THPDC0101 - Poster Discussion
Sexual mixing patterns between men who have sex with men in southern India: implications for modelling the HIV epidemic and predicting the impact of targeted oral pre-exposure prophylaxis
Presented by Kate M. Mitchell (United Kingdom).
K.M. Mitchell1, A.M. Foss1, H.J. Prudden1, M. Pickles1,2, J.R. Williams2, H.C. Johnson1, B.M. Ramesh3,4, R. Washington3,5, S. Isac3, S. Rajaram6, A.E. Phillips2, J. Bradley6,7, M. Alary6,7, S. Moses4, C.M. Lowndes7,8, C.H. Watts1, M.-C. Boily2, P. Vickerman1,9
1London School of Hygiene and Tropical Medicine, London, United Kingdom, 2Imperial College London, London, United Kingdom, 3Karnataka Health Promotion Trust, Bangalore, India, 4University of Manitoba, Winnipeg, Canada, 5St. John's Research Institute, Bangalore, India, 6CHARME-India Project, Bangalore, India, 7Centre Hospitalier Affilié Universitaire de Québec, Québec, Canada, 8Health Protection Agency, London, United Kingdom, 9University of Bristol, Bristol, United Kingdom
Background: In southern India, the identity of men who have sex with men (MSM) is closely related to role taken in anal sex, but little is known about sexual mixing between identity groups. Both role segregation and assortative (within-group) mixing are known to affect HIV epidemic size in other settings. This study aimed to explore how different mixing patterns affect estimated HIV trends and intervention impact for MSM in Bangalore.
Methods: Deterministic models describing HIV transmission between three MSM identity groups (mostly insertive panthis/bisexuals (PB), mostly receptive kothis/hijras (KH) and versatile double deckers (DD)), were parameterised with data collected in Bangalore for the evaluation of the Avahan intervention. These models were used to explore four different mixing patterns (table). 300,000 randomly sampled parameter sets were obtained from data ranges and used to find multiple fits to group-specific HIV prevalence data in 2006 and 2009. Model fits were used to compare predicted HIV time trends. To compare the impact of a new intervention scenario, condom use was assumed to decline from high levels in 2010 due to condom-intervention fatigue. Oral pre-exposure prophylaxis (PrEP) was introduced in 2015, assuming 42% effectiveness (efficacy x adherence) and 60% coverage, targeted at KH and DD (the groups easiest to reach).
[Table: mixing patterns]
|Mixing pattern||How mixing was determined||Number of fits||Median Q* for fits|
|Maximum assortative||KH, DD, PB all have as many acts with members of the same group as constraints allow.||108||0.16|
|Setting plausible||DD have as many acts with other DD as constraints allow. KH have as many acts with PB as constraints allow and vice versa.||159||0.00|
|Proportionate||Receptive acts distributed between groups in proportion to the number of insertive acts offered by each group.||301||-0.16|
|Disassortative||PB and DD have as few acts with members of the same group as constraints allow.||430||-0.40|
|*measures assortativeness of mixing in the whole population. 1 = completely assortative; -0.5 = completely disassortative|
Results: Large differences in levels of assortative mixing were seen for fits identified using different mixing patterns (table), but little difference was projected in HIV prevalence trends (figure A). Different mixing patterns gave somewhat different estimates for group-specific impact of the PrEP intervention (figure B), but overall impact in the whole MSM population was very similar (< 10% difference in % infections averted).
Conclusions: A variety of different mixing patterns are consistent with the data. However, model predictions of future HIV epidemic trends and overall impact of a targeted intervention are robust to the different mixing patterns and intervention scenario explored here.
Download the e-Poster
Back to the Programme-at-a-Glance