Psychosocial Mechanisms Linking Intersecting Stigmas and Condomless Anal Sex among Transgender Women in India

Abstract

Transgender women in India experience a disproportionate HIV burden, but the pathways linking intersecting stigmas and condomless receptive anal sex (CAS) have not been elucidated. Using minority stress theory, and psychological mediation and intersectionality frameworks, we tested the mediating and synergistically interacting pathways linking these stigmas to HIV risk. Between November 2020 and March 2022, we conducted a prospective three-wave cohort study among 500 trans women in Mumbai and Chennai. We assessed stigmas (transgender identity, perceived HIV, sex work), five psychosocial variables (depression, anxiety, problematic alcohol use, internalized transprejudice, social support), and CAS with male nonregular partners. We used causal mediation analysis to examine indirect pathways between stigma and CAS. We also tested for synergistic effects. Baseline levels of all three stigmas were high. In Wave-2, we noted relatively high prevalence rates of anxiety (38%), internalized transprejudice (46%), low social support (46%), depression (15%), alcohol use (15%), and CAS (61%, Wave-2; 81%, Wave-3). In mediation analyses, we found that stigmas increased anxiety and depression while eroding protective social support, each of which subsequently increased CAS risk. Social support and anxiety showed the strongest and most consistent indirect effects across stigma types, whereas depression mediated associations primarily for transgender identity and HIV-related stigmas. We demonstrated strong synergistic interactions between specific stigma combinations and anxiety, depression, problematic alcohol use, and low social support. The findings underscore the need for multicomponent interventions that simultaneously address stigmas and their psychosocial consequences, including screening and treatment for depression and anxiety, while bolstering social support.

Introduction

HIV continues to disproportionately affect marginalized and stigmatized communities. Globally, transgender (trans) women bear one of the highest HIV burdens of any population group, with a global estimated HIV prevalence of 20% [1]. India has a concentrated HIV epidemic, with a disproportionate HIV burden faced by marginalized communities such as trans women, men who have sex with men (MSM), people who inject drugs and female sex workers, and an estimated 2.4 million people living with HIV [2]. HIV prevalence among trans women in India ranges between 3% and 8% [2–4], which is many-fold higher than the 0.2% estimated for the general adult population and higher than key populations such as MSM (3.3%) and female sex workers (1.8%) [2]. Condomless receptive anal sex (CAS) is the primary mode of HIV acquisition for trans women in India [5], with a 2014-15 government survey reporting CAS of 35-48% across partner types [3]. CAS is a modifiable risk behavior known to be shaped by psychosocial factors and stigma experiences, making it a critical outcome for understanding the mechanisms through which structural and social conditions translate into HIV risk.

Trans women in India deal with multilayered structural and psychosocial stressors such as pervasive physical and sexual violence, police harassment, and limited livelihood options, which interact to elevate the likelihood of CAS despite high levels of HIV knowledge [3, 5]. These stressors contribute to a syndemic of depression, substance use, and HIV in this population [6–8]. Accordingly, CAS among trans women in India should be understood as a behavioral manifestation of the syndemic rather than as lapses in individual motivation or awareness [5].

Multiple Stigmas Experienced by Trans Women

Stigma remains a fundamental structural driver of the HIV epidemic, posing significant barriers to HIV prevention behaviors, testing, and treatment [9–12]. Trans women in India experience multiple, intersecting forms of stigma related to their gender identity, engagement in sex work, and HIV status [13], each uniquely shaping their health experiences and outcomes [14].

Transgender identity stigma is a form of enacted stigma [15] that encompasses a variety of behavioral manifestations of deep-seated cisnormativity and transprejudice within Indian society, including societal rejection, discrimination, and violence directed at individuals whose gender identity differs from their assigned sex at birth. Specific manifestations include verbal harassment; physical and sexual violence; police brutality; and exclusion from family, healthcare, and employment opportunities [6, 16]. Empirical evidence from India demonstrates that transgender identity stigma is associated with depression, often through pathways involving reduced social support and impaired resilient coping [17]. Stigma related to transgender identity also reflects traditional gender hierarchies and misogyny within the cultural context of India.

Sex work stigma involves both structural aspects, such as criminalization of public solicitation and associated activities under India’s Immoral Traffic Prevention Act, 1956 (rather than criminalization of sex work itself), and interpersonal aspects, such as moral condemnation and social exclusion of sex workers. A nationwide study from India reported that 33% of trans women engage in sex work as their main occupation, with significant geographic variation ranging from less than 10% in Kolkata to 86% in Mumbai [3]. For trans women in India, entry into sex work often occurs due to lack of education, limited employment opportunities, and systemic discrimination, creating a cycle where economic vulnerability compounds their stigmatization [6]. Qualitative research from India has shown that sex work stigma undermines trans women’s self-efficacy to negotiate condom use and fosters psychological distress and substance use as coping mechanisms [16]. International evidence corroborates this link between internalized sex work stigma and increased depression and anxiety among trans women in sex work [8].

HIV-related stigma among trans women manifests as enacted discrimination, anticipated rejection, and internalized stigma associated with having HIV or perceived HIV risk [18]. Anticipated HIV stigma, which is the expectation of rejection or discrimination if one’s HIV seropositivity were to become known, serves as a significant barrier to regular HIV testing and timely engagement in care among marginalized populations [19, 20], including trans women in India [13, 21]. This stigma is particularly salient given the disproportionate HIV burden among trans women in India.

Theoretical Frameworks Guiding the Present Study

This study employs compatible theoretical frameworks to examine the complex relationships between intersecting stigmas, psychosocial problems, and HIV-related risk behaviors among trans women in India.

Minority Stress Theory and Psychological Mediation Framework: Pathways from Stigma to Psychological Distress and HIV Risk

Minority stress theory posits that sexual and gender minorities experience chronic stress due to their stigmatized social position, leading to adverse health outcomes through distinct pathways [22–24]. This framework distinguishes between distal stressors such as enacted stigma, discrimination, and violence and proximal stressors, including internalized stigma, identity concealment, and rejection sensitivity, that occur within the individual as psychological responses to minority status. Hatzenbuehler’s psychological mediation framework [25] extends minority stress theory by specifying the cognitive, emotional, and interpersonal mechanisms through which these stressors influence health behaviors, proposing that minority stressors operate through emotion dysregulation, social and interpersonal problems, and cognitive processes to increase risk for psychological distress (e.g., depression, anxiety) and maladaptive coping behaviors (e.g., substance use).

While these frameworks were not originally designed to explain HIV risk specifically, mounting empirical evidence demonstrates their utility in understanding HIV risk behaviors among sexual and gender minorities [19, 26, 27]. Research shows that heterosexism and identity concealment predict increased CAS among trans women and men who have sex with men, mediated through identity conflict [19]. Among gay and bisexual men, distal minority stress processes confer risk for proximal stressors and emotion dysregulation, which predict sexual compulsivity through anxiety and depression [26]. Sexual compulsivity and low self-esteem have also been significantly associated with CAS among trans women [27]. Taken together, these findings show that stigma generates psychological distress and maladaptive coping, which compromise decision-making and elevate HIV risk.

Framework Differences and Social Support Pathways

These frameworks differ in their treatment of social support: minority stress theory considers it a moderator that buffers individuals from the adverse effects of stressors, while the psychological mediation framework positions it as a mediator between stressors and outcomes. Empirical evidence shows that social support mediates relationships between stigmas and HIV risk behavior through multiple pathways. Indian studies have demonstrated that social support and resilient coping serve as protective mediators against stigma’s effects among trans women and MSM [17, 28]. Social support can directly reduce HIV risk [29] and reduce condomless sex in low-stigma environments [30], but stigma undermines these benefits by decreasing social support because minorities self-isolate to prevent discrimination [31]. Conversely, a lack of social support may increase internalized stigma and decrease condom use [32], while HIV self-stigma reduces social support and condom use intentions [33].

