Which phase of the natural history of HIV infection is more transmissible?

Sir: It is now accepted that the viral concentration in the circulating blood (viraemia) and other organic ̄ uids determines the probability of transmission and the clinical course of the infection. So, a large variation in the probability of transmission of HIV should be expected during the natural course of the infection. It has been argued that a substantial fraction of the total HIV transmission in a population is due in the ® rst few months after the infection, to high level viraemia seen in the initial phase of HIV infection. However, there have been several clinical± epidemiological reports of distinct intensity of HIV transmission at different phases in the natural course of the infection. In a recent publication1 we proposed a general theoretical framework for the estimation of the probability of transmission per potentially infective contact and per unit of HIV viraemia, taking into account the course of the infection inside individuals. From what is currently known about the temporal evolution of HIV infection2, it is possible to identify three distinct phases of viraemia, varying from 102 to 107 copies/mL: an initial phase of high viral replication rate and concentration, which lasts for about 4± 6 weeks, followed by a long period of low viral concentration in the blood, lasting for about 10 years, and a ® nal phase, lasting for about two years, when the immune system breaks down and the viral expression is maximized again. We also divided the population into an arbitrary number of classes with distinct sexual exposure patterns. We then calculated the relative contribution of each of the above phases of viraemia taking into account both HIV viral load levels and different scenarios of sexual exposure. As the relation between the viral load and HIV transmissibility is not yet completely understood, we considered for the purpose of calculations, two extreme assumptions regarding this dependence: the probability of transmission as directly proportional to the viral load or as proportional to the logarithm of the viral load. Of course other forms could be considered for this dependence. We argue that very recent data favour a log dependence of infectiousness on the viral load3,4. We analysed three scenarios of sexual activity behaviour considering in each case the dependence of infectivity on viral load as linear or logarithmic. If infectivity is assumed as directly proportional to the viral load, then either phase 1 (® rst weeks after infection) or phase 3 (pre full-blown AIDS) account for the majority of HIV transmission, depending on the populational pattern of sexual behaviour. When the contacts are predominantly among individuals with practically the same age and the peak of sexual activity occurs early in sexual life (for instance 20 years), then phase 1 predominates. This was the scenario analysed by Koopman et al.5 and it may hold true for some speci® c communities but it is not realistic for the majority of cases. When the contacts are spread among individuals with a broader age range and the peak of sexual activity occurs later in life, there will be enough infectives in phase 3 infecting susceptible contactants and, therefore, phase 3 predominates. This scenario is more realistic and contrasts with the former scenario in which most people in phase 3 have already ceased sexual activity. However, in all scenarios of sexual behaviour considered, if infectivity is assumed as proportional to the logarithm of the viral load, phase 2 (asymptomatic period) always predominates. Therefore, if the relationship between the viral concentration in organic ̄ uids and the chance per contact of transmitting HIV is better described by the logarithm approximation, there may be serious epidemiological implications on planning HIV therapeutic interventions in a population. In this sense, as the objective of current therapy is to reduce viral load to undetectable levels, decisions regarding implementation of mass therapy should necessarily take into account the possible epidemiological consequences as well as the individual bene® ts. In addition, with respect to the epidemiological bene® ts of antiretroviral therapy, the model by Blower et al.6 suggests that it would decrease both AIDS deaths rate and HIV incidence rate. Finally, answering the title question, it is now clear that the ® rst and the third phases are the most transmissible per contact but, as the empirical evidences are supporting a log relationship between viral load and transmissibility, the second, asymptomatic phase becomes the one which accounts for the majority of new cases.