Backgound Current epidemiological evidence suggests that receptive anal intercourse (RAI) considerably increases the risk of HIV infection per sex act (RRRAI) compared to vaginal intercourse (VI). RAI may increase HIV risk to a similar extent or more than primary HIV infection (PI)(ie, recent infection) increases infectivity compared to asymptomatic infection (RRPI). Considerable attention has been placed to understand the role of PI to HIV epidemics. However, the potential role of RAI to heterosexual HIV epidemics has never been assessed even if it seems to be a relatively common practice in many settings. We aim to compare the fraction of HIV infections due to AI and PI in a generalised heterosexual epidemic.

Methods A deterministic model of heterosexual HIV transmission during to VI, RAI and insertive AI, incorporating three HIV infectiousness stages was developed (Abstract O1-S07.01 Table 1). Behaviour and HIV prevalence data from Kalishman's et al (2009) study in South Africa was used to define plausible ranges of parameter values (Abstract O1-S07.01 table 1). As it is unknown, the degree of mixing during VI between those who engage and do not engage in AI (non-AI) was varied. 20 parameter sets that best fitted HIV prevalence data by AI/non-AI were identified following exploration of 1000 parameter sets selected by uniformly sampling the plausible parameter ranges.

Results Overall, 17%–40% of annual infections (PAF) may be due to RAI and insertive AI (IAI). The PAF due to AI is 2–2.6 larger for female than male (Abstract O1-S07.01 Table 1). In comparison, the overall PAF due to PI is between 25 and 31%, and more similar between gender (PAF female: male 1.0–1.2). Under our assumptions, the PAF due to AI was always larger (smaller) than the PAF due PI for females (males) (Abstract O1-S07.01 table 1). The PAF due to AI and PI was positively associated with increases in the overall fraction of all sex acts which are AI (%AI), whereas the latter depended on the level of mixing. In order to be able to relax the mixing to make it less assortative, the %AI needed to be reduced to allow more VI between AI and non-AI to occur.

Conclusion Our preliminary results suggest that even a small fraction of AI (<10%) in a population may be as important, to overall HIV transmission in generalised epidemics, as the primary phase of infection, especially for women. Our results are based on the likely assumption that RRRAI equal or larger than RRPI. Focusing prevention to reduce AI may be more cost-effective than to test and treat for recent infections.