Statistics from Altmetric.com
It has now become possible to measure fractional flow reserve from standard coronary computed tomographic angiography.
The rapid accumulation in clinical data behind this technique has culminated in a Food and Drug Administration approval and also approval by the National Institute for Health and Clinical Excellence (NICE).
The aims of the current manuscript are to explore the scientific principles behind this technique, the published data validating its use, the potential benefits to healthcare systems and also the challenges it is likely to face as it attempts to enter the clinical domain.
Coronary computed tomographic angiography (CTA) is now established as a clinically valuable non-invasive anatomical test for the detection and exclusion of significant coronary disease. A number of prospective multicentre trials have shown coronary CTA to be an ideal test for the exclusion and detection of coronary disease using invasive angiography as the reference.1–3 Despite this, owing to its relatively low positive predictive value of 48% and inability to determine functional significance,1 its use in international guidelines has generally been restricted to patients with chest pain at a low-intermediate risk of having coronary artery disease (CAD).4 For patients at an intermediate risk of CAD, functional testing is generally indicated, and for high-risk patients, invasive coronary angiography (ICA) remains the recommended diagnostic test.
Although this strategy is designed to determine whether a patient’s symptoms are attributable to CAD, and specifically myocardial ischaemia, recent studies indicate that this approach has important flaws. In a study of almost 400 000 patients, Patel et al showed that up to 62% of the patients who underwent ICA in the USA were subsequently found to have no significant obstructive disease. Furthermore, of those patients with a positive stress test, approximately two-thirds had no obstructive disease and, conversely, 28% of the patients with a negative stress test were found to have obstructive disease.5 Chinnaiyan et al also showed a poor relationship between abnormal stress testing and obstructive disease detected by coronary CTA and ICA.6 Given these findings, it is reasonable to consider whether alternative strategies may exist that confer an advantage to current clinical practice, especially taking into account the implications for both health service resource use and patient convenience of a pathway that may include two tests to arrive at a management conclusion.
Of all of the tests available to evaluate the physiological significance of a coronary stenosis, invasive fractional flow reserve (FFR) represents a gold standard test. The technique enables assessment of the ischaemic potential of a stenosis at the point of diagnostic angiography and is defined as the ratio of distal coronary to aortic pressure during maximal coronary vasodilation provided via adenosine-induced hyperaemia.7 A number of studies have demonstrated both the clinical utility and prognostic value of FFR-guided revascularisation.8–11 These data support the notion that plaque-related ischaemia may be a more important marker of clinical outcome than anatomical features alone. As a consequence, we are increasingly observing revascularisation for lesion-level ischaemia as the dominant target for revascularisation.12 This then raises an important, and uncomfortable, truth about anatomical assessment of coronary lesions at angiography: the correlation between the angiographic severity of a lesion and its physiological significance is relatively poor. Thus, there is a consistent discrepancy between the estimated ‘significance’ of a lesion based on the angiographic appearance and the FFR in around 30% of the lesions described in stable,8 acute coronary syndrome13 and mixed populations.14 The consequence of this is that when FFR data are available, the management of patient populations is altered from the initial angiogram-derived plan in between 22% and 48% of the cases.8 13 15–17 The potential value of routine physiological assessment of patients at the time of invasive angiography is high and being tested systematically in the RIPCORD 2 trial (NCT02892903). However, the clinical potential of a screening test that provides both anatomical and physiological coronary data non-invasively on patients presenting with chest pain is unquestionably even greater.
FFR by coronary CTA (FFRCT)
In recent years, it has now become possible to measure FFR from standard coronary CTA datasets (figure 1). This technique has been developed by the company HeartFlow (Redwood City, California, USA) and is termed FFRCT. The derivation of FFRCT is based on computational fluid dynamics (CFD) which involves applied mathematics, physics and computational software to visualise how fluids interact with adjacent objects. CFD relies on the Navier-Stokes equations which describe how the velocity, pressure, temperate and density of a moving fluid are related.
The scientific principles behind the calculation of FFRCT have been extensively published elsewhere18–21 and are beyond the scope of the current manuscript. A step-by-step approach to the process is summarised in figure 2.
