Past-year violence experienced by transgender women living with HIV with or without recent involvement in transactional sex: the French national cross-sectional ANRS-Trans&HIV study

Abstract

Introduction

Stigma and violence experienced by transgender women and women engaged in transactional sex may negatively impact health and access to HIV care. We investigated typologies of past-year violence experienced by transgender women living with HIV (TWLH), with or without involvement in transactional sex (TWLH-TS and TWLH-nTS, respectively) in France.

Methods

We used data from the 2020–2022 French national cross-sectional ANRS-TRANS&HIV survey, which included TWLH recruited in 36 HIV care centres. We performed latent class analyses using variables representing combinations of past-year violence type and perpetrator separately for TWLH-TS and TWLH-nTS. Correlates of the resulting past-year violence typologies among TWLH-TS were assessed using multinomial logistic regression.

Results

Among TWLH-TS (66% of the 501 participants), three typologies emerged: low reported violence (42% of the 501 participants), multiple-perpetrator violence (8%) and client and street violence (17%). Among TWLH-nTS (34% of the 501 participants), two typologies emerged: low reported violence (30% of the 501 participants) and multiple-perpetrator violence (3%). In TWLH-TS, correlates of client and street violence were inadequate housing, ART interruption, stimulant use and daily cannabis use, while those of multiple-perpetrator violence were depression and a lack of emotional support. Client and street violence and multiple-perpetrator violence were correlated with meeting clients in outdoor locations.

Conclusion

TWLH in France face intersecting forms of systemic violence and structural vulnerability. One in ten experienced multi-form, multiple-perpetrator violence, regardless of transactional sex involvement, with potential consequences for mental health. Client and street violence affected one in four TWLH-TS, and was associated with precarious housing and working conditions, as well as substance use and ART adherence difficulties. Ensuring equitable access to HIV care and mental health services for transgender women is essential and inseparable from protecting their fundamental human rights and improving their socioeconomic conditions.

Trial Registration

NCT04849767.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Nationwide recruitment with high participation provided good representativeness of transgender women living with HIV in care in France.

• Typologies of recently experienced violence were established based on multiple dimensions of violence, detailed by form and perpetrator.

• Participants were recruited primarily from hospital-based HIV care, which may limit generalisability to transgender women not engaged in care or receiving HIV care elsewhere, and introduces potential selection bias due to physician-led recruitment.

• The identified violence typologies support targeted interventions to improve safety and service access, especially among transgender women engaged in transactional sex exposed to client and street violence.

• Due to the cross-sectional design, causality could not be inferred. Small sample sizes limited the analysis of correlates of violence among transgender women not engaged in transactional sex.

Introduction

Transgender women are a key HIV population, with a disproportionately high prevalence, estimated at 20% globally, and an incidence rate of 0.43 per 100 person-years in 2022.^{1 2} Throughout their lives, they face structural and interpersonal stigma, as well as verbal, physical and sexual violence in healthcare facilities, in the workplace, within families, from partners and from passersby, all of which negatively affect health outcomes and access to care.^3–5 More generally, among marginalised groups like transgender women, violence, stigma, poverty and insufficient healthcare services have cumulative and interacting adverse effects on substance use, mental and overall health, and high-risk sexual behaviours. These effects increase the risk of HIV acquisition, and create barriers to HIV-care access and uptake, collectively referred to as the HIV syndemic.^6–9

Although no national estimate of HIV prevalence among transgender women is currently available in France,¹⁰ they are likely disproportionately affected. Studies in transgender women, including those involved in transactional sex, in other European countries have reported prevalence rates ranging from 12% to 22%.¹¹ A small study found that one-third of 49 transgender women receiving care at a tertiary hospital in Paris were living with HIV.¹² While being HIV-positive can constitute an additional source of potential stigma and exposure to violence,^{13 14} French data on violence and the associated social and health indicators among transgender women living with HIV (TWLH) are scarce. In the French national survey, ANRS-Trans&HIV, which was conducted between 2020 and 2022 among 536 TWLH engaged in hospital-based HIV care, 84% were migrants (ie, non-French nationals), 65% engaged in transactional sex and 30% were undocumented (ie, had no resident permit). Additionally, 33% lived in precarious housing and 34% had recently relied on food assistance.^{15 16} Moreover, despite France’s historically strong social and health protection system, recent years have seen a political trend toward restricting access to rights for migrants and reducing public health services.^17–19 Finally, France’s abolitionist sex work policy, based on the Nordic model which criminalises sex buyers, may fail to adequately protect individuals who engage in transactional sex from violence and limits their access to services providing related support.^20–23

Given this context in France, which likely increases exposure to violence and limits protection for migrant populations, transgender individuals and those involved in transactional sex, combined with the high socioeconomic vulnerabilities reported among TWLH engaged in care, we hypothesised that TWLH are highly exposed to violence, which may further exacerbate health vulnerabilities. However, it remains unclear which types of violence TWLH experience, who the perpetrators are and how these experiences are distributed across social spheres, particularly in relation to involvement in transactional sex. Using data from ANRS-Trans&HIV, the present study aimed to (i) identify latent patterns of violence (ie, violence typologies) of TWLH with and without recent involvement in transactional sex, based on the perpetrators and type of violence they experienced; (ii) examine how these violence typologies differed in terms of the accumulation of other health-related and social vulnerabilities including difficulties in ART adherence, poor mental health, socioeconomic, housing and migration-related vulnerabilities

Methods

The ANRS-Trans&HIV survey

Design and settings

ANRS-Trans&HIV was a cross-sectional, nationwide and community-based survey among TWLH receiving care in hospital-based HIV services in France, conducted between August 2020 and June 2022. It aimed to identify vulnerabilities and barriers to healthcare access and retention, as well as TWLH’s HIV-related and gender affirmation needs. It was developed and implemented through close collaboration between academic researchers and community organisations led by and for transgender people and those involved in HIV advocacy and support. The detailed methodology can be found elsewhere.¹⁵ Of 232 French hospital-based HIV services, 54 had at least one TWLH on their active patient roster in the previous 12 months, and 36 of them (with 777 TWLH receiving care) participated in ANRS-Trans&HIV.

