Lost in dichotomization: Preserving statistical power and sensitivity in MS cognitive research

Abstract

Background:

Research on cognitive impairment in multiple sclerosis (MS) often reduces performance across diverse domains—such as processing speed, working memory, and executive functioning—into a single dichotomous variable (e.g. cognitively impaired vs cognitively intact). While this simplification may enhance clinical meaningfulness, it undermines efforts to develop sensitive and specific cognitive biomarkers.

Objective:

In this perspective, we argue that dichotomizing continuous cognitive measures, although useful for clinical classification, significantly reduces statistical power. Furthermore, collapsing across different cognitive domains obscures the multidimensional nature of cognitive dysfunction in MS and we risk overlooking important patterns of impairment and mischaracterizing the underlying neural mechanisms.

Conclusion:

We recommend assessing cognitive data as continuous variables and to consider domain-specific approaches. Such strategies enhance statistical sensitivity, without increasing false positives and/or false negatives, and provide a more nuanced understanding of cognitive impairment—critical steps toward meaningful biomarker discovery and targeted intervention.