| Background and objectiveAcute coronary syndrome(ACS)is a series of common and serious cardiovascular diseases characterized by acute myocardial ischemia.Percutaneous coronary intervention(PCI)is the most commonly used method of revascularization in patients with ACS.Compared with stable coronary heart disease,PCI is more frequently used in patients with ACS and the clinical evidence is more reliable.In addition,the incidence of revascularization in patients underwent PCI for ACS was significantly higher than that for stable coronary heart disease.Revascularization is a usual cardiovascular adverse event in patients underwent PCI for ACS.As one of the effective indicators to evaluate PCI,revascularization will offset the benefits from PCI in a part of ACS patients.Currently,coronary angiography is the gold standard to judge whether patients underwent PCI need revascularization.However,as an invasive examination,coronary angiography may leads to a variety of potential complications,hospitalization and high cost.It is a common clinical situation for ACS patients underwent their first PCI over 5 years to visit the department of cardiology.To our knowledge,there is no simple and effective method to calculate individualized probability of revascularization before angiography in such patients.The aim of this retrospective cohort study was to develop and validate the prediction method of pre-angiographic individualized probability of revascularization in patients underwent first PCI for ACS over 5 years.Methods and resultsFrom October 1,2017 to May 31,2019,clinical data of patients successfully underwent their first PCI for ACS over 5 years in our hospital was collected from the second clinic of Cardiology Department of Xijing Hospital,and made up the training dataset for the prediction model.From June 1,2019 to September 30,2019,clinical data of patients successfully underwent their first PCI for ACS over 5 years in our hospital was collected from the Department of Cardiology of Xijing Hospital,and made up the validation dataset for the prediction model.According to the necessity of revascularization,patients were divided into two groups: revascularization group(case group)and non-revascularization group(control group).Inclusion criteria for the case group: the necessity of revascularization for patients was confirmed to be positive according to the result of coronary angiography or further fractional flow reserve(FFR)test.Inclusion criteria for the control group: the necessity of revascularization for patients was confirmed to be negative according to the result of coronary angiography(adding FFR test if necessary)or coronary artery enhanced CT;or according to the judgment by cardiologists based on relatively recent coronary angiography or coronary artery enhancement CT adding all other clinical data,and corrected by one-month follow-up.Exclusion criteria:(1)patients were prepared to undergo planned revascularization;(2)patients refused necessary diagnostic procedure such as coronary angiography or coronary enhancement CT recommended by cardiologists;(3)insufficient clinical data or lost to follow-up.Two independent data collectors obtained patient-related data through a systematic review of hospital electronic cases.Significance test of the difference between groups of baseline characteristics for patients: continuous data were compared using t-test or Wilcoxon-test as appropriate.Categorical data were compared using chi-square test or Fisher exact test as appropriate.To develop the prediction model: Logistic regression analysis was performed in the training dataset,and step-by-step method was used to screen the candidate predictors.The regression equation was established according to the independent variables and their corresponding regression coefficients in the results of Logistic regression analysis.The optimal regression model is selected according to the Bayesian information criterion.The collinear diagnosis of the regression equation is performed to confirm the independence of the residual.The Pre-angiographic prediction model is presented by nomogram.To validate the prediction model,the discrimination of the prediction model was assessed in both training dataset and validation dataset.By taking the predicted need for revascularization as the test variable and the actual need for revascularization as the state variable,the receiving operating characteristic(ROC)curve was drawn and the area under curve(AUC)was calculated.The calibration of the prediction model was assessed in both training dataset and validation dataset.The calibration curve of the prediction probability of revascularization by prediction model and the actual probability of revascularization was drawn.The calibration of the prediction model was evaluated by unreliability U test.Decision curve was plotted.Decision curve analysis was performed to judge the clinical use of the prediction model by quantifying the net benefits with different probability thresholds in the independent validation dataset.P<0.05 was considered to be statistically significant.A total of 308 patients were enrolled in the training cohort,including 133 patients in the case group and 175 patients in the control group.A total of 168 patients were selected in the validation cohort,including 93 patients in the case group and 75 patients in the control group.In the training cohort,64 candidate predictors were screened by Logistic regression analysis,and chest pain,troponin I(Tn I),age at first PCI,initial diameter stenosis≥90% at first PCI and prior multi-vessel disease were selected as predictors.The probability of the necessity of revascularization = 1 /(1 + exp(-(-2.167193 + chest pain× 3.104935 + Tn I × 1.209669-age at the first PCI × 0.0572564 + initial PCI stenosis ×0.9607565 + previous multi-vessel disease × 0.7433985))).The results of variance inflation factor showed no multi-collinearity among independent variables in the prediction model.The nomogram drawn according to the regression equation was competent to visually evaluate the specific probability of the necessity of revascularization.The prediction model showed good discrimination as the AUC in the training dataset is0.830(95% CI,0.786 ～ 0.874)and the AUC in the validation dataset is 0.820(95% CI,0.756 ～ 0.884).The prediction model showed good calibration(p>0.05 in unreliability U test)in both training dataset and validation dataset.The decision curve analysis showed that when the probability threshold was set between 0.05 and 0.95,the decision curve of the prediction model did not coincide with the two invalid lines,and the net benefit from using this prediction model had real benefit.Practically,the false negative harm of judging the necessity of revascularization exceeded the false positive harm.Therefore the probability threshold of this nomogram should be set between 0.05 and 0.5.Conclusion1.Chest pain,Tn I>0.03ng/ml,age at first PCI,initial diameter stenosis≥90% at first PCI and prior multi-vessel disease are independent predictors of revascularization in patients underwent first PCI for ACS over 5 years.2.The nomogram shows acceptable predictive ability to the individualized probability of revascularization in patients underwent first PCI for ACS over 5 years that composed of five variables including chest pain,Tn I level,age at first PCI,initial diameter stenosis at first PCI and number of prior diseased vessels. |
PDF Full Text Request