Author + information
- Received September 7, 2017
- Revision received October 26, 2017
- Accepted October 29, 2017
- Published online December 18, 2017.
- Marcel van't Veer, MSc, PhDa,b,∗ (, )
- Nico H.J. Pijls, MD, PhDa,b,
- Barry Hennigan, MB BCh BAO, BMedScic,d,
- Stuart Watkins, MBChB, MDc,d,
- Ziad A. Ali, MD, PhDe,
- Bernard De Bruyne, MD, PhDf,
- Frederik M. Zimmermann, MDa,
- Lokien X. van Nunen, MD, PhDa,
- Emanuele Barbato, MD, PhDf,g,
- Colin Berry, MBChB, PhDc,d and
- Keith G. Oldroyd, MDc,d
- aDepartment of Cardiology, Catharina Hospital Eindhoven, Eindhoven, the Netherlands
- bDepartment of Biomedical Engineering, Eindhoven University of Technology, Eindhoven, the Netherlands
- cBritish Heart Foundation, Glasgow Cardiovascular Research Centre, University of Glasgow, Glasgow, Scotland, United Kingdom
- dWest of Scotland Heart and Lung Center, Golden Jubilee National Hospital, Clydebank, Scotland, United Kingdom
- eColumbia University Medical Center, New York, New York
- fCardiovascular Center Aalst, Aalst, Belgium
- gDepartment of Advanced Biomedical Science, University of Naples, Naples, Italy
- ↵∗Address for correspondence:
Dr. Marcel van't Veer, Department of Cardiology, Catharina Hospital Eindhoven, Michelangelolaan 2, 5623 EJ Eindhoven, the Netherlands.
Background Pressure measurement for the duration of the wave-free period (WFP) is considered essential for resting-state physiological assessment of coronary stenosis severity using the instantaneous wave-free ratio (iFR).
Objectives The aim of this study was to compare other diastolic resting indexes to iFR.
Methods In the population of the VERIFY2 (Pd/Pa vs iFR in an Unselected Population Referred for Invasive Angiography) study, iFR calculated by proprietary software (Volcano Harvest, Volcano Corporation, Rancho Cordova, California) was compared with the ratio of resting distal coronary pressure and aortic pressure during the complete duration of diastole (dPR), 25% to 75% of diastole (dPR25–75), and midpoint of diastole (dPRmid), along with Matlab calculated iFR (iFRmatlab) and iFR-like indexes shortening the length of the WFP by 50 and 100 ms (iFR−50ms and iFR−100ms), respectively. Mutual differences, Spearman correlations, area under the curve values from receiver-operating characteristic analyses, and diagnostic performance with respect to iFR and fractional flow reserve (FFR) were calculated for all indexes.
Results Median iFR in 197 patients with 257 vessels was 0.91 with an interquartile range of 0.87 to 0.95. The mutual differences (± SD) with iFR were 0.006 ± 0.011 (dPR), 0.001 ± 0.007 (dPR25–75), 0.001 ± 0.008 (dPRmid), 0.005 ± 0.009 (iFRmatlab), 0.003 ± 0.008 (iFR−50ms), and 0.001 ± 0.009 (iFR−100ms). Correlations for all indexes with iFR were >0.99 (p < 0.001 for all). Area under the curve values for predicting iFR were >0.99 for all indexes as well. Diagnostic accuracy compared with FFR was 76% to 77% for all indexes including iFR.
Conclusions All diastolic resting indexes tested were identical to iFR, both numerically and with respect to their agreement with FFR. A numerically equal value to iFR can be determined without restriction to the WFP. Cutoff values, guidelines, and clinical recommendations for iFR can therefore be extended to these other indexes. (Pd/Pa vs iFR in an Unselected Population Referred for Invasive Angiography [VERIFY2]; NCT02377310)
Assessment of the physiological severity of coronary artery disease by fractional flow reserve (FFR) is superior in guiding coronary revascularization compared with angiography-based strategies and is recommended in European and American guidelines (1,2) based on randomized controlled clinical outcome trials (3–5). Measurement of FFR requires administration of a vasodilatory drug (commonly adenosine) to obtain a reproducible and steady state of coronary hyperemia (3,6,7), which is associated with transient and generally well tolerable symptoms, prolongation of the procedure by a few minutes, and in some countries, extra costs.
