provide stronger evidence for sexual minority-targeted public health interventions.15
In light of these gaps in the existing literature, the primary objective of the current systematic review and meta-analysis was to summarise the evidence for disparities in the prevalence of cigarette use among bisexual people relative to gay/lesbian and heterosexual people. Specifically, we report bisexual-specific data on three cigarette smoking measures: current cigarette use, past-30 days cigarette use and lifetime cigarette smoking. Where sufficient data were available, we report the pooled estimates of each outcome across sexual orientation groups, overall and stratified by gender. Using meta-regression analysis, we also examine whether study-level and sample-level characteristics have any role in explaining the heterogeneity in each study outcome.
This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses recommendations for reporting on systematic reviews.16 The protocol of this review was registered in the PROSPERO database (registration number: CRD42016053991). We searched the following databases: Medline (OVID, including e-pub ahead of print, in process and other non-indexed citations), PsycINFO (OVID), CINAHL (EBSCO), LGBT Life (EBSCO) and Scopus, from 1995 to September 2019. As recommended,7 we chose 1995 as the inception date due to the concern that sexual minority research before 1995 might lack rigour (eg, insufficient coverage of bisexual individuals in sampling). Subject headings and keywords were selected as appropriate for each database (online supplementary appendix A).
Eligible studies for inclusion in the meta-analysis were original quantitative studies published in peer-reviewed sources in English, French or Spanish, carried out in a community or population-based setting, reporting the prevalence of cigarette smoking specific to bisexual people of all age groups. Studies were excluded if they recruited participants in clinical settings, did not report bisexual specific data, did not separate cigarette smoking from other tobacco products (eg, reporting any tobacco use as the study outcome), did not specify a clear definition of cigarette smoking or a clear period when defining cigarette use or were conducted among specialised populations (eg, injection drug users) in which a higher prevalence of the study outcomes are expected. Detailed inclusion and exclusion criteria are shown in online supplementary appendix B.
All the retrieved records were imported to Covidence, a systematic review application (www.covidence.org). After removing duplicates, titles and abstracts were screened by two authors (MS and BA) to remove irrelevant studies (eg, qualitative studies, case reports and no primary/original data). Studies passing this step were assessed by the same two authors at the full-text screening phase to confirm eligibility; conflicts were resolved by the senior author (LR). Full texts of the included studies were again assessed by two authors to determine whether original and necessary data (eg, prevalence and sample size in each study group) on cigarette smoking specific to bisexual individuals were reported. In this final phase, studies with incomplete data were removed.
Data were extracted into a standardised data extraction spreadsheet specific to bisexual, gay or lesbian, and heterosexual people. Study and sampling characteristics were also recorded, as described below. Other key data concerning the prevalence of cigarette smoking included sample size, prevalence and/or the total number of individuals in each group who reported cigarette smoking. Data extractions were completed by the two authors, and the accuracy of the data was checked by the senior author if needed.
We extracted data specific to three sexual orientation groups: bisexuals, lesbians/gays and heterosexuals. Bisexual people were the main study group, meaning that we first collected data on the prevalence of cigarette smoking among bisexuals, and then sought data on the other two groups for the purpose of comparison for bisexuals versus lesbians/gays and bisexuals versus heterosexuals.
We included data on three binary cigarette smoking measures: (1) lifetime cigarette smoking, defined as having ever smoked cigarette over the lifetime (with or without the 100-cigarette threshold that was commonly but not universally applied as part of this definition), (2) past 30 days cigarette smoking, defined as any cigarette smoking in the past 30 days prior to the start of the survey/interview (with or without the 100-cigarette threshold as noted above), (C) current cigarette smoking, self-declaration of ‘current smoking’ behaviour, defined as now smoking every day or some days (similarly with or without the 100-cigarette threshold). These three timeframes for cigarette smoking were considered because they are the most common definitions of cigarette smoking in the tobacco literature, including among the studies identified in this review.
