Author + information
- Received July 8, 2013
- Revision received September 22, 2013
- Accepted September 22, 2013
- Published online April 8, 2014.
- Allen Jeremias, MD, MSc∗,†∗ (, )
- Akiko Maehara, MD†,‡,
- Philippe Généreux, MD†,‡,§,
- Kaleab N. Asrress, MA, BM BCh‖,
- Colin Berry, MBChB, PhD¶,
- Bernard De Bruyne, MD#,
- Justin E. Davies, MBBA∗∗,
- Javier Escaned, MD††,
- William F. Fearon, MD‡‡,
- K. Lance Gould, MD§§,
- Nils P. Johnson, MD, MS§§,
- Ajay J. Kirtane, MD, SM†,‡,
- Bon-Kwon Koo, MD‖‖,
- Koen M. Marques, MD, PhD¶¶,
- Sukhjinder Nijjer, MBBA∗∗,
- Keith G. Oldroyd, MBChB, MD¶,
- Ricardo Petraco, MD∗∗,
- Jan J. Piek, MD##,
- Nico H. Pijls, MD∗∗∗,
- Simon Redwood, MD‖,
- Maria Siebes, PhD##,
- Jos A.E. Spaan, PhD##,
- Marcel van 't Veer, MSc, PhD∗∗∗,
- Gary S. Mintz, MD†,‡ and
- Gregg W. Stone, MD†,‡
- ∗Division of Cardiovascular Medicine, Stony Brook University Medical Center, Stony Brook, New York
- †Cardiovascular Research Foundation, New York, New York
- ‡Department of Medicine, Columbia University Medical Center, New York, New York
- §Department of Medicine, Hôpital du Sacré-Coeur de Montréal, Montreal, Quebec, Canada
- ‖Cardiovascular Division, British Heart Foundation Centre of Research Excellence, King's College London, St. Thomas' Hospital, London, United Kingdom
- ¶West of Scotland Regional Heart and Lung Centre, Golden Jubilee National Hospital, Glasgow, Scotland, United Kingdom
- #Cardiovascular Center Aalst, OLV Clinic, Aalst, Belgium
- ∗∗International Centre for Circulatory Health, National Heart and Lung Institute, Imperial College London, London, United Kingdom
- ††Hospital Clinico San Carlos, Madrid, Spain
- ‡‡Division of Cardiovascular Medicine, Stanford University Medical Center, Stanford, California
- §§Weatherhead PET Center for Preventing and Reversing Atherosclerosis, Division of Cardiology, Department of Medicine, University of Texas Medical School and Memorial Hermann Hospital, Houston, Texas
- ‖‖Seoul National University Hospital, Seoul, Republic of Korea
- ¶¶Department of Cardiology, VU University Medical Center, Amsterdam, the Netherlands
- ##Departments of Cardiology and Biomedical Engineering and Physics, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands
- ∗∗∗Department of Cardiology, Catharina Ziekenhuis, Eindhoven, the Netherlands
- ↵∗Reprint requests and correspondence:
Dr. Allen Jeremias, Division of Cardiology, Department of Medicine, Health Sciences Center, T16-080, Stony Brook, New York 11794-8160.
Objectives This study sought to examine the diagnostic accuracy of the instantaneous wave-free ratio (iFR) and resting distal coronary artery pressure/aortic pressure (Pd/Pa) with respect to hyperemic fractional flow reserve (FFR) in a core laboratory–based multicenter collaborative study.
Background FFR is an index of the severity of coronary stenosis that has been clinically validated in 3 prospective randomized trials. iFR and Pd/Pa are nonhyperemic pressure-derived indices of the severity of stenosis with discordant reports regarding their accuracy with respect to FFR.
Methods iFR, resting Pd/Pa, and FFR were measured in 1,768 patients from 15 clinical sites. An independent physiology core laboratory performed blinded off-line analysis of all raw data. The primary objectives were to determine specific iFR and Pd/Pa thresholds with ≥90% accuracy in predicting ischemic versus nonischemic FFR (on the basis of an FFR cut point of 0.80) and the proportion of patients falling beyond those thresholds.
