Article Text


Epidemiology oral session 9: Use of observational data and mathematical modelling for impact evaluation
O1-S09.06 Assessing the impact of a FSW targeted HIV intervention programme on incidence and prevalence in Cotonou, Benin
  1. J Williams1,
  2. M C Boily1,
  3. C Lowndes2,
  4. S Camden3,
  5. E Demers3,
  6. I Minani4,
  7. M Zannou4,
  8. S Anagonou5,
  9. A C Labbé6,
  10. M Alary7
  1. 1Imperial College London, London, UK
  2. 2Health Protection Agency, London, UK
  3. 3Centre Hospitalier Affilié Universitaire de Québec CHA, Quebec City, Canada
  4. 4Dispensaire IST, Cotonou, Benin
  5. 5Centre National Hospitalier Universitaire, Cotonou, Benin
  6. 6Hopital Maisonnneuve-Rosemont, Montreal, Canada
  7. 7Centre Hospitalier Affilié Universitaire de Québec CHA and Université Laval, Quebec City, Canada


Background From 1993 to 2005 female sex workers (FSW) were the focus of a HIV intervention in Cotonou where most local FSW originated from three neighbouring countries. Intervention impact was assessed using a HIV transmission model parameterised and fitted to empirical data within a Bayesian framework, to compare predicted epidemic outcomes with and without the intervention.

Methods An age-structured model was developed of HIV transmission in Cotonou, coupled with a non-age structured Gc model. The model included realistic demography, heterogeneous sexual risk groups including local and migrant FSW to and from Cotonou, clients and general population (GP); each FSW nationality was parameterised separately. Programs of condom use and STI treatment modelled reflected levels of uptake pre- and post-intervention. Plausible parameter ranges based on local multiple survey data from different risk groups and time points and literature reviews (ie, prior parameter distribution) were sampled repeatedly by Latin hypercube sampling. Model predictions from parameter sets matching (ie, fitting") observed HIV prevalence data from Cotonou (FSW by nationality, their clients, and broad GP age-bands) at different time points were accepted as good fits (ie, posterior parameter distribution (PPD)) and used to estimate impact (95% credibility interval (CrI)) by comparing HIV incidence predicted based on the PPD with that using the same parameter values except for the assumption that condom use and STI treatment remained at 1993 levels (control group/counterfactual).

Results From 100 000 parameter combinations tested, 18 produced results that agreed with observed HIV prevalences in FSW (Abstract O1-S09.06 figure 1) and other groups over time. Results with no intervention suggest a peak HIV prevalence in FSW 10–12 years later than observed and at around 75%, compared with the observed level of just over 53% (corresponding GP prevalences were 9.6% & 6.7% respectively). Results also suggest that about 33%(CrI: 28, 37) and 25%(CrI: 17, 33) potential HIV cases may have been averted in Cotonou FSW and GP respectively, by increased condom use and STI services after the start of the intervention.

Conclusions Coupled with observed HIV trends, the analysis indicates that FSW targeted intervention was effective in curtailing the HIV epidemic in Cotonou-based FSW despite high turnover of foreign FSW. There was also substantial impact in the general population by impeding onward transmission to clients/boy-friends.

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.