Fractional flow reserve (FFR) computed from standard coronary CT scans (FFRCT
) is a novel noninvasive method for determining the functional significance of coronary artery lesions. Although coronary CT angiography (CTA) alone can accurately detect obstructive coronary artery disease (CAD) when compared with invasive coronary angiography [1
], coronary CTA cannot define the hemodynamic significance of CAD. This results in poor positive predictive value owing to an unreliable relationship between a simplistic measure of reduction in coronary artery diameter and ischemia [2
]. The diagnostic performance of FFRCT
compared with coronary CTA alone using invasively measured FFR as the reference standard has recently been published [3
]. The Determination of Fractional Flow Reserve by Anatomic CTA (DeFACTO) study showed improved diagnostic accuracy and improved discriminatory power of FFRCT
as compared with coronary CTA alone for the diagnosis of lesion-specific ischemia in stable patients with CAD [3
]. The DeFACTO study represented the first large-scale validation of the application of computational fluid dynamics for the derivation of FFR from a typically acquired coronary CTA. Importantly, the sensitivity and negative predictive value of FFRCT
were high, suggesting a low rate of false-negative studies. The overall accuracy of FFRCT
was, however, limited by more modest specificity.
Patient preparation for coronary CTA has long been emphasized in societal guidelines to help ensure image data optimization and achieve optimal diagnostic accuracy [4
]. Beta-blockade and sublingual nitroglycerin administration are recommended [4
], because the use of both enhances image quality and diagnostic accuracy in coronary CTA [6
]. Moreover, these agents may also affect coronary motion (owing to inadequately controlled heart rates) and luminal diameter (from inadequate vasodilation of the coronary artery), both of which may influence the fluid dynamic models used to calculate FFRCT
because it depends on accurate image input data to assign correct boundary conditions for physiologic computations. In addition to these medications, CT scan protocols may also affect coronary CTA image quality—with tube current and kilovoltage being the most commonly adjusted parameters—and that may have profound effects on image artifacts, including motion, misalignment artifact, and visualization of severe coronary artery calcifications.
The impact that adherence to best practices in coronary CTA has on the diagnostic performance of FFRCT has not yet been established.
In a substudy of the prospective multicenter international DeFACTO study, we thus sought to evaluate the influence of technical factors related to coronary CTA acquisition, patient preparation, and image artifacts on the diagnostic performance of FFRCT as compared with invasive coronary angiography.
The present results, derived from a large multinational prospective study, suggest that whereas FFRCT provides superior diagnostic performance over coronary CTA alone across a broad range of image quality and patient preparation methods, significant improvements in diagnostic performance of FFRCT can be expected with adherence to best practices for CT image acquisition. These include the use of β-blockade and nitroglycerin administration and coronary CTA scan protocol optimization to reduce misalignment artifacts.
The integration of computational fluid dynamics to typically acquired coronary CTA provides a unique opportunity to define the physiologic significance of CAD without the need for either additional studies or radiation (or both) [3
]. This extends the potential of coronary CTA to not only detect obstructive disease but also help identify those patients who may benefit from an invasive strategy with intended revascularization. Although FFRCT
has previously been shown to be relatively impervious to issues of image quality [12
], the current data suggest that there are significant opportunities to improve accuracy and specificity of FFRCT
by adhering to established best practices of cardiac CTA as defined by guidelines produced by specialty societies [4
]. Given that FFR is the only physiologic method for ischemia detection that has been shown to be useful in improving short-term freedom from unplanned intervention [14
], the ability to predict FFR from a noninvasive test renders FFRCT
a potentially “disruptive” technology—offering for the first time the opportunity to identify those specific lesions that cause ischemia by using a single imaging test that is performed at rest. Given this potential, it is vital to determine the potential variables of coronary CTA acquisition, image quality, and interpretation to ensure optimal computation of FFRCT
It has been previously established that the diagnostic performance of coronary CTA as compared with invasive coronary angiography is improved by both β-blockade and nitroglycerin administration [17
]. Approximately 3–5 minutes after the administration of nitroglycerin, the coronary arteries are significantly larger in both diameter and area, which allows more accurate assessment of anatomic details [18
]. As a further benefit, nitroglycerin administration appears also to improve the assessment of FFRCT
, probably because its computation is based on an anatomic model of the epicardial coronary arteries created from CT data that is assumed to match the degree of vasodilation that might be induced during intracoronary administration of nitroglycerin at the time of FFR measurement. The pressure loss along an epicardial artery is approximately equal to the product of the flow and the segmental resistance of the coronary artery. The segmental resistance is approximately inversely proportional to the fourth power of the radius; thus, a 10% increase in the radius of the vessel attainable by administration of nitroglycerin before coronary CTA acquisition will reduce resistance by approximately 30%.