Participant recruitment and data collection

Eligibility criteria were being 18 years old or older, receiving care in a hospital-based HIV unit in France, identifying as a transgender woman (ie, individuals assigned male at birth but identifying as women, female or trans feminine) and providing informed consent. Physicians in participating services were briefed to invite all eligible TWLH to participate, based on medical records in which the sex assigned at birth and current gender identity were recorded.

Physicians completed a refusal report form in which they anonymously recorded basic demographic information, including age, nationality and education level, of those who declined to participate. Participants completed a standardised questionnaire in a face-to-face interview conducted by trained community-based interviewers, proficient in French, Spanish or Portuguese. The questionnaire collected current and lifetime information on participants’ socioeconomic status, living and psychosocial conditions, health status, behaviours and healthcare utilisation. Clinical data were obtained from medical records.

Patient and public involvement

ANRS-Trans&HIV was developed with community organisations led by and for transgender people and HIV advocacy groups, who contributed to study design, recruitment and questionnaire development. All study results are shared and discussed with these community partners to ensure accurate interpretation and to support dissemination to their communities.

Data analysis

Study population

For the present analysis, we selected data for participants in ANRS-Trans&HIV with available medical records and questionnaires, and who had complete data on violence (see below).

We described the demographic, socioeconomic, psychosocial, behavioural, HIV-related and transactional sex characteristics for (i) the present study’s sample, (ii) participants with recent involvement in transactional sex (TWLH-TS), defined as having exchanged sex for money, housing, goods or services in the previous 12 months and (iii) those without recent involvement in transactional sex (TWLH-nTS).

Construction of violence typologies in TWLH-TS and TWLH-nTS

The questionnaire assessed four dimensions of lifetime violence (1) verbal violence, (2) physical violence (3) sexual violence involving psychological pressure or threats and (4) sexual violence involving physical force, with one question for each dimension. Respondents who answered ‘yes’ to any question were then asked when the violence occurred (in the previous 12 months vs more than 12 months ago), as well as the perpetrator (for verbal and physical violence: passersby, clients (ie, persons who purchased sex), co-workers, law enforcement, non-paying sexual partners, family members, friends or other persons; for sexual violence: clients, casual partners, primary partners and violence by participants themselves against their primary partners).

We constructed 24 binary indicators corresponding to all combinations of violence type and perpetrator, each indicating whether participants had experienced a given form of violence from a specific perpetrator in the previous 12 months (yes/no), and the optimal solution was selected based on Bayesian Information Criterion, entropy, class sizes and substantive interpretability.

From these responses, we created 24 dichotomous variables corresponding to all combinations of violence type and perpetrator, each indicating whether participants had experienced a given form of violence from a specific perpetrator in the previous 12 months or not (eg, having experienced verbal violence from passersby in the previous 12 months: ‘yes’ vs ‘no’: ie, never or more than 12 months ago).

We then conducted two latent class analyses (LCA)—one for each subgroup (ie, TWLH-TS and TWLH-nTS)—using the aforementioned dichotomous variables to identify typologies of participants with similar experiences of recent violence.^{24 25} Latent class models with increasing numbers of classes were estimated, and the optimal number of typologies for each subgroup was determined according to the best-fit LCA model, using the lowest Bayesian Information Criterion (BIC) value as the primary criterion, complemented by entropy and class size to ensure interpretability and classification quality.²⁶

Correlates of violence typologies in TWLH-TS

We aimed to further distinguish violence typologies in terms of cumulative health and social vulnerabilities. After identifying violence typologies using latent class analysis, we conducted exploratory multinomial regression analyses to examine correlates of membership in each typology, with violence typology modelled as a categorical dependent variable. Analyses were restricted to the TWLH-TS group because of the small number of participants in one of the two TWLH-nTS typologies identified in the LCA (see Violence typologies in the Results section). Candidate independent variables were selected based on theoretical relevance and prior literature documenting associations or co-occurrence between these vulnerabilities and violence exposure among women:

• Demographic characteristics and socioeconomic status ­

  • Age²⁷ (years). ­

  • Education level^{27 28} (ie, high school diploma or lower vs higher than high school diploma). ­

  • Legal status²⁹ (European citizenship or resident permit vs undocumented). ­

  • Net monthly income²⁷ (≥1000€; 500–1000€; <500€). ­

  • Employment²⁸ (declared; undeclared or informal; unemployed). ­

  • Perceived housing conditions²⁸ (unsatisfactory vs satisfactory or acceptable).

