Abstract
INTRODUCTION: Hepatocellular carcinoma is the most common primary neoplasia of the liver. Microvascular invasion predicts outcome and defines tumor staging. However, its diagnosis is still a challenge. The present study aims to evaluate inter and intraobserver agreement in identifying the presence of microvascular invasion using conventional and immunohistochemistry histology. METHODS: Three pathologists performed the analysis of 76 hepatocellular carcinoma explants to characterize the presence of microvascular invasion using the hematoxylin/eosin stain and immunohistochemistry for CD34. The evaluations were made individually, in two distinct moments. Results were analyzed by the Kappa's coefficient and ROC curves. RESULTS: Our study demonstrated similar agreement for microvascular invasion between hematoxylin/eosin and CD34 methods. However, the intraobserver agreement values for both methods were higher than the interobserver ones. The accuracy of CD34 in relation to hematoxylin/eosin by ROC curves in intraobserver analysis tends to a high specificity, ranging from 82.1 to almost 100%, with sensitivity of 46.9% to 81.1%. In interobserver analysis, CD34 also has a high specificity (84.3% to 85.5%) while its sensitivity is a little shorter (81.2% to 84.3%). CONCLUSION: Intraobserver higher agreement allows us to suppose that pathologists employed own criteria to evaluate vascular invasion, reinforcing the need of standardization. ROC Curves analysis showed that the CD34 method is more specific than sensitive. Therefore, immunohistochemistry for CD34 should not be used routinely, but it could be useful to help confirming invasion previously seen by conventional histology.