Article type
Year
Abstract
Background: Many traditional meta-analyses summarize the results of studies that compare an intervention with placebo. Often there is limited or insufficient evidence from direct head-to-head comparisons of competing interventions. A network meta-analysis enables the comparison among different interventions in the absence of direct head-to-head comparisons. Evidence from a network meta-analysis is reliable if studies with very similar study characteristics and eligibility criteria (study designs, populations, interventions, comparisons and outcomes) are combined.
Objectives: To discuss the methodological and process challenges involved in network meta-analyses and how to overcome them.
Methods: We used the example of two Cochrane overviews of reviews on biologics involving network meta-analyses to illustrate the methodological challenges encountered and how we overcame them. We compared the efficacy and safety of six different biologics in patients with rheumatoid arthritis in the first overview. We found limited evidence on the adverse effects such as tuberculosis, life threatening fungal infections and lymphomas that may be common with biologics, so we undertook a second overview to assess the adverse effects of nine biologics across different indications of use. The unit of analysis was different in the two overviews: Cochrane systematic reviews in the first, and clinical trials in the second.
Results: We faced different methodological considerations, including eligibility criteria and methods of assessment of risk of bias and analysis depending on the type of overview most appropriate to answer our question. Challenges included managing a large volume of data for screening and extraction; achieving consensus with a large team of extractors; preparing data for analysis; and presenting results using the Review Manager overview template.
Conclusions: Our experience showed that where limited evidence is available for direct comparisons of different interventions, well-conducted indirect comparisons based on network meta-analyses are challenging to undertake, but necessary to guide health decision making.
Objectives: To discuss the methodological and process challenges involved in network meta-analyses and how to overcome them.
Methods: We used the example of two Cochrane overviews of reviews on biologics involving network meta-analyses to illustrate the methodological challenges encountered and how we overcame them. We compared the efficacy and safety of six different biologics in patients with rheumatoid arthritis in the first overview. We found limited evidence on the adverse effects such as tuberculosis, life threatening fungal infections and lymphomas that may be common with biologics, so we undertook a second overview to assess the adverse effects of nine biologics across different indications of use. The unit of analysis was different in the two overviews: Cochrane systematic reviews in the first, and clinical trials in the second.
Results: We faced different methodological considerations, including eligibility criteria and methods of assessment of risk of bias and analysis depending on the type of overview most appropriate to answer our question. Challenges included managing a large volume of data for screening and extraction; achieving consensus with a large team of extractors; preparing data for analysis; and presenting results using the Review Manager overview template.
Conclusions: Our experience showed that where limited evidence is available for direct comparisons of different interventions, well-conducted indirect comparisons based on network meta-analyses are challenging to undertake, but necessary to guide health decision making.