Real-Time Detection of Analytical Outliers on the Roche Troponin T Generation 5 Assay

• Published in: American journal of clinical pathology Vol. 152; no. Supplement_1; p. S83
• Main Authors: Wockenfus, Amy, Hartung, Katherine, Kelley, Brandon, Katzman, Brooke, Donato, Leslie, Jaffe, Allan, Karon, Brad
• Format: Journal Article
• Language: English
• Published: US Oxford University Press 11.09.2019
• Subjects: Calcium-binding protein; Troponin; Troponin T
• ISSN: 0002-9173; 1943-7722
• DOI: 10.1093/ajcp/aqz116.007

Abstract Introduction Analytical outliers occur with most troponin assays and can adversely affect patient management. We developed a system to detect analytical outliers for the Roche Troponin T (cTnT) Gen 5 assay using repeat analysis of samples manifesting changing values over 2 to 6 hours. Methods In our ED practice, troponins are collected at baseline, 2 hours, and at times at 6 hours in plasma separator tubes and processed on the Roche Cobas e411 (Indianapolis, IN). When troponin values change by ≥10 ng/L or ≤10 ng/L in the 2- or 6-hour sample compared to baseline, the sample that produced the result is repeated. If initial and repeat values differ by > ±5 ng/L (cTnT <100 ng/L) or ±5% (cTnT ≥100 ng/L), the initial value is deemed to be an analytical outlier. Outliers are confirmed by a third measurement. Results Since March 2018, 19 of 17,154 (0.11%) troponin samples analyzed have been labeled outliers. Outliers were falsely elevated in 14 cases and falsely decreased in 5. Mean (range) of initial cTnT values for outliers was 39 ng/L (18-621 ng/L), with only one initial value >100 ng/L. All changes could affect analysis of change over time. Eleven of 19 outliers were outside gender-specific reference interval on both samples and thus did not cause a change from abnormal to normal. Repeat measurements on 8 of 19 outliers were within the URL, which may have greater clinical consequences. Conclusions Repeat analysis upon observation of changing values allows detection of analytical outliers in real time, prior to reporting results. These outliers are almost always clinically significant in terms of interpretation of changing values and at times change patients from above to below the URL. If we tested more than just changing samples, we might detect still more outliers.