Methods: We compare the performance of three estimators: the unadjusted estimator (which ignores baseline variables), and two adjusted estimators called the standardized and logistic regression estimators (which leverage information in baseline variables). Our comparisons are based on re-analyzing data from a phase 3 trial for treating severe stroke, called the CLEAR III trial, and by a simulation study that mimics features from that dataset.

Results: Re-analysis of the CLEAR III data shows that confidence intervals from the standardized estimator are 10% narrower than when the unadjusted estimator is used. In the simulations the standardized estimator requires 29% less sample size to achieve the same power as the unadjusted estimator. The simulations also show that the standardized estimator has slightly better precision compared to the logistic regression estimator.

]]>