In this large and longitudinally well-characterized group of MWH and MWOH MSM, we found that use of marijuana, poppers, stimulants, and testosterone was correlated with ED drug use. Likewise, depression medication usage was weakly positively correlated with ED drug use. Furthermore, ED drug use and engagement in sexual activities, such as having more than two partners since last visit, as well as reporting insertive anal intercourse with and without condom, increased together over time.

Our findings were consistent with those of other studies that examined associations between ED drug usage, recreational drug use, and lifetime sexual partners (Chan et al., 2015; Fisher et al., 2006; Garin et al., 2017; Halkitis & Green, 2007; Harte & Meston, 2012; Paul et al., 2005). Our findings regarding the association between ED drug use and insertive anal intercourse without condom corroborated those from a study by Schwarcz et al. (2007) and Mitchell et al. (2016). Fisher et al. (2011) noted that men who were using recreational drugs were more likely to be using ED drugs; they also reported an association between ED drug use and GHB and cocaine use among men who were taking both sildenafil and methamphetamine, while another study found that users of sildenafil, along with other drugs, reported higher numbers of sex partners than those who did not use them together (Kim et al., 2002). In a descriptive analysis of users of both marijuana and ED drugs, men used ED drugs to offset the impact of marijuana on libido, since marijuana acts on the same cytochrome P450 enzyme as sildenafil (Eloi-Stiven et al., 2007; Hubbard et al., 1999). Furthermore, men were more likely to be engaged in anal intercourse without condom if they were using more than two drugs simultaneously (Halkitis & Parsons, 2002), and sildenafil users were more likely to have condomless anal sex (Swearingen & Klausner, 2005).

Within individuals, on average, use of ED drugs and other recreational drugs was positively associated. Use of testosterone, depression medication, and sexual behaviors were associated with visit-level within-person ED drug use. This suggests that increased ED drug use on average was associated with increased use of recreational drugs and sexual behaviors within individuals. To our knowledge, there are no studies to date that have examined patterns of ED drug use in association with other behavioral factors at an individual level. One study that investigated the longitudinal pattern of recreational drug use and sexual behavior among MSM found that even sporadic use of drugs was associated with a higher risk of engaging in unprotected anal sex (Colfax et al., 2005). Although the study did not look at ED drugs, it did suggest that some of the recreational drugs used alone contributed to sexual behaviors, thereby increasing the risk of HIV transmission. Most studies that have assessed the use of ED drugs, recreational drugs, and sexual behaviors have suggested that an intervention on one of the factors may not be enough to help reduce transmission. Our study supports such a hypothesis, since concurrent use of ED and recreational drugs, as well as sexual behavior, was positively associated over time.

The strengths of our study include the use of a large, well-defined longitudinal cohort. Most prior studies were of cross-sectional design, or only evaluated baseline characteristics without examining ED drug use longitudinally. In the MACS, ED drug use was assessed longitudinally, and included socially sensitive data collected through a standardized method for data collection. The statistical method used in our analysis has several advantages. The bivariate random-intercept model shows whether average person-level usage of two variables are correlated, as well as whether these variables move up and down together at the visit-level within the individual. Joint modeling of the two variables as a bivariate response that was repeatedly measured over time allowed for such analysis. In addition, variables with different numbers of observations or different times of observation can be analyzed by this model.

Our findings have several limitations. We could not evaluate ED drug frequency or dosage information. Likewise, we could not separately evaluate specific ED drug usage, i.e., we could not determine whether ED drug use was for prescription or recreational purposes. Prescription drug information was not available, so ED drug use by participants could not be linked to confirm self-reported responses. Hence, we could not perform subgroup analysis for prescription and recreational use of ED drugs. Social demographic information, as well as data regarding recreational drug use and sexual practices, were also from self-reported data. Even though the use of audio computer-assisted self-interviewing (ACASI) helps participants respond to sensitive questions more accurately, there can still be misreporting of information. Finally, sexual partner HIV serostatus was usually unknown, data on sexually transmitted infections were not collected, and pre-exposure prophylaxis (PrEP), which was approved by the FDA in 2012, was unavailable for most of our study period. Therefore, we could not determine whether insertive anal intercourse with or without condom was in the context of PrEP use.