• Psychosocial characteristics ­

  • Living alone (yes vs no).³⁰ ­ Having a primary partner³¹ (yes vs no). ­

  • Having someone to rely on for support in case of difficulty³² (ie, yes vs no to the question: ‘currently, is there someone you can count on in a difficult situation or problem?’). ­

  • Receiving needed emotional support in the previous 12 months³² (ie, yes, or not concerned/not in need vs no, or yes but not enough, to the question: ‘during the past 12 months, have you received the support you needed to face difficult situations or to lift your spirits?’). ­

  • Moderate to severe depression³³ (ie, defined as a score ≥10 on the PHQ-9 scale³⁴). ­

  • Gender affirmation (GAF) profile³⁵ (ie, patterns of engagement across psychological, social, medical and administrative dimensions of gender affirmation—such as hormone use, gender-affirming surgeries, gender-concordant identity documents and experiences of misgendering—identified through clustering analyses, and published elsewhere³⁶). ­

  • Past-month stimulant drug use³⁷ (ie, having used cocaine, crack, free-base, amphetamines or methamphetamines at least once in the past month, yes vs no). ­

  • Past-month dissociative drug use³⁸ (ie, having used ketamine, LSD or other hallucinogens at least once in the past month, yes vs no). ­

  • Past-month daily cannabis use³⁹ (ie, reporting daily or almost daily cannabis use in the past month, yes vs no). ­

  • At risk-alcohol consumption⁴⁰ (defined as a score ≥3 on the AUDIT-C scale, as recommended for transgender women,⁴¹ yes vs no).

• HIV-related outcome­

  • ≥2 days ART interruption in the previous month⁴² (yes vs no).

• Conditions of transactional sex­

  • Length of time engaging in transactional sex⁴⁰ (in years).

  • Methods used to contact clients and locations for paid sexual encounters (combined variable).^{40 43}

As a pre-screening step, univariate analyses were conducted, and covariates with a p value <0.2 were considered for inclusion in the multivariable model. First, we applied Bayesian Model Averaging (BMA) to account for model uncertainty and assess the robustness of associations across competing plausible models (online supplemental table 1). Variables identified as robust through BMA were subsequently entered into a multinomial logistic regression model. Model selection was based on the Bayesian Information Criterion (BIC), which favours parsimonious models while penalising model complexity. The final model corresponded to the model with the lowest BIC among the competing models, ensuring an optimal balance between goodness of fit and parsimony given the sample size (n=333). Adjusted odds ratios (aORs) from the final model were interpreted as conditional descriptive correlates of violence-typology membership, given the other variables retained in the selected model, rather than as independent effects or causal associations. Given the cross-sectional design and empirical variable selection, the magnitude and statistical significance of aORs were interpreted cautiously. To assess the robustness of the main findings, each statistically significant variable was examined in a separate multivariable model adjusted for predefined potential confounders⁴⁴ (online supplemental table 2).

Results

Selection and characteristics of participants

Of the 777 TWLH followed-up in the 36 participating HIV services, 624 were eligible and reachable; of these, 536 agreed to participate in ANRS-Trans&HIV. 30 had missing questionnaires or medical records, and 6 had incomplete data on violence in the questionnaire; accordingly, 501 participants were included in the present study.

Median age of the 501 participants was 43 years (table 1). The sample comprised 333 (66%) TWLH-TS and 168 (34%) TWLH-nTS. Most participants lived in the Paris region (422, 84%). One-third (170) lived in hostels, hotels, squats or were homeless, 23% (114) reported inadequate housing conditions, 32% (158) were undocumented, 35% (175) were unemployed and 70% (349) earned less than 1000 euros per month. 65% percent (327) were classified as ‘Predominant use of medical GAF dimension’, characterised by lifetime—but not current—hormone use, silicone injections and feminising—but not genital—surgery, with no significant use of the administrative, psychological or social GAF dimensions.³⁶ 52% (259) were living alone, 61% (306) had no primary partner, 39% (194) did not receive needed emotional support, and 21% (107) reported moderate-to-severe depression. Past-month stimulant drug use, at-risk alcohol use and daily cannabis use were reported by 27% (135), 38% (186) and 9% (43) of participants, respectively. 28% (138) and 13% (65), respectively, had interrupted their ART treatment in the previous month and had an unsuppressed viral load (ie, ≥50 copies/mL).

Table 1. Demographic, socioeconomic, psychosocial, HIV and transactional sex characteristics of transgender women living with HIV, with or without engagement in transactional sex (ANRS-Trans&HIV).

