Calculating the consequences: HAART and risky sex

Since the widespread availability of highly active antiretroviral therapy (HAART) in 1996, many have wondered what effect the widespread usage of these therapies will have on the incidence of HIV. In an extremely interesting paper in this issue of AIDS, Law et al. ^[1] predict what effect HAART will have on reducing new HIV infections in the gay community in Australia. To make their predictions Law and colleagues ^[1] used a mathematical model that specifies the transmission dynamics of HIV and the potential treatment effects of HAART. Transmission models can be useful health policy tools, because they can be used to provide a quantitative evaluation of the likely range of epidemiological outcomes. Such models can also be coupled with sophisticated uncertainty and sensitivity analysis techniques in order to be used as predictive tools [2]. The first transmission model was constructed by Daniel Bernoulli over 200 years ago, and was used to evaluate the consequences of different interventions for the control of smallpox. Transmission models are dynamic process models, because they are constructed to contain the causal processes that generate epidemic-level data such as incidence rates, death rates and prevalence. Dynamic process models are commonly used, and widely accepted, in the pure sciences, such as physics and chemistry. Such models are also used widely in many applied fields such as economics, demography, fisheries management, geology, hydrology and climatology. However, transmission models have not yet been widely applied as either health policy or predictive tools in the study of infectious diseases.

Law et al. ^[1] modelled the effect of HAART on decreasing infectiousness and the potential competing effect of increases in unsafe sex, and then predicted the net outcome of these two effects on the HIV incidence rate. They developed a simplified mathematical model of HIV transmission for gay men in Australia, and used an uncertainty analysis to predict the incidence rate. They assumed that treatment (by reducing viral load) could reduce infectiousness, anywhere from no decrease to a 100-fold decrease, and that risky sex could increase anywhere from no increase to doubling. On the basis of these assumptions they then used their model to evaluate these competing effects and to predict the effect on the incidence rate for a one year period. Their analysis showed that the greater the reduction in infectiousness that HAART induced, then the greater the reduction in the incidence rate would be. However, their results also showed that the beneficial effect on therapy of reducing the incidence rate could be counterbalanced by the effect of increases in risky behavior. Increases in risky behavior could result in the incidence rate actually increasing. These results of Law et al. ^[1] for the gay community in Australia confirmed the previous findings of Blower et al. ^[2] for the gay community in San Francisco. Blower et al. ^[2] demonstrated that whereas a high usage of HAART would decrease the incidence rate, such a beneficial effect would fairly easily be counterbalanced by an increase in risky behavior. Their predictions that the incidence rate in the San Francisco gay community would increase ^[2] have recently been confirmed by empirical data ^[3].

The current results of Law et al. ^[1] and the previous results of Blower et al. ^[2] have significant public health implications. These results imply that there will be a significant public health benefit (i.e. incidence rates of HIV will fall) as more HIV-positive individuals gain access to treatment, but that this public health benefit will only occur if the levels of risky behavior do not increase ^[1,2]. Incidence rates will increase, even with a high usage of HAART, if levels of risky behavior increase ^[1,2]. The majority of HIV prevention intervention efforts for gay men since the beginning of the HIV epidemic have been focused on trying to prevent HIV-negative individuals from becoming infected. Recently, the focus of prevention efforts has shifted towards trying to prevent HIV-positive individuals from transmitting the infection. With the widespread usage of HAART, we should use this opportunity to try and ensure that HIV-positive individuals who come in for treatment reduce (rather than increase) their risky behaviors. Significant and substantial efforts should be made to try to link the medical and the behavioral interventions together in a way that does not currently occur. Programs should be developed to simultaneously treat HIV-positive individuals and to counsel these individuals to reduce the number of sex partners that they have, to use condoms and to select sex partners of matching sero-status (i.e., to sero-sort).

Transmission models can be viewed as translational research tools, because they provide a means of linking the results from laboratory studies with the results from behavioral science and clinical trial data in order to predict the epidemic-level consequences of medical and behavioral interventions. Models can thus be used to predict the public health implications of different treatment strategies ^[1,2,4]. Recently, it has been suggested that HAART should be used as a public health prevention strategy for reducing incidence rates. However, when making public health decisions regarding HAART it is necessary to consider carefully the beneficial and detrimental effects to the individual patients ^[2,4]. It may often be in the best clinical interests of the individual patient to delay treatment rather than to treat early, although doctors should not withhold HAART from certain patients simply because they assume that the patients will not be able to adhere to the complex treatment regimen ^[4]. Therefore the usage rate of HAART at any time should be determined by the number of individuals in the population who would receive a medical benefit from these therapies. Any public health decision regarding HAART should also consider multiple epidemiological goals, not only the reduction of incidence rates. Although the results from the models show that a high usage of HAART could reduce HIV incidence rates ^[1,2,4] and AIDS death rates ^[2,4], the modelling studies also show that a high usage rate of HAART over time would result in a high prevalence of drug-resistant cases ^[2,4,5]. This build-up of drug-resistant cases could well be detrimental to public health, as well as to the individual patients who develop resistance. The current modelling results ^[1,2,4] predict what would happen given certain usage rates of HAART. These results should not be used as a rationale for increasing population-level treatment rates of HAART, unless such a decision would also result in clinical benefits to the individual patients. However, it is very clear from the current modelling results ^[1,2] that every effort should be made to try and stop the levels of risky sex increasing.