The following information considered to be important for the purpose of data quality was also extracted17: sample type and/or mechanism (population-based sampling applying a complex, random procedure vs other non-random methods such as community or convenience sampling) and data collection mode (self-administered (telephone) interview, online or multimode). We also recorded the country in which the study participants were sampled (US based, as the most common country from which studies originated, vs other countries) and the median year of data collection.
We recorded (A) gender: genders pooled and gender-specific prevalence of cigarette smoking were recorded if a study reported either both or only one of these two. For the studies reporting genders pooled and gender-specific data, we used genders pooled data for the overall estimates of the prevalence of cigarette smoking while using the gender-specific data when reporting estimates stratified by gender; (B) age groups (adolescents, young adults, adults or older adults): we kept these original categories as reported in each included study given that there was some overlap between the first and second categories and the second and third categories; and (C) sexual orientation: estimates were reported stratified by self-identification (mostly measured as the following options: straight/heterosexual vs gay/lesbian vs bisexual), behaviour (mostly measured as the following options for sexually active individuals: having sex with same sex partner vs having sex with same sex and opposite partners) and attraction (mostly measured as the following options: feeling sexually attracted to same sex only vs same sex and opposite sex vs opposite sex only) definitions of sexual orientation, even though self-identification was the most common definition. Thus, genders-pooled and gender-specific data for self-identified sexual orientation were reported, while only gender pooled data were reported for the other two constructs of behaviour and attraction due to insufficient records (ie, <3 studies).
Risk of bias assessment
Consistent with the literature,17 no risk-of-bias scale was used to assess the validity of the included studies as available scales do not address the particular methodological concerns with regard to sexual orientation research. However, in line with the Meta-analysis of Observational Studies in Epidemiology guidelines,18 we included key specific variables that are typically used in risk-of-bias scales (eg, sample type, data collection mode and sexual orientation construct) as covariates in the analyses.
We reported the pooled prevalence (proportion) of each cigarette smoking outcome with 95% CIs for bisexuals, lesbians/gays and heterosexuals. The pooled prevalence estimates were computed after Freeman-Tukey double arcsine transformation to stabilise the variances when constructing the 95% CIs.19 Data were pooled by applying a random-effects model using the method of DerSimonian and Laird, with the estimate of between-study variability (heterogeneity) being taken from the inverse-variance fixed-effect model.20 A random-effects model was used a priori given the expected high between-study heterogeneity in prevalence studies. We used Stata’s metaprop package to provide the pooled estimates of cigarette smoking prevalences.21
In addition, we reported the crude ORs and 95% CIs examining the likelihood of each cigarette smoking outcome among bisexuals compared with lesbians/gays and heterosexuals. For the purpose of public health importance, we also reported prevalence differences (shown in online supplementary appendix D1, E1, F1). Stata’s metan package22 was used to estimate these measures. We then used univariate and multivariable random-effects meta-regression using aggregate-level data to assess the relationship between sample and study characteristics with each study outcome. These analyses were done to explore whether these characteristics can explain variations across the included studies. Due to the fundamental role of gender/sex in sexual orientation health disparities, as described above, subgroup analysis was also done for gender/sex. We included the following covariates: gender/sex, age groups, sampling methods, study regions, data collection mode and sexual orientation construct. These analyses were performed using Stata’s metareg package23 by applying the Knapp-Hartung modification24 to the variance of the estimated coefficients in multivariable models, accompanied by the use of t distribution instead of the standard normal distribution when constructing the 95% CIs and p values for the unstandardised beta coefficients. The presence of heterogeneity was assessed using I2, interpreting heterogeneity as low (25%), moderate (50%) and high (75%).25 Other measures of heterogeneity including adjusted R2 and Cochrane Q test were also reported. Results of these meta-regression analyses are shown in online supplementary appendix D3, D4 for lifetime cigarette smoking, online supplementary appendix E3, E4 for the past 30 days smoking and online supplementary appendix F3, F4 for current smoking. Finally, publication bias was visually assessed using funnel plots26 using Stata’s metafunnel package,27 and the Egger test for asymmetry using Stata’s metabias package,28 with p values <0.05 interpreted as the probable presence of publication bias.29 All statistical analyses were performed in Stata V.15.