FFRCT diagnostic accuracy
Following on from the basic validation of the model,12–15 a series of studies designed to test the diagnostic performance of FFRCT against invasive FFR have been performed including the Diagnosis of ISChemia-Causing Stenoses obtained via NoninvasivE Fractional FLOW Reserve (DISCOVER-FLOW), The Determination of Fractional Flow Reserve by Anatomic Computed Tomographic Angiography (DeFACTO) study and Analysis of Coronary Blood Flow Using CT Angiography: Next Steps (NXT trial).
The DISCOVER-FLOW trial was a study of 103 patients with suspected or known CAD who underwent coronary CTA, ICA and FFR at 4 sites between 2009 and 2011.22 The inclusion criteria included a coronary CTA stenosis with severity ≥50% in a major coronary artery >2 mm in diameter. Of the 156 vessels for which invasive FFR was performed, an FFRCT≤0.8 was shown to have an accuracy of 84.3%, sensitivity of 87.9%, specificity of 82.2%, a positive predictive value of 73.9% and a negative predictive value of 92.2% against an invasive FFR measurement of ≤0.8. This compared favourably to a coronary CTA assessment of stenosis severity ≥50% which had a significantly lower accuracy of 58.6% a positive predictive accuracy of 48.6%. This trial was conducted with the first iteration of FFRCT algorithms (V.1.0). As in all the published studies, coronary CTA scans that were deemed to be non-evaluable by the FFRCT Corelab were excluded.
The DeFACTO study23 was designed to evaluate the diagnostic performance of FFRCT against invasive FFR in patients with stable suspected native CAD who were scheduled for ICA within a 60-day timeframe. Of the 252 patients originating from 17 centres in five countries, 137 (54.5%) were found to have an abnormal FFR determined by ICA. The diagnostic accuracy of FFRCT and coronary CTA was 73% (95% CI 67% to 78%), which did not meet the primary endpoint of 70% of the lower bound of the 95% CI. Despite this, FFRCT showed superior discrimination against a coronary stenosis severity of ≥50% (AUC 0.81 vs 0.68). When vessels interrogated by FFR were compared against FFRCT and coronary CTA stenosis alone, FFRCT showed superior discriminatory ability with a diagnostic sensitivity of 80% and specificity of 61%.
The third trial evaluating the diagnostic performance of second generation FFRCT was the NXT trial.24 This was designed to evaluate the diagnostic accuracy of FFRCT in patients with suspected CAD against invasive FFR as the reference standard. From the 365 patients screened, there were eventually 251 patients included in the final analysis, each with at least one luminal stenosis on coronary CTA of between 30% and 90% in a vessel ≥2 mm. The diagnostic accuracy, sensitivity, specificity, PPV and NPV for FFRCT on a per-patient basis against invasive FFR was 81%, 86%, 79%, 65% and 93%, respectively. On a per-vessel basis, the values were 86%, 84%, 86%, 61% and 95%. A notable difference from the DeFACTO trial was that the NXT study used an improved iteration of the FFRCT algorithms (V.1.04—improved image quality analysis, improved image segmentation, refined physiological models and increased automation). In addition, the cardiac CT acquisition protocols and invasive FFR measurement protocols were standardised across all sites, factors which were believed to have adversely impacted on the results of the DeFACTO study.
