Should Network Meta-analysis Become the Standard in Evidence-based Clinical Practice?

Network Meta-analysis (NMA) is a statistical tool that allows comparing and estimating the relative effects of multiple interventions at the same time. Although its popularity has increased in the latest years, many questions remain unsolved before it can be proposed as the standard method of evidence synthesis for informing evidence clinical practice. In this thesis, using systematic surveys of the literature, we aimed to determine 1) the extent to which NMA can be used to answer current clinical questions; 2) whether systematic reviews (SRs) using NMA report the same results as SRs using head-to-head comparisons (HTHC); and 3) the robustness of the rankings obtained from NMA to the exclusion randomized clinical trials from the network, and the impact of increasing decision thresholds on the ranking probabilities. We observed that only 25.3% (205 out of 809 Cochrane SRs published in a 1-year period) of the SRs had questions in which an NMA was necessary; that SRs using NMA to assess the effects to stents in patients undergoing percutaneous coronary intervention reported the same results as SRs using HTHC in 44.8% to 83.3% of the cases; that rank probabilities and the rankings remain reasonably constant when excluding trials from the analysis (with an overall mean absolute change of 4.3% in rank probabilities across NMAs), but there are cases in which dramatic increases or decreases of the rank probabilities, and switches in the treatments ranked first and second can be observed; and that increasing the threshold to claim superiority may result in important changes in rank probabilities, which in some cases lead to the first treatment having extremely low probabilities of being distinguishable as the best. These issues suggest that because NMA is still in its infancy compared to HTHC, more research and guidance for its use are necessary before it can be claimed that NMA should become the standard for comparing treatment effectiveness.Ph.D