J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S VO L . 1 0 , N O . 1 4 , 2 0 1 7

ª 2 0 1 7 B Y T H E A M E R I C A N CO L L E G E O F C A R D I O L O G Y F O U N DA T I O N

P U B L I S H E D B Y E L S E V I E R

I S S N 1 9 3 6 - 8 7 9 8 / $ 3 6 . 0 0

h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j c i n . 2 0 1 7 . 0 3 . 0 3 1

Accuracy of Fractional Flow ReserveMeasurements in Clinical PracticeObservations From a Core Laboratory Analysis

Mitsuaki Matsumura, BS,a Nils P. Johnson, MD,b William F. Fearon, MD,c Gary S. Mintz, MD,a

Gregg W. Stone, MD,a,d Keith G. Oldroyd, MD,e Bernard De Bruyne, MD,f Nico H.J. Pijls, MD, PHD,g,h

Akiko Maehara, MD,a,d Allen Jeremias, MD, MSCa,i

ABSTRACT

Fro

He

CefCa

of

ini

he

(fo

sen

Me

Me

Me

fro

Me

res

OBJECTIVES The aim of this study was to compare site-reported measurements of fractional flow reserve (FFR) with

FFR analysis by an independent core laboratory (CL).

BACKGROUND FFR is an index of coronary stenosis severity that has been validated in multiple trials and is widely used

in clinical practice. However, the incidence of suboptimal FFR measurements is unknown.

METHODS Patients undergoing FFR assessment within the CONTRAST (Can Contrast Injection Better Approximate FFR

Compared to Pure Resting Physiology) study had paired, repeated measurements of multiple physiological metrics per

local practice. An independent central physiology CL analyzed blinded pressure tracings off-line in a standardized fashion

for comparison.

RESULTS A total of 763 patients were included in the study; 4,946 distal coronary artery pressure/aortic pressure

(nonhyperemic) and FFR tracings were analyzed by the CL (mean 6.5 tracings per patient). Pull-back data were available

for 616 patients (80.7%), of whom 108 (17.5%) had signal drift, defined as distal coronary artery pressure/aortic pressure

(nonhyperemic) <0.97 or >1.03. Among the remaining 4,217 tracings without evidence of signal drift, 222 (5.3%) were

noted to have ventricularization of the aortic waveform, and 168 (4.0%) had aortic waveform distortion. Excluding cases

with signal drift and waveform distortion, there was excellent agreement between CL-calculated and site-reported FFR,

with a mean difference of 0.003 � 0.02. Predictors of distorted waveforms were smaller guiding catheter size (odds

ratio: 6.30; 95% confidence interval: 3.22 to 12.32; p < 0.001) and intracoronary adenosine use (odds ratio: 0.13; 95%

confidence interval: 0.05 to 0.33; p < 0.001).

CONCLUSIONS This FFR CL analysis showed that almost 10% of tracings demonstrated waveform artifacts, and an

additional 17.5% had signal drift. Among adequate tracings, there was a close correlation between site-reported and

CL-analyzed FFR values. Attention to detail is critical for FFR studies to ensure adequate technique and optimal results.

(J Am Coll Cardiol Intv 2017;10:1392–401) © 2017 by the American College of Cardiology Foundation.

m the aCardiovascular Research Foundation, New York, New York; bMcGovern Medical School at UT Health and Memorial

rmann Hospital, Houston, Texas; cStanford University Medical Center, Stanford, California; dColumbia University Medical

nter, New York, New York; eWest of Scotland Heart and Lung Centre, Golden Jubilee Hospital, Clydebank, United Kingdom;

rdiovascular Center Aalst, OLV Clinic, Aalst, Belgium; gCatharina Hospital, Eindhoven, the Netherlands; hEindhoven University

Technology, Eindhoven, the Netherlands; and the iSt. Francis Hospital, Roslyn, New York. This study was an investigator-

tiated study and supported financially by St. Jude Medical. Dr. Johnson has received internal funding from the Weather-

ad PET Center for Preventing and Reversing Atherosclerosis and significant institutional research support from St. Jude Medical

r this study, NCT02184117) and Volcano/Philips Corporation (for NCT02328820), makers of intracoronary pressure and flow

sors. Dr. Fearon has received institutional research support from St. Jude Medical and Medtronic; has received honoraria from

dtronic; and has served as a consultant to HeartFlow. Dr. Mintz has received honoraria from Boston Scientific and ACIST

dical; and fellowship or grant support from Boston Scientific and St. Jude. Dr. Oldroyd has received speaking fees from St. Jude

dical, AstraZeneca, and Volcano Corporation. Dr. De Bruyne has received institutional consultancy fees and research support

m St. Jude Medical. Dr. Pijls has received institutional grant support from St. Jude Medical; serves as a consultant for St. Jude

dical, Boston Scientific, and Opsens; and possesses equity in Philips, General Electric, and HeartFlow. Dr. Maehara has received

earch grants from and is consultant for Boston Scientific; and speaking fees from St. Jude Medical. Dr. Jeremias has served as a

