Aims: In the era of targeted therapeutics, histological typing of hepatobiliary carcinomas has major clinical implications. Little is known about the reproducibility of the pathological diagnosis of primary liver carcinomas. Therefore, this study aimed to evaluate the worldwide variation in the pathological expert diagnoses of primary liver carcinomas with fibrous stroma in patients who did not have cirrhosis.
Methods: A single set of slides was selected from 25 tumours, and this set was reviewed independently by 12 pathologists who have worldwide expertise in liver tumours. Reproducibility of the diagnoses was evaluated by Light's kappa, and diagnoses were clustered by multidimensional scaling. Immunohistochemistry was performed after histological review.
Results: The interobserver reproducibility for diagnosis of hepatocellular carcinoma subtypes and cholangiocarcinomas was poor (kappa 0.23-0.52), even when the experts considered that the diagnosis required no additional stains or clinical information. Interestingly, multidimensional scaling revealed three main clusters of tumours: hepatocellular carcinoma with no other specifications (n = 13), fibrolamellar hepatocellular carcinoma (n = 3) and cholangiocarcinoma (n = 9). Using immunohistochemistry, these histological clusters correlated with expression of anti-hepatocyte and anti-cytokeratin 19 (p<0.001).
Conclusions: The results demonstrate the poor reproducibility among experts of the pathological diagnosis of primary liver carcinomas with fibrous stroma in patients who did not have cirrhosis, and highlight that the systematic use of immunohistochemistry may improve the diagnostic accuracy.