Gender Abuse, Depressive Symptoms, and HIV and Other Sexually Transmitted Infections Among Male-to-Female Transgender Persons: A Three-Year Prospective Study

Abstract

Objectives. We examined gender abuse and depressive symptoms as risk factors for HIV and other sexually transmitted infections (HIV/STI) among male-to-female transgender persons (MTFs).

Methods. We conducted a 3-year prospective study of factors associated with incident HIV, syphilis, hepatitis B, chlamydia, and gonorrhea among 230 MTFs from the New York Metropolitan Area. Statistical techniques included Cox proportional hazards analysis with time varying covariates.

Results. Among younger MTFs (aged 19–30 years), gender abuse predicted depressive symptoms (Center for Epidemiologic Studies Depression score ≥ 20), and gender abuse combined with depressive symptoms predicted both high-risk sexual behavior (unprotected receptive anal intercourse) and incident HIV/STI. These associations were independent of socioeconomic status, ethnicity, sexual orientation, hormone therapy, and sexual reassignment surgery.

Conclusions. Gender abuse is a fundamental distal risk factor for HIV/STI among younger MTFs. Interventions for younger MTFs are needed to reduce the psychological impact of gender abuse and limit the effects of this abuse on high-risk sexual behavior. Age differences in the impact of gender abuse on HIV/STI suggest the efficacy of peer-based interventions in which older MTFs teach their younger counterparts how to cope with this abuse.

Extremely high rates of HIV have been detected among male-to-female transgender persons (MTFs). Community-based studies using nonprobability sampling have observed an HIV prevalence of 22% to 35%, with yearly incidence rates ranging from 3.5% to 7.8%.^1–9 Established risk factors for HIV in this population include ethnicity (African Americans and Hispanics compared with Whites) and sexual orientation (those attracted to men only compared with other categories of sexual attraction).^1,4

The dominant behavioral mode by which MTFs contract HIV and transmit the virus to others, including the general population,¹⁰ is unprotected receptive anal intercourse (URAI) with committed, casual, or commercial partners.¹¹ MTFs report frequencies of high-risk sexual behavior (including URAI) with noncommercial and commercial sex partners that are much higher than those for the general population¹² and higher than for sexual minorities.¹³

A potentially significant proximal risk factor, which may combine with URAI to cause HIV in this population, is depressed affect. MTFs report levels of depressive symptomatology that are much higher than in the general population,^14,15 and some previous studies suggest that depressive symptoms are intertwined with high-risk sexual behavior.^16–20

A more fundamental distal risk factor, which may cause depressive symptoms and ultimately high-risk sexual behavior and HIV among MTFs, is abuse associated with an atypical presentation of gender. Because they transgress basic gender norms, many MTFs are taunted or beaten by family members, neighbors, coworkers, strangers, or the police,²¹ and a recent study showed that this abuse is linked to depressive symptomatology.²² For some MTFs, gender abuse may be highly traumatizing and intertwined with depressed affect, which may erode prevention consciousness and the use of condoms to prevent HIV in particular.^23–27

We present the findings of a community-based prospective study designed to evaluate social, psychological, and behavioral risk factors for incident HIV and other sexually transmitted infections (HIV/STI) among MTFs. We tested 3 interrelated hypotheses, reflecting the literature reviewed here: (1) gender abuse is associated with depressive symptoms as measured by the Center for Epidemiologic Studies Depression scale (CES-D); (2) gender abuse is associated with URAI with committed, casual, and commercial partners, with the effects partially mediated by depressive symptoms; and (3) gender abuse is associated with incident HIV/STI, with the effects partially mediated by depressive symptoms and URAI.

Psychological or physical abuse associated with an atypical presentation of gender (gender abuse) is the result of a binary gender system in which all individuals are expected to conform to a single gender role (male or female) consistent with their sexual anatomy at birth.²³ Although it is subjectively experienced, this abuse is ultimately the product of social forces beyond the perceptions of individuals²⁸; following the Institute of Medicine’s conceptualization,²⁹ we posited that it was a distal social risk factor for HIV/STI in this population. We conceptualized depressive symptoms as a proximal psychological risk factor for URAI. We included URAI as a behavioral risk factor that directly causes HIV/STI. We hypothesized that gender abuse and depressive symptoms affect HIV/STI via their effects on URAI.

