ABC | Volume 114, Nº2, February 2020

Original Article Vieira et al. Evaluation myocardial ischemia with iFR Arq Bras Cardiol. 2020; 114(2):256-264 Figure 1 – Flow chart of the study showing the heart team decision making for the stenotic lesions evaluated. iFR: instantaneous wave-free ratio; TAVI: transcatheter aortic valve implantation. No functional exams or discordance between non-invasive functional tests and coronary angiograpy. 78 patients Heart team decision making Invasive analysis by iFR 52 patients (iFR ≤ 0.89 positive for ischemia) Other diagnostic or therapeutic approach 26 patients Inclusion in the study 52 patients/96 lesions 56 moderate lesions (median iFR = 0.92), 15.5% treated with stent 40 severe lesions (median iFR = 0.79), 52.5% treated with stent 3 (3.1%) lesions treated with surgical revascularization after functional analysis 3 (3.1%) lesions with positive iFR not treated with revascularization (of any type) (pre-TAVI patient) coronary anatomy (3.1%) and transcatheter aortic valve implantation (TAVI) (3.1%) – these therapeutic decisions were made by the Heart Team. Regarding the localization of the stenotic lesions, 52 lesions were located in the anterior descending artery (54.1%), 11 in the circumflex artery (11.4%), 9 in the right coronary artery (9.3%), 10 in the diagonal branch (10.4%), 9 in the marginal branch (9.3%), 1 in the left posterior descending coronary artery (1.1%), 1 in the right posterior descendent artery (1.1%), 1 in the intermediate artery (1.1%), 1 in the posterior ventricular branch (1.1%), and 1 in the left main (1.1%). Characteristics of the patients are described in Table 1. There was a predominance of men and a high frequency of coronary risk factors, especially diabetes mellitus and smoking. The frequency of clinical manifestations was not different between chronic CAD and acute CAD patients. Most patients showed significant lesion in only one vessel, and approximately two thirds of them were not treated with stent placement. Statistical modelling and graphic analysis Two logistic regression models were constructed to evaluate the need for stent placement. Model 1 was implemented using the variables selected by the elastic net - diabetes mellitus, dyslipidemia, presence of moderate stenosis and positive iFR. Model 2 was composed by the variables that showed statistical significance in the previous model - presence of moderate stenosis and positive iFR. Both dyslipidemia and diabetes mellitus lost statistical significance in the second model (Table 2). Classification trees were developed to evaluate interactions between thepredictors identifiedby logistic regressionand facilitate their interpretation. (Figures 2 and3). An iFR≤0.87was statistically associated with the occurrence of stent implementation, in nearly 37.5% of moderate stenotic lesions. Figure 4 shows the box plot of the distribution of iFR values for moderate and severe lesions treated with stent placement. Among these, median iFR was 0.92 (0.82-0.94) for moderate lesions and 0.79 (0.61–1.00) for severe lesions, i.e ., there was a higher variability in iFR values in severe lesions. The ROC curve evaluated sensitivity and 1-specificity of the iFR cut-off determined using the classification tree. Figure 5 depicts the ROC curve for the iFR, with an area under the curve of 0.7933 (95%CI, 0.6918-0.8949). A sensitivity of 0.57 and a 1-specificitity of 0.88 were obtained for an iFR cut-off of 0.87. Discussion Previous studies have validated the iFRmethod in comparison with the FFR. the iFRwas shown to be non‑inferior to the FFR for 258