Clinical utility and patient outcomes from FFRCT
Although the DISOVER-FLOW, DeFACTO and NXT trials suggested that FFRCT had the potential ability to enhance the diagnostic capabilities of coronary CTA, its use as a clinical tool had not been formally evaluated. Curzen et al, therefore, elected to study whether the routine availability of FFRCT would change clinical decision compared with the results from coronary CTA alone.25 In this study, three interventional cardiologists assessed the CTA (with access to a specialist report) in order to describe presence and severity of angiographic lesions and, hence, reached a consensus management strategy for 200 consecutive patients who had undergone coronary CTA in the NXT trial. The cardiologists were then shown the FFRCT data for each case and again produced a consensus management according to their assessment of which vessels and lesions were significant. The prescribed CTA-derived management strategy was changed in 36% of the cases, with a reduction in PCI rates seen in 30%, and a change in the PCI target vessel treated in 18%. These results were very similar to those derived from the same protocol performed using invasive angiography and FFR in the original RIPCORD study.8
The largest FFRCT study to date is PLATFORM, a prospective, consecutive cohort study of 585 patients.26 It was designed to test the hypothesis that the evaluation of stable patients with suspected CAD by coronary CTA/FFRCT as their default test, would result in lower rates of ICA showing no obstructive disease, without any consequence in terms of major adverse cardiac event rate compared with routine clinical care. The study participants were enrolled in two consecutive cohorts assigned to receive standard care testing or CTA/FFRCT testing. Each cohort was subdivided into two groups based on patient pathway allocation that was made before enrolment into the study. Thus, for patients deemed to require a non-invasive test, standard care represented any functional test (including exercise tolerance, stress echo, nuclear perfusion, stress CMR) while the ‘non-standard’ care was coronary CTA/FFRCT. For those patients requiring ICA (in most cases after a non-invasive test) the standard care was an invasive coronary angiogram ± revascularisation, while the ‘non-standard’ care was coronary CTA/FFRCT first as a gatekeeper to invasive angiography plus revascularisation. In both cohorts, FFRCT was performed whenever there was a coronary stenosis of severity ≥30% on the coronary CTA.
In the invasive cohort, 100% of the patients underwent ICA in the prespecified routine care invasive arm, with 73.3% showing no significant obstructive disease. By contrast, only 39.4% of the patients allocated to CTA/FFRCT required ICA based on this test, of whom only 31.6% showed no obstructive disease. Thus, the primary endpoint (the rate of ICA at 90 days showing no obstructive disease) occurred substantially more often in the planned invasive testing arm (73.3%) vs the coronary CTA FFRCT arm (12.4%, p<0.0001). PLATFORM showed that coronary CTA/ FFRCT has the ability to reduce the number of patients with new-onset chest pain who require ICA and that the proportion subsequently found to have no significant coronary disease at invasive angiography is significantly lower.
The other important result from PLATFORM was that at 1 year, no MACE events had occurred in the 117 patients whose coronary angiograms were deemed unnecessary based on reassuring coronary CTA/FFRCT findings.27 The investigators also showed a lower mean cost (−33%) of medical care with an FFRCT-guided strategy against an invasive strategy at 1 year ($8127 vs $12 145, p<0.0001), with similar clinical outcomes and quality of life indices.
Modelled economic data
There is a lack of contemporary economic data for the use of FFRCT in the UK. In a retrospective review of 410 patients attending a rapid access chest pain clinic over a 12-month period, Rajani et al demonstrated a poor adherence to the 2010 National Institute for Health and Clinical Excellence (NICE) guidelines for the evaluation of patients with chest pain of recent onset (<50%). When this was scaled to a total of 1000 patients and NHS tariffs were applied, it was predicted that the use of coronary CTA and FFRCT as the primary diagnostic test would result in a saving of £2 00 000 per annum for the cohort, a 48% reduction in invasive coronary angiograms and a 49% reduction in PCI procedures.28 Similar economic savings have been predicted in other healthcare systems. Hlatky et al used clinical data from 96 patients in the DISCOVER-FLOW trial to project initial management costs and 1-year outcomes associated with contemporary management strategies against coronary CTA and FFRCT in USA. They showed that a strategy using FFRCT to guide the selection of patients for ICA and PCI had the potential to result in a 30% reduction in cost and 12% fewer events at 1 year, compared with a strategy using ICA and an ‘eyeball’ decision to proceed to percutaneous coronary intervention.29 Similarly, Kimura et al used clinical data from 254 patients derived from the NXT trial and estimated 32% lower costs and 19% fewer event rates using a coronary CTA/FFRCT strategy compared a primary invasive management strategy within the Japanese Healthcare sector.30 These studies send a consistent signal that management strategies incorporating coronary CTA and FFRCT may confer benefits beyond economic savings alone.
