Construction Of Risk Prediction Model For Acute Type A Aortic Dissection

Objectives:The acute type A aortic dissection(ATAAD)is a dangerous aortic disease which is characterized by an acute onset,rapid progression,and high mortality.Some patients died from aortic rupture during transfer or in the hospital due to lack of timely and effective intervention.Computed tomography angiography(CTA)clearly and precisely shows the severity and range of lesions in the dissected aortic,suggesting that it plays a significant role in early assessment of the risk of rupture.Based on the characteristics of CTA imaging andcombined clinical manifestations,color Doppler ultrasound examination and laboratory examination results,this study aimed to analyze the related rupture risk factors of ATAAD,and established an effective predictive scoring model to provide an objective basis for clinically judging the possible rupfture of the dissection and selecting an optimal treatment plan.Methods:Two hundred patients with ATAAD in the department of emergency,Guangdong Provincial People’s Hospital between 2012.4.1 and 2017.3.31 were retrospectively recruited as a modeling group.There were 160 males and 40 females with an average age of 53.300±13.19 years.Patients were divided into two groups:ruptured group and unruptured group.Their data were repeatedly analyzed at 24 hours(model A),48 hours(model B),and 72 hours(model C)after CTA.The steps for predictive model construction were as follows:1.Establishing the rupture risk prediction scoring model:The factors with significant difference between the two groups were determined,then the relevant risk factors of ATAAD rupture was analyzed by univariate analysis.The factors of P<0.20 were selected for the two-class logistic regression.The ATAAD rupture risk scoring model was established according to the β value of the relevant risk factors in regression model.2.Model predictive performance evaluation:The degree of discrimination in the prediction model was evaluated by the area under the curve(AUC)of the receiver operating characteristic(ROC)which exhibited a good predictive value when the AUC value is>0.7.The predicted calibration degree was Hosmer－Lemeshow(H-L)goodness of fit test.P>0.05 means the model calibration degree was good.3.Model risk stratification.The risk score is a continuous mathematical score.According to the optimal threshold of its ROC curve and taking an integer,the risk stratification(low-risk and high-risk of rupture)was carried out,and the accuracy and test effectiveness of the layered system were evaluated.4.Prospective validation of the model.80 patients with ATAAD diagnosed in the department of emergency between 2017.4.1 and 2018.5.31 were prospectively recruited.A total of 57 males and 23 females with an average age of(54.00± 13.68)years were included.The predictive and discriminative performance of the score system in other populations was validated.5.Analysis of differences in survival of ATAAD patients with different CTA imaging parameters.To analyze the difference of survival time of ATAAD patients with different CTA imaging parameters:the CTA imaging parameters of the patients were collected and the quartile spacing(P25,P75)of all the measurement data was taken as the dividing point.The measurement data were converted into three classification variables data.The Survival of classified variables was analyzed using Kaplan-Meier.The Log-rank test was statistically significant at P<0.05.From a single perspective,some risk factors associated with aortic dissection have been explained and have certain clinical value.Results:1.Six factors were selected in model A:diabetes(2 scores),celiac dry vascular ischemia(2 scores),branch vascular involvement number>6(2 scores),ascending aorta wide diameter>56 mm(3 scores),white blood cell count(WBC)>14.5×109/L(2 scores)and cardiac troponin T(cTnT)>82.5pg/ml(2 scores).The results showed high predictive value(AUC＝0.811,95%CI:0.687 to 0.935,P<0.001)and goodness of fit(H-L χ2=9.682,P = 0.207);Low risk of rupture(0 to 5 scores)and high risk of rupture(6 to 13 scores)with sensitivity 60.0%(30/50)and specificity 95.3%(143/150).2.The factors were included in model B:female(1 score),time of onset for≤37h(1 score),branch vascular involvement>6(1 score),ascending aorta widest diameter Cross-sectional false lumen area>11 cm2(2 scores),lactic acid(Lac)>2.0 mmol/L(1 scores)and cardiac troponin T(cTnT)>83 pg/ml(2 scores).The results demonstrated high predictive value(AUC = 0.820,95%CI:0.728-0.912,p<0.001)and goodness of fit(H-L χ2=2.003,P=0.960);Low risk of rupture(0 to 3 scores)and high risk of rupture(4 to 8 scores)with sensitivity of 77.3%(58/75)and specificity of 81.6%(102/125).3.The factors were included in model C:age>63 years(2 scores),femal(2 scores),mechanical ventilation(3 scores),aspartate aminotransferase(AST)>80 U/L(2 scores),no distortion of the endometrium(2 scores),diameter of aortic sinus>41 mm(1 score),widest diameter of ascending aorta>48 mm(1 score),ratio of false lumen area/true cavity area>2.12(2 scores),lactic acid(Lac)>1.9 mmol/L(3 scores)and white blood cell count(WBC)>14.2×109/L(1 score).The results revealed high predictive value(AUC =0.928,95%Cl:0.872 to 0.984,P<0.001)and goodness of fit(H-L χ2=8.331,P=0.402);Low risk of rupture(0 to 6 scores)and high risk of rupture(7 to 19 scores)with sensitivity 83.0%(83/100)and specificity 86.0%(86/100).4.Survival analysis of ATAAD patients with different CTA imaging parameters,the following differences in survival between the groups were statistically significant:grouping of the widest diameter of the ascending aorta,grouping of the ratios of false lumen area/true cavity area,grouping of branch vessel involvement,grouping of sinus tube junction diameters,grouping of false lumen areas of the widest cross section of the ascending aorta,grouping of false lumen arc lengths of the widest cross section of the ascending aorta,and grouping of false lumen thrombosis.With the increase of the above variables,the cumulative survival rate of patients with ATAAD after CTA was decreased accordingly.The cumulative survival rate of patients with non-thrombosis in the pseudocavity was lower than that of patients with thrombosis.Conclusion:In the current this study,we constructed three sets of models:Patients with ATAAD were examined for a predictive model of the risk of dissection fractures within 24 h after CTA(A model),patients with ATAAD were examined for a predictive model of the risk of dissection fractures within 48 h after CTA(B model),and patients with ATAAD were examined for a predictive model of the risk of dissection fractures within 72 h after CTA(C model).The accuracy of the A model is 86.5%,the sensitivity is 60%,the specificity is 95.3%,and the AUG is 0.808.The accuracy of the B model is 80.00%,the sensitivity is 77.3%,the specificity is 81.6%,and the AUG is 0.845.The accuracy of the C model is 84.5%,the sensitivity is 83.00%,the specificity is 86.00%,and the AUG is 0.908.Based on the above results,the value of clinical application of C model is the highest.With the increasing value of some CTA imaging parameters,the cumulative survival rate of patients with ATAAD after CTA was decreased.These three models can help doctors assess the rupture risk and predict the prognosis of ATAAD patients which enhances the current rupture risk assessment method for ATAAD patients,and is of great significance for improving the survival rate of ATAAD patients.