AB BR E V I A T I O N S

AND ACRONYM S

cFFR = contrast fractional flow

reserve

CL = core laboratory

FFR = fractional flow reserve

IC = intracoronary

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7 Matsumura et al.J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1 Accuracy of FFR

1393

T he use of a pressure wire to calculate frac-tional flow reserve (FFR) is a technique toassess the functional severity of an epicardial

coronary artery stenosis (1). Multiple randomizedclinical trials and observational data have demon-strated that FFR-guided revascularization improvesclinical outcomes (2–5), and both U.S. and Europeanguidelines have endorsed its use (6,7).

SEE PAGE 1402

IV = intravenous

OR = odds ratio

Pd/Pa = distal coronary artery

pressure/aortic pressure

(nonhyperemic)

Because of the relative simplicity of the FFR tech-nique and need for an immediate guide to treatment,none of the clinical trials on which the recommen-dations have been based used a centralized core lab-oratory (CL). Although a CL adds to the cost andcomplexity of a study, this investment can beworthwhile if the CL serves as quality control andthus reduces the sample size of a trial by boostingsignal or reducing noise for a primary or secondaryoutcome. Other considerations that favor the use of aCL include regulatory requirements and the potentialfor expanded post hoc analyses.

Every FFR assessment contains a small amount ofuncertainty due to biologic variability (8). Addition-ally, bias due to pressure signal drift, waveformartifacts, or operator interpretation error can besuperimposed (9). Real-world data on the incidenceof these factors on the overall accuracy of FFR arelacking. The aim of this study was to compare theaccuracy of clinically obtained FFR measurementsversus a standardized FFR analysis in a central CL.

METHODS

PATIENT POPULATION. The present investigationwas a multicenter, international study comparing theaccuracy of contrast-induced hyperemia (contrastfractional flow reserve [cFFR]) versus standard FFRobtained with intravenous (IV) and intracoronary (IC)adenosine (CONTRAST [Can Contrast Injection BetterApproximate FFR Compared to Pure RestingPhysiology]; NCT02184117). Details of the studymethodology and results have been published previ-ously (10). Briefly, patients underwent routine physi-ological lesion assessment for clinical indications, andsubsequent care proceeded on the basis of the clinicalFFR measurement. Each subject gave informed con-sent as approved by the local Institutional Review

consultant and member of the Speakers Bureau for Volcano/Philips Corpo

reported that they have no relationships relevant to the contents of this pap

Manuscript received November 6, 2016; revised manuscript received March

Board of that participating center. Recruit-ment took place between June 2014 and April2015. This study was investigator initiated andsupported by funding from St. Jude Medical(St. Paul, Minnesota). The funding source wasnot involved in the design of the protocol orthe analysis and interpretation of the results.

Patients with prior coronary bypass surgery,left ventricular ejection fractions <30%, leftventricular hypertrophy (septal wall thickness>13 mm), contraindications to adenosine, orrenal insufficiency were excluded from thestudy. Only the first lesion interrogated with

FFR in each subject was included in the analysis. Culpritlesions for an acute myocardial infarction wereexcluded, but nonculprit lesions were permitted.

PHYSIOLOGY PROTOCOL. For FFR measurements, aCertus or Aeris pressure wire and the QUANTIENacquisition unit (St. Jude Medical) were used. Lesionselection for FFR was left to the individual operatorson the basis of clinical necessity and study inclusioncriteria. The pressure wire was prepared for FFRmeasurements according to the manufacturer’s rec-ommendations. Vessel preparation included admin-istration of IC nitroglycerin and anticoagulation perlocal practice. Equalization of the pressure wire andthe aortic pressure was performed at the tip of theguiding catheter before all measurements. The pres-sure wire was then advanced distal to the stenosis in astable location to ensure high-quality tracings.

As detailed previously, the complete physiologyprotocol consisted of duplicate measurements ofresting physiology (whole-cycle distal coronary arterypressure/aortic pressure [nonhyperemic] [Pd/Pa] aswell as the instantaneous wave-free ratio), cFFR, andFFR obtained with IC and IV adenosine (10). Therewas a minimum period of 1 min between measure-ments. Theoretically, the maximum number of mea-surements per subject was 8 if the entire protocol wascarried out. However, not all FFR measurements weremandatory, and thus the number of physiology trac-ings submitted to the CL was typically <8 per subject.The IC adenosine dose was left to operator discretion,but a strong recommendation was made for 100 to200 mg (11). Adenosine infusion was administered at astandard rate of 140 mg/kg/min via either a central or aperipheral IV line. The duration of the infusion wasapproximately 2 min but could be prolonged as

ration and St. Jude Medical. All other authors have

er to disclose.

15, 2017, accepted March 23, 2017.