We further hypothesized that these associations are modified by age. Some studies suggest that younger MTFs are particularly vulnerable to gender abuse and victimization.^30–33 Other studies suggest that older MTFs, after years of coping to this adversity, develop attitudes and skills to better cope with it.^29,34 Age differences in vulnerability to gender abuse were demonstrated in a recent retrospective study by our research team: gender abuse was strongly associated with depressive symptomatology during adolescence and early adulthood, but the strength of this association declined markedly during later stages of life.²² Building on this finding, we examined age differences in the effects of gender abuse on depressive symptom longitudinally in this study, with further predictions of age differences in the effects of gender abuse on URAI and HIV/STI.

An analysis of HIV among MTFs must also recognize the fact that this is a diverse population with regard to socioeconomic status (SES) and stage of gender transition,²⁹ both of which could confound observed associations between gender abuse and HIV/STI. Because of this potential confounding, we included indicators of SES (education and income) and variation along a spectrum of gender transition (hormone therapy, preoperational transsexual identity, and sexual reassignment surgery) in the analysis.

METHODS

Transgender or gender-variant individuals were actively involved in all aspects and phases of the research study, including the design of the instrument, data collection, data analysis, and dissemination of the findings.

Selection of Study Participants

We recruited 591 study participants for a community-based retrospective–prospective study of MTFs in the New York Metropolitan Area. Approached individuals were eligible for the study if they were assigned as “male” at birth but subsequently did not regard themselves as “completely male” in all situations or roles (reflecting an MTF–transgender spectrum). Eligibility criteria also included age 19 years or older and the absence of psychotic ideation.

Study participants were recruited via transgender organizations in the New York Metropolitan Area (e.g., Society for the Second Self [or TRIESS], Cross Dressers International, and the Mid Hudson Valley Transgender Association), the Internet, newspaper advertisements, the streets, clubs, client referrals of other clients, and paid assistants from transgender communities who worked on a day-to-day basis with the field staff. Recruitment initially emphasized transgender organizations (which disproportionately included older Whites with lower HIV risk behavior), with later recruitment relying more on study assistants from transgender communities (which included more younger non-Whites with higher HIV risk behavior).

From the 591 MTFs in the retrospective component of the study, 40.1% (n = 237) and 59.9% (n = 354) biologically assayed as HIV positive and HIV negative, respectively. From the latter HIV negatives, we included 230 in the prospective component of the study designed to evaluate risk factors for new cases of HIV/STI.

To increase the efficiency of the prospective research design, we oversampled individuals with high-risk behavior (younger age and recent URAI). Those older than 30 years and those with no recent unprotected anal sex were selected with a 50% probability; those who were aged 30 years or younger and those who reported recent unprotected anal sex were selected with a 100% probability. Since the initial sample was nonrandomly selected, with a possible overweighting of older MTFs and those with lower HIV risk behavior, we did not incorporate data analytic corrections for an oversampling of younger MTFs and those with higher risk behavior in Results.

Because of time constraints in this 5-year study, there was variation in the number of years study participants could potentially be followed. We extended the recruitment phase, which began in December 2004, to September 2007 so that all participants could potentially be followed for at least 1 year. From the initial pool of 230 HIV negatives included in the prospective component, we followed and interviewed 171 (74.3%), 92 (40.0%), and 56 (24.3%) at years 1, 2, and 3, respectively. The percentages of potentially available study participants that we reinterviewed and biologically tested for new HIV/STI were 171 of 230 (74.3%), 92 of 135 (68.1%), and 56 of 74 (75.7%) at years 1, 2, and 3, respectively. Of those followed across the yearly assessment points, we tested more than 98.5% for HIV/STI, with appropriate counseling.

Measurements of Client Characteristics and Risk Factors

At baseline, study participants completed face-to-face interviews in conjunction with the Life Review of Transgender Experiences (LRTE), an instrument designed to collect biographical information about transgender experiences, including HIV risks, since early adolescence.⁴ At yearly assessment points after baseline, for up to 3 years, the subset of 230 study participants included in the prospective study completed follow-up versions of the LRTE designed to document short-term changes in lifestyle factors and HIV risks. The English version of the LRTE was fully translated into Spanish, and 19% (44 of 230) were interviewed in Spanish by a fluent interviewer. Study participants were compensated $40 for completing all of the protocols associated with a given assessment period.

