Mean Difference, Standardized Mean Difference (SMD), and Their Use in Meta-Analysis: As Simple as It Gets

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Evidence Synthesis

Background:

• Randomized controlled trials (RCTs) frequently analyze endpoint or change scores to compare experimental and control groups.
• Comparing mean differences (MD) across studies is feasible only when outcomes share identical measurement units.
• Discrepancies in measurement units across studies necessitate alternative statistical approaches for meta-analysis.

Purpose of the Study:

• To elucidate the concepts of mean difference (MD) and standardized mean difference (SMD) in the context of RCTs.
• To explain the conditions under which MD and SMD can be pooled in meta-analysis.
• To differentiate between weighted mean difference (WMD), Cohen's d, and Glass' delta.

Main Methods:

• The study defines MD as the direct comparison of mean outcome differences between groups when units are consistent.
• It introduces SMD as a method to standardize mean differences using standard deviations (SDs) when units vary across studies.
• Examples illustrate the calculation and interpretation of Cohen's d and Glass' delta as specific types of SMD.

Main Results:

• When measurement units are uniform, mean differences (MD) can be pooled directly, often termed weighted mean difference (WMD).
• When measurement units differ, mean differences are standardized by their SDs to calculate SMD, enabling meta-analysis.
• Standardized mean difference (SMD) values of 0.2, 0.5, and 0.8 signify small, medium, and large effect sizes, respectively.

Conclusions:

• Understanding MD and SMD is crucial for accurately synthesizing evidence from multiple RCTs.
• SMD provides a standardized metric for meta-analysis when studies employ different instruments to measure the same outcome.
• This explanation aids researchers in selecting appropriate statistical methods for evidence synthesis.