Multiple Stigma Effects

These frameworks have been applied to understand how multiple stigmas influence health outcomes among sexual and gender minorities in India. A South Indian study of MSM living with HIV found that sexual stigma predicted depression through internalized HIV stigma, while enacted HIV stigma affected internalized homonegativity but not depression directly [34]. Qualitative research among MSM and trans women living with HIV in India revealed that enacted stigma manifested as familial shame and healthcare discrimination, reducing support access and care engagement, while anticipated stigma created disclosure concerns [13, 35]. However, trans women communities also provide resilience resources offering emotional and tangible support [13]. These studies suggest that multiple stigmas affect health behaviors through psychological distress, with social support serving as protection.

Together, minority stress theory and psychological mediation frameworks inform our mediation hypothesis that multiple stigmas (transgender identity stigma, HIV-related stigma, and sex work stigma) will influence condomless anal sex indirectly through psychosocial mediators (depression, anxiety, problematic alcohol use, and internalized transprejudice) and social support while also examining the potential direct effects of multiple stigmas on HIV risk behavior.

Intersectionality: Examining Synergistic Interactions between Multiple Stigmas

The intersectionality framework provides a critical lens for understanding how multiple, interlocking systems of oppression create unique experiences that cannot be understood by examining single identities or social positions in isolation [36]. Rather than simple additive effects, intersectionality posits that the combination of multiple forms of stigma creates qualitatively different experiences of marginalization, with effects that are not reducible to the sum of their individual contributions [37]. For trans women experiencing intersecting stigmas related to gender identity, sex work, and HIV status, this framework predicts that the sum total of their adverse impacts on health would be greater than what otherwise would be expected on the basis of a simple summation of their individual adverse impacts on health.

Methodological Applications and Research Gaps

Despite its theoretical importance, applying intersectionality theory in quantitative health research has faced methodological challenges, with qualitative approaches predominating. However, recent advances enable intersectional interpretation through statistical interaction terms, latent variable modeling, and multilevel approaches [38, 39]. These methodological developments emphasize examining results based on additive versus multiplicative scales, as the former focus on absolute risk, which is more important from a public health perspective.

Indian studies examining multiple stigmas among sexual and gender minorities have predominantly relied on qualitative methods, focusing on intersectional stigma effects on HIV care engagement among MSM [35] and trans women living with HIV [13]. While one study among trans women demonstrated synergistic effects of violence victimization, alcohol use, and drug use on CAS [40], the formal testing of stigma interactions using additive interaction metrics such as RERI and attributable proportion has not previously been applied in this context. Furthermore, indian studies examining stigma and HIV-related outcomes in this population have relied on qualitative or cross-sectional designs, which can document associations but cannot establish temporal ordering or decompose mechanisms [5]. Causal mediation analysis, which allows rigorous decomposition of stigma effects into direct and indirect pathways through specific psychosocial mediators, also remains underutilized in transgender health research in India.

Together, intersectionality theory and related methodological advances inform our interaction (moderation) hypothesis that multiple stigmas (transgender identity stigma, HIV-related stigma, and sex work stigma) will interact synergistically to influence both psychosocial problems (depression, anxiety, problematic alcohol use, internalized transprejudice and social support) and CAS, such that their combined effects exceed the sum of their individual contributions. Understanding whether stigmas operate through specific psychosocial mechanisms – and whether their effects are amplified when stigmas co-occur – has direct clinical and programmatic relevance: if anxiety, depression, and eroded social support are the key pathways, these are actionable targets for mental health screening and community-based support programs that could simultaneously reduce HIV risk.

Objective and Hypotheses

The objective of this prospective cohort study was to test two complementary hypotheses.

Mediation Hypothesis

Guided by the Minority Stress and Psychological Mediation Frameworks, we hypothesized that stigmas would influence CAS through both direct and indirect pathways.

Interaction Hypothesis

Guided primarily by the intersectionality framework, we hypothesized that multiple stigmas interact synergistically to worsen psychosocial problems and increase CAS while simultaneously acting antagonistically to erode the protective effects of social support.

Methods

Study Design

Between November 2020 and March 2022, we conducted a prospective observational cohort study among trans women in India. A total of 630 participants were enrolled and followed across three waves of data collection, with each wave spaced approximately six months apart. Data were collected via a structured questionnaire administered in a hybrid format, incorporating both offline (face-to-face) and online (web-based) surveys to enhance reach and inclusivity [41, 42]. Given systematic differences in sociodemographic and stigma-related characteristics between offline and online participants, the current data analysis was restricted to the offline sample (n = 500).

Recruitment and Survey Administration

Participants were eligible if they self-identified as transgender women, were aged 18 years or older, had been sexually active in the past two months, and provided written (offline) or electronic (online) informed consent.

Offline recruitment targeted trans women already registered with community-based organizations (CBOs) providing HIV prevention services in Mumbai (The Humsafar Trust) and Chennai (Sahodaran). Peer recruiters affiliated with these CBOs identified and approached eligible participants. Informed consent was obtained in writing, and surveys were administered in person by trained peer research interviewers using digital tablets (SmartPads). Surveys were conducted in participants’ preferred language (Tamil, Hindi, Marathi, or English). The survey completion time was between 35 and 45 minutes.

For the online sample, peer recruiters disseminated study information and survey links via social media platforms (e.g., Facebook, WhatsApp) and dating apps used by trans women (e.g., Grindr, Blued). Potential participants accessed a secure web page, first completing an eligibility screener and, if eligible, then proceeding to the full self-administered survey in their chosen language. The online consent process paralleled the offline procedure. To prevent duplicate participation, online participants were required to enter a one-time password sent to their mobile number (for which they provided consent during study enrollment), each time they logged in to complete the self-administered survey.

After the first survey (offline or online), in accordance with participant preferences and to protect confidentiality, peer recruiters communicated with participants via phone and WhatsApp, sending reminder messages that did not disclose study details. During periods of COVID-19-related restrictions, interviewers adhered to safety protocols for face-to-face interviewing, including maintaining physical distancing, use of masks by both interviewer and participant, avoidance of physical contact, and frequent hand hygiene.

Ethics

The study protocol (Reference Nos. 43-12/2018, 47-12/2019, 47/1-12/2020) was approved by the Institutional Review Board of The Humsafar Trust, Mumbai. All procedures adhered to national guidelines for research with human participants, including the protection of confidentiality and sensitive data. All participants received an honorarium of INR 500 (~ USD 6) for each survey wave completed.

Measures

Sociodemographic Characteristics

Participants were asked to provide sociodemographic information including age in years; highest level of educational status achieved; sex work engagement in the past month; HIV status; and their gender identities, including indigenous transgender identities (e.g., thirunangai, hijra, kinnar).

Stigma Measures

Three types of stigmas were assessed. For mediation analyses, continuous stigma scores were used; for interaction analyses, each stigma score was dichotomized at the sample median (Table 1).

Table 1. Sociodemographic Characteristics and Key Variables.

