Clinical significance of inter-assay discrepancies and regression-derived correction of serum CA19-9 values: a comparison between CLIA and ECLIA

Serum carbohydrate antigen 19-9 (CA19-9) is widely used in the management of pancreatic diseases, but measurement results often vary across immunoassay platforms. This study sought to describe how the values obtained by chemiluminescent immunoassay (CLIA) and electrochemiluminescent immunoassay (ECLIA) differ in practice, to assess whether discrepancies vary across concentration ranges, and to identify clinical factors that might contribute to this variability. METHODS: We examined 1,760 paired serum samples measured simultaneously by CLIA (Abbott Architect i2000SR) and ECLIA (Roche Cobas e801) during a laboratory transition period. To assess agreement between assays, correlation and regression analyses were performed on log₁₀-transformed data. Subgroup comparisons were made by CA19-9 concentration, disease category, and hepatobiliary enzyme status. Multivariable regression was used to explore independent factors associated with the magnitude of inter-assay differences. RESULTS: Overall correlation between CLIA and ECLIA was strong (r = 0.94, ρ = 0.89), although CLIA tended to yield slightly higher values. Discrepancies were greater in pancreatic cancer samples, in those with elevated ALP or γ-GT, and at higher concentrations (CLIA ≥ 1000 U/mL). Around 8 % of samples were classified differently near the clinical cutoff (37 U/mL). Older age, male sex, pancreatic cancer, and elevated hepatobiliary enzymes were independently associated with larger assay differences (adjusted R² = 0.13). CONCLUSIONS: Despite good overall agreement, the two methods showed variability that could affect clinical interpretation, especially in low or high concentration ranges. When assay changes are unavoidable, using method-specific reference ranges or confirming results near decision thresholds may help ensure consistent interpretation.