STATISTICAL METHODS FOR ASSESSING AGREEMENT BETWEEN TWO METHODS OF CLINICAL MEASUREMENT¶
Why this mattered¶
Bland and Altman’s 1986 paper mattered because it changed how clinical researchers judged whether two measurement methods could be used interchangeably. Before it, method-comparison studies commonly relied on correlation coefficients or regression, which can show strong association even when two instruments disagree systematically. Bland and Altman reframed the question: not “are the measurements related?” but “how far apart are they, and is that difference clinically acceptable?” Their proposed plot of pairwise differences against pairwise means, with bias and limits of agreement, made disagreement visible rather than hiding it behind a single association statistic.
The paradigm shift was practical as much as statistical. The paper gave clinicians, laboratory scientists, and device evaluators a simple diagnostic framework for deciding whether a new, cheaper, faster, or less invasive measurement method could replace an established one. It exposed fixed bias, proportional bias, changing variability across the measurement range, and outliers in a form that non-statisticians could interpret. That made rigorous agreement assessment newly routine in clinical measurement, where decisions often depend on whether two readings are close enough for patient care, not merely whether they move together.
Its influence carried into later work on diagnostic devices, laboratory assays, imaging measurements, observer reliability, and biomarker validation. The Bland-Altman plot became a standard part of reporting agreement studies, shaping guidelines and statistical practice well beyond medicine. Subsequent breakthroughs in portable monitoring, automated laboratory systems, digital imaging, and wearable sensors all depended on the same methodological distinction the paper clarified: a new measurement technology is useful only if its errors relative to the reference method are understood and acceptable for the intended clinical decision.
Abstract¶
(no abstract available)
Related¶
- enables → Bias in meta-analysis detected by a simple, graphical test — Bland-Altman plotting used differences against averages to reveal bias graphically, a template for Egger's funnel-plot test of meta-analysis bias.
- cite ← Bias in meta-analysis detected by a simple, graphical test — Egger’s meta-analysis bias test cites Bland-Altman because both use simple graphical methods to detect systematic statistical discrepancies.