The Effect of Psychosocial Syndemic Production on 4-Year HIV Incidence and Risk Behavior in a Large Cohort of Sexually Active Men Who Have Sex With Men

Abstract

Background: 

Cross-sectional studies have suggested that co-occurring epidemics or “syndemics” of psychosocial health problems may accelerate HIV transmission among men who have sex with men (MSM) in the United States. We aimed to assess how 5 syndemic conditions (depressive symptoms, heavy alcohol use, stimulant use, polydrug use, and childhood sexual abuse) affected HIV incidence and sexual risk behavior over time.

Methods: 

Eligible men in a large prospective cohort of sexually active HIV-uninfected MSM completed HIV testing and behavioral surveys at baseline and every 6 months for 48 months. We examined interrelationships between psychosocial problems and whether these interactions increased the odds of HIV risk behaviors and risk of seroconversion over study follow-up.

Results: 

Among 4295 men, prevalence of psychosocial conditions was substantial at baseline and was positively associated with each other. We identified a statistically significant positive dose–response relationship between numbers of syndemic conditions and HIV seroconversion for all comparisons (with the greatest hazard among those with 4–5 conditions, adjusted hazard ratio = 8.69; 95% confidence interval: 4.78 to 15.44). The number of syndemic conditions also predicted increased HIV-related risk behaviors over time, which mediated the syndemic–HIV seroconversion association.

Conclusions: 

The accumulation of syndemic psychosocial problems predicted HIV-related sexual risk behaviors and seroconversion in a large sample of US MSM. Given the high prevalence of syndemic conditions among MSM and the moderate effect sizes attained by traditional brief behavioral interventions to date, the HIV prevention agenda requires a shift toward improved assessment of psychosocial comorbidities and stronger integration with mental health and substance abuse treatment services.

INTRODUCTION

Despite the development of empirically grounded behavioral interventions to reduce HIV transmission, incidence continues to increase among men who have sex with men (MSM) in the United States.¹ Most HIV prevention interventions focus on reducing specific sexual risk behaviors; yet, MSM experience disproportionate burdens of depression, victimization, and distress^2–4 that promote sexual risk taking.^5–7 Several of these highly prevalent psychosocial problems also interact with substance abuse to produce additive or multiplicative effects on sexual risk.^8,9 Over the past decade, research has suggested that these co-occurring epidemics among MSM may be intertwined, representing “syndemics” that accelerate HIV transmission.^10,11

The term syndemic was conceived to describe co-occurring and mutually reinforcing epidemics of substance abuse, violence, and AIDS in poor urban communities.¹² Although traditional epidemiological approaches conceptualize diseases as distinct entities, diseases often present in clusters and interact with each other to affect the health of particular groups in unique and detrimental ways.¹³ Syndemic theory emphasizes how intertwined epidemics arise from adverse social conditions (eg, poverty, social marginalization, political oppression) and interact to adversely affect health outcomes within disadvantaged communities. Recognizing the utility of this perspective in understanding HIV disparities, researchers have increasingly focused on identifying associations between co-occurring psychosocial problems and HIV risk taking among MSM and other sexual minority populations who are subject to high levels of social pressure, marginalization, and stress.^4,11,14

While investigating syndemic factors in a cross-sectional sample of adult MSM, Stall et al¹¹ documented that greater numbers of psychosocial conditions were significantly associated with an increased odds of high-risk sexual behaviors and already having HIV infection. Mustanski et al¹⁵ identified a similar phenomenon among younger MSM,¹⁰ recently demonstrating the consequences of syndemics to be more severe for sexual minority youth than those in the majority heterosexual population. In addition to depression, victimization, and substance abuse, additional cross-sectional analyses have provided evidence that sexual compulsivity, stimulant use, intimate partner violence, stress, and childhood sexual abuse also interact to increase HIV risk among MSM.^5,16–18 To improve the efficacy of behavioral interventions to prevent HIV among US MSM, which have only achieved small to moderate effect sizes to date,^19,20 a better understanding of how syndemics promote HIV risk among MSM over time could lead to the development of improved interventions that adequately target underlying interacting psychosocial conditions and the broader social contexts from which they originate.

