Research On Predictive Value Of Imaging-based Predictive Models For The Prognosis Of Transient Ischemic Attack

Part Ⅰ Research on Predictive Value of Modified ABCD3-I Scores based on DWI patterns for the Prognosis of Transient Ischemic AttackBackground and Objective:China bears a heavy transient ischemic attack(TIA)burden.TIA is a generally recognized stroke risk warning predictor.It is crucial to make accurate risk prediction and stratification for TIA patients to formulate individualized sencondary prevention.ABCD3-I score is a prognostic model with the highest predictive value.Recent studies demonstrated that the predictive value of dual TIA is controversial,and different diffusion-weighted imaging(DWI)patterns portended various risk of recurrent stroke.The present study aims to develop a novel evaluation model by incorporation of DWI patterns with ABCD3-I score in a prospective derivation cohort,and externally validate it in a multicenter TIA cohort.Methods:The derivation of the modified ABCD3-I(mABCD3-I)score was conducted in the TIA database of the First Affiliated Hospital of Zhengzhou University.The validation cohort was from the Henan Transient Ischemic Attack Registry Program.All the patients were enrolled by trained neurology physicians.All registered patients were suggested to be followed-up at 90-day,6-month and 1-year.The predictive outcome was stroke occurrence(including both ischemic and hemorrhagic stroke)and all-cause mortality.Statistical analysis was done with SPSS(version 25.0).The comparisons of baseline characters between continuous variables were presented by mean ± SD,and the categorical variables were expressed by percentage.Cox proportional hazards regression model was conducted to determine the risk factors associated with stroke/death.The mABCD3-I score was developed by the results of multivariate analysis.The receiver operating characteristics(ROC)curves were plotted,and area under the curve(AUC)was calculated as a measure of discriminative ability.The risk reclassification was evaluated by calculating the net reclassification improvement(NRI).The risks of stroke/death of different prognosis scores for each time interval were presented by cluster bars.P ≤ 0.05(two-sided)was deemed statistically significant.Results:Total of 964 TIA patients were enrolled in the derivation cohort,5.2% of patients were lost to follow-up;the stroke/death rates at 90-day,6-month and 1-year were 11.2%,12.7% and 13.9%,respectively.Based on the results of the multivariate analysis,the risk of stroke/death for paitents with multiple DWI lesions of different age was 4-5 times in comparison with patients without.However,the multiple DWI lesions of different artery were not the predictive factor.Three modified ABCD3-I models by incorporation of different DWI patterns were developed: mABCD3-I(1)model was established by replacing dual TIA with multiple DWI lesions of different age,mABCD3-I(2)model was established by replacing dual TIA with multiple DWI lesions of different artery and mABCD3-I(3)model was established by replacing dual TIA with multiple DWI lesions of different age/artery.The AUCs for the mABCD3-I(1)were higher than ABCD3-I score at 90-day(P = 0.002),6-month(P = 0.008)and 1-year(P = 0.018).The similar results were found in the mABCD3-I(3)score at 90-day(P = 0.005),6-month(P = 0.021)and 1-year(P = 0.050).The differences of AUCs between mABCD3-I(2)and ABCD3-I score for any time interval were not statistically significant.Total number of 1030 TIA patients were enrolled in the validation cohort with a lost to follow-up rate of 5.0%.The mABCD3-I(1)score had equal predictive values at 90-day(P = 0.056)and at 6-month(P = 0.054),and a relatively higher value at 1-year(P = 0.044),compared with ABCD3-I score.Z-testing revealed that the AUCs for mABCD3-I(3)score were superior to ABCD3-I score at 6-month(P = 0.033)and 1-year(P = 0.029).Identically,the discriminative abilities between the mABCD3-I(2)and ABCD3-I score had no difference at 90-day,6-month or 1-year.Conclusion:The mABCD3-I score by incorporation with multiple DWI lesions of different age could predict short-and long-term stroke/death risk of TIA patients.Part Ⅱ Research on Predictive Value of Modified ABCD2-I Score based on DWI patterns for the Prognosis of Transient Ischemic AttackBackground and Objective:A transient ischemic attack(TIA)can be considered as a prodrome for stroke.Recent studies observed a lower risk of cardiovascular events after TIA than previously reported.It was speculated that the decreased risk was associated with timely therapy in the acute phase,accurate risk stratification and standardized secondary prevention strategies.The ABCD2-I score,adding brain imaging into the ABCD2 score,had a relatively higher predictive value than ABCD2 score.And the development of imaging technology makes it possible to use ABCD2-I score to triage TIA patients rapidly.The present study aims to develop a novel evaluation model/system by adding diffusion-weighted imaging(DWI)patterns into ABCD2 score(modified ABCD2-I,m ABCD2-I)in a prospective derivation cohort,and externally validate it in a multicenter TIA cohort.Methods:The