What patient groups should undergo FFRCT?
It remains uncertain as to which groups of patients FFRCT should be offered. The DISCOVER-FLOW trial included only those patients with a coronary CTA stenosis of >50%, DeFACTO had no clear stenosis severity cut-offs, NXT used a coronary stenosis severity of 30%–90% and the PLATFORM trial calculated FFRCT for only those patients with a stenosis severity of ≥30%. Despite this, data suggests that angiographic stenosis severity bears a poor relationship to functional significance by invasive FFR.8 31 32 In the FFRCT RIPCORD study, 2.5% of the lesions with a stenosis severity 1%–29% and 6.3% with a stenosis severity between 30% and 50% had a positive FFRCT≤0.8. This indicates that up to 8.8% of the coronary CTA lesions ≤50% may be of haemodynamic significance. Given that these findings were derived from patients already scheduled for ICA, it is unknown whether similar findings would be observed in patients at a lower risk of coronary disease, who undergo coronary CTA as a rule-our test.
DeFACTO, DISCOVER-FLOW and NXT evaluated patients already scheduled to undergo ICA and in whom the pre-test probability of CAD was already high. The diagnostic accuracy of FFRCT in this group of patients, where the prevalence of coronary disease is lower and symptoms are less typical, has not been proven. Similarly, outcome data derived from invasive FFR in the FAME and FAME-2 trials were from different patient cohorts compared with those studied by FFRCT.10 11 With FFRCT having limited direct outcome data, much of its projected benefit on health outcomes is derived from its comparison as a surrogate marker of invasive FFR.
Coronary CTA data quality and coronary calcification
Data from the NXT trial show promising agreement between FFRCT and invasive FFR measurements. However, there remains concern as to whether these data can be reproduced in routine clinical practice. Since FFRCT is ultimately derived from patient-specific coronary anatomy, results are inherently dependent on good quality coronary CTA data. In the DeFACTO, NXT and PLATFORM studies, 11%, 13% and 12% of coronary CT data sets, respectively, were unsuitable for FFRCT analysis. This was primarily due to phase misalignment artefact, coronary motion artefact and extensive coronary calcification. The minimum stipulated requirement is for coronary CTA to be performed on at least a 64 row MDCT scanner and for scans to be according to the Society of Cardiovascular Computed Tomography Guidelines.33 Respiratory and coronary motion artefact should be minimised with breath-hold instructions, and beta blockers should be used where appropriate to achieve a heart rate <65 bpm.
Phase misalignment is recognised as a limiting factor to the diagnostic accuracy of FFRCT. The DeFACTO trial observed a reduction in sensitivity from 86% to 43% and accuracy from 71% to 56% in cases with significant phase misalignment artefact.34 For those centres which use older 64 MDCT scanners, a heart rate <60 bpm is advisable. With regards to coronary artery calcification, there is no arbitrary Agatston score cut-off beyond which FFRCT is not possible. Nørgard et al evaluated the Agatston score in 214 patients from the HeartFlow NXT trial who underwent coronary CTA, FFRCT and invasive FFR.35 Despite a mean Agatston score of 302±468 (range 0–3599), there was no difference in diagnostic accuracy, sensitivity or specificity for FFRCT across Agatston score quartiles. The AUC for the highest quartile of Agatston score (416–3599) was 0.86 vs 0.92 for the low-mid Agatston scores (0–415).
As FFRCT gains traction for routine clinical use, the real-world rejection rate is likely to be significantly higher than that reported by the selected centres who participated in the FFRCT trials. It is, therefore, likely that FFRCT will be offered initially only to centres with established coronary CTA services and/or to those units who have passed an initial quality standard. For more extensive use of FFRCT, the availability of coronary CTA will need to increase, which is inevitable given the recent revision of the NICE CG 95 on Chest Pain of Recent Onset, which recommends CTA for a much higher proportion of this population than is currently undertaken.
