Scale-up of antiretroviral therapy (ART) has led to increased survival of people living with HIV (PLWH). As result, life expectancy of PLWH in high income countries approaches that of the population not living with HIV [1]. A study using data from multiple clinical cohorts in Latin America and the Caribbean including over 30,000 participants estimated that life expectancy at age 20 years among PLWH increased from 31.0 years in 2003–2008 to 69.5 years in 2013–2017 [2]. Moreover, PLWH who initiated ART with CD4 counts greater than 200 cells/mm³ had similar life expectancy as the population not living with HIV [2], highlighting the continued need for early HIV diagnosis and ART therapy initiation to fully realize the life expectancy gains.

Nonetheless, disparities in mortality rates and survival among PLWH stratified by demographic characteristics such as sex assigned at birth, gender and sexual orientation exist. Several observational studies found that cisgender women had lower mortality rates and longer life expectancy than cisgender men [2,3], although mortality rates among cisgender women vary largely across several regions [4]. A recent Dutch study reported that transgender women had higher mortality risk than cisgender women (standardized mortality ratio 2.8, 95% confidence interval [CI] 2.5–3.1), and cisgender men (standardized mortality ratio SMR 1.8, 95% CI 1.6–2.0) [5]. Additionally, heterosexual cisgender men have lower life expectancy than cisgender women and gay, bisexual and other cisgender men who have sex with men (MSM) in Latin America [2].

These disparities may possibly be explained by social determinants of health (i.e., income level/poverty, employment, education, social support, discrimination, and stigma) interdependently with late presentation to care. Social determinants of health are the root of health inequalities, hindering access to health care and impacting all steps of the HIV care continuum, including HIV testing, early diagnosis, ART initiation, virologic suppression, and retention to care [6–9]. Moreover, the burden of social determinants varies across gender and sexual orientation groups [10–12], with sexual and gender minorities being additionally affected by minority stress (i.e., stigma, prejudice and discrimination against their identities), thus increasing their risk for negative health outcomes [13].

Brazil has made steady progress towards achieving UNAIDS 95-95-95 targets by 2030 [14], and in 2020, 88% of PLWH were diagnosed, 71% were receiving ART, and 63% were virally suppressed [15]. However, late diagnosis prevails and in 2016, 44% of PLWH in Brazil initiated ART with CD4 counts below 350 cells/mm3[6]. Furthermore, two Brazilian studies have shown disparities in late presentation (i.e., starting ART with CD4 < 350 cells/mm3 or with AIDS defining illness) by sex, gender, and sexual orientation. Rodrigues et al. found that, compared to cisgender women, MSM and heterosexual cisgender men had higher odds for late presentation (Rio de Janeiro, Brazil, 2005–2018) [8], while Pacheco et al. found that compared to MSM, cisgender women, heterosexual cisgender men, and men with unknown sexual orientation had higher odds for late presentation (Goiania, Brazil, 2009–2012) [16].

In this study, we estimated mortality rates and survival curves among PLWH over the past 20 years and investigated possible disparities in the hazard of death by gender and sexual orientation while controlling for confounders.

In this study, we found significant disparities in mortality among PLWH according to gender and sexual orientation, with Munk and MSWo having higher hazards of death than MSM. Munk had 3-fold higher hazard of death than MSM after controlling for sociodemographic and clinical confounders, suggesting that other unmeasured variables, such as socioeconomic status (i.e., poverty, employment, housing) and stigma might play a role on the association between sexual orientation and mortality among PLWH. Munk although accounting for less than 5% of our study population represented 11% of the deaths, and had the highest mortality rate, which was 4.5 times higher than MSM. Remarkably, one third of Munk died after cohort entry and their poor survival experience is most pronounced during the first year of follow-up. Munk were older at cohort entry, had the highest proportion of non-tuberculosis OIs and AIDS malignancies diagnosis at enrollment and during follow-up; conversely, they had the lowest proportion of ART use. Though we do not know which sexual orientation would be most applicable to the men in the Munk category, we hypothesize that this group was heterogenous, consisting of a combination of MSM and MSWo. For instance, the high frequency of AIDS malignancies observed in Munk (even higher than MSM) suggests that part of the group consisted of MSM, since Kaposi Sarcoma (historically the most common AIDS malignancy in our cohort) [21,22] is much more common among MSM than other groups due to human herpes virus-8 sexual transmission [23]. Conversely, the Munk group was older and less educated, characteristics that are more similar to the MSWo participants. The unknown sexual orientation status might constitute missing information resulting from severe health conditions at cohort entry, with consequent rapid progression to early death, thus hindering the capture of a comprehensive medical and social history. Alternatively or concomitantly, it might result from concealment of sexual orientation to health care professionals, possibly driven by stigma and discrimination. The rationales to support both hypotheses are detailed below.