The use of β-blockade for coronary CTA is also supported by best-practice recommendations [4
]. Recent real-world published data suggest that adherence to these recommended guidelines is variable [19
]. A number of potential explanations exist for this finding, including the time needed for administration and effect of β-blockers and the perception of limited additional value. Given the opportunity to realize the benefit of accurate physiologic data when CT examinations are performed in accordance to societal guidelines, our data support that adherence to these guidelines will improve diagnostic performance, both for anatomic stenosis and for FFRCT
. Interestingly, in our analysis, heart rate at the time of the scan did not affect the accuracy of FFRCT
. This finding may be explained, in part, by other effects of β-blockers beyond heart rate lowering, including reduction in heart rate variability or reduced ectopy during coronary CTA. This potential benefit of β-blockade is often underappreciated by those performing coronary CTA, and as a result, a β-blocker is often not administered when the heart rate is below specified thresholds at presentation. Our data provide further support for the use of β-blockade in coronary CTA, particularly given the growing evidence suggesting the potential to provide information regarding the hemodynamic significance of an anatomic lesion identified on coronary CTA.
has been shown to be less sensitive to issues of image quality than coronary CTA alone [12
]. Unlike coronary CTA, which relies only on the image quality in any individual segment, FFRCT
is determined through the solving of millions of equations thousands of times throughout the cardiac cycle, relying on a significantly larger pool of data that includes not only a specific coronary segment in question but also the remainder of the coronary vasculature, myocardium, and the proximal aorta. Although this added complexity adds to the robustness of FFRCT
, our data suggest that image artifacts commonly seen in the setting of suboptimal patient preparation may significantly reduce the accuracy of FFRCT
. These data constitute evidence that guidelines-based CT acquisition and performance should be adhered to for optimal derivation of FFRCT
This study is not without limitations. The present results are a post hoc analysis of the DeFACTO trial. Although the trial is large and multicenter in nature, some of the presented subanalyses are limited by small sample size. However, the differences that we identified in the diagnostic performance of FFRCT were statistically significant and can be explained by the premises of the FFRCT methods. The findings were meant as exploratory, with the aim of deriving important clues for the optimization of accuracy of FFRCT, and in the future, larger studies should be performed to confirm the findings of the present study. In addition, we lack baseline data about the number of patients who received β-blockade at the time of CT. Further, the coronary CTAs were acquired using various scan platforms without a difference in diagnostic performance; however, we recognize that the use of varied scan platforms somewhat limits our analysis. The importance of β-blockade before coronary CTA acquisition is emphasized in this article. We do not, however, have complete baseline β-blockade use in all patients, which we acknowledge to be a limitation of our assessment. Nevertheless, our analysis does suggest incremental improvement in diagnostic performance of FFRCT. Our data suggest that the diagnostic performance of FFRCT is stable in the setting of coronary motion. It is important, however, to recognize that the coronary CTA examinations included in the DeFACTO study were deemed interpretable by the coronary CTA core laboratory and that the impact of severe coronary motion that would render a coronary CTA uninterpretable has not been evaluated in our study. Finally, we recognize that our analysis did not adjust for all potential confounders. Although this is a limitation given the large number of variables that could potentially impact diagnostic performance of FFRCT—including that one individual performed the invasive FFR and another performed the FFRCT analysis—we do not think it is feasible to create an appropriate multivariable model.