	Total	TWLH-nTS	TWLH-TS
	n=501	n=168	n=333
	n (%), median (IQR)
Demographic and socioeconomic characteristics
Geographic region
Paris region (Ile-de-France)	422 (84%)	129 (77%)	293 (88%)
Other region	79 (16%)	39 (23%)	40 (12%)
Age (years)	43 (36–50)	50 (43–56)	41 (34–47)
Education level
High school diploma or higher	98 (20%)	49 (29%)	49 (15%)
No schooling/primary school/some secondary education	402 (80%)	118 (71%)	284 (85%)
Housing
Stable (owner or renter of own home)	328 (66%)	133 (80%)	195 (59%)
Unstable (hosted/hotel/squat/homeless)	170 (34%)	34 (20%)	136 (41%)
Perceived housing conditions
Adequate (ie, responding very satisfactory/satisfactory/acceptable)	387 (77%)	141 (84%)	246 (74%)
Inadequate (ie, responding unsatisfactory/very unsatisfactory)	114 (23%)	27 (16%)	87 (26%)
Legal status in France
Documented (European citizenship or resident permit)	338 (67%)	144 (87%)	194 (59%)
Undocumented	158 (32%)	22 (13%)	136 (41%)
Type of health insurance
General public health insurance	347 (69%)	149 (89%)	198 (60%)
Medical state assistance for undocumented migrants	123 (25%)	12 (7%)	111 (33%)
No insurance	29 (6%)	6 (4%)	23 (7%)
Employment
Declared employment	64 (13%)	50 (30%)	14 (4%)
Undeclared or informal employment	262 (52%)	17 (10%)	245 (74%)
Unemployed/inactive	175 (35%)	101 (60%)	74 (22%)
Estimated monthly net income
≥1000€	152 (30%)	70 (42%)	82 (25%)
500–1000€	204 (41%)	67 (40%)	137 (41%)
<500€	145 (29%)	31 (19%)	114 (34%)
Psychosocial characteristics
Typology of gender affirmation (GAF) dimensions*
Current and comprehensive use of all GAF dimensions: medical, administrative, psychological and social	63 (13%)	41 (24%)	22 (7%)
Past and predominant use of the medical dimension of GAF	327 (65%)	91 (54%)	236 (71%)
Sparse use of medical and administrative dimensions of GAF	111 (22%)	36 (21%)	75 (23%)
Living alone
No	242 (48%)	60 (36%)	182 (55%)
Yes	259 (52%)	108 (64%)	151 (45%)
Had a primary partner
Yes	192 (39%)	74 (45%)	118 (35%)
No	306 (61%)	91 (55%)	215 (65%)
Had someone to rely on for support when needed
Yes	402 (80%)	146 (87%)	256 (77%)
No	99 (20%)	22 (13%)	77 (23%)
Received needed emotional support (in previous 12 months)
Yes/did not need support	307 (61%)	115 (69%)	192 (58%)
No/did not receive enough support	194 (39%)	53 (32%)	141 (42%)
Receive needed financial help (in previous 12 months)
Yes/did not need material help	213 (43%)	88 (52%)	125 (38%)
No/did not receive enough material help	288 (57%)	80 (48%)	208 (63%)
Depression
None to mild	386 (79%)	136 (84%)	250 (76%)
Moderate to severe†	104 (21%)	26 (16%)	78 (24%)
Use of stimulants (in previous month)
No	366 (73%)	138 (82%)	228 (69%)
Yes	135 (27%)	30 (18%)	105 (32%)
Use of dissociative drugs (in previous month)
No	482 (96%)	164 (98%)	318 (96%)
Yes	18 (4%)	3 (2%)	15 (5%)
Use of cannabis (in previous month)
No use or not daily	458 (91%)	158 (94%)	300 (90%)
Daily	43 (9%)	10 (6%)	33 (10%)
Alcohol consumption
No consumption or low risk	309 (62%)	120 (73%)	189 (57%)
At-risk consumption‡	186 (38%)	45 (27%)	141 (43%)
HIV-related outcomes
ART interruption (≥2 days in previous month)
No	356 (71%)	128 (78%)	228 (69%)
Yes	138 (28%)	37 (22%)	101 (31%)
ART treatment adherence
Good adherence	289 (58%)	112 (68%)	177 (54%)
Poor to moderate adherence§	206 (42%)	53 (32%)	153 (46%)
HIV viral load suppression (ie, <50 copies/mL) at most recent visit
Yes	429 (87%)	149 (90%)	280 (85%)
No	65 (13%)	17 (10%)	48 (15%)
Transactional sex conditions
Length of time engaging in transactional sex (years)	–	–	20 (13–26)
Contacting clients and locations for sexual encounters
Meeting clients at participant’s home after contacting them on the internet	–	–	90 (27%)
Meeting clients at participant’s home after contacting them using other means	–	–	66 (20%)
Meeting clients in outdoor locations	–	–	135 (41%)
Meeting clients in other locations¶	–	–	39 (12%)

^*Type of gender affirmation (GAF) domain use (defined and described elsewhere: medical, administrative, psychological, social.

^†Defined as PHQ-9≥10.

^‡Defined as AUDIT-C≥3.

^§Defined according to previously published categorisation.

^¶In a hotel, at the client’s home, in a vehicle, in a location rented with other people engaged in transactional sex.

ART, antiretroviral therapy; AUDIT-C, Alcohol Use Disorders Identification Test–Consumption; PHQ-9, Patient Health Questionnaire-9.; TWLH-nTS, TWLH without recent involvement in transactional sex; TWLH-TS, TWLH with recent involvement in transactional sex.

Among TWLH-TS (333), the median length of time they had engaged in transactional sex was 20 years (IQR 13–26) (table 1). 41% (135) reported meeting and having sex with clients outdoors, 27% (90) met clients at their home after contacting them on the internet, 20% (66) met clients at their home after contacting them using other means and 12% (39) in other locations (eg, hotel, client’s home, vehicle).

Comparison of Trans&HIV participants with eligible individuals who declined participation has been described previously¹⁶ and showed similar age, region of follow-up and level of education, but a higher proportion of French nationals among non-participants.

Past-year verbal, physical and sexual violence

In the past 12 months, verbal violence from passersby was reported by 29% of participants, from clients by 11%, from co-workers by 7%, from law enforcement by 7%, from sexual partners by 3%, from family members by 2% and from friends by 2%. Physical violence was experienced from passersby by 14% of participants, from clients by 8%, from co-workers by 2%, from law enforcement by 2%, from sexual partners by 2%, from friends by 2% and from other perpetrators by 2%. Psychological or threatening sexual coercion was reported by 20% of participants from clients, 3% from casual partners and 3% from primary partners. Physical sexual coercion was reported by 14% from clients, 3% from casual partners and 1% from primary partners.

