Patrick Sullivan of Emory University next presented similar data from larger cohorts in the cities of Lusaka and Kigali, the capitals (and largest cities) of Zambia and Rwanda respectively. A total of 2993 serodiscordant couples were followed up between 2002 and late 2008 in these two cities.
In these couples, the HIV-infected partner started antiretroviral therapy when their CD4 cell count was below 200 cells/mm3 or if they developed symptomatic HIV disease (WHO stages III and IV). Every three months, the HIV-negative partners were tested for the virus and received risk-reduction counselling. Information on sexual risk behaviour was collected at every screening visit. Vaginal smears were also used to check for the presence of sperm, and information was gathered on pregnancies.
The median duration of follow-up was 512 days, during which a total of 175 HIV infections were recorded. New infections were confirmed (by sequencing) as originating with the HIV-positive partner. Those that could not be confirmed (roughly 15%) were not considered in the analysis as their source was uncertain.
Of the confirmed within-couple transmissions, the overwhelming majority (171) occurred in couples where the HIV-infected partner was not taking HIV treatment. The remaining four infections occurred in couples during HIV treatment. This resulted in an HIV incidence rate, during antiretroviral therapy, of 0.7% per 100 couple-years, compared to an incidence of 3.4% per 100 couple-years when HIV treatment was not being taken – a fivefold risk reduction (relative risk = 0.21; 95% CI, 0.08-0.59).
Another, more conservative, estimate included two people who seroconverted in the three-month period after their partner began taking ART. Considering these two as 'infections while on ART' increased the count of on-ART infections to six, for an incidence rate of 1.0 per 100 couple-years, and a threefold reduction in risk (relative risk = 0.32; 95% CI, 0.14-0.73).
The investigators found comparable rates of pregnancy and self-reported condom use between the two groups, although sexual-risk-level assessment largely depended on self-reporting. To try to control for possible underreporting of risky sex, vaginal smears were also taken from female participants to test for the presence of semen. Semen was actually found in somewhat fewer women in on-ART couples (2.2% vs 3.1%), and combined indicators of risk were also somewhat lower in on-ART couples (19% vs 25%, p<.05 for both).
There were also significant differences between the two cities, with people in Kigali at half the risk of those in Lusaka. This was not explained, although it was consistent with several previous study findings. Viral-load measurements were also not routinely done and were not available for analysis; nor were adherence data.
In summary, Sullivan concluded that three- to fivefold lower rates of HIV transmission were seen in mixed-status heterosexual couples on ART in the capital cities of Rwanda and Zambia. However, both presenters also stressed that ART should not be considered a primary means of HIV prevention, but part of a combination approach to prevention. In addition, most HIV-positive persons in Africa are either not aware of their status, or are not on ART because they do not qualify (based on CD4 cell counts).
The reduction in risk of transmission seen in the Rwanda/Zambia study falls considerably short of the efficacy assumed for antiretroviral therapy in preventing transmission by a World Health Organization modelling exercise published in late 2008. That model projected a very substantial decline in HIV incidence if all people with HIV could be treated, assuming that treatment reduced the risk of transmission by 99%.
However, speaking at a CROI symposium on the global epidemic on Sunday, epidemiologist Christophe Fraser of Imperial College, London, noted that his own modelling, using the other assumptions contained in the WHO model, suggests that if the preventive efficacy of ART is even a little less than 99%, there is a much less substantial decline in HIV incidence, while an efficacy of 80% would result in a paradoxical increase in HIV incidence in the short term.
He called for multiple groups of epidemiologists to explore the issue of 'treatment as prevention' very carefully before policy is made, with a particular focus on determining whether all the assumptions in models are based on robust evidence.