Use of Resting Non-hyperemic Indices for Avoidance of Fractional Flow Reserve Measurement: The Goal of 100% Accuracy

Abstract: Objectives. Recent studies have suggested that fractional flow reserve (FFR) measurement can be avoided by using similar ranges of baseline mean coronary pressure (Pd) to mean aortic pressure (Pa) ratio (0.88-0.95). Further studies have suggested that too many significant coronary stenoses are misclassified based on these ranges. We hypothesized that with a certain range of baseline Pd/Pa, 100% positive predictive value (PPV) and negative predictive value (NPV) can be achieved to avoid misclassification. Methods. We retrospectively evaluated the pressure tracings of 555 consecutive intermediate coronary stenotic lesions that had undergone FFR measurement in the cardiac catheterization laboratory of a tertiary-care center. The baseline Pd/Pa was manually measured and correlated with final FFR. The operating test characteristics were calculated using an abnormal FFR of ≤0.80 as the criterion standard for the presence of hemodynamic, significant coronary stenosis. Results. The area under the receiver-operating characteristics curve of baseline Pd/Pa for predicting FFR was 0.89, very similar to published results for instantaneous wave-free ratio and Pd/Pa. However, a significant number of lesions were mischaracterized (ie, using a baseline Pd/Pa of ≤0.88 to >0.95, there were 22 misclassifications, with 6 false-positive and 16 false-negative results). At a Pd/Pa of ≤0.86, 100% PPV was achieved, and 100% NPV was achieved at >1.00. Conclusion. A baseline Pd/Pa of ≤0.86 is associated with a PPV of 100%, which can avoid the misclassification errors seen in prior studies. This provides a more clinically useful application of baseline Pd/Pa. 

J INVASIVE CARDIOL 2016;28(7):265-270

Key words: adenosine, blood flow, stenosis, fractional flow reserve

Coronary fractional flow reserve (FFR) measurement provides an estimate of the physiological significance of intermediate coronary stenosis (50%-70%), obviating the need for stress testing and bringing patients back to the cardiac catheterization laboratory if the presence of ischemia is demonstrated and reducing interventions on lesions that are not physiologically significant.¹ 

In an effort to further expedite the decision-making process regarding such patients, recent studies suggested that FFR measurement and adenosine administration can be avoided in a subgroup of patients if a resting non-hyperemic index of baseline instantaneous wave-free ratio (iFR) measurement is employed. Only those patients with a resting iFR >0.86 and <0.93 would be subjected to adenosine.² While this concept further streamlines the care of coronary patients with intermediate stenoses in catheterization laboratories, its utilization is limited by the availability of commercial iFR software. In addition, a study in 2013 suggested that too many significant coronary stenoses are misclassified based on the currently proposed iFR/FFR hybrid approach ranges,³ further reducing the enthusiasm for this strategy.

Other studies have evaluated another non-hyperemic resting index – the ratio of coronary pressure (Pd) to aortic pressure (Pa) in the range of 0.88-0.95 – in eliminating the need for adenosine administration, and, therefore, preventing its side effects.^4,5 These studies have found nearly identical operating test characteristics of Pd/Pa to iFR (positive predictive value [PPV] of 91%; negative predictive value [NPV] of 98%; area under the curve [AUC] of 0.89). The use of the Pd/Pa ratio obviates the need to purchase iFR software, because Pd/Pa is free on all monitors. More recent data in 2015 from Kwon et al⁶ suggest better utilization of resting Pd/Pa data by widening the adenosine administration Pd/Pa range.^4,5

We believe the additional cost, side effects, and time delay associated with adenosine administration cause less harm to these patients than a missed opportunity to intervene in a physiologically significant stenosis due to a misclassification error. Therefore, adenosine infusion should be avoided only if a resting non-hyperemic index provides an accuracy of 100%. We carried out this study to find the baseline Pd/Pa range with 0% misclassification error, 100% PPV, and 100% NPV.

Methods

This study was carried out in the cardiac catheterization laboratory of Aurora St. Luke’s Medical Center in Milwaukee, Wisconsin. The institutional review board reviewed and approved the protocol of this retrospective study.

