基于LASSO回归和Nomogram预测经皮肾活检术后出血风险

Predicting the risk of postoperative hemorrhage in percutaneous renal biopsy based on LASSO regression and nomogram

摘要:

目的 探讨超声引导下经皮肾穿刺活检术后出血的危险因素，并构建列线图预测模型。方法 回顾性收集2022年6月1日至2023年6月1日在十堰市太和医院超声医学科进行实时超声引导下经皮肾穿刺活检住院患者的临床及影像资料。采用LASSO回归、Logistic回归分析，构建超声引导下肾穿刺活检术后出血的列线图预测模型，利用受试者工作特征（ROC）曲线、校正曲线（calibration curve）和决策曲线分析（DCA）三个层面对模型进行评估。结果 最终纳入206例超声引导下肾穿刺活检的患者。LASSO回归及Logistic回归分析结果显示，高血压病史（O^R=5.339，P<0.001）、穿刺肾下极皮质厚度（O^R=0.410，P<0.001）、穿刺肾皮髓质分界不清（O^R=6.133，P<0.001）和穿刺时患者不能配合（O^R=4.525，P=0.004）是超声引导下肾穿刺活检后出血的独立危险因素。列线图预测模型具有良好的诊断效能AUC=0.891，95%CI为0.842~0.941，绘制模型校准曲线，平均绝对误差为0.026，理想曲线和校正曲线贴合较好。Hosmer-Lemeshow 检测χ2=6.599，P=0.580（P>0.05），表明该模型的准确度较好。绘制临床决策曲线显示当列线图预测模型的阈概率小于 89%时，该模型的临床净收益率最高。结论 高血压病史、穿刺肾下极皮质厚度、穿刺肾皮髓质分界不清、穿刺时患者不能配合是超声引导下肾穿刺活检术后出血的危险因素；基于列线图模型预测超声引导下肾穿刺活检术后出血具有可行性，可以为临床评估肾穿刺活检后出血风险提供可视化依据。

Abstract:

Objective To explore the risk factors for postoperative bleeding after ultrasound-guided percutaneous renal biopsy and establish a nomogram model.Methods Clinical and imaging data of inpatients who underwent real-time ultrasound-guided percutaneous renal puncture biopsy were retrospectively collected in the Department of Ultrasound Medicine of Taihe Hospital, Shiyan City, between June 1, 2022 and June 1, 2023. LASSO regression and Logistic regression analysis were used to establish a nomogram model for postoperative bleeding after ultrasound-guided percutaneous renal biopsy. The model was evaluated at three levels: receiver operation characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA).Results A total of 206 patients who underwent ultrasound-guided percutaneous renal biopsy were included in this study. LASSO regression and logistic regression analysis showed that a history of hypertension (O^R=5.339, P<0.001), cortical thickness of the lower pole of the punctured kidney (O^R=0.410, P<0.001), unclear boundary between the cortex and medulla of the punctured kidney (O^R=6.133, P<0.001), and patient non-cooperation during puncture (O^R=4.525, P=0.004) were independent risk factors for postoperative bleeding after ultrasound-guided renal puncture biopsy. The nomogram model had good diagnostic performance with an AUC of 0.891 and 95% CI of 0.842–0.941. The calibration curve showed a mean absolute error of 0.026, indicating a good fit between the ideal curve and the calibration curve. The Hosmer-Lemeshow test χ2=6.599, P=0.580 (P>0.05), indicating good accuracy of the model. The clinical decision curve showed that the model had the highest net benefit when the threshold probability of the nomogram model was less than 89%.Conclusion A history of hypertension, cortical thickness of the lower pole of the punctured kidney, unclear boundary between the cortex and medulla of the punctured kidney, and patient non-cooperation during puncture are risk factors for postoperative bleeding after ultrasound-guided percutaneous renal biopsy. The nomogram model can feasibly predict postoperative bleeding after ultrasound-guided renal puncture biopsy and provide visual evidence for the clinical assessment of bleeding risk.