AI indicates National American Indian Adolescent Health Survey; BC, British Columbia Adolescent Health Survey; MAHS, Minnesota Adolescent Health Survey; M-H, Mantel-Haenszel; MN, Minnesota Student Survey; NR, not reported; SE, Seattle Adolescent Health Survey; YRBS, Youth Risk Behavior Surveillance System.
eBox. Definitions of Sexual Orientation
eFigure 1. Funnel Plot of Attempted Suicide in Sexual Minority Adolescents Compared to Heterosexual Peers
eFigure 2. Funnel Plot of Attempted Suicide in Each Group of Sexual Minority Adolescents Compared to Heterosexual Peers
eFigure 3. Forest Plot of Attempted Suicide in Each Group of Sexual Minority Adolescents Compared With Heterosexual Peers
eFigure 4. Forest Plot of Attempted Suicide in Sexual Minority Adolescents Compared to Heterosexual Peers According to the Country of Sampling
eFigure 5. Forest Plot of Attempted Suicide in Sexual Minority Adolescents Compared to Heterosexual Peers According to the Grouping of Country of Sampling
eFigure 6. Forest Plot of Attempted Suicide in Sexual Minority Adolescents Compared to Heterosexual Peers According to the Grouping of Year of Sampling
Customize your JAMA Network experience by selecting one or more topics from the list below.
di Giacomo E, Krausz M, Colmegna F, Aspesi F, Clerici M. Estimating the Risk of Attempted Suicide Among Sexual Minority Youths: A Systematic Review and Meta-analysis. JAMA Pediatr. 2018;172(12):1145–1152. doi:10.1001/jamapediatrics.2018.2731
Are sexual minority youths at higher risk of attempting suicide?
In this systematic review and meta-analysis of nearly 2.5 million adolescents, sexual minority youths were found to have greater risk of life-threatening behaviors compared with their heterosexual peers. Transgender youths were the most affected followed by bisexual and homosexual teens.
Sexual minority adolescents may have a greater risk of life-threatening behaviors compared with heterosexual peers.
Suicide is the second-leading cause of death among adolescents. Sexual minority individuals are at a higher risk of suicide and attempted suicide, but a precise and systematic evaluation of this risk among sexual minority youths has not been documented to our knowledge.
To examine the risk of attempted suicide among sexual minority adolescents, differentiating for each sexual minority group.
Electronic databases (PubMed, Embase, and PsycINFO) were searched for articles published through April 30, 2017, with the following search terms: heterosexual, homosexual, bisexual, transgender, adolescents, teens, and attempted suicide.
Studies that reported attempted suicide in sexual minority adolescents compared with heterosexual peers were included. Thirty-five studies satisfied criteria for inclusion of 764 records identified.
Data Extraction and Synthesis
Pooled analyses were based on odds ratios (ORs), with relevant 95% CIs, weighting each study with inverse variance models with random effects. Risk of publication bias and analysis of heterogeneity through univariable and multivariable meta-regressions were also rated.
Main Outcomes and Measures
The evaluation of increased odds of attempted suicide among sexual minority youths compared with heterosexual peers.
Thirty-five studies reported in 22 articles that involved a total of 2 378 987 heterosexual and 113 468 sexual minority adolescents (age range, 12-20 years) were included in the analysis. Sexual minority youths were generally at higher risk of attempted suicide (OR, 3.50; 95% CI, 2.98-4.12; c2 = 3074.01; P < .001; I2 = 99%). If estimated in each sexual minority group, the OR was 3.71 in the homosexual group (95% CI, 3.15-4.37; c2 = 825.20; P < .001; I2 = 97%) and 4.87 in the bisexual group (95% CI, 4.76-4.98; c2 = 980.02; P < .001; I2 = 98%); transgender youths were described as an individual group in only 1 study, which reported an OR of 5.87 (95% CI, 3.51-9.82). Meta-regressions weighted for the study weight highlighted that the presence of young participants (12 years old) was associated with heterogeneity in the bisexual group, whereas the year of sampling was associated with heterogeneity in the whole group when combined with other covariates.
Conclusions and Relevance
Our findings suggest that youths with nonheterosexual identity have a significantly higher risk of life-threatening behavior compared with their heterosexual peers. Public awareness is important, and a careful evaluation of supportive strategies (eg, support programs, counseling, and destigmatizing efforts) should be part of education and public health planning.