Violence typologies

TWLH-nTS

Two typologies emerged from the LCA for the TWLH-nTS group (table 2). The first (152 participants, representing 90% of TWLH-nTS or 30% of the total sample) was characterised by low probabilities of all violence type-perpetrator combinations. We call this typology Low reported violence (figure 1). The second typology (16 participants, representing 10% of TWLH-nTS or 3% of the total sample) was characterised by high probabilities of verbal and physical violence from passersby and, to a lesser extent, from co-workers and law enforcement, as well as high probabilities of verbal, physical and sexual violence from sexual partners. We call this typology Multiple-perpetrator violence, as it reflects violence occurring across several social contexts, including public spaces, intimate relationships, and interactions with authorities, suggesting generalised exposure to violence from multiple perpetrator groups.

Table 2. Results of latent class analyses (LCAs) using ‘violence-type perpetrator’ variables.

LCA in TWLH-nTS
Model	Obs	ll(model)	df	AIC	BIC	Entropy	LRT χ² (Δdf)	P value
2 classes	168	−250.4508	135	566.9015	669.9924	0.73	–	–
3 classes	168	−232.2977	118	564.5953	720.7935	0.81	36.3062 (17)	0.004
4 classes	168	−221.5874	101	577.1748	786.4804	0.76	37.038 (17)	0.003
LCA in TWLH-TS
Model	Obs	ll(model)	df	AIC	BIC	Entropy	LRT χ² (Δdf)	P value
2 classes	333	−1511.065	284	3120.131	3306.73	0.7	–	–
3 classes	333	−1433.717	259	3015.435	3297.237	0.78	154.696 (25)	<0.001
4 classes	333	−1413.424	234	3024.848	3401.854	0.74	40.586 (25)	0.03

The two-class solution was retained for TWLH-nTS based on acceptable fit indices (BIC and entropy) and class size distribution. Although the three-class model showed slightly improved AIC and entropy, one class contained only a very small proportion of participants (2–3%), limiting statistical stability and interpretability. For TWLH-TS, the three-class solution was selected as the optimal model, supported by lower BIC, higher entropy and balanced class sizes. The four-class solution did not meaningfully improve model fit (non-significant LRT for TWLH-TS; small and unstable classes in both groups) and therefore was not retained.

AIC, Akaike information criterion; BIC, Bayesian information criterion; LRT, likelihood ratio test; TWLH-nTS, TWLH without recent involvement in transactional sex; TWLH-TS, TWLH with recent involvement in transactional sex.

Figure 1. Violence typologies in transgender women living with HIV (TWLH) with or without recent involvement in transactional sex. TWLH-nTS, TWLH without recent involvement in transactional sex; TWLH-TS, TWLH with recent involvement in transactional sex.

TWLH-TS

Three typologies emerged from the LCA for the TWLH-TS group (table 2). The first (208 participants, representing 62% of TWLH-TS or 42% of the whole study sample) was characterised by low probabilities of all type-perpetrator violence variables. Again, we call this low reported violence (figure 1). The second (39 participants, representing 12% of TWLH-TS or 8% of the whole study sample) was characterised by high probabilities of sexual violence from clients and casual partners, as well as verbal and physical violence from passersby, clients, co-workers, law enforcement, and to a lesser extent, from non-paying sexual partners, family members and friends. As before, we call this typology Multiple-perpetrator violence, reflecting pervasive violence across multiple environments, including sex work settings, public spaces, intimate relationships and interactions with authorities, indicating broad exposure to victimisation from diverse perpetrator groups. The third typology (86 participants, representing 26% of TWLH-TS or 17% of the whole sample) was characterised by high probabilities of verbal and physical violence from passersby and clients, as well as sexual violence from clients, but with low probabilities of violence from other perpetrators. We call this typology Client and street violence, as it reflects a more context-specific pattern of violence concentrated in public spaces and sex work-related interactions, with comparatively limited exposure to violence from other social networks or institutions.

Correlates of violence typologies in TWLH-TS

Table 3 presents the results of the final multivariable multinomial logistic regression model (see detailed results of model selection procedures in online supplemental table 1 and table 2). TWLH-TS from the client and street violence typology were more likely to report inadequate housing conditions (adjusted OR (95% CI): 2.3 (1.2 to 4.3)), stimulant use (2.3 (1.2 to 4.3)) and ART interruption of more than 2 days in the previous month (1.9 (1.1 to 3.5)), compared with TWLH-TS from the Low reported violence typology (table 3).

Table 3. Correlates of violence typologies in TWLH engaged in transactional sex (multinomial logistic regression, ref: typology ‘low reported violence’).

