A great concern in respect to the deployment of Pre-Exposure Prophylaxis (PrEP) as an HIV prevention tool is risk compensation – i.e. the possibility that the protection afforded by PrEP would itself encourage sexual risk-taking (STI blogs/Sugarman & Mayer). Recent studies of PrEP efficacy would seem an unpromising context in which to assess the likely impact of such a factor, given that a dependable level of HIV protection – such as might encourage risk-taking behaviour – is precisely what these studies are seeking to demonstrate. This difficulty has not deterred the researchers, however. The most recent attempt to evaluate risk compensation in such a context (Marcus & Grant) (M&G) is a spin-off from the recent iPrEx study of the efficacy of PrEP using metricitabine/tenofovir disoproxil fumarate (FTC/TDF) in 2,500 MSM and transgender women (recruited at 11 sites in Peru, Ecuador, South Africa, Brazil, Thailand and the US) (Grant and Glidden (2010)). Marcus and Grant’s study follows closely on another spin-off risk-compensation study arising from the Partners PrEP study covered in an earlier STI blog (STI blog/Mugwanya & Baeten).
Not only were participants in the Marcus & Grant study uncertain of the degree of protection that PrEP would afford (the eventual outcome was a 44% reduced risk of HIV acquisition, and 92% reduction for participants with detectable drug levels); they were also, of course, unaware of whether they had been assigned to the intervention or placebo arm of the study. Marcus & Grant therefore use perceived treatment assignment and belief in PrEP efficacy as surrogates for the assurance of HIV protection which future PrEP users might be tempted to see as offsetting any increase in their risk-taking behaviour. Sexual practices were assessed at baseline and quarterly thereafter; perceived treatment assignment and belief in PrEP efficacy were assessed at 12 weeks. Practice of receptive anal intercourse without condom (ncRAI) was taken as the indicator of sexual risk-taking behaviour.
Findings of the study were as follows. Mean number of RAI partners decreased (10-6) in the first twelve weeks from baseline, then remained more or less constant, while condom use with RAI partners rose in the same period (c.50-c.75%), and then plateau-ed. The researchers found no indication of any association between risk-taking and participation in the trial – we may wonder though whether anyone would have expected there to be one. However, a clear association emerged between ncRAI/no ncRAI at base line and perceived assignment to PrEP/placebo (53%/47% for PrEP, and 61%/39% for placebo). Marcus and Grant are inclined to interpret this association as a reflection of the impact of risky behaviour on attitudes to treatment (“the belief could have served as a psychological mechanism to reduce concerns about sexual behaviour”), rather than that of treatment optimism on sexual behaviour. However, the key finding for our authors was the absence of any difference in risk behaviour (ncRAI) between the treatment optimists and the treatment pessimists – and between those who claimed to have greater confidence in PrEP efficacy and those who claimed to have less. This suggests the absence of a risk compensation effect.
A number of study limitations are discussed in the paper (not least of which that of reporting bias), but not the limitation that is patently the most serious: namely the questionable adequacy of perceived treatment assignment as a surrogate for the assurance of a certain level of PrEP efficacy.
On this point Mugwanya and Baeten (W&B) offer an interesting comparison (STI blog/Mugwanya & Baeten). Participants in the intervention arm of the Partners PrEP study were informed of the study outcome (indicating a 96% reduction in HIV transmission between infected and uninfected partners), and follow-up, including monitoring of STI incidence, was continued for a further year, allowing comparison between sexual behaviour of participants before and after knowledge of PrEP efficacy. M&B seem to corroborate the Markus & Grant’s finding of no risk compensation – and on the basis of far more solid evidence. However, M&B’s finding relates to the behaviour of heterosexual couples, and is prima facie unlikely to be generalizable to the behaviour of MSM and transgender women.