As early mortality was most frequent among Munk, this may partially explain the missing information for sexual orientation. The high proportion of Munk participants with missing CD4 at enrollment, missing CD4 nadir, no ART use prior to end of follow-up supports the hypothesis that for at least a fraction of these participants, advanced disease and early mortality soon after cohort entry might have prevented us to collect the sexual orientation information as well as laboratory data. The second hypothesis, concealment of sexual orientation, might have been driven by stigma and discrimination. Sexual concealment may act as a functional adaptation to hostile structural contexts; in societies where heteronormativity norms prevail (including Brazil), sexual and gender minorities are often victim of stigma, discrimination and violence [24]. In Brazil, a multicity population-based survey of MSM conducted in 2016, found that 65% of the respondents had been discriminated against because of their sexual orientation [25]. Interestingly, a prior survey, conducted in 2009 [26], with a similar rationale to that of the 2016 study, had reported a much lower frequency of discrimination based on sexual orientation (28%), pointing to an increase in societal conservatism in most recent years, coupled with increased awareness of what constitutes discriminatory acts [27]. Fear of disclosure of sexual orientation delays access to care [28] as suggested in a previous study with PLWH in Brazil that showed cisgender men with unknown sexual orientation had 2.5 times the odds for late presentation relative to MSM [16].

In our study, we also found that MSWo had a higher hazard of death compared to MSM. This finding was expected as life expectancy among PLWH is lower in heterosexual men compared to women and MSM in Latin America [2]. Moreover, in a previous analysis of our cohort (2000–2011), hazard of AIDS-related death among heterosexual men was twice higher than that of women [18]. Studies in Brazil have shown that late presentation is frequent among heterosexual men. A study conducted in Salvador in 2010–2011 showed that 79% of heterosexual men had first CD4 count < 350 cell/mm3[29]. Also, heterosexual men had higher odds for late presentation than MSM and women in Goiania (2009–2012) [16]. On the other hand, a recent study with a sample of young conscripts in Brazil, mostly heterosexual men, has found that 85% of the participants had never been tested for HIV and that anticipated HIV stigma was associated with delay in HIV testing [30]. Altogether, these findings can be partially explained by cultural and gender norms that prevail in Brazil, hegemonic masculinity is often associated with men's risky behavior (feelings of invulnerability), which include alcohol and drug use, pleasure seeking, lack of self-care, and denial of health information and services [31–34]. Nonetheless, heterosexual men are often left out of HIV prevention and care policies and programs [31,34].

We found that cisgender women had an adjusted hazard of death similar to that of MSM. This finding is somewhat different from a study with PLWH in Latin America and the Caribbean that found lower mortality rate and higher life expectancy among cisgender women than MSM [2], and different from a previous study of AIDS and non-AIDS mortality in our cohort (2000–2011) [18]. Compared to MSM, cisgender women in our study were more likely to be Black (25% of the women, the highest proportion across the groups), less educated (60% had not completed primary school) and had low CD4 counts (nadir and at enrollment). In addition, they were less likely to use ART, had more OIs (tuberculosis and non-tuberculosis) and were more likely to have a hospital admission during follow-up than MSM. These factors probably contributed to cisgender women's higher mortality rates. Results from the Cox regression models suggest that adjustment for these factors resulted in a hazard of death that was not different from MSM.