Results Of 1,974 submitted lesions, 381 (19.3%) were excluded because of suboptimal acquisition, leaving 1,593 for final analysis. On receiver-operating characteristic analysis, the optimal iFR cut point for FFR ≤0.80 was 0.90 (C statistic: 0.81 [95% confidence interval: 0.79 to 0.83]; overall accuracy: 80.4%) and for Pd/Pa was 0.92 (C statistic: 0.82 [95% confidence interval: 0.80 to 0.84]; overall accuracy: 81.5%), with no significant difference between these resting measures. iFR and Pd/Pa had ≥90% accuracy to predict a positive or negative FFR in 64.9% (62.6% to 67.3%) and 48.3% (45.6% to 50.5%) of lesions, respectively.
Conclusions This comprehensive core laboratory analysis comparing iFR and Pd/Pa with FFR demonstrated an overall accuracy of ∼80% for both nonhyperemic indices, which can be improved to ≥90% in a subset of lesions. Clinical outcome studies are required to determine whether the use of iFR or Pd/Pa might obviate the need for hyperemia in selected patients.
Fractional flow reserve (FFR) is an index of the hemodynamic significance of a coronary stenosis that is calculated directly from measurements of hyperemic pressure (1,2). The physiological basis of FFR has been extensively validated in animal and human studies, and FFR shows good correlation to noninvasive ischemia testing with perfusion scintigraphy (3) and positron emission tomography (4). FFR has been shown in 3 randomized trials to identify coronary stenoses that will benefit from early revascularization (those with a positive FFR) (5) and conversely those lesions with a negative FFR for which revascularization may be safely deferred (6,7). To measure FFR, a vasodilator (most commonly intravenous or intracoronary adenosine) is administered to minimize microvascular resistance and the effect of resting hemodynamics such that coronary pressure becomes proportional to myocardial flow.
Interest has recently emerged as to whether 2 nonhyperemic measures of pressure might be useful to assess the severity of coronary stenosis. Pd/Pa is the ratio of distal coronary artery pressure to aortic pressure over the entire cardiac cycle. Conversely, the instantaneous wave-free ratio (iFR) measures coronary pressure during a specific period of diastole when resting resistance is the lowest (8). By reducing procedural time and cost, avoiding patient-related discomfort from pharmacological hyperemia, and allowing continuous online measurements (thereby facilitating multivessel interrogation), assessment of the severity of coronary stenosis without induction of hyperemia is intuitively appealing, provided diagnostic accuracy is preserved. However, in prior reports, the diagnostic accuracy of iFR compared with FFR has ranged widely from 60% to 91% (8–11), and its relative accuracy compared with Pd/Pa has been debated. Previous comparative studies to date have been limited by different study methodologies, modest sample sizes, and the use of different algorithms to calculate iFR. Given these conflicting reports, we formed a collaborative group of investigators to perform a large-scale, physiology core laboratory–based analysis with standardized methods to compare the diagnostic accuracy of iFR and Pd/Pa with respect to FFR as the reference standard and to determine the proportion of patients in whom the accuracy of iFR and Pd/Pa is at least 90%.
Patient population and study inclusion criteria
The present investigation was an international, multicenter, nonrandomized, retrospective, core laboratory–based analysis in patients with coronary artery disease undergoing physiological lesion assessment by FFR, iFR, and Pd/Pa. The principal investigators representing all of the published iFR/FFR comparative studies agreed to collaboratively participate in this effort, including the ADVISE (ADenosine Vasodilator Independent Stenosis Evaluation) study and registry (8,11), VERIFY (VERification of Instantaneous Wave-Free Ratio and Fractional Flow Reserve for the Assessment of Coronary Artery Stenosis Severity in EverydaY Practice) (9), and Johnson et al. (10). In addition, 6 other study sites contributed unpublished data to the analysis. All studies included in this analysis were approved by the institutional review boards of the individual sites. Original raw phasic pressure waveforms from each patient were submitted digitally to the Physiology Core Laboratory at the Cardiovascular Research Foundation (New York, New York) for independent off-line analysis. In addition, selected baseline patient demographic and procedural data were supplied to the core laboratory. This study was an investigator-sponsored study by the Cardiovascular Research Foundation and was supported by funding from Volcano Corp. (San Diego, California). The funding source was uninvolved with the design of the protocol and the analysis and interpretation of the study results.
