Tumor-derived proliferative CTCs and CTC clusters predict aggressiveness and early recurrence in hepatocellular carcinoma patients.

Abstract

Circulating tumor cells (CTCs), an indispensable liquid biopsy classifier, can provide extra information for the diagnosis and prognosis of hepatocellular carcinoma (HCC). We systematically analyzed the peripheral blood of preoperative HCC patients (n = 270) for CTC number, Ki67 index reflecting the proliferative CTC percentage (PCP), and CTC clusters correlated with the characteristics of malignant HCC tumors. Driver gene mutations were found with high levels of consistency between CTCs and primary tumors (n = 73). CTC number and PCP were associated with tumor size, microvascular invasion (MVI), presence or absence of multiple tumors, and thrombosis significantly. CTC number and PCP robustly separated patients with and without relapse, with a sensitivity of 88.89%/81.48% and a specificity of 72.73%/94.81% in the training set (n = 104) and corresponding values of 80.00%/86.67% and 78.38%/89.19% in the validation set (n = 52), showing a better performance than that associated with the alpha-fetoprotein (AFP) level. CTC number, PCP, CTC clusters, and MVI were independent significant risk factors for HCC recurrence (P = 0.0375, P < 0.0001, P = 0.0017, and P = 0.0157). A nomogram model based on these risk factors showed a considerable prediction ability for HCC recurrence (area under the curve = 0.947). PCP (training: log-rank P < 0.0001; hazard ratio [HR] 30.13, 95% confidence interval [CI] = 11.12-81.62; validation: log-rank P < 0.0001; HR 25.73, 95% CI = 5.28-106.60) and CTC clusters (training: log-rank P < 0.0001; HR 17.34, 95% CI = 7.46-40.30; validation: log-rank P < 0.0001; HR 9.92, 95% CI = 2.55-38.58) were more significantly correlated with worse recurrence-free survival than CTC number (training: log-rank P < 0.0001; HR 14.93, 95% CI = 4.48-49.78; validation: log-rank P = 0.0007; HR 9.03, 95% CI = 2.53-32.24). PCP and CTC clusters may predict HCC recurrence and improve the performance of the serum biomarker AFP.