FIGURE 1 Flow Chart for Core Laboratory Analysis

Matsumura et al. J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7

Accuracy of FFR J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1

1394

necessary or abbreviated if a steady state had beenreached or if not tolerated by the patient. At the endof the procedure, a drift check was recommended bypulling the pressure wire back to the tip of the guid-ing catheter to the same location as the initialequalization.

CL ANALYSIS. All pressure tracings were sent to theCardiovascular Research Foundation (New York, NewYork) Physiology Core Laboratory for standardized andcentralized review. Each subject’s physiology studywas separated into individual pressure tracings (i.e.,resting Pd/Pa, cFFR, or FFR with IC and IV adenosine)and submitted separately in random order to the CL inbatches of 10 individual patients. The CL was blindedto both individual patients and their pressure tracings;thus, the CL carried out a post hoc analysis withoutknowledge of the locally determined Pd/Pa or FFRvalue, method of hyperemia, enrolling site, or subjectand lesion characteristics. Because each section of thetracing was blinded and uncoupled from the rest, the

CL remained unbiased by knowledge of the othermeasurements for that subject.

The RadiAnalyzer Xpress instrument (St. JudeMedical) was used for coronary pressure measure-ments. The Physiology Core Laboratory assessed eachindividual tracing for quality based on pre-specifiedcriteria that included evaluation of the aortic andcoronary pressure signal for waveform distortion orloss, aortic pressure ventricularization, andarrhythmia. Each tracing received a binary decisionregarding adequate quality for inclusion, and Pd/Paor FFR was calculated independently for each tracing.In cases in which a final drift check was performed,the quality of the pull-back was assessed along withthe amount of drift. All tracings were overread by aphysician experienced in physiology measurements(A.J. or A.M.) to ensure data quality.

FFR was independently calculated at the CL andcompared with operator-reported measurements.Because measurements were performed in duplicateto assess reproducibility, the first FFR value with IV

FIGURE 2 Representative Case Examples

Patterns of variability in aortic and distal pressure waveforms. (A) Fractional flow reserve (FFR) pull-back with significant signal drift (left, asterisk) versus no signal

drift (right). (B) Representative example of aortic pressure ventricularization (left) versus normal adequate aortic pressure waveform (right). Note loss of dicrotic

notch (arrow) and deep diastolic dipping of the aortic waveform during ventricularization. (C) Aortic waveform distortion (left) versus normal waveform (right). Note

loss of dicrotic notch, lower amplitude of the aortic versus the distal waveform, and a sinus wave–like pattern of distorted aortic waveform (arrow). Green pressure

tracing is from the pressure wire and red pressure tracing is from the guiding catheter.

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7 Matsumura et al.J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1 Accuracy of FFR

1395

adenosine was used for FFR calculation. If the firstmeasurement was not available or rejected by the CL,the second FFR value with IV adenosine was used. Ifboth of these measurements were rejected, anacceptable measurement with IC adenosine was used

for FFR calculation. Figure 1 displays the study flowchart for the CL analysis.

CL DEFINITIONS. Pressure drift is defined as a sepa-ration from the initially equal aortic and coronary

FIGURE 3 Frequency Distribution of Signal Drift

Distribution of signal drift among 616 patients who had final pull-back performed.

A value of 1.00 indicates no drift at all.

TABLE 1 General Characteristics of the Study Population

(n ¼ 763)

Age, yrs 66 � 10

Male 547 (71.7)

Body mass index, kg/m2 27.3 � 4.7

Target vessel

Left main coronary artery 25 (3.3)

Left anterior descending coronary artery 460 (60.3)

Left circumflex coronary artery 138 (18.1)

Right coronary artery 140 (18.3)

Medical history

Prior myocardial infarction 198 (26.0)

Prior PCI 114 (14.9)

Family history of premature CAD 191 (25.0)

Hypertension 545 (71.4)

Dyslipidemia 508 (66.6)

Diabetes mellitus 219 (28.7)

Smoking (current or past) 363 (47.6)

Renal insufficiency(GFR <60 ml/min/1.73 m2)

74 (10.0)

Clinical presentation

Stable coronary artery disease 598 (78.4)

Unstable angina 84 (11.0)

NSTEMI 73 (9.6)

STEMI 8 (1.0)

Values are mean � SD or n (%).

CAD ¼ coronary artery disease; GFR ¼ glomerular filtration rate; NSTEMI ¼non–ST-segment elevation myocardial infarction; PCI ¼ percutaneous coronaryintervention; STEMI ¼ ST-segment elevation myocardial infarction.

Matsumura et al. J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7

Accuracy of FFR J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1

1396

pressures (measured at the tip of the guiding cath-eter) at the end of the procedure. Before theadvancement of the coronary wire distal to the ste-nosis, the 2 pressures are “equalized” to ensure thatthe pressure recordings agree (Pd/Pa ¼ 1). After theFFR measurement, the coronary wire is pulled backinto the guiding catheter, and Pd/Pa is recordedagain. In the absence of any drift, Pd/Pa should be 1.An arbitrary range of 0.97 to 1.03 was used asacceptable values, and drift was defined asmeasurements <0.97 or >1.03 (Figure 2A).