In part of the analysis (to better interpret and compare hazard rate with HIV/STI), we dichotomized and reverse coded age, with younger age (19–30 years) compared with older age (31–59). We determined ethnicity using preestablished census categories. In part of the analysis, this was grouped as non-Hispanic White versus all other categories (minority ethnicity). We classified education as less than high school, high school graduate, some college, and college graduate or higher. We classified legitimate income during the prior 6 months as $1999 or less, $2000 to $9999, $10 000 to $29 999, and $30 000 or more. We categorized sexual orientation as attraction to men only, women only, men and women, and neither men nor women. In some of the analysis, this was grouped as sexual attraction to men only versus all other categories.

We defined hormone therapy as any lifetime use of female hormone supplements. Self-identification as a preoperative transsexual reflected whether the term “pre-op” was selected as an appropriate description of identity. We coded sexual reassignment surgery as having undergone any type of surgery designed to transform sexual anatomy.

High-risk sexual behavior included the number of episodes of receptive anal intercourse (RAI) with committed, casual, and commercial partners (assessed separately) during the prior 6 months, and whether any of these episodes was unprotected (URAI). We quintiled the right-skewed frequencies of RAI into roughly equal categories. URAI (emphasized in the analysis below) was classified as none or any.

We measured depressive symptoms with the widely used 20-item CES-D, which assesses depressive symptoms experienced during the prior week and has a theoretical range of 0 to 60.³⁵ Alpha reliabilities of the CES-D items ranged from 0.92 at baseline to 0.97 at the 2-year assessment point. Rest–retest correlations of the CESD across assessment points were 0.44 (baseline to year 1), 0.38 (year 1 to year 2), and 0.31 (year 2 to year 3). In some of the analyses, we dichotomized depressive symptoms using a cut-score of 20, which has been shown to predict clinical depression in high-risk populations.³⁶

We queried study participants about whether they were “verbally abused or harassed” (psychological abuse) and whether they thought it was because of their gender identity or presentation during the prior 6 months. A parallel item asked about being “physically abused or beaten” (physical abuse). An index of gender abuse during the prior 6 months was scored as neither psychological nor physical abuse, either psychological or physical abuse, or psychological and physical abuse. Rest–retest correlations of gender abuse across assessment points were 0.40 (baseline to year 1), 0.29 (year 1 to year 2), and 0.27 (year 2 to year 3).

Biological Assays

Laboratory tests for HIV/STI, performed by Bendiner Schlesinger (Brooklyn, NY), were conducted at baseline for the broader pool of study participants, and at yearly intervals thereafter for up to 3 years for those included in the prospective component. We defined incident HIV as newly observed HIV antibodies using an enzyme immunoassay screen with a Western blot confirmation.

We included non-HIV STIs as biological markers for HIV infection.³⁷ Syphilis was determined with a rapid plasma reagin screen and a fluorescent treponema antibody confirmation. We defined incident syphilis as a newly observed infection after 1, 2, or 3 years of follow-up (suggesting an initial infection) or a doubling of the titer ratio during 12 months of follow-up (suggesting a reinfection). We defined incident hepatitis B by the presence of newly observed surface antigens. We detected recent untreated exposures to chlamydia and gonorrhea by DNA amplification using ligase chain reaction of urine specimens. We defined incident cases as an absence of recent treatment or laboratory indication of these bacterial infections followed by reported treatment or laboratory confirmation.

We removed those found to be HIV positive after 12 or 24 months of follow-up from further study participation and referred them for HIV treatment. Those found to be infected with a non-HIV STI were referred for treatment but not removed from the study.

Statistical Techniques and Modeling

We analyzed data with Cox proportional hazards analysis implemented with Release 9 of Stata Statistical Software (StataCorp LP, College Station, TX).³⁸ We estimated hazard ratios with 95% confidence intervals using standard errors calculated with jackknife resampling.³⁹ The Cox analysis is based on time until the occurrence of identified outcome events, defined here as a high score on CES-D (≥ 20); URAI with committed, casual, or commercial partners; and HIV/STI. The event history analysis incorporated the possibility of multiple occurrences of events across time (high CES-D score, URAI, or HIV/STI). To estimate overall effects, we conducted the analyses using the total sample. To fully describe the hypothesized age differences, we conducted separately age-stratified analyzes among the younger (aged 19–30 years) and older (aged 31–62 years) study participants. We examined confounding associated with ethnicity, sexual orientation, SES, and stage of gender transition only for the associations of gender abuse with HIV/STI.

