Clinical Utility of Novel Fractional Flow Reserve Pullback for Individual Lesion Contribution in Serial Disease

Objectives. Fractional flow reserve (FFR) pullback is frequently used to assess serially diseased arteries, but has been shown to be inaccurate due to physiological interaction between individual lesions. We evaluated the clinical utility of a novel solution that improves estimation of true FFR contribution of each stenosis in the presence of serial disease. Methods. Ten interventional cardiologists were presented with tiered information for 18 elective patients with serial coronary disease and submitted revascularization strategies and assessment of lesion significance. Operators were first shown clinical and angiographic information only (Angio); then, conventional practice FFR (FFRnorm); and finally, pullback with corrected FFR contributions of each stenosis (FFRpred). Results. The treatment strategy agreement between operators was k=0.39, k=0.64, and k=0.77 using Angio, FFRnorm, and FFRpred, respectively (P<.001). Lesion significance uncertainty was 26%, 28%, and 3%, respectively. The number of stents per patient was 1.49 ± 0.57, 1.50 ± 0.57, and 1.3 ± 0.5, respectively (P<.001). In total, percutaneous coronary intervention (PCI) strategy changed in over 50% of cases analyzed, with participants opting for shorter stent length with FFRpred (29.5 ± 15.2 mm) compared with FFRnorm (34.1 ± 14.4 mm; P<.001) and Angio (34.6 ± 14.3; P=.04). This was accompanied by significantly less interobserver variability. Conclusion. The ability to quantify the contribution of individual lesions with the novel FFR pullback-based solution significantly increases operator confidence regarding PCI strategy, reduces heterogeneity in practice, and can reduce the planned number of stents and total stent length.

J INVASIVE CARDIOL 2021;33(7):E491-E496. Epub 2021 June 7. doi:10.25270/jic/20.00550

Key words: diffuse disease, fractional flow reserve, stable angina

Over the last decade, there has been a growing body of evidence demonstrating improved clinical outcomes in patients with epicardial coronary artery disease (CAD) when revascularization is guided by ischemia and pressure-derived physiological indices.^1–4 However, until recently, it has been unclear whether invasive coronary indices such as fractional flow reserve (FFR) can reliably assess individual lesions in serial CAD due to the hemodynamic interplay between stenoses.⁵ Attempts have been made to address this potential error with complex formulae that incorporate occlusive coronary wedge pressure and other parameters, but the complexity of these techniques has led to limited adoption in clinical practice.^6-8 To overcome this limitation, we developed and validated a mathematical solution that can be applied to routine FFR pullback measurements, without the need to measure coronary occlusive pressure.⁹ While these in vitro and clinical validation studies have demonstrated a reduction in errors and stenosis misclassification when using this novel correction equation,¹⁰ it is unclear whether this will make a difference to routine clinical practice or is purely of academic interest. We therefore sought to assess how the availability of this FFR solution, which incorporates a smoothed FFR pullback trace and allows prediction of effect of individual lesions in isolation, alters revascularization strategy and resource utilization in real-world clinical scenarios of serial CAD.

Ten interventional cardiologists, working at several European sites, were presented with progressive tiers of physiological information from patients with serial CAD who presented for elective percutaneous coronary intervention (PCI). All operators worked in high-volume tertiary PCI centers and routinely performed FFR measurements for intermediate stenosis assessment as per international guidelines.¹¹ Four operators worked in London (United Kingdom) and 6 in Milan (Italy).

Study patient population and physiological data acquisition. Eighteen serially diseased vessels were evaluated and included 11 left anterior descending (LAD) coronary arteries, 4 LAD coronary arteries with left main coronary artery (LMCA) involvement, 2 right coronary arteries, and 1 left circumflex artery with LMCA involvement. Serial CAD was defined by 2 stenoses of at least 30% diameter stenosis separated by at least 10 mm, where the operator would consider treating them separately by PCI. Fourteen patients (78%) were male, and mean age was 64.7 ± 10.7 years. The average diameter stenosis by quantitative coronary angiography (QCA) was 64.6 ± 11.2% and 60.6 ± 8.9% for proximal and distal stenoses, respectively. Mean distal cumulative FFR of the vessels was 0.72 ± 0.10.