Suicide is the second-leading cause of death during adolescence.1,2 The most recognized risk factors for suicide and attempted suicide are previous attempts, psychiatric illness, bullying, and childhood abuse and trauma.3-6 Among these risk factors, sexuality and related issues have been investigated, particularly in connection with abuse and gender identity.7-9 Gender identity is part of the self-identity that contributes to the full development of an adult human. Even if its roots are in childhood, gender orientation fully expresses itself during adolescence, especially because of the onset of sexual drive. Homosexual, bisexual, and transgender people are generally at higher risk of isolation, exposure to violence, and stigmatization, both self-inflicted and from peers or family members.10-13 The calculation of the real extent of suicide and attempted suicide during adolescence often stems from research led by lesbian, gay, bisexual, and transgender (LGBT) communities or focused on bullying within the school context.14-17 Neither of these situations is appropriate for the precise quantification of suicide among sexual minority youths. Studies within LGBT communities may have incorrect estimations because they are composed of people with higher self-consciousness about their sexual orientation and represent a place of aggregation and support. On the other hand, focusing on bullying may increase quantification because of a biased selection for the implication of an explicit risk factor.
Despite the increasing evidence of a greater risk of attempted suicide among sexual minority adolescents,12,18-22 a systematic analysis of the extent of this phenomenon has not been performed to our knowledge. This study examined the risk of attempted suicide among sexual minority adolescents, differentiating for each sexual minority group, to identify preventive strategies.
This systematic review and meta-analysis was performed according to the Meta-analyses of Observational Studies in Epidemiology guidelines.23 Procedures and study inclusion criteria were defined a priori and registered in PROSPERO (an international prospective register of systematic reviews).
A systematic search for articles published in electronic databases (PubMed, Embase, and PsycINFO) through April 30, 2017, was performed with no language or time restrictions. Search phrases combined thesaurus and free-search indexing terms related to age, attempted suicide, self-threatening behavior, and sexual orientation, using combinations of the following search terms: heterosexual, homosexual, bisexual, transgender, adolescents, teens, and attempted suicide. We contacted corresponding authors of selected studies if additional information was required.
We included observational studies with a comparative estimation on rates of heterosexual, homosexual, bisexual, and transgender among high school students. We included studies that defined LGBT and attempted suicide according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria.4 We excluded studies without an appropriate comparison group, those that did not clearly specify gender orientation or adherence to criteria for attempted suicide, and case reports. To reduce the risk of misclassification errors, we included only high-quality studies, with unequivocal definitions. If data from the same sample were published in multiple works, we retained only the study with more exhaustive information. Finally, we included only the studies published in peer-reviewed journals, excluding conference abstracts and dissertations.
After 1973, the American Psychological Association no longer considered LGBT orientation as a mental disorder and has since worked intensively to eradicate the stigma historically associated with homosexual orientation. Therefore, a precise and impartial definition of sexual orientation was derived from several publicly recognized, free, online dictionaries and studies24-27 performed on LGBT orientation after its declassification from mental illness (eBox in the Supplement).
Three authors (E.d.G., F.C., and F.A.) preliminarily reviewed titles and abstracts of traced articles. The initial screening was followed by the analysis of full texts to check compatibility regarding inclusion and exclusion criteria. Discordances were analyzed and disagreements were resolved by discussion among all the authors. When reported information was unclear or ambiguous or numerical data were not obtainable by percentages, the relevant corresponding author was contacted for clarification.
A standardized form was used to extract data, including information on year of publication, country, setting, characteristics of study participants (sample size, age, and percentages of men and women), sexual orientation, and life-threatening behavior. If raw numerical data were not reported, they were calculated by percentages, deriving crude odds ratios (ORs). Two authors (E.d.G. and F.C.) conducted data extraction independently; extraction sheets for each study were cross-checked for consistency, and any differences were resolved by discussion among the coauthors.
