Incorporating durability information in the comparison of proportions of patients with HIV suppression.

In some HIV clinical trials, the proportion of patients who achieve treatment success at a clinically meaningful time point (e.g., 16 or 24 weeks) and the subsequent durability of the treatment success after that time point are collected from two exclusive followup intervals. Two treatments are usually compared in terms of the proportion of patients achieving treatment success at the pre-defined time point and the subsequent durability information is ignored. However, combining the failure/success proportion at the pre-defined time point and the subsequent durability information in one test statistic could be more powerful if the experimental treatment is more efficacious than the control in that either fewer patients fail the experimental treatment at this time point or the responding patients have longer duration of viral suppression. In this paper, we propose a time-to-event type potency/durability endpoint which captures the information from the two exclusive followup intervals. A linear rank statistic to compare the two treatments in terms of this potency/durability endpoint can be interpreted as a weighted statistic to incorporate the potency information at a clinically meaningful time point (e.g., 16 or 24 weeks) and the durability information after that time point. The statistical hypotheses being tested by using this potency/durability endpoint and their clinical interpretations are discussed. A clinical endpoint study in anti-retroviral treatment naive patients is used to illustrate this method. Simulation studies show that this method is more powerful than comparing the success rates alone when the experimental treatment is also more durable in maintaining long term success.