Avoiding causal fraud in the evaluation of clinical benefits of treatments for Alzheimer's disease

Recent regulatory approvals of three amyloid-lowering monoclonal antibody therapies for the treatment of Alzheimer's disease (AD) have triggered a polarizing debate in the field on the clinical meaningfulness of their reported effects. The question of how to define clinical meaningfulness for any treatment that has a modest effect size is important and will likely be subject to influence from interested stakeholders. We warn of claims of evaluating meaningful within-individual change from randomized parallel-group trials of AD treatments, sometimes purportedly assessed by a commonly recognized "responder" analysis approach, and explain why it is likely to mislead and should simply be avoided. The average between-group difference in score change is where the debate and research efforts should be focused to contextualize and evaluate the clinical meaningfulness of the true treatment effect. The statistical and communication principles we consider and would recommend are applicable to the evaluation of most interventions in medicine. HIGHLIGHTS: Dichotomized outcome analysis approaches purporting to evaluate within-individual meaningful change are highly likely to mislead. In our view, the most valid statistical approach to understanding the true treatment effect is to analyze the average between-group difference in outcome scores. The average between-group difference in score change is where the debate and research efforts should be focused to contextualize and evaluate the clinical meaningfulness of the true treatment effect.