CT Angiography–derived Fractional Flow Reserve: The Global Game of Thrones

See also the article by van Hamersvelt et al in this issue.

Joe Schoepf, MD, FACR, FAHA, FNASCI, FSCBT-MR, FSCCT, is a professor with appointments in radiology, medicine, and pediatrics at the Medical University of South Carolina (MUSC) in Charleston. At MUSC, Dr Schoepf serves as the director of the division of cardiovascular imaging and vice chair for research, as well as assistant dean for clinical research. His main scientific interest is the use of advanced CT, MRI, image postprocessing, and artificial intelligence techniques for diagnosing disorders of the heart and lung.

Hunter Gray, BS, is a member of Dr Joseph Schoepf’s research group at the Medical University of South Carolina, department of radiology, and serves as a coordinator and researcher.

Christian Tesche, MD, is an assistant professor in cardiology at the Heart Center Munich-Bogenhausen in Munich, Germany. He serves as the head of cardiovascular imaging and his main scientific interest is the use of CT and artificial intelligence algorithms for optimizing public health by integration of technical innovations into streamlined clinical workflows for improved patient care.

The mere anatomic evaluation of coronary stenosis with invasive conventional angiography or coronary CT angiography is an insufficient means to steer patient management toward or away from revascularization. This explains the vast global interest in functional approaches based on coronary CT angiography, which may yield a more relevant noninvasive diagnosis of coronary artery disease (CAD). Fractional flow reserve (FFR) derived from standard coronary CT angiographic data sets (CT FFR) has been validated in previous trials as a reliable method for the noninvasive detection of lesion-specific ischemia in comparison to invasive FFR (1–4). Previous results have shown improved therapeutic guidance to streamline and rationalize the management of patients suspected of having CAD and improved outcomes, while overall health care costs are reduced (5–7). Newer trials focusing on the clinical applicability of CT FFR in real-world scenarios have largely confirmed the evidence created on the use of CT FFR since its inception; the 1-year outcomes from the ADVANCE (Assessing Diagnostic Value of Noninvasive FFR_CT in Coronary Care) registry (real-world clinical utility and impact on clinical decision-making of coronary CT angiography–derived FFR) show low rates of events in all patients, with less revascularization and a trend toward lower rate of major adverse cardiac events and significantly lower rates of cardiovascular death or myocardial infarction in patients with a negative result at CT FFR compared with patients with abnormal CT FFR values (8). CT FFR modified treatment recommendation in two-thirds of patients as compared with coronary CT angiography alone, was associated with fewer negative results at invasive angiography, predicted revascularization, and identified those at low risk of adverse events through 90 days (7). Gaps in knowledge and evidence still exist, such as large-scale comparative-effectiveness studies on the application of CT FFR to a broader patient population with low-to-intermediate risk and more rationalized guidance on the management of patients with lesions in the “gray-zone” FFR range of 0.75 to 0.80 (7,8). Practicing clinicians may also be confused when confronted with abnormal CT FFR values isolated to small-diameter distal segments in the most extreme portions of the coronary circulation; however, these should rarely lead to clinical uncertainties.

Regardless of the areas that warrant additional study, the prospect of rationalized, noninvasive, safe, and cost-effective management of patients suspected of having CAD enabled by CT FFR has prompted health care carriers in the United States and in certain European and Asian economies to initiate health insurance coverage of CT FFR. So far, the evidence-creation toward the clinical implementation of CT FFR has been largely, if not exclusively, driven by a single vendor (HeartFlow; Redwood City, Calif), which holds the intellectual property rights behind the only solution available for clinical use, which has gained approval by the U.S. Food and Drug Administration (FDA). Some practitioners are voicing their wariness and discomfort with a single vendor providing this type of analysis in the health care market and decry the lack of competition; yet, it is important to note that such monopolies are far from unique, nor are they necessarily detrimental per se. In fact, there is a long history of highly disruptive technologies initially coming to market as single-vendor solutions, most notably drug-eluting stents, transcatheter aortic valves, and, more recently, transcatheter mitral valves.

Because the prospect of replacing invasive FFR with a noninvasive method is so attractive and foreseeably becoming a reimbursable service, this application is fascinating researchers, developers, and industries around the world to propose their own technical approaches toward CT FFR. For instance, we, among others, were instrumental in the refinement and validation of deep machine learning–based, artificial intelligence solutions of this clinical problem (3,9).

The investigation led by van Hamersvelt et al (10), which is featured in this issue of Radiology: Cardiothoracic Imaging, is the latest in a series of proposals toward the development of algorithms to identify lesion-specific ischemia at coronary CT angiography. In this single-center study, a total of 57 patients with 77 vessels were included. For CT FFR analysis, an on-site CT FFR algorithm based on patient-specific lumped parameter models was tested (FFR CT, IntelliSpace Portal; Philips Healthcare, Cambridge, Mass). This algorithm uses a lumped parameter model to represent the complete coronary artery tree with boundary conditions being represented as inflow and outflow resistors and are determined based on the ostium and outflow segment geometry. On a per-vessel level, the algorithm demonstrated an area under the curve (AUC) for CT FFR of 0.87 (95% confidence interval [CI]: 0.77, 0.94), which was superior to that of coronary CT angiography (AUC, 0.70; 95% CI: 0.58, 0.80). This is in line with previous investigations on different CT FFR algorithms, showing AUCs ranging from 0.87 to 0.93 for CT FFR (2, 3). Moreover, in the present study the performance of CT FFR in intermediate lesions (25%–69% stenosis) was assessed, demonstrating a high accuracy with an AUC of 0.89 (95% CI: 0.79, 0.95). Additionally, the diagnostic performance of CT FFR depending on the Agatston calcium score was investigated with no significant differences in the diagnostic performance between vessels with low Agatston score (0–100, accuracy 78%) and vessels with high Agatston score (≥101, accuracy 86%). However, authors stated that outliers with a large difference between CT FFR and invasive FFR (>20% difference) were found in vessels with Agatston score of 101 or greater, which has recently been validated in a multicenter study showing a significant impact of calcium burden on the diagnostic accuracy of CT FFR (11).

By using this CT FFR algorithm based on patient-specific lumped parameter models, operating time was found to be 36 minutes per patient, which is comparable to prior studies showing computational times from 15 minutes up to 4 hours (2,9).

The major limitation of the current study, as in many early CT FFR investigations, is the high rate of cases excluded due to poor image quality and inability to investigate ostial lesions (7% and 19%, respectively, in the present study). This limitation, together with the high prevalence of CAD in the study cohort, hampers the generalizability of the results to a broader patient population, as it does not truly reflect the real-world outpatient population with low-to-intermediate risk undergoing coronary CT angiography.

Despite these limitations, the results are further testimony to the global excitement that surrounds the refinement and clinical use of CT FFR. The worldwide ingenuity and zeal to develop ever more effective and accurate CT FFR algorithms is certainly inspiring, healthy, and helpful for furthering this field. However, there can be little doubt that the current FDA-approved commercial solution based on computational fluid dynamics will dominate the clinical market for the foreseeable future, at least in the United States and in those economies that have made recent coverage decisions or are at the brink of doing so. So protective are intellectual property regulations that erstwhile contenders in the medical imaging solutions market have tabled their own CT FFR ambitions and chosen to engage in partnerships rather than competition. Whether there is a space for competing approaches to create alternative commercial opportunities here or in areas of the world that espouse different health care market dynamics than our own will remain to be seen.