We coded gender abuse in 3 ways. “Lagged” abuse referred to abuse experienced the prior year. “Contemporaneous” abuse was abuse suffered that same year. “Changed” abuse was an increase in abuse from the prior to the current year. All the measurements at baseline were necessarily “contemporaneous.”

We tested hypothesis 1 as the effects of gender abuse on high CES-D score. We evaluated hypothesis 2 as the effects of gender abuse and CES-D score on URAI with committed, casual, and commercial partners. This analysis included only the contemporaneous measurement of gender abuse, with a continuous measurement of CES-D score included as a time-varying covariate. We estimated the main effects of gender abuse and depressive symptoms on URAI, and also created mediation models in which gender abuse affected URAI because of the intervening effects of depressive symptoms. We assumed such mediation if the bivariate associations of gender abuse and URAI were reduced by more than 10% with CES-D score controlled for.⁴⁰

We assessed hypothesis 3 as the effects of background characteristics, gender abuse, CES-D score (continuous measurement), and high-risk sexual behavior (RAI and URAI) on HIV/STI. We examined bivariate effects followed by the effects of gender abuse on HIV/STI, controlling for statistically significant background characteristics (to assess and control confounding), CES-D score, and high-risk sexual behavior (to assess mediation).

RESULTS

We compared the subsets of study participants followed at years 1, 2, and 3 with those not so followed with regard to baseline measurements of the HIV/STI risk factors described in Methods and other variables included in the data set. Two of these correlations were statistically significant: age with study completion at years 1 (r = 0.15; P ≤ .05) and 3 (r = 0.16; P ≤ .05).

Study Participants

Study participants were aged 19 to 62 years (mean = 34.0), with 43.5% between 19 and 29 years, 23.8% between 30 and 39 years, 17.9% between 40 and 49 years, and 14.8% between 50 and 62 years (Table 1). Ethnicity was 35.2% non-Hispanic White, 35.7% Hispanic, 17.4% non-Hispanic Black, and 11.7% other. Most (58.9%) were attracted to men only, 25.4% to women only, 13.8% to men and women, and 1.8% to neither men nor women.

TABLE 1—

Baseline Description of Male-To-Female Transgender Study Participants: New York Metropolitan Area, 2004–2007

Variable	%
Age, y
19–29	43.5
30–39	23.8
40–49	17.9
50–62	14.8
Ethnicity
Non-Hispanic White	35.2
Hispanic	35.7
Non-Hispanic Black	17.4
Other	11.7
Sexual orientation
Attracted to men only	58.9
Attracted to women only	25.4
Attracted to men and women	13.8
Not attracted to men or women	1.8
Education
< high school graduate	24.8
High school graduate	27.4
Some college	27.4
≥ college graduate	20.4
Legitimate income, $
≤ 1999	16.8
2000–9999	34.4
10 000–29 999	22.9
≥ 30 000	26.0
Hormone replacement therapy	67.9
Preoperative transsexual	34.4
Sexual reassignment surgery	7.0
Gender abuse
0 (none)	45.2
1 (psychological or physical)	46.5
2 (psychological and physical)	8.3
CES-D score ≥ 20	63.0
URAI partner
Committed	20.0
Casual	13.5
Commercial	5.1

Note. CES-D = Center for Epidemiologic Studies Depression scale; URAI = unprotected receptive anal intercourse. For baseline sample, n = 230.

Education ranged from 24.8% who were not high school graduates to 20.4% who were college graduates or higher. Legitimate income from all sources during the prior 6 months ranged from 16.8% who reported less than $1999 to 26.0% who reported $30 000 or more. About two thirds (67.9%) reported lifetime hormone therapy, 34.4% identified as a preoperative transsexual, and 7.0% indicated prior sexual reassignment surgery.

Risk Factors

At baseline, 45.2% reported no gender abuse, 46.5% reported either physical or physical gender abuse, and 8.3% reported both psychological and physical gender abuse. At this time point, 63.0% scored 20 or higher on the CES-D, which is a cutpoint consistent with clinical depression³⁶ (mean CES-D score = 24.0; SD = 12.0; empirical range = 5–50). At baseline, URAI with committed, casual, and commercial partners was 20.0%, 13.5%, and 5.1%, respectively.