Aortic pressure ventricularization was defined asdiastolic dipping of the waveform, similar to a leftventricular pressure tracing (Figure 2B). This mayoccur in the presence of an ostial stenosis, largeguiding catheter, small vessel size, or deep catheterengagement and can be rectified by partial disen-gagement of the guiding catheter.

Aortic waveform distortion is defined as a blunt-ing of the aortic waveform with loss of the dicroticnotch and sinusoid appearance of the waveform(Figure 2C). Typically this is related to residualcontrast in the guiding catheter or injector system,small catheters, or luminal obstruction by a device(second guiding wire or balloon catheter) inside theguiding catheter.

STATISTICAL ANALYSIS. Data were summarized bydescriptive statistics. Linear regression analysis andintraclass correlation analysis were performed toexamine the relationship between operator and CLPd/Pa and FFR measurements. Receiver-operatingcharacteristic curves were constructed to identifythe concordance between the measurements. Agree-ment between operator and CL measurements wasassessed by Bland-Altman plots with corresponding95% limits of agreement. To explore the predictors ofsuboptimal FFR measurements, we used a logisticregression model that incorporated clinically relevantparameters. The clustering of pressure tracingswithin patients was modeled by including the patientas a random effect. Similarly, the clustering of pa-tients within sites was modeled by including the siteas a random effect. SAS version 9.2 (SAS Institute,Cary, North Carolina) was used for all analyses, and a2-tailed p value <0.05 was considered to indicatestatistical significance.

RESULTS

STUDY POPULATION. A total of 763 patients wereincluded in the study, and 4,946 pressure tracingswere analyzed by the CL (mean 6.5 tracings per

FIGURE 4 Comparison of Fractional Flow Reserve Between Operator and

Core Laboratory

(A) Linear regression analysis of fractional flow reserve (FFR) measurements comparing

operator-reported FFR with core laboratory analysis. There was a strong and linear

correlation between both FFRmeasurements. (B) Bland-Altman agreement between 2 sets

of measurements. Difference between measurements with core laboratory FFR and

operator FFR plotted against mean. ICC ¼ intraclass correlation coefficient.

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7 Matsumura et al.J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1 Accuracy of FFR

1397

patient), including resting measurements (Pd/Paand instantaneous wave-free ratio), cFFR, and FFRwith IC or IV adenosine. Pull-back data were availablefor 616 patients (80.7%), among whom 108 (17.5%)had evidence of signal drift (Figure 3). Includingthe 147 patients who had no pull-back available,4,217 tracings from 655 patients underwentwaveform analysis and FFR calculation. Proceduralcharacteristics and details of the study populationare presented in Table 1.

WAVEFORM ANALYSIS. Among the 4,217 tracings(n ¼ 655) without evidence of signal drift, 222 tracings(5.3%) were noted to have ventricularization of theaortic waveform, and 168 (4.0%) had aortic waveformdistortion. In a patient-level analysis, 130 patients(19.8%) had at least 1 disqualifying tracing demon-strating either aortic pressure ventricularization or adistorted aortic waveform. A total of 16 patients(2.4%) were identified, who had all analyzed wave-forms meet at least 1 exclusion criteria.

Overall, 238 patients (31.2%) had either signal driftor abnormal waveforms affecting at least 1 tracing,and 124 patients (16.3%) had signal drift or abnormalwaveforms in all tracings. However, repeating theanalysis of the CONTRAST study, no significant dif-ference was noted in the overall agreement betweencFFR and adenosine FFR when including all tracings(area under the curve ¼ 0.934) or only CL-acceptedtracings (area under the curve ¼ 0.928), indicatingthat overall study results in larger samples were notsignificantly affected by these abnormalities.

COMPARISON OF FFR CALCULATIONS. A total of598 patients had acceptable waveforms and absenceof significant drift and were included in the FFRanalysis. Of those, 330 FFR measurements were ob-tained with IV adenosine and 268 tracings with ICadenosine. CL-calculated and operator-reported FFRwere 0.79 � 0.11 and 0.80 � 0.11, respectively. Therewas a strong and linear correlation between CL andoperator reported FFR (R2 ¼ 0.969; intraclass corre-lation coefficient ¼ 0.984; p < 0.001) (Figure 4A). Theagreement between the FFR measurements wassimilarly good, with a mean difference of 0.003 �0.020 (Figure 4B). However, variation was noted inindividual cases, with 39 patients (6.5%) demon-strating an FFR difference of 0.02, 16 patients (2.7%) adifference of 0.03, and 26 patients (4.3%) a differenceof >0.03. Based on an FFR cutoff point of 0.80, 14patients (2.3%) were recategorized based on Physi-ology Core Laboratory FFR calculation from ischemicto nonischemic or vice versa, but only 2 patients(0.3%) crossed over the ischemic gray zone of >0.80

to #0.75 or vice versa. The differences in FFR calcu-lations between the CL and the clinical sites arisefrom the CL’s manually selecting the optimal cardiaccycle to determine the FFR, excluding any possibleartifacts.