	Univariate analysis*				Multivariate analysis
	n=333				n=321
	Client and street violence		Multiple-perpetrator violence		Client and street violence		Multiple-perpetrator violence
	OR (95% CI)	P value	OR (95% CI)	P value	aOR (95% CI)	P value	aOR (95% CI)	P value
Demographic and socioeconomic characteristics
Age (years)	0.97 (0.94 to 1.00)	0.049	1.01 (0.97 to 1.05)	0.795
Education level
Upper secondary school diploma or third level education	Ref		Ref
No schooling/primary school/lower secondary school	2.18 (0.97 to 4.89)	0.059	2.68 (0.78 to 9.17)	0.116
Perceived housing condition
Adequate	Ref		Ref		Ref		Ref
Inadequate	2.18 (1.25 to 3.82)	0.006	2.83 (1.37 to 5.84)	0.005	2.29 (1.22 to 4.30)	0.01	1.76 (0.78 to 3.98)	0.173
Estimated monthly net income
≥1000€	Ref		Ref
500–1000€	0.89 (0.46 to 1.71)	0.725	4.89 (1.40 to 17.08)	0.013
<500€	1.62 (0.85 to 3.09)	0.141	4.30 (1.17 to 15.72)	0.028
Psychosocial characteristics
Had someone to rely on for support when needed
Yes	Ref		Ref
No	1.27 (0.69 to 2.34)	0.437	3.25 (1.58 to 6.67)	0.001
Had received needed emotional support (in previous 12 months)
Yes/did not need support	Ref		Ref		Ref		Ref
No/had not received enough support	0.75 (0.44 to 1.27)	0.286	3.76 (1.77 to 7.96)	0.001	0.67 (0.36 to 1.25)	0.206	3.08 (1.33 to 7.12)	0.009
Depression
None to mild	Ref		Ref		Ref		Ref
Moderate to severe†	1.56 (0.85 to 2.88)	0.151	5.48 (2.65 to 11.31)	<0.001	1.37 (0.69 to 2.71)	0.361	4.90 (2.23 to 10.79)	<0.001
Use of stimulants
No	Ref		Ref		Ref		Ref
Yes	2.10 (1.24 to 3.54)	0.006	0.91 (0.42 to 1.99)	0.82	2.25 (1.19 to 4.28)	0.013	1.66 (0.66 to 4.20)	0.282
Use of dissociative drugs
No	Ref		Ref
Yes	3.60 (1.11 to 11.67)	0.033	3.38 (0.77 to 14.78)	0.105
Use of cannabis
No use or not daily	Ref		Ref		Ref		Ref
Daily	3.73 (1.68 to 8.28)	0.001	2.40 (0.80 to 7.25)	0.12	4.48 (1.83 to 10.95)	0.001	3.76 (1.08 to 13.04)	0.037
HIV-related outcomes
ART interruption (≥2 days in previous month)
No	Ref		Ref		Ref		Ref
Yes	2.35 (1.38 to 4.01)	0.002	1.19 (0.55 to 2.57)	0.655	1.93 (1.06 to 3.50)	0.03	0.92 (0.39 to 2.14)	0.839
Transactional sex conditions
Length of time engaging in transactional sex (years)	1.00 (0.97 to 1.02)	0.917	1.05 (1.01 to 1.08)	0.01
Locations for sexual encounters and client-contact method
Meeting clients at participant’s home after contacting them on the internet	Ref		Ref		Ref		Ref
Meeting clients at participant’s home after contacting them using other means	1.12 (0.50 to 2.47)	0.786	1.43 (0.44 to 4.73)	0.553	1.42 (0.58 to 3.46)	0.446	1.98 (0.53 to 7.36)	0.342
Meeting clients in outdoor locations	2.52 (1.32 to 4.78)	0.005	3.61 (1.38 to 9.48)	0.009	4.00 (1.85 to 8.66)	<0.001	4.40 (1.41 to 13.72)	0.011
Meeting clients in other locations‡	0.95 (0.36 to 2.54)	0.919	2.04 (0.57 to 7.24)	0.272	1.02 (0.33 to 3.14)	0.969	1.83 (0.45 to 7.43)	0.299

^*The following variables were tested in univariate analysis but were not significantly associated with violence typologies: legal status in France, employment, gender affirmation domain use clusters.

^†Defined as PHQ-9 ≥10.

^‡In a hotel, at the client’s home, in a vehicle, in a location rented with other people engaged in transactional sex.

Multiple-perpetrator violence was significantly associated with a lack of emotional support when needed (3.1 (1.3 to 7.1)) and moderate-to-severe depression (4.9 (2.2 to 10.8)). TWLH-TS in the client and street violence and the multiple-perpetrator violence typologies were significantly more likely to engage with clients outdoors (4.0 (1.9 to 8.7) and 4.4 (1.4 to 13.7), respectively) and to use cannabis daily (4.5 (1.8 to 11.0) and 3.8 (1.1 to 13.0), respectively), compared with TWLH-TS in the Low reported violence typology.

Discussion

Violence against TWLH

Our findings reveal alarming levels of past-year violence experienced by TWLH in France. 11% of our study sample (12% of TWLH-TS and 10% of TWLH-nTS) was exposed to multiple-perpetrator violence, and 17% (26% of TWLH-TS) to client and street violence. Multiple-perpetrator violence was associated with a lack of support and mental health problems, while client and street violence was associated with precarious housing and working conditions, increased substance use and ART interruptions among TWLH-TS.

While not directly comparable, levels of violence in our study are much higher than those reported in the general population, where national data indicate that 6% of women experienced sexual violence in the past year, and 1.4% reported verbal, physical or sexual violence from an intimate partner during the same period.⁴⁵ In France, where negative attitudes and stereotypes about transgender people are prevalent,⁴⁶ an online survey conducted in 2016 indicated that, over their lifetime, 53% of transgender individuals reported experiencing psychological violence, 38% physical violence and 14% sexual violence—levels higher than those reported among cisgender LGBT people.⁴⁷ Our findings concerning the exposure of TWLH to multiple types of violence from various perpetrators—including sexual partners, family members, co-workers, law enforcement and passersby in public spaces—further underscore the systemic violence against transgender women in France in private and public spaces, including institutional violence by law enforcement.