Recently, the instantaneous wave-free ratio (iFR) was introduced as an adenosine-free alternative for FFR (8). This index is calculated as the ratio of resting distal coronary pressure (Pd) and aortic pressure (Pa) over a specific period in late diastole, the wave-free period (WFP), during which intracoronary resistance is purportedly naturally constant and minimal. The physical basis and even the existence of the WFP have been questioned (9). Moreover, experimental data to support this theory are lacking, and previous studies have shown that both iFR and resistance during the WFP decreased markedly during adenosine infusion (10–12). In fact, in the IDEAL (Iberian–Dutch–English) study, it was even shown that resistance during the WFP is higher than whole-cycle hyperemic resistance (13).
Nonetheless, several clinical studies have reported that use of adenosine can indeed be avoided in a number of lesions using a hybrid approach with a dual iFR cutoff value (10,11,14). Moreover, in 2 recent randomized controlled trials, an iFR-based revascularization approach using a binary cutoff value of 0.89 was noninferior with respect to a composite clinical endpoint in a low-risk population compared with an FFR-based approach (15,16).
In contrast to FFR, which is generic, has been granted open-access, and is not commercially restricted, iFR measurements are limited to proprietary software of 1 single vendor, curtailing clinical application and research more generally. However, other resting indexes can be measured or calculated directly using any measuring equipment available in the catheterization laboratory. Accordingly, the aim of the current study was to compare several nonhyperemic indexes obtained during different parts of diastole or the whole heart cycle to iFR.
This post hoc analysis was performed using the prospectively acquired hemodynamic recordings from the VERIFY2 (Pd/Pa vs iFR in an Unselected Population Referred for Invasive Angiography; NCT02377310) study; the specific details of the study have been described previously (10).
Briefly, Pa (guiding catheter) and Pd (Prestige or Verrata Wire, Volcano Corporation, Rancho Cordova, California) curves were acquired simultaneously in 197 near-consecutive patients from 257 vessels at baseline (resting condition) and during administration of intravenous adenosine (hyperemic condition). For repeatability, all measurements were repeated after a 2-min resting period.
In addition, in this study the following diastolic indexes were pre-specified (Figure 1):
• Diastolic pressure ratio (dPR): average Pd/Pa over the entire diastole.
• dPR25–75: average Pd/Pa from 25% to 75% into diastole.
• dPRmid: Pd/Pa at the single point in time at mid-diastole.
• iFRmatlab: average Pd/Pa from 25% into diastole until 5 ms before end of diastole, following the definition of the WFP described by Sen et al. (8); this was not calculated using the proprietary software (Volcano Corporation), but instead by using standard mathematical methods.
• iFR−50ms: average Pd/Pa from 25% into diastole until 50 ms before end of diastole.
• iFR−100ms: average Pd/Pa from 25% into diastole until 100 ms before end of diastole.
Furthermore, as explained in the Online Appendix, the pressure ratios during the intervals defined in the previous text but derived from the curves during hyperemia were also calculated to investigate whether the observed correlations were only confined to resting conditions or would also exist during hyperemia.
Finally, in addition to the diastolic indexes, some other resting indexes that are not confined to diastole were analyzed. Rest Pd/Pa over the whole heart cycle and Pd/Pamin were compared with iFR. This work is further described in the Online Appendix.
Data handling and statistics
All curves and iFR values, calculated with the proprietary Volcano Harvest software, were stored in the Volcano s5 console HDD (Volcano Corporation). Pressure curves were analyzed offline by using Matlab version R2012b (Mathworks, Inc., Natick, Massachusetts) to obtain the different diastolic indexes as described in the previous text. The start of diastole was defined at the nadir of the dicrotic notch on the aortic pressure signal (Pa) and the end of diastole was defined at 50 ms before the upstroke in the aortic pressure signal from the subsequent ventricular contraction. All indexes were determined in a fully automated manner for 5 consecutive beats and were then averaged and compared with acquired iFR values. Spearman’s correlation coefficients were determined for all indexes with iFR, and linear regression was applied as well as determination of mutual differences.
Receiver-operating characteristic (ROC) curves with area under curve (AUC) values were determined with iFR as reference standard using a threshold of ≤0.89. Because repeated measurements were available, the average value of the 2 measurements was used for the analyses. To put the AUC values for the different indexes into perspective, an ROC curve analysis was also performed for the second iFR measurement, with the first measurement as the reference standard using a threshold of ≤0.89.
Diagnostic values with respect to FFR ≤0.80 (gold standard) were determined using an identical cutoff value of ≤0.89 for iFR, dPR, dPR25–75, dPRmid, iFRmatlab, iFR−50ms, and iFR−100ms. Diagnostic values with respect to iFR ≤0.89 were determined similarly, and the results are shown in the Online Appendix.
Continuous variables are presented as mean ± SD or median with interquartile range (IQR) as appropriate. Analyses were done with Matlab version R2012b.
