Gene-expression phenotypes for vascular invasiveness of hepatocellular carcinomas

Abstract

Gross vascular invasion is a well-established prognostic indicator in hepatocellular carcinoma (HCC), but the biological significance of microscopic invasion remains unclear. Curatively resected primary HCCs were classified retrospectively into 3 groups: HCCs without vascular invasion (V0), HCCs with microvascular invasion (V1), and HCCs with macrovascular invasion (V2). Microarray profiling of patients with V0, V1, and V2 using Jonckheere-Terpstra (JT) tests and Wilcoxon rank sum tests was performed. Distinct patterns of gene expression were demonstrated between V0 and V2 groups; less (L) and highly (H) invasive phenotypes, respectively. It is noteworthy that 2 dendrograms by the hierarchical clustering provided exactly the same assignment results for V1 cases that were thus separated into L and H gene-expression phenotypes. Marked differences were found in overall (P < .001) and tumor-free survival (P < .001) between L and H gene-expression phenotypes. Multivariate analyses indicated that the phenotypes of the patterns of gene expression, rather than the clinicopathologic markers of vascular invasion, were independent predictors of tumor recurrence (P = .031). Using the gene-expression patterns identified by both JT and Wilcoxon rank sum test analyses, other V1 cases validated these differences in tumor-free survivals between gene-expression phenotypes within the group (P = .039). Gene profiling suggested that microvascular invasiveness consisted of a classable mixture of 2 distinct phenotypes. Thus, gene-array analyses may have clinical benefit, because they may in fact be more predictive than other clinical factors.