The data used for calculating ORs for the outcome of safety also puzzled us. The authors1 reported that intracerebral bleeding occurred in 16 patients. It seems to us that data for the trials of Hillbom et al,3Diener et al,4and Sherman et al5 for this outcome were 1/76 vs 0/72, 2/272 vs 3/273, and 4/877 vs 6/872, respectively. But, we cannot replicate the OR for the trial of Hillbom et al.3 We think the continuity correction for sparse tables, if used, should have been reported, since this information is important for calculating ORs. And for this outcome, the number of patients in the trial of Sherman et al5 is not 1,335. Overall, we think that in order to make the analysis more transparent to readers, the raw data need to be reported, and the time horizon for each outcome (during treatment period or at the end of follow-up) needs to be specified.