The problem of multiple adjustments in the assessment of minimal clinically important differences

INTRODUCTION

Anthropometric, demographic, genetic, and clinical features may affect cognitive, behavioral, and functional decline, while clinical trials seldom consider minimal clinically important differences (MCIDs) in their analyses.

METHODS

MCIDs were reviewed taking into account features that may affect cognitive, behavioral, or functional decline in clinical trials of new disease-modifying therapies.

RESULTS

The higher the number of comparisons of different confounders in statistical analyses, the lower P values will be significant. Proper selection of confounders is crucial to accurately assess MCIDs without compromising statistical significance.

DISCUSSION

Statistical adjustment of the significance of MCIDs according to multiple comparisons is essential for the generalizability of research results. Wider inclusion of confounding variables in the statistics may help bring trial results closer to real-world conditions and improve the prediction of the efficacy of new disease-modifying therapies, though such factors must be carefully selected not to compromise the statistical significance of the analyses.

Highlights

• Anthropometric, demographic, and clinical features may affect cognitive, behavioral, and functional decline.
• Clinical trials seldom take minimal clinically important differences (MCIDs) or their confounders into account.
• Generalizability of research results requires the assessment of multiple confounding factors.
• The higher the number of comparisons involved, the lower P values will be considered significant.
• Use of MCIDs adjusted for confounding factors should be implemented when outcomes are not susceptible to translation into absolute benefits.