We screened titles and abstracts of 4663 unduplicated studies, of which 392 full-text studies were checked for eligibility criteria; of them, only 47 unique studies met the study eligibility criteria (figure 1). No bisexual-specific data were the most frequent reasons for being excluded within the review process, followed by the wrong outcome, duplicate data and research on specialised populations. Of the 47 unique studies/publications, 14, 23 and 22 studies reported data on lifetime, past 30 days and current cigarette smoking, respectively. Only 4 out of 12 studies examining lifetime smoking, 0 out of 14 studies examining past 30 days smoking and 14 out of 22 studies examining current smoking considered the 100-cigarette smoking cut-off. This threshold did not moderate the associations of interest between bisexuality and cigarette smoking outcomes (data are not shown).
Lifetime cigarette smoking
Of the 14 included studies, overall, we analysed data from a total of 939 366 individuals, of whom 28 527 were bisexuals, 19 003 were lesbians/gays and 891 836 were heterosexuals. Study-specific information is summarised in table 1 and detailed in online supplementary appendix C table 1.
The overall pooled prevalence of lifetime smoking among self-identified bisexuals was 59.8% (95% CI 54.9 to 64.6), with 63.6% (95% CI 57.5 to 69.5) among women versus 54.2% (95% CI 47.6 to 60.7) among men; among lesbians/gays was 56.5% (95% CI 51.1 to 61.8), with 59.8% (95% CI 50.4 to 68.8) among women versus 51.8% (95% CI 42.4 to 61.2) among men; and among heterosexuals was 39.2% (95% CI 35.2 to 43.2), with 36.1% (95% CI 30.8 to 41.5) among women versus 39.7% (95% CI 33.9 to 45.6) among men. Pooled estimates by other sexual orientation constructs are reported in table 2 (and online supplementary appendix D1,D5)
Overall, self-identified bisexuals were 19% more likely to report lifetime cigarette smoking than lesbians/gays, OR: 1.19 (95% CI 1.00 to 1.43), with estimates for men and women as 1.10 (95% CI 0.98 to 1.24) and 1.16 (95% CI 1.01 to 1.32), respectively. Bisexuals were 2.26 (95% CI 1.97 to 2.59) times more likely to report lifetime cigarette smoking than heterosexuals, with the pooled estimate for women 2.82 (95% CI 2.42 to 3.28) higher than that among men 1.54 (95% CI 1.25 to 1.89). Details can be seen in table 3 (and online supplementary appendix D2,D6).
Compared with lesbians/gays, no covariates were identified as significant moderators explaining the slightly higher prevalence of lifetime cigarette smoking among bisexuals. Compared with heterosexuals, however, gender was a moderator of the stronger association between bisexuality and lifetime cigarette smoking among women compared with that in men, explaining approximately R2=60% of the between-study variabilities in univariate meta-regression. Gender remained significant in the multivariable meta-regression (online supplementary appendix D3,D4), meaning that except for gender, other characteristics did not statistically explain variations across the included studies.
The funnel plots for studies reporting lifetime smoking among bisexuals versus lesbians/gays were symmetrical (Egger test p value=0.413) and versus heterosexual were almost symmetrical (Egger test p value=0.099) (online supplementary appendix D7), suggesting no significant publication bias.
Past 30 days cigarette smoking
Of 23 included studies for this outcome, overall, we analysed data from a total of 854 258 individuals, of whom 49 709 were bisexuals, 21 951 were lesbians/gays and 782 598 were heterosexuals. Study-specific information is summarised in table 1 and detailed in online supplementary appendix C table 2.
