Development and validation of GADA model for diagnosing hepatocellular carcinoma in Chinese hepatitis B patients

BACKGROUND: Chronic hepatitis B virus (HBV) infection accounts for more than 40% of hepatocellular carcinoma (HCC) cases worldwide. The GALAD (Gender, Age, alpha fetoprotein heterogeneity [AFP-L3], AFP, and des-gamma-carboxy prothrombin [PIVKA-II/DCP]) model has potential for enhancing HCC diagnosis. However, its diagnostic performance in HBV-associated chronic liver disease requires further optimization. This study aimed to establish a diagnostic model for HBV-associated HCC. METHODS: The data from 2190 patients with HBV-related liver diseases, including chronic hepatitis B (CHB), cirrhosis, and HCC were analysed. The GADA model was developed by omitting AFP-L3 for improved accessibility, and its performance was assessed via the area under the receiver operating characteristic curve (AUC). Validation was performed on 20,908 patients across ten hospitals. RESULTS: The simplified GADA model demonstrated exceptional diagnostic performance in identifying HBV-related HCC in the modeling group, achieving an AUC of 0.940 (95% CI, 0.927-0.952), with a sensitivity and specificity of 82.57% and 91.53%, respectively, surpassing the original GALAD model's AUC of 0.907 (95% CI, 0.891-0.921). In the internal validation cohort, the GADA model had an AUC of 0.935 (95% CI, 0.914-0.953), sensitivity of 82.04%, and specificity of 92.65%. For early-stage HCC detection, the AUC in the modeling group was 0.858 (95% CI, 0.837-0.877), and in the internal validation group was 0.868 (95% CI, 0.836-0.895), significantly better than the original GALAD model. External validation with 20,908 participants yielded an AUC of 0.919 (95% CI, 0.916-0.923), with sensitivity of 86.00% and specificity 81.88%. For early-stage HCC diagnosis, the AUC was 0.903 (95% CI: 0.898-0.907), with sensitivity of 88.16% and specificity 74.50%. CONCLUSIONS: The GADA model shows exceptional promise for diagnosing HBV-related HCC, enhancing early detection and reducing misdiagnosis rates. Its clinical utility is supported by robust validation, indicating significant implications for patient prognosis and treatment options.