Comparisons of the mean differences and standardized mean differences for continuous outcome measures on the same...

Area of Science:

• Biostatistics
• Evidence Synthesis
• Medical Research Methodology

Background:

• Systematic reviews and meta-analyses commonly use mean differences (MDs) and standardized mean differences (SMDs) for continuous outcomes.
• SMDs are typically preferred when studies use different measurement scales, standardizing data for synthesis.
• Concerns exist regarding SMD interpretability and potential for increased heterogeneity.

Purpose of the Study:

• To compare the performance of MDs and SMDs using simulation studies.
• To evaluate their utility in scenarios where continuous measures are on the same scale.
• To assess robustness to violations of the normality assumption.

Main Methods:

• Conducted simulation studies comparing MD and SMD.
• Varied settings including cases where normality assumption does not hold.
• Evaluated bias, mean squared errors, and confidence interval coverage probability.

Main Results:

• In scenarios with comparable scales, SMDs showed poorer performance than MDs.
• SMDs exhibited greater biases, higher mean squared errors, and lower CI coverage.
• MDs demonstrated greater robustness to violations of the normality assumption.

Conclusions:

• SMDs are valuable for synthesizing continuous data on different scales.
• When scales are comparable, MDs are generally preferred due to better statistical properties.
• MDs offer a more robust and interpretable effect measure for similar continuous data.