The overall pooled prevalence of past 30 days smoking among self-identified bisexuals was 29.0% (95% CI 24.7 to 33.5), with 23.3% (95% CI 17.1 to 30.1) among men and 29.6% (95% CI 21.5 to 38.4) among women; among lesbians/gays was 25.6% (95% CI 20.4 to 31.0), with 23.1% (95% CI 14.3 to 33.3) among men and 22.9% (95% CI 13.6 to 33.9) among women; and among heterosexuals was 15.3% (95% CI 12.9 to 17.8), with 17.4% (95% CI 13.0 to 22.4) among men and 13.3% (95% CI 10.2 to 16.6) among women. Pooled estimates by other sexual orientation constructs are reported in table 2 (and online supplementary appendix E1,E5).
Overall, self-identified bisexuals had higher odds of past 30 days cigarette smoking than lesbians/gays, OR: 1.17 (95% CI 1.08 to 1.27), with 1.27 (95% CI 1.11 to 1.44) among women and 1.00 (95% CI 0.88 to 1.14) among men. Bisexuals were 2.49 (95% CI 2.20 to 2.83) times more likely to report smoking than heterosexuals, with the pooled estimate for women 3.27 (95% CI 2.83 to 3.78) higher than that among men 1.65 (95% CI 1.32 to 2.07). Details can be seen in table 3 (and online supplementary appendix E2,E6).
Female gender was identified as a moderator explaining the estimated differences for cigarette smoking prevalence in the past 30 days between bisexuals versus lesbians/gays. Female gender, median year of the surveys, and US studies positively and non-complex survey types negatively moderated the association between bisexuality (as compared with heterosexuals) and past 30 days smoking (online supplementary appendix E3,E4). This means that these characteristics had the potential to statistically explain variations across the studies.
The funnel plots for studies reporting past 30 days cigarette smoking among bisexuals versus lesbians/gays appeared asymmetrical (Egger test p value=0.012), in particular, with regard to the data for women (Egger test p value=0.001) but not men (Egger test p value=0.921), while symmetrical versus heterosexuals (Egger test p value=0.992) (online supplementary appendix E7).
Current cigarette smoking
For this outcome, 22 studies were included; overall, we analysed data from a total of 2 268 438 individuals, of whom 32 664 were bisexuals, 43 501 were lesbians/gays and 2 192 273 were heterosexuals. Study-specific information is summarised in table 1 and detailed in online supplementary appendix C table 3.
The overall pooled prevalence of current cigarette smoking among self-identified bisexuals was 30.4% (95% CI 28.7 to 32.2), with 28.5% (95% CI 25.4 to 31.6) among men and 32.0% (95% CI 30.0 to 34.1) among women; among lesbians/gays was 25.7% (95% CI 23.8 to 27.7), with 26.0% (95% CI 22.5 to 29.6) among men and 25.9% (95% CI 23.1 to 28.8) among women; and among heterosexuals was 18.1% (95% CI 15.7 to 20.6), with 21.4% (95% CI 19.9 to 23.0) among men and 13.8% (95% CI 13.0 to 14.6) among women. Insufficient data were available to provide pooled estimates by other sexual orientation constructs (See table 2, and online supplementary appendix F1,F5).
Overall, self-identified bisexual people had higher odds of current cigarette smoking than lesbians/gays, OR: 1.25 (95% CI 1.15 to 1.37), with 1.30 (95% CI 1.17 to 1.44) for women and 1.11 (95% CI 0.97 to 1.26) among men. Bisexuals were OR: 2.18 (95% CI 1.84 to 2.59) times more likely to report smoking than heterosexuals, with the pooled estimate for women 3.02 (95% CI 2.59 to 3.51) greater than that among men 1.61 (95% CI 1.32 to 1.97) (details in table 3, and online supplementary appendix F2,F6).
In models comparing bisexuals to lesbians/gays, no covariates were identified as moderators; however, in those comparing bisexuals with heterosexuals, female gender positively and median year of the surveys negatively moderated the association between bisexuality and current smoking (online supplementary appendix F3,F4), indicating that other than gender/sex, other variables did not explain heterogeneities across the included studies.
The funnel plots of the outcome among bisexuals versus lesbians/gays (Egger test p value=0.854) and bisexuals versus heterosexuals (Egger test p value=0.793) were symmetrical (online supplementary appendix F7).