Following standard diagnostic angiography, all patients went on to have pressure-wire pullback measurements. This was performed after guidewire pressure sensors were normalized in the aorta before advancement into the distal vessel, with the distal wire position documented fluoroscopically. A resting ratio of distal coronary to aortic pressure (Pd/Pa) was recorded, after which intravenous adenosine was commenced to ensure steady-state hyperemia (140 µg/kg/min); onset of hyperemia was confirmed using invasive pressure waveform changes.¹² A conventional FFR pullback maneuver was then performed whereby the pressure wire was pulled back at a fixed rate from distal to proximal through the serially diseased vessel. Operators were free to manage the patient in keeping with their conventional practice (heart team discussion, PCI using angiographic ± physiological information or medical management).

Offline corrected, smoothed FFR pullback curves were created using our novel mathematical solution (FFRpred) using Matlab software (Mathworks). In order to assure the quality of the trace, artifactual data points giving non-physiologically high or low values were filtered out and then the trace was plotted using the smoothing spline algorithm, which gave a polynomial fit output according to an adjustable smoothing parameter. The output was then assessed for goodness of fit. Figure 1 is an example of typical FFR pullback output. The corrected contributions were derived using our recently developed mathematical solution for serial stenosis interplay, which involves input variables that are confined to data derived solely from pressure-wire pullback.¹³ This simple and easy-to-use solution assumes a linear relationship between pressure and flow across a stenosis so that the hemodynamic equivalent of Ohm’s law can apply, whereby the individual resistance of stenoses and the distal circulation stay fixed regardless of other stenoses being removed. Under these conditions, we have shown that the theoretical FFR can be reliably derived in all configurations of tandem CAD without the need for coronary occlusive pressure, if the variability of collateral flow across intermediate stenoses is assumed to be minimal.¹⁰ 

Assessment of clinical utility. Following the catheter laboratory procedure, 10 operators were shown the following information in sequence for each case (Figure 1):

(1) Angio = clinical and angiographic information only.

(2) FFRnorm = cumulative FFR value (derived with sensor in the vessel beyond the distal stenosis) and a standard FFR pullback trace.

(3) FFRpred = smoothed FFR pullback trace with corrected contributions of each stenosis to the total vessel FFR.

At each stage, operators were asked to assess whether either of the 2 stenoses were significant or not (yes, no, or unsure), state which of the 2 stenoses were more significant (proximal stenosis [1] or distal stenosis [2]), decide revascularization strategy (optimal medical therapy [OMT], coronary artery bypass graft surgery [CABG], or PCI) and, if PCI were planned, to state the predicted total stent length and number of stents that would be implanted.

Statistical analysis. Continuous data are presented as mean ± standard deviation. Categorical data are presented as numbers and percentages and compared using the Chi-square test. In order to determine agreement in lesion assessment between operators (interobserver variability), Fleiss Kappa coefficients were calculated. This coefficient describes the degree of agreement between the 10 operator decisions over that which would be expected by chance. A value of 0 denoted random agreement while a coefficient of 1 demonstrated perfect agreement in all cases. Statistical analysis was performed using IBM SPSS 24.0 for Apple Macintosh (SPSS).

Using angiographic information alone (Angio), operators concluded that 5% of vessels should be managed with OMT, 73% with PCI, and 22% with CABG. With conventional pressure-wire methods (FFRnorm) these proportions changed to 15%, 68%, and 17%, respectively. When presented with the predicted FFR of each stenosis superimposed on a smoothed pullback trace (FFRpred), 17%, 68%, and 15% were planned for OMT, PCI, and CABG, respectively. The agreement about treatment strategy between operators was found to be 0.39, 0.64, and 0.77 with Angio, FFRnorm, and FFRpred, respectively (P<.001).

Assessment of individual lesion significance. When asked about whether a lesion was significant, 26% were unsure using the Angio value alone. Supplementing operators with FFRnorm did not significantly affect their uncertainty (28%). However, when presented with FFRpred, operator uncertainty about the significance of a lesion fell to 3% of all lesions (P<.001) (Figure 2). This was accompanied by a significant reduction in interobserver variability (Table 1).

