Diagnostic performance of angiography‐derived fractional flow reserve and CT‐derived fractional flow reserve: A systematic review and Bayesian network meta‐analysis

Abstract Objective Accumulating evidence has demonstrated that fractional flow reserves (FFRs) derived from invasive coronary angiograms (CA‐FFRs) and coronary computed tomography angiography‐derived FFRs (CT‐FFRs) are promising alternatives to wire‐based FFRs. However, it remains unclear which method has better diagnostic performance. This systematic review and meta‐analysis aimed to compare the diagnostic performances of the two approaches. Methods The Cochrane Library, PubMed, Embase, Medline (Ovid), the Chinese China National Knowledge Infrastructure Database (CNKI), VIP, and WanFang Data databases were searched for relevant studies that included comparisons between CA‐FFR and CT‐FFR, from their respective database inceptions until January 1, 2023. Studies where both noninvasive FFR (including CA‐FFR and CT‐FFR) and invasive FFR (as a reference standard) were performed for the diagnosis of ischemic coronary artery disease and were designed as prospective, paired diagnostic studies, were pulled. The diagnostic test accuracy method and Bayesian hierarchical summary receiver operating characteristic (ROC) model for network meta‐analysis (NMA) of diagnostic tests (HSROC‐NMADT) were both used to perform a meta‐analysis on the data. Results Twenty‐six studies were included in this NMA. The results from both the diagnostic test accuracy and HSROC‐NMADT methods revealed that the diagnostic accuracy of CA‐FFR was higher than that of CT‐FFR, in terms of sensitivity (Se; 0.86 vs. 0.84), specificity (Sp; 0.90 vs. 0.78), positive predictive value (PPV; 0.83 vs. 0.70), and negative predictive value (NPV; 0.91 vs. 0.89) for the detection of myocardial ischemia. A cumulative ranking curve analysis indicated that CA‐FFR had a higher diagnostic accuracy than CT‐FFR in the context of this study, with a higher area under the ROC curve (AUC; 0.94 vs. 0.87). Conclusions Although both of these two commonly used virtual FFR methods showed high levels of diagnostic accuracy, we demonstrated that CA‐FFR had a better Se, Sp, PPV, NPV, and AUC than CT‐FFR. However, this study provided only indirect comparisions; therefore, larger studies are warranted to directly compare the diagnostic performances of these two approaches.