Improving tumor marker comparability: A case series on harmonizing carbohydrate antigen 19‐9 and carbohydrate antigen 15‐3 measurements across diverse measurement systems

Abstract

Serum tumor markers play a critical role in the early screening, treatment monitoring, and prognostic assessment of malignant tumors, yet poor comparability across different measurement systems frequently undermines their clinical value. We evaluated the comparability of carbohydrate antigen 19‐9 (CA19‐9) and carbohydrate antigen 15‐3 (CA15‐3) across seven contemporary measurement systems—Abbott Alinity I, Autobio A2000PLUS, Beckman Access Dxl 800, Mindray CL6000i, Roche Cobas E801, Snibe Maglumi X8, and Sysmex HISCL‐5000 using 464 native serum samples for CA19‐9 and 499 for CA15‐3. Spearman correlations were higher for CA15‑3 ( r : 0.927‒0.997) than for CA19‑9 ( r : 0.777‒0.985). With the all‐method median as a pragmatic comparator, Passing‒Bablok regression and difference plots were generated for each system. Against analytical performance specifications derived from biological variation, all systems except Abbott Alinity (22.5%) met the minimum bias criterion of ±21.6% for CA19‐9, whereas for CA15‐3 only Abbott Alinity (‒10.159%), Roche Cobas (13.01%), and Snibe Maglumi (−3.923%) fell within the minimum bias criterion of ±13.9%. CA19‐9 and CA15‐3 concentrations were recalibrated using three mathematical approaches: principal component analysis and two nonlinear regression models. After recalibration, the variability of CA19‐9 and CA15‐3 values decreased significantly. Our findings indicate that CA19‐9 and CA15‐3 results are not directly interchangeable across different systems, which poses a real‐world challenge for longitudinal patient follow‐up. However, mathematical models derived from individual patient serum samples can effectively reduce system‐related discrepancies, thereby enabling more consistent serial interpretations and supporting more informed clinical decisions in routine practice.