eBooks View all | By AtoZ | By category
Statistical methods for comparing limits of detection in molecular diagnostics a practical guide
By Jesse A Canchola
Roche Molecular Systems, Inc., 4300 Hacienda Drive
Published: Feb 27, 2026 | pg. no: 1-16
Abstract: The limit of detection (LoD) is a critical performance characteristic in molecular diagnostics, defining the lowest analyte concentration at which an assay achieves reliable detection.1,2 While numerous approaches exist for estimating LoD in both quantitative and qualitative assays, confusion often arises regarding appropriate methods for statistical comparison between a reference formulation and a new test condition.3,4 This paper provides a comprehensive overview of theoretical frameworks, statistical models, and practical considerations for LoD comparison, with additional attention to multi-target assay scenarios common in contemporary molecular diagnostics. We contrast quantitative assays, where LoD is a numeric estimate derived from continuous data, with qualitative assays, where detection probabilities are assessed across dilution levels. Worked examples illustrate correct applications of the delta method,5 Fieller's theorem,6 bootstrap,7 two one-sided test (TOST) procedures8 and non-inferiority procedures when Just Enough Testing (JET) is the objective. The discussion emphasizes common pitfalls, particularly the inappropriate reliance on overlapping confidence intervals as evidence of equivalence,9,10 and offers regulatory-compliant recommendations for both single-target and multi-target comparison studies.11,12 Guidance on sample size determination, multiplicity considerations, and interpretation frameworks ensures that practitioners can design and execute statistically sound LoD comparison studies that meet regulatory expectations or, when using the JET method, for delivery to assay development for optimization.
View eBook