Rollout of Avahan programme activities began in January reac
Rollout of Avahan programme activities began in January, 2004, reaching almost all districts by mid-2005, and with rapid scale-up within each district. By December, 2008, more than 75% of the estimated target populations of female sex workers (total population 217 000) and high-risk men who have sex with men (total population 80 000) were being contacted monthly. In 28 districts Avahan was the first and only intervention; in the remaining 41 districts it worked alongside or took over from existing NGO interventions. From an assessment standpoint, this rapid rollout, the presence of other interventions, and ethical considerations mitigated against the use of community-based randomised controlled trials or a stepped-wedge study design.
This independent assessment (the CHARME-India project) was planned as an integral part of Avahan. In place of community-based randomised controlled trials, we used mathematical modelling with detailed HIV and STI prevalence and behavioural data to obtain plausible evidence for the effectiveness of the intervention. We first investigated the effect of Avahan in the high-risk groups targeted by the intervention, and then traced the effect on their long-term partners in the general population, reflecting the intended causal pathway of the intervention. We believe that this approach improves on a previous analysis of the population-level effect of Avahan, which did not take into account the high-risk groups on which programme activities focused, but instead used a static approach to model effectiveness through district-level differences in HIV prevalence trends in women attending antenatal care clinics. That analysis was also limited by the fact that antenatal clinic data can be subject to transient biases, leading to estimated HIV time trends that are unrepresentative of the general population prevalence.
By means of a Bayesian inference method, we aimed to use Zalcitabine testing to examine whether observed prevalence trends in high-risk groups were suggestive of evidence for condom use increasing faster during Avahan than beforehand, and to estimate, using the mathematical model, the number and proportion of HIV infections averted by Avahan because of these increases in condom use (ranked by the strength of evidence from the hypothesis testing).
Results Figure 1 provides a summary of the results of the hypothesis-testing analysis. Two of the 24 IBBA districts (Salem and Thane) could not be analysed because they had no fits with either the intervention or the control condom hypothesis, due to conflicting trends in HIV prevalence data between different risk groups. Seven of the remaining 22 districts had strong evidence (Bayes factor >5) and six had moderate evidence (Bayes factor >2 to 5) that consistent condom use by female sex workers increased during Avahan. Nine districts had weak evidence (Bayes factor ≤2). In districts with weak evidence, the overall difference in consistent condom use between the intervention and control condom hypotheses in 2010 was generally smaller (16%) than for districts with moderate or strong evidence (57%; p=0·01 [Mann-Whitney test of the medians]; figure 2). This result was partly caused by the estimated baseline consistent condom use being higher (48%) in the districts with weak evidence than in the districts with moderate to strong evidence (12%). In four of the five modelled districts in which Avahan was the first intervention targeted at female sex workers, there was moderate or strong evidence for the intervention condom hypothesis; the exception was Yevatmal, where there was weak evidence, probably because its small IBBA sample size of female sex workers resulted in less informative estimates of condom use and HIV prevalence. Figure 1 also shows the overall district-specific median proportions and numbers of infections averted in all population groups (including the general population) over the first 4 years of Avahan. New HIV infections decreased substantially in most IBBA districts, with 42% (95% CrI 33–51) of infections averted across the modelled districts. Generally, a larger proportion of infections was averted in districts with moderate to strong evidence than in those with weak evidence (median 51% vs 30%). Intervention effectiveness varied across states, largely because of differences in baseline consistent condom use, with 67% of infections averted in Tamil Nadu, 49% in Andhra Pradesh, 36% in Karnataka, and 12% in Maharashtra. Over 10 years, effectiveness increased in all districts, with 57% (46–68%) of HIV infections averted across the modelled districts.