Background: Computational fractional flow reserve (FFR) was recently developed to expand the use of physiology-guided percutaneous coronary intervention (PCI). Nevertheless, current methods do not account for plaque composition. It remains unknown whether the numerical precision of computational FFR is impacted by the plaque composition in the interrogated vessels.Methods: This study is an observational, retrospective, cross-sectional study. Patients who underwent both optical coherence tomography (OCT) and FFR prior to intervention between August 2011 and October 2018 at Wakayama Medical University Hospital were included. All frames from OCT pullbacks were analyzed using a deep learning algorithm to obtain coronary plaque morphology including thin-cap fibroatheroma (TCFA), lipidic plaque volume (LPV), fibrous plaque volume (FPV), and calcific plaque volume (CPV). The interrogated vessels were stratified into three subgroups: the overestimation group with the numerical difference between the optical flow ratio (OFR) and FFR >0.05, the reference group with the difference ≥−0.05 and ≤0.05, and the underestimation group with the difference <−0.05. Results:In total 230 vessels with intermediate coronary artery stenosis from 193 patients were analyzed.The mean FFR was 0.82±0.10. Among them, 21, 179, and 30 vessels were in the overestimation, the reference, and the underestimation group, respectively. TCFA was higher in the underestimation group (60%) compared with reference (36.3%) and overestimation group (19%). Besides, it was not associated with numerical difference between OFR and FFR (NDOF) after multilevel linear regression. LPV was associated with NDOF as OFR underestimated FFR with −0.028 [95% confidence interval (CI): −0.047, −0.009] for every 100 mm 3 increase in LPV.Conclusions: High lipid burden underestimates FFR when OFR is used to assess the hemodynamic importance of intermediate coronary artery stenosis. TCFA, FPV, and CPV were not independent predictors of NDOF.

No abstract