Setting clinical performance specifications to develop and evaluate biomarkers for clinical use. (September 2019)

• Record Type: Journal Article
• Authors: Lord, Sarah J; St John, Andrew; Bossuyt, Patrick MM; Sandberg, Sverre; Monaghan, Phillip J; O'Kane, Maurice; Cobbaert, Christa M; Röddiger, Ralf; Lennartz, Lieselotte; Gelfi, Cecilia; Horvath, Andrea R
• Is Part Of: Annals of clinical biochemistry. Volume 56:Number 5(2019:Sep.)
• Page Start: 527
• Page End: 535
• Publication Date: 2019-09
• Subjects: Test evaluation -- biomarker -- medical tests -- clinical performance -- clinical accuracy -- research methods; Clinical chemistry -- Periodicals; Clinical biochemistry -- Periodicals; 616.075
• DOI: 10.1177/0004563219842265
• Languages: English
• ISSNs: 0004-5632

Abstract

Background: Biomarker discovery studies often claim 'promising' findings, motivating further studies and marketing as medical tests. Unfortunately, the patient benefits promised are often inadequately explained to guide further evaluation, and few biomarkers have translated to improved patient care. We present a practical guide for setting minimum clinical performance specifications to strengthen clinical performance study design and interpretation. Methods: We developed a step-by-step approach using test evaluation and decision-analytic frameworks and present with illustrative examples. Results: We define clinical performance specifications as a set of criteria that quantify the clinical performance a new test must attain to allow better health outcomes than current practice. We classify the proposed patient benefits of a new test into three broad groups and describe how to set minimum clinical performance at the level where the potential harm of false-positive and false-negative results does not outweigh the benefits. (1) For add-on tests proposed to improve disease outcomes by improving detection, define an acceptable trade-off for false-positive versus true-positive results; (2) for triage tests proposed to reduce unnecessary tests and treatment by ruling out disease, define an acceptable risk of false-negatives as a safety threshold; (3) for replacement tests proposed to provide other benefits, or reduce costs, without compromising accuracy, use existing tests to benchmark minimum accuracy levels. Conclusions: Researchers can follow these guidelines to focus their study objectives and to define statistical hypotheses and sample size requirements. This way, clinical performance studies will allow conclusions about whether test performance is sufficient for intended use.