We included randomized controlled trials, quasi-randomized controlled trials (that is, studies assigning patients to treatment arms by birthday or date of admission), nonrandomized cohort studies, and studies that used a historical cohort.