Incidence

Forty new cases (17.4%) of HIV/STI were observed during the 3-year follow-up. The overall incidence of HIV across the yearly assessment points was 2.8% (9 new cases over 316 person-years of follow-up). The year-specific rates for HIV were 3.0% during year 1, 3.3% during year 2, and 1.8% during year 3. New infections of syphilis were 3.6% during year 1, 1.1% during year 2, and 1.8% during year 3. Hepatitis B was 2.4% during year 1, 6.3% during year 2, and 5.5% during year 3. Gonorrhea declined from 4.2% during year 1 to 0.0% during year 3. Chlamydia declined from 4.5% during year 1 to 1.1% during year 3.

Gender Abuse as a Predictor of Depressive Symptoms

As shown in Table 2, the contemporaneous (hazard ratio [HR] = 3.01; 95% confidence interval [CI] = 2.24, 4.05) and change (HR = 3.23; 95% CI = 2.03, 5.14) measurements of gender abuse were predictive of depressive symptoms as measured by a high CES-D score. Among younger MTFs, the lagged (HR = 1.54; 95% CI = 1.03, 2.30), contemporaneous (HR = 2.94; 95% CI = 1.76, 4.94), and change (HR = 5.72; 95% CI = 1.73, 19.01) measurements were consistently and more strongly predictive of high CES-D score. The contemporaneous (HR = 2.94; 95% CI = 1.69, 5.22) and change (HR = 2.08; 95% CI = 1.19, 3.63) measurements were, for the most part, less strongly predictive of high CES-D score among older MTFs. In sum, the hypothesized association of gender abuse with depressive symptoms was supported by the expected differences in the strength and consistency of this association between age groups.

TABLE 2—

Gender Abuse as a Predictor of Depressive Symptoms Among Male-To-Female Transgender Persons During 3 Years of Follow-Up: New York Metropolitan Area, 2004–2007

	Depressive Symptoms
Gender Abusea	Scale Range	HR (95% CI)
Total Sample (n = 230)
Lagged	0–2	1.31 (0.99, 1.70)
Contemporaneous	0–2	3.01 (2.24, 4.05)
Change	0–1	3.23 (2.03, 5.14)
Aged 19–30 y (n = 100)
Lagged	0–2	1.54 (1.03, 2.30)
Contemporaneous	0–2	2.96 (1.76, 4.94)
Change	0–1	5.72 (1.73, 19.01)
Aged 31–62 y (n = 130)
Lagged	0–2	1.16 (0.81, 1.67)
Contemporaneous	0–2	2.94 (1.69, 5.22)
Change	0–1	2.08 (1.19, 3.63)

Note. CI = confidence interval; HR = hazard ratio. Depressive symptoms were defined as a score of ≥ 20 on the Center for Epidemiologic Studies Depression scale.

^a“Lagged” abuse was experienced the prior year, whereas “contemporaneous” abuse was suffered that same year. “Change” was an increase in abuse from the prior to the current year. All the measurements at baseline were necessarily “contemporaneous.”

Predictors of Unprotected Receptive Anal Intercourse

As shown in Table 3, gender abuse was associated with an increased rate of URAI with committed (HR = 1.79; 95% CI = 1.21, 2.65), casual (HR 2.55; 95% CI = 1.73, 3.75), and commercial (HR = 3.72; 95% CI = 2.27, 6.09) partners. Higher depressive symptoms (continuous measurement on the CES-D scale) increased the rate of URAI with committed (HR = 1.03; 95% CI = 1.01, 1.06), casual (HR = 1.05; 95% CI = 1.03, 1.07), and commercial (HR = 1.07; 95% CI = 1.05, 1.10) partners. The associations of gender abuse and URAI with different types of partners was reduced by about 40% with CES-D score controlled for (mediation effect). This pattern of effects was generally found in the age-stratified analysis, with the effects of gender abuse and CES-D score on URAI, for the most part, stronger among the younger respondents (especially with regard to causal and commercial partners).

