The influence of haemolysis in patient samples on biochemical tests analysed using Roche Cobas ® 8000 analyzer

Abstract

Background

Modern analyzers employ the haemolysis index (HI) to identify interference in biochemical assays, yet manufacturer-defined HI thresholds may be inappropriate for true haemolysis effects, resulting in unnecessary sample rejections. This study aimed to validate these thresholds using non-simulated hemolyzed patient samples.

Methods

Paired samples (hemolyzed primary and non-hemolyzed recollected) from 678 patients were analysed for haemolysis interference. Biochemical analytes and serum indices were measured using a Roche Cobas^® 8000 analyzer. Haemolysis effects on test results and lipaemia index (LI) were assessed. HI thresholds were derived from reference change value (RCV) limits and regression of HI versus percentage bias, then compared to the conventional 10% deviation criterion and Roche-defined cut-offs.

Results

Samples exhibited predominantly moderate haemolysis (72.3%, HI: 101–300). Strong HI correlations were observed for lactate dehydrogenase (51% change per 100-unit HI, R ² = 0.6524, P < .0001), potassium (14% per 100-unit HI, R ² = 0.5630, P < .0001), and sodium (−0.6% per 100-unit HI, R ² = 0.5414, P < .0001). Elevated biases exceeded the RCV for these analytes, plus ammonia, aspartate aminotransferase, creatine kinase, γ−glutamyltransferase, and bilirubin-direct, whereas sodium showed a clinically significant reduction at heavy haemolysis (HI 560). RCV-derived thresholds exhibited comparable or higher than 10% change and Roche cut-offs. The elevated LI in hemolyzed samples with HI greater than 100 decreased significantly after recollection.

Conclusions

Patient-based haemolysis data indicated that biases for most analytes remain within clinically acceptable limits, suggesting the manufacturer’s HI thresholds may overestimate interference, supporting lab-validated, RCV-based cut-offs enhance clinical relevance and decrease unnecessary sample rejection.