Modelling HIV incidence in gay men: increased treatment, unsafe sex and sexually transmissible infections

Several authors have now used mathematical models to assess the effect that combination antiretroviral treatments might have on the incidence of HIV by decreasing the HIV viral load and therefore, presumably, reducing the risk of HIV transmission [^1-3]. These different models have produced qualitatively similar results, indicating that apparently very large reductions in the risk of HIV transmission in individuals receiving antiretroviral treatments can be counterbalanced in terms of new HIV infections by more modest, albeit quite substantial, increases in risk behaviours. Our model suggested that, among homosexual men in Australia, two-, five- and 10-fold decreases in infectiousness as a result of antiretroviral treatment could be counterbalanced in terms of new HIV infections by increases in unprotected anal intercourse (UAI) between HIV-negative and positive men of approximately 40, 60 and 70%, respectively.

One aspect that none of these models has investigated is that any increases in unprotected intercourse could lead to increased rates of sexually transmissible infections (STI), which independently increase the risk of HIV sexual transmission [⁴]. In Australia, there is continuing evidence of increases in UAI among HIV-infected and uninfected homosexual men [⁵]. There has also been evidence of increases in the rates of gonorrhoea among homosexual men in Sydney [⁶], and a doubling in the reports of rectal gonorrhoea among men in New South Wales [⁷].

To investigate how increased STI rates (as a result of increased UAI) among homosexual men in Australia might independently effect HIV transmissions, we extended the mathematical model described previously [¹]. Data from the Sydney Men and Sexual Health Study, a large cohort study of homosexual men in Sydney, indicated that before the availability of effective antiretroviral treatment in Australia, self-reported gonorrhoea 6 month period prevalence rates in homosexual men were approximately 2% (unpublished data on file). Since the widespread availability of combination antiretroviral treatments in mid-1996, the prevalence rates of rectal gonorrhoea in men in New South Wales have at least doubled [^6,7]. The increase in prevalence rates of all STI, including gonorrhoea, were therefore simulated from a uniform distribution corresponding to an increase of 2% of homosexual men (range 1-3%). The increase in STI rates was assumed to occur in all homosexual men, regardless of HIV infection status or disease stage. This magnitude of increase in the percentage of homosexual men with STI was assumed to be 100%, correlated with the magnitude of the simulated increases in UAI. The effect of being infected with an STI was assumed to correspond to a 3.5-fold (range two- to fivefold) increase in the risk of HIV transmission [⁴]. A combination of these parameters resulted in the effect of STI on increasing HIV transmission being 73%, correlated with increases in UAI. In all other respects the model was the same as previously described [¹]. As before, the effects of antiretroviral treatment and increases in UAI were assumed to occur instantaneously. One thousand simulations of the model were run over a time-horizon of a single year, and as before the model-predicted rate of HIV incidence was compared with a null model in which there was no effect of antiretroviral treatment on reducing HIV infectiousness, and no increase in UAI.

With this model, the dynamic competing effects of reduced HIV infectiousness as a result of antiretroviral treatment and the increased rates of UAI and STI on HIV incidence are illustrated in Fig. 1. In this figure, black dots and black squares correspond to simulations in which the incidence of HIV was found to increase by 0-20%, and over 20%, respectively, compared with the null model. This figure suggests that two-, five- and 10-fold decreases in infectiousness as a result of antiretroviral treatment could be counterbalanced in terms of new HIV infections by increases in UAI of approximately 30, 50 and 65%, respectively.

Fig. 1

These models suggest that even very modest increases in STI, as a result of increased UAI, could have an important multiplicative effect increasing the incidence of HIV. Our results serve to underscore that the dynamic between interventions that may decrease HIV transmissions, such as combination antiretroviral treatment or vaccines, and changes in risk behaviours, which may increase HIV transmission, is complex and may not be easily predictable.

Matthew G. Lawa

Garrett Prestagea

Andrew Grulicha

Paul Van de Venb

Susan Kippaxb