Spectral CT predicts initial bleeding risk in cirrhotic patients with esophagogastric varices

To develop a multiparametric model integrating spectral CT quantitative metrics, venous morphometrics, and clinical biomarkers for the noninvasive prediction of initial bleeding risk in cirrhotic patients with esophagogastric varices (EGV).

This retrospective study evaluated a cohort of 195 cirrhotic patients with EGV who underwent portal venous spectral CT. Univariable and multivariable logistic regression analyses were conducted to screen spectral CT metrics, venous morphometrics, and clinical biomarkers, aiming to identify independent predictive factors for initial bleeding. A joint diagnostic model was subsequently constructed based on these independent predictors and formulated as a nomogram. Model discrimination was quantified via the area under the receiver operating characteristic curve (AUC) and compared using the DeLong test, calibration was visually assessed using calibration plots, and clinical utility was evaluated using decision curve analysis (DCA). Internal validation was performed using bootstrap resampling (200 repetitions).

Following feature selection, six predictors of initial bleeding were identified: EGV normalized iodine concentration (EGV NIC), liver effective atomic number (Liver Eff-Z), superior mesenteric vein diameter (SMVD), portal vein diameter (PVD), lymphocyte count, and alanine aminotransferase (ALT). The integrated multiparametric model demonstrated high apparent discriminative performance, with an AUC of 0.912 (95% CI, 0.864-0.960), significantly exceeding the AUCs of all individual predictors (AUC range, 0.636-0.835; all P < 0.05 by pairwise DeLong tests). Calibration analysis showed good agreement between predicted probabilities and observed bleeding outcomes. Bootstrap internal validation yielded an optimism-corrected AUC of 0.880 (95% CI, 0.787-0.951). Decision curve analysis suggested that the joint model provided greater net benefit across a range of clinically relevant threshold probabilities.

A multiparametric nomogram synergizing selected spectral CT hemodynamics, portal morphometrics, and clinical profiles provides a highly accurate and noninvasive tool for predicting initial bleeding risk in cirrhotic patients with EGV.