A statistical model for the evaluation of barrier contraceptive efficacy

This paper describes an approach for the analysis of barrier contraceptive efficacy trials that accounts for timing frequency of intercourse and compliance. We allow exposure variables to vary for each act of intercourse and we control for timing of each act through a specific parametric function of the day of the act relative to the last day of the follicular phase of the cycle. The model can be used to examine the level of protection provided by a barrier versus no contraceptive method even when no control group of non-users is studied, as long as there are acts with no barrier use during the fertile window. We present results of a simulation study which examines performance of estimators and power under a variety of scenarios, including situations where an accurate benchmark for ovulation day is not available. As compared to the survival analysis approach commonly used in this setting, simulation results show that the new approach yields considerable gains in power to detect differences between the efficacy of contraceptive methods. An application to data from the FemCap versus diaphragm trial show results consistent with previous findings suggesting superiority of the diaphragm but also provides new evidence of the per act protection provided by both methods.