Despite this expanding evidence, research is needed to examine the syndemic production of HIV longitudinally. Thus, we drew from of large prospective cohort of sexually active HIV-uninfected MSM enrolled in a behavioral intervention in the United States to assess the interplay between multiple psychosocial health problems and determine the extent to which these syndemic conditions affected HIV seroconversion (becoming infected with HIV) over study follow-up and intervention efficacy. Based on previous work supporting the centrality of syndemics in heightening HIV risk among MSM,^5–7 we hypothesized that 5 specific psychosocial health conditions would interact over time (depressive symptoms, heavy alcohol use, stimulant use, polydrug use, and childhood sexual abuse) to increase sexual risk behavior and HIV seroconversion. We also sought to identify the most relevant syndemic psychosocial permutations with the overall goal of refining current HIV prevention intervention targets.

METHODS

Data originated from the EXPLORE study (HIVNET 015; previously detailed), a randomized controlled trial of a counseling intervention to prevent HIV acquisition among MSM in Boston, Chicago, Denver, New York, San Francisco, and Seattle enrolled from January 1999 to February 2001.²¹ Institutional review boards at each institution approved study procedures.

Study Population

Eligible men were negative for HIV antibodies, ≥16 years old, and reported having anal intercourse with ≥1 man in the previous year. Men were excluded if they were in a mutually monogamous relationship for ≥2 years with a male partner known to be HIV seronegative. Recruitment included active entertainment venue–based outreach, printed and electronic advertising, and referrals. Of 4862 men screened, 4296 were enrolled and randomized; 4295 were included in the final analysis.

Procedures

Participants completed HIV counseling and testing (through blood samples for HIV antibody detection by enzyme-linked immunosorbent assay) and behavioral surveys at baseline and every 6 months through 48 months. Reactive serum samples on a first test were retested; repeatedly reactive samples were confirmed by Western blot or immunofluorescence assays. Participants with positive results at any visit were referred to appropriate services, and their participation in the study was discontinued.

Measures

HIV Seroconversion and Sexual Risk Behaviors

We assessed time from baseline to HIV seroconversion (or last visit with a negative HIV test) and sexual behaviors (past 6 months) with partners of presumed negative, positive, and unknown HIV serostatus. High-risk sex (past 6 months) included (1) any unprotected insertive or receptive anal sex (UAS) with any partner and (2) any UAS with partners of known positive or unknown serostatus (hereafter serodiscordant UAS, SDUAS).

Syndemic Psychosocial Conditions

Depressive symptoms were assessed using a 7-item Center for Epidemiologic Studies Depression scale measuring clinically significant distress as a marker for clinical depression (scores ≥ 13). Childhood sexual abuse was defined as having any sexual experience before 13 years of age with someone >5 years older or between 13 and 17 years of age with someone >10 years older and considering this experience to be sexual abuse; this was only measured at baseline. Heavy alcohol use was defined as having ≥4 drinks everyday or ≥6 drinks on a typical day when drinking. Stimulant use was separated from other drugs because of its strong independent association with UAS and included any use of crack, cocaine, or amphetamines (past 6 months). Polydrug use was defined as using ≥3 nonprescription drugs excluding stimulants (marijuana, amyl nitrite “poppers,” heroin, phencyclidine, or other noninjectables; past 6 months). The syndemic measure was calculated as a count score based on the number of psychosocial conditions endorsed, resulting in scores ranging from 0 to 5. If participants were missing data on any of the items, the scale was not scored.

Baseline Covariates

Baseline covariates included age (categorized into 16–19, 20–25, 26–30, 31–35, 36–40, or >40 yrs), race (White, Black, Latino, or other), highest education completed (high school or less, some college, college degree, or postcollege), employment status (student, employed full time or part time, unemployed, or other), and annual income (<$6,000, $6000–$11,999, $12,000–$29,999, $30,000–$59,999, ≥$60,000). Adjusted models included these covariates. Although the randomized trial did not find a statistically significant effect on risk of HIV infection between the intervention and control conditions,²² adjusted models also controlled for study site location and randomization arm.

Statistical Analyses

To describe the syndemic, we first examined frequencies and proportions of each psychosocial problem separately at baseline and over follow-up. In line with our syndemic framework, we also examined interrelationships between psychosocial problems at baseline through multiple multivariable logistic regression models of each psychosocial problem as the outcome. We then tested whether the syndemic measure changed over follow-up using generalized estimating equations with a multinomial link function for the nominal syndemic outcome measure.