Patients with stable angina, unstable angina, or non–ST-segment elevation myocardial infarction undergoing coronary angiography with or without percutaneous coronary intervention in whom FFR of a single stenosis in a major epicardial coronary artery was performed during the procedure were considered for inclusion in the study. Two or more lesions could be present in a single patient if in different epicardial vessels. Exclusion criteria included left main disease, heart failure as defined by New York Heart Association class III or IV, respiratory failure requiring intubation or supplementary oxygen, cardiogenic shock, significant arrhythmia precluding waveform analysis (e.g., excessive premature ventricular contractions or atrial fibrillation), and tachycardia with a heart rate >120 beats/min.
Pressure measurements and analysis
Physiological measurements of coronary stenoses were performed according to existing study protocols. The RadiAnalyzer Xpress instrument with the Certus coronary pressure wire (St. Jude Medical, Uppsala, Sweden), the Volcano s5 imaging system with the PrimeWire (Volcano Corp., Rancho Cordova, California), or earlier-generation equipment from these manufacturers was used for measurements of coronary pressure. After the pressure sensor was zeroed and equalized to aortic pressure, it was positioned at least 5 mm distal to the stenosis and a recording of the baseline distal coronary and aortic pressures was obtained. After the administration of intracoronary nitroglycerine as per the operators' discretion, hyperemia was induced by the administration of either intravenous adenosine at a dosage of 140 μg/kg/min or intracoronary adenosine at various doses and FFR was calculated. All pressure tracings were submitted directly to the Cardiovascular Research Foundation physiology core laboratory for analysis.
FFR is the ratio of mean distal coronary pressure (Pd) to mean aortic pressure (Pa) during maximum hyperemia. The Pd signal is obtained from a guidewire with a piezoresistive pressure transducer, and the Pa signal is obtained from a fluid-filled guiding or diagnostic catheter. FFR is taken as the lowest stable value of the Pd/Pa ratio during maximal hyperemia. To ensure accuracy of the analysis, waveform analysis of all pressure tracings was performed to confirm that none of the following exclusion criteria were present: significant arrhythmia that may preclude appropriate waveform analysis, loss of Pa or Pd pressure signal at any point during the run apart from intracoronary vasodilator administration, inappropriate recording of Pa or Pd (e.g., only a flat signal is present at some point during the recording), dampened Pa or Pd waveform, reversed gradient during hyperemia (i.e., Pd pressure signal elevated above Pa, resulting in an FFR >1.00), or sensor drift defined as FFR ≤0.97 or ≥1.03 after pullback of the pressure wire transducer into the guiding catheter. In addition to the waveform analysis, the FFR recording had to have an adequate baseline tracing before administration of adenosine. Specifically, a minimum of 5 waveforms of uninterrupted recording adequate for analysis without significant artifact of the tracing was required. FFR was calculated independently from the original readout as the lowest artifact-free Pd/Pa during maximal hyperemia.
iFR is the ratio of Pd/Pa measured during a pre-specified period in mid to late diastole of the cardiac cycle without hyperemia (8). The onset of diastole was identified from the dicrotic notch, and the diastolic window was calculated beginning 25% into diastole and ending 5 ms before end diastole. iFR was calculated off-line in the core laboratory using the Volcano Harvest software package, which contains the iFR computational algorithm developed at the Imperial College of London (8). All analyses were performed in a fully automated manner, eliminating the need for manual selection of data time points. This automated analysis is based on a synchronized electrocardiographic (ECG) signal to determine the appropriate diastolic intervals for pressure measurements. If the ECG signal was missing, the core laboratory manually inserted R-wave markings based on the pressure waveform into the baseline tracing from which iFR was calculated.