We identified 926 FFR measurements due to the intermediate nature of coronary stenosis (50%-70%) done in the catheterization laboratory database. The FFR tracings are saved in Synapse (Fujifilm), a DICOM imaging system for coronary angiograms. The saved image is a zoomed-out capture of the entire tracing, which made it difficult to evaluate the pressure tracings in detail and introduced a retrospective data measurement bias. Therefore, the original FFR tracings, in their entire detail, were procured directly from the original machines, giving us the same information that was available to the clinicians at the time of the test. These machines include the PrimeWire in the CORE system (Volcano Corporation) and pressure guidewire in the Ilumien FFR system (St. Jude Medical). 

The patients’ original FFR tracings were reviewed in zoomed-in view, which provided the ability to measure the baseline Pd/Pa (distal Pd/proximal Pa), as well as all the pressures before, during, and after the adenosine measurement. Baseline Pd/Pa was identified in the tracings, after normalization, before adenosine infusion, based on the profile of pressure curves (Figure 1). The final FFR measurement was similarly identified. Then, both the Pd/Pa and FFR zoomed-in, or magnified, snapshots were printed and filed for later review by the review committee.

Each pressure tracing was evaluated by one of four physicians (SSK, AR, FH, or MNA), and all tracings with potential artifacts were compiled. These tracings were adjudicated by a committee of five investigators (SSK, AR, MNA, KAA, and AG), and the type of error was characterized. The committee was led by the director of the cardiac catheterization laboratory. In the case of controversial tracings, we referred to the prior publication by Pijls et al,¹ guidelines issued by the American Heart Association,⁷ and numerous other resources in order to reach consensus. The types of artifacts are well described by Pijls et al and Lockie et al.^1,8 Briefly, these were:

(a)  recording artifact, defined as incomplete recording by the catheterization laboratory technician so that either the FFR tracing for baseline Pd/Pa was missing or there was a lack of a recording during or after adenosine infusion; 

(b)  normalization artifact, defined as an inability to verify equal pressures when the pressure sensor is close to the tip of the guiding catheter and correct them through an electronic normalization maneuver; 

(c)  motion artifact;

(d)  catheter whip, seen as abnormally large spikes in the distal pressure curve due to the pressure wire sensor striking the coronary wall;

(e)  electronic drift, defined as identical aortic and distal coronary pressure wave forms and persistence of the aortic notch in the distal curve;

(f)  zeroing error, defined as the presence of an FFR <0.80 without ventricularization of the distal pressure curve due to the failure to correct the error by pulling back the pressure guidewire to the catheter tip; and

(g)  ventricularization of the catheter guidewire, as seen when the catheter guidewire migrates into the left main stem during hyperemia, resulting in a drop in Pa. This leads to an underestimation of FFR.

FFR was manually measured on all the tracings while the investigator was blinded to the FFR measurements in the electronic medical record systems. A discrepancy between the FFR in a medical record and the manually measured FFR was recorded as “FFR measurement error” and adjudicated by the same review committee.

Statistical analysis. The operating test characteristics of Pd/Pa were evaluated by comparing them against an FFR <0.80. Receiver operating curves, scatter plots, and Spearman correlation r were calculated. We used JMP version 10.0 (SAS Institute, Inc). 

Results

Of the 926 total FFR measurements, a total of 371 were excluded because of data recording errors, measurement error, or the presence of artifacts. Of those excluded, the majority were due to the technician not saving pressure tracings for the entire length of the study. Since significant effort went into delineation of these errors and artifacts, prevalence of errors and artifacts will be described in detail elsewhere in a separate manuscript. The remaining 555 adequate tracings were from 483 studies performed on 477 patients. This was due to a minority of the patients having multiple studies and multiple lesions (Table 1). The demographics of this retrospective study are presented in Table 2. The majority of our patients were middle-aged men, with most having a history of hypertension, dyslipidemia, and smoking. 

The mean percentage of coronary artery stenosis was 60 ± 12%, which is consistent with the main premise of the study that only patients with intermediate coronary stenosis were interrogated. Most patients had only one vessel involved (Table 1). The most common lesion was in the left anterior descending coronary artery (Table 1). The mean FFR was 0.85 ± 0.09, with 28% abnormal FFRs. The mean baseline resting Pd/Pa was 0.95 ± 0.05. 

The scatter plot shown in Figure 2 demonstrates a strong, positive, linear, and statistically significant relationship between FFR and Pd/Pa (r = 0.76; P<.001). 

