Development and validation of a prediction model for esophageal varices by changes in spleen size after Rex surgery

Abstract Purpose To develop a noninvasive prediction model for esophageal varices (EVs) based on changes in spleen size after Rex surgery. Method The clinical data of children with cavernous transformation of the portal vein who underwent Rex surgery at the Department of Hepatobiliary Surgery of our hospital from 2014-09 to 2021-12 were collected, and the children were divided into a no-to-mild group and a moderate-to-severe group according to the EV status on postoperative gastroscopy. Variables related to changes in spleen size were included in logistic regression models. Construction and internal validation of a postoperative moderate-to-severe EV risk prediction model. Results A total of 78 children were included, 55 in the no-mild group and 23 in the moderate-severe group. The splenic thickness difference (STD), splenic long diameter difference (SDD), and splenic volume difference (SVD) were included in the multifactorial logistic regression analysis, and the regression equation obtained was modeled as logit(P)= -STDx0.18-SVDx0.011+0.502. The STD and SVD are independent risk factors for moderate-to-severe EV after surgery. The area under the ROC curve was 89.73%, the optimal threshold point was -0.952, and its specificity and sensitivity were 82.6% and 83.3%, respectively.The model was internally validated, and the C-index was 0.897, indicating good discrimination and calibration. Conclusion The model constructed by multifactorial logistic regression is valuable and effective for the noninvasive detection of postoperative EVs, and deserves further research.