Characteristic	Mean (SD) or N (%)
Wave-1 (N=500)
Sociodemographics
Age in Years	27.58 (5.67)
Highest Level of Education Completed
No schooling	31 (6.2%)
Completed primary (5th grade)	59 (11.8%)
Completed elementary (8th grade)	75 (15.0%)
Completed high school (10th grade)	124 (24.8%)
Completed higher secondary (12th grade)	100 (20.0%)
Completed college	97 (19.4%)
Completed diploma	14 (2.8%)
Educational Status
Low (12th grade or less)	389 (77.8%)
High (Diploma/College degree)	111 (22.2%)
Self-identification of gender
Hijra	109 (21.8%)
Kinnar	35 (7.0%)
Thirunangai	250 (50.0%)
Transgender woman (English term)	71 (14.2%)
Jogta/Jogti	31 (6.2%)
Woman	4 (0.8%)
Sex work engagement (past two months)
No	199 (39.8%)
Yes	301 (60.2%)
Self-reported HIV-positive status
No	459 (95.03%)
Yes	24 (4.97%)
Wave-1 Stigma Scores
Transgender Identity Stigma Score	15.27 (6.47)
Perceived HIV Stigma Score	6.06 (2.94)
Sex Work Stigma Score	14.48 (10.28)
Wave-2 (N=424)
Wave-2 Psychosocial problems (binary)
Depression	63 (14.9%)
Anxiety	161 (38.0%)
Problematic Alcohol Use	65 (15.3%)
Internalized Transprejudice	96 (46.2%)
Wave-2 Psychosocial problems (continuous)
Depression Score	6.39 (6.24)
Anxiety Score	2.29 (1.65)
Alcohol Score	1.73 (1.77)
Internalized Transprejudice Score	23.67 (7.76)
Social Support Score	8.17 (3.24)
Wave-2 Consistency in Condom Use with male nonregular partners (past 2 months)
Always/Every time	153 (38.9%)
More than half of the time	10 (2.5%)
About half of the time	210 (53.4%)
Less than half of the time	17 (4.3%)
Never	3 (0.8%)
Wave-3 Consistency in Condom Use with male nonregular partners (past 2 months)
Always/Every time	53 (19%)
More than half of the time	6 (2.1%)
About half of the time	165 (59%)
Less than half of the time	56 (19.9%)
Never	1 (0.4%)
Never had anal sex in the past 2 months	1 (0.4%)

Note. Totals may not sum to the sample size for each wave due to missing data

Transgender Identity Stigma

We administered the 9-item Transgender Identity Stigma Scale, validated in South India [43], to elicit study participants’ experiences of discrimination and rejection related to their transgender identities (e.g., “How often have you heard that transgender people are not normal? “; “How often has your family not accepted you because of your transgender identity?”). Responses ranged from 0 (“Never”) to 3 (“Many times”), with higher scores indicating greater stigma. The measure demonstrated good internal reliability (Cronbach’s alpha = 0.84).

Perceived HIV-Related Stigma

We administered the 3-item “concerns about public attitudes” subscale, drawn from the 12-item short form of the Berger’s HIV Stigma Scale [44, 45], to measure anticipated stigma relating to HIV status. A sample item is “Most people believe that a person who has HIV is dirty.” Responses ranged from 0 (“Strongly disagree”) to 3 (“Very strongly agree”). Cronbach’s alpha = 0.96.

Sex Work Stigma

We administered a 6-item scale adapted from the Internalized Sex Work Stigma Scale [46], which was itself an adaptation of the Internalized AIDS-Related Stigma Scale [47]. The scale assessed internalized stigma related to sex work (e.g., “I am ashamed that I engage in sex work”). Responses ranged from 1 (“Very strongly disagree”) to 5 (“Very strongly agree”). While this scale was originally designed to measure internalized sex work stigma, two of the 6 items (“It is difficult to tell people that I sell sex” and “I hide that I sell sex from friends and family”) were found to measure anticipated stigma rather than internalized stigma [46]. Hence, we use the broader term “sex work stigma” to reflect the multidimensional nature of stigma experiences captured by this measure. Cronbach’s alpha = 0.93.

Psychosocial Variables

Depression Symptom Severity

Depression symptoms over the past two weeks were measured using the 9-item Patient Health Questionnaire (PHQ-9) [48]. Responses ranged from 0 (“Not at all”) to 3 (“Nearly every day”), with a total score ranging from 0 to 27. Scores above or equal to 5 indicated a positive screen for major depressive disorder. Cronbach’s alpha = 0.78.

Anxiety Symptom Severity

Anxiety symptoms over the past two weeks were assessed using the 2-item Generalized Anxiety Disorder scale (GAD-2) [49]. Response options ranged from 1 (“Not at all”) to 4 (“Nearly every day”), with a total score ranging from 0 to 8. Scores above or equal to 3 classified participants as screening positive for generalized anxiety disorder. Cronbach’s alpha = 0.96.

Alcohol Use Disorder

Alcohol use was measured using the 3-item Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) questionnaire [50]. Responses ranged from 0 to 4, with higher values indicating greater use (total score range: 0 to 12). Problematic alcohol use was defined as scores above or equal to 4. Cronbach’s alpha = 0.56.

Internalized Transprejudice

We used an 8-item internalized transprejudice scale [22], which was adapted from the 26-item internalized transphobia scale [51]. Example items included “I resent my gender identity” and “My gender identity makes me feel like a freak”, each scored on a five-point Likert scale (1=“Strongly disagree” to 5=“Strongly agree”), with a total score ranging from 5 to 40. Cronbach’s alpha = 0.90.

Social Support

Social support was assessed using the 3-item Oslo scale [52], with items such as “How many people are so close to you that you can count on them if you have great personal problems?” The items were scored on a four- to five-point Likert-type scale ranging from 1 to 4 or 5. The total score ranged from 3 to 14, with higher scores indicating greater social support. Cronbach’s alpha = 0.88.

Condomless Receptive Anal Sex (CAS)

CAS with male (nonregular) partners over the preceding two months was assessed using the question “How often have you used condoms when you had anal sex with nonregular male partners in the past 2 months?” [53]. Responses were scored on a five-point Likert scale, ranging from “Always/Every time” to “Never”. CAS was operationalized as a binary variable, with any response other than “Always/Every time” coded as inconsistent condom use (CAS=1) and consistent condom use coded as 0.

Analytic Framework

As described earlier, our analytic framework integrates minority stress and psychological mediation frameworks to test the mediation hypothesis and an intersectionality framework to test our interaction hypothesis. Diagram 1 presents the mediation model, in which the three stigma variables measured in Wave-1 (transgender identity stigma, perceived HIV stigma, and sex work stigma) serve as exposures (predictors), CAS in Wave-3 serves as the outcome, and the Wave-2 psychosocial variables – depression, anxiety, problematic alcohol use, internalized transprejudice, and social support – serve as mediators. For the interaction analysis, we assessed two-way interactions between stigmas (measured at Wave-1) in their impacts on psychosocial problems (measured at Wave-2) and on CAS (measured at both Wave-2 and Wave-3) in an effort to distinguish between proximal (immediate) and distal (persistent) effects of stigma combinations, in addition to establishing temporality.

Data Analysis

We used descriptive statistics to characterize the sample and conducted bivariate analyses (t-tests and chi-square tests) to examine unadjusted associations with CAS. The analytic sample (n = 500) included all offline participants regardless of HIV status since CAS represents HIV transmission or acquisition risk behavior, and perceived HIV stigma affects the entire trans women population.

Causal mediation analysis

We conducted causal mediation analysis [54, 55] to investigate the direct, indirect, and total effects of stigma exposures (measured in Wave-1) on CAS with nonregular male partners (measured in Wave-3), with psychosocial mediators measured in Wave-2. This analysis was performed using the mediate command in Stata Statistical Software (Release 18; StataCorp LLC, College Station, TX).

