Analytical similarity is foundational to biosimilar development, yet modern comparability programs generate heterogeneous evidence that cannot be represented adequately by one composite score. This article proposes an original hierarchical evidence model that preserves attribute identity across structural, physicochemical, functional, clinical-consequence, residual-uncertainty, and decision-response layers. The model rejects context-free compensation because extensive agreement among low-criticality attributes could conceal an unresolved difference in a clinically important characteristic. Critical quality attributes are therefore interpreted according to product-specific mechanism, assay confidence, reference-product variability, and plausible consequence for pharmacokinetics, pharmacodynamics, immunogenicity, safety, or efficacy. Residual uncertainty is defined as the decision-relevant question that remains after available evidence has been integrated, not as the number of studies still unperformed. Additional evidence is selected according to that question: targeted analytical reassessment, mechanism-relevant functional testing, sensitive pharmacokinetic or pharmacodynamic evaluation, focused immunogenicity investigation, or comparative clinical evidence when clinically meaningful uncertainty cannot be resolved more directly. Retrospective validation would require blinded reconstruction of comparability cases, explicit treatment of missing information, inter-rater assessment, falsification cases, lifecycle lot evaluation, and calibration against independent outcomes, followed by prospective testing. The contribution is a proposed non-empirical architecture rather than a validated scoring system, regulatory standard, clinical recommendation, or deployment-ready tool. Its principal boundaries arise from incomplete mechanism knowledge, assay limitations, reference-product drift, product-class differences, and jurisdiction-specific interchangeability decisions.
IJPRAS