Resting Pd/Pa was calculated in similar fashion to iFR except that Pd/Pa was time averaged over the entire cardiac cycle, thus including both systole and diastole. In addition to the exclusion criteria for measurement of FFR, iFR and Pd/Pa recordings with any of the following characteristics were excluded from the analysis: insufficient baseline recording before administration of adenosine (recording had to contain at least 5 cardiac cycles from the start of the recording to the onset of hyperemia), significant arrhythmias including supraventricular tachycardia or premature ventricular contractions within the baseline tracing, or heart rate <50 or >120 beats/min.
Core laboratory analyses were performed in a blinded fashion at 3 separate workstations by different technicians in sequential, independent phases. First, a thorough waveform analysis was performed of all baseline and hyperemic tracings, and pressure recordings meeting any of the previously outlined exclusion criteria were removed from the analysis. Second, an independent calculation of FFR was performed blinded to the original FFR readout. Third, fully automated, computerized calculations of Pd/Pa and iFR were performed by a physician unaware of the waveform analysis and computation of FFR. All tracings were over-read by a physician experienced in physiology measurements (A.M., P.G., or A.J.) to ensure data quality. FFR, iFR and Pd/Pa data were recorded on separate case report forms that were not merged until the completion of the blinded analyses.
The primary objective of this study was to evaluate the level of diagnostic accuracy of iFR and Pd/Pa compared with FFR in a variety of clinical settings in the largest population studied to date using rigorous, pre-specified core laboratory–based processes. Using FFR as the reference standard, the primary endpoint of the study was to identify the iFR thresholds that most strongly correlated with an FFR cut point of 0.80 and to determine the proportion of lesions for which these thresholds applied. Thresholds with ≥90% diagnostic accuracy were calculated (pre-specified as representing the minimal thresholds required for the potential clinical utility of iFR), and the proportions of lesions that fell beyond those thresholds were determined (defined as the adenosine-free zone).
Secondary study objectives included determining the iFR thresholds necessary to achieve >90% to 99% diagnostic accuracy, construction of receiver-operating characteristic (ROC) curves for iFR to assess the optimal cutoff point with respect to the clinical threshold of FFR ≤0.80, assessment of the overall correlation between iFR and FFR using regression techniques, and assessment of the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall diagnostic accuracy at that cutoff point. All of the preceding analyses were also performed with the cycle-averaged resting pressure ratio Pd/Pa, and iFR and Pd/Pa were directly compared with respect to their diagnostic accuracy. In addition, subgroup analyses were performed for both iFR and Pd/Pa with respect to coronary vessel (left anterior descending [LAD] coronary artery vs. non-LAD), route of adenosine administration (intravenous vs. intracoronary), and study site (to assess center variability).
Data are summarized by descriptive statistics. Pearson's correlation and linear regression analysis were performed to examine the relationship between iFR and FFR and Pd/Pa and FFR, respectively. ROC curves were constructed to identify the concordance between FFR, iFR, and Pd/Pa. Agreement between the methods was assessed by Bland-Altman plots with corresponding 95% limits of agreement. Sensitivity, specificity, PPV, NPV, and overall diagnostic accuracy of measurement of iFR and Pd/Pa relative to an FFR cutoff of ≤0.80 were determined, C statistics were generated, and optimal cutoff values for iFR and Pd/Pa were computed based on maximizing the sum of sensitivity plus specificity. Binary variables were compared using chi-square testing. From the raw data examining the relationship between iFR (or Pd/Pa) and FFR, separate iFR (Pd/Pa) thresholds were determined for which the PPV and NPV were each ≥90% (corresponding to an FFR of ≤0.80 and >0.80, respectively), and the proportion of lesions meeting these criteria was determined. Similar analyses were performed using different thresholds from ≥90% to ≥99%. SAS software version 9.1 (SAS Institute, Cary, North Carolina) was used for all analyses, and a 2-tailed p value of <0.05 was regarded as statistically significant.
Patient demographics and procedural data
A total of 1,768 patients with 1,974 lesions from 15 clinical sites were submitted for analysis. Of these lesions, 381 (19.3%) met at least one of the pre-defined core laboratory exclusion criteria, leaving 1,593 lesions for final analysis. The most common reasons for exclusion were insufficient baseline recording or artifact during recording (n = 227), lesions not meeting study entry criteria (n = 56), pressure drift or incorrect calibration (n = 42), and other technical factors (n = 56).