To determine whether the number of psychosocial conditions increased seroconversion hazard over follow-up, we fit a Cox proportional hazards regression model treating the syndemic measure as time varying by using the value of the syndemic measure just before the event (ie, HIV seroconversion or censoring). Because childhood sexual abuse was only measured at baseline, this measure was carried forward to each time point. We fit an unadjusted model and a model adjusting for baseline demographics, randomization arm, and study site. Using generalized estimating equations with an unstructured covariance matrix, we also assessed whether the number of psychosocial conditions increased the odds of UAS and SDUAS over follow-up. We also examined whether using only the baseline assessment of the syndemic measure affected the results; however, because the measure did not change meaningfully or statistically over follow-up, the regression results also did not differ meaningfully. As such, only the repeated measures analyses are presented.

After the Baron and Kenny test of mediation, we examined whether sexual risk acted as a mediator between number of psychosocial conditions and HIV seroconversion. To determine whether having more psychosocial conditions modified the effect of the intervention on HIV seroconversion (the primary intervention outcome), we included the syndemic-by-treatment condition interaction in a Cox model. Additionally, we included a syndemic-by-race/ethnicity interaction to determine if race/ethnicity modified the syndemic effect on HIV seroconversion.

To examine whether certain patterns of psychosocial problems played a larger role in the syndemic effect, we assessed differences in rates of seroconversion and population attributable fraction for 32 unique patterns of psychosocial problems. Because stimulant use was the most prevalent syndemic condition seen among the 32 unique patterns that we most attributable to HIV seroconversion, we also included stimulant use in our final adjusted Cox regression model to examine the independent impact of both stimulant use and syndemic measure on hazard of HIV seroconversion. For this model, the syndemic measure was recalculated to exclude stimulant use (range, 0–4).

RESULTS

Among 4295 men, mean age was 34 years (SD: 9.4), 19.0% were ≤25 years of age, 72.5% were White, 15.2% were Latino, 6.5% were African-American, 35.8% had education less than a college degree, and 40.3% had an annual household income <$30,000, as previously described.²² Psychosocial health conditions prevalent at baseline included depressive symptoms (47.3%), childhood sexual abuse (39.3%), stimulant use (25.1%), polydrug use (13.7%), and heavy alcohol use (10.5%; Table 1). There was no statistically significant change in prevalence of the count of psychosocial health conditions over follow-up, and the correlations for the count measure at each time point were high.

TABLE 1: Frequencies and Prevalence of Individual Psychosocial Conditions at Baseline and Over Follow-up

After adjusting for baseline covariates, every psychosocial condition at baseline was positively associated with all other psychosocial conditions with the exception of polydrug use, which was only significantly associated with stimulant use (see Table S1, Supplemental Digital Content, https://links.lww.com/QAI/A607). At baseline, one-quarter (24.5%) did not have any psychosocial health problems (see Table S2, Supplemental Digital Content, https://links.lww.com/QAI/A607). Over one-third (34.6%) had 1, 24.0% had 2, 10.3% had 3, 4.3% had 4, and 0.7% had 5. The distribution of these conditions did not change significantly over follow-up (P = 1.00).

Overall, 6.0% (n = 259) seroconverted over follow-up. Seroconversion rates were highest among those with more baseline psychosocial health problems: 15.2% of those with 4 or 5 conditions seroconverted, 10.0% of those with 3, 7.7% of those with 2, 5.1% of those with 1, and 3.3% of those with no psychosocial health conditions.

The hazard of HIV seroconversion was highest for those with more psychosocial health problems over follow-up, with a trend toward lower hazard with fewer conditions (Fig. 1). Specifically, in both unadjusted and adjusted models, compared with those with no psychosocial health conditions, those with 4 or 5 conditions had 8.7 times the hazard of seroconverting over follow-up [adjusted hazard ratio (aHR) = 8.69; 95% confidence interval (CI): 4.78 to 15.44], those with 3 had 5.3 times the hazard of seroconverting (aHR = 5.28; 95% CI: 3.31 to 8.44), those with 2 had 2.4 times the hazard (aHR = 2.41; 95% CI: 1.55 to 3.76), and those with 1 condition had 1.7 times the hazard (aHR = 1.68; 95% CI: 1.09 to 2.59; Table 2). A similar trend of increased risk for those with more psychosocial conditions was seen for the sexual risk behavior analysis as well (Table 2). Moreover, the sexual risk behaviors seemed to partially mediate the syndemic effect on HIV seroconversion; however, although the risk of HIV was substantially attenuated at each level of psychosocial conditions, it remained high (eg, when accounting for SDUAS, aHR for 4/5 vs. 0 conditions = 5.06; 95% CI: 2.79 to 9.18).

FIGURE 1: Hazard of HIV seroconversion by syndemic condition.