TABLE 3—

Gender Abuse and Depressive Symptoms as Predictors of Unprotected Receptive Anal Intercourse (URAI) With 3 Types of Partners Among Male-To-Female Transgender Persons During 3 Years of Follow-Up: New York Metropolitan Area, 2004–2007

	URAI by Type of Partner
	Committed, HR (95% CI)	Casual, HR (95% CI)	Commercial, HR (95% CI)
Total Sample (n = 230)
Depressive symptoms	1.03 (1.03, 1.06)	1.05 (1.03, 1.07)	1.07 (1.05, 1.10)
Gender abuse	1.79 (1.21, 2.65)	2.55 (1.73, 3.75)	3.72 (2.27, 6.09)
Depressive symptoms controlled for	1.13 (0.71, 1.81)	1.57 (1.01, 2.44)	1.88 (0.99, 3.57)
Aged 19–30 y (n = 100)
Depressive symptoms	1.04 (1.02, 1.06)	1.08 (1.05, 1.11)	1.11 (1.07, 1.14)
Gender abuse	1.56 (0.96, 2.54)	3.46 (2.14, 5.62)	4.39 (2.37, 8.13)
Depressive symptoms controlled for	0.97 (0.47, 1.97)	1.88 (0.93, 3.80)	1.80 (0.79, 4.06)
Aged 31–62 y (n = 130)
Depressive symptoms	1.05 (1.02, 1.07)	1.04 (1.02, 1.07)	1.06 (1.00, 1.11)
Gender abuse	2.08 (1.07, 4.05)	1.85 (0.96, 3.58)	2.99 (1.25, 7.14)
Depressive symptoms controlled for	1.25 (0.59, 2.65)	1.20 (0.63, 2.30)	1.65 (0.60, 4.51)

Note. CES-D = Center for Epidemiologic Studies Depression scale; CI = confidence interval; HR = hazard ratio. Depressive symptoms were defined as a score of ≥ 20 on the CES-D. The table shows Cox proportional hazards analysis predicting time until any unprotected receptive anal intercourse with different types of partners, with contemporaneous measurement of CES-D score and gender abuse incorporated as time-varying covariates.

In sum, the hypothesized associations of gender abuse and depressive symptoms with URAI, and the predicted age differences in the strength of these associations, were broadly supported by the results.

Predictors of HIV and Other Sexually Transmitted Infections

Associations of the background variables, high-risk sexual behavior, CES-D score, and gender abuse with HIV/STI are summarized in Table 4.

TABLE 4—

Predictors of Incident HIV or Other Sexually Transmitted Infection (STI) in the Total Sample of Male-To-Female Transgender Persons: New York Metropolitan Area, 2004–2007

Variable	Incident HIV/STI, HR (95% CI)
Background
Education	0.78 (0.53, 1.16)
Legitimate income	0.75 (0.54, 1.03)
Hormone therapy	1.73 (0.77, 3.91)
Preoperative transsexual	1.47 (0.76, 2.85)
Sexual reassignment surgery	0.69 (0.10, 4.56)
Younger age	2.28 (1.01, 5.35)
Non-White ethnicity	3.37 (1.34, 8.46)
Sexually attracted to men only	4.96 (1.79, 13.78)
High-risk sexual behavior
Committed partners
RAI	1.55 (1.06, 2.24)
URAI	2.23 (1.06, 4.70)
Casual partners
RAI	2.60 (1.85, 3.66)
URAI	2.02 (1.03, 4.48)
Commercial partners
RAI	3.10 (2.09, 4.61)
URAI	3.42 (1.21, 9.66)
CES-D score ≥ 20	1.05 (1.03, 1.07)
Gender abusea
Lagged	1.64 (1.02, 2.63)
Change	2.62 (1.14, 6.01)
Contemporaneous	2.44 (1.58, 3.76)
Variables controlled For
Non-White ethnicity	2.25 (1.48, 3.43)
Attracted to men only	2.28 (1.47, 3.55)
High-risk sexual behavior	1.53 (0.94, 2.51)
CES-D score ≥ 20	1.58 (0.90, 2.78)

Note. CES-D = Center for Epidemiologic Studies Depression scale; CI = confidence interval; HR = hazard ratio; RAI = receptive anal intercourse; URAI = unprotected receptive anal intercourse. For baseline sample, n = 230.

^a“Lagged” abuse was experienced the prior year, whereas “contemporaneous” abuse was suffered that same year. “Change” was an increase in abuse from the prior to the current year. All the measurements at baseline were necessarily “contemporaneous.”

Background factors.

Education (HR = 0.78; 95% CI = 0.53, 1.16), income (HR = 0.75; 95% CI = 0.54, 1.03), hormone therapy (HR = 1.73; 95% CI = 0.71, 3.91), identification as a preoperative transsexual (HR = 1.47; 95% CI = 0.67, 2.85), and sexual reassignment surgery (HR = 0.69; 95% CI = 0.10, 4.56) were not associated with HIV/STI.