While the greatest accuracy, ie, 84%, was achieved at a single baseline Pd/Pa value of 0.91, it came at an unacceptable clinical price with a specificity, PPV, sensitivity, and NPV of 92%, 76%, 60%, and 86%, respectively. 

The area under the receiver-operating curve (Figure 3) of baseline Pd/Pa for predicting the FFR was 0.89, which is very similar to prior published results for iFR (AUC = 0.87)³ and Pd/Pa (AUC = 0.89).⁴

In order to further improve the single baseline Pd/Pa value, we evaluated our data in the previously proposed Pd/Pa range of 0.88-0.95. Not only was the AUC of Pd/Pa to predict FFR <0.80 in our study nearly identical to previously published results for Pd/Pa and iFR, we noted a similar specificity of 98% and PPV of 89% for Pd/Pa ≤0.88, and similar sensitivity of 94% and NPV of 96% for Pd/Pa of >0.95. 

Despite these seemingly high operating test characteristics, we noticed a significant prevalence of misclassification of lesions within the prior proposed Pd/Pa range (Pd/Pa ≤0.88 and >0.95), with 6 false-positives and 16 false-negatives. The review of the receiver operating curve revealed that a PPV of 100% was achieved at a Pd/Pa of ≤0.86 while 100% NPV was achieved at ≥1.00 (Figure 4), obviating the need to run adenosine on 18.8% of our 553 patients.  

Discussion

Our results demonstrate that a baseline Pd/Pa of ≤0.86 or ≥1.00 obviates the need for adenosine infusion and has 100% PPV and NPV, respectively, for functionally significant stenosis, defined as an FFR of <0.80.⁹ If perfect accuracy is desired, adenosine challenge can be eliminated for a Pd/Pa of <0.87 (all had FFR <0.80), but there is no safe upper limit for the baseline ratio since, technically, Pd/Pa cannot be >1.00. 

The use of intravenous adenosine is essential for the induction of maximal coronary hyperemia, needed for an accurate measurement of FFR. The reduction in use of adenosine based on the suggested Pd/Pa range of <0.86 will not only reduce the cost and time of the procedure but also spare the patient the benign side effects, eg, transient chest pain and flushing, and severe side effects, eg, bronchospasm, associated with intravenous adenosine administration.¹ 

The concept of iFR was introduced by Sen et al in the ADVISE study.¹⁰ Subsequent studies demonstrated a weak and hyperemia-dependent correlation between FFR and iFR.³ Similarly, the more recent RESOLVE study demonstrated that iFR and Pd/Pa had >90% accuracy to predict a positive FFR in 64.9% (62.6%-67.3%) of lesions and negative FFR in 48.3% (45.6%-50.5%) of lesions.¹¹ These results expose the limitations of iFR in accurately identifying a functionally significant coronary stenosis and the resultant mischaracterization of lesion significance at both ends of the iFR range.

The concept of adenosine-range iFR has been applied to only one other resting non-hyperemic index, ie, baseline Pd/Pa, to come up with an adenosine range Pd/Pa.^4,5 In 2010, Mamas et al⁵ suggested an adenosine-free range of baseline Pd/Pa from a baseline Pd/Pa of ≤0.87 (PPV, 94.6%) or resting Pd/Pa of ≥0.96 (NPV, 93%). In 2013, Kim et al proposed a similar range (Pd/Pa ≤0.88 or >0.95) to defer the use of adenosine for FFR measurement.⁴ They observed that when the resting baseline Pd/Pa was >0.95, sensitivity was 97.3% (95% confidence interval [CI], 93.7%-99.1%) and NPV was 98.1% (95% CI, 95.5%-99.4%). The baseline Pd/Pa of ≤0.86 was associated with a specificity of 98.6% (95% CI, 97.0%-99.5%) and a PPV of 90.8% (95% CI, 81.0%-96.5%). The findings of Kim et al were very similar to those seen in our study population. The highest accuracy in our study, 84%, was seen at a baseline Pd/Pa of 0.91 and was nearly identical to their highest accuracy of 82% at a baseline Pd/Pa of ≤0.92.⁴ Similarly, our data revealed that at the Pd/Pa of ≤0.88, the specificity and PPV were 98% and 89%, respectively, and at the baseline Pd/Pa of >0.95, the sensitivity was 94% and the NPV was 96%. However, after applying the previously proposed criteria by Kim et al to our data, we still found 22 misclassifications of coronary stenoses, with 6 false-positives and 16 false-negatives (Figure 4). Hence, it is not surprising to see such guidelines not being used to make clinical decisions in cardiac catheterization laboratories, as the risk of misclassification of a significant lesion or unnecessary coronary stenting heavily outweighs the benefits of deferring the use of adenosine outside this proposed baseline Pd/Pa range. 

