Introduction Cesar Victora's “inverse equity hypothesis” predicts that socioeconomic inequality in a health outcome will increase as new interventions are introduced, but suggests the gap will close over time. We explored this in relation to the HIV epidemic in Tanzania.

Methods Literature review and analysis of two nationally-representative HIV prevalence surveys conducted among adults aged 15–49 in 2003–2004 (N=10934) and 2007–2008 (N=15542). We explored whether changes over time in HIV prevalence and sexual-behaviours differed according to socioeconomic position (measured by educational attainment). In our analysis we pooled the survey data and fitted time-period*SEP interactions within appropriately adjusted regression models for each outcome.

Results Literature review suggested that when HIV prevention activities were relatively undeveloped during the 1990s, HIV prevalence was higher among more educated individuals in Tanzania. However, between 2003/4 and 2007/8 HIV prevalence was stable among those with no education (adjusted OR 2007–2008 vs 2003–2004 1.03, 0.72–1.47), whereas prevalence declined among those with secondary education (adjusted OR 0.53, 0.34–0.84). Time-trends by education differed with respect to different aspects of sexual behaviour. There was some evidence that the association between educational attainment and recent condom use was now moving from a large poor-rich gap to a smaller one.

Conclusion The data are partially supportive of the inverse equity hypothesis. The data suggest a widening over time of the poor-rich gap in prevalent HIV infections (which may have been acquired some time ago), but there was some evidence that trends in recent condom use reflect the final “gap closing” component.