Reilly and Evans correctly note that the mortality reduction for LMWHs was a critical variable in our cost-effectiveness analysis. That is why we used meta-analysis to derive estimates of treatment effectiveness based on the best available evidence. The weight of evidence from clinical trials does not support their conclusion that LMWHs and unfractionated heparin have the same effects on mortality rates in the treatment of deep venous thrombosis.

Deleting studies one at a time is a standard procedure used in meta-analysis to determine whether one or more studies exerted a disproportionate influence on the treatment effect. Deletion of the study by Hull and colleagues (1) resulted in loss of statistical significance but little change in the estimated treatment effect size. It is not logical for Reilly and Evans to suggest that this study should have been singled out for exclusion because it did not use a weight-based nomogram for the dosing of unfractionated heparin when none of the other studies used such a dosing regimen. Although we agree that evidence supports the use of the weight-based nomogram, our findings are likely to apply directly in community settings, where the weight-based nomogram may not be widely used.

Reily and Evans correctly point out that time costs associated with unfractionated heparin administration might be important if one adopts a cost-minimization approach to decision making. In a post hoc analysis, LMWH treatment became cost-saving when we assumed that unfractionated heparin administration required more than 1.4 hours of additional nursing or pharmacy time per hospital day. We suspect that less time is required in most instances of deep venous thrombosis treatment with unfractionated heparin. We agree with Reilly and Evans that this issue should be addressed in future studies. In the meantime, we stand by our conclusion that LMWHs are highly cost-effective, if not cost-saving, for the inpatient treatment of deep venous thrombosis.