This tutorial focuses on standardised mean differences (SMD) as effect
masures in meta-analyses. We will explain what they are, when they
should be used, how to correctly compute and interpret them, and some of
the most common error made within evidence synthesis. Additionally, we
have created a micro-learning module to accompany this article which
includes a real-life example of meta-analysis of SMDs. Moreover, there
is the opportunity to test out how to use a tool to correctly calculate
SMDs, and how to interpret them.
Standardised
mean differences micro-learning module.
INTRODUCTION
In this tutorial, we focus on standardised meandifferences (SMD); what they are, when they should be used, how
to correctly compute and interpret them, and some of the common errors
made within systematic reviews.
Review authors use the SMD as a summary statistic in meta-analyses of
continuous outcomes when the studies all assess the same outcome but
measure it in a variety of ways [1]. For example, we can look at a
meta-analysis in which included studies have measured the depressive
symptoms of their participants using different scales/questionnaires
[2] (e.g., the Beck Depression Inventory, the Geriatric Depression
Scale, the Hamilton Rating Scale, or the Montgomery-Asberg Depression
Scale). From a statistical perspective, it is virtually impossible to
quantitatively synthesise the results directly. Instead, a solution
would be standardising all these data to a common effect size measure:
the SMD (Figure 1).