20th International AIDS Conference - Melbourne, Australia


WEPE298 - Poster Exhibition

Multi-encounter analysis of substance use, situational characteristics, and typologies of sexual outcomes in single-partner sexual encounters reported by MSM in England

G.J. Melendez-Torres1,2, F. Hickson3, D. Reid3, P. Weatherburn3, C. Bonell2

1Centre for Evidence-Based Intervention, Department of Social Policy and Intervention, University of Oxford, Oxford, United Kingdom, 2Social Science Research Unit, Institute of Education, University of London, London, United Kingdom, 3Sigma Research, London School of Hygiene and Tropical Medicine, London, United Kingdom

Background: Men who have sex with men (MSM) in England use recreational substances at rates higher than the overall population. This substance use commonly co-occurs with sex. Though multi-encounter evidence from the United States suggests substance use is associated with sexual risk among MSM, findings are inconsistent across studies and no such evidence exists for MSM in England. Compared to single-encounter or person-level analyses, analysis of multiple sexual encounters per respondent better captures encounter-level associations between substance use, situational variables (e.g. place of sex), and sexual behaviour by addressing person-level confounding and encounter-specific characteristics.
Methods: Throughout 2011 and early 2012, MSM in England took part in a longitudinal web-based survey in which they reported up to five sexual encounters with another man. Multilevel models measured whether respondent substance use, place of sex, relationship with partner, or seroconcordance were associated with unprotected anal intercourse (UAI). A multilevel latent class model with substance use and situational variables as manifest indicators and UAI as a distal outcome was also specified.
Results: 2,142 MSM reported 6,742 eligible encounters. In univariate multilevel models, respondent substance use (OR=2.02, p< 0.001), number of substances used (OR=1.64, p< 0.001), and use of each of poppers, crystal methamphetamine, and GHB (all p< 0.05) were positively associated with UAI. UAI was less probable with anonymous (OR=0.16, p< 0.001) or serodiscordant (OR=0.31, p< 0.001) partners, and in sex-on-premises venues (OR=0.48, p< 0.001). Multivariate models including a substance use measure and situational variables yielded similar effects.

A three-class model was selected. Class 1 (31.5% of encounters, UAI probability 41.5%) reflected at-home encounters, steady partners and little substance use. Class 2 (61.8% of encounters, UAI probability 23.0%) reflected non-steady/anonymous partners and little substance use. Class 3 (6.7% of encounters, UAI probability 52.9%) was defined by polysubstance use. Overall differences in UAI probability were statistically significant (p< 0.001).
Conclusions: This study mirrors other multi-encounter studies finding significant relationships between substance use and UAI. It extends previous studies by examining substance use and situational characteristics through latent class modelling. Future studies should examine specific substances more closely and continue developing interventional approaches to reduce substance use-associated sexual risks.

Download the e-Poster

Back to the Programme-at-a-Glance