Use of covariates in randomized controlled trials