The pitfalls of CA19-9: routine testing and comparison of two automated immunoassays in a reference oncology center

We evaluated CA19-9 as a marker of various malignancies and compared the results of 2 commercial immunoassays. The Abbott ARCHITECT i2000 and Roche cobas 410 immunoassays were used on 500 consecutive samples to evaluate the frequency of positive results by cancer type and the correlation between assays. The patients were tested before or after surgery and/or during chemotherapy. The rate of results exceeding conventional thresholds was 92.3% in pancreatic cancer, 36.8% in gastric cancer, and ranged from 3.0% to 35.9% in other tumors. Agreement (90.6%) and correlation (R(2) = 0.865) between the 2 assays were good and the frequency of highly discordant results was low (6/500). In some cases, interference by heterophilic antibodies was demonstrated. The 2 methods were comparable in diagnostic accuracy and had good correlation but are not interchangeable. Patients should always be monitored for CA19-9 with the same method and it should be indicated in the report.