An Empirical Study of Bias in Randomized Controlled Trials and Non-randomized Studies of Surgical Interventions