VOICE study reports low adherence as reason for lack of efficacy for PrEP: anal sex common in African heterosexuals

Simon Collins, HIV i-Base

The five-arm placebo controlled VOICE HIV prevention study randomised over 5000 HIV negative women in South Africa (n= 4077), Zimbabwe (n=630) and Uganda (n=322) to one of three active (oral TDF, oral TDF/FTC and TDF gel) or two placebo arms (oral and gel).

Interventions by the study DSMB stopped both the tenofovir oral and gel arms in late 2011 due to lack of efficacy, but the data analysis from the full study, including the efficacy data for the discontinued arms, was presented for the first time as an oral presentation at CROI 2013. As previously suggested, low adherence drove the lack of efficacy.

The mean age at baseline was approximately 25 years, with 20% married, >90% using oral (>20%) or injectable (70%) contraceptives. More than 20% or participants reported having more than two male partners in the previous three months but condom use was high (85% on last occasion).

Perhaps most notably, with information collected using computer assisted self-reported questionnaires (A-CASI), rather that by interview, 18% of women had anal sex in the previous three months. This has rarely been listed as a risk factor in previous prevention studies.

Approximately 10% of participants were lost to follow up, but >95% of the projected person years of follow up (PYFU) were available. Adherence by self-report and returned medication bottles was close to 90% in all arms.

Overall, 344 women became HIV positive during the study, 22 of whom were excluded due to likely seroconversion at study enrollment leading to an incidence of 5.7%.

Primary efficacy results in terms of new infections were not statistically different between arms and are detailed in Table 1. There were also no safety differences between arms. The overall pregnancy rate was 7.8%.

The PK substudy, essential to explain both positive and negative findings in PrEP studies, included over 3,200 samples from 773 participants (median 4 samples, range 1-12 per participant), including all seroconvertors. Less than 40% of women in any of the active arms had detectable tenofovir (test sensitive to >0.3 ng/mL) during the first three months of the study and this fell steadily to 20% by the sixth quarterly visit. More than 50% of the women had no detectable tenofovir at any point during the study.

Although there was insufficient PK data to determine efficacy rates for active arms, women older than 25 and married women were both more likely to have detectable tenofovir, and these factors correlated to lower risk of catching HIV. In South African sites, the incidence of HIV acquisition per 100 PY was 8.7 (7.6, 10.0) vs 4.7 (3.8, 5.8) in women younger vs older than 25 and 0.9 (0.2, 2.7) vs 7.5 (6.6, 8.4) in married vs unmarried women.

Table 1: Primary efficacy results in VOICE study
PYFU No. infections Incidence /100 PY (95%CI) HR (95%CI)(vs placebo) p-value
Oral TDF 823 52 6.3(4.7, 8.3) 1.49(0.97, 2.3) 0.07
Oral TDF placebo 837 35 4.2(2.9, 5.8)
Oral TDF/FTC 1285 61 4.7(3.6, 6.1) 1.04(0.7, 1.5) > 0.2
Oral TDF/FTC placebo 837 60 4.6(3.5, 5.9)
TDF gel 1026 61 5.9(4.5, 7.6) 0.85(0.6, 1.2) > 0.2
Gel placebo 1030 70 6.8(5.3, 8.6)


This major prevention study failed to show benefit from potentially effective interventions due to minimal adherence, with rates that were especially low in the people at the highest risk of infection, even when incidence rates were higher than expected.

This highlights the urgency of more acceptable and effective interventions, including long-acting parenteral formulations.

Although minimal data are collected on partners in the study, this will be one of the many factors that will try to explain why participants were not actively retained in the study.


Marrazzo J et al. Pre-exposure prophylaxis for HIV in women: daily oral tenofovir, oral tenofovir/emtricitabine, or vaginal tenofovir gel in the VOICE study (MTN 003). 20th CROI, 3-6 March 2013, Atlanta. Oral abstract 26LB.

Webcast (third presentation)

Links to other websites are current at date of posting but not maintained.