PREDICTORSOFAORTICPRESSUREVENTRICULARIZATION.

No significant differences were noted with respect toage or body surface area for aortic pressure ven-tricularization. However, there was a modest trendfor more frequent aortic pressure ventricularizationin women (adjusted odds ratio [OR]: 1.72; p ¼ 0.06)

TABLE 2 Predictors of Aortic Pressure Ventricularization

Aortic PressureVentricularization

UnadjustedOdds Ratio (95% CI) p Value

AdjustedOdds Ratio (95% CI)* p ValuePresent Absent

Per patient (N ¼ 655) n ¼ 72 n ¼ 583

Age, yrs 64.4 66.0 0.98 (0.96–1.01) 0.21 0.98 (0.96–1.01) 0.20

Body surface area, m2 1.92 1.90 1.45 (0.47–4.43) 0.52 1.49 (0.44–5.07) 0.52

Female, % 37.5 28.3 1.54 (0.91–2.58) 0.11 1.72 (0.98–3.00) 0.06

Male, % 62.5 71.7 1.00 (reference) – 1.00 (reference) –

$6-F guiding catheter, % 88.9 80.8 1.89 (0.88–4.00) 0.10 1.75 (0.79–3.89) 0.17

5-F guiding catheter, % 11.1 19.2 1.00 (reference) – 1.00 (reference) –

Right coronary artery, % 13.9 18.9 0.67 (0.33–1.36) 0.27 0.69 (0.34–1.39) 0.30

Left coronary artery, % 86.1 81.1 1.00 (reference) – 1.00 (reference) –

Per tracing† (N ¼ 4,049) n ¼ 222 n ¼ 3,827

FFR with contrast, % 68.1 49.3 2.97 (1.76–5.01) <0.001 3.04 (1.79–5.17) <0.001

FFR with IC adenosine, % 76.0 38.7 7.19 (4.37–11.85) <0.001 7.30 (4.41–12.10) <0.001

FFR with IV adenosine, % 52.4 34.1 2.36 (1.35–4.12) 0.003 2.54 (1.44–4.49) 0.001

Resting Pd/Pa, % 13.5 32.0 1.00 (reference) – 1.00 (reference) –

*Adjusted by all listed parameters in this table. †Per-tracing analyses performed with tracings clustered within subjects in multilevel models. FFR with contrast, IC adenosine,and IV adenosine were each compared separately with resting Pd/Pa in 3 separate models.

CI ¼ confidence interval; FFR ¼ fractional flow reserve; IC ¼ intracoronary; IV ¼ intravenous; Pa ¼ aortic pressure; Pd ¼ distal pressure.

Matsumura et al. J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7

Accuracy of FFR J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1

1398

and with the use of larger guiding catheters (OR: 1.75;p ¼ 0.17). Analyzing individual tracings, FFR withcontrast, IV adenosine, and IC adenosine werestronger predictors of aortic pressure ventriculariza-tion compared with resting Pd/Pa (adjusted OR: 3.04,7.30, and 2.54, respectively) (Table 2).

PREDICTORS OF AORTIC WAVEFORM DISTORTION.

The use of a smaller guiding catheter size (i.e., 5-F)was the only predictor of aortic waveform distortion(OR: 6.30; 95% confidence interval: 3.22 to 12.32;

TABLE 3 Predictors of Distorted Waveforms

Distorted WaveformUn

Odds RPresent Absent

Per patient (N ¼ 655) n ¼ 63 n ¼ 592

Age, yrs 63.9 66.1 0.98

Body surface area, m2 1.86 1.91 0.49

Female, % 19.1 30.4 0.57

Male, % 80.9 69.6 1.00

5-F guiding catheter, % 52.4 14.7 6.12

$6-F guiding catheter, % 47.6 85.3 1.00

Right coronary artery, % 14.3 18.8 0.76

Left coronary artery, % 85.7 81.2 1.00

Per tracing† (N ¼ 3,995) n ¼ 168 n ¼ 3,827

FFR with contrast, % 54.7 49.7 1.30

FFR with IC adenosine, % 7.9 41.8 0.12

FFR with IV adenosine, % 37.6 34.5 0.93

Resting Pd/Pa, % 34.5 30.9 1.00

*Adjusted by all listed parameters in this table. †As in Table 2.

Abbreviations as in Table 2.

p < 0.0001). Female sex (OR: 0.46; 95% confidenceinterval: 0.21 to 1.01; p ¼ 0.05) and the use of ICadenosine (OR: 0.13; p < 0.001) were associated withsignificantly fewer tracings demonstrating aorticwaveform distortion (Table 3).