The analysis was based on a counterfactual (potential outcomes) framework, which allows for the decomposition of the total effect of an exposure into natural direct effects (NDE) and natural indirect effects (NIE) through a specific mediator. All outcome and mediator models were estimated using logistic regression, appropriate for their binary nature. For each stigma exposure, a separate mediation model was estimated for each of the following Wave-2 psychosocial mediators: depression, anxiety, problematic alcohol use, and internalized transprejudice.

To estimate mediation effects for the continuous stigma exposures, we contrasted the levels of stigma scores at the 25th percentile (low-level reference) vs. 75th percentile. Each causal mediation model was adjusted for two a priori selected covariates, age (in years) and education status (low: grade 12 or below; high: diploma or college education). These covariates were included in both the outcome model (for CAS) and the respective mediator models (for depression, anxiety, alcohol use, internalized transprejudice, and social support). All direct, indirect, and total effects from the mediation analysis were reported as odds ratios (ORs). Additionally, to understand the associations of all covariates within the underlying regression models, we fitted logistic regression models with robust standard errors for each outcome and mediator, with results presented as ORs.

Causal mediation analysis assumes that: (1) all relevant confounders have been measured and controlled; (2) the initial exposure is not affected by unmeasured factors that would also alter the outcome or mediators; and (3) the mediating variables are not influenced by unmeasured factors that also affect the outcome [55]. Statistical significance was set at p < .05.

Analysis of Interactions

We investigated potential two-way interactions between three types of stigmas in their associations with CAS and psychosocial outcomes. We modeled seven outcomes: Wave-2 and Wave-3 CAS with male nonregular partners; Wave-2 depression, anxiety, problematic alcohol use, and internalized transprejudice (all specified as binary outcomes); and Wave-2 social support (specified as a continuous outcome). The primary exposures were Wave-1 stigma scores: transgender identity stigma (TGS), perceived HIV stigma (HPS), and sex work stigma (SWS). To assess additive interactions between stigma domains, the continuous stigma scores were dichotomized at their medians. For each binary outcome, we then evaluated all two-way stigma interactions (TGS × SWS, TGS × HPS, and SWS × HPS), adjusting for age, education, and including the stigma main effects. From the models, we calculated the Relative Excess Risk due to Interaction (RERI), the Attributable Proportion due to interaction (AP), and the Synergy Index (SI).

An additive interaction was considered statistically significant if the 95% confidence interval for the RERI excluded 0; a positive RERI indicated synergy, and a negative RERI indicated antagonism. The additive scale is prioritized over the multiplicative scale because it is more relevant for assessing public health implications (Rothman et al., 1980; Knol & VanderWeele, 2012). While no standard guidelines exist for qualitatively reporting the magnitude of synergy, we adopted a descriptive heuristic for interpretation. For statistically significant positive RERIs, the strength was categorized based on the point estimate of the AP as follows: weak synergy (AP < 0.25), moderate synergy (AP 0.25–0.49), and strong synergy (AP ≥ 0.50). All interaction analyses were conducted in R (version 4.4.2; R Core Team, Vienna, Austria), and model estimates were derived using the emmeans package (version 1.10.6).

Handling of Missing Data

Participants were lost to follow-up with attrition rates of 15.2% at Wave-2 (n=424/500) and 40.8% at Wave-3 (n=296/500). The overall missingness across all variables (excluding structural missingness) was 14.4%. Missing data patterns were assessed using descriptive statistics and Little’s MCAR test, which was significant (p < 0.001), indicating that data were not missing completely at random (Little, 1988). Further analyses demonstrated that missingness in key variables was associated with observed participant characteristics, supporting the missing at random (MAR) assumption rather than missing not at random (MNAR). Multiple imputation by chained equations (20 datasets) was then performed in Stata-18 for mediation and in the mice package (Version 3.18.0) in R (Version 4.4.2) for interactions. Imputation models included all analysis variables, with convergence and imputation diagnostics conducted to ensure the validity of the imputed datasets (Azur et al., 2011). The imputation models addressed item-level missingness only among participants retained at each wave; data for participants who were entirely lost to follow-up were not imputed. Multiply imputed datasets were subsequently used to perform causal mediation analysis (CMA) and assess additive interactions. All reported results are based on analysis using multiply imputed datasets.

Results

Sociodemographic and Other Characteristics

Table 1 summarizes the sociodemographic, stigma, and psychosocial characteristics of participants in the two study waves. In Wave-1 (N=500), the mean age was 27.6 years (SD=5.7); most (77.8%) had secondary education or less, while 22.2% had a college degree or diploma. Gender self-identifications included Thirunangai (50.0%), hijra (21.8%), transgender woman (14.2%), kinnar (7.0%), jogta/jogti (6.2%), and woman (0.8%). Sixty percent (60.2%) reported recent engagement in sex work. Participants’ mean stigma scores were as follows: transgender identity stigma = 15.27 (SD = 6.47), perceived HIV stigma = 6.06 (SD = 2.94), and sex work stigma = 14.48 (SD = 10.28). Pearson correlation coefficients demonstrated that the different forms of stigma scores were correlated with each other: transgender identity stigma and perceived HIV stigma, r=0.24 (p<.001); transgender identity stigma and sex work stigma, r=0.18 (p<.001); and perceived HIV stigma and sex work stigma, r=0.12 (p=0.004). In Wave-2 (N=424), the prevalence rates of psychosocial problems were as follows: depression (14.9%), anxiety (38.0%), internalized transprejudice (46.2%), and problematic alcohol use (15.3%). The participants had a mean social support score of 8.17 (SD=3.24).

Bivariate analyses compared participants based on their reported CAS in Wave-3 (CAS group: n=229, 81.2%; No CAS group: n=53, 18.8%) (Table 2). Individuals who reported CAS in Wave-3 had significantly higher levels of stigma in Wave-1. This was reflected in higher mean scores for transgender identity, perceived HIV, and sex work stigma (all p<0.001) and a greater proportion of participants with high levels of stigma: transgender identity stigma (62% vs. 32%; χ² = 16.2, p<0.001), sex work stigma (70% vs. 13%; χ² = 57.2, p<0.001), and perceived HIV stigma (48% vs. 30%; χ² = 5.29, p=0.021). Furthermore, the group reporting CAS in Wave-3 also showed a significantly higher prevalence of psychosocial problems in Wave-2, including depression (26.0% vs. 6.0%; χ² = 9.43, p=0.002), anxiety (56.0% vs. 22.0%; χ² = 19.4, p<0.001), and internalized transprejudice (52% vs. 34%; χ² = 5.56, p=0.018), along with low social support (12% vs. 74%; χ² = 67.4, p<0.001).

Table 2. Bivariate Associations between Sociodemographic and Key Variables and Outcome (Wave-3 Condomless Anal Sex).

