Validation of longitudinal biomarker screening algorithms for HCC detection in patients with cirrhosis

BACKGROUND: Blood-based biomarker panels, including GALAD, have been proposed as an alternative to abdominal ultrasound for hepatocellular carcinoma (HCC) surveillance. Studies suggest longitudinal evaluation of biomarkers can improve test performance; however, no large studies have validated these findings. METHODS: We leveraged the HCC Early Detection Strategy (HEDS) study (n=1019 patients with cirrhosis; 99 incident HCC) and Texas HCC Consortium (THCCC) (n=2345 patients with cirrhosis; 126 incident HCC) to compare the performance of fixed-threshold GALAD (cutoff of -1.36), and 4 longitudinal algorithms: longitudinal GALAD, PALAD, mFB-ALAD, and mPEB-ALAD. The HEDS cohort was used to derive longitudinal algorithms, and external validation was performed in THCCC. Patient-level sensitivity and test-level false-positive rate (FPR) were examined overall and across subgroups by age, sex, and cirrhosis etiology. RESULTS: In THCCC, fixed-threshold GALAD had higher sensitivity in the 6 months before HCC diagnosis (71.4%, 95% CI: 62.9%-81.7%) than longitudinal GALAD (55.1%, 95% CI: 49.3-66.4), PALAD (55.1%, 95% CI: 50.0%-70.5%), mPEB-ALAD (53.1%, 95% CI: 43.6%-62.0%), and mFB-ALAD (42.9%, 95% CI: 32.3-48.5). However, fixed-threshold GALAD had a higher FPR (25.4%, 95% CI: 23.7%-26.9%) compared with longitudinal GALAD (15.0%, 95% CI: 12.8%-15.8%), PALAD (13.0%, 95% CI: 10.2%-14.6%), and mFB-ALAD (11.2%, 95% CI: 9.6%-12.1%) but comparable to mPEB-ALAD (22.9%, 95% CI: 21.2%-23.8%). In subgroup analyses, fixed-threshold GALAD had the highest FPR in males (35.6%) and those aged ≥65 (43.5%); longitudinal algorithms had significantly lower FPRs in subgroups but with lower sensitivity for HCC. CONCLUSIONS: Improvements in FPRs with longitudinal GALAD algorithms, as compared with fixed-threshold GALAD, are offset by significantly decreased sensitivity for HCC detection.