LETTER TO THE EDITOR Conditioning on the propensity score can result in biased estimation of common measures of treatment effect: A Monte Carlo study (p n/a)