What is it about?

Two explanations for apparent observed treatment by trial interactions in meta-analysis are discussed, neither of which depend on effect modification of the treatment by the 'trial'. The first is that the treatments may differ from trial to trial, the second is that variances may be being deflated because weighting by observed precision underestimates variances of estimated effects.

Featured Image

Why is it important?

This paper alerts researchers to explanations for observed heterogeneity that do not depend on the effects of treatments varying due to variations in the populations treated,


My work in the pharmaceutical industry (1987-1995) involved developing various formulations of a treatment. There were surprising differences between the formulation. On the other hand many meta-analysists seem to be prepared to pool different molecules when investigating the effects of treatment. This was the inspiration for the first explanation. The second explanation involved a more technical mathematical investigation.

Professor Stephen J Senn
Consultant Statistician

Read the Original

This page is a summary of: A note regarding alternative explanations for heterogeneity in meta‐analysis, Statistics in Medicine, September 2022, Wiley, DOI: 10.1002/sim.9403.
You can read the full text:



The following have contributed to this page