Positive impact of increases in condom use among female sex workers and clients in a medium HIV prevalence epidemic: modelling results from Project SIDA1/2/3 in Cotonou, Benin

Abstract

Background: A comprehensive, HIV prevention programme (Projet Sida1/2/3) was implemented among female sex workers (FSWs) in Cotonou, Benin, in 1993 following which condom use among FSWs increased threefold between 1993 and 2008 while FSW HIV prevalence declined from 53.3% to 30.4%.

Objective: Estimate the potential impact of the intervention on HIV prevalence/incidence in FSWs, clients and the general population in Cotonou, Benin.

Methods and findings: A transmission dynamics model parameterised with setting-specific bio-behavioural data was used within a Bayesian framework to fit the model and simulate HIV transmission in the high and low-risk population of Cotonou and to estimate HIV incidence and infections averted by SIDA1/2/3. Our model results suggest that prior to SIDA1/2/3 commercial sex had contributed directly or indirectly to 93% (84-98%) of all cumulative infections and that the observed decline in FSWs HIV prevalence was more consistent with the self-reported post-intervention increase in condom use by FSWs than a counterfactual assuming no change in condom use after 1993 (CF-1). Compared to the counterfactual (CF-1), the increase in condom use may have prevented 62% (52-71%) of new HIV infections among FSWs between 1993 and 2008 and 33% (20-46%) in the overall population.

Conclusions: Our analysis provides plausible evidence that the post-intervention increase in condom use during commercial sex significantly reduced HIV prevalence and incidence among FSWs and general population. Sex worker interventions can be effective even in medium HIV prevalence epidemics and need to be sustained over the long-term.

Figure 1. Observed and predicted HIV prevalence trends by risk groups.

Observed and Predicted HIV prevalence among: A) all FSW nationalities combined, B) clients, C) all males (including FSW clients) and D) all females (excluding FSWs). Data points show the mean and 95% CIs of overall HIV prevalence from available data over time (solid red squares); open red squares (A) indicate data points for which CIs were not available. Solid lines (dashed lines) show the medians and 90%CrI (5th and 95th percentiles) of the modelling results. Good model fits under SIDA1/2/3 to the 38 targets used at the fitting stage (see methods) are shown as greyed areas. Also shown are results for the counterfactuals CF-1 (red), assuming FSW condom use remains at 1993 levels, and CF-2 (blue) assuming that FSW condom use mirrors that by moderate risk females in the general population. Note that overall HIV prevalence data, which were used to cross-validate model predictions because they were not used at the fitting stage (A, C & D), show good agreement with the modelled trends.

Figure 2. Condom use trends and infections prevented.

A) 95%CrI of the modelled trends in condom use during: i) SIDA1/2/3 scenario (dark grey); ii) counterfactual CF-1 (red) in which FSW condom use remains at estimated 1993 level; and iii) counterfactual CF-2 (blue) in which FSW condom use mirrors that of the moderate risk females in the general population. Prior ranges assumed for condom use by FSW nationality (Table S1) reflect potential reporting biases in self reported condom use data. B) The proportion of HIV infections prevented under modelled SIDA1/2/3 condom use trends among FSWs (pink), clients (green) and the general population (black) compared to the counterfactual scenario (CF-2) cumulatively over 5 year time period: 1993-1998, 1998–2003, 2003–2008.

Figure 3. Long term impact on HIV prevalence of sustained condom use.

Predicted HIV prevalence from the start of the epidemic to year 2040 among A) FSWs, B) clients and C) the general population assuming the pattern of condom use and migration patterns remains at current levels (median: solid line; 90%CrI: dashed lines).

Figure 4. Contribution of commercial sex to heterosexual transmission.

A) Cumulative population attributable fraction (PAF) calculated as described in Methods: i) from the start of the epidemic to the start of the intervention; and ii) from the start of the intervention in 1993 to 2008. B) Annual fraction of new HIV infections due to commercial sex work following the start of the SIDA1/2/3 intervention in 1993 (i.e. compares total number of new HIV infections in the general population every single year with and without FSW-client transmission). Box plots show the minimum and maximum, 25^th & 75^th percentiles and median.