Outcome (Wave-3 Condomless Anal Sex)
Characteristic	Overall N = 282	No CAS N = 53	CAS N = 229	Statistic (t or χ2)2a
Sociodemographics	Mean (SD) or n (%)	Mean (SD) or n (%)	Mean (SD) or n (%)
Age in Years	27.95 (5.36)	29.02 (5.86)	27.70 (5.22)	1.50
Educational Status (low [≤12th grade] vs. High [diploma/college])	89 (32%)	8 (15%)	81 (35%)	8.19**
Sex Work Engagement (past month)	236 (83.6%)	14 (6%)	222 (94%)	57.9***
Wave-1 Stigma Scores
Transgender Identity Stigma Score	15.92 (6.10)	13.38 (6.17)	16.51 (5.94)	-3.35**
Perceived HIV Stigma Score	6.24 (2.39)	4.04 (3.68)	6.76 (1.60)	- 5.26***
Sex Work Stigma Score	17.20 (8.74)	7.36 (8.69)	19.47 (7.02)	- 9.46***
Wave-1 Binary Stigmas (Median-split) (Low vs. High)
High Transgender Identity Stigma	160 (57%)	17 (32%)	143 (62%)	16.2***
High Perceived HIV Stigma	125 (44%)	16 (30%)	109 (48%)	5.29*
High Sex Work Stigma	167 (59%)	7 (13%)	160 (70%)	57.2***
Wave-2 Psychosocial binary variables (No/Yes)
Depression	62 (22%)	3 (6%)	59 (26%)	9.43**
Anxiety	139 (50%)	11 (22%)	128 (56%)	19.4***
Problematic Alcohol Use	31 (11%)	5 (10%)	26 (11%)	0.090
Internalized Transprejudice	136 (49%)	17 (34%)	119 (52%)	5.56*
Low Social Support	174 (62.8%)	6 (12%)	168 (74%)	67.4***
Wave-2 Psychosocial Variable Scores (Continuous)
Depression Score	8.04 (6.57)	4.20 (5.10)	8.89 (6.56)	-
				5.56***
Anxiety Score	2.69 (1.64)	1.92 (1.89)	2.86 (1.53)	-3.30**
Alcohol Score	1.58 (1.61)	1.40 (1.70)	1.62 (1.59)	--0.84
Internalized Transprejudice Score	24.05 (6.74)	19.60 (9.88)	25.04 (5.38)	- 3.77***
Social Support Score	7.09 (2.83)	10.88 (3.21)	6.25 (1.91)	9.82***

^ap values are based on Welch’s two-sample t test statistic (continuous variables)or χ² = Chi-square test statistic (categorical variables)

^*p<0.05,

^**p<0.01,

^***p<0.001

Associations between Stigma and Condomless Anal Sex (Wave-3) and between Stigma and Psychosocial Problems (Wave-2)

CAS had a statistically significant association with sex work stigma (OR=1.145, 95% CI 1.095–1.197) but not with transgender identity stigma or perceived HIV stigma (Table 3). Anxiety emerged as a significant predictor of CAS (OR=3.192, 95% CI 1.075–9.482), while depression, alcohol use, and internalized transprejudice showed no significant associations with CAS. Higher social support scores were associated with lower odds of CAS (OR=0.640, 95% CI 0.541–0.757).

Table 3. Stigma-specific Effects on Wave-3 Condomless Anal Sex (CAS) and Wave-2 Psychosocial Problems: Logistic and Poisson Regression Results.

Predictors	CAS OR (95% CI)	Depression OR (95% CI)	Anxiety OR (95% CI)	Problematic Alcohol Use OR (95% CI)	Internalized Transprejudice OR (95% CI)	Social Support IRR (95% CI)
Psychosocial problems
Depression (1 vs. 0)	1.512 (0.322–7.107)	-	-	-	-	-
Anxiety (1 vs. 0)	3.192 (1.075–9.482)*	-	-	-	-	-
Alcohol Use (1 vs. 0)	0.601 (0.155–2.331)	-	-	-	-	-
Internalized Transprejudice (1 vs. 0)	1.123 (0.440–2.863)	-	-	-	-	-
Social Support Score	0.640 (0.541–0.757)***	-	-	-	-	-
Stigmas
Transgender Identity Stigma Score	0.962 (0.899–1.029)	1.087 (1.030–1.146)**	1.089 (1.049–1.130)***	1.037 (0.991–1.085)	1.017 (0.985–1.050)	0.992 (0.986–0.997)**
Perceived HIV Stigma Score	1.180 (0.947–1.470)	1.098 (0.969–1.244)	1.004 (0.929–1.084)	1.002 (0.908–1.107)	1.010 (0.942–1.082)	0.966 (0.955–0.978)***
Sex Work Stigma Score	1.145 (1.095–1.197)***	1.054 (1.014–1.096)**	1.037 (1.015–1.060)**	0.982 (0.955–1.009)	1.018 (0.997–1.039)	0.990 (0.987–0.994)***
Confounding variables
Age in Years	1.005 (0.936–1.079)	0.997 (0.946–1.051)	1.033 (0.993–1.074)	1.046 (0.995–1.100)	0.998 (0.963–1.033)	0.998 (0.992–1.004)
Education (1 vs. 0)	2.009 (0.763–5.287)	1.528 (0.844–2.765)	1.754 (1.089–2.827)*	0.816 (0.410–1.625)	0.792 (0.500–1.255)	0.940 (0.866–1.022)

Each column represents the output of a single multivariable logistic or Poisson regression model, with the variable in the column header specified as the outcome; thus, this table represents the output of six different regression models.

^*p<0.05,

^**p<0.01,

^***p<0.001, OR: Odds Ratio, IRR: Incidence Rate Ratio

In the analysis of psychosocial problems (specified as outcomes), higher transgender identity stigma scores were associated with both depression (OR=1.087, 95% CI 1.030–1.146) and anxiety (OR=1.089, 95% CI 1.049–1.130). A similar pattern was observed with sex work stigma and its associations with both depression (OR=1.054, 95% CI 1.014–1.096) and anxiety (OR=1.037, 95% CI 1.015–1.060). Perceived HIV stigma showed no significant associations with any of the psychosocial problems. Social support was negatively associated with transgender identity stigma (Incidence Rate Ratio [IRR]=0.992, 95% CI 0.986–0.997), perceived HIV stigma (IRR=0.966, 95% CI 0.955–0.978), and sex work stigma (IRR=0.990, 95% CI 0.987–0.994).

Mediation Models

The mediation models yielded evidence that social support, depression, and anxiety played mediating roles in the associations between stigma and CAS (Wave-3). Lower social support scores mediated the relationships between all stigmas and CAS: for transgender identity stigma (NIE OR=1.02, 95% CI 1.007-1.04, p=0.004), HIV-related stigma (NIE OR=1.33, 95% CI 1.10-1.61, p=0.004), and sex work stigma (NIE OR=1.03, 95% CI 1.01-1.04, p<0.001) (Table 4). Similarly, anxiety mediated the relationships between all stigmas and CAS: for transgender identity stigma (NIE OR=1.07, 95% CI 1.03-1.11, p=0.001), HIV-related stigma (NIE OR=1.07, 95% CI 1.01-1.13, p=0.03) and sex work stigma (NIE OR=1.04, 95% CI 1.01-1.08, p=0.02). In contrast, depression only mediated the associations between CAS and transgender identity stigma (NIE OR=1.02, 95% CI 1.001-1.05, p=0.04) and between CAS and HIV-related stigma (NIE OR=1.03, 95% CI 1.002-1.07, p=0.04). Alcohol use and internalized transprejudice showed no mediation effects.

Table 4. Effect of Individual Stigmas on Condomless Anal Sex (CAS): Causal Mediation Analysis Results.

