Scholarlink

Objective <p>Postoperative pancreatic fistula (POPF) following splenectomy in patients with Wilson’s disease (WD) complicated by portal hypertension adversely affects prognosis. Therefore, developing and validating an individualized nomogram to predict POPF risk is crucial.</p> Methods <p>This retrospective study included 519 patients with WD-associated splenomegaly and hypersplenism who underwent splenectomy between January 2016 and June 2025. Patients were divided by center into a training cohort (<i>n</i> = 411) and a validation cohort (<i>n</i> = 108). Least absolute shrinkage and selection operator (LASSO) regression was used to select candidate variables, which were then incorporated into multivariable logistic regression to construct a predictive model and develop a nomogram. Shapley additive explanations (SHAP) were applied to interpret the selected variables. Restricted cubic spline (RCS) curves were used for threshold analysis of continuous variables. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration was assessed with calibration curves and the Hosmer-Lemeshow (HL) test, while clinical net benefit was determined via decision curve analysis (DCA). Predictive performance was compared with preoperative composite indices, including C-reactive protein to albumin ratio (CAR), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR).</p> Results <p>The overall incidence of POPF was 18.3% (95/519). LASSO regression identified seven variables: operation time, history of abdominal surgery, type of splenic hilum division, body mass index (BMI), splenomegaly grade, main pancreatic duct (MPD) diameter, and Clavien-Dindo classification (CDC) complications. Multivariable analysis revealed operation time (OR = 1.039), history of abdominal surgery (OR = 2.223), BMI (OR = 1.078), grade III splenomegaly (OR = 2.521), and CDC complications (OR = 1.726) as risk factors; type II splenic hilum division (OR = 0.494) and MPD diameter &gt; 3&#xa0;mm (OR = 0.509) as protective factors (all <i>P</i> &lt; 0.05). SHAP interpretation indicated that operation time (+ 0.075) and history of abdominal surgery (+ 0.059) were the primary contributors to POPF prediction. RCS analysis highlighted BMI exceeding 24&#xa0;kg/m² and operation time exceeding 200&#xa0;min as key thresholds for increased POPF risk. The nomogram demonstrated AUC values of 0.761 (95% CI: 0.697–0.825) in the training cohort and 0.814 (95% CI: 0.702–0.927) in the validation cohort. The HL test yielded <i>P</i> = 0.615 and <i>P</i> = 0.410, respectively, indicating good calibration. DCA showed high net benefit in the training cohort at threshold probabilities of approximately 12%-91% and in the validation cohort at 15%-98%. Compared with CAR, NLR, and PLR, the nomogram exhibited superior AUC (0.761 vs. 0.621, 0.662, 0.633; DeLong test, all <i>P</i> &lt; 0.05).</p> Conclusion <p>The nomogram, constructed from a multivariable model, effectively discriminates and calibrates POPF risk following splenectomy in patients with WD, offering potential clinical utility. Particular attention should be paid to history of abdominal surgery and operation time, with preoperative individualized interventions tailored accordingly.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Risk factors and prediction model development for pancreatic fistula following splenectomy in Wilson’s disease patients with portal hypertension

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月25日

查看全文
  • Zhou Zheng,
  • Yi Shen,
  • Hui-cong Min,
  • Hui Peng,
  • Long Huang,
  • Wan-zong Zhang,
  • Hui Feng,
  • Qing-sheng Yu

摘要

Objective

Postoperative pancreatic fistula (POPF) following splenectomy in patients with Wilson’s disease (WD) complicated by portal hypertension adversely affects prognosis. Therefore, developing and validating an individualized nomogram to predict POPF risk is crucial.

Methods

This retrospective study included 519 patients with WD-associated splenomegaly and hypersplenism who underwent splenectomy between January 2016 and June 2025. Patients were divided by center into a training cohort (n = 411) and a validation cohort (n = 108). Least absolute shrinkage and selection operator (LASSO) regression was used to select candidate variables, which were then incorporated into multivariable logistic regression to construct a predictive model and develop a nomogram. Shapley additive explanations (SHAP) were applied to interpret the selected variables. Restricted cubic spline (RCS) curves were used for threshold analysis of continuous variables. Model discrimination was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration was assessed with calibration curves and the Hosmer-Lemeshow (HL) test, while clinical net benefit was determined via decision curve analysis (DCA). Predictive performance was compared with preoperative composite indices, including C-reactive protein to albumin ratio (CAR), neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR).

Results

The overall incidence of POPF was 18.3% (95/519). LASSO regression identified seven variables: operation time, history of abdominal surgery, type of splenic hilum division, body mass index (BMI), splenomegaly grade, main pancreatic duct (MPD) diameter, and Clavien-Dindo classification (CDC) complications. Multivariable analysis revealed operation time (OR = 1.039), history of abdominal surgery (OR = 2.223), BMI (OR = 1.078), grade III splenomegaly (OR = 2.521), and CDC complications (OR = 1.726) as risk factors; type II splenic hilum division (OR = 0.494) and MPD diameter > 3 mm (OR = 0.509) as protective factors (all P < 0.05). SHAP interpretation indicated that operation time (+ 0.075) and history of abdominal surgery (+ 0.059) were the primary contributors to POPF prediction. RCS analysis highlighted BMI exceeding 24 kg/m² and operation time exceeding 200 min as key thresholds for increased POPF risk. The nomogram demonstrated AUC values of 0.761 (95% CI: 0.697–0.825) in the training cohort and 0.814 (95% CI: 0.702–0.927) in the validation cohort. The HL test yielded P = 0.615 and P = 0.410, respectively, indicating good calibration. DCA showed high net benefit in the training cohort at threshold probabilities of approximately 12%-91% and in the validation cohort at 15%-98%. Compared with CAR, NLR, and PLR, the nomogram exhibited superior AUC (0.761 vs. 0.621, 0.662, 0.633; DeLong test, all P < 0.05).

Conclusion

The nomogram, constructed from a multivariable model, effectively discriminates and calibrates POPF risk following splenectomy in patients with WD, offering potential clinical utility. Particular attention should be paid to history of abdominal surgery and operation time, with preoperative individualized interventions tailored accordingly.