Well Morenike a composite score is certainly on the card. But I believe there is room for evaluation of individual indicators. For example the proportion of all discordant couples actually accessing the service, proportion of those who take up the service and achive a threshold adherence level, and of course the big one is what proportion of couples retain the sero-discordant status. It would be necessary to have the full description of the three models to provide a more comprehensive menu of type of measurements and method of measurements.

I believe we are dealing with extermaly low, if any, incidence. At the same time the only accurate measure of best PreP delivery would always be keeping the serodicordant status of a couple unchanged.

As with other rare end-points I would personally value study collaborations over any composite scores. Maybe it would be better to set standars for PreP studies, especially adherence measures and data collection, and merge datasets from different studies for analyses?

Thanks Justyna for bringing in the collaboration and merging data sets, as indeed the occurrence of end point will always be low. And with this the need for standardization of processes and procedures is critical to derive reliable and valid findings

Thanks for the responses. I actually asked the question because I needed guidance as I develop a study protocol. My the points raised, I have proposed to look at the study HIV incidence - an effective model would result in 0 incidence. The less effective the model is, the higher the rate of seroconversion. All the parameters critical for adherence to drug regimen adherence would then be evaluated.