Prediction of number and length of stents. The operators predicted that the average number of stents per patient would be 1.49 ± 0.57 using Angio alone, 1.50 ± 0.57 with FFRnorm, and 1.3 ± 0.5 with FFRpred (P<.001). Moreover, the interobserver variability of operators for the predicted number of stents was significantly reduced when presented with FFRpred (k=0.55) vs FFRnorm (k=0.26) or Angio (k=0.08) (P<.001). The predicted total length of stent required was significantly different with each tier of information, with a shorter total stent length using FFRpred (29.5 ± 15.2 mm) vs FFRnorm (34.1 ± 14.4 mm; P<.001) and Angio alone (34.6 ± 14.3; P=.04) (Figure 3). Moreover, the interobserver variability was significantly reduced when presented with FFRpred (k=0.93) vs FFRnorm (k=0.85) or Angio (k=0.83) (P<.001). Overall, compared with Angio, FFRpred changed decision making in 68% of vessels, with a reduction in stent length from 34.6 ± 14.2 mm to 29.5 ± 15.2 mm; P=.04. In comparison with conventional FFRnorm-based decision making, FFRpred changed decision making in 60% of cases.

In this study, we demonstrated the clinical utility of a smoothed FFR pullback trace together with corrected pressure gradients for each stenosis, and showed how this novel method can alter PCI strategy in real-world clinical scenarios of serial CAD. The main findings are: (1) FFRpred increased the agreement between operators for treatment strategy in cases of serial coronary artery disease; (2) FFRpred had a significant impact on the adjudication of stenosis severity, with operators feeling more certain about whether a stenosis required revascularization and also which stenosis was more significant; and (3) FFRpred resulted in significantly less stent implantation, with a significant reduction in mean stent length and number.

We previously demonstrated that applying this correction equation (FFRpred) to routine FFR pullback data reduced the error in estimating the true FFR contribution of individual lesions and associated misclassification according to an FFR threshold of 0.80 for revascularization.¹⁰ We have also shown a reduction in relative error and misclassification in comparison with resting physiological indices of instantaneous wave-free ratio (iFR) and resting Pd/Pa.¹⁰ The current study demonstrates the clinical applicability of this mathematical solution through reduced ambiguity and increased operator confidence across 10 separate attending operators who were previously naïve to the novel correction method.

This novel solution requires smoothing of conventional FFR pullback data together with incorporation of the correction equation to allow the identification of the true FFR contribution of each stenosis before PCI (Figure 1). Importantly, for clinical application in PCI planning, FFR pullback can be rapidly performed in nearly all cardiac catheterization laboratories and requires no additional hardware. This is important, as an element of serial/diffuse disease is thought to exist in over 25% of all CAD patients,^14-16 meaning the utilization of such a novel solution should be widespread and easy to adopt. Our FFR pullback-based solution allows the pressure tracing itself to identify the lesions with the greatest influence upon flow impairment, without the need to adopt a new physiological index, use new hardware, or measure coronary occlusive pressures. An update to existing pressure-wire consoles would be sufficient to incorporate this solution by displaying a smoothed Pd/Pa pullback curve together with the true FFR contribution of a given stent location and the residual FFR in the vessel, with the rest of the disease left alone. Such an update has been investigated, and an example of clinical utilization in a case of serial CAD is shown in Figure 4.

This is an important improvement in FFR methodology, and is likely to enable wider and more accurate adoption of FFR for complex CAD with serial/diffuse disease. In addition, this study goes on to describe significant clinical utility of an even greater magnitude than that described from the iFR gradient registry of iFR pullback in serial CAD.¹⁶ The novel method now needs prospective assessment vs contemporary methods (including iFR pullback) to see whether it improves clinical outcomes and resource utilization in a larger, multicenter patient cohort.

Study limitations. First, in this utility study, operators outlined treatment strategy based on tiered data; therefore, it is possible that revealing FFRnorm may have led to bias of FFRpred interpretation. However, the reduction in interoperator variability demonstrated with FFRpred indicates this is unlikely. Second, our study had a small sample size; however, this was overcome with the large number of lesions each operator evaluated and was able to demonstrate statistical significance. Finally, this is a virtual study; hence, the accuracy of FFRpred needs to be evaluated in vitro. 

In patients with CAD, hemodynamic interplay between serial lesions is a source of error and stenosis misclassification. Our study shows the clinical utility of a novel method to correct routine FFR pullback output by demonstrating increased operator certainty regarding revascularization and PCI strategy, together with a significant increase in agreement between interventional cardiologists regarding the physiological assessment of serial CAD.