Diagnostic Performance Comparison of AFP, PIVKA-II, GALAD Model, and ASAP Model Across Two Chemiluminescence...

This study aimed to evaluate the diagnostic performance of individual serum biomarkers [alpha-fetoprotein (AFP), protein induced by vitamin K absence or antagonist-II (PIVKA-II)] and composite models (GALAD, ASAP) for hepatocellular carcinoma (HCC) across two immunoassay platforms.

From 2011 to 2021, 518 serum samples were selected from a liver-related disease biobank at Peking Union Medical College Hospital (Beijing, China), including 102 HCC patients, 117 with benign liver disease, 38 with cholangiocarcinoma, 96 with colorectal cancer, 65 with metastatic hepatic carcinoma, and 100 healthy controls. AFP and PIVKA-II levels were measured on both the Hotgen and Abbott ARCHITECT platforms. The GALAD and ASAP scores were calculated based on the data from each platform. Receiver operating characteristic (ROC) curve analysis and the corresponding areas under the curves (AUCs) were used to evaluate and compare the diagnostic value of the individual biomarkers and the two composite models.

For HCC diagnosis, AFP exhibited comparable efficacy between Hotgen (AUC: 0.821) and Abbott (AUC: 0.846), whereas PIVKA-II performed better on Abbott (AUC: 0.863) than Hotgen (AUC: 0.787). GALAD and ASAP models exhibited significantly better diagnostic performance than individual serum biomarkers on both platforms (P < 0.05): on Hotgen, both models achieved an AUC of 0.872, while on Abbott, ASAP (AUC: 0.913) was marginally superior to GALAD (AUC: 0.901, P = 0.0569). Notably, both models performed better on Abbott than Hotgen (GALAD: 0.901 vs 0.872, P = 0.0001; ASAP: 0.913 vs 0.872, P = 0.0003). Spearman correlation analysis showed moderate inter-platform correlations for AFP (r = 0.573) and PIVKA-II (r = 0.460). Bland-Altman analysis indicated poor inter-platform consistency, with mean biases of 44.32% (AFP) and -92.02% (PIVKA-II).

GALAD and ASAP models demonstrate superior diagnostic efficacy for HCC compared to individual biomarkers, and their performance is significantly influenced by the immunoassay platform employed.