Meta-analysis of the overall comparison of attempted suicide rates among sexual minority youths and each sexual minority group compared with their heterosexual peers was performed, and pooled ORs with 95% CIs were generated using inverse variance models (DerSimonian-Laird28,29) with random effects. Results were summarized using conventional forest plots. Standard χ2 tests and the I2 statistic (ie, the percentage of variability in prevalence estimates attributable to heterogeneity rather than sampling error or chance, with values ≥75% indicating high heterogeneity) were used to assess between-study heterogeneity.30 To test for publication bias, we performed funnel plot analysis and the Egger test on all studies stratified by sexual orientation (homosexual, bisexual, transgender, or LGB). Thus, 4 separate Egger tests were performed. The Egger test quantifies bias captured in the funnel plot analysis with linear regression using the value of effect sizes and their precision (SE) and assumes that the quality of study conduct is independent of study size. If analyses showed a significant risk of publication bias, we would use the trim and fill method to estimate the number of missing studies and the adjusted effect size.31-34 Meta-regression analysis was performed to examine sources of between-study heterogeneity if of a high level (I2>75%) on a range of study prespecified characteristics (ie, sample size, age, and country).
All analyses were performed using R, version 3.2.3 (meta and metaphor packages; R Foundation for Statistical Computing). Statistical tests were 2-sided and used a significance threshold of P < .05.
Thirty-five1,7,12,18,20-22,35-48 studies reported in 24 articles that involved a total of 2 378 987 heterosexual and 113 468 sexual minority adolescents (age range, 12-20 years) were included in the analysis (Figure 1). The studies were conducted in 10 countries (Iceland, United States, Ireland, Canada, Switzerland, Norway, New Zealand, China, Taiwan, and Korea). Ten of the articles were published between 2005 and 2010, 13 in the 1990s, 4 after 2010, 5 between 2000 and 2005, and 1 in 1986, and 2 do not report timing of sampling. Eleven studies enrolled participants between 2005 and 2010, 14, in the 1990s; 4, after 2010; 5, between 2000 and 2005; and 1, in 1986. Most of the studies had sample weights of less than 5%, whereas 5 studies had sample weights between 10% and 40%. All the study characteristics are summarized in the Table.
Sexual minority adolescents had an increased risk of attempted suicide compared with their heterosexual peers, with significant evidence of between-study heterogeneity (OR, 3.50; 95% CI, 2.98-4.12; c2 = 3074.01; P < .001; I2 = 99%). Sensitivity analysis, in which the meta-analysis was serially repeated after the exclusion of each study, showed that most studies affected the overall OR at an SD of 0.07 except for 3 studies: exclusion of the study by Lucassen et al18 from the total sample revealed a decrease in OR to 3.31 (95% CI, 2.82-3.90), exclusion of National American Indian Adolescent Health Survey of 1990 from the study by Saewyc et al42 revealed an increase in OR to 4.05 (95% CI, 3.58-4.59), and exclusion of the Minnesota Student Survey of 1992 in the study by Saewyc et al42 revealed an increase in OR to 3.61 (95% CI, 3.10-4.22). Exclusion of the study by Peter et al22 from the homosexual group revealed an increase in the homosexual OR to 4.22 and global OR to 4.55, exclusion of the British Columbia Adolescent Health Surveys of 1992 in the study by Saewyc et al42 from the bisexual group revealed an increase in bisexual OR to 5.13 and global OR to 4.66, and exclusion of the Youth Risk Behavior Surveillance System39 from the LGBT group revealed a decrease in the LGBT OR to 2.73 and an increase of the global OR to 4.4 (Figure 2 and eFigures 1 and 2 in the Supplement).
All the groups had increased odds of attempted suicide compared with their heterosexual peers. Transgender youths were the most highly weighted (OR, 5.87; 95% CI, 3.51-9.82). Homosexual adolescents had an OR of 3.71 for attempted suicide (95% CI, 3.15-4.37; c2 = 825.20; P < .001; I2 = 97%), and bisexual youths had an OR of 3.69 (95% CI, 2.96-4.61; c2 = 980.02; P < .001; I2 = 98%). Both these analyses showed high heterogeneity (eFigures 1 and 3 in the Supplement).