The mean age of the population was 63.4 ± 10.3 years, and 74.9% were male. A total of 21.2% had prior myocardial infarction and 28.1% had diabetes mellitus, and 29.4% were current smokers. A small fraction had prior coronary artery bypass grafting (3.4%), chronic kidney disease (8.3%), and congestive heart failure (6.3%). The clinical presentation was most commonly chronic stable angina (68.6%), with 14.4% having unstable angina and 8.4% non–ST-segment elevation myocardial infarction. More than half of the population had multivessel coronary artery disease (53.8%); the LAD coronary artery was the most commonly interrogated target lesion (63%), followed by the right coronary artery (20%) and the left circumflex artery (17%). FFR studies were performed with intravenous adenosine in 80.1% of cases and intracoronary adenosine in the remainder.
Relationships between FFR, iFR, and Pd/Pa
The median (interquartile range) FFR, iFR, and Pd/Pa in this study population was 0.79 (0.70 to 0.86), 0.90 (0.83 to 0.95), and 0.93 (0.86 to 0.96), respectively. A scatter plot between iFR and FFR is shown in Figure 1A, demonstrating moderate overall linear correlation between the 2 measures, with an R2 of 0.66 (95% confidence interval [CI]: 0.64 to 0.70) (p < 0.001). Similarly, the correlation of resting Pd/Pa and FFR demonstrated an R2 of 0.69 (95% CI: 0.67 to 0.72) (p < 0.001) (Fig. 1B). Although the overall correlations between Pd/Pa versus FFR and iFR versus FFR were similar, the data points were more clustered around the regression line with a flatter slope and greater intercept for the Pd/Pa versus FFR relationship. The area under the ROC curve (C statistic) to predict an FFR ≤0.80 was 0.81 (95% CI: 0.79 to 0.83) for iFR and 0.82 (95% CI: 0.80 to 0.84) for Pd/Pa, indicating moderate to good discrimination for both (Fig. 2). The optimal cutoff value for an FFR ≤0.80 derived from ROC analyses was 0.90 for iFR and 0.92 for Pd/Pa.
Bland-Altman plots for iFR and Pd/Pa are shown in Figures 3A and 3B, respectively. On average, iFR exceeded FFR by +0.10 (95% CI: −0.06 to +0.26) and Pd/Pa exceeded FFR by +0.14 (−0.01 to +0.29). However, both iFR and Pd/Pa demonstrated a substantial degree of scatter, particularly below the threshold of 0.80.
The correlation between iFR and Pd/Pa is shown in Figure 4A. There was a strong correlation between these 2 parameters (R2 = 0.95; p < 0.001), showing that 95% of the variation in iFR was accounted for by Pd/Pa. However, Bland-Altman analysis showed that Pd/Pa overestimates iFR on average by 0.04 and substantially more when the iFR is <0.80 (Fig. 4B).
Diagnostic accuracy of iFR
The overall sensitivity, specificity, PPV, and NPV for iFR ≤0.90 versus FFR ≤0.80 was 78.9%, 82.4%, 85.2%, and 73.3%, respectively, with an overall diagnostic accuracy of 80.4%. To achieve ≥90% diagnostic accuracy at each extreme, the overall iFR range had to be restricted to ≤0.88 (to predict an FFR ≤0.80) and ≥0.97 (to predict an FFR >0.80), comprising 1,034 of the 1,593 study lesions (64.9%). In other words, if a ≥90% diagnostic accuracy compared with FFR is deemed sufficient for therapeutic interchangeability, 64.9% (95% CI: 62.6% to 67.3%) of the study lesions would fall within the adenosine-free zone and not require hyperemia for the diagnosis of ischemia. Figure 5 demonstrates the association between the adenosine-free zone and diagnostic accuracy. The adenosine-free zone narrows as increasing diagnostic accuracy of iFR is required, such that only 28.6% (26.4% to 30.8%) and 18.0% (16.1% to 19.8%) of lesions would achieve ≥95% and ≥99% diagnostic accuracy, respectively.
