Background and Aim: Hepatocellular carcinoma (HCC) is unique in that the presence of vascular invasion significantly changes tumor stage. Even though searching for vascular invasion is a common practice in surgical pathology, there appears to be a great variation among pathologists in its recognition. This study was designed to assess whether HCC could be accurately staged using vascular invasion as a staging parameter. Methods: The interobserver and intraobserver agreement for vascular invasion was analyzed in 126 liver resections for HCC. Selected slides were circulated twice among six pathologists for independent review using their own criteria. One to three representative images from 26 equivocal cases selected by one of the authors were re-evaluated by the pathologists. The presence or absence of vascular invasion on each slide or image was recorded as yes or no. The results were analyzed using unweighted kappa statistic analysis for multiple raters. Results: The interobserver agreement was moderate on two slide circulations with kappa values of 0.50 (95% confidence interval 0.45-0.55) and 0.43 (0.38-0.47), respectively. The kappa value dropped significantly to 0.19 (0.09-0.29) on selected images photographed from controversial cases. The intraobserver agreement was moderate, with kappa values ranging from 0.23 to 0.56 (mean = 0.45). Conclusions: Pathologists can reproducibly recognize vascular invasion in many HCC cases but may have difficulty in equivocal cases, which may lead to either understaging or overstaging of the tumors. This may have a significant impact on prognostic assessment and therapeutic decision making. Our observations indicate the need for improved definition for vascular invasion in HCC. © 2010 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd.