Text this: The number of patients and events required to limit the risk of overestimation of intervention effects in meta-analysis--a simulation study.