Diagnostic accuracy of Pd/Pa
The sensitivity, specificity, PPV, and NPV for a Pd/Pa ≤0.92 for an FFR ≤0.80 was 76.3%, 88.1%, 89.2%, and 74.4%, respectively, resulting in an overall diagnostic accuracy of 81.5%. A diagnostic accuracy of ≥90% was achieved when the Pd/Pa range was restricted to ≤0.92, with 769 of the 1,593 (48.3%; 45.6% to 50.5%) lesions falling in that range. However, in contrast to iFR, there was no upper boundary of Pd/Pa that predicted with ≥90% accuracy a negative FFR value (i.e., >10% of lesions with a Pd/Pa of 1.00 had an FFR ≤0.80). Figure 5 shows the association between the adenosine-free zone and diagnostic accuracy for Pd/Pa. Similar to iFR, there was a trade-off between diagnostic accuracy and the size of the adenosine-free zone. Only 36.0% (33.7% to 38.4%) and 19.5% (17.5% to 21.4%) of lesions would achieve a diagnostic accuracy of ≥95% and ≥99%, respectively.
There was no significant difference in the diagnostic accuracy of iFR compared with FFR with intravenous versus intracoronary administration of adenosine (81.5% vs. 78.2%; p = 0.07) or among patients presenting with stable versus unstable angina (80.4% vs. 80.2%; p = 0.97). Similarly, no significant differences in diagnostic accuracy were noted when LAD coronary artery stenoses were compared with non-LAD coronary artery stenoses (79.9% vs. 81.9%; p = 0.34) or for tracings with versus without an embedded ECG signal (83.7% vs. 80.2%; p = 0.39). Finally, the variation in overall accuracy between iFR and FFR at individual study sites ranged from 78.6% to 82.7%, and the correlation varied from an R2 of 0.54 to an R2 of 0.72 (Table 1). For Pd/Pa, the overall accuracy ranged from 72.6% to 89.5% and the correlation varied from an R2 of 0.61 to an R2 of 0.75.
In this large, core laboratory–based analysis, the overall linear correlation between both iFR and Pd/Pa and FFR was moderate (R2 = 0.66 and 0.69, respectively), with an overall diagnostic accuracy of ∼80% for both nonhyperemic indices (using the optimal ROC-determined cutoff points of 0.90 and 0.92 to predict an FFR ≤0.80). The diagnostic accuracy was independent of vessel, embedded versus core laboratory–generated ECG gating signal, use of intravenous versus intracoronary adenosine to induce hyperemia, and clinical site. Accepting FFR as the reference method (in the absence of outcome studies with iFR or Pd/Pa), this level of accuracy is insufficient to use either parameter for procedural guidance in all cases because ∼20% of therapeutic decisions would be discordant from FFR.
Although iFR and Pd/Pa are imperfect surrogates of FFR close to the clinically used cutoff value of 0.80 (11), they may still provide acceptable accuracy at greater or lesser degrees of functional stenosis severity. The fundamental principle of FFR, justifying pressure-derived estimation of coronary flow impairment, is that the translesional pressure ratio approximates flow when microvascular resistance is minimized (12,13), requiring the use of a potent vasodilator. However, microvascular resistance is influenced by many factors, including capacitive, inertial, and resistive forces as well as the complex effects of systolic contraction. Nonhyperemic pressure ratios may theoretically have adequate concordance with hyperemic pressure measurements when there is a large baseline gradient (i.e., obvious impairment of coronary flow) or no gradient at all (i.e., absence of any resting flow disturbance). In this regard, a recent retrospective analysis of almost 500 patients demonstrated a good correlation between Pd/Pa and FFR with an area under the curve of 0.86 (14). When only translesional resting pressure ratios of <0.88 and >0.95 were considered, the PPV and NPV increased to >95%, with more than half of the study population falling in these categories. The present larger, multicenter, core laboratory–based analysis shows that if 90% accuracy compared with the FFR reference standard is accepted at the margins (the pre-specified precision limit for therapeutic interchangeability in the present study), use of iFR and Pd/Pa might avoid hyperemia in 65% and 48% of lesions, respectively. If 95% accuracy is required, however, use of iFR and Pd/Pa might avoid hyperemia in only 29% and 36% of lesions, respectively. In addition, the percentage of lesions falling into the adenosine-free zone will vary based on the spectrum of lesions being studied. If only intermediate lesions are investigated (i.e., with an FFR near 0.80 in a greater proportion of patients), the adenosine-free zone may be smaller compared with the findings of the current study.
