Conditioning on the propensity score can result in biased estimation of common measures of treatment effect: a Monte Carlo study | Synapse