TABLE 2: Associations Between Number of Psychosocial Conditions and Time to HIV Seroconversion, Unprotected Anal Sex, and Serodiscordant Unprotected Anal Sex Over Follow-up

The null effect of the EXPLORE intervention on HIV seroconversion was not modified by the syndemic measure (interaction P = 0.722). With one exception, the control group was at higher risk of seroconverting within each strata of the syndemic measure (as were those who received the intervention within each strata of the syndemic measure compared with the control condition with no syndemic conditions); however, these differences were not statistically different (Table 3). Additionally, the association between the syndemic measure and HIV seroconversion was not modified by race, indicating that although minority MSM have higher syndemic levels the impact of the syndemic measure on HIV seroconversion does not differ by race/ethnicity.

TABLE 3: Hazard of HIV Seroconversion, by Randomization Arm and Number of Psychosocial Conditions Over Follow-up

To determine whether certain combinations of psychosocial conditions drove the syndemic effect, we examined all 32 (2⁵) possible patterns of the psychosocial conditions (Table 4). We calculated population attributable hazard fractions (for this cohort) to assess the proportion of HIV seroconversion attributable to each distinct pattern of psychosocial conditions, assuming the association is causal. Using the overall seroconversion rate of 6.3% among those with non-missing syndemic data (n = 4228), we classified 18 patterns as “high risk” (seroconversion rate > 6.3%) and 14 patterns as “low risk” (seroconversion rate ≤ 6.3%). Of the 18 high-risk patterns, 15 included stimulant use. Of the 14 low-risk patterns, only one included stimulant use.

TABLE 4: Rate of HIV Seroconversion Over Follow-up by Pattern of Psychosocial Conditions at Baseline and PAHF, Ordered by Rate of Seroconversion

To assess whether stimulant use was driving the association between the syndemic measure and hazard of HIV seroconversion, we conducted post hoc analyses including stimulant use and the syndemic measure (after removing stimulant use) simultaneously in an adjusted Cox regression model. Both stimulant use (aHR = 2.31; 95% CI: 1.73 to 3.08) and the syndemic variable (for 3/4 vs. none: aHR = 3.84; 95% CI: 2.27 to 6.51; for 2: aHR = 2.07; 95% CI: 1.40 to 3.07; for 1; aHR = 1.38; 95% CI: 0.95 to 2.01) were associated with an increased hazard of HIV seroconversion. Thus, the syndemic effect was attenuated but remained large and had a larger independent effect on seroconversion than stimulant use alone.

DISCUSSION

This study demonstrates prospectively that the accumulation of syndemic or overlapping psychosocial problems predicts HIV seroconversion among US MSM. In particular, we found that depressive symptoms, childhood sexual abuse, heavy alcohol use, stimulant use, and polydrug use interacted to produce additive effects on HIV-related sexual risk behaviors and seroconversion over 48 months of follow-up, greatly extending previous cross-sectional research.^10,11 We identified a dose–response relationship between increasing numbers of syndemic conditions and elevated HIV risk (eg, those with 4 or 5 interacting psychosocial problems experienced 8.6 times the hazard of seroconversion while those with fewer interacting conditions experienced a lower hazard). Our study provides compelling evidence for the directionality of the effect of syndemics on increased HIV risk among MSM.

The preponderance of overlapping psychosocial problems identified in our large sample of MSM in 6 US cities likely reflects the extent of mental health and substance abuse disparities in this population.^2–4 Despite 30 years of research on HIV risk among MSM, only recently have investigators begun to link these syndemic vulnerabilities, many of which can be traced back to social marginalization and disadvantage,¹² to the sexual health of MSM.^10,11 That co-occurring psychosocial health problems persist into adulthood would contribute to HIV disparities among gay, bisexual, and other MSM is broadly consistent with Meyer's minority stress model.⁴ This developmental model argues that early stressors associated with sexual minority status produce external and internal stress processes (eg, antigay prejudice, discrimination, internalized homophobia, expectations of rejection) that contribute to mental health and substance use disorders later in life.¹⁵ In support of this model, programs targeting discrimination and internalized homophobia among MSM have effectively reduced sexual risk.²³

