Analysis Of Risk Factors And Prediction Model Of Perioperative Cerebral Infarction In Adult Ischemic Moyamoya Disease

Objective:Risk factors were filtered to stablish a risk prediction score system and a risk stratification of perioperative cerebral infarction in adult ischemic moyamoya disease and to complete validation of scoring system and risk stratification.Methods:This study is divided into two parts:the analysis of risk factors of perioperative cerebral infarction as well as the creation and validation of a predictive scoring system.Analysis of risk factors of cerebral infarction among the perioperative period:A retrospective analysis was performed on the clinical data of 235 adult patients(322 hemispheres)with ischemic moyamoya disease who underwent revascularization in the First Affiliated Hospital of Zhengzhou University from January 2019 to December 2020.Risk factors were included in multivariate Logistic regression analysis which screened out by Chi-square test,rank-sum test,and univariate Logistic regression analysis.In order to identify the independent risk factors of perioperative cerebral infarction,a Logistic regression model was established.Finally,the ROC curve was established to evaluate the modeling effect.Creation and validation the predictive scoring system:Based on the regression coefficient β and OR value obtained in part 1,each independent risk factor was assigned to establish a Logistic scoring system and an Additive scoring system.After that,the modeling cohort was scored.ROC curves were established to assess the discriminative ability of the two scoring systems,and the Hosmer-Lemeshow test was performed to assess the calibrative ability.The clinical data of 112(136 hemispheres)adult patients with ischemic moyamoya disease who underwent intracranial and extracranial revascularization in the First Affiliated Hospital of Zhengzhou University from March 2021 to October 2021 were collected for external validation and the distinguishing correction abilities were evaluated.Finally,the risk stratification of perioperative cerebral infarction was performed based on the approximate trisection method and the rank-sum test as well as the chi-square test was used to verify the ability of differentiation.Results:The results of the risk factor analysis are as follows.(1)Rank sum test:The difference in age and mRS score between the cerebral infarction group and the control group is statistically significant.(2)Chi-square test:The difference in age≥45 years old,age≥50 years,old blood pressure control,hypertension,blood glucose control,preoperative cerebral infarction,the longest diameter of a recent cerebral infarction,recent TIA attack,collateral compensation,the anterior cerebral circulation compensation,and posterior cerebral circulation compensation score between the cerebral infarction group and the control group is statistically significant.(3)Univariate Logistic regression analysis:Age ≥45 years old,age≥50 years old,high admission mRS Score,poor blood pressure control,hypertension,poor blood glucose control,recent cerebral infarction,longest diameter of recent cerebral infarction,recent TIA attack,no circulatory compensation,no anterior and poor posterior cerebral circulatory compensation are risk factors in adult ischemic moyamoya disease(OR>1).(4)Multivariate Logistic regression analysis:High admission mRS score(OR=2.983,1.051～8.466),age≥50 years(OR=4.830,1.365～17.089),poor blood pressure control(OR=4.579,1.225～17.121),longest diameter of recent cerebral infarction(OR=3.025,1.180～7.751),recent TIA attack(OR=2.702,1.164～6.273)and poor posterior cerebral circulation compensation(OR=2.446,1.143～5.234)are independent risk factors for perioperative cerebral infarction in adult ischemic moyamoya disease.(5)The discriminative effect of the Logistic regression model is excellent(AUC=0.908),and the Hosmer-Lemeshow test model fits well(P=0.534).The parameters of the prediction scoring systems and the validation results are shown below.(1)The scoring range of the Logistic scoring system is 0～12 points while the scoring range of the Additive scoring system is 6～41 points.The internal verifications of the two scoring systems show excellent distinguishing ability(AUC>0.8).The results of Hosmer-Lemeshow test show good calibration(P>0.05).(2)According to the Logistic scoring system,patients are divided into low-risk group(<4 points),intermediate-risk group(4～7 points),and high-risk group(>7 points).In the modeling cohort,there are statistically significant differences in the perioperative cerebral infarction among the low-risk group,the intermediate-risk group,and the high-risk group(P<0.05).(3)The external verifications of the two scoring systems show excellent distinguishing ability(AUC>0.8).The results of Hosmer-Lemeshow test show good calibration(P>0.05).(4)In the validation cohort,there were statistically significant differences in the perioperative cerebral infarction among the low-risk group,the intermediate-risk group,and the high-risk group(P<0.05).Conclusions:The risk prediction model of perioperative cerebral infarction can effectively predict the risk of perioperative cerebral infarction,and provide personalized and precise treatment plans for patients according to different risk stratifications.