HIV prevalence may decline because the most vulnerable are infected and die first

The declines in HIV prevalence and incidence seen in recent years in some countries may be largely due to differences in people’s susceptibility to the virus rather than behaviour change, according to a mathematical model based on a survey of Kenyan sex workers, published in the January 14^th edition of AIDS.

The model suggests that HIV infections in the early stage of an epidemic occur more frequently in a subgroup of the population who are genetically most vulnerable to HIV. This means that as time goes by and most of these people become infected, HIV can only be passed on to those less genetically vulnerable, resulting in drastic declines in incidence and – as the first generation dies – declines in prevalence.

HIV prevalence in a number of African countries has declined substantially in the last decade, for instance from 31 to 16% in Zimbabwe and 14 to 5% in Kenya. Estimates of HIV prevalence in India have also been more than halved – see this report. When this phenomenon was originally seen in Uganda in the 1990s, it was attributed to the success of early prevention programmes conducted in the late 1980s.

However in a long-standing longitudinal study of initially HIV-negative Nairobi sex workers (Kimani), HIV incidence declined despite no apparent changes in sexual risk behaviour. In this study the risk of acquiring HIV per sex act declined fourfold between 1985 and 2005, from one infection per 225 sex acts in 1985 to one per 1000 in 2005. This cannot be explained by a difference in HIV prevalence among the sex workers’ clients, as it pre-dated HIV prevalence declines in the Kenyan male population by a decade.

The authors hypothesised that HIV incidence might be declining because the virus was disproportionately infecting the most genetically vulnerable women first. They devised a model of a typical African population. They divided the population into high-risk persons (namely female sex workers and their clients) and the general population, and modelled movement between these groups. They then modelled the way the epidemic would develop if they divided this population into three groups of 30% each, who could be respectively infected with HIV easily, quite easily, and with difficulty, and 10% who were completely resistant to infection.

Ten per cent deliberately overestimates the proportion likely to be totally resistant to HIV because the authors wanted to build the most conservative possible estimates of susceptibility to infection into their model in order to find out if it still explained observed declines in incidence. For the same reason they assumed there was no difference in infectiousness between people in early infection and chronic infection, and made no allowances for other factors that influence infectiousness and susceptibility, such as sexually transmitted infections.

They found that a version of this model almost completely explained the observed decline in infections seen in the Nairobi female sex workers and fitted fairly closely the observed increase and later decline in prevalence in the general Kenyan population.

It is difficult to directly measure people’s susceptibility to HIV, as they can be infected only once. This may explain why differing susceptibilities of people to infection has received less attention as a possible driver of the epidemic than differential infectiousness in those already with the virus. However it may provide a better explanation of why, for instance, the risk of infection after ten heterosexual contacts with a person with HIV has been found to be as high as 10%, but only increases to 23% after 2000 contacts.

This model implies that predictions of the spread of HIV in a mature epidemic may be drastically overestimated if they are based on the infection rate seen in the first few years. It may also explain the lower-than-expected infection rates seen in some HIV prevention trials. It also implies that the effectiveness of some HIV prevention programmes may have been overestimated.

The authors comment: “We propose that the phenomenon of heterogeneity in HIV susceptibility may have contributed to the observed declines in HIV incidence and prevalence.”

They add that although there is compelling evidence that at least part of the HIV prevalence declines in parts of Africa were caused by changes in risk behaviour, “some of this decline may have occurred without behavioural change, confounding our ability to attribute HIV epidemic shifts to specific interventions.”