42 Title: Large-Scale Retrospective Evaluation of Hemolysis, Icterus, and Lipemia Index Reproducibility on Roche cobas Analyzers

Analytical interference caused by hemolysis, icterus, and lipemia (HIL) is recognized as a significant source of error in clinical laboratory testing. Estimates of HIL severity (i.e., HIL indices) are measured spectrophotometrically on automated chemistry analyzers and utilized to guide result reporting. HIL indices are not reported to the patient’s chart and are not subject to the same quality control and external quality assessment requirements as reportable assays. Therefore, the analytical performance of HIL indices is less well understood. The objective of this study was to evaluate the reproducibility of HIL index measurements across different Roche cobas analyzers and sample types. HIL index results generated from two Roche cobas c702 analyzers and two Roche cobas c502 analyzers over a 24-month period were retrieved from the middleware information system. Reproducibility of HIL indices was evaluated as a function of HIL severity, analyzer type, and sample type. Over the 24-month period, 928,484 specimens had at least one set of HIL indices measured. Of these, 63,523 (6.8%) had two or more sets of HIL index measurements completed within 60 minutes (median: 13 minutes) and were included in the retrospective evaluation. Nearly all specimens were lithium heparin plasma (98.8%), followed by EDTA (0.9%), and serum (0.4%). The initial HIL measurements were more common on the two c702 analyzers (64%) than the two c502 analyzers (36%). The two sets of HIL indices were measured on the same instrument model for 63% of samples and on the same exact instrument for 51% of samples. Hemolysis index demonstrated strong agreement (slope = 0.996) and correlation (R² = 0.995) across samples with varying hemolysis indices (median: 13, range: 0 – 4,688); 72% of measurements differed by ≤15%. Icteric index demonstrated similarly strong agreement (slope = 0.997) and correlation (R² = 0.993) across samples with varying icteric indices (median: 1, range: 0 – 87); 94% of measurements differed by ≤15%. Lipemia index demonstrated the weakest agreement (slope = 0.970) and correlation (R² = 0.927) across samples with varying lipemia indices (median: 14, range: 0 – 1,741); 45% of measurements differed by ≤15%. The hemolysis index remained the most consistent across different analyzer models, whereas minor bias between analyzer models was observed for the icteric index and, to a lesser extent, the lipemia index. Non-lithium heparin sample types exhibited increased variability for all HIL indices compared to lithium heparin samples. Overall, our findings demonstrate good reproducibility of HIL indices within and between different Roche cobas analyzers and across common sample types. The most variability is introduced when HIL index measurements are performed on different instrumentation models or on non-lithium heparin sample types. This study reinforces the reliability of automated HIL assessments and supports their utility in routine clinical chemistry testing.