We have found that a high proportion of Ugandan women in sero-discordant partnerships used the vaginal gels consistently throughout the trial. The average reported gel use at last sex act (92.9%) was higher than that reported in many other trials, although comparisons are complicated by the use of different measurements. The COL1492 trial, reported approximately 85% of the women using gel in 95% or more vaginal sex acts with clients . In the FHI trials of SAVVY in Ghana and Nigeria, women reported using the gel during approximately 76% and 78% of sex acts respectively, in the last 7 days, based on the participant’s mean gel use across all of their follow-up visits [5, 9]. Whereas in the Carraguard trial, in which on average 96% of the women reported using the gel at the last sex act, an applicator stain test to detect vaginal insertion indicated that women only used gel on average during 42% of their sex acts . In the FHI Cellulose Sulphate trial, women reported using gel in more than 80% of sex acts in the previous 7 days at each quarterly visit . Furthermore, in the CONRAD Cellulose Sulphate trial, women reported using the gel in approximately 87% of all sex acts  and in the same (CONRAD) trial, adherence was slightly lower in their Uganda site than in the other African sites. The HPTN 035 trial reported women using gel in 81% of their last sex acts, based on self-reported data collected at the quarterly study visits . The CAPRISA 004 trial estimated that on average 72% (median 60%) of self-reported sex acts in the last 30 days were covered by two doses of gel, based on the return of used applicators . Interestingly, the average reported gel use at the last sex act was also higher at the Uganda site than in the overall MDP 301 trial, which was reported at 89% . The reported high average gel use in this analysis may reflect that the women in the MDP 301 trial at Masaka knew that their partner was HIV positive. Known sero-discordance may go some way to explaining differences in adherence, and effectiveness, in recent PrEP trials [22, 23]. However, the higher adherence could also have been influenced by the fact that couples were enrolled into the trial together, not just the women. The evidence suggests that generally couples jointly decided on the use of gel and women reported feeling supported by their partners .
Few trials have reported on consistent gel use based on a composite measure of self-reported gel use, return of applicators and visit attendance, in the manner used in this analysis. This analysis demonstrated that over two-thirds of women reported consistently using gel for between one and two years. While some trials have observed an increase in gel use over time  and others have observed a decrease , we did not observe any changes in gel use over time. The main predictors of consistent gel use were age and household space. Few studies to date have reported on predictors of gel adherence [16, 17], although in a surrogate study  higher age was associated with higher adherence to gel use.
An important finding from this analysis is that consistent gel use was independently associated with the number of rooms in the household used for sleeping. Household space has not been shown to be associated to gel use in any previous studies and was not a predictor of consistent gel use in the overall MDP 301 analysis . Further qualitative research is necessary to understanding the reason for this finding. A possible explanation is that couples living in households with more space enjoy more privacy, thereby making it easier to apply gel before sex. None of the other variables that we evaluated in this analysis were predictive of consistent gel use. Our finding that educational attainment was not significantly associated with consistent gel use is supported by previous findings . It is of particular interest that condom use at the last sex act was not associated with consistent gel use, as gel adherence has been linked to condom use previously . There is substantial evidence that the use of vaginal microbicide gels enhance the sexual pleasure of women and their partners [26–29]. In this analysis, although consistent gel use was slightly higher among women who reported that gel improved sexual pleasure, the impact of the gel on sex was not significantly associated with consistent use.
The strength of this analysis is that we have used a composite measure of consistent gel use. However, the main limitation is that this measure still relies on self-reported sexual behavior and gel use data. Self-reported data are subject to both recall and social desirability bias. In the MDP 301 clinical trial we used a mixed methods and triangulation model to increase the accuracy of adherence and sexual behavior data [30, 31]. This model found that women were more likely to under-report sexual activity, condom and gel use in the administered questionnaires used for this analysis, when compared to self-completed diaries and in-depth interviews. Although this model found that most inaccuracies in the self-reported quantitative data were unintentional, there is a chance that consistent gel use has been overestimated in this analysis. Conversely, the fact that the composite measure required women to have attended at least 13 of their expected visits and to have remained in follow-up for at least a year, may have underestimated consistent gel use among women in the trial. Furthermore, underestimation of the consistency of gel use could result from the fact that women who completed fewer visits were more likely to be regarded as inconsistent gel users by the definition; for example, a woman who used gel at the last sex act in 13 out of 15 visits would be regarded as an inconsistent gel user compared to one who completed all the 26 visits and reported using gel at the last sex act in 24 visits. If the women who had attended fewer visits had completed all their scheduled visits, they may well have been considered consistent gel users.
This study suggests that long-term consistent gel use is high among sero-discordant rural couples in south-west Uganda. The fact that adherence was higher in this cohort than previously reported, suggests that factors affecting risk reduction behavior for women in sero-discordant relationships may be distinct from women who are not aware of their partners’ status. The HPTN 052 trial recently demonstrated that early initiation of antiretroviral use by HIV-infected individuals can substantially protect their HIV-uninfected sexual partners from acquiring HIV infection, with a 96% reduction in risk of HIV transmission . Women’s preferences for their own use of vaginal microbicides versus their positive partners’ use of treatment as prevention warrants further research in Uganda. This is the only microbicide trial to report on adherence among sero-discordant couples. However, the evidence regarding the benefit of treatment as prevention may prevent future HIV prevention trials from being able to ethically recruit sero-discordant couples, thereby making it more difficult to assess the use of alternative prevention modalities.