FFRCT clinical workstream
Currently, the process of obtaining FFRCT involves uploading coronary CT data sets to HeartFlow. This is followed by the generation of a per-patient specific anatomical digital three-dimensional (3D) model of the coronary artery. Powerful computer CFD algorithms are then applied to the 3D model to generate FFRCT values for each coronary vessel. This is then made available to the referring clinician using secure webservers while maintaining patient anonymity. The time for complete processing of the coronary CT data set has fallen over the last 5 years to <8 hours with clinical results now being available within 24 hours as a result of newer iterations of the CFD algorithms. It is expected that within the next 12 months, processing times will fall to approximately 1 hour, hence opening up the possibility to use FFRCT for patients presenting with acute chest pain.
FFRCT vs workspace-based computed FFR
A number of CT vendors have developed alternative workspace-based solutions to derive FFR from coronary CTA data sets. Coenen et al investigated the performance of an on-site clinician operated prototype CFD Algorithm (cFFR version 1.4; Siemens Healthcare). The authors demonstrated a moderate correlation against invasive FFR (Pearson′s correlation coefficient=0.59). The sensitivity of cFFR against invasive FFR for all lesions was 87.5%, specificity 65.1%, PPV of 64.8%, NPV of 87.7% and overall accuracy of 74.6%.36 In a further paper, the authors demonstrated an overall accuracy of 75% against invasive FFR and documented various factors that potentially could influence the diagnostic performance of the reduced order hybrid algorithm.37 Of note is that even workspace-based solutions require anything from 25 to 115 minutes for a clinician/technician to generate an accurate 3D coronary model. It, thereafter, takes an additional 5–10 minutes to apply the cFFR algorithm to deriver results.37 Although similar such approaches are emerging in the literature, these are not interchangeable with FFRCT. The techniques for deriving CT-based FFR substantially differ by virtue of using 1D models, lumped parameter models and lower resolution coronary segmentation and meshing. As such, the trial data supporting FFRCT and regulatory approval is not perceived to be transferrable to these newer techniques.
HeartFlow FFRCT was selected for a medical technology evaluation with the National Institute of Clinical Excellence in December 2014. Following an independent appraisal process that included a review of the scientific, clinical and economic data pertaining to FFRCT, guidance was from the Medical Technology Advisory Committee (MTAC) of NICE in February 2017.38 In this, three recommendations were made:
Heartflow FFRCT for estimating FFR from coronary CTA is supported by the evidence. The technology is non-invasive and safe and has a high level of diagnostic accuracy.
Heartflow FFRCT should be considered as an option for patients with stable, recent chest pain who are offered coronary CTA as part of the NICE pathway on chest pain. This may avoid the need for invasive angiography and revascularisation. For correct use, Heartflow FFRCT requires access to 64-slice MDCT (or above).
Heartflow FFRCT may lead to cost savings of £214 per patient. Adopting this technology, the NHS in England may save a minimum of £9.1 million by 2022 through avoiding invasive investigation and treatment, achieving a saving of around £7.7 million to the NHS in England each year.
Although coronary CTA is ideally suited to the detection of coronary plaque, there has been in tandem significant interest in its ability to characterise plaque and to determine whether specific plaque morphological features are related to future plaque rupture and myocardial infarction. Well-established markers of high-risk plaque such as low attenuation, positive remodelling, spotty calcification39 40 and the napkin ring sign41 42 remain useful markers, but are restricted to being pure anatomical descriptors of plaque morphology. With the capability of CFD to evaluate pressure, flow and stresses within coronary arteries, there has emerged considerable interest as to whether further physiological descriptors of plaque may better predict future rupture and cardiac events.
Studies have demonstrated that alterations in wall shear stress (WSS) may contribute to the focal development of atherosclerotic plaque.43 Low WSS is sensed by luminal endothelial mechanoreceptors after which a downstream signalling cascade is triggered which includes a reduction in the bioavailability of nitric oxide, a downregulation of prostacyclin and an upregulation in endothelin-1. These events contribute to the development of atherosclerosis.44 Whereas WSS describes the tangential stress-derived from the friction of flowing blood on the endothelial surface, it does not describe the entirety of the forces to which the endothelial surface and plaque is subject to. This total force is known as the traction force which is related not only the tangential force (WSS), but also the axial pressure. Recent studies have shown that using CFD may be possible to measure the WSS within coronary vessels45 and the axial plaque stress46 (figure 3). Initial data from the Exploring the Mechanism of the Plaque Rupture in Acute Myocardial Infarction (EMERALD) study (NCT02374775) indicate that the incorporation of these advanced CFD parameters may take us one step closer to being able to predict vulnerable plaques that subsequently rupture and lead to myocardial infarction.
