探讨超声联合血液生化指标，构建肝门区囊肿(HHC)患儿发生囊肿型胆道闭锁(CBA)风险列线图预测模型。

选择2016年1月至2022年3月于广州市妇女儿童医疗中心，进行肝门-空肠吻合术(Kasai手术)治疗的134例HHC患儿(日龄<120 d)为研究对象。采取回顾性分析法，按照患儿腹腔镜下胆道造影术或Kasai手术中切除组织的组织病理学检查结果，将其分别纳入CBA组(n＝54)与胆总管囊肿(CC)组(n＝80)。收集2组患儿的年龄、性别、术前超声征象及血液生化指标。采用logistic回归分析方法筛选预测HHC患儿发生CBA的独立预测因子，并使用R软件rms程序包绘制HHC患儿发生CBA风险的列线图。采用受试者工作特征曲线(ROC)及曲线下面积(AUC)、校准曲线、Hosmer-Lemeshow检验及决策曲线分析(DCA)评价该列线图模型预测效能。采用bootstrap方法对该列线图模型进行内部验证。2组患儿进行HHC超声检查时日龄、性别构成比等一般临床资料比较，差异均无统计学意义(P>0.05)。本研究遵循的程序符合2013年修订的《世界医学协会赫尔辛基宣言》要求。

①2组患儿HHC最大长径、最大横径，空腹胆囊形态异常、肝门区纤维斑块、肝门淋巴结肿大、肝内胆管扩张、HHC内胆泥沉积发生率，以及血清总胆红素(TBIL)、直接胆红素(DBIL)、γ-谷氨酰氨基转移酶(GGT)浓度分别比较，差异均有统计学意义(P<0.05)。②多因素非条件logistic回归分析结果显示，HHC最大长径、空腹胆囊形态异常、肝内胆管扩张、血清DBIL浓度均为HHC患儿发生CBA的独立影响因素(OR＝0.871、70.251、0.007、1.089，95%CI：0.780~0.972、2.445~2 018.581、0~0.530、1.026~1.156，P＝0.014、0.013、0.025、0.005)，并根据这4项因素建立预测HHC患儿发生CBA风险的列线图模型。③该列线图模型预测HHC患儿发生CBA风险的AUC为0.996(95%CI：0.991~1.000，P<0.001)，说明该模型对HHC患儿发生CBA和CC区分度好，其预测HHC患儿发生CBA的最佳临界值为0.512，敏感度为97.5%，特异度为96.3%。④该模型的Hosmer-Lemeshow检验的P值为0.992；绘制该模型的校准曲线图分析结果显示，其校准曲线与理想曲线非常接近，该模型预测的HHC患儿CBA发生率与实际发生率一致性高，校准度好。⑤DCA曲线显示，该模型曲线显著高于所有HHC患儿均发生CBA与均未发生CBA这2条极端曲线，表明该模型预测HHC患儿发生CBA的临床净收益大。⑥采用bootstrap方法对该模型进行内部验证结果显示，其预测HHC患儿发生CBA的AUC为0.983(95%CI：0.976~1.000，P<0.001)。

本研究基于HHC最大长径、空腹胆囊形态异常、肝内胆管扩张及血清DBIL浓度4项指标，构建了预测HHC患儿发生CBA风险的列线图预测模型。该模型区分度、校准度好，具有临床适应性。由于本研究仅为单中心研究，未对该模型进行外部验证，该模型对HHC患儿发生CBA的预测效能仍然有待大样本、多中心研究进行验证。

To establish a prediction model based on ultrasonic features combined with biochemical parameters to predict the risk for cystic biliary atresia (CBA) in neonates and infants with hepatic hilar cysts (HHC).

A total of 134 children aged under 120 days with HHC who received Kasai operation in Guangzhou Women and Children′s Medical Center from January 2016 to March 2022 were selected as research subjects. According to results of intraoperative laparoscopic cholangiography and histopathological examination of Kasai operation, they were divided into the CBA group (n=54) and choledochal cyst (CC) group (n=80) by retrospective method. The age, gender, preoperative ultrasonographic features and biochemical indicators of two groups were retrospectively collected. Logistic regression analysis method was used to identify the independent predictors of CBA in children with HHC and a nomogram was developed by using R software with package " rms". The performance of the nomogram was assessed by receiver operator characteristic curve (ROC) and area under curve (AUC), calibration curve, Hosmer-Lemeshow test and decision curve analysis (DCA). Bootstrap was performed for internal model validation of the nomogram. There were no statistical differences in age of ultrasonography for HHC and gender constituent ratio between two groups (P>0.05). The procedures followed in this study were in accordance with the Helsinki Declaration of the World Medical Association revised in 2013.

①There were significant differences in the maximum length diameter of HHC, the maximum width diameter of HHC, incidence of abnormal gallbladder morphology, portal fibrous plaque, enlargement of portal lymph node, intrahepatic bile duct dilatation, and biliary sludge in HHC, as well as serum total bilirubin (TBIL), direct bilirubin (DBIL), γ-glutamyltransferase (GGT) levels between two groups (P<0.05). ②Multivariate logistic analysis indicated that the maximum length diameter of HHC, abnormal gallbladder morphology, intrahepatic bile duct dilatation and serum DBIL level all were independent influencing factors of CBA in children with HHC (OR=0.871, 70.251, 0.007, 1.089; 95%CI: 0.780-0.972, 2.445-2 018.581, 0-0.530, 1.026-1.156; P=0.014, 0.013, 0.025, 0.005), and a nomogram was developed based on these four prognostic factors. ③The AUC of this nomogram was 0.996 (95%CI: 0.991-1.000, P<0.001) for predicting the risk of CBA in children with HHC, indicating that the prediction model had a good discrimination between CBA and CC, and the optimal cut-off value of this model for predicting the risk of CBA in children with HHC was 0.512, with a sensitivity of 97.5% and a specificity of 96.3%. ④The P-value of the Hosmer-Lemeshow test for this model was 0.992. The calibration curve of this model showed that the calibration curve was very close to the ideal curve, the predicted incidence of CBA in HHC children by this model was consistent with the actual incidence and this model had a good calibration. ⑤DCA curve showed that the model was significantly higher than the two extreme curves of CBA in all HHC children and neighter CBA in HHC children, indicating that the clinical net benefit of predicting CBA in HHC children by this model was significant. ⑥Internal validation of the model using bootstrap method showed that the AUC of predicting CBA in children with HHC was 0.983 (95%CI: 0.976-1.000, P<0.001).

In this study, we constructed a nomogram model to predict the risk of CBA in children with HHC based on four indexes: the maximum length diameter of HHC, abnormal gallbladder morphology, intrahepatic bile duct dilatation and serum DBIL level. The model has good discrimination and calibration, and clinical adaptability. Because this study is only a single-center study, and the model has not been validated externally, it still needs large-sample, multi-center studies to validate the predict efficiency of this model for predicting CBA in children with HHC.