A total of 197 patients were included, of whom 136 (69%) were men, 31 (16%) had diabetes, and 73 (37%) had prior MI. Coronary pressure measurements were acquired in 257 coronary arteries, of which 148 (58%) were in the left anterior descending artery, and the mean diameter stenosis was 48 ± 13%. The most frequent indications for angiography were stable angina (50.1%), non–ST-segment elevated myocardial infarction (30.1%), unstable angina (7.0%), and atypical chest pain (5.1%). The median FFR value was 0.81 (IQR: 0.75 to 0.87) and the median iFR was 0.91 (IQR: 0.87 to 0.95). For further characteristics of the VERIFY2 patient cohort, please refer to Hennigan et al. (10).
Mutual differences and correlation between iFR and other diastolic resting indexes
Table 1 shows the median values of the various indexes, their respective differences compared with iFR, and their respective correlation with iFR. The differences between any of the indexes and iFR are similar to the test-retest repeatability of iFR, which was found to be −0.0001 ± 0.0156 in this study. Spearman’s correlation coefficients are >0.99 (p < 0.001) for all of the indexes. The scatterplots for the relation between iFR and the other indexes are displayed in Figure 2.
ROC analyses with iFR as the reference standard
AUC values from the ROC analyses with iFR as the reference standard for the data at rest are >0.99 (Table 1) for all of the indexes, indicating near-exact matches (Figure 3). The AUC value from the ROC analysis of the second Volcano iFR measurement with the first Volcano iFR measurement as the reference standard was found to be 0.995, corroborating the former statement.
Diagnostic value of all diastolic resting indexes with FFR as the reference standard
The diagnostic accuracy to predict an FFR value with a binary cutoff value of 0.80 was 77% for iFR, which is equal to all other indexes (76% to 77%), all with the same cutoff value of 0.89. Negligibly small differences were observed for sensitivity, specificity, and negative and positive predictive values (Figure 4).
Correlation between the different diastolic pressure ratios in the hyperemic curves
As displayed in Online Figure 1 (and in Online Table 1), under hyperemic conditions, the Pd/Pa ratios corresponding with the time intervals of the investigated indexes showed the same correlation with identical numerical values compared with iFR, showing that the equality of all indexes is not dependent only on resting conditions.
Nondiastolic resting indexes
Comparison of nondiastolic resting indexes with iFR showed an equally similar linear correlation for Pd/Pa at rest during the whole heart cycle and for Pd/Pamin, but with slightly different numerical values (Online Table 2). ROC curves with iFR as the reference standard are shown in Online Figure 2.
Diagnostic value of all diastolic indexes with iFR as the reference standard
The diagnostic accuracy of all indexes with iFR as the reference standard was equal to the diagnostic accuracy of the second iFR measurement compared to the first iFR measurement as measured by the Volcano console (Online Figure 3).
The present analysis shows that all diastolic indexes are essentially identical to iFR both numerically and in terms of diagnostic performance. In addition, all indexes had the same agreement with FFR as the reference standard with a diagnostic accuracy of 76% to 77%. This implies that all of these resting indexes can be used interchangeably, and all cutoff values, clinical recommendations, and guidelines used for iFR are directly applicable to the diastolic indexes investigated in this study (Central Illustration). Our results simplify the clinical translation of coronary physiology measurements in the catheterization laboratory and should facilitate adoption of functional testing.
The negligible differences between iFR and the diastolic indexes presented in this study (Table 1) indicate that measurements during a specific period of diastole and the exact definition of the end of diastole are not essential for the clinical utility of diastolic resting indexes of stenosis severity. These differences are within the test-retest limits of each parameter. The AUC values of >0.99 for every diastolic index from ROC curve analyses when using iFR as the reference standard, questions the proposed uniqueness of the iFR algorithm. Shortening the definition of the WFP by 50 or 100 ms (or extended to earlier diastole), does not influence the calculated values. Resting Pd/Pa from the whole heart cycle and Pd/Pamin can be used, but with a slightly different cutoff value. This was shown in a recent study by Kobayashi et al. (17), which demonstrated an AUC value of 0.98 when the resting whole cycle Pd/Pa was compared to iFR, which is similar to values from this study (Online Appendix). An iFR value can be predicted from whole cycle resting Pd/Pa by a simple linear transformation (18). Moreover, both of these indexes are independent of a specific period in diastole, but need to be further investigated in further studies.
The fact that the equivalence between investigated indexes and iFR was maintained when also making the calculations from the hyperemic curves indicates that there is no specific physiological basis for preferring one index over another. In other words, the dependency of all of these indexes on hyperemia is identical, consistent with the findings from the VERIFY study (11). Finally, repeatability of all indexes was equally good.
