Invasive coronary angiography (ICA) is the final common pathway for patients with probable coronary artery disease (CAD) being considered for revascularisation. Traditionally, decision-making involves inspection of the ICA images and inferring the effect of any stenosis seen upon blood flow through the arterial lumen. This is both subjective and inaccurate,1 2 resulting in unnecessary procedures or undertreatment. Fractional flow reserve (FFR) or related indices can assess the ischaemic potential of a stenotic lesion3 and guide percutaneous coronary intervention (PCI), with benefits in terms of reduced adverse events4–7 and healthcare costs.8 FFR also changes management in both stable9 and unstable syndromes.10 Guidelines support the use of FFR and instantaneous wave-free ratio(iFR).11 12 However, logistic, practical and financial reasons prevent physiological guidance being widely employed in clinical practice.13 Therefore, several systems which compute ‘virtual’ FFR (vFFR) from ICA have been developed. The first of these was the VIRTUheart system (University of Sheffield, UK), employing computational fluid dynamic (CFD) modelling to calculate vFFR.14 Validation of vFFR against invasive FFR is acceptable,15 and data on clinical outcomes are emerging.16 17 However, their applicability and impact on decision-making in ‘real-world’ settings are lacking.

The VIRTU-4 Study aimed to investigate the impact on decision-making of vFFR in patients with acute coronary syndrome (ACS) or chronic coronary syndrome (CCS) undergoing ICA.

This was an investigator-initiated, multicentre, cohort, prospective observational study in which patients with ACS or CCS, who were scheduled to undergo ICA at a tertiary centre (patients with ACS) and four district hospitals (patients with CCS) in Northern England, were enrolled over the 2-year period (2020–2021).

Patients were over the age of 18 years, presenting with CCS or ACS, requiring ICA, ±on-site invasive FFR assessment in the ACS group. In cases where FFR was measured invasively, this was done after intracoronary administration of nitrate, with a pressure-sensitive wire, with the transducer positioned at least 15 mm distal to the lesion, during maximal stable hyperaemia, induced by an intravenous infusion of adenosine at 140 µg/kg/min.3 Across the five hospital sites, 24 interventional cardiologists performed ICA. Exclusion criteria were serum creatinine >180 µmol/L, refractory ischaemia, haemodynamic instability, prior coronary artery bypass grafting (CABG), significant valvular disease, intolerance to antiplatelet drugs, life-threatening comorbidity and failure to consent. A second phase of exclusion, based on the angiographic requirements for modelling, included chronic total occlusion as the only lesion, left main stem or aorto-ostial disease (which is difficult to model), normal coronary arteries (<30% stenosis), lesions with >90% diameter stenosis (FFR being both unnecessary and difficult to model), vessel diameter <2.25 mm and inadequate images (eg, vessel overlap, inadequate contrast or inability to obtain two views). Angiographic images were acquired ensuring that each lesion was clearly displayed in at least two views, at least 30° apart. Angiographic disease was classified as non-significant (0VD) or one, two or three-vessel disease (1, 2 or 3VD), based on what the cardiologists believed to be potentially physiologically significant lesions (≥30% visual stenosis) by visual estimation as this most closely represented standard clinical practice. These were then reclassified following disclosure of vFFR, using a vFFR value of ≤0.80 as the threshold for functional significance. VIRTUheart was, therefore, deployed in all vessels with a visual stenosis of 30–90% as assessed visually by the research team (MG and HH) after ICA.

The VIRTUheart system that employs a three-dimensional (3D) quantitative coronary angiography segmentation is derived from two angiographic images, at end-diastole, displaying the lesion of interest. Arteries were reconstructed, by one of two operators, both experienced in the use of VIRTUheart and CFD modelling (MG or HH), from coronary ostium to at least six vessel radii distal to the lesion of interest. A CFD solver then resolves the Navier-Stokes and continuity equations, from which a pseudo-transient vFFR is calculated. The vFFR is superimposed in colour upon an image of the 3D anatomy.14 18 See central illustration for an example case of vFFR.

