颅脑创伤后垂体功能减退的列线图预测模型构建

Establishment of nomograph model for individualized prediction of hypopituitarism after craniocerebral trauma

目的:构建颅脑创伤后垂体功能减退的预测模型并验证。方法:选取2021年1月至2022年5月我院进行治疗的620例颅脑创伤患者为研究对象,按照7∶3的比例随机分为建模组(434例)和验证组(186例),其中建模组根据是否发生垂体功能减退分为减退组和正常组;收集患者临床资料,分别采用单因素和多因素Logistic回归分析影响颅脑创伤后垂体功能减退的危险因素;构建预测颅脑创伤后垂体功能减退的列线图模型,采用校正曲线、H-L拟合优度检验评价列线图模型的校准度,绘制受试者工作特征(ROC)曲线验证模型的区分度。结果:建模组434例颅脑创伤患者有143例发生垂体功能减退,发生率为32.95%;多因素Logistic回归分析显示入住ICU(OR=12.644,95%CI 5.800~27.566)、入院GCS评分≤8分(OR=8.168,95%CI 2.478~26.927)、弥漫性脑水肿(OR=5.759,95%CI 2.329~14.241)、脑疝(OR=2.220,95%CI 1.035~4.762)、中线移位≥5 mm(OR=13.479,95%CI 6.640~27.360)、颅内压增高(OR=6.957,95%CI 2.459~19.682)、颅底骨折(OR=2.538,95%CI 1.083~5.950)、住院天数(OR=1.136,95%CI 1.079~1.197)均为影响垂体功能减退的独立危险因素(P<0.05)。构建列线图模型并进行内外部验证,H-L拟合优度检验结果显示,建模组的χ^(2)=7.287,P=0.506,验证组的χ^(2)=7.202,P=0.515,一致性较好;ROC曲线分析结果显示建模组和验证组预测颅脑创伤患者垂体功能减退的ROC曲线下面积(AUC)分别为0.929(95%CI 0.906~0.953)、0.892(95%CI 0.843~0.942)。结论:入住ICU、入院GCS评分≤8分、弥漫性脑水肿、脑疝、中线移位≥5 mm、颅内压增高、颅底骨折、住院天数均为影响颅脑创伤后垂体功能减退的危险因素,基于以上危险因素构建的预测模型可有效预测颅脑创伤后垂体功能减退的风险,有助于临床医师早期识别颅脑创伤后垂体功能减退患者。

Objective:To establish and validate a personalized prediction model of hypopituitarism after craniocerebral trauma.Methods:A total of 620 patients with craniocerebral trauma who were treated in our hospital from January 2021 to May 2022 were selected as the research objects.They were randomly divided into modeling group(434 cases)and verification group(186 cases)according to the ratio of 7∶3.In the modeling group,patients were divided into hypopituitarism group and normal group according to whether they had hypopituitism;the clinical data of patients were collected,and the risk factors of hypopituitarism after craniocerebral trauma were analyzed by single factor and multi factor Logistic regression;an nomogram model was constructed to predict hypopituitarism after craniocerebral trauma,the calibration of the nomogram model was evaluated by calibration curve and H-L goodness of fit test,and the discrimination of the model was verified by drawing the receiver operating characteristic(ROC)curve.Results:143 of 434 patients with craniocerebral trauma had hypopituitarism,the incidence was 32.95%;multivariate Logistic regression analysis showed that admission to ICU(OR=12.644,95%CI 5.800-27.566),admission GCS score≤8 points(OR=8.168,95%CI 2.478-26.927),diffuse cerebral edema(OR=5.759,95%CI 2.329-14.241),cerebral hernia(OR=2.220,95%CI 1.035-4.762),midline displacement≥5 mm(OR=13.479,95%CI 6.640-27.360),increased intracranial pressure(OR=6.957,95%CI 2.459-19.682),skull base fracture(OR=2.538,95%CI 1.083-5.950)and length of stay(OR=1.136,95%CI 1.079-1.197)were all independent risk factors for hypopituitarism(P<0.05).The column line graph model was constructed and internally and externally validated,and the results of the H-L goodness-of-fit test showed thatχ^(2)=7.287,P=0.506 for the modeling group andχ^(2)=7.202,P=0.515 for the validation group,with good agreement;the results of ROC curve analysis showed that the area under the ROC curve(AUC)for predicting hypopituitarism in patients with craniocerebral trauma in the modeling and validation groups were 0.929(95%CI 0.906-0.953)and 0.892(95%CI 0.843-0.942).Conclusion:Admission to ICU,GCS score≤8 points at admission,diffuse brain edema,cerebral hernia,centerline displacement≥5 mm,increased intracranial pressure,skull base fracture,and length of hospital stay are all risk factors affecting hypopituitarism after craniocerebral trauma,the prediction model based on the above risk factors can effectively predict the risk of hypopituitarism after craniocerebral trauma,which is helpful for clinicians to early identify patients with hypopituitarism after craniocerebral trauma.