DISCUSSION

In this systematic CL analysis on the prevalenceof erroneous or suboptimal FFR measurements inclinical practice, the principal findings were that:

adjustedatio (95% CI) p Value

AdjustedOdds Ratio (95% CI)* p Value

(0.95–1.01) 0.13 0.97 (0.94–1.00) 0.06

(0.13–1.85) 0.29 0.48 (0.10–2.35) 0.36

(0.28–1.15) 0.12 0.46 (0.21–1.01) 0.05

(reference) – 1.00 (reference) –

(3.24–11.53) <0.001 6.30 (3.22–12.32) <0.001

(reference) – 1.00 (reference) –

(0.35–1.65) 0.48 0.71 (0.32–1.57) 0.40

(reference) – 1.00 (reference) –

(0.86–1.97) 0.21 1.33 (0.87–2.04) 0.19

(0.05–0.31) <0.001 0.13 (0.05–0.33) <0.001

(0.57–1.53) 0.77 0.79 (0.47–1.31) 0.36

(reference) – 1.00 (reference) –

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7 Matsumura et al.J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1 Accuracy of FFR

1399

1) almost one-fifth of FFR measurements had evi-dence of signal drift, defined as Pd/Pa <0.97 or >1.03at the final pull-back; 2) almost 10% of tracings hadeither ventricularization or distortion of the aorticwaveform; 3) after the exclusion of subjects withsignal drift and waveform abnormalities, the overallagreement between CL-analyzed and operator-reported FFR was excellent; and 4) the predictors ofabnormal pressure tracings included smaller guidecatheter size and the use of IC adenosine.

FFR is a simple invasive index for the assessmentof physiological lesion severity that can be rapidlyand reliably obtained during cardiac catheterization.It is well validated against noninvasive ischemiatesting (12) and has been shown to provide superiorrisk stratification and clinical outcomes comparedwith angiography alone (2). In addition, currentguidelines recommend its use when evidence ofinducible ischemia from noninvasive testing is notavailable (13–15). However, no guideline or consensusdocument on the appropriate measurement tech-nique existed until recently (9), in part becauseobtaining an FFR measurement is considered a rela-tively simple procedure compared with other coro-nary interventions. In fact, until recently, no CL wasavailable to analyze and verify these measurements,and none of the randomized physiology studies haveused a CL for data analysis. This is in stark contrast tomyriad studies in interventional cardiology thatroutinely submit data for a comprehensive CL anal-ysis of quantitative coronary angiography, intravas-cular ultrasound, optical coherence tomography, andother techniques to independently verify the studyresults (16,17).

The present study demonstrates the value of anindependent CL, as almost one-third of tracings hadeither significant drift or waveform abnormalities,diminishing the quality of the measurements. Inclinical practice, alterations in waveform, drift, orFFR misinterpretations may have a significant impacton an individual basis, depending on the closeness tothe cutoff point of 0.80. In FFR measurements thatare clearly ischemic (i.e., FFR <0.70) or clearly non-ischemic (i.e., FFR >0.90), minor ventricularizationof the aortic waveform or even moderate drift maynot alter clinical decision making. However, becauseFFR measurements are recommended predominantlyfor intermediate lesions, it is expected that the ma-jority of measurements fall into the clinically impor-tant area of 0.7 to 0.9. Meticulous technique,including a thorough visual analysis of the aortic anddistal waveforms as well as a pull-back to excludewire drift at the completion of the measurement, istherefore critical to ensure an accurate and reliable

FFR value on which clinical decisions can be confi-dently based. Other benefits of CL analysis includedata accuracy, particularly for U.S. Food and DrugAdministration studies, as well as rapid feedback tostudy sites to continuously improve data quality andthus reduce subject exclusions.

The clinical predictors of aortic pressure ven-tricularization showed a strong trend toward femalesex and larger size guiding catheters, which could beexplained by smaller coronary ostia in women andpotentially significant obstruction from larger cathe-ters (a 6- or 8-F guiding catheter potentially creates astenosis of 43% or 77%, respectively) (18). Also, theuse of contrast and adenosine versus resting Pd/Pawas associated with a significantly higher prevalenceof aortic pressure ventricularization, likely because ofthe necessity of adequate catheter engagement toeffectively deliver the drug into the coronary circu-lation or by catheter suction into the vessel fromincreased coronary blood flow. Aminian et al. (19)reported that disengagement of the guiding catheterduring adenosine infusion led to a decrease in meanFFR values that was driven predominantly by an in-crease in proximal aortic pressure. It is important toremember to partially disengage the guiding catheterto avoid this problem. In contrast, smaller guidingcatheters (i.e., 5-F) were associated with a signifi-cantly greater likelihood of a distorted aortic wave-form. Meticulous flushing of the guiding catheterwith saline after the injection of contrast will becritical to clear the catheter of contrast residue andthus avoid distortion of the waveform, especially insmaller caliber catheters.