The initial management strategy of the patient’s cardiologist was recorded after ICA, immediately after angiography, while the patient was on the table, before any PCI/FFR had been performed. This was categorised as optimal medical therapy (OMT), PCI, CABG or ‘more information required’, as this best reflects management categorisation of standard practice. The cardiologist was encouraged to commit to a strategy, formulated from the patient’s clinical history, comorbidities and angiographic images. The vFFR results were then disclosed to the cardiologist, and any changes in plan documented. Actual changes were not permitted because VIRTUheart is a research tool. The confidence level of the cardiologist in making their treatment decision was recorded at both stages.

The primary endpoint was an intended change in management strategy after vFFR disclosure. This was defined as a change of plan between the categories of OMT, PCI, CABG or ‘more information required’ (invasive FFR, PCI plus invasive FFR or multidisciplinary team). Secondary endpoints included the number of vessels classified as significant, the confidence levels of the cardiologists, the vFFR failure rate, interobserver vFFR variability and agreement with invasively measured FFR (when measured).

Sample size was calculated based on the RIPCORD9 and FAMOUS-NSTEMI10 Studies, which showed that invasive FFR at the time of ICA changed management in >20% of patients. Assuming that vFFR might be less sensitive than measured FFR, we set our power level more stringently. 412 patients were required to provide 85% power at 5% significance to reject a change in treatment in <10% of patients. Categorical variables were presented as counts and percentages. Normally distributed data were presented as mean (±SD) or median (IQR), as appropriate. Normality of data distribution was assessed using the Shapiro-Wilk test. Primary and secondary outcomes were assessed using McNemar-Bowker, χ2, paired Student’s t-test and Mann-Whitney U tests, as appropriate. Differences in confidence levels before and after vFFR disclosure were assessed using one-way repeated measures analysis of variance. Interobserver variability between the two vFFR operators was assessed on randomly selected 10% of the patient cohort and compared using the intraclass correlation coefficient with a two-way mixed model. When comparing invasive and virtual FFR, correlation was assessed using Pearson’s correlation coefficient, agreement with Bland-Altman plots with associated 95% limits of agreement and diagnostic test performance with receiver operating characteristic (ROC) curve analysis.

The entire VIRTUheart Programme, which started in 2009, is reviewed each year by the Sheffield National Institute of Health Research Cardiovascular Patient Panel. For this study, they reviewed the protocol, provided advice concerning how to approach potential participants, and corrections for the patient information sheet and the consent form.

This study demonstrates that ICA-derived vFFR is applicable in 320 (62%) of 517 ‘all-comer’ patients undergoing clinically indicated ICA. Use of vFFR changed the management strategy in 71 (22%) patients. While there was some movement between groups, it did not alter the number of lesions deemed to be haemodynamically significant per patient (1.16 vs 1.18). In the ACS cohort, there was a 62% increase in the number proposed for OMT, whereas in the CCS cohort, there was a 27% decrease. In both groups, the use of vFFR overall increased the cardiologists’ confidence in their decision-making (central illustration).

The strength and originality of this study are its exploration of the ‘real-world’ impact of vFFR on everyday practice in both diagnostic and interventional cardiac catheter laboratories. The most important finding is the proportion of patients in whom it provided a change in management; 14% of those potentially suitable before the ICA was performed, and 22% of those in whom vFFR quantification was possible after ICA. In order to identify the 71 patients in whom vFFR would alter management, it was necessary to apply the software to 320. Nevertheless, screening five to benefit one seems a reasonable clinical yield. It is this ‘intermediate’ group, in which visual ambiguity around ICA interpretation exists, who may benefit from the addition of a physiological test. While invasive FFR is most widely used in clinical practice, application of vFFR may reduce procedural challenges, time and cost.

The overall proportion of patients in each category of management (OMT, PCI, CABG, ‘more information required’) before and after vFFR was similar for both all participants and the CCS subgroup, but there was significant crossover between treatment groups, so this may therefore represent more appropriate treatment allocation (figure 3), allowing for better targeting of effort and expenditure to achieve the best possible results in all patients. This is in accord with previous studies of measured4–7 9 and computationally derived16 FFR, the latter of which showed vFFR changed management decisions in 23% of patients with epicardial disease and is very similar to our finding of an overall change in management for 22% of patients.