Our results are consistent with research worldwide reporting significant exposure of TWLH to violence from different perpetrators. Specifically, compared with another Western setting, we observed similar proportions of verbal and physical violence among TWLH to those reported in a recent Canadian study, where 29% of TWLH experienced verbal violence from a client or another sexual partner, and 18% experienced physical violence (all perpetrators combined) in the previous 6 months.⁴⁸ In a Brazilian study, 41% of TWLH reported psychological, physical or sexual violence by a sexual partner.⁴⁹ The Brazilian study was conducted between May 2020 and August 2020, during the first wave of the COVID-19 pandemic and associated mobility restriction measures, whereas Trans&VIH began in August 2020, after the strict lockdown period in France, which may partly explain the higher levels of partner violence observed in the Brazilian study. Lockdown measures likely increased women’s vulnerability to intimate partner violence within cohabiting partnerships,⁵⁰ and exposure to violence from clients of those engaged in transactional sex.⁵¹

In our study, in addition to the multiple-perpetrator violence typologies, common to both TWLH-TS and TWLH-nTS, a third typology, client and street violence, emerged for TWLH-TS. These patterns of exposure among TWLH-TS are similar to those reported in a study conducted in the Dominican Republic. This study found that, although about half of TWLH-TS fell into a ‘low-violence’ typology, a significant minority (approximately one fifth) experienced multiple types of violence from multiple perpetrators, including physical and sexual violence from partners, new or existing clients and law enforcement, as well as verbal harassment by law enforcement.⁵² Another typology, affecting 31% of participants, was primarily characterised by verbal law enforcement harassment with little violence from other perpetrators. Similarly, in our study, we identified a typology of violence affecting 26% of TWLH-TS that involved a limited set of perpetrators, primarily clients and passersby. Across the two studies, it appears that a substantial proportion of TWLH-TS who otherwise do not experience violence are exposed to harm specifically in the context of transactional sex: from passersby and clients in France, and from law enforcement in the Dominican Republic. This difference likely reflects contextual factors, including the high level of law enforcement corruption in the Dominican Republic,⁵³ which may place TWLH at risk of law enforcement violence despite sex work not being illegal, whereas the penalisation of clients in France⁵⁴ potentially increases client violence against sex workers.^{22 23}

Social and health correlates of violence in TWLH-TS

Client and street violence was associated with inadequate housing conditions among TWLH-TS in our study. Economic hardship, including debt and precarious housing, may place sex workers at a higher risk of violence, because it reduces their ability to select clients and to choose the conditions of transactional sex encounters.⁴⁸ Additionally, client and street violence and multiple-perpetrator violence were both associated with meeting clients outdoors. Previous research on cisgender and transgender female sex workers found that individuals working in outdoor contexts were at higher risk of experiencing violence, not only from clients but also from law enforcement, co-workers and local residents.^55–57 However, indoor sex workers can also experience violence from clients met online,⁵⁸ and given the cross-sectional design of our study, we cannot exclude the possibility that some participants shifted to meeting clients outdoors after experiencing violence in indoor or home-based settings, which may contribute to the observed association between outdoor client meetings and exposure to violence.

The higher proportion of ART interruption in the client and street violence typology in our study is consistent with previous studies in the USA, Uganda and the Dominican Republic, highlighting associations between exposure to street, client and community-based violence and poor HIV-care outcomes among cisgender and transgender female sex workers.^59–62 The violence that TWLH experience in the context of transactional sex may have multiple adverse effects on their practical and psychological ability to regularly access and adhere to ART. Insecurity in one’s livelihood likely leads to challenges in managing one’s daily life—in terms of time, energy and resources—as well as psychological distress and trauma. This may make it difficult to maintain consistent health behaviours. Moreover, TWLH in our study from the client and street violence typology were more likely to use stimulants and to consume cannabis daily. This may reflect coping mechanisms to frequent exposure to violence experienced by TWLH engaged in transactional.^63–65

Finally, TWLH-TS in the multiple-perpetrator violence typology in our study had higher rates of depression, which reflects previous research in this population.⁵² Multiple types of violence from multiple perpetrators have cumulative adverse effects on mental health.^{66 67} In our study, one possible reason for the correlation between multiple-perpetrator violence and not receiving emotional support when needed could be that victims of violence generally have a greater, yet unmet, need for emotional support. Additionally, experiencing violence can lead to social isolation, either directly through coercion by perpetrators,⁶⁸ or indirectly, by fostering distrust and insecurity in social interactions, which in turn can trigger coping mechanisms such as social withdrawal.⁶⁹ Transgender individuals have significantly higher rates of suicide and mortality compared with the general population in European countries.^{70 71} Our findings emphasise the mental health difficulties of TWLH who are victims of multiple-perpetrator violence, and their unmet needs for support.