So far, research with iFR has been restricted to users of Volcano/Philips equipment. The results from the current study indicate that any diastolic resting index can be used with the same advantages and disadvantages inherent to iFR. VERIFY2 is the only comparatively large, publicly available database that includes iFR measured with proprietary software and rest Pd/Pa. This resource becomes all the more relevant given the lack of published data supporting rest Pd/Pa.
Merits of iFR
The merit of iFR, and the subsequent discussions and studies investigating whether hyperemia is mandatory for clinical decision making for ischemia and necessity of revascularization, is that use of adenosine can be omitted in some of the patients undergoing invasive functional testing of stenosis severity. Using a hybrid approach, hyperemia can be avoided in about 50% of the population without seriously affecting diagnostic accuracy. These results represent a worthwhile advance, provided safety is not compromised (19). The longer-term follow-up results from DEFINE-FLAIR (Functional Lesion Assessment of Intermediate Stenosis to Guide Revascularisation) (15) and SWEDE-HEART (Instantaneous Wave-free Ratio versus Fractional Flow Reserve in Patients with Stable Angina Pectoris or Acute Coronary Syndrome) (16) and application of resting indexes in higher-risk patients will be important in this regard.
Shortcomings of iFR
iFR has been described as a very specific index based upon particular pathophysiological phenomena present during a select part of diastole. Experimental studies supporting the hypothesis are lacking. The existence of the WFP, upon which iFR is based, has been questioned (9), and our data indicate, in fact, that a particular WFP does not exist.
iFR has been shown to be strongly influenced by hyperemia, with resistance during the WFP falling between 150% and 200% after administration of adenosine. The numerical correspondence of iFR to FFR was described in the ADVISE (ADenosine Vasodilator Independent Stenosis Evaluation) study (8) but was later discarded, and the best cutoff value for indicating ischemia increased from 0.83 to the range of 0.89 to 0.91. The reasons for the comparatively low iFR values in the ADVISE study and the shift to a higher cutoff value are not completely understood, but were attributed to patient selection (20).
The present analyses show that any resting index, taken from different parts of diastole, yields factually identical diagnostic results, both numerically and in relation to FFR. Our results also show that microvascular resistance in rest is minimized not only during the WFP but also in other periods of diastole, supporting the critique of Westerhof et al. (9).
Caveats of resting indexes in general
Although the present study shows that iFR and all other diastolic indexes are “clinically equal” and that concordance is present between all resting indexes and FFR in 76% to 77% of cases, some caveats exist when using resting indexes only.
First, independent superiority studies to show improved outcome with iFR have never been performed. Resting indexes including iFR have only been compared with FFR. Second, outcome studies in comparison to FFR have only been performed in low-risk populations (15,16). Especially in proximal stenoses in large coronary arteries (left main, proximal left anterior descending, proximal right coronary artery), false-negative iFR is present in up to 30% of patients (21). Also, in large perfusion territories, short tight lesions, young patients with good microvascular function, and dominant right coronary arteries, minimal resting gradients can be associated with significant coronary artery disease, as reflected by FFR. Third, hyperemic gradients in a coronary artery are usually 2 to 3× larger than resting gradients. Therefore, avoiding adenosine decreases the signal-to-noise ratio of the pressure pull-back recording and the sensitivity to unmask the significance of diffuse disease and cross-talk between different lesions in 1 coronary artery. Fourth, the decreased signal-to-noise ratio at rest also enhances confounding by drift, which is a troublesome phenomenon associated with most pressure wires. As drift is an absolute phenomenon (mm Hg/unit of time), its influence is larger when using resting indexes. Fifth, post-PCI assessment by physiology—an increasingly important issue (22)—is not possible by resting indexes because a variable state of submaximal hyperemia often exists for a prolonged time after a complex PCI. Finally, resting indexes do not allow quantification of the rate of improvement by stenting. So, when using hyperemia and FFR, an increase of FFR from 0.60 to 0.90 indicates an improvement of 50% of maximum myocardial perfusion by PCI. Such useful information is lost when only using resting indexes.
First, this study was not a pre-specified post hoc analysis. However, the results indisputably show that selecting the exact WFP is not essential for calculating iFR. Second, one could argue about the definition of the end of diastole. Physiologically, the start of the consecutive left ventricular contraction precedes the upstroke of the aortic pressure by about 30 to 50 ms, the isovolumetric contraction period. Therefore, the end of diastole was chosen 50 ms before the upstroke of the next heartbeat.
