AUTHOR=Zheng Zhou , Yu Qingsheng , Peng Hui , Huang Long , Zhang Wanzong , Shen Yi , Feng Hui , Jing Wenshan , Zhang Qi TITLE=Nomogram-based prediction of portal vein system thrombosis formation after splenectomy in patients with hepatolenticular degeneration JOURNAL=Frontiers in Medicine VOLUME=10 YEAR=2023 URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2023.1103223 DOI=10.3389/fmed.2023.1103223 ISSN=2296-858X ABSTRACT=Objective

Splenectomy is a vital treatment method for hypersplenism with portal hypertension. However, portal venous system thrombosis (PVST) is a serious problem after splenectomy. Therefore, constructing an effective visual risk prediction model is important for preventing, diagnosing, and treating early PVST in hepatolenticular degeneration (HLD) surgical patients.

Methods

Between January 2016 and December 2021, 309 HLD patients were selected. The data were split into a development set (215 cases from January 2016 to December 2019) and a validation set (94 cases from January 2019 to December 2021). Patients’ clinical characteristics and laboratory examinations were obtained from electronic medical record system, and PVST was diagnosed using Doppler ultrasound. Univariate and multivariate logistic regression analyses were used to establish the prediction model by variables filtered by LASSO regression, and a nomogram was drawn. The area under the curve (AUC) of receiver operating characteristic (ROC) curve and Hosmer–Lemeshow goodness-of-fit test were used to evaluate the differentiation and calibration of the model. Clinical net benefit was evaluated by using decision curve analysis (DCA). The 36-month survival of PVST was studied as well.

Results

Seven predictive variables were screened out using LASSO regression analysis, including grade, POD14D-dimer (Postoperative day 14 D-dimer), POD7PLT (Postoperative day 7 platelet), PVD (portal vein diameter), PVV (portal vein velocity), PVF (portal vein flow), and SVD (splenic vein diameter). Multivariate logistic regression analysis revealed that all seven predictive variables had predictive values (P < 0.05). According to the prediction variables, the diagnosis model and predictive nomogram of PVST cases were constructed. The AUC under the ROC curve obtained from the prediction model was 0.812 (95% CI: 0.756–0.869) in the development set and 0.839 (95% CI: 0.756–0.921) in the validation set. Hosmer–Lemeshow goodness-of-fit test fitted well (P = 0.858 for development set; P = 0.137 for validation set). The nomogram model was found to be clinically useful by DCA. The 36-month survival rate of three sites of PVST was significantly different from that of one (P = 0.047) and two sites (P = 0.023).

Conclusion

The proposed nomogram-based prediction model can predict postoperative PVST. Meanwhile, an earlier intervention should be performed on three sites of PVST.