Comparison of chronic kidney disease trial designs and analysis strategies.

Abstract

BackgroundDespite the large burden of chronic kidney disease (CKD), it is challenging to conduct adequately powered clinical trials in this setting. Sound and efficient trials are needed to advance treatment. Various analysis strategies can be used to compare the efficacy of a parallel trial design with that of three two period trial designs.MethodsThe type 1 error rates and powers of various trial designs were calculated using simulated data from models fit to two recent CKD trials. In addition, we assessed the influences of a variety of analysis strategies and of the presence of a carryover effect.ResultsThe parallel and crossover designs (with analysis of change from baseline to the off treatment value) maintained the target type 1 error rate in all scenarios. In some scenarios, an open label design yielded inflated type 1 error rates. In many scenarios, the open label and delayed start designs had unacceptably low power and high type 1 error rates. Overall, the crossover design had the highest power by far, and always controlled the type 1 error rate.ConclusionThe recommended approach to a CKD trial is a two period design with an endpoint that is the rate of change in estimated glomerular filtration rate from pretreatment to off treatment. As compared to a parallel trial, a crossover study involves a considerably smaller sample size and shorter total follow-up duration. A crossover design may also be preferable for patients, and facilitates recruitment of a sufficient number of subjects.