Objective The biological heterogeneity of hepatocellular carcinoma (HCC) makes prognosis difficult. We translate the results of a genome-wide high-throughput analysis into a tool that accurately predicts at presentation tumour growth and survival of patients with HCC.

Design Ultrasound surveillance identified HCC in 78 (training set) and 54 (validation set) consecutive patients with cirrhosis. Patients underwent two CT scans 6 weeks apart (no treatment in-between) to determine tumour volumes (V₀ and V₁) and calculate HCC doubling time. Baseline-paired HCC and surrounding tissue biopsies for microarray study (Agilent Whole Human Genome Oligo Microarrays) were also obtained. Predictors of survival were assessed by multivariate Cox model.

Results Calculated tumour doubling times ranged from 30 to 621 days (mean, 107±91 days; median, 83 days) and were divided into quartiles: ≤53 days (n=19), 54–82 days (n=20), 83–110 days (n=20) and ≥111 days (n=19). Median survival according to doubling time was significantly lower for the first quartile versus the others (11 vs 41 months, 42, and 47 months, respectively) (p<0.0001). A five-gene transcriptomic hepatic signature including angiopoietin-2 (ANGPT2), delta-like ligand 4 (DLL4), neuropilin (NRP)/tolloid (TLL)-like 2 (NETO2), endothelial cell-specific molecule-1 (ESM1), and nuclear receptor subfamily 4, group A, member 1 (NR4A1) was found to accurately identify rapidly growing HCCs of the first quartile (ROC AUC: 0.961; 95% CI 0.919 to 1.000; p<0.0001) and to be an independent factor for mortality (HR: 3.987; 95% CI 1.941 to 8.193, p<0.0001).

Conclusions The hepatic five-gene signature was able to predict HCC growth in individual patient and the consequent risk of death. This implies a role of this molecular tool in the future therapeutic management of patients with HCC.

Trial registration number ClinicalTrials.gov Identifier: NCT01657695.