In-depth analyses based on the country of selection, grouping of countries (North America and Canada, New Zealand, Asia, and Northern Europe [Iceland and Norway], and Europe [Switzerland and Ireland]), and the year of sampling as grouping of years (1990, 2000 and 2005, 2005-2010, and after 2010) were conducted to determine whether those factors had an association with the increased odds of attempted suicide among sexual minorities. The highest odds for attempted suicide among sexual minority groups was in New Zealand (OR, 11.70; 95% CI, 3.09-44.25). This result might be influenced by the presence of the study by Clark et al,37 which is the only one focused on transgender youths, who have the most increased odds. The other study18 from New Zealand reported the highest odds compared with all the other studies included in this meta-analysis. If grouped on a continental basis, New Zealand had the highest odds as previously reported, whereas Europe (which included only 2 studies from Switzerland and Ireland) had the least increased odds (OR, 2.69; 95% CI, 1.25-5.81). Studies from Europe had no between-study heterogeneity (I2 = 0%). Studies with participant enrollment between 2005 and 2010 had the most increased odds (OR, 5.20; 95% CI, 4.32-6.26). All the subgroups had high between-study heterogeneity except for the studies from Europe (eFigures 4-6 in the Supplement).
There was no evidence of publication bias in studies focused on the homosexual groups (Egger test; SE, 1.57; 95% CI, −6.23 to −0.24; P = .09) and for studies focused on LGB youths (Egger test; SE, 2.81; 95% CI, −8.88 to 4.49; P = .50). Studies focused on bisexual youths found small evidence of publication bias (Egger test; SE, 2.40; 95% CI, −10.40 to −0.15; P = .04). When all the studies were taken into account, the evidence for publication bias increased (Egger test; SE, 1.97; 95% CI, −8.63 to −0.62; P = .02). The trim-and-fill method trimmed, with a random-effect model, 10 studies to the left of the mean (OR, 2.47; 95% CI, 1.93-3.16; Q = 14 603.96).
Univariable and multivariable meta-regressions weighted for the study weight were performed on the following variables that were potentially associated with heterogeneity: (1) the country where the study was conducted, (2) grouping of those countries for their continent, (3) the inclusion in the sample of the youngest participants (12 years), (4) the inclusion in the sample of the oldest participants (20 years), and (5) grouping of the year of patient recruitment. The mean age of the samples and the percentage of females within each group were not available or calculable in most of the studies. The sex percentage was present in only 9 of 35 studies and the mean age in 13 of 35 studies.
The presence of young participants (12 years old) was associated with the variance in the bisexual group both as a single covariate or in combination with other covariates (country where the study was conducted, year of sampling, and oldest participants). This association was highly significant, explaining 60% of the true variance as a single covariate and 60% to 72% when combined with the other covariates (P < .001). The year of sampling was associated with the variance in the whole group in combination with other covariates; however, the true variance was less than 20% (eTable in the Supplement).
This is the first extensive and comprehensive meta-analysis, to our knowledge, that attested the risk of attempted suicide among sexual minority adolescents compared with heterosexual peers with a specific, in-depth evaluation of risk in each sexual minority group (homosexual, bisexual, and transgender youths). A previous meta-analytic study was performed by Marshal et al,49 but it had a smaller total sample compared with our meta-analysis (122 955 vs 2 390 455 adolescents) and was not exclusively focused on attempted suicide but included studies that reported attempted suicide and suicide ideation. The process of selection of the studies with a rigorous definition of self-threatening behaviors and sexual orientation in adolescents with ages ranging from 12 to 20 years identified 35 studies that involved nearly 2 500 000 participants. Sexual minority youths had an increased risk of attempted suicide compared with heterosexual peers. Among each sexual minority group, transgender youths appeared to be the most highly weighted, followed by homosexual and bisexual adolescents. The analysis of the forest plot showed that all the singular estimations were coherent in direction, but the CIs often did not overlap. High heterogeneity was attested with each I2. Despite the between-study heterogeneity, the risk of publication bias was detected only if all the studies were analyzed together, whereas subgroup analysis within each sexual minority group demonstrated no risk of publication bias. In analysis of the source of heterogeneity with univariable and multivariable meta-regression, the specific weight of each sample was a significant contributor if all the studies were taken into consideration or in studies focused on bisexual youths. The country of selection was another important factor that influenced heterogeneity if combined with other factors, especially in the homosexual group in the presence of the youngest group in the sample. The LGB group was the only group influenced by the historical moment of sampling, particularly in the presence of the oldest groups in the samples. The influence of country and historical moment of selection might show the evolution in acceptance of the LGBT identify, especially when the influence is combined with the ages at the extremes of the range that we considered (young [12 years old] or almost young adults [20 years old]). These age limits seemed to be more influenced by self-acceptance and capability of disclosure about sexual orientation, and social acceptance of sexual minority orientation might affect these individuals more because of their varying self-esteem and consciousness.
