Interpretation

Higher GALAD (gender, age, AFP-L3, AFP, des-gamma-carboxy prothrombin) model scores correlate with increased risk of hepatocellular carcinoma (HCC). The area under the curve (AUC) of a receiver operating characteristic (ROC) curve of the GALAD score was 0.95 for all HCC detection and 0.92 for the detection of early-stage HCC. Additionally, the AUC of the GALAD score (0.95) was higher than that of ultrasound alone for all HCC detection (AUC of 0.82, P <0.01).

The sensitivity and specificity performance characteristics of the GALAD score for HCC will be influenced by the selected GALAD score cut-off. For example, at an optimal AUC cutoff of -0.76, the GALAD score had 91% sensitivity and 85% specificity for HCC detection. At a more specific GALAD score cutoff of 0.88, the observed sensitivity was 80% for HCC detection, with an observed specificity of 97%.

The GALAD model was developed and validated in patient cohorts with a prevalence of HCC ranging from 35% to 49%. The performance of the model may be altered in populations with different HCC prevalence. In addition, the clinical performance of the GALAD score varies by etiology of HCC and, therefore, may be different in different regions of the world.