Initially, electrocardiogram signals were used as a marker for determination of the end of diastole in the iFR algorithm. Later, some authors have described that left ventricular contraction observed in the distal coronary signal may be sufficient, making the electrocardiogram superfluous (23). The results presented in this study, based on the aortic signal alone, show that a simple time difference of 50 or even 100 ms before the Pa upstroke yields exactly the same results.
Third, the choice for the different diastolic periods investigated was arbitrary but intuitive. Investigating the complete diastole instead of part of it is straightforward, but many other (diastolic) resting indexes perform equally well. Calculating Pd/Pa during other parts of diastole makes little difference.
Finally, the number of investigated curves to conclude that iFR and other diastolic indexes are similar is arbitrary. But, we believe that the similarity in this dataset of 250 × 2 curves will not change when extending this comparison to thousands of curves (24).
Many diastolic resting indexes are identical to iFR, both numerically and with respect to agreement with FFR. Consequently, all cutoff values, guidelines, and clinical recommendations for iFR can be extended 1:1 to any of these other diastolic resting indexes.
COMPETENCY IN MEDICAL KNOWLEDGE: The nonhyperemic iFR can be used for physiological assessment of coronary artery stenosis severity at rest. The iFR can also be calculated from the ratio of mean Pd/Pa during specific periods in diastole other than the WFP.
TRANSLATIONAL OUTLOOK: Further studies of functional tests are needed to expand the utility of physiological measurements in the catheterization laboratory.
Dr. van’t Veer has received a speaker fee from Abbott. Dr. Pijls has served as a consultant for Abbott and Opsens; and has equity in Philips, GE, HeartFlow, and ASML. Dr. Ali has received institutional research grants to Columbia University from St. Jude Medical and Cardiovascular Systems Inc.; and has served as a consultant to St. Jude Medical, ACIST Medical, Cordis, and Boston Scientific. Dr. De Bruyne is shareholder for Siemens, GE, Bayer, Philips, HeartFlow, Edwards Lifesciences, Sanofi, and Omega Pharma; his institution, the Cardiovascular Center Aalst, has received grant support from Abbott, Boston Scientific, Biotronik, and St. Jude Medical; and his institution has received consulting fees on behalf of Dr. De Bruyne from St. Jude Medical, Opsens, and Boston Scientific outside of the submitted work. Dr. Barbato’s institution, the Cardiovascular Research Institute Aalst, has received speaker and consultancy fees from Abbott Vascular and Boston Scientific on his behalf. Dr. Berry and the University of Glasgow hold research and consultancy agreements with Abbott. Dr. Oldroyd has received consultancy and speaker fees from Abbott; and has received research support from Boston Scientific. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Berry and Oldroyd contributed equally.
- Abbreviations and Acronyms
- area under the curve
- diastolic pressure ratio
- fractional flow reserve
- instantaneous wave-free ratio
- interquartile range
- aortic pressure
- distal coronary pressure
- receiver-operating characteristics curve
- wave-free period
- Received September 7, 2017.
- Revision received October 26, 2017.
- Accepted October 29, 2017.
- 2017 The Authors
- Windecker S.,
- Kolh P.,
- Alfonso F.,
- et al.
- Fihn S.D.,
- Blankenship J.C.,
- Alexander K.P.,
- et al.
- De Bruyne B.,
- Pijls N.H.J.,
- Barbato E.,
- et al.
- Sen S.,
- Escaned J.,
- Malik I.S.,
- et al.
- Hennigan B.,
- Oldroyd K.G.,
- Berry C.,
- et al.
- Berry C.,
- van ’t Veer M.,
- Witt N.,
- et al.
- Johnson N.P.,
- Kirkeeide R.L.,
- Asrress K.N.,
- et al.
- Nijjer S.S.,
- de Waard G.A.,
- Sen S.,
- et al.
- Jeremias A.,
- Maehara A.,
- Généreux P.,
- et al.
- Götberg M.,
- Christiansen E.H.,
- Gudmundsdottir I.J.,
- et al.
- Davies J.E.,
- Sen S.,
- Dehbi H.-M.,
- et al.
- Kobayashi Y.,
- Johnson N.P.,
- Zimmermann F.M.,
- et al.
- Kern M.J.,
- Seto A.H.
- Berry C.,
- McClure J.D.,
- Oldroyd K.G.
- Petraco R.,
- Escaned J.,
- Davies J.
- Kobayashi Y.,
- Johnson N.P.,
- Berry C.,
- et al.
- Piroth Z.,
- Toth G.G.,
- Tonino P.A.L.,
- et al.
- Petraco R.,
- Sen S.,
- Nijjer S.,
- et al.