Conceptualizing the production of syndemic conditions in this way helps maintain focus on these important targets of the broader HIV prevention agenda; yet, the social and structural causes of syndemics among MSM are typically out of the reach of traditional brief behavioral interventions.^19,20 However, our study identified specific syndemic patterns that contribute to the highest levels of population attributable risk for seroconversion, providing specific intervention targets for prevention programs. MSM characterized by the higher risk patterns (eg, stimulant users with depressive symptoms) may only benefit from sexual risk reduction counseling that is integrated into mental health and substance abuse treatment programs. Although the “seek, test, treat, and retain” strategy and other brief interventions have greater scalability and continue to play an important role in the prevention agenda, they may not be relevant for MSM experiencing multiple psychosocial comorbidities (approximately 40% of our sample). Thus, integrating sexual risk reduction counseling into other services and treatments may provide the greatest benefit for MSM experiencing the highest risk syndemic patterns. Expanding the scope of HIV prevention in this way will, however, require embracing greater integration of services and a broader range of referrals. It will also require implementing improved assessment and diagnostic tools to identify those who require additional services.

The syndemic production of HIV suggests a need for more refined comprehensive clinical assessments of psychosocial comorbidities. Diagnostic assessments would allow greater specification of treatment targets to support referrals or linkage to care for individuals at heightened risk. Key syndemic indicators among MSM, including those identified in our study, are typically assessed through self-report. However, the use of clinician-administered diagnostic assessments would help identify those who meet diagnostic criteria for particular disorders or have characteristics of significant functional impairment (eg, major depressive disorders, methamphetamine abuse/dependence). Similarly, our findings and previous research support the need to carefully differentiate between specific syndemic factors. In this sample, stimulant use is an important predictor of HIV seroconversion and is prevalent in the highest risk patterns of psychosocial conditions. Moreover, the influence of stimulant abuse on sexual compulsivity,¹⁸ sexual assault, depression, and general distress leading to sexual risk taking¹⁰ is well documented. Notably, however, above and beyond stimulant use, the syndemic effect remained, suggesting that it is not stimulant use that is alone driving these findings. This further supports our study's distinction between heavy alcohol, stimulant, and polydrug use.

Refining syndemic assessments will allow identification of individuals at heightened risk and improved treatment targets for integrated prevention programs. Improved specification of adult mental health and substance abuse treatment targets may be particularly helpful for HIV prevention because there are established efficacious behavioral and psychopharmacological treatments for many of these disorders (eg, depression, anxiety, substance abuse).²⁴ For example, focusing prevention efforts on treating the persistent symptoms of posttraumatic stress in adulthood resulting from childhood sexual abuse (rather than just identifying the abuse) may yield significant improvements in reducing sexual risk taking. In other words, understanding the syndemic production of HIV risk from a developmental perspective helps clarify that the target for integrated HIV prevention programs should be the current dysfunction in adulthood rather than the distal developmental stressor of childhood sexual abuse itself. Emerging behavioral treatment technology has already yielded promising results by integrating behavioral treatments and sexual risk reduction counseling for MSM^25,26 and other populations at heightened risk for HIV acquisition and transmission.^27,28 Such a shift, informed by syndemic theory, will help interventions to consider but move beyond antecedent developmental vulnerabilities to more appropriately address the consequences, in this case for HIV seroconversion, of syndemic production.

There are limitations to consider with this study. First, our syndemic indicators were based on self-report; thus, the extent of the impairment or distress associated with these measures is unknown. Recruitment for the intervention study occurred in large urban centers, and we obtained a sample of relatively high socioeconomic status, possibly preventing the generalizability of our findings to nonurban settings. Assessment of the higher level structural variables highlighted in syndemic theory was also limited. Finally, as the focus of our study was on initially HIV-uninfected MSM, future research is also needed to characterize the syndemic production of HIV transmission among HIV-infected MSM, which may occur through direct influence on risk behaviors²⁹ or by interfering with viral load suppression via suboptimal treatment adherence and engagement in care.³⁰ Future syndemic research among HIV-infected MSM may also help identify appropriate targets for secondary prevention interventions.

Despite these limitations, this study is an important addition to the literature in that it establishes the prospective relationship between syndemic mental health and substance abuse problems and subsequent HIV sexual risk and seroconversion in a large sample of US MSM. In addition to addressing the common social and structural causes of these co-occurring epidemics, the HIV prevention agenda will require greater integration of risk reduction counseling into existing mental health and substance abuse treatment services. Recent advances in behavioral health and prevention science research demonstrate that this approach is feasible; however, given the high prevalence of syndemic conditions among MSM and dramatically increased hazard of HIV seroconversion among those with multiple presenting syndemic conditions, now is the time to act. The efficacy of HIV prevention interventions will be enhanced by incorporating treatment components designed to address the specific presenting syndemic conditions among MSM.