derivation cohort was from the TIA database of the First Affiliated Hospital of Zhengzhou University.And the validation cohort was based on the Henan Transient Ischemic Attack Registry Promgram.Trained neurology physicians enrolled the patients strictly.All registered patients were suggested to be followed-up at 90-day,6-month and 1-year for stroke occurrence(including both ischemic and hemorrhagic stroke)and all-cause mortality.Statistical analysis was done with SPSS(version 25.0).P ≤ 0.05(two-sided)was deemed statistically significant.The comparisons of baseline characters between continuous variables were presented by mean ± SD,and the categorical variables were expressed by percentage.Cox proportional hazards regression model was conducted to determine the risk factors associated with predictive outcome.Based on the results of multivariate analysis,the m ABCD2-I score was developed.The receiver operating characteristics(ROC)curves were plotted,and area under the curve(AUC)was calculated as a measure of discriminative ability.Results:A total of 981 TIA patients were enrolled in the derivation cohort,a percentage of 5.3% patients lost to follow-up.The stroke/death rates at 90-day,6-month and 1-year were 11.1%,12.6% and 13.8%,respectively.The 90-day/6-month/1-year risk of stroke/death for paitents with DWI positive was 2 times higher in comparison with patients with DWI negative and the risk for patients with multiple DWI lesions of different age was 8-9 times.The m ABCD2-I score was developed by adding DWI patterns(DWI negative,DWI positive,multiple DWI lesions of different age)into the ABCD2 score,based on the results of the multivariate analysis.In the derivation cohort,the predictive value of m ABCD2-I was higher than ABCD2-I score at 90-day(AUC 0.79 VS 0.76,P<0.001),6-month(AUC 0.79 VS 0.76,P=0.001)and 1-year(0.78 VS 0.76,P=0.003).Total number of 1077 patients were enrolled in the validation cohort with 4.7% patients lost to follow-up.The stroke rates were 7.2%,9.1% and 10.3% at 90-day,6-month and 1-year.Patients with DWI positive were 2 times stroke/death risk of 90-day/1-year,and the risk of patients with multiple DWI lesions of different age was 8-9 times,in comparison with patients with DWI negative.The predictive value of m ABCD2-I was higher at 6-month(AUC 0.74 VS 0.72,P=0.019)and 1-year(0.73 VS 0.71,P=0.041)than ABCD2-I score in the validation cohort.However,the effect was not existed at 90-day(0.76 VS 0.75,P=0.099).Conclusion:The m ABCD2-I score by adding DWI patterns into ABCD2 score could predict short-and long-term stroke/death risk of TIA patients.Part Ⅲ Research on Comparison of Predictive Value of Modified ABCD3-I Score and Modified ABCD3-I Score for the Prognosis of Transient Ischemic AttackBackground and Objective:China bears a heavy transient ischemic attack(TIA)burden.The clinical-based ABCD2 score had limited predictive value.Although the tissue-based ABCD3-I score had a highest predictive value,it was difficult to evaluate TIA patients rapidly.ABCD2-I score was a tissue-based prognosis model easy to use.However,the discriminative ability of ABCD2-I score was lower than the ABCD3-I score.Hence,clinicians are in desperate need of a simple and accurate predictive model.The study aims to validate the modified ABCD2-I(m ABCD2-I)and modified ABCD3-I(m ABCD2-I)scores developed in the former 2 parts in a prospective multicenter cohort.Methods:The study population was from the TIA database the First Affiliated Hospital of Zhengzhou University and Henan TIA registry program.Statistical analysis was done with SPSS(version 25.0).Statistical significance was set at P ≤0.05(two-sided).We used χ2 test for categorical variables and t-test for continuous variables.Cox proportional hazards regression model was conducted to determine the risk factors associated with stroke/death.The receiver operating characteristics(ROC)curves were plotted,and area under the curve(AUC)was calculated as a measure of discriminative ability.Results:A total of 1994 patients were included in the final analysis with a 5.1% lost to follow-up rate.The mean age of the patients was 58.29 ± 12.69 years,and 60.5% were male.In the multivariate cox analysis,ABCD2 ≥ 4,DWI positive and multiple DWI lesions of different age were independent risk factors.However,dual TIA,carotid artery stenosis ≥ 50% or multiple DWI lesions of different artery were not statistical significant.The predictive values of both the m ABCD2-I score(AUC 0.78,0.77,0.76)and m ABCD3-I score(0.77,0.76,0.76)were higher than the ABCD3-I(0.73,0.73,0.73)score at 90-day,6-month and 1-year.The difference between the predictive value of m ABCD3-I and m ABCD2-I score was not statistically significant at 90-day(Z = 1.824,P = 0.068)or 1-year(Z = 0.735,P = 0.463).However,the predictive value of m ABCD2-I score was higher than the m ABCD3-I score at 6-month(Z = 2.031,P = 0.042).Conclusion:Both the predictive values of m ABCD2-I and m ABCD3-I scores were much higher than the original ABCD3-I score.Considering the discrimination ability of m ABCD2-I was equal to the m ABCD3-I,the much easier m ABCD2-I was optimal for the emergency assessment for TIA patients.