The final published paper on the VOICE trial in women in three African countries mainly reinforces what conference presentations have already shown: this ambitious trial failed to demonstrate the effectiveness of either oral pre-exposure prophylaxis (PrEP) or of a tenofovir-containing vaginal microbicide gel, and the reason for this was that only 25-30% of women actually used the study product, despite 88% claiming they used it.
There was one crumb of comfort for the researchers: in women randomised to use tenofovir microbicide gel, there was a 66% reduction in HIV infection in women who actually did use it. However, as the researchers themselves admit, the gel users were not a random sample of trial participants, and could have been at less risk of HIV. So the actual efficacy of the gel may have been lower.
The main results of the VOICE trial were reported by Aidsmap in 2013 when they were presented at the CROI conference in Atlanta. Briefly, roughly 5000 trial participants from South Africa, Uganda and Zimbabwe were randomised into five groups of just under 1000 each. These five groups were given to take or use daily:
- a tenofovir pill plus a placebo (dummy) tenofovir/emtricitabine (Truvada) pill
- a Truvada pill plus a placebo tenofovir pil
- two placebo pills
- a vaginal microbicide gel containing 1% tenofovir
- a placebo microbicide gel.
Retention in the trial was excellent, with 91% of women completing follow-up visits, and adherence appeared to be excellent too with 86% adherence reported on the basis of women's returns of empty pill boxes and gel applicators.
Low adherence – and high concealment of it
The effectiveness results told a different story, though. The oral tenofovir and tenofovir-microbicide arms were discontinued in autumn 2011 when it became clear there was not going to be a positive result and although the oral Truvada arm continued till August 2012, that did not show efficacy against HIV either. The end result was that there were actually 49% more HIV infections in participants taking oral tenofovir than in ones taking placebo, 4.4% more in patients taking Truvada, and 14.5% fewer in volunteers using tenofovir gel than in ones using a placebo. None of these differences were statistically significant.
This was not surprising when drug levels were assessed in a randomly-selected 22% of participants who took or used active product. Averaged through the course of the study, measurable drug levels were only seen in the blood of 30% of volunteers on tenofovir, 29% on Truvada and 25% using the microbicide. Even at the first quarterly visit, only 40% of participants had any drug in their blood and over 50% of participants allocated to active product never had a single blood test that showed they had actually used it. Not only were these low adherence levels startling; so were the lengths participants had gone to to conceal their low adherence. Participants clearly wanted to be in the VOICE trial, but didn’t want the thing it was testing.
Interviews and qualitative studied with the participants after the end of the trial showed a number of different reasons participants both had low adherence and concealed it. See this Aidsmap report for a full account, but in general four themes emerged
- participants joined the trial for the pregnancy and HIV testing and the medical monitoring it offered (and possibly for the stipend offered per visit, which was as much as $15);
- participants were afraid of being seen to have HIV drugs and assumed to have HIV;
- they knew it was a placebo-controlled trial, so stockpiled product until there was proof it worked, or gave it to others;
- or they believed one of the rumours circulating - that the drugs were harmful, that they made you infertile or actually gave you HIV.
As the final report says, unfortunately it was those most in need of PrEP who were least likely to adhere to it. Young women under 25 had 40% lower adherence than women over this age but HIV incidence in younger women taking placebo was nearly three times higher than in women over 25.
Similarly, women who had had one or more children and married women were, respectively, only half and one-eighth as likely to acquire HIV as childless and unmarried women, but were twice and three times more likely to take oral PrEP.
Did the microbicide gel have some efficacy?
This, however, is where a slight hint of advantage for the microbicide gel starts to creep in. Although the mean 25% adherence to the gel was, by drug levels in the blood, even lower than in other groups, it was no lower in younger women under 25 than in older women. And although adherence to it was higher in married women and ones who had had more than one child, the difference was not as great as it was for oral PrEP.
Furthermore, blood levels of tenofovir may have underestimated adherence to the gel. Vaginal swabs taken found tenofovir in a mean of 49% of samples. It is possible that the blood levels could tell the true story and, given the elaborate efforts trial participants made to conceal low adherence, the vaginal swab levels could be due to so-called ‘white-coat dosing’ – using the product solely before clinic visits. Alternatively, if the vaginal levels are reliable, the tenofovir in the gel could simply have failed to penetrate through the tissues and into the bloodstream in sufficient amounts to be detected.
Either way, women allocated to the microbicide gel with detectable blood levels of tenofovir were the only group in the whole study who benefited from significant protection against HIV. HIV incidence in members of the microbicide arm with no drug detected was 9.1% a year; in those with drug detected it was 1.9% year. After this was adjusted for the fact that women with drug detected were more likely to be married and have had children and therefore less at risk of HIV, there remained nonetheless a significant 66% reduction in HIV infections in women with evidence in their blood that they had used the gel (hazard ratio, 0.34: 95% confidence interval, 0.13 to 0.87; p=0.02.).
In contrast, having detectable drug in the blood was not protective for oral PrEP users. This at first seems a greater puzzle. However the threshold for drug detection was set at very low levels so detectable drug may not have indicated adherence sufficient to protect.
The 66% reduction in HIV infections in gel users is not a randomised result. As the researchers say, although they can adjust for the type of person gel users were, they cannot adjust for actual HIV exposure. It is possible that better adherence to the gel was correlated with lower vulnerability to HIV for a reason the researchers did not control for: lower levels of domestic violence or forced sex, for instance.
However the finding that the tenofovir gel, if used, may have offered some protection is one positive result in a trial that was in other ways a shock and, as a recent New York Times report emphasised, a challenge to researchers to remodel how they did prevention studies with disadvantaged groups. It may also bode well for the results of the FACTS 001 tenofovir-gel microbicide trial in South Africa, which will report its results at this year’s CROI conference in two weeks’ time.
Marrazzo JM et al. Tenofovir-based pre-exposure prophylaxis for HIV infection among African women. NEJM 372(6): 509-518. 2015.