The problem of incidence

At the 2009 International Aids Society (IAS) Conference in Cape Town,¹ early data from the PrEP trial in Thai injecting drug users (CDC 4370) was unveiled. This contained a strong hint that, as in the Ghana PrEP trial, risk behaviour declined drastically during the trial, possibly predicting lack of power to generate a statistically significant result.

Injecting and needle-sharing both declined precipitously during the trial. Three months after enrolment, 32% of participants were injecting and only 4% reported having shared needles; by twelve months after enrolment, the figures were down to 25% injecting and only 2% sharing needles.

Because of this decline in risk behaviour, and also because HIV prevalence in injecting drug users in Bangkok declined from 54% in 2002 to 29% in 2007, a decision was therefore taken to increase the number enrolled to 2400. It was calculated that the number of HIV seroconversions needed for this trial to have any chance of demonstrating efficacy is 40.

In December 2009, the US CDC announced that it was experiencing the same problem with the TDF2 (CDC 494) trial in Botswana.

HIV incidence in the study had been much lower than expected, due in part to declining HIV incidence in Botswana as a whole, but also due to the intensive monthly prevention counselling received by all trial participants.

The study was originally designed to recruit 1200 participants, but investigators concluded that even if recruitment were doubled to 2400 participants, the study would be unlikely to show a difference in the risk of HIV infection between those receiving Truvada and the placebo group.

Another development that was difficult to quantify in the context of this study was the impact of widening availability of antiretroviral therapy in Botswana. It has been difficult to measure the extent to which antiretroviral treatment has reduced the rate of new infections and therefore affected the outcome in prevention trials like TDF2. If infection rates decline, and if accompanying prevention measures such as counselling and condom provision have a consistent impact in trials of prevention technologies such as PrEP, microbicides and vaccines, future trials may need to be much larger in order to detect differences.

The power some prevention studies to detect a statistically meaningful difference has also been undermined by challenges in retaining participants in the study. Pregnancy, for example, usually results in suspension from studies. If participants move out of the area, or cannot attend clinic visits, they are also lost to follow-up.

In the end, TDF2 did find statistically significant efficacy for PrEP in men, and also more adherent women, but its problems illustrate a general dilemma for researchers.