Answers are grounded in this paper and cite their sources.
Background Biomarkers increasingly gate Alzheimer disease trials and clinical access, raising risk of design-induced bias when biomarker data are reused for both selection and inference. Objective To quantify circularity patterns in Alzheimer disease biomarker trials and trial-ready cohorts and, using mechanistic experiments, estimate effects on discrimination, calibration/transportability, and false-positive risk. Methods We audited PubMed-indexed biomarker interventional trials and trial-ready cohorts (2020-2024) using dual independent full-text review and mechanism classification. Two simulation experiments examined (1) enrichment-induced range restriction with transport to an unselected population and (2) selection-inference reuse (“double dipping”) versus split-sample confirmation. Results Among 72 studies, 10 (13.9%) raised possible/definite non-independence; enrichment/range restriction predominated (7/10). In simulation experiment 1, enrichment preserved nominal type I error under the null but shifted the estimand under signal, attenuating discrimination and degrading calibration when enriched-fit models were applied to unselected populations. In simulation experiment 2, double dipping produced P ( p < 0.05) = 0.994 under the null ( n = 200, p = 100) versus 0.047 with split-sample confirmation, with winner’s-curse effect inflation. Conclusions Circularity was uncommon but nontrivial and most often reflected enrichment with implicit generalization beyond the enriched population. Safeguards include explicit target populations/estimands and separation of discovery from confirmation (or prespecified multiplicity control), with clear validation and leakage prevention when biomarker claims are intended to guide decisions.
No takes yet. Share an insight, caveat, or question.
Verma et al. (2025) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: