We purposely did not include the NNT and the NNH because, as we explicitly stated in our article, our meta-analysis was done to "support hypotheses rather than provide directives for clinical practice."1
The NNT and NNH are useful indices to describe the trade-off between the benefit and the harm produced by a new treatment or regimen.2 In the case of our meta-analysis, the concept of NNH as a trade-off for NNT is less useful because both the benefit (thrombosis prevented) and harm (bleeding complications) favor the use of preoperative prophylaxis. With this in mind and using the data extrapolated by Dr Ghosh (ie, NNT for all thromboses, 21; NNH for major bleeding episodes, 41; and NNH for minor bleeding, 25), it is evident that treatment of 100 patients with preoperative prophylaxis rather than postoperatively initiated prophylaxis results in about 5 incidents of DVT prevented, 2 major bleeding episodes avoided, and 4 minor bleeding episodes avoided. In other words, treating 9 patients with preoperatively rather than postoperatively initiated prophylaxis prevents 1 "bad outcome" of either DVT or bleeding. This is of clear clinical importance if the results of our hypothesis-forming meta-analysis are confirmed by a randomized trial that directly compares LMWH prophylaxis administered either 12 hours preoperatively or 18 to 24 hours postoperatively. The timing of prophylaxis is potentially important. The concept of a trade-off between NNT and NNH may be more important when prophylaxis is administered immediately prior to surgery (when the potential for increased bleeding is greater), rather than 12 hours preoperatively.