Exposurea	Mediatorb	Natural Indirect Effect (NIE) OR (95% CI), p	Natural Direct Effect (NDE) OR (95% CI), p	Total Effect (TE) OR (95% CI), p
Transgender Identity Stigma Score (TGS)	Depression	1.023 (1.001, 1.046), p=0.040	1.047 (0.965, 1.136), p=0.270	1.071 (0.978, 1.176), p=0.143
	Anxiety	1.069 (1.026, 1.113), p=0.001	1.035 (0.946, 1.133), p=0.455	1.106 (1.010, 1.213), p=0.030
	Problematic Alcohol Use	0.993 (0.977, 1.010), p=0.412	1.019 (0.953, 1.089), p=0.581	1.012 (0.940, 1.090), p=0.755
	Internalized Transprejudice	1.004 (0.994, 1.014), p=0.413	1.029 (0.965, 1.098), p=0.386	1.034 (0.965, 1.107), p=0.347
	Social Support Score	1.022 (1.007, 1.038), p=0.004	1.019 (0.985, 1.055), p=0.274	1.042 (1.005, 1.080), p=0.027
Perceived HIV Stigma Score (HPS)	Depression	1.033 (1.002, 1.065), p=0.038	1.077 (0.932, 1.245), p=0.316	1.112 (0.970, 1.275), p=0.128
	Anxiety	1.066 (1.006, 1.130), p=0.032	1.063 (0.890, 1.267), p=0.500	1.132 (0.954, 1.344), p=0.155
	Problematic Alcohol Use	0.990 (0.962, 1.020), p=0.517	1.067 (0.889, 1.280), p=0.482	1.058 (0.877, 1.274), p=0.559
	Internalized Transprejudice	0.999 (0.990, 1.008), p=0.797	1.072 (0.901, 1.274), p=0.433	1.071 (0.900, 1.272), p=0.443
	Social Support Score	1.329 (1.097, 1.610), p=0.004	1.210 (0.894, 1.640), p=0.217	1.608 (1.292, 2.002), p<0.001
Sex Work Stigma (SWS)	Depression	1.006 (0.985, 1.027), p=0.577	1.288 (1.110, 1.493), p=0.001	1.296 (1.111, 1.511), p=0.001
	Anxiety	1.042 (1.007, 1.079), p=0.020	1.362 (1.111, 1.672), p=0.003	1.421 (1.155, 1.745), p=0.001
	Problematic Alcohol Use	1.000 (0.998, 1.002), p=0.982	1.303 (1.120, 1.517), p=0.001	1.303 (1.120, 1.517), p=0.001
	Internalized Transprejudice	1.000 (0.996, 1.004), p=0.981	1.281 (1.105, 1.486), p=0.001	1.282 (1.104, 1.488), p=0.001
	Social Support Score	1.027 (1.013, 1.042), p<0.001	1.035 (1.019, 1.050), p=<0.001	1.063 (1.042, 1.084), p<0.001

This table represents the output of 15 different mediation analyses, where the variable in column 1 is specified as the primary exposure of interest (at Wave-1), the variable in column 2 is specified as the putative mediator of interest (at Wave-2), and CAS at Wave-3 is specified as the outcome.

NIE: Natural Indirect Effect; NDE: Natural Direct Effect; TE: Total Effect; OR: Odds Ratio

^aExposure contrasts are 50th vs. 75th percentile: TGS: 20 vs. 15; HPS: 9 vs. 6; SWS: 24 vs. 18.

^bAll mediators, except social support score, are binary mediators. Exposure models adjusted for age and education. Outcome model adjusted for age, education, and all mediators.

Statistically significant direct effects (NDE) and total effects (TE) of sex work stigma on CAS were present across depression, anxiety, problematic alcohol use, internalized transprejudice and social support pathways, indicating that unmeasured mechanisms contribute to CAS.

Additive Interactions

Analyses of additive interactions, with stigma variables dichotomized at the median, revealed that specific pairs of stigma experiences were associated with synergistic or antagonistic effects on outcomes (Table 5). For Wave-2 CAS, HIV-related stigma and transgender identity stigma had a strong synergistic interaction in their impacts on the outcome (HPS × TGS: AP = 0.31, 95% CI 0.11–0.52), while HIV-related stigma and sex work stigma had a weak synergistic interaction (HPS x SWS: AP = 0.14, 95% CI 0.03–0.26). There were no statistically significant additive interactions for Wave-3 CAS.

Table 5. Additive Interactions Results: Two-way Interactions between Wave-1 Stigmas on Wave-2 Psychosocial Problems and Condomless Anal Sex.

Outcomea	Interactionsb	RERI (95% CI), p	AP (95% CI), p	SI (95% CI), p	Interpretationc
Wave-2 CAS	HPS× SWS	0.13 (0.02– 0.23), p=0.017	0.14 (0.03– 0.26), p=0.018	1.22 (0.98– 1.45), p=<.001	Weak Synergy
	HPS × TGS	0.24 (0.08– 0.41), p=0.004	0.31 (0.11– 0.52), p=0.004	2.21 (0.01– 4.41), p=0.053	Moderate Synergy
	SWS × TGS	0.05 (-0.07– 0.16), p=0.43	0.06 (-0.09– 0.20), p=0.43	1.10 (0.84– 1.36), p=<.001	No Synergy
Wave-3 CAS	HPS × SWS	0.03 (-0.22– 0.28), p=0.806	0.03 (-0.23– 0.29), p=0.806	1.07 (0.60– 1.54), p=<.001	No Synergy
	HPS × TGS	0.09 (-0.24– 0.42), p=0.598	0.10 (-0.28– 0.49), p=0.599	2.14 (-4.29– 8.56), p=0.516	No Synergy
	SWS × TGS	0.06 (-0.08– 0.20), p=0.403	0.06 (-0.08– 0.21), p=0.399	1.13 (0.79– 1.47), p=<.001	No Synergy
Depression	HPS × SWS	0.03 (-0.04– 0.11), p=0.353	0.16 (-0.18– 0.50), p=0.347	1.35 (0.18– 2.53), p=0.027	No Synergy
	HPS × TGS	0.10 (0.02– 0.18), p=0.012	0.50 (0.11– 0.89), p=0.013	-0.58 (– 49.17– 48.01), p=0.981	Strong Synergy
	SWS × tgs	0.14 (0.08– 0.21), p=<.001	0.56 (0.28– 0.85), p=<.001	3.97 (-2.10– 10.04), p=0.204	Strong Synergy
Anxiety	HPS × SWS	0.24 (0.14– 0.35), p=<.001	0.50 (0.28– 0.72), p=<.001	-2.47 (– 19.43– 14.50), p=0.777	Strong Synergy
	HPS × TGS	0.17 (0.01– 0.33), p=0.044	0.32 (0.01– 0.63), p=0.046	3.01 (– 10.63– 16.64), p=0.667	Moderate Synergy
	SWS × TGS	0.29 (0.18– 0.40), p=<.001	0.45 (0.27– 0.63), p=<.001	3.90 (-3.52– 11.33), p=0.306	Moderate Synergy
Problematic Alcohol Use	HPS × SWS	0.09 (0.02– 0.16), p=0.019	0.65 (0.11– 1.20), p=0.022	0.47 (0.19– 0.75), p=0.001	Strong Synergy
	HPS × TGS	0.00 (-0.08– 0.08), p=0.948	0.02 (-0.50– 0.53), p=0.945	0.14 (-3.62– 3.91), p=0.941	No Synergy
	SWS × TGS	-0.04 (-0.12– 0.04), p=0.352	-0.25 (-0.79– 0.29), p=0.364	0.45 (-4.58– 5.48), p=0.861	No Synergy
Internalized Transprejudice	HPS × SWS	-0.09 (-0.20– 0.02), p=0.100	-0.19 (-0.41– 0.03), p=0.095	0.62 (0.40– 0.84), p=<.001	No Synergy
	HPS × TGS	0.09 (-0.09– 0.26), p=0.327	0.18 (-0.18– 0.53), p=0.329	-0.24 (- 29.17– 28.69), p=0.987	No Synergy
	SWS × TGS	0.07 (-0.06– 0.19), p=0.307	0.13 (-0.12– 0.38), p=0.303	3.88 (– 14.50– 22.25), p=0.680	No Synergy
Social Support Score	HPS × SWS	-0.07 (-0.22– 0.07), p=0.318	-0.21 (-0.63– 0.20), p=0.313	1.19 (0.76– 1.63), p=<.001	No Synergy
	HPS × TGS	-0.09 (-0.19– -0.003), p=0.047	-0.23 (-0.46– -0.00), p=0.052	1.36 (0.88– 1.84), p=<.001	Subadditive Antagonism
	SWS × TGS	0.07 (-0.01– 0.16), p=0.093	0.14 (-0.03– 0.31), p=0.108	0.79 (0.58– 0.99), p=<.001	No Synergy

Note. This table represents the output of 21 different regression models, where the variable in column 1 was specified as the outcome and column 2 contains the product term of primary interest in assessing for a potential interaction between the variables shown. All regression models included the two variables shown and their product term, and adjusted for age and education.