The current systematic review showed considerable variations in smoking behaviours across sexual orientation populations, with higher prevalence among bisexuals, regardless of how sexual orientation is measured. Bisexual individuals were at greater risk of smoking cigarettes—for all three measures—than their lesbian/gay and heterosexual counterparts. Of particular interest, cigarette smoking was found to be particularly high among bisexual women, highlighting the intersection of sexual orientation and gender contributing to differential risk of cigarette smoking; for example, self-identified bisexuals were 2.18 times more likely to be current cigarette smokers than heterosexuals in the gender pooled model; however, subgroup analysis by gender showed that women (pooled OR 3.02) were at greater risk than men (pooled OR 1.61). Compared with the lesbian/gay population, bisexual women were also at significantly greater risk of cigarette smoking but such comparisons were not significant among men. These findings are consistent with previous research examining other health outcomes7 8 suggesting the poorest outcomes among bisexual people compared with all other sexual orientation groups and particularly among bisexual women.
Although the dearth of evidence on bisexual women’s health and health behaviours makes it difficult to explain the reasons for the elevated rates of cigarette smoking among bisexual women, other literature suggests that the disproportionate burden of smoking may be due to a coping behaviour arising from exposure to everyday stressors and psychological distress associated with gender-based and sexuality-based social rejection, discrimination and stigmatization5 30 31—as can be seen within the Minority Stress Model31 32—and lifetime victimisation experiences (eg, partner and non-partner violence)33 that are more prevalent among bisexuals, particularly among women, than other sexual identity groups. In their qualitative research, Ross et al,34 delineated that health and behavioural disparities among bisexual people can occur at multiple levels within a sociological framework: important health determinants were identified at the social structure level (eg, experiences of biphobia and monosexism), interpersonal level (eg, partner and intimate relationships and family members) and individual level (eg, struggles with identity and self-acceptance). Bisexual-specific discrimination experiences such as biphobia (ie, negativity and prejudice) and monosexim (ie, structural dismissal or disallowal of bisexuality) can help explain disparities observed in health outcomes and behaviours among these individuals.34–37 Previous research has linked bisexual-specific stressors with an elevated risk of substance use,30 prescription opioid misuse38 and poor mental health7 among bisexual individuals.
Supported by an intersectionality framework,39 in which intersecting social positions may lead to unique experiences of inequalities, and consistent with previous reviews,7 8 the findings of our systematic review also suggest the intersection (or interaction) of sexual orientation and gender/sex contributes to the elevated vulnerabilities. Specifically, we found that cigarette smoking among bisexual people was amplified in intersection with female sex/gender resulting in a higher burden among bisexual women relative to other sexual orientation and gender/sex subgroups. Such a finding is supported by existing evidence highlighting the non-homogeneity of the experiences of sexual minority individuals in intersection with other socially significant identities.34 40 41 Smoking prevention programmes among bisexual individuals should, therefore, consider the implications of smoking as a behaviour that can be attenuated across other socially important positions (eg, gender/sex as observed in our study), suggestive of exploring the role of sexual orientation-based and gender-based intermediating factors in explaining the elevated burden of cigarette smoking among bisexual women.6 42 43
Tobacco control programmes should attend to the substantial variation in smoking behaviours across sexual minority groups in order to reduce smoking inequalities and their associated consequences among bisexuals.9 Also, given the variations across genders, and as the literature emphasises,42 research on targeted tobacco control programmes addressing cigarette smoking among sexual minorities should consider the role of gender—and possibly other social positions such as race/ethnicity—that intersect to produce the elevated likelihood of behavioural inequalities among sexual minorities.6 10 44 This is important to be taken into account as behaviour change patterns can be different across sexual minority groups based on other socially important identities they hold. For example, Cochran and Mays45 noted that the underlying reasons for the reduced smoking among sexual minority men observed in their research may not have the same effect on sexual minority women ‘for reasons we do not yet understand’. We also agree with Gruskin et al 46 who have previously emphasised the importance of involving and working closely with minority groups to reduce tobacco-related morbidity and mortality among sexual minority people and gain the full benefits from current multilevel best practice models of tobacco control. Prior reviews have identified promising evidence for the efficacy of the minority-specific group cessation interventions15 47; however, additional resources are needed to develop and implement evidence-based and policy interventions to promote health equity of sexual minorities, including bisexuals.47 However, as McQuoid et al 48 indicated in their qualitative research, efforts are better to identify the unique stressors of cigarette smoking specific to bisexuals if they aim to reduce disparities in smoking among these individuals.
