Inter-reader agreement for the assessment of portal vein thrombosis on computed tomography

To evaluate the inter-reader reliability of CT-based measurements of portal vein thrombosis (PVT) to support the development of standardized response criteria endorsed by the Vascular Liver Disease Group (VALDIG) group.

In this retrospective bi-centric study, 44 patients with PVT who underwent contrast-enhanced CT (2020-2024) were included. Two independent abdominal radiologists evaluated thrombus characteristics using pre-specified standardized measurements on portal venous phase CT acquisition. For the analysis of the main PVT, quantitative continuous measurement included thrombus length, thickness, and remnant lumen assessment based on diameter (%RL_diam) and surface area (%RL_surf). Categorical classifications were derived using both three-tier and four-tier occlusion scales. Qualitative categorical assessments were performed for involvement of the right and left portal branches, splenic vein, and superior mesenteric vein. Inter-reader agreement was assessed using intraclass correlation coefficients (ICC), weighted and unweighted Cohen's kappa (κ), and Bland Altman plots.

Inter-reader agreement for thrombus thickness and length yielded ICC values of 0.92 and 0.84 respectively. Agreement was highest for diameter-based remnant lumen assessment (%RL_diam; ICC = 0.93), outperforming surface-based measurements (%RL_surf; ICC = 0.83). Bland Altman plots confirmed narrower limits of agreement for diameter-based metrics (±22%) compared with surface-based metrics (-32% to +40%). Categorical classification based on diameter showed substantial agreement (weighted κ = 0.71 and 0.78 for three- and four-tier systems, respectively). Semi-quantitative assessment demonstrated substantial concordance, particularly for the right portal branch (κ = 0.77) and superior mesenteric vein (κ = 0.72).

Diameter-based quantitative assessment of PVT on CT demonstrates higher inter-reader reproducibility than surface-based measurements. The three-tier diameter-based categorization achieved substantial agreement and aligns with current hepatology guidelines, supporting its applicability in clinical practice and research settings.