Our study focuses on eliminating this doubt by proposing a limited range of baseline Pd/Pa of 0.87-1.00 for adenosine use. This range has 100% PPV and 100% specificity for a baseline Pd/Pa ≤0.86 and a 100% NPV and 100% sensitivity for a baseline Pd/Pa ≥1.00. The cost savings and time benefits provided by our proposed strategy would have been offered to 18.8% of our 553 patients if the new proposed strategy of a baseline Pd/Pd of 0.87-1.00 for adenosine administration had been followed. While 18.8% is much lower than the percentage in whom adenosine administration would be avoided if the narrower, previously proposed adenosine Pd/Pa ranges are used, the higher PPV and NPV create the necessary confidence to use this strategy clinically. If we can avoid adenosine administration in 1 out of 5 patients, then it is still better than letting the hybrid strategy die in clinical practice due to misclassification errors. This is where the rubber meets the road, ie, the application of theoretical concepts faces the practical medical, ethical, and financial realities of misclassification errors.

In this study, we propose a “no compromise” strategy, which is more in concert with clinical practice because it is more damaging for an interventional cardiologist to miss the therapeutic opportunity due to misclassification than to administer adenosine for a few minutes and have 100% accuracy. A study in 2015⁶ independently proposed a near-identical range of baseline Pd/Pa of 0.87-0.99 for the use of adenosine, with a baseline Pd/Pa of ≤0.86 with 100% PPV or a baseline Pd/Pa >0.99 resulting in 100% NPV. Their data closely parallel our data, with a near-identical distribution of baseline Pd/Pa and FFR values, ie, mean ± standard deviation of 0.83 ± 0.08 and 0.94 ± 0.05, respectively.⁶ The reproducibility of these findings across multiple institutions indicates real physiology that should continue to reproduce in future studies as well as in clinical practice. If the concept of resting non-hyperemic indices to predict FFR is to survive and prosper in clinical catheterization laboratories, then only those guidelines with 100% PPV and NPV will be able to provide the necessary confidence to warrant consistent clinical implementation. Extrapolating our data to clinical practice – since physiologically a Pd/Pa of >1 is unlikely to happen and is likely a measurement error – there is no upper limit of confidence to rule out a need for adenosine infusion, but the lower limit is identifiable at <0.87 as clearly ischemic, obviating the need for adenosine.

In the original study by Pijls et al, a total of 45 consecutive patients underwent three different forms of stress testing and if ischemia was detected by any one of the stress tests, then that was considered positive for ischemia.¹² In that landmark study, the NPV and sensitivity of FFR were 88% and PPV and specificity were both 100%, with an overall accuracy of 93%. Our study revitalizes this concept by bringing back the value of a “no compromise” strategy on PPV. Our data suggest 100% PPV for those with a baseline Pd/Pa of <0.87, which is identical to the 100% PPV observed by the landmark study in 1996.¹² The NPV and sensitivity of 88% in that study indicated that even in that setting, the strength of FFR was not in sensitivity and NPV, but in specificity and PPV, which is the same theme that our data bring forth. The only difference is that the prior landmark study utilized post-stress indices, whereas our study has utilized resting coronary pressure indices.

Study limitations and strengths. Our study was based on retrospective data collection, which creates biases such as documentation bias, measurement bias, etc. We endeavored to negate the effects by going back to the original pressure recordings from St. Jude Medical System, as the company has been storing the data in its complete detail. While we may have improved the data collection and measurement process, the results cannot parallel a well-designed prospective study in which it would be unlikely to have a sizable minority (40%) of incompletely recorded tracings. 

While other studies have shown similar findings,^4,6 both were similarly retrospective studies, and we deployed a very strict methodology of repeat manual measurement of Pd/Pa of FFR, which has not been previously performed, and increases the scientific validity of our findings.

Given the relatively low numbers when compared with a multicenter study, the finding of 100% NPV and PPV must be viewed with caution. We suggest a prospective look at this method before it can be accepted as a high-level, evidence-based recommendation.

