Contribution of virtual biopsy to the screening of microvascular invasion in hepatocellular carcinoma: A pilot study.

Abstract

Microvascular invasion (mVI) is a major prognostic factor in hepatocellular carcinoma (HCC) that cannot be detected before surgery. Predictive biomarkers of mVI are thus urgently needed. We have developed an original approach of virtual biopsy to assess the performance of an immunohistochemical panel comprising three biomarkers of mVI (H4K16ac, H4K20me2, PIVKA-II) for the prediction of mVI in HCC core needle biopsies (CNB). A test set of HCC surgical specimens (n=64) and an independent validation set of HCC CNB (n=42) were retrospectively constituted. Immunostainings were first quantified in the test set on the whole tissue section, to determine optimal cut-off values for each marker. From the digitised image of the whole section, three virtual biopsies were provided. Immunostainings and accuracy of the panel for the prediction of mVI were further assessed in virtual biopsies and in the validation set of CNB. In virtual biopsies, PIVKA-II/H4K16ac had the best performance for prediction of mVI, with sensitivity, specificity, predictive positive value (PPV), and predictive negative value (PNV) of 30%, 97%, 91%, 56%, respectively. In CNB, PIVKA-II/H4K20me2 showed the best accuracy for prediction of mVI, with sensitivity, specificity, PPV, and NPV of 43%, 95%, 90%, and 62%, respectively. The two panels were independent predictive factors of mVI (PIVKA-II/H4K16ac, P=.037; PIVKA-II/H4K20me2, P=.026). This study shows that a panel of two markers is able to predict mVI in HCC CNB, and pave the way for the future development of prognostic biomarkers in HCC that could guide the therapeutic strategy.