Each week, we will be publishing layman’s abstracts of new articles from our prestigious portfolio of journals in statistics. The aim is to highlight the latest research to a broader audience in an accessible format.
Matching‐adjusted indirect comparisons: Application to time‐to‐event data. Statistics in Medicine. 2021; 40: 566–577. https://doi.org/10.1002/sim.8789, , .
The gold standard to compare a drug A to another drug B, in a particular population, is to conduct a Randomized Clinical Trial (RCT) and perform a head to head comparison in this setting. This option is not always feasible but an indirect comparison of A versus B can be conducted when drug A is assessed in a RCT ”A” and drug B is assessed in another RCT ”B”, in particular when both RCTs ”A” and ”B” have the same control arm. However such indirect comparison must be performed with caution as it can be biased by cross trial population differences. Even though indirect comparisons can be performed in the context of meta-analyses, unbiased indirect comparisons require some adjustments that generally necessitate availability of Individual Patient Data (IPD) for both trials. In many situations, researchers do not have access to IPD of both studies but have access to the IPD of one study, say study A, and only to aggregate data from the other study, say study B, through a publication. The Matching-Adjusted Indirect Comparison (MAIC) method has been developed with this specific purpose: it is a recent methodology that allows to perform indirect comparisons between two drugs assessed in two different studies, where individual patients data are available in only one of the two studies, the data of the other one being available in an aggregate format only; it enables an indirect comparison of A versus B in adjusting average patient characteristics in trials with IPD (study A) to match those reported for study B.
The purpose of this work is to assess the properties of the MAIC method, and to compare, through simulations, various approaches and several ways of practical implementation of the method. These methods are assessed in a context where the study outcome is a time-to-event variable and the various comparisons are done using (potentially stratified) Cox models. It is also assumed that the two studies are similar as well-controlled phase III trials, with a reasonably large sample size.
It is concluded that it is more efficient to match the treatment arms separately (match the two drugs to compare on one hand, and the control arms on the other hand). Moreover, using the Lasso technique to select the covariates for the matching step is better than matching a maximal set of covariates.