Conclusion

Our data support the use of a resting non-hyperemic index of Pd/Pa in the range of <0.87 to avoid the use of adenosine administration, with a focus on a “no compromise” strategy as opposed to prior proposed strategies that result in a misclassification of a sizable number of lesions.

Acknowledgments. The authors gratefully acknowledge Jennifer Pfaff and Susan Nord of Aurora Cardiovascular Services for editorial preparation of the manuscript and Brian Miller and Brian Schurrer of Aurora Research Institute for help with the figures.

References

1.     Pijls NH, Kern MJ, Yock PG, De Bruyne B. Practice and potential pitfalls of coronary pressure measurement. Catheter Cardiovasc Interv. 2000;49:1-16. Erratum in 2000;50:157. 

2.     The iFR Modality Fundamentals. Volcano Precision Guided Therapy. Accessed at https://www.volcanocorp.com/products/pdf-files/The-iFR-Modality-Fundamentals.pdf. Accessed October 8, 2015.

3.     Berry C, van ’t Veer M, Witt N, et al. VERIFY (VERification of Instantaneous Wave-Free Ratio and Fractional Flow Reserve for the Assessment of Coronary Artery Stenosis Severity in EverydaY Practice): a multicenter study in consecutive patients. J Am Coll Cardiol. 2013;61:1421-1427. 

4.     Kim JS, Lee HD, Suh YK, et al. Prediction of fractional flow reserve without hyperemic induction based on resting baseline Pd/Pa. Korean Circ J. 2013;43:309-315. 

5.     Mamas MA, Horner S, Welch E, et al. Resting Pd/Pa measured with intracoronary pressure wire strongly predicts fractional flow reserve. J Invasive Cardiol. 2010;22:260-265.

6.     Kwon TG, Matsuzawa Y, Li J, et al. Clinical usefulness of nonhyperemic baseline Pd/Pa as a hybrid baseline Pd/Pa-fractional flow reserve strategy. Coron Artery Dis. 2015;26:49-55. 

7.     Vranckx P, Cutlip DE, McFadden EP, Kern MJ, Mehran R, Muller O. Coronary pressure-derived fractional flow reserve measurements: recommendations for standardization, recording, and reporting as a core laboratory technique. Proposals for integration in clinical trials. Circ Cardiovasc Interv. 2012;5:312-317.

8.     Lockie T, Rolandi MC, Piek JJ. Dynamic damping of the aortic pressure trace during hyperemia: the impact on fractional flow reserve measurement. J Invasive Cardiol. 2013;25:549-550.

9.     Tonino PA, De Bruyne B, Pijls NH, et al. Fractional flow reserve versus angiography for guiding percutaneous coronary intervention. N Engl J Med. 2009;360:213-224. 

10.     Sen S, Escaned J, Malik IS, et al. Development and validation of a new adenosine-independent index of stenosis severity from coronary wave-intensity analysis: results of the ADVISE (ADenosine Vasodilator Independent Stenosis Evaluation) study. J Am Coll Cardiol. 2012;59:1392-1402. 

11.     Jeremias A, Maehara A, Généreux P, et al. Multicenter core laboratory comparison of the instantaneous wave-free ratio and resting Pd/Pa with fractional flow reserve: the RESOLVE study. J Am Coll Cardiol. 2014;63:1253-1261. 

12.     Pijls NHJ, de Bruyne B, Peels K, et al. Measurement of fractional flow reserve to assess the functional severity of coronary artery stenosis. N Engl J Med. 1996;334:1703-1708.

^*Joint first authors.

From ¹Aurora Cardiovascular Services, Aurora Sinai/Aurora St. Luke’s Medical Centers, Milwaukee, Wisconsin.

Disclosure: The authors have completed and returned the ICMJE Form for Disclosure of Potential Conflicts of Interest. The authors report no conflicts of interest regarding the content herein.

Manuscript submitted November 30, 2015, provisional acceptance given January 21, 2016, final version accepted March 1, 2016.

Address for correspondence: Anjan Gupta, MD, Aurora Cardiovascular Services, Aurora St. Luke’s Medical Center POB, 2801 W. Kinnickinnic River Parkway, Suite 840, Milwaukee, WI 53215. Email: publishing5@aurora.org