Abstract
HbA1c is interpreted using fixed clinical decision limits, but measurement uncertainty (MU) may affect analytical interpretation of results near these thresholds. This study estimated ISO/TS 20914-based MU for two HbA1c analytical systems and evaluated the patient-level impact of including calibration- and traceability-related uncertainty. In this retrospective two-center study, 30,791 routine HbA1c results from independent patient cohorts were obtained using Roche cobas c501 and Adams HA-8180 V systems. MU was estimated from within-laboratory reproducibility uncertainty (uRw), calibration- and traceability-related standard uncertainty (ucal), and uncertainty associated with the IFCC-to-NGSP master equation (uME). Decision-limit-specific coverage intervals were calculated at 5.7% and 6.5% HbA1c using a calibration-included MU model (uRw, ucal, and uME) and a calibration-excluded sensitivity analysis (uRw and uME). Patient-level reclassification was evaluated using McNemar's test with Holm adjustment. Roche showed higher uRw than Adams, whereas ucal was higher for Roche (2.35% vs 1.75%). Calibration-included standard MU was 3.77% and 3.76% for Roche at 5.7% and 6.5%, respectively, and 2.47% and 2.45% for Adams. Inclusion of ucal increased the proportion of results within MU coverage intervals from 38.81% to 48.05% for Roche at 5.7%, from 15.45% to 20.76% at 6.5%, and from 10.69% to 15.90% for Adams at 6.5% (all Holm-adjusted P < 0.001). No reclassification occurred for Adams at 5.7%. Inclusion of calibration- and traceability-related uncertainty materially affected analytical interpretation around clinical decision limits in three of four system-threshold analyses. Patient-level impact depended on both MU budget composition and the distribution of results around the relevant threshold.