The syndemic of violence victimisation, drug use, frequent alcohol use, and HIV transmission risk behaviour among men who have sex with men: Cross-sectional, population-based study in India

Abstract The theory of syndemics has been used to explain elevated HIV risk facing men who have sex with men (MSM). However, few studies have employed suitable analytical methods to test this theory. Using data from a probability-based sample of MSM in India, we tested three proposed models linking the co-occurring epidemics of violence victimisation, drug use, and frequent alcohol use to HIV risk: 1) the syndemic model of synergistically interacting epidemics; 2) the “chains of risk” model; and 3) the model of mutually causal epidemics. The primary outcome was inconsistent condom use with male or hijra (transgender women) partners in the past month. For the syndemic model, we included product terms between the exposures and assessed for interaction on the additive (linear probability regression) and multiplicative (logistic regression) scales. Path analysis was used to test the models of serially causal epidemics and mutually causal epidemics. Among 22,297 HIV-negative MSM, violence victimisation (24.7%), frequent alcohol use (27.5%), and drug use (10.9%) frequently co-occurred. We found evidence for a three-way interaction between violence victimisation, drug use and frequent alcohol use on both the multiplicative (semi-elasticity = 0.28; 95% CI 0.10, 0.47) and additive (b = 0.14; 95% CI 0.01, .27) scales. We also estimated statistically significant two-way interactions between violence victimisation and frequent alcohol use on the multiplicative (semi-elasticity = .10; 95% CI 0.008, 0.20) and additive (b = 0.05, 95% CI 0.002, 0.107) scales, and between drug use and frequent alcohol use on the multiplicative (semi-elasticity = 0.13, 95% CI 0.02, 0.24) and additive (b = 0.06, 95% CI 0.007, 0.129) scales. Thus, we found strong evidence for the syndemic model. The models of serially causal and mutually causal epidemics were partially supported. These findings highlight the need to sharpen how syndemic models are specified so that their empirical predictions can be adequately tested and distinguished from other theories of disease distribution.