Age coded as a continuous variable from 19 through 62 years (HR = 0.96; 95% CI = 0.93, 0.99) and coded as younger (aged 19–30 years) versus older (aged 31–62 years; HR = 2.28; CI = 1.01, 5.35) was associated with HIV/STI, as was sexual orientation (attracted to men only vs other categories; HR = 4.96; CI = 1.78, 13.78). Compared with those sexually attracted to men only (reference category), those attracted to women only (HR = 0.12; 95% CI = 0.02, 0.91) and those attracted to both men and women (HR = 0.20; 95% CI = 0.06, 0.66) were less likely to become HIV/STI infected (the small numbers of those attracted to neither men nor women were excluded from this analysis).

We detected an association between non-White (minority) ethnicity and HIV/STI (HR = 3.37; 95% CI = 1.34, 8.46). Compared with non-Hispanic Whites (reference group), Hispanics (HR = 3.43; 95% CI = 1.30, 9.02), non-Hispanic Blacks (HR = 2.48; 95% CI = 1.02, 7.81), and others (HR = 3.67; 95% CI = 1.07, 11.53) were more likely to become HIV/STI infected. A multivariate model of ethnicity and sexual orientation as predictors of HIV/STI showed that sexual orientation remained significant (HR = 4.10; 95% CI = 1.35, 12.40), but ethnicity was no longer significant (HR = 1.27; 95% CI = 0.45, 3.53).

High-risk sexual behavior and depressive symptoms.

RAI (HR = 1.55; CI = 1.06, 2.24) and URAI (HR = 2.23; CI = 1.06, 4.70) with committed partners were associated with HIV/STI. RAI (HR = 2.60; CI = 1.85, 3.66) and URAI (HR = 2.02; CI = 1.03, 4.48) with casual partners were more strongly associated with HIV/STI. RAI (HR = 3.10; CI = 2.09, 4.61) and URAI (HR = 3.42; CI = 1.21, 9.66) with commercial partners were still more strongly associated with HIV/STI.

Depressive symptoms (high CES-D score; continuous scale) predicted new HIV/STI (HR = 1.05; CI = 1.03, 1.07).

Gender abuse.

Lagged (HR = 1.64; 1.02, 2.63), change (HR = 2.62; CI = 1.14, 6.01), and contemporaneous (HR = 2.44; CI = 1.58, 3.76) measures of gender abuse were consistently predictive of HIV/STI. The observed effects of contemporaneous gender abuse on HIV/STI were little changed when either ethnicity (HR = 2.25; CI = 1.48, 3.43) or sexual orientation (HR = 2.28; CI = 1.47, 3.55) was controlled for (suggesting a lack of confounding). As expected, the effects of gender abuse on new HIV/STI were much reduced (and no longer statistically significant) with high-risk sexual behavior controlled for (abuse affects infection resulting from sexual behavior) (HR = 1.53; 95% CI = 0.94, 2.51). Finally, the effects of gender abuse on HIV/STI were reduced by about 40% with CES-D score controlled for (HR = 1.58; CI = 0.90, 2.78; mediation effect).

Age Differences as Predictors

As predicted, the effect of gender abuse on HIV/STI was significantly stronger among younger than among older MTFs. With the main effects of age and gender abuse included, the interaction product of younger age multiplied by gender abuse was a statistically significant predictor of HIV/STI (interaction term HR = 2.82; CI = 1.06, 7.64).

The age-stratified analysis of gender abuse and CES-D score as predictors of a new HIV/STI is summarized in Table 5. Among younger MTFs, contemporaneous gender abuse (HR = 3.37; CI = 2.12, 5.36) and CES-D score (5–50) (HR = 1.06; CI = 1.03, 1.09) were both associated with HIV/STI. In this age group, the association between gender abuse and HIV/STI was reduced by about 40% when CES-D was controlled for (mediation effect). Among older MTFs, the associations of gender abuse and HIV/STI were not significant statistically (HR = 1.10; 95% CI = 0.44, 2.73).

