Average viral load in the community predicts HIV incidence in drug users

A study from Vancouver has found a strong correlation between average viral load in the HIV-positive members of a group of injecting drug users and the rate of subsequent HIV infections in the HIV-negative members of the same community, irrespective of risk behaviours. The study by the British Columbia Centre for Excellence in HIV/AIDS, published in the British Medical Journal, found that the average ‘community viral load’ was a more reliable predictor of the rate of HIV infection over the subsequent six months than any behavioural or social factor such as ethnicity, the rate of needle sharing, unsafe sex, homelessness, or the numbers actively injecting heroin or cocaine. However the association only held in pre-HAART days when community viral load was high.

The study strengthens evidence that bringing down the average viral load among HIV-positive people by putting as many people as possible on antiretroviral drugs (ARVs) would be a very powerful HIV prevention tool. Last year, a mathematical model by the same team (Lima et al.) predicted that if it were possible to put all HIV-positive people on ARVs within a year of their infection the rate of new infections would fall by 70%.

It might seem obvious that reducing average viral load might reduce infections, but it is less easy to prove, because behavioural, socioeconomic and demographic changes in the population studied may also profoundly affect HIV transmission rates. It is also problematic to measure variables like viral load and HIV acquisition that change over time in a mobile population of people.

Because there is only one HIV testing laboratory and ARV dispensing clinic in the area, the Canadian team were able to overcome these barriers and observe the viral load and HIV acquisition rates in a group of over 2000 injecting drug users, 30% of them already HIV-positive, in the Downtown Eastside neighbourhood of Vancouver, an area that has long had an injecting drug use and HIV problem.

The researchers started the study in January 1996 and eventually recruited 2051 people, 622 already with HIV and 1429 initially HIV-negative. The average at recruitment was 36. Those with HIV were more likely to be female (40% of the HIV-positive group versus 32% of the HIV-negative group) and non-white (43% versus 37%). The researchers estimate that 30% of the entire injecting drug population of the neighbourhood took part in this study and that the HIV-positive participants had an average of 20 viral load tests.

The researchers measured viral loads in the HIV-positive group and tested people in the negative group for HIV every six months. Participants were encouraged to recruit other injecting drug users and were paid CA$20 for every test done.

In the HIV-positive group, average viral load declined continuously between January 1996 and mid-1998: at the start of 1996 it was about 55,000, in mid-1997 17,000 and in mid-1998 8,000. After this it declined more slowly to under 1000 by mid-2000 and has stayed under 5000 ever since. The proportion of people who were on triple combination therapy (HAART) increased from 3.6% in 1996 to 69% in 2007.

In the HIV-negative group, 155 out of 1429 people acquired HIV during the study. The incidence density (the average rate at which people acquired HIV) was just under 2.5% a year. After an initial reading of 1.5% a year in mid-1996 it soared to the extremely high figure of 11.7% a year in early 1997 but then sharply declined. Rates of between 4.5% and 3.8% a year were measured intermittently till mid-2000, but after that incidence never exceeded 2% a year until the last measurement in early 2007, when it was 2.3%.

The raw data indicated that viral load in HIV-positive participants was strongly correlated with HIV incidence six months later in the HIV-negative participants. It was calculated that for every tenfold increase in community viral load the likelihood of HIV infection was multiplied by 3.57.

Because other factors could also affect HIV incidence, the researchers then adjusted this raw data by accounting for changes in other behavioural and demographic data in the HIV-negative cohort, as gathered in six-monthly questionnaires. Several other factors were found to influence HIV incidence. Being of white ethnicity reduced the likelihood of acquiring HIV by a third, being homeless increased it by 40%, and injecting cocaine (but not heroin) multiplied HIV incidence by 2.5 - the second most significant factor.

Once these were all taken into account, HIV viral load among the positive population remained the strongest predictor of HIV incidence in the negative group. In this adjusted model, every tenfold increase in community viral load multiplied HIV incidence by 3.32.

A post-hoc analysis of the data showed that the association was only statistically meaningful in the pre-HAART era when average viral loads were high. After January 1998, once the community viral load had fallen below 20,000, every tenfold increase in viral load raised HIV incidence by 70%, but this was no longer significant and could have been a chance finding. This limits the strength of the study’s findings because it was only able to confirm a link between community viral load and infection in the early years when viral loads were high.

Nonetheless this study strengthens the evidence base for the idea of ‘treatment as prevention’. The research team, which is headed by International AIDS Society president Julio Montaner, persuaded the provincial government of British Columbia last autumn to adopt such a policy – see this report – and the use of ARVs to prevent HIV has been identified by the World Health Organization as a top research priority.