Adjustment for baseline characteristics in randomized trials using logistic regression: sample-based model versus true model.

Sample-based adjustment yields similar results to exact adjustment in estimating the treatment effect. Sample-based adjustment is preferable to no adjustment.