A secondary goal of the present study was to compare and contrast iFR and Pd/Pa. By restricting measurements to a specific segment of diastole in which the maximum achievable coronary flow occurs during resting conditions, iFR has a theoretical advantage compared with Pd/Pa. However, using FFR as the reference standard, we found no significant differences between iFR and Pd/Pa with respect to sensitivity, specificity, PPV, NPV, or diagnostic accuracy. Although modest differences were noted between the iFR and Pd/Pa versus FFR regression patterns, the overall similar results are consistent with a prior retrospective analysis by Johnson et al. (10). Prospective studies are required to determine whether the differences between iFR and Pd/Pa are practically or clinically relevant.
The present study has several strengths but also some limitations. Prior studies examining the relationship between iFR, Pd/Pa, and FFR showed significant variability and thus reached strikingly different conclusions (8–10). In this regard, it is reassuring to note that by applying a rigorous study methodology, common inclusion and exclusion criteria, and a standardized physiology assessment methodology, the data from these prior studies showed relatively little variation, with diagnostic accuracy ranging from 79% to 83%. We have applied linear models to our data, although the complete physiological relationship between FFR and iFR or rest Pd/Pa may best be described by a curvilinear relationship. RESOLVE is the first coronary physiology study that used a core laboratory for analysis of hyperemic and resting pressure–derived indices of the severity of stenosis. Surprisingly, 19% of measurements were found to be suboptimal and were excluded from the analysis (perhaps explaining the reduced site-to-site variability in the present report compared with previously reported individual studies). Future clinical trials should consider including core laboratory analysis to assess the validity of hemodynamic measurements, as is currently the standard for quantitative coronary angiography and intravascular ultrasonography. An additional strength is the size of the present study, encompassing all iFR studies published to date as well as several nonpublished clinical experiences, which provides incremental power to accurately locate point estimates while reducing CI width and affording subgroup analysis. However, the present retrospective analysis is limited by nonuniform patient and lesion characteristics at each site and varying FFR acquisition protocols. Despite the fact that all studies underwent rigorous analysis by an independent core laboratory to eliminate potential erroneous measurements and minimize variability, we cannot fully exclude selection bias and other sources of inconsistencies. A final pullback of the pressure wire into the guiding catheter confirming the absence of pressure drift was not required and was performed in only a small minority of cases.
As with any diagnostic test FFR, iFR and Pd/Pa have inherent variability (9,15,16). On the basis of 3 randomized trials showing superior clinical outcomes with FFR guidance compared with angiographic guidance alone (5–7), FFR is justifiably accepted as the standard in both US and European guidelines for invasive physiological lesion assessment and clinical decision making (17,18). On the basis of the present report and consistent with prior studies (9,10), the universal adoption of iFR or Pd/Pa with use of a single cutoff point cannot be recommended (19). However, using a hybrid approach wherein Pd/Pa or iFR are accepted at the 2 outer tails of the spectrum with FFR-based decisions required in the gray area in between (20) may be feasible and might avoid the use of hyperemia in approximately 48% to 65% of lesions, respectively, if ≥90% correlation with an FFR cutoff ≤0.80 is accepted. Although there will always be a trade-off for greater diagnostic accuracy (e.g., if >99% accuracy compared with FFR is desired, the adenosine-free zone would shrink to <20% of patients), a small (≤10%) degree in variability between nonhyperemic physiological measurements and FFR in a large proportion of patients may be acceptable to many physicians in daily clinical practice given the cost, inconvenience, and potential side effects associated with administration of adenosine (21,22) and the relatively low major adverse cardiac event rate around the FFR 0.80 cut point (5), where most classification errors are likely to occur. However, the iFR and Pd/Pa cutoff values identified in the present retrospective study require validation, and prospective randomized trials are required to determine whether a hybrid strategy results in noninferior clinical outcomes to the routine use of FFR.