RERI = Relative Excess Risk due to Interaction; AP = Attributable Proportion due to interaction; SI = Synergy Index; TGS = Transgender Identity Stigma, SWS = Sex Work Stigma, HPS = Perceived HIV Stigma

^aAll outcomes, except the social support score, were binary outcomes.

For depression, the interactions of both HIV-related stigma and transgender identity stigma (HPS × TGS: AP = 0.50, 95% CI 0.11–0.89) and sex work and transgender identity stigma (SWS × TGS: AP = 0.56, 95% CI 0.28–0.85) demonstrated strong synergy. Similarly, for anxiety, all three two-way interactions showed significant synergistic effects, with the strongest interactions observed for HPS × SWS (AP = 0.50, 95% CI 0.28–0.72) and SWS × TGS (AP = 0.45, 95% CI 0.27–0.63).

The interaction between HPS and SWS on problematic alcohol use also showed evidence of strong synergy (AP = 0.65, 95% CI 0.11–1.20). In contrast, the analysis of social support pointed toward antagonism, with a statistically significant negative interaction observed between HPS and TGS (RERI = -0.09, 95% CI -0.19 to -0.003). There were no statistically significant additive interactions for internalized transprejudice. Overall, these findings suggest that the combined burden of specific stigmas synergistically increased the risk for depression, anxiety, and problematic alcohol use, while one combination had an antagonistic relationship with social support.

Discussion

This study advances the understanding of how stigma drives HIV risk behavior among transgender women in India by integrating two distinct but complementary theory-driven approaches (minority stress theory and a psychological mediation framework) to guide our mediation analysis and an intersectionality framework to model stigma interactions. The mediation hypothesis investigates the mechanisms by which Wave-1 stigmas (transgender identity, perceived HIV status, and sex work) contribute to Wave-2 psychosocial challenges (depression, anxiety, problematic alcohol use, internalized transprejudice and social support), which subsequently increase Wave-3 condomless anal sex. In contrast, the interaction hypothesis tests whether the co-occurrence of multiple Wave-1 stigmas synergistically amplifies the psychosocial burden and HIV risk behavior beyond the sum of their individual effects, focusing on interactions among stigma types rather than external moderators. The theoretical innovation of this study lies in simultaneously examining how (mediation) and whether (interactions) stigmas compound vulnerability, while its methodological rigor is underscored by a prospective cohort design spanning three waves, which clarifies temporal ordering and strengthens causal inference.

This study provides evidence for both our mediation and interaction hypotheses. The findings from the mediation analyses demonstrated that stigma influences CAS through both direct and indirect pathways, with social support emerging as the most robust and consistent mediator across all three stigma types. We also observed differential patterns of mediation: anxiety mediated the effects of all three stigma types, while depression mediated only the effects of transgender identity and HIV-related stigma, suggesting distinct psychological pathways through which different stigmas operate. Complementing these findings, additive interaction analyses revealed strong synergistic effects between Wave-1 stigma combinations, with the perceived HIV stigma and transgender identity stigma interaction showing the most pronounced synergy on Wave-2 condomless anal sex, indicating that trans women experiencing both stigmas had substantially elevated risk beyond their additive effects. Conversely, we found that stigma combinations showed antagonistic effects on social support (negative RERI), demonstrating that intersecting stigmas systematically undermine protective resilience resources. Together, these results offer critical insights into the dual mechanisms (mediation and interactions) by which stigmas influence condomless sex and mental health, which have practical implications for HIV prevention and mental health promotion interventions targeting multiply marginalized trans women.

Our findings provide nuanced support for minority stress theory and the psychological mediation framework. The significant and consistent mediation by social support and anxiety across all three stigmas provides support for their roles as pivotal pathways linking distal minority stressors (stigmas) to CAS. Stigma likely undermines perceived or actual social support by straining family relationships or prompting self-isolation to avoid discrimination in social spaces [31], thereby eroding protective resources. For example, afraid of ‘stigma by association’, the fear that societal backlash will affect them or their cisgender children, family members may abuse and disown their transgender children and force them from the home [56]. This illustrates how societal stigma and discrimination directly and indirectly contribute to reduced social support, including loss of family support. We found that multiple stigmas increase condomless anal sex through anxiety, potentially via an arousal-mediated pathway: anxiety’s physiological arousal components create subjective excitement that increases susceptibility to sexual risk-taking, based on prior research among other populations [57–59]. Future research should investigate whether this mechanism operates uniformly across populations or is influenced by transgender-specific contextual factors.

Our finding that depression mediated only transgender identity and HIV-related stigma, but not sex work stigma, suggests that different stigmas may activate distinct affective and cognitive responses. The lack of significant mediation by problematic alcohol use and internalized transprejudice could reflect cultural (e.g., hijra communities/subculture) or contextual factors in India or potential measurement limitations (given the suboptimal reliability of the AUDIT-C alcohol use scale). Although internalized transprejudice may lead to depression and depression to problematic alcohol use (or vice versa), we did not test such sequential or reciprocal pathways, as all mediators were measured at the same timepoint [60]. Future research should examine these pathways and temporal relationships to clarify their roles.

In comparing these results to existing studies, our demonstration of social support as a key mediator mirrors findings from research among sexual and gender minorities globally and in India, where social support has been shown to buffer the effects of stigma on health outcomes and to promote safer sexual practices [17, 29, 30]. Similarly, the significant mediation through anxiety is in line with prior work that has identified psychological distress as a mechanism through which minority stress elevates HIV risk [61]. However, the lack of mediation through problematic alcohol use contrasts with findings from both India and abroad. A large-scale Indian study reported that sexual violence victimization increased condomless sex through alcohol and drug use [40], while several Western studies have identified alcohol or drug use as pathways linking stigma to HIV-related risk behaviors among vulnerable communities [62, 63]. This discrepancy may reflect cross-cultural differences and contexts. For instance, among trans women engaged in sex work in India, alcohol consumption may occur before standing at work spots [6], not primarily as a means of coping with stigma or psychological distress but to bolster courage against potential threats from ruffians or police or to deter male clients from cheating them out of payment, another potential explanation that aligns with the first author’s engagement with trans communities over the past two decades. Additionally, COVID-19 lockdown restrictions may have restricted alcohol availability and altered usual consumption patterns during our data collection period (especially Wave-1 and Wave-2), potentially disrupting the alcohol-mediated pathways reported in other contexts. Notably, internalized transprejudice did not mediate the relationship between multiple stigmas and condomless anal sex, contrary to our theory-guided hypotheses. It is possible that collective identities or community-based forms of resilience among hijra and other indigenous trans communities (e.g., thirunangai) may offset, or at least complicate, the role of internalized transprejudice, a hypothesis deserving further investigation [7].

