A Dynamic Nomogram Predicting Portal Vein Thrombosis in Cirrhotic Patients During Primary Prophylaxis for Variceal Hemorrhage

Abstract

Background: Portal vein thrombosis (PVT) would exert a further increase in resistance to portal blood flow, resulting in worsening portal hypertension and poor outcome. This study aimed to identify risk factors and develop an clinically applicable dynamic nomogram predicting the occurrence of PVT in cirrhotic patients during primary prophylaxis for variveal hemorrhage (VH). Methods: The multi-center retrospective study enrolled cirrhotic patients with high-risk varices, which were further divided into training and validation cohorts for 3 years follow-up. A dynamic nomogram based on the Cox proportional hazard regression model was developed with the cutoff value calculated by X-title analysis. The performance of the nomogram was evaluated with Harrell’s concordance index (C-index), calibration curve and decision curve analysis. Results: 91 (34.0%) of the whole cohort were diagnosed with PVT during 3-year follow-up. Variables including carvedilol (P < 0.001), low portal vein velocity (P < 0.001), increased size of esophageal varices (P = 0.005), and high HbA1c (P < 0.001) and procalcitonin (P = 0.015) were identified to be independently associated with PVT, which were further incorporated into the dynamic nomogram with optimal cutoff (8.8 and 14.6) for risk-stratification. The C-indexes (0.894 of internal validation and 0.892 of external validation) and calibration curves demonstrated ideal discrimination and calibration. The thresholds for more reasonable application of the nomogram were 0–0.27, 0–0.66, and 0.04–1.00 at 1, 2, and 3-year, respectively. Conclusion: The dynamic nomogram could be accurately and reliably used for clinical risk-stratification of PVT in cirrhotic patients during primary prophylaxis for VH.