The authors thank Liang Dong, MD, Lin Wang, MD, Shinji Inaba, MD, PhD, Shigeo Saito, MD, Tomotaka Dohi, MD, PhD, Nobuaki Kobayashi, MD, PhD, and Elias Sanidas, MD, PhD, for their invaluable efforts with data analysis.
Dr. Jeremias has served as a consultant and member of the Speakers' Bureau for Volcano Corp. Dr. Berry has served as a consultant for and received a research grant from St. Jude Medical. Dr. De Bruyne has received institutional consulting fees from St. Jude Medical. Dr. Davies has received study support from and served as a consultant with licensed intellectual property for Volcano Corp. Dr. Escaned has served as a member of the Speakers' Bureau for St. Jude Medical and Volcano Corp. Dr. Fearon has received research support from St. Jude Medical. Dr. Gould holds a nonfinancial, mutual nondisclosure agreement with Volcano Corp. and is a 510(k) applicant for cfrQuant, a software package for quantifying absolute flow using cardiac PET; all royalties will go to a University of Texas scholarship fund and the University of Texas has a commercial, nonexclusive agreement with Positron Corporation to distribute and market cfrQuant in exchange for royalties; however, Dr. Gould retains the ability to distribute cost-free versions to selected collaborators for research. Dr. Johnson holds a nonfinancial, mutual nondisclosure agreement with Volcano Corp. Dr. Koo has received honorarium and a research grant from St. Jude Medical. Dr. Oldroyd has served as a member of the Speaker's Bureau for St. Jude Medical and Volcano Corp. Dr. Piek has served as a consultant for MAB Abbott Vascular and Miracor. Dr. Pijls has served as a consultant for St. Jude Medical; received institutional research grants from St. Jude Medical; and served as an advisory board member for Heart Flow. Dr. Mintz has served as a consultant and received grant support from Volcano Corp. Dr. Stone has served as a consultant for Volcano Corp., InfraReDx, and Boston Scientific. This study was an investigator-sponsored study by the Cardiovascular Research Foundation and was supported by funding from Volcano Corp. (San Diego, California). The funding source was uninvolved with the design of the protocol and the analysis and interpretation of the study results. All other authors have reported that they have no relationships relevant to the content of this paper to disclose.
- Abbreviations and Acronyms
- confidence interval
- fractional flow reserve (hyperemic by definition)
- instantaneous wave-free ratio (nonhyperemic)
- left anterior descending
- negative predictive value
- distal coronary artery pressure/aortic pressure (nonhyperemic)
- positive predictive value
- receiver-operating characteristic
- Received July 8, 2013.
- Revision received September 22, 2013.
- Accepted September 22, 2013.
- 2014 American College of Cardiology Foundation
- Pijls N.H.,
- Van Gelder B.,
- Van der Voort P.,
- et al.
- De Bruyne B.,
- Baudhuin T.,
- Melin J.A.,
- et al.
- Pijls N.H.,
- van Schaardenburgh P.,
- Manoharan G.,
- et al.
- Sen S.,
- Escaned J.,
- Malik I.S.,
- et al.
- Berry C.,
- van 't Veer M.,
- Witt N.,
- et al.
- Johnson N.P.,
- Kirkeeide R.L.,
- Asrress K.N.,
- et al.
- Pijls N.H.,
- van Son J.A.,
- Kirkeeide R.L.,
- et al.
- Kern M.J.,
- Lerman A.,
- Bech J.W.,
- et al.
- Spaan J.A.,
- Piek J.J.,
- Hoffman J.I.,
- Siebes M.
- Levine G.N.,
- Bates E.R.,
- Blankenship J.C.,
- et al.
- Wijns W.K.P.,
- Danchin N.,
- Di Mario C.,
- et al.
- Samady H.,
- Gogas B.D.
- Cerqueira M.D.,
- Verani M.S.,
- Schwaiger M.,
- Heo J.,
- Iskandrian A.S.
- Kleiman N.S.