Even if a single factor does not explain the heterogeneity, the observation of the characteristics of the studies indicate that they might have points of discrepancies among them. First, the year of sampling was different (from the 1990s to 2017), and it is possible to presume that gradual social acceptance might be influential on self-acceptance and, consequently, on self-harming behaviors. Second, different sample sizes appear to be the most explanatory factor. It is possible that greater samples would attest the rate of attempted suicide more precisely because of the greater number of participants or that smaller communities might contribute to self-acceptance as supporting or ostracizing. Third, the influence of including very young people or the oldest participants (both these categories are not present in all the studies) is highly plausible. Self-consciousness and self-acceptance are influenced by personal maturity, and the presence of ages at the extremes of the range could be a factor in the consciousness and disclosure of sexual orientation. Nevertheless, the high heterogeneity detected would not seem to be a limitation but instead an indication of the trend of self-harming in LGBT people through time and in its possible declination at different ages. The low detection of publication bias, not detectable when a meta-analysis included a specific sexual minority group alone, seems to support this interpretation.
The social involvement in comprehension and protection of adolescents passes through the analysis of factors that might influence their well-being. The present study aimed to examine the association of nonheterosexual orientation with the risk of attempting suicide. Our comprehensive meta-analysis revealed an increase in the risk of attempted suicide if youths’ sexual orientation was nonheterosexual, independently from sexual minority group (homosexual, bisexual, or transgender). Transgender youths had the highest risk, but all the subgroups were considerably burdened. Therefore, any of these sexual orientations should be considered an important risk agent. Our quantification was planned to avoid bias, such as bullying or LGBT-community peer support, which might overrate or underrate the risk of attempted suicide. Community and in-school surveys provide a reliable representation of adolescent population, producing a trustworthy quantification of self-threatening behaviors. The magnitude and importance of being a youth with sexual minority identity justify social attention and effort in supplying preventive strategies. Avoiding the possible influence of bullying or peer support to provide an unbiased evaluation of the rates of attempted suicide, our results are suitable to sustain possible explanations other than exclusion or oppression. One of the possible explanatory keys might involve personal resources of self-comprehension and acceptance. Most adolescents have to face the management of the activation of sexual impulsions. The understanding and acceptance of personal sexual orientation, if different from most youths, might require greater self-esteem and psychic resources. Nonheterosexual youths also lack comparison with peers, a mechanism on which most of adolescents rely, because of their different experiences compared with their heterosexual peers. Consequently, at least at the beginning of their self-disclosure, nonheterosexual adolescents might perceive themselves as different and isolated. Those sensations, often accompanied by discomfort and fear, could contribute to personal difficulties that might exhibit in self-threatening behavior if not adequately managed and supported.
One limitation of this study is that, although most of the studies involved large samples, all the data are self-reported. In addition, even though questionnaires were accurate, self-disclosure might have limited the findings.
Our systematic review and meta-analysis found that adolescents with nonheterosexual orientation had a significantly higher risk of life-threatening behavior compared with their heterosexual peers. Public awareness appears to be needed, and a careful evaluation of support strategies (eg, support programs, counseling, and destigmatizing efforts) should be part of education and public health planning.
Accepted for Publication: June 27, 2018.
Corresponding Author: Ester di Giacomo, MD, Department of Neuroscience, Doctorate School of the University of Milano-Bicocca, Via Cadore, 48-20090 Monza MB, Italy (firstname.lastname@example.org).
Published Online: October 8, 2018. doi:10.1001/jamapediatrics.2018.2731
Author Contributions: Dr di Giacomo had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: di Giacomo, Krausz, Clerici.
Acquisition, analysis, or interpretation of data: di Giacomo, Colmegna, Aspesi.
Drafting of the manuscript: di Giacomo.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: di Giacomo, Krausz, Aspesi, Clerici.
Administrative, technical, or material support: di Giacomo, Colmegna.
Supervision: di Giacomo, Colmegna, Clerici.
Conflict of Interest Disclosures: None reported.
Additional Contributions: We thank Angela Tanzi, EdD, for editing the manuscript. She was not compensated for her work.