Alongside new indices derived from CFD, a number of additional multicentre studies are currently underway which will significantly enhance the evidence base of FFRCT. Notably, the Computed TomogRaphic Evaluation of Atherosclerotic DEtermiNants of Myocardial IsChEmia (CREDENCE) trial (NCT02173275) will evaluate the diagnostic accuracy of vessel territory specific ischaemia of integrated stenosis-adverse plaque characteristics-FFRCT against perfusion imaging with invasive FFR as the gold standard. The Dual Energy CT for Ischemia Determination compared to ‘Gold Standard’ Techniques (DECIDE-Gold) Study (NCT02178904) will, on the other hand, evaluate the diagnostic performance of FFRCT against dual energy CT perfusion with invasive FFR as the gold standard.
FFRCT is based on sound scientific bioengineering principles. In the last 5 years, there have been in excess of 140 peer-reviewed articles reporting its application in a multitude of different settings. Outcome, economic and quality of life data have emerged that suggest FFRCT has the potential to change patient management pathways and streamline the use of valuable economic resources. Having been granted a De Novo 510 (k) clearance by the Food and Drug Administration in November 2014, it has now received approval by the MTAC of NICE in February 2017. Commercially available, and already in use by a number of centres worldwide, more than 6000 patients worldwide have now received the HeartFlow analysis with a number of centres now using the technology in routine clinical practice. Future data are likely to further enhance the evidence base behind FFRCT to a critical level at which sceptics change to enthusiasts.
It is possible to evaluate fractional flow reserve from coronary computed tomographic (CT) angiography (FFRCT) using technology that is based on sound bioengineering principles.
FFRCT can be derived from standard coronary CTA data sets without the need for additional medications or additional ionising radiation.
The addition of FFRCT to standard coronary CTA provides a simultaneous assessment of coronary anatomy and lesion functional significance.
Initial data suggest that FFRCT improves the overall accuracy of coronary CTA.
FFRCT offers potential savings to healthcare systems, more streamlined patient pathways and a reduction in the rate of normal coronary angiograms.
FFRCT is available for commercial use and has Food and Drug Administration approval. Draft guidance from National Institute for Health and Clinical Excellence (NICE) indicates support for the use of FFRCT in the UK.
You can get CPD/CME credits for Education in Heart
CME credits for Education in Heart
Education in Heart articles are accredited for CME by various providers. To answer the accompanying multiple choice questions (MCQs) and obtain your credits, click on the ‘Take the Test’ link on the online version of the article. The MCQs are hosted on BMJ Learning. All users must complete a one-time registration on BMJ Learning and subsequently log in on every visit using their username and password to access modules and their CME record. Accreditation is only valid for 2 years from the date of publication. Printable CME certificates are available to users that achieve the minimum pass mark.
We thank Dr Campbell Rogers and Christopher Zarins from HearFlow (Redwood City, California, USA) for figures 1–3.
Contributors All authors have contributed significantly to the manuscript in either principal drafting of the manuscript, scientific literature review, critical review and editing.
Competing interests NC has received unrestricted research grants from HeartFlow, Boston Scientific, St. Jude Medical, Haemonetics, and Medtronic; honoraria/speaker fees from HeartFlow, St. Jude Medical, and Haemonetics; and travel sponsorships from Biosensors, Lilly/D-S, and Abbott Vascular. RR and NC are expert advisers to the NICE MTAC for FFRCT.
Provenance and peer review Commissioned; externally peer reviewed.
Correction notice Since this article was first published online the corresponding authors email address has been updated. The abbreviation FFA has been removed from two sections in the paper and replaced with the correct abbreviation FFR.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.