Our results align with the notion of the HIV syndemic, whereby multiple vulnerabilities—such as socioeconomic hardship, social exclusion, mental health issues, substance use and exposure to violence, co-occur in marginalised communities, including transgender women, and may mutually reinforce each other, leading to increased engagement in high-risk sexual behaviours and exposure to HIV, while also creating barriers to effective participation in the HIV-care cascade.^{5–7 52 72 73}

Study strengths and limitations

This study has several strengths. First, to our knowledge, it is the first French nationwide study to establish typologies of violence experienced by TWLH, based on multiple dimensions of violence and their perpetrators, and to explore mental health, social and HIV-related correlates. 86% of all TWLH who were solicited participated in our research, which provides valuable insight into experiences of violence among transgender women, both those who engage in transactional sex and those who do not.

Our study also has limitations. Some eligible TWLH may have been missed during recruitment due to missing or inaccurate information in the medical records regarding gender and sex assigned at birth. Moreover, our study is not representative of all TWLH in France as it did not include individuals not engaged in care, lost to follow-up or followed outside of hospital-based HIV-care settings—including potential victims of violence who may be less connected to care. However, in France, HIV follow-up is primarily managed in hospital-based services⁷⁴ and ANRS-TRANS&HIV was a national study with a high participation rate and a near-exhaustive sampling of TWLH identified in participating HIV care services, which strengthens the robustness of the findings. No differences in basic demographics were identified between eligible TWLH who participated in ANRS-TRANS&HIV and those who did not (region of residence, age, education level), except for a slight overrepresentation of non-French nationals among participants.¹⁶ While this may reflect a selection bias, possibly because more socially integrated individuals have less time to participate in research studies,⁷⁵ it also allowed greater inclusion of migrant TWLH—potentially more exposed to situations of vulnerability. The analysis was based on self-reported data, which may have introduced different biases, including social desirability bias,⁷⁶ particularly for sensitive topics such as transactional sex and violence. Using community-based interviewers likely helped mitigate these biases. Our questions did not evaluate the intensity or recurrence of each form of violence and may not have captured all forms or perpetrators of violence experienced by participants. Exposure to violence among TWLH may be even higher than estimated, which could mean that the true association between violence and health or social outcomes is stronger than observed, as misclassification of exposure would likely bias the effect toward the null. Finally, the cross-sectional design of ANRS-TRANS&HIV precludes causal inference. The exploratory multivariable model was selected using univariable screening, Bayesian Model Averaging and BIC to limit model complexity and account for model uncertainty. As estimates are model-dependent, adjusted ORs should be interpreted as conditional descriptive correlates rather than independent effects. These analyses were intended to describe co-occurring vulnerabilities and generate hypotheses rather than identify causal determinants. Violence experienced by TWLH who do not engage in transactional sex should be investigated in greater depth in future longitudinal studies with a larger sample of these individuals.

National and global context, future perspectives

Our study highlights the pervasive and multifaceted violence experienced by TWLH in France, revealing both common and distinct patterns among those with or without involvement in transactional sex. The abolitionist approach to sex work and the growing restrictions on migrant rights and access to health in France,^{77 78} the attacks on transgender rights and HIV science in the USA, along with major funding cuts in the global HIV response, all create a particularly threatening political context both nationally and internationally for the rights and well-being of transgender people, migrants, sex workers and people living with HIV, as well as for public health more generally ^{79 80 81 82} . Our findings further contribute to the growing body of evidence on the syndemic of violence, HIV and other vulnerabilities for transgender women. It underscores the urgent need for policy changes and comprehensive, multi-sectoral targeted interventions to address systemic violence toward TWLH in France, and its potential impact on mental health and HIV-care outcomes, an issue that is not only a public health necessity but also a crucial step toward safeguarding fundamental human rights.

Policies addressing structural determinants—such as stable housing, social and legal protections, including protection against law enforcement abuse as well as community-based support networks—are critical to reducing vulnerability to violence and improving the overall well-being of TWLH. HIV and other health services providing care to transgender women, as well as community-based organisations should implement routine screening for exposure to violence, and adopt trauma-informed care approaches to ensure sensitive, non-stigmatising interactions with TWLH, with referral to social and legal support services. Community-based outreach in outdoor transactional sex venues may help reach TWLH, particularly those engaged in transactional sex under precarious conditions or with high social vulnerability. Practical tools and interventions to prevent, respond to and protect against violence are essential. For example, the Jasmine alert and information system in France enables individuals involved in transactional sex to report incidents, receive alerts about potentially dangerous clients by verifying an email, phone number or vehicle registration number and quickly locate nearby support services.⁸³ Such tools complement routine screening and care by providing TWLH-TS with immediate, practical ways to enhance their safety and access essential resources. Evidence from studies in other countries highlights the effectiveness of multi-sectoral and peer-based interventions in reducing violence against cisgender women engaged in transactional sex and mitigating its harmful effects. These interventions—including community mobilisation combined with legal support, crisis response teams, peer outreach programmes, skills training for risk assessment, harm reduction and trauma-informed support—have been shown to enhance safety and improve access to essential services.⁸⁴ Peer-delivered, trauma-informed programmes have similarly demonstrated promise among transgender women, improving mental health, empowerment and safety and illustrating the benefits of combining peer support with targeted psychosocial care.⁸⁵

Future research should provide detailed analyses of exposure to violence across the life trajectories of transgender women, taking into account the different forms of violence, the perpetrators and the frequency and severity of exposure, and should examine its long-term health implications. Research should also assess the effectiveness of interventions aimed at reducing violence while improving socioeconomic conditions and overall quality of life for TWLH.

Data availability statement

Data may be obtained from a third party and are not publicly available.