Construction and validation of prognostic model for cerebral venous and dural sinus thrombosis based on Lasso regression
Objective To construct a prediction model for the prognosis of cerebral venous and dural sinus thrombosis(CVST)based on least absolute shrinkage and selection operator(Lasso)regression and to verify its effectiveness.Methods The clinical data of 98 CVST patients admitted to the Neurosurgery Department of the First Affiliated Hospital of Chongqing Medical University from December 2012 to May 2023 were retrospectively analyzed.The patients'outcomes were assessed based on the modified Rankin Scale(mRS).We compared the baseline data,symptoms at onset,and symptoms on admission of patients in the good prognosis group(mRS＜3 points,81 cases)and the poor prognosis group(mRS ≥ 3 points,17 cases)based on the International Study on Cerebral Vein and Dural Sinus Thrombosis-Risk Score(ISCVT-RS),cerebral venous and dural sinus thrombosis grading scale(CVT-GS)on admission,Glosgow Coma Scale(GCS)score at admission,treatment plan,mRS score on discharge,imaging data and laboratory test data.Binary logistic regression was used to screen out predictors to build a prognosis prediction model 1,and then Lasso regression was used to screen out predictors to build a prediction model 2.The receiver operating characteristic curve(ROC)and area under the curve(AUC)were used to evaluate and compare the performance of the two prediction models.Calibration curves are used to evaluate the calibration of a predictive model.The model was internally verified using Bootstrap self-sampling method.Decision curve analysis(DCA)was used to evaluate the clinical benefit of predictive models.Results Compared with the good prognosis group,the poor prognosis group had higher proportions of patients with consciousness disorder,limb weakness,cerebral parenchymal hemorrhage,venous infarction,sigmoid sinus thrombosis,transverse sinus thrombosis and internal jugular vein thrombosis,and ISCVT-RS score on admission,CVT-GS score,GCS,white blood cell count,neutrophil count,blood glucose concentration,and D-dimer were all increased,while lymphocyte count and serum albumin concentration were all decreased,and the differences were all statistically significant(all P＜0.05).Multivariate logistic regression analysis showed that limb weakness(OR=9.40,95％CI:1.13-78.42),elevated peripheral blood neutrophil count(OR=1.70,95％CI:1.18-2.44)and elevated blood glucose concentration(OR=1.36,95％CI:1.03-1.79)were risk factors for the prognosis of patients with CVST,while increased peripheral blood lymphocyte count(OR=0.05,95％CI:0.01-0.19)and serum albumin concentration(OR=0.82,95％CI:0.70-0.96)were protective factors for the prognosis of patients with CVST.The prediction model 1 was constructed based on the above five predictors.Prediction model 2 was constructed based on six predictors:venous infarction,peripheral blood neutrophil count,lymphocyte count,serum albumin concentration,blood glucose concentration and D-dimer.The AUCs of the two prediction models were 0.95(95％CI:0.90-1.00)and0.94(95％CI:0.89-0.99)respectively.The calibration curve showed good homogeneity between predictions and actual observations.DCA showed that prediction model 2 had a larger normalized net benefit.Conclusion The Lasso regression model based on venous in farction,peripheral blood neutrophil caunt,lymphocyte count,serum albumin concentration,blood glucose concentration and D-dimer provides a more clinically realistic prognosis prediction for CVST patients,and can more quickly and accurately assess the prognosis of CVST patients,thereby guiding clinical treatment.
NETL
NSTL
万方数据
