Background BV is the most common female genital infection, particularly in developing countries and reaches its highest prevalence among female sex workers (FSWs). Over the past decade, evidence has accumulated as to its role in HIV acquisition. If BV actually plays such a role, even a modest relative risk would yield an important attributable risk for HIV, particularly in developing countries where FSWs play a pivotal role in the dynamics of HIV epidemics. Yet, data on this association among FSWs are still scanty. We analysed data from high risk women screened prior to participation in a microbicide trial to estimate BV and HIV respective prevalence and to study their association, in the presence or not of other STIs.

Methods Data from Kampala, Cotonou, Chennai and Mudhol/Jhamkandi sites were analysed. Socio-demographic, behavioural and medical data were collected through individual interviews. Blood was taken for HIV and syphilis antibody testing. Genital samples were collected for BV diagnosis using Nugent scoring system, of gonorrhoea and chlamydias by SDA and of trichomoniasis (TV) and candidiasis on wet mount. Binomial log regression was used to estimate HIV prevalence ratio (PR) in relation to BV. Fitting of multivariable models was done with backward selection using the approach of proportional change of the PR. Product terms from BV and each of other STIs were included in the final model. Significance level was set at 5% for testing associations and 15% for interactions.

Results Out of 1491 FSWs, BV data were available for 1367 among whom, BV and HIV prevalences were 47.6% (95% CI=45.0% to 50.3%) and 27.0% (95% CI=24.6% to 29.3%) respectively. In multivariable analysis (Abstract O1-S05.06 table 1), adjusting for site, age, years of education, occupation, current contraceptive method, oral sex, past history of STI, gonorrhoea, candidiasis and syphilis, BV was significantly associated to HIV (adjusted PR=1.20, 95% CI=1.01% to 1.42%, p=0.03). In addition, the PR was negatively modified by TV, whose prevalence was 6.7%: PR was 1.25 (1.05 to 1.48) and 0.76 (0.41 to 1.38) in the absence and the presence of TV respectively (p for interaction =0.12).

Conclusions Though its cross-sectional design precludes all directional interpretation of the findings, this study confirms the relationship between BV and HIV among FSWs and warrants prospective studies in this population. The negative modifying effect of TV on this association's measure needs further investigation.