Transgender women had the second lowest mortality rate in our study (only after MSM). At first, this finding may be surprising since Brazil leads the global statistics for numbers of death and violence against transgender women [35], and the number of homicides against sexual and gender minorities has been increasing steadily over the years [36]. Transgender women are disproportionately affected by HIV, with an estimated 30% prevalence of HIV infection in Rio de Janeiro in 2016 [37]. Moreover, transgender women face barriers to HIV testing, care and treatment, and stigma and discrimination play an essential role in reducing the access of transgender women to medical visits and HIV testing services [38]. To situate our results within the broader context of high mortality risk among transgender women living with HIV [5] we need to highlight some critical findings. First, 72% of transgender women included in our study enrolled after 2014 as a result of a series of interventions developed and implemented to reach and engage transgender women in HIV prevention and care in our clinic [39,40] (i.e., gender neutral space, transgender staff, community engagement, partnership with stakeholders and non-governmental organizations). A large population-based study aiming to recruit transgender women in Rio de Janeiro was conducted in 2015–2016 and those participants were offered routine HIV prevention, HIV care and endocrinological cross-gender care [37,39]. As a consequence, compared to other groups, transgender women included in our study were younger, had the highest CD4 counts (at enrollment and nadir) and the lowest proportion of non-tuberculosis OI at enrollment. Altogether, our findings suggest that the low mortality observed among transgender women enrolled in our cohort may be partly explained by their relative better health when compared to the other groups. Moreover, our findings suggest that active interventions to reach, test and engage transgender women in care may counteract the deleterious structural factors driving mortality in transgender women living with HIV.

Finally, our results underscore the association between social determinants of health and late presentation to HIV care with the risk of death among PLWH. Pardo/mixed and Black race and low education were strongly associated with a higher death hazard. Moreover, proxies of late presentation (i.e., lower CD4 counts and AIDS defining illnesses) were associated with an increased risk of death, while ART use reduced the risk of death. In addition, a worrisome upward trajectory of the hazard of death in most recent years (between 2015 and 2018) was observed, though this finding was not completely unexpected. A previous study from our cohort (2004–2015) has shown that very early mortality (i.e., deaths occurring up to 90 days after enrollment) remained high over the years, unlike deaths occurring between 90 and 365 days after enrollment [41]. Likewise, in the present study, survival curves, particularly for Munk, show a steep drop in survival probabilities in early follow-up. The spread of the HIV epidemic among vulnerable populations, the persistence of late presentation, and the increased of poverty and social inequalities in Brazil over the past decade [42] likely contributed to this finding.

Our study has limitations. First, 8.4% of our study population were deemed lost to follow-up; however, we periodically update vital status for our cohort participants through linkage with the State of Rio de Janeiro mortality database, allowing us to assume with reasonable confidence that participants classified as lost to follow-up were indeed alive as of December 31, 2018 (further corroborated by the sensitivity analysis shown in the Supplementary Material). Second, we did not differentiate causes of death, preventing us to assess gender and sexual orientation disparities related to specific causes of death (AIDS, non-AIDS, violent deaths). Third, we did not have information on potential confounders, mainly socioeconomic status (i.e., income, employment, housing), behavior factors (alcohol, smoking) and stigma. Fourth, most of the transgender women were recently enrolled in the cohort (from 2014 onwards) and future studies will be needed to better assess mortality trends in upcoming years, particularly for this group. Finally, our results are likely not representative of the entire population of PLWH in Brazil given that our participants represent a convenience sample. In addition, we found higher overall mortality rates (period 2000–2018, 20.3/1000 PY) than a recent study including PLWH in Latin America and the Caribbean (the Caribbean, Central and South America network for HIV epidemiology [CCASAnet], period 2003–2017, 17.9/1000 PY) [2] and a previous study from our clinical cohort (INI-Fiocruz, period 2000–2011, 10.2/1000 PY) [18]. It is very likely that differences in the studies’ specific eligibility criteria explain the different rates. Specifically, the current study population includes individuals who did not use ART and whose follow up was minimal. In contrast, the CCASAnet study restricted the study population to ART users [2]. Likewise, the study population of the prior INI-Fiocruz analysis excluded participants with less than 60 days of follow-up and also excluded IDU, Munk, and individuals with no CD4 count [18]. These criteria potentially excluded population groups with higher hazards of death from the two aforementioned studies.

Our results highlight important disparities in mortality risk by gender and sexual orientation among PLWH, and that Munk and MSWo have increased risk of death compared to MSM. Future studies should aim to better characterize and understand this population of men with concealed sexual orientation and their outcomes, including in-depth qualitative research, to help tailor policies and interventions to improve outcomes in this population. Finally, we did not observe an increased risk of death among transgender women, which may be explained by specific strategies (gender neutral space, transgender staff, community engagement, partnership with stakeholders and non-governmental organizations) implemented in our service to reach, engage, retain, and support transgender women.