Drift of the pressure sensor has been describedsince the introduction of the pressure wire (20). It isbelieved that this is related primarily to changes inthe piezoelectric sensor during the measurement andcan be minimized by flushing the wire with salinebefore its use. However, the prevalence of this prob-lem has never been formally examined, and althoughconsidered a nuisance, it is not believed to have amajor impact on the measurements. The presentstudy indicates that this is not an uncommon prob-lem, affecting nearly 20% of FFR measurements.Although the exact threshold for “significant” drift isarbitrary in a population, large drift could change adecision over the “gray zone” of 0.75 to 0.80, andhence a pull-back at the completion of the FFR studyis critically important to validate the accuracy of theresults. Nevertheless, within a large sample, studyresults are unlikely to be affected, as previouslydemonstrated in CONTRAST (10), as drift is a randomnoise with a median value of 0.99 and an SD of about0.03. FFR values scatter around their true value

PERSPECTIVES

WHAT IS KNOWN? FFR is a relatively simple pro-

cedure within the realm of interventional cardiology

and has been shown to improve clinical outcomes and

reduce cost.

WHAT IS NEW? In this CL analysis, we have

demonstrated that suboptimal FFR measurements

occur in almost one-third of tracings. Attention to

detail is critical for the procedure, including a careful

assessment of the waveforms as well as exclusion of

signal drift.

WHAT IS NEXT? Development of automated soft-

ware detecting waveform abnormalities and further

refinement of the pressure sensor would substantially

aid in increasing the accuracy of FFR measurements.

Also, the use of a CL could help standardize the

procedure and provide adequate quality control.

Matsumura et al. J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7

Accuracy of FFR J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1

1400

without introducing a bias and, because the scatter isrelatively small, the overall impact in CONTRASTwas not affected by making tighter drift thresholds(0.05, 0.03, and 0.01).

STUDY LIMITATIONS. Pull-backs were available inonly about 80% of the population, and patientswithout pull-back were included in the analysis.Assuming a similar rate of drift among patients whodid not have recorded pull-backs, some tracings wereincluded that may have significant drift. Also, theclinical significance of these findings was not estab-lished, and it is unknown to what extent clinicaldecision making would have been altered by theseresults. The comparison of site-reported andCL-measured FFR values was limited to CL-acceptedtracings, excluding subjects with significant drift orwaveform abnormalities because the “true” FFR(i.e., FFR from artifact-free tracings) cannot bereproduced. Thus, the excellent correlation betweenthe FFR assessments may represent an overestimate.Also, we were not able to determine the reason for thefew cases of large discrepancy between site-reportedand CL-determined FFR measurements, becausethe exact images on which the site-reported FFRmeasurements were based were not collected.Finally, we used only 1 commercially available pres-sure wire; it is not known whether the results wouldhave been similar with other devices.

CONCLUSIONS

This Physiology Core Laboratory analysis of FFRmeasurements demonstrates a relatively high

prevalence of imperfect FFR measurements eitherfrom signal drift or artifacts in the pressure wave-form. This may have important implications on clin-ical decision making, and attention to detail and astrict standardization of methods are critical whenmeasuring FFR to ensure optimal results.

ADDRESS FOR CORRESPONDENCE: Dr. AllenJeremias, St. Francis Hospital, Department of Cardi-ology, 100 Port Washington Boulevard, #105, Roslyn,New York 11576. E-mail: allen.jeremias@chsli.org.

RE F E RENCE S

1. Pijls NH, van Son JA, Kirkeeide RL, et al.Experimental basis of determining maximum cor-onary, myocardial, and collateral blood flow bypressure measurements for assessing functionalstenosis severity before and after percutaneoustransluminal coronary angioplasty. Circulation1993;87:1354–67.

2. Pijls NH, van Schaardenburgh P, Manoharan G,et al. Percutaneous coronary intervention offunctionally nonsignificant stenosis: 5-yearfollow-up of the DEFER Study. J Am Coll Cardiol2007;49:2105–11.

3. Tonino PA, De Bruyne B, Pijls NH, et al. Frac-tional flow reserve versus angiography for guidingpercutaneous coronary intervention. N Engl J Med2009;360:213–24.

4. De Bruyne B, Pijls NH, Kalesan B, et al. Frac-tional flow reserve-guided PCI versus medicaltherapy in stable coronary disease. N Engl J Med2012;367:991–1001.

5. Li J, Elrashidi MY, Flammer AJ, et al. Long-termoutcomes of fractional flow reserve-guided vs.angiography-guided percutaneous coronary inter-vention in contemporary practice. Eur Heart J2013;34:1375–83.