First, we only included data from published literature in our systematic review and quantitative meta-analysis, which results in the potential for publication bias, particularly given that the past 30 days smoking outcome had a significant Egger test, specifically due to the female data. However, neither of the other two smoking outcomes had a significant Egger test, and there is no reason to expect that unpublished data would be particularly important for one smoking outcome over another. Second, a large proportion of the data included in the meta-analysis came from the USA, with limited data being available from contexts outside the US. This may indicate that our findings can be largely generalised to the US population. The results of this review should be updated in the future once enough data from other contexts are available. Third, we reported crude (unadjusted) estimates for the pooled prevalences and pooled odds ratios rather than adjusted estimates. This was necessary given that individual studies either did not report adjusted estimates or in case when they were reported: (a) different sets of covariates were adjusted for in each study, making it difficult for us to draw clear conclusions from the adjusted estimates or (B) some papers provided adjustments for covariates such as socioeconomic status that could potentially be considered to be a result of sexual orientation status, given strong evidence that bisexual people have poorer economic outcomes than other sexual orientation groups, rather than being confounding factors for the association between bisexuality and cigarette smoking. Future studies should make adjusted estimates more explicit such that true confounders are controlled for in multivariable regression models. Finally, our prevalence estimates relied on the common approaches measuring sexual orientation status in health surveys; however, we believe that future studies investigating sexual minorities’ health disparities and health behaviours should account for missing sexual orientation responses, according to what Stepleman et al 49 suggest.
The results of this first systematic review and meta-analysis of cigarette smoking across sexual orientation groups reveal that bisexuals, and in particular, bisexual women, are at greater risk for cigarette smoking compared with both lesbians/gays and heterosexuals. It is therefore important for tobacco control programmes to consider sexual orientation-specific risk for smoking, and additional research is needed to better understand the patterns and aetiology of smoking behaviours, as well as to develop sexual orientation-specific intervention strategies45 that attend to these risk factors. Given that bisexual women also have higher rates of poor mental health,7 interventions are recommended to be multifaceted and address women’s mental health concerns together with smoking behaviour. Our findings suggest that greater attempts should be made to identify the pathways through which individual, interpersonal, community, social and environmental factors contribute to the elevated smoking rates among bisexual people and thus how prevention and smoking cessation programmes can address these pathways.50 Gender-based approaches are also needed in tobacco control programmes to reduce the high prevalence of cigarette smoking—and its related consequences—among bisexual women. While we focused only on cigarette smoking in the current review, future research should also explore inequalities of other forms of tobacco use (eg, electronic nicotine products, cigars) as evidence indicates that sexual and gender identity intersectionalities might have differing risks by tobacco product type.51
What this paper adds
Bisexual people experience greater disparity in cigarette smoking than lesbian/gay and heterosexual individuals.
Bisexuality and gender intersected such that bisexual women had the highest prevalence of cigarette smoking.
Future research on disparities in cigarette smoking should ensure to disaggregate data for bisexual people.
Patient consent for publication
Contributors LR conceptualised and oversaw all aspects of the study. Under the supervision of LR, MS and BA screened and reviewed all the identified records, assessed the eligibility criteria and extracted data from all of the included studies. MS conducted the statistical analysis and prepared the first draft of the manuscript. LR and TS contributed to the interpretation of the findings. All authors have reviewed and approved the final version of the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.