Our findings regarding the synergistic interactions among stigmas bring methodological and theoretical advancements to the intersectionality literature on HIV risk. While the framework of intersectionality has long posited that the impacts of multiple stigmas are beyond additive, a principle often cited but rarely tested with appropriate quantitative methods in India, this study provides direct evidence for such synergistic effects of intersecting stigmas using measures such as the RERI and attributable proportion. At the same time, the antagonistic interaction that we found for social support highlights that not only can multiple stigmas compound risk, but they can also erode protective factors, a pattern less often documented but of high theoretical importance. Taken together, this study advances theoretical consensus by formally demonstrating that intersecting stigmas operate through multiple, simultaneous pathways, thus amplifying psychological distress and directly undermining resilience resources in ways greater than the sum of individual effects.

Methodologically, this study addresses key gaps raised in prior research: the use of a large prospective cohort, the application of causal mediation analysis, and the simultaneous modeling of main, mediated, and interaction effects constitute significant advances over prior cross-sectional or qualitative designs predominant in the Indian context. By applying additive interaction metrics rarely used in stigma research in India, this study offers a template for future quantitative explorations of intersectional hypotheses in other settings and populations [64–66].

Implications for Theory, Policy and Practice

The synergistic effects documented in this study provide compelling evidence that interventions need to address more than one form of stigma and multiple marginalized statuses, taking into account intersectionality. These findings support the implementation of comprehensive, multicomponent, multilevel intervention frameworks that simultaneously target individual, interpersonal, and structural determinants of HIV risk [67, 68]. The strong synergistic interactions suggest that coordinated approaches targeting multiple stigmas simultaneously may yield disproportionately large benefits. Critically, the effective implementation of India’s Transgender Persons (Protection of Rights) Act, 2019, and accompanying Rules, 2020, which explicitly prohibit discrimination against transgender persons in healthcare, employment, and education settings, requires active monitoring and enforcement mechanisms. Our findings underscore that legal protections alone are insufficient without concurrent efforts to address intersectional stigma within healthcare systems, families, and communities.

The substantial indirect effects through anxiety and depression for intersectional stigma, combined with anxiety mediating all three individual stigma types, demonstrate that HIV prevention interventions must integrate evidence-based mental health components. This finding aligns with WHO recommendations for integrated HIV and mental health interventions, which have shown effectiveness in improving both treatment adherence and stigma reduction outcomes. NACO, in its white paper on ‘Comprehensive Health-Related Services for Transgender Persons’ [69], explicitly states the importance of mental health and stigma reduction components in reducing HIV risk and improving treatment adherence. Given the importance of social support documented in our study, interventions must extend beyond individual-level mental health screening to include family acceptance programs. The Indian context particularly necessitates culturally adapted family interventions that address the complex dynamics within hijra communities and traditional family structures, where family rejection significantly impacts mental health outcomes. Mental health screening should be integrated into all HIV prevention programs for trans women, with services available to address depression and anxiety given that they have been documented as significant mediators of stigma effects on HIV risk behaviors.

Given the persistently high levels of CAS documented in this cohort and the well-established challenges to consistent condom use among trans women (e.g., including forced sex encounters and condom negotiation barriers in sex work) [16], pre-exposure HIV prophylaxis (PrEP) is a critical complementary prevention strategy that warrants urgent scale-up in India [70]. However, realizing PrEP’s potential will require addressing the psychosocial barriers identified in the present study, as multifaceted stigma, depression, and eroded social support are likely to undermine awareness, acceptability, and adherence among trans women, particularly those in sex work [70].

Strengths and Limitations

Several limitations should be considered when interpreting these findings. Our recruitment strategy, which primarily relied on community-based organizations in two metropolitan cities, may limit generalizability to trans women in rural areas or those not connected to formal support networks. As is common in community-based studies using peer-led recruitment, systematic data on the number of trans women approached and screened for eligibility prior to enrollment were not collected, precluding the reporting of a complete participant flow from initial contact to enrollment; future studies should prospectively track recruitment yield at each stage to enable assessment of selection bias. In addition, because data collection coincided with India’s COVID-19 waves, pandemic-related mobility restrictions may have further limited participation among trans women who were most socially and economically marginalized. Our stigma measures focused on the individual and interpersonal levels, while structural-level stigma, such as the implementation of anti-discrimination policies, remains unmeasured. Although this study operationalized intersectionality analytically [71] – by measuring three theoretically distinct stigma constructs separately and formally testing their statistical interactions using additive interaction metrics – we did not employ a composite intersectional stigma instrument (e.g., the Intersectional Discrimination Index) [72]. Such intersectional stigma tools capture qualitatively distinct experiences of simultaneous marginalization that may not be fully represented by interacting individually measured stigma scores; their absence means we may have underestimated the lived experience of compounded stigma and its effects on CAS. Future research should assess multiple dimensions (enacted, anticipated, internalized, and perceived) of various stigmas (or the intersectional stigma construct) to elucidate their distinct effects on health outcomes. Another limitation is that although drug use was assessed via brief self-report items listing commonly used substances, the absence of a validated scale precluded its inclusion as a formal mediator alongside the psychometrically validated measures used for other psychosocial variables. This is noteworthy because stigma-related substance use coping may operate through drug use rather than, or in addition to, alcohol in this population, which may partly explain the null mediation finding for problematic alcohol use [40]. Additionally, we did not assess gender expression or the degree of ‘outness’ about transgender identity, factors that can significantly moderate stigma experiences. Sample size constraints prevented reliable estimation of three-way interactions, which require substantially larger samples than two-way interactions, limiting our ability to examine higher-order synergistic effects. The self-reported nature of CAS introduces potential social desirability bias. However, participants reported notably high levels of CAS that increased steadily from Wave-1 to Wave-3, coinciding with the COVID-19 pandemic when economic vulnerability among trans women sex workers was heightened, lending plausibility to these reports. Finally, although the approximately 6-month intervals between waves support the assumed temporal sequence from stigmas (Wave-1) to psychosocial mediators (Wave-2) to CAS (Wave-3), this spacing cannot fully rule out time-varying confounding or bidirectional relationships (for example, CAS influencing subsequent anxiety or depression), and our mediation results should therefore be interpreted as consistent with, but not conclusive of, causal pathways.

We implemented several strategies to address these limitations and enhance study robustness. Predictable attrition inherent to longitudinal cohort studies was addressed through multiple imputation analyses. While we did not test all possible alternative mediation pathways, our selected models were guided a priori by the Minority Stress and Psychological Mediation Frameworks, providing strong theoretical justification for the analytical approach. Sensitivity analyses using E-values demonstrated that the effects of stigmas on CAS through anxiety and social support were robust to unmeasured confounding, strengthening confidence in our findings. This study contributes to advancing intersectional stigma research in India by examining both individual stigma pathways and composite intersectional approaches in mediation analyses and using additive interaction metrics for assessing synergy and demonstrates that interconnected stigmas function both as unified constructs and through synergistic interactions.

Conclusion

This study illuminates the complex, multidimensional ways in which intersecting stigmas heighten HIV-related risk among trans women in India. By testing both mediation and interaction hypotheses, grounded in minority stress, psychological mediation, and intersectionality frameworks, we provide complementary and corroborative evidence for how distinct and combined stigmas contribute to psychological distress and HIV risk behavior and how they undermine protective social support. The concordance between mediation and interaction models underscores the central importance of social support as both a pathway and a resilience resource. Our findings highlight the urgent need for intersectional approaches to transgender health, moving beyond single-issue strategies to recognize and address the consequences of experiencing multiple and overlapping stigmas at the individual, interpersonal, and structural levels. Meaningful progress in reducing health disparities requires coordinated, multilevel stigma reduction and psychosocial interventions. Such efforts are essential for improving HIV outcomes and advancing health equity and social inclusion for trans women worldwide.

Figure 1. Conceptual model to test the mediation hypothesis.

Data availability

Data will not be shared publicly.