Investigating the relationship between relative accuracy and linearity acceptance criteria in bioassay validation for clinical studies

Abstract

BACKGROUND: Relative accuracy (RA) and linearity are distinct validation characteristics in bioassay validation, yet quantitative relationships between their acceptance criteria are rarely defined. This study investigated how predefined RA tolerances determine admissible linearity acceptance ranges. RESEARCH DESIGN AND METHODS: A theoretical framework for assessing dilutional linearity was established using a b-fold serial dilution design. Relative accuracy was defined as the ratio of observed to expected geometric means, evaluated against a tolerance threshold (θ). Linearity was determined through linear regression of logb(GMobserved) against logb(Dilution). Analytical expressions were derived for acceptable slope intervals based on θ and the number of valid dilution levels. A simulated bioassay dataset illustrated its application. RESULTS: Analytical derivations showed that admissible slope intervals depend exclusively on θ and the number of valid dilution levels. The resulting limits were invariant with respect to both the initial geometric mean and the dilution fold factor. In the illustrative example, the proposed framework supported acceptance of linearity while retaining all valid dilution levels. CONCLUSIONS: This study provides a mathematically consistent method for deriving linearity acceptance criteria directly from predefined RA requirements. Although illustrated using simulated data, it offers a tool for dilution-based bioassay validation and may support more coherent validation decisions.