6. Levine GN, Bates ER, Blankenship JC, et al. 2011ACCF/AHA/SCAI guideline for percutaneous coro-nary intervention. A report of the American Col-lege of Cardiology Foundation/American HeartAssociation Task Force on Practice Guidelines andthe Society for Cardiovascular Angiography andInterventions. J Am Coll Cardiol 2011;58:e44–122.

7. Windecker S, Kolh P, Alfonso F, et al. 2014 ESC/EACTS guidelines on myocardial revascularization:the Task Force on Myocardial Revascularization ofthe European Society of Cardiology (ESC) and theEuropean Association for Cardio-Thoracic Surgery(EACTS) developed with the special contributionof the European Association of PercutaneousCardiovascular Interventions (EAPCI). Eur Heart J2014;35:2541–619.

8. Johnson NP, Johnson DT, Kirkeeide RL, et al.Repeatability of fractional flow reserve despitevariations in systemic and coronary hemody-namics. J Am Coll Cardiol Intv 2015;8:1018–27.

9. Toth GG, Johnson NP, Jeremias A, et al. Stan-dardization of fractional flow reserve measure-ments. J Am Coll Cardiol 2016;68:742–53.

10. Johnson NP, Jeremias A, Zimmermann FM,et al. Continuum of vasodilator stress from rest tocontrast medium to adenosine hyperemia forfractional flow reserve assessment. J Am CollCardiol Intv 2016;9:757–67.

11. Adjedj J, Toth GG, Johnson NP, et al. Intra-coronary adenosine: dose-response relationshipwith hyperemia. J Am Coll Cardiol Intv 2015;8:1422–30.

12. Danad I, Szymonifka J, Twisk JW, et al. Diag-nostic performance of cardiac imaging methods todiagnose ischaemia-causing coronary artery dis-ease when directly compared with fractional flow

J A C C : C A R D I O V A S C U L A R I N T E R V E N T I O N S V O L . 1 0 , N O . 1 4 , 2 0 1 7 Matsumura et al.J U L Y 2 4 , 2 0 1 7 : 1 3 9 2 – 4 0 1 Accuracy of FFR

1401

reserve as a reference standard: a meta-analysis.Eur Heart J 2017;38:991–8.

13. Lotfi A, Jeremias A, Fearon WF, et al. Expertconsensus statement on the use of fractional flowreserve, intravascular ultrasound, and optical coher-ence tomography: a consensus statement of the So-cietyofCardiovascularAngiographyand Interventions.Catheter Cardiovasc Interv 2014;83:509–18.

14. Wijns W, Kolh P, Danchin N, et al., Task Force onMyocardial Revascularization of the European Soci-ety of Cardiology (ESC) and the European Associa-tion for Cardio-Thoracic Surgery (EACTS), EuropeanAssociation for Percutaneous Cardiovascular In-terventions (EAPCI). Guidelines on myocardialrevascularization. Eur Heart J 2010;31:2501–55.

15. Patel MR, Dehmer GJ, Hirshfeld JW, Smith PK,Spertus JA. ACCF/SCAI/STS/AATS/AHA/ASNC/HFSA/SCCT 2012 appropriate use criteria for coro-nary revascularization focused update: a report ofthe American College of Cardiology Foundation

Appropriate Use Criteria Task Force, Society forCardiovascular Angiography and Interventions,Society of Thoracic Surgeons, American Associationfor Thoracic Surgery, American Heart Association,American Society of Nuclear Cardiology, and theSociety of Cardiovascular Computed Tomography.J Am Coll Cardiol 2012;59:857–81.

16. Cheng JM, Garcia-Garcia HM, de Boer SP, et al.In vivo detection of high-risk coronary plaquesby radiofrequency intravascular ultrasoundand cardiovascular outcome: results of theATHEROREMO-IVUS study. Eur Heart J 2014;35:639–47.

17. Kubo T, Shinke T, Okamura T, et al. Opticalfrequency domain imaging vs. intravascular ul-trasound in percutaneous coronary intervention(OPINION trial): study protocol for a randomizedcontrolled trial. J Cardiol 2016;68:455–60.

18. De Bruyne B, Stockbroeckx J, Demoor D,Heyndrickx GR, Kern MJ. Role of side holes in

guide catheters: observations on coronary pres-sure and flow. Cathet Cardiovasc Diagn 1994;33:145–52.

19. Aminian A, Dolatabadi D, Lefebvre P, et al.Importance of guiding catheter disengagementduring measurement of fractional flow reserve inpatients with an isolated proximal left anteriordescending artery stenosis. Catheter CardiovascInterv 2015;85:595–601.

20. Cook CM, Ahmad Y, Shun-Shin MJ, et al.Quantification of the effect of pressure wire drifton the diagnostic performance of fractional flowreserve, instantaneous wave-free ratio, andwhole-cycle Pd/Pa. Circ Cardiovasc Interv 2016;9:e002988.

KEY WORDS coronary physiology,fractional flow reserve, percutaneouscoronary intervention