TABLE 5—

Age-Stratified Analysis of Gender Abuse and Depressive Symptoms as Predictors of Incident HIV or Other STI Among Male-To-Female Transgender Persons: New York Metropolitan Area, 2004–2007

Variable	Incident HIV/STI, HR (95% CI)
Aged 19–30 y (n = 100)
Gender abuse	3.37 (2.12, 5.36)
Depressive symptoms	1.06 (1.03, 1.09)
Gender abuse, controlling for depressive symptoms	2.22 (1.05, 4.73)
Aged 31–62 y (n = 130)
Gender abuse	1.10 (0.44, 2.73)
Depressive symptoms	1.04 (1.02, 1.06)
Gender abuse, controlling for depressive symptoms	0.67 (0.26, 1.74)

Note. CI = confidence interval; HR = hazard ratio; STI = sexually transmitted infection. Depressive symptoms were defined as a score of ≥ 20 on the Center for Epidemiologic Studies Depression scale.

In sum, the hypothesized associations of gender abuse and depressive symptoms with incident HIV/STIs, and the predicted age differences in the strength of these associations, were consistently supported by the results.

DISCUSSION

The 2.8% yearly incident rate for HIV was extremely high but less than the 3.5% to 7.8% yearly rates for HIV observed in some prior studies of this population.^2,8,9 The downturn in the HIV rate during year 3 (1.8%) suggests some decline in risk behavior as a result of multiyear study participation (repeated interviewing and counseling) and an artificially low estimate of HIV incidence. By contrast, the oversampling for high-risk sexual behavior may have produced an HIV incidence that was artificially high compared with that of the broader MTF population. The HIV rate computed here, although perhaps biased, was nonetheless extremely high. The incidence rates for syphilis, hepatitis B, gonorrhea, and chlamydia were also high. These findings underscore the continuing critical need for improved HIV/STI prevention aimed at this population.

Background Factors

An early study of MTFs in San Francisco reported a much higher prevalence of HIV among Hispanics and African Americans compared with Whites,¹ as did a later study in the New York Metropolitan Area.⁴ In the latter study, White versus non-White differences in known HIV risk factors were largely explained by a small set of factors distinguishing these ethnic categories (especially sexual orientation). In the current prospective study, the difference between Whites and non-Whites in incident HIV/STI was not significant with sexual orientation controlled for. This suggests that HIV prevention should be targeted at younger MTFs who are sexually attracted only to men (most of whom are non-White).

High-Risk Sexual Behavior and Depressive Symptoms

This study showed that URAI with all types of partners—committed, casual, and commercial—is associated with incident HIV/STI among MTFs. Increased episodes of RAI may increase the likelihood of 1 or more of these episodes being unprotected.

An association between negative affect and HIV risk behavior or infection has been inconsistently observed over the years.^41,42 One review concluded that this association is best examined with a prospective study in a well-defined population with a clearly defined measure of negative affect.⁴³ The current prospective study of MTFs suggests that depressed affect (high CES-D score) is indeed predictive of both high-risk sexual behavior and a new HIV/STI.

Gender Abuse

Gender abuse among younger MTFs apparently causes both depressive symptoms and high-risk sexual behavior, which then cause HIV/STI. Younger MTFs would appear to be extremely vulnerable psychologically to gender abuse. This may reflect their limited coping skills for blunting the effects of gender abuse on their emotional well-being. Older MTFs, by contrast, would appear to be comparatively resilient psychologically to gender abuse. This may reflect their improved coping skills for blunting the effects of gender abuse on their emotional well-being. The psychological vulnerability of younger MTFs to gender abuse apparently causes them to engage in high-risk sexual behavior and ultimately become HIV/STI infected. Younger MTFs who are depressed and abused may not possess the ability or inclination to negotiate condom use. By contrast, older MTFs who are depressed or abused may possess a higher ability and inclination to negotiate condom use.

Conclusions

Even with developments in preexposure prophylaxis for HIV, future interventions will need to contain elements aimed at reducing demonstrated risk factors.⁴⁴ Future HIV interventions among MTFs should be designed in part to confront gender abuse. The lower incidence of HIV/STI infection among older MTFs (aged 30–62 years), combined with their apparent skills for dealing with gender abuse, suggest that they may be enlisted to teach their younger counterparts what they have learned about coping with this abuse. The feasibility of such an intervention is suggested by a qualitative study showing that older “trans mothers,” who have gained experience about practicing HIV prevention in the context of social and economic adversity, are eager to share their experiences with younger MTFs.⁴⁵

Our findings, and their interpretations, should be understood in the context of the limitations of this study. The sample of MTFs used in this study, although diverse, was not randomly selected, and STIs were used as biological markers or proxies for HIV infection. Moreover, causal inferences about risk factors for HIV/STIs were necessarily limited by the nonexperimental research design.