Giving treatment to specific high-risk people may bring down HIV faster than distributing it equally

In a concentrated epidemic, HIV is mainly restricted to – or at least more common in – specific risk groups. In this situation, and if universal antiretroviral treatment (ART) cannot be provided, it may have more effect in public health terms to give ART to people at high risk of acquiring and transmitting HIV.

If the correct risk group is targeted, suppressing viral load in that group may reduce HIV prevalence in other groups they are connected to; on the other hand, targeting ART to a group that is larger but more peripherally connected to the epidemic may not help to reduce prevalence and incidence within other populations.

These are the implications of a model developed by Brian Williams of the South African Centre for Epidemiological Modelling and Analysis (SACEMA) and presented to the 2014 Treatment as Prevention Workshop in Vancouver last week.

The reproduction number

The model was developed using data from Can Tho, a province in southern Vietnam, but could in theory be applied everywhere with sufficient surveillance data. It provides a new way of looking at HIV infection/transmission risk: not centring on the number of infections or the proportion infected in specific groups, but on what is called the reproduction number, or R₀, of members of each group. 

The R₀ number, which Williams described as “the most important number in epidemiology”, is the number of other people each individual in an epidemic will infect before they recover or die. If the R₀ number is over 1, the epidemic grows; if it is below 1, it shrinks. With good enough data this can be calculated not just for the epidemic as a whole but for specific groups.

Williams warned that models are not predictions, but tests of the implications of what current data say. He also said that the model he presented was not one that could be calculated in many areas, as it depended on very good surveillance of HIV infections and of knowing the risk factors not only of those diagnosed but of the people who infected them. Vietnam, which had a network of local doctors with good knowledge of their patients and their local community, was just such a country.

The Can Tho, Vietnam model

Can Tho has a concentrated epidemic, but one that is very mixed. Among its approximately 5000 people diagnosed with HIV are 109 women classed as being female sex workers (FSW), 131 men whose risk was being a client of a female sex worker (MCF), 615 men who have sex with men (MSM), 1446 people who inject drugs (PWID) and 1776 low-risk women (LRW).

Because there are far more low-risk women in the province (464,000) than ones known to be sex workers (just over 2000) or people who inject drugs (about 3150) and even MSM (about 62,000), HIV prevalence in the groups is very different from the absolute number of infections. It is currently estimated as 0.4% in low-risk women, 1% in MSM, 5.4% in sex workers, 8.8% in their clients, and no less than 46% in people who inject drugs.

Using surveillance data, Williams and colleagues were able to map out a complex network of exactly how common HIV transmission was between different groups of people – who was infecting whom – and calculate the R_{0­ ­­­­}number of each group. He divided MSM and FSWs into ones who were, and were not, injecting drugs.   

The overall R₀ within the HIV-positive population was 22. This is very high: by comparison, the R₀ in sub-Saharan Africa is about 5. This implies that if ART was given uniformly, 96% or more of the HIV-positive population would have to be on ART and virally suppressed for treatment as prevention alone to end the epidemic. 

There are, however, vast disparities in the R₀ of different groups. Williams estimated that in terms of infections occurring within the epidemic at any one time, 41% of infections involved a person who injects drugs either as a transmitter or recipient and 30% a female sex worker. People who belonged to more than one group bore an even greater burden of infections/transmissions: 43% of infections involved an MSM who injected drugs, either as receiver or transmitter.

The most striking disparities became clear when the direction of infection was ascertained: whether particular groups were more likely to be receivers or transmitters, and who they primarily passed HIV on to or acquired HIV from. This varied by a factor of over 100,000 according to which group a person belonged to.

Because infection in low-risk women was rare, the number of infections each individual transmitted to them was itself low. The R₀ number of male clients of sex workers towards low-risk women, for instance, was only 0.01: this means that only one in 100 male clients of sex workers would transmit HIV to a low-risk woman in their lifetime. The same was true of people who inject drugs who might also be partners of low-risk women, and for bisexual MSM, only one in 1000 would transmit HIV to a low-risk woman. Infection by low-risk women was so low as to be impossible to estimate. This means that a generalised epidemic involving the heterosexual population is unlikely to happen in Vietnam.

