Anonymous comparative assessment by an academic core laboratory of four angiography-derived fractional flow ratio software in patients with pressure derived iFR and FFR

Abstract

Abstract Background/Introduction Functional assessment of coronary artery stenoses plays an important role in guiding the clinical management of patients with ischemic heart disease. Software to compute angiography-derived fractional flow ratio (angio-FFR) have been validated against pressure-wire-derived FFR (PW-FFR) with an area under the receiver operator characteristic curve (AUC) of 0.93-0.97. However, it is unknown whether commercially available angio-FFR software are equally reliable, and to what extent each might over- or under-estimate the true functional severity of a coronary artery lesion. Purpose The aim of this study was to investigate the diagnostic accuracies of five angio-FFR software/methods by an independent core-lab in a prospective cohort of 390 vessels with carefully documented sites of PW instantaneous-wave-free ratio (PW-iFR) and PW-FFR. Methods One "matcher investigator (MI)" colocalized on angiography the sites of PW measurement with angio-FFR measurements and provided the same two optimal angiographic views and frame selection to independent analysts, who were blinded to invasive physiological results and results from other software. Results are anonymized and randomly presented. The AUC of each angio-FFR was compared with 2D-quantative coronary angiography percent diameter stenosis (2D-QCA%DS) using a 2-tailed paired comparison of ROC (DeLong’s method). Results All 5 software/methods yielded a high proportion of analysable vessels (A:100%, B:100%, C:92.1%, D:99.5%, E:92.1%). The AUC’s for predicting an FFR≤0.8 for software A, B, C, D, E, and 2D-QCA%DS were 0.75, 0.74, 0.74, 0.73, 0.73, and 0.65, respectively (Figure A). The AUC for each angio-FFR was significantly greater than that for 2D-QCA%DS. Furthermore, binary logistic regression analyses showed the predictors of false positives and false negative: severity of lesion stenosis, lesion location, microvascular resistance, and intermediate zone of angiography-derived FFR potentially reduce the diagnostic accuracy (Figure B). Conclusions This head-to-head comparison by an independent core-lab demonstrated that the diagnostic accuracy of various angio-FFR software for predicting a PW-FFR≤0.80 was useful, with a higher discrimination compared to 2D-QCA%DS, however it didn’t reach the diagnostic accuracy (AUC≥0.90) previously reported in validation studies of various vendors. The intrinsic clinical value of "angiography-derived FFR" therefore requires confirmation in large clinical trials.Figure AFigure B