The implications of vFFR in the subgroups diverged. In the ACS cohort, vFFR resulted in an 88% relative (6% absolute) reduction in the intended use of invasive FFR and a 63% relative (10% absolute) increase in the proportion assigned to OMT. The mean invasive FFR in this cohort was 0.86, perhaps reflecting the intended use of FFR as a tool to safely defer PCI in those considered for revascularisation.4 These findings complement those of RIPCORD 2, which showed that routine use of invasive FFR at the diagnostic stage (ICA only) is without benefit in selected patients with CCS and ACS.19 Unlike in RIPCORD 2, in which a pressure wire was deployed in all major epicardial vessels (median four), regardless of the presence of disease, in our study, vFFR was only deployed in lesions of uncertain severity, which accords better with normal practice. Therefore, vFFR, which is comparatively inexpensive on a patient-by-patient basis, and devoid of complications, could reap the benefits seen in previous, wire-measured FFR studies.

Somewhat different findings emerged in the CCS cohort. This group underwent cardiac catheterisation in non-PCI hospitals, where vFFR resulted in a 31% relative (8% absolute) increase in the proportion of patients allocated to PCI, despite an overall reduction in the extent of significant CAD. This finding was predominantly a consequence of 23% of medically managed patients being reallocated to single-vessel PCI after vFFR disclosure, 50% of whom had prior evidence of ischaemia on non-invasive testing—the extra evidence of localised ischaemia perhaps tipping the balance in favour of revascularisation in patients with CCS, despite current trends.20–22 Furthermore, in the CCS cohort, there was a 75% relative increase in the rate of invasive FFR referral after vFFR disclosure. Although the planned rate of invasive FFR use was lower in the CCS cohort compared with the ACS cohort, deploying vFFR in this group (in a non-PCI setting) could avoid a second procedure and attendant delays.23 This underlines the importance of education about the use of FFR and the ‘zone of uncertainty’ among non-interventional cardiologists to allow for cost-effective gate-keeping.24

The practicalities of using vFFR were encouraging. It was applicable in ‘real time’ in diverse cardiac catheterisation laboratories and allowed decisions to be made on the day of the procedure, which is an advantage for both the patient and cardiologist. Once deployed, the vFFR failure rate was low (3%) although, at the second screening stage, many ICAs did not provide the necessary lesion clarity in two views in diastole. vFFR was well received by consultants, and this translated into a small but significant increase in confidence in clinical decision-making. It was noticeable that confidence increased in cases in which clinical opinion accorded with the vFFR, and decreased when it differed. Validation of vFFR against invasive FFR values in the ACS cohort confirmed excellent diagnostic accuracy (94%).

First, this was a virtual trial, with changes in intended rather than actual management, and this may have affected the readiness of the cardiologists to propose changes. Second, the design did not provide an opportunity to assess actual clinical outcomes, because VIRTUheart is not clinically approved. Third, there were fewer patients with CCS than ACS, but that accurately reflects current clinical practice. Fourth, the study excluded patients with prior CABG, aorto-ostial and left main stem lesions, severe stenoses and diffuse CAD, all of which are regularly encountered. Fifth, vFFR makes assumptions about microvascular resistance, which can affect accuracy, unlike invasive FFR.25 26 Sixth, the cardiologists caring for the patients with ACS were interventionists and those caring for the patients with CCS were non-interventionists, so handling of the information may have differed. This, however, also reflects clinical pathways in many centres. Seventh, despite the consistency and high success rate of the two experts, their assessment of physiological significance differed in 15% cases, reinforcing doubts about the robustness and repeatability of this technology, especially if used by non-experts. Selection of angiographic views, timing of the selected frames, correction of the arterial outline and setting of the proximal and distal boundaries may all have contributed to this finding.27 Despite this, accuracy of the VIRTUheart system, which personalises microvascular resistance to patient-specific parameters,28 compared well with other systems.15 25