An overview of methods for network meta-analysis using individual participant data: when do benefits arise?

Thomas PA Debray, Ewoud Schuit, Orestis Efthimiou, Johannes B Reitsma, John PA Ioannidis, Georgia Salanti, Karel GM Moons
  • Statistical Methods in Medical Research, August 2016, SAGE Publishing
  • DOI: 10.1177/0962280216660741

Network meta-analysis using individual participant data

What is it about?

Network meta-analysis (NMA) is a common approach to synthesize the efficacy of multiple therapeutic interventions, and to compare their relative efficacy. Most NMA are based on aggregate data, that is, published summary estimates of relative treatment effect (e.g. as identified through a systematic literature search). It is, however, also possible to conduct an NMA with the raw individual participant data (IPD) from each study. We we set out to explore common challenges and potential advantages of NMA based on IPD.

Why is it important?

Our findings show that access to raw participant data may help to resolve inconsistency and reduce heterogeneity in network meta-analysis. This, in turn, may enhance the validity and clinical usefulness of summary estimates of comparative treatment effects.

Perspectives

Dr. Thomas Debray (Author)
Julius Center for Health Sciences and Primary Care

Network meta-analyses are used increasingly often to inform healthcare professionals and assist medical decision making. Over the past few years, however, it has become apparent that their results can sometimes be misleading. Although access to IPD may help to address potential concerns, obtaining and combining such data tends to be very difficult. For this reason, I think that any NMA should start with summarizing aggregate data. Obtaining IPD could then be prioritized for those trials that compare the treatments of primary interest and for which direct and indirect evidence are in disagreement, or for which corresponding effect estimates are heterogeneous. When IPD are eventually used in NMA, thorough description and explanation is essential for the acceptance and understanding at regulatory and HTA levels.

The following have contributed to this page: Dr. Thomas Debray