Rule-based ranking schemes for antiretroviral trials

Endpoints are a continuing source of controversy in clinical trials of antiretroviral (specifically, anti-HIV) treatments. The most visible disagreement is about the respective roles of morbidity and mortality as endpoints, and laboratory measurements as endpoints. Laboratory measurements have been intensely examined as possible surrogates for clinical outcomes, but the definition of the usual clinical outcome - first occurrence or recurrence of an AIDS-defining condition or death - has received little critical scrutiny. First disease progression has serious weakness as an endpoint, and one should consider alternatives. In this paper, we suggest using rule-based schemes to rank patients' post-randomization histories and then using the ranks as an outcome measure, an extension of the work by Follmann et al. on heart disease. We evaluated six rule-based ranking schemes for antiretroviral trials by applying them to 60 participants in CPCRA 002 and comparing the results to subjective rankings given by five experts. The expert's rankings were in good agreement with each other, and the six rule-based schemes were clearly differentiated by their degree of agreement with the expert's rankings. The ranking scheme most in accord with the experts ranked patients first by seriousness of their most serious AIDS-defining disease, second by the timing of that disease, and the third by the total number of AIDS-defining diseases they experienced. Finally, we used this rule-based rankings to re-analyse CPCRA 002.