This review addressed a clearly defined question and utilised appropriate methods to minimise bias in the acquisition, selection and appraisal of evidence.
The authors' conclusions reflect the evidence presented but the reliability of the findings are moderate to low because of uncertainties inherent in the evidence base. As the author's acknowledged, applicability of the findings to all patients with VTE was questionable as the randomised controlled trials on which the analysis was based were likely to have excluded patients with more severe disease status. Most of the linkages in the network were based on individual trials (maximum two) so the estimates of heterogeneity were imprecise and uncertain. Any biases related to reporting, trial conduct or trial publication could have a profound effect on network geometry with potentially important impacts on the conclusions.
The authors acknowledged other uncertainties related to the problems of assuming exchangeability between studies (variable follow-up and lack of individual patient data). They also highlighted the lack of direct comparisons, need for longer follow-up and problems with low power associated with sparse data particularly with (relatively) rare adverse events. Thus although the conclusion that oral anticoagulants were effective compared to placebo or standard care was likely to be reliable, conclusions regarding relative effectiveness were probably unreliable, particularly where the ratio of benefits to harms was concerned.