Interobserver study on histologic features of idiopathic non-cirrhotic portal hypertension

Background Histologic features of idiopathic non-cirrhotic portal hypertension (INCPH) may overlap with those without INCPH. Recently, these features have been recognized as part of the larger spectrum of porto-sinusoidal vascular disease (PSVD). We assessed interobserver agreement on histologic features that are commonly associated with INCPH and studied whether a provision of relevant clinical history improves interobserver agreement. Methods The examined histologic features include lobular (such as anisocytosis, nodular regeneration, sinusoidal dilatation, increased parenchymal draining veins, and incomplete fibrous septa) and portal tract changes (such as paraportal shunting vessel(s), portal tract remnant, increased number of portal vessels, and obliterative portal venopathy). Thirty-four archived liver samples from patients with (group A) and without (group B) INCPH were retrieved. A total of 90 representative images of lobules (L) and portal tracts (P) were distributed among 9 liver pathologists blinded to true clinical history. Each pathologist answered multiple choice questions based on the absence (Q1) or presence (Q2) of clinical history of portal hypertension. Fleiss’ kappa coefficient analysis (unweighted) was performed to assess interobserver agreement on normal versus abnormal diagnosis, in L and P, based on Q1 and Q2. Results The kappa values regarding normal versus abnormal diagnosis were 0.24, 0.24, 0.18 and 0.18 for L-Q1, L-Q2, P-Q1, and P-Q2, respectively. With true clinical history provided, the kappa values were L- 0.32, P-0.17 for group A and L-0.12, P-0.14 for group B. Four pathologists changed their assessments based on the provided history. Interobserver agreement on the interpretation of L and P as normal versus abnormal was slight to fair regardless of provision of clinical history. Conclusions Our findings indicate that the histologic features of INCPH/PSVD are not limited to patients with portal hypertension and are subject to significant interobserver variation.