When conducting network meta-analysis (NMA) – that is, a technique that involves comparing multiple treatments simultaneously in one analysis by combining both direct comparisons of treatments trialed against each other in randomized control trials and indirect comparisons, based on a common comparator – bias can be caused by effect modification, potentially leading to inaccurate conclusions.
Generally speaking, effect modification occurs when the relative treatment effect differs between different levels of patient characteristics. For example, if the treatment is less effective in younger patients, then age can be considered an effect modifier. And in the context of NMA, bias may occur when the distribution of such effect modifiers differs across trials in the network.
A number of population-adjustment NMA methods exist to handle the complexities caused by effect modification. For example, when individual patient data (IPD) are available for all trials in the network, effect modification can be modeled through IPD network meta-regression (NMR), though this situation is rare. For NMAs that include at least one trial with available aggregate-level data (AgD), methods to deal with effect modification include matching-adjusted indirect treatment comparison, simulated treatment comparison, and multilevel NMR. But each of these methods have various limitations that make them unsuitable for certain scenarios.
Additionally, commonly used approaches rely on the assumption that the effect modifiers impact all treatments in the evidence network in the exact same way, that is, they rely on the shared effect modification (SEM) assumption. In a previous webinar on the topic, Dr. Harari argued that the assumption of SEM is “extremely limiting” and “rarely reasonable to make.”
What’s more is that they disregard available subgroup information from aggregated data in the evidence network.
In a recently published article for Research Synthesis Methods, Cytel Senior Research Principal, Statistics, Ofir Harari, Senior Biostatistician Mohsen Soltanifar, Associate Director & Research Principal and Econometrician Andre Verhoek, Senior Research Consultant, Statistics, Caitlin Daly, and Vice President, RWA, Bart Heeg, along with their co-authors, propose instead using Network Meta-Interpolation (NMI), which uses subgroup analyses to adjust for effect modification but does not assume SEM.
Network Meta-Interpolation, as they explain in the article, is “a novel approach to balancing patient populations from various studies prior to NMA, using regression techniques to relate outcomes and effect modifiers. NMI consists of a data enrichment step, followed by two regression steps. The output is a balanced, NMA-ready AgD, evaluated at an effect modifier configuration of the researcher’s choice, which could reflect the pivotal trial or a real-world population of interest.”
Through their analysis, they conclude that Network Meta-Interpolation is a simple, accurate, and robust option for balancing studies in an NMA, provided its data requirements are met.
To learn more about Network Meta-Interpolation and the methods Cytel’s researchers used to arrive at these conclusions, read the full article here:
Read more from Perspectives on Enquiry & Evidence:
Sorry no results please clear the filters and try again