Log in to MyACC
CCTA, FFR, and Perfusion PET for Ischemia Diagnosis
Study Questions:
How does the diagnostic accuracy of fractional flow reserve (FFR) computation from coronary computed tomography angiography (CTA) datasets (FFRCT) compare to that of CTA, single-photon emission CT (SPECT), and positron emission tomography (PET) myocardial perfusion imaging (MPI)?
Methods:
The authors conducted a non–prespecified secondary analysis of 208 patients enrolled in the prospective PACIFIC trial. Coronary CTA scans were sent for quantification of FFRCT and this was compared with blinded interpretations of coronary CTA alone, SPECT MPI, and PET MPI for diagnosis of FFR ≤0.80. Importantly, although the FFRCT core lab was blinded to FFR values, they were not blinded to the location of measured FFR. For the primary analysis, datasets not evaluable by FFRCT were excluded, although a secondary intention-to-diagnose analysis of all datasets was performed.
Results:
The primary results showed FFRCT was evaluable in all vessels in 157 (75%) patients and 505 (83%) vessels. In the primary analysis, FFRCT had the highest per-vessel sensitivity (90%, p = 0.03) with marginally lower specificity compared to SPECT (86% vs. 97%, p = 0.098). Overall area under the receiver-operating characteristic curve (AUC) for FFRCT on a per-vessel basis, AUC for FFRCT was also greatest at 0.94 with PET as next closest at 0.87 (p < 0.001). On a per-vessel basis, FFRCT and PET had similar AUC (0.92 for FFRCT vs. 0.91 for PET, p = 0.559). Notably, PET had overall higher per patient diagnostic accuracy (88% vs. 78% for FFRCT, p = 0.012).
When the one in four nonevaluable FFRCT cases were also included in the intention-to-diagnose analysis, FFRCT no longer increased per-vessel AUC relative to coronary CTA (0.83 vs. 0.80, p = 0.261) alone and was numerically lower than PET (0.86, p = 0.157). On a per-patient basis, PET significantly outperformed FFRCT, coronary CTA and SPECT with AUCs of 0.90, 0.79, 0.76 and 0.74, respectively (p = 0.005).
Conclusions:
Overall, although one in four coronary CTA scans is unevaluable for FFRCT, this method may have slightly higher diagnostic accuracy compared to coronary CTA and SPECT, but not compared to PET. Importantly, this analysis was fully blinded for CTA, SPECT, and PET, but only partly blinded for FFRCT.
Perspective:
This paper continues a series of investigations into FFRCT. Unfortunately, the research methods leave much to be desired, placing emphasis on analyses that substantially inflate the accuracy of FFRCT by ignoring 25% of patients in whom FFRCT cannot be fully calculated. Readers should focus only on the intention-to-diagnose results, which are substantially less favorable to FFRCT. The reason for this is that when a patient is referred for coronary CTA, one cannot know in advance whether the image data will be useable for FFRCT. Consequently, for the purposes of comparing different diagnostic methods, physicians and patients should account for the substantial likelihood that the data will not be usable. This is especially important as patients with contraindications for coronary CTA are already excluded, already selecting for a relatively favorable patient population.
Furthermore, the study makes a serious error in favor of FFRCT by unblinding the core lab as to where to make measurements. In clinical practice, physicians will not be able to time travel to know where to make the FFRCT measurements relative to an invasive FFR that may or may not be performed in the future and will thus almost certainly perform less well than in this study.
Keywords: Angina Pectoris, Coronary Angiography, Coronary Artery Disease, Diagnostic Imaging, Fractional Flow Reserve, Myocardial, Myocardial Ischemia, Myocardial Perfusion Imaging, Perfusion Imaging, Positron-Emission Tomography, Tomography, Emission-Computed, Single-Photon, Tomography, X-Ray Computed
< Back to Listings
