A Laboratory-Based Triage Model for Hepatocellular Carcinoma Early-Detection Pathways in Hospital-Evaluated Populations: Development and External Validation
Related Experiment Video
Updated: Nov 9, 2026
A Three-Dimensional Spheroid Model to Investigate the Tumor-Stromal Interaction in Hepatocellular Carcinoma
Published on: September 30, 2021
Liqiao Zhang1, Xiaozhen Xu2, Xingfen Zhang2
1Department of Clinical Laboratory, The Affiliated Lihuili Hospital of Ningbo University, Ningbo, Zhejiang, 315040, People's Republic of China.
Background:
Hepatocellular carcinoma (HCC) remains a major clinical and health-system burden in East and South-East Asia, and early-detection pathways remain vulnerable when alpha-fetoprotein (AFP) is low or non-informative. We developed and externally validated Extended-7 as a routine laboratory-based triage aid to prioritize diagnostic attention within hospital HCC early-detection and referral pathways.
Methods:
This two-center retrospective study evaluated a hospital-based laboratory triage model in a Ningbo development cohort (n=523; 267 HCC and 256 controls) and an independent Lihuili external-validation cohort (n=269; 150 HCC and 119 controls). Using these data, we developed and validated an HCC risk-stratification model intended to support prioritization for further diagnostic evaluation, not stand-alone diagnosis or replacement of imaging/pathology. Discrimination, calibration, decision-curve profiles, cohort-specific Youden operating characteristics, and exploratory AFP-low subgroups were evaluated against AFP, AFP-L3, PIVKA-II, a three-marker model, GALAD, and ASAP.
Results:
Extended-7 combined ln(AFP+1), AFP-L3, ln(PIVKA-II+1), age, sex, platelet count, and albumin. Extended-7 achieved a 5-fold cross-validated AUC of 0.900 (95% CI, 0.871-0.924) in the development cohort and an external AUC of 0.738 (95% CI, 0.681-0.793). In external validation, discrimination was higher than GALAD (AUC 0.694; DeLong p=0.006), ASAP (AUC 0.643; p<0.001), and the three-marker model (AUC 0.602; p<0.001). At the external Youden cutoff, Extended-7 identified 119 of 150 HCC cases while classifying 47 of 119 non-HCC controls as false-positive triage referrals. In the external AFP-low subgroup, Extended-7 achieved an AUC of 0.808; the paired bootstrap comparison vs GALAD did not reach conventional statistical significance. External calibration was weak (Brier score 0.224; calibration slope 0.262).
Conclusions:
Extended-7 showed moderate external discrimination and may support laboratory-based prioritization in hospital HCC early-detection pathways, including patients with low AFP, while subgroup findings remain exploratory. However, transported probabilities require local recalibration, and prospective multicenter evaluation of referral thresholds, decision impact, and cost-effectiveness is needed before clinical implementation.