A-055 The Current Discrepancies Among Commercial Insulin Assays

Abstract Background Accurate insulin testing is crucial for diagnosing and managing conditions affecting insulin regulation, including diabetes, metabolic syndrome, insulinoma, hypoglycemia, and polycystic ovary syndrome (PCOS). However, discrepancies across insulin assays limit their broader clinical utility. This study evaluates variations among widely used insulin assays to support future standardization efforts. Methods Serum samples covering a broad range of insulin concentrations (n = 40) were distributed to nine major manufacturers, representing 12 distinct methods. For eight methods the samples were analyzed in duplicate while single values were submitted for the other methods . Each manufacturer followed its standard calibration protocols. Assays were deidentified per manufacturer*s’ recommendations. The same samples were analyzed using liquid chromatography-mass spectrometry (LC-MS) at the University of Washington. LC-MS analysis involved acetonitrile precipitation, solid-phase extraction, Glu-C digestion, and quantitation with a surrogate labeled peptide (RGFFYTPKT) via LC-MS/MS. Due to skewed residuals, a non-parametric intraclass correlation coefficient (ICC) was used to assess the reliability of replicate results for assays run in duplicate. Intra-assay variability was assessed via absolute differences (AbsDiff) and relative percent absolute differences (PctAbsDiff). Passing-Bablok regression was used to evaluate each assay*s agreement with LC-MS; results were considered partially equivalent to those of LC-MS if the confidence intervals for the slope or intercept included 1 or zero respectively, or fully equivalent if both conditions were satisfied Results ICC values ranged from 0.992 to 0.999, indicating high reliability. The largest AbsDiff (107.2 units) occurred in assay J, corresponding to an 8.6% relative difference. The smallest AbsDiff was 0.00 across four assays. Maximum PctAbsDiffs ranged from 3.0% (assay F) to 10.8% (assay D). Passing-Bablok regression revealed full equivalence for only one method (H) and partial equivalence by intercept for four methods (A, B, D, E). Graphical analysis showed that most assays aligned with LC-MS at lower values but underestimated insulin at higher concentrations. Despite calibration against the same WHO 66/304 standard, significant discrepancies persisted across methods. Conclusion These findings highlight the urgent need for a standardized calibration scheme incorporating commutable certified reference materials (CRMs) to improve agreement among insulin assays.