Article Text


Epidemiology poster session 4: Modelling
P1-S4.18 Using mathematical modelling to investigate the role of the hidden “population of men who have sex with men (MSM) on the HIV epidemic in Southern India”
  1. H Prudden1,
  2. A Foss1,
  3. K Mitchell1,
  4. M Pickles2,
  5. A Phillips2,
  6. B M Ramesh3,4,
  7. R Washington3,4,5,
  8. M Alary6,
  9. C Lowndes7,
  10. P Vickerman1
  1. 1London School of Hygiene and Tropical Medicine, London, UK
  2. 2Imperial College London, London, UK
  3. 3Karnataka Health Promotion Trust, Bangalore, India and
  4. 4University of Manitoba Winnipeg, Canada
  5. 5St John's Research Institute, Bangalore, India
  6. 6Centre hospitalier affilié universitaire de Québec, Québec, Canada
  7. 7Health Protection Agency, London, UK Canada


Background Biological and behavioural data for men who have sex with men (MSM) in Bangalore, Karnataka, India, have mainly been collected from sites where commercial sex is prevalent. Consequently, the survey data may fail to capture the behaviour of a larger lower-risk “hidden” MSM population. Mathematical modelling is used to explore the potential bias in the survey data and better quantify the characteristics of this hidden population.

Methods A dynamic model of HIV transmission among MSM was developed and parameterised using detailed data* from high risk MSM in urban Bangalore. The MSM were categorised into three subgroups: Kothi and Hijra(KH): who mostly take the receptive role in anal sex, Panthis and Bisexuals(PB) who are predominantly insertive partners and Double Deckers(DD) who take both roles. Due to the sampling methods used, it was thought the MSM survey data were more representative of KH and DD than PB, although the extent of this bias is unknown. Therefore, no fitting constraint was applied to the PB HIV prevalence and instead the model was used to explore what PB HIV prevalence values are projected if the model was only fit to the 95% CIs of the prevalence data for KH and DD. One million randomly sampled model simulations were undertaken to find model fits.

Results Abstract P1-S4.18 figure 1 shows that, although the model can produce HIV prevalence estimates consistent with the survey estimates, overall the model projections suggest a lower PB HIV prevalence is more consistent with the survey estimates for KH and DD. In addition, 80% of the model fits to the KH and DD HIV prevalence data had a sampled frequency of sex acts for PB in the lower half of the uncertainty interval suggesting PB's sexual activity may be lower than the median reported in the MSM survey. As expected, an inverse relationship occurs between PB population size and their level of sexual activity, with the median PB population size being 55 400 (2.7% of the total urban male population) and varying between 28 000 and 73 000 (1.5 to 3.9% of the total urban male population) for the IQR of the model fits Conclusions: Survey data imply MSM are a small, highly active group, many of whom regularly sell sex, have very high numbers of partners and typically take the receptive role. As demonstrated here, modelling can be used to provide insights into the likely HIV prevalence, population size and sexual activity of hidden “MSM not reached in surveys.”

Abstract P1-S4.18 Figure 1

Number and range of projected HIV prevalence estimates for the PB population (from the model fits to KH and DD data. *Integratedbiological and behavioural assessment (IBBA) 2009, collected within the monitoring and evaluation of Avahan, the India AIDS initiative.

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.