Original research articleWill microbicide trials yield unbiased estimates of microbicide efficacy?
Introduction
The clearest scientific design for assessing the efficacy of a candidate microbicide in reducing acquisition of sexually transmitted infections — that is, how well it works under ideal conditions when used correctly at every coital act — would be random assignment to the microbicide or a placebo. Trial participants would use the assigned product at every act of intercourse for the intended duration of their participation in the trial, and access to other means of preventing infections, such as condoms, would be denied. If there were equal numbers of participants in each group and no early discontinuations from the trial, the relative risk (RR) would be estimated as the number of infections in the microbicide group divided by the number of infections in the placebo group. Efficacy would be estimated as 1−RR, the proportionate reduction in the risk of infection. For example, if 60 infections occurred in the microbicide group and 100 in the placebo group, then RR=0.60, and the estimated proportionate reduction in the risk of infection through the end of the follow-up period is 40%.
In reality, of course, such a trial cannot be implemented. Since all participants in a microbicide trial are likely to be at risk for sexually transmitted infections, withholding condoms, which are now the only proven method for preventing these infections, would be unethical; in fact, current standards require that condom use be strongly encouraged in both study arms. Moreover, we cannot force trial participants to follow instructions exactly; intercourse without the assigned microbicide or placebo will inevitably occur. Thus, in a typical trial [1], participants are randomly assigned to a microbicide-plus-condom arm or to a placebo-plus-condom arm, and some nonadherence with these assignments is expected.
In this paper, we explore the potential impact of nonadherence to microbicide use and different patterns of condom use on the interpretation of the results of such a study.
Section snippets
Effect of nonadherence
First, consider the effect of nonadherence in a situation in which condoms (and other preventive measures) are prohibited or simply not used. Let:
- N
the number of participants in each arm (assumed to be equal in the microbicide and placebo groups)
- R
infection rate in the absence of microbicide (or placebo) use
- em
microbicide efficacy (the reduction in infection risk caused by the microbicide)
- ep
placebo efficacy (the reduction in infection risk caused by the placebo)
- γ
proportion of those who use
Effect of addition of condoms
We consider next the trial design in which participants in both the microbicide and the placebo arms are counseled to use condoms. Let:
- N
the number of participants in each arm
- R
infection rate in the absence of microbicide (or placebo) or condom use
- ec
condom efficacy (regardless of whether microbicides are also used)
- em
microbicide efficacy (regardless of whether condoms are also used)
- α
proportion of those who use condoms in the placebo arm
- β
proportion of those who use condoms in the microbicide arm
- γ
Conclusions
We have shown that both microbicide nonadherence and some patterns of condom use may seriously impair our ability to identify an effective microbicide. Under many plausible assumptions, effectiveness results of typical trials will substantially underestimate the efficacy of the candidate microbicide product being studied. Therefore, we could have a microbicide that is highly efficacious but not be able to tell that from our trials. The underestimation of effectiveness cannot be resolved merely
Acknowledgment
We are grateful to Elizabeth Raymond for helpful comments on a draft of this paper.
References (3)
- et al.
Effectiveness of COL-1492, a nonoxynol-9 vaginal gel, on HIV-1 transmission in female sex workers: a randomised controlled trial
Lancet
(2002)
Cited by (16)
Estimating contraceptive efficacy: The case of spermicides
2013, ContraceptionMicrobicide drug candidates to prevent HIV infection
2007, LancetCitation Excerpt :The provision of lunch or entertainment are mechanisms that may help keep women motivated.124–127 Another concern is the time that women are “off-product”,128 which decreases the statistical power of the study. Women might stop using the product because they become pregnant due to non-compliance with the study protocol.
HIV prevention research: the ecstasy and the agony
2008, The LancetEfficacy of Carraguard for prevention of HIV infection in women in South Africa: a randomised, double-blind, placebo-controlled trial
2008, The LancetCitation Excerpt :The ability to detect a protective effect could have been compromised by poor adherence. If the estimated proportion of sex acts with gel use is accurate, then the trial was underpowered to detect anything other than high product efficacy levels of 70% or more.36 The low adherence that we recorded in this clinical trial could have been be due to feasibility issues—such as running out of or forgetting to use gel, which were the most common reasons for non-use given by a subset of women (n=1601) whom we interviewed at study exit—and could suggest that even a highly efficacious coitally-dependent product will have insufficient effectiveness in real-life settings.