The R₀s of transmission to and from other groups were much higher, and all over 1, implying a concentrated epidemic that will be self-sustaining without more intervention.

The model found that the average number of HIV infections transmitted by one MSM in Can Tho in his lifetime was 6.1: 4.1 to non-drug-injecting MSM and 2 to MSM/people who inject drugs. MSM who inject drugs transmitted more: their total R₀ was 27.4, of which 19 infections would be transmitted to other people who inject drugs, 4 to other MSM who did not inject drugs, 3.3 to other MSM who also injected drugs, and 1 to a female sex worker who injected drugs.

Again, however, this is the average number transmitted by individuals in specific group: because there are far fewer MSM who inject drugs than MSM, the absolute number of infections transmitted by MSM will be bigger than the number transmitted by MSM who inject drugs.

People who inject drugs had an R₀ per individual of 21.5: 19 to other people who inject drugs who were not MSM or FSWs, 2 to other people who inject drugs who were MSM and 0.5 to other people who inject drugs who were FSWs.

Female sex workers had a larger R₀ still. Because there were not many sex workers, they were only the source for 12% of transmissions of HIV: 10% to male clients and 2% to people who inject drugs. However, the infections were almost entirely one way – from sex workers to their clients. A male client only had a 1-in-17 chance of passing HIV on to a female sex worker during his lifetime. On the other hand, the R₀ of a female sex worker was a startling 77: because female sex workers had so many clients, each one was passing HIV on to an average of 77 other people during her lifetime, and female sex workers who were also injecting drugs even more, a total of 99: 77 infections to male clients, 19 to people who inject drugs, two to people who inject drugs who are also MSM and one to another female sex worker who injects drugs.

So, according to the model, the average R₀, the number of individuals each person with HIV will infect during their lifetime, is as follows:

• Female sex workers who also inject drugs: 99
• Female sex workers: 77
• MSM who also inject drugs: 27.4
• Other people who inject drugs: 21.5
• MSM who do not inject drugs: 6.1
• Clients of female sex workers: 0.061
• Low-risk women: negligible

Targeting antiretroviral therapy

From these findings, it might look as if female sex workers are the obvious people to prioritise for HIV treatment at any CD4 count. However, this turned out not to be the case in the model. The reason is that people who inject drugs are much more central to this epidemic than female sex workers: they are in contact with, and pass HIV on to, people from most of the other risk groups, at first or second hand, whereas female sex workers, in the main, only pass HIV on to their clients who in turn only occasionally pass HIV on to other women.

If ART was not available, Williams calculated, the epidemic would stabilise with HIV prevalence in most groups remaining at their present rate except in MSM, where prevalence is increasing and would eventually stabilise at 8%.

In a “counterfactual” scenario, i.e. what is not likely to happen, providing universal ART and needle exchange to all HIV-positive people who inject drugs (and no one else) would have a dramatic effect on the epidemic: if instituted today it would virtually abolish HIV in people who inject drugs by 2025 and would bring prevalence down to 1% in MSM; by 2035 it would bring prevalence in female sex workers down to 1% and by 2050 it would bring down prevalence in sex worker clients to 0.6% and low-risk women to 0.1%.   

If instead ART (and condoms) were exclusively targeted at female sex workers, it would bring down infections dramatically in them and in their clients and modestly in low-risk women: but there would be hardly any change in prevalence in MSM and people who inject drugs. 

If MSM, sex worker clients, or low-risk women were exclusively targeted, the result would be falls in HIV prevalence in those specific groups but only relatively modest or no falls in the other groups.

It would seem to make sense, then, within the context of this particular province in a country with a complex concentrated epidemic, to prioritise people who inject drugs as a group and female sex workers as individuals when it comes to offering ART and other prevention services. This does not mean not offering ART to others, but it does suggest people who inject drugs and female sex workers should have a high priority if resources are stretched – the opposite of what happens in some countries, where people who inject drugs, in particular, are more likely to be excluded from prevention services and ART than some other groups.

Each country or locality will be different and different individuals and groups may need to be prioritised. But as a recent phylogenetic study presented at the Conference on Retroviruses and Opportunistic Infections in March also suggested, Williams’ model, using a quite different methodology, finds that targeting specific highly connected individuals for prevention resources may have a disproportionately powerful effect in bringing down HIV prevalence.