OUP user menu

Three-dimensional and two-dimensional quantitative coronary angiography, and their prediction of reduced fractional flow reserve

Andy S.C. Yong , Austin C.C. Ng , David Brieger , Harry C. Lowe , Martin K.C. Ng , Leonard Kritharides
DOI: http://dx.doi.org/10.1093/eurheartj/ehq259 345-353 First published online: 12 August 2010

Abstract

Aims We investigated whether three-dimensional (3D) and two-dimensional quantitative coronary angiography (2D-QCA) measurements differed in their accuracy in predicting reduced fractional flow reserve (FFR), and how this varied with stenosis severity and the FFR cut-off used.

Methods and results Three-dimensional and 2D-QCA were compared in their measurements of minimum luminal area (MLA), percentage area stenosis, lesion length, minimum luminal diameter (MLD) and percentage diameter stenosis, and in their prediction of functionally significant FFR. In total, 63 target lesions were interrogated in 63 patients undergoing elective percutaneous coronary intervention. Of all measurements of lesion severity obtained by 3D-QCA, MLA best correlated with FFR (R = 0.63, P< 0.001), and was the most accurate predictor of FFR <0.75 (C statistic 0.86, P< 0.001). Of 2D-QCA measurements, MLD correlated best with FFR (R = 0.58, P< 0.001), and best predicted FFR <0.75 (C statistic 0.80, P<0.001). Overall, 3D-QCA showed a non-significant trend towards more accurate prediction of FFR than 2D-QCA, especially in intermediate lesions. The relationship between FFR and apparent stenosis severity was found to be curvilinear. Both 3D- and 2D-QCA were less accurate in intermediate lesions, and in predicting FFR ≤0.80 than in predicting FFR <0.75.

Conclusions The accuracy of QCA in predicting functionally significant FFR is limited and is dependent on FFR cut-off used and lesion severity. Where FFR is not available or contraindicated, 3D-QCA may assist in the evaluation of coronary lesions of intermediate severity.

  • Three-dimensional
  • Quantitative coronary angiography
  • Fractional flow reserve

See page 261 for the editorial comment on this article (doi:10.1093/eurheartj/ehq314)

Introduction

Three-dimensional quantitative coronary angiography (3D-QCA) uses standard images acquired during routine coronary angiography to reconstruct a 3D model of a coronary artery by fusing two or more orthogonal angiographic images. Three-dimensional quantitative coronary angiography reportedly allows a more accurate depiction of true vessel geometry when compared with standard two-dimensional (2D) QCA in phantom models1 and has been validated against intravascular ultrasound (IVUS), but only in a small sample of ten patients.2 Although IVUS currently yields the most accurate measurements of vessel geometry and lesion severity,3 3D-QCA measurements can be performed on existing standard coronary angiography images without the need for additional time or equipment during the procedure.4

Myocardial fractional flow reserve (FFR) is a measure of the functional significance of epicardial coronary stenoses. It represents the ratio of maximal blood flow in the myocardium perfused by a stenosed coronary artery to the maximal blood flow in the same territory without stenosis.5

Earlier studies showed that FFR <0.75 is an accurate predictor of ischemia when used to interrogate coronary lesions.5 Fractional flow reserve <0.75 is associated with poorer outcomes, and deferral of percutaneous coronary intervention for lesions with FFR ≥0.75 appeared to be safe.6 More recent studies however, have used FFR ≤0.80 as a cut-off to guide revascularisation.7,8

Importantly, FFR-guided coronary intervention was shown to be superior to 2D-QCA driven coronary intervention in preventing myocardial infarction, revascularisation or death.8 However, 2D-QCA was previously shown to have only a modest correlation with FFR.9,10 Two-dimensional quantitative coronary angiography is known to overestimate lesion area stenosis severity and underestimate lesion length because it only allows characterisation of lesion severity in single planes.4,11

Few studies have compared 3D- and 2D-QCA in native coronary stenoses,1114 and one study reported that 3D-QCA has reasonable accuracy in predicting FFR <0.75 in patients with visually determined intermediate stenosis.15 However, no studies to date have compared the differential utility of the anatomical analyses of 3D- and 2D-QCA in predicting functionally significant stenosis as assessed by FFR, or assessed the effects of FFR cut-point or lesion severity as determinants of this predictive value. We investigated the accuracy of 3D-and 2D-QCA measurements in predicting both FFR <0.75 and ≤0.80, and compared the correlation of FFR with 3D- and 2D-QCA measurements.

Methods

Patients from two tertiary referral hospitals scheduled for elective FFR-guided PCI of single-target lesions within one of the major epicardial coronary arteries were recruited for this study. The study was approved by the human research ethics committees of both institutions.

Consecutive patients who fulfilled the following criteria were included: target lesion was within one of the major epicardial coronary arteries excluding the left main coronary artery [left anterior descending (LAD), left circumflex (LCX) or right coronary artery (RCA)], no more than one lesion ≥20% diameter stenosis within the target vessel (lesions were considered separate if they were more than three reference vessel diameters apart), and no previous myocardial infarction in the target vessel territory.

Quantitative coronary angiography

Patients first underwent routine coronary angiography. Angiographic cine images were acquired at 15 frames per second (Axiom Artis, Siemens, Forchheim, Germany). Two-dimensional quantitative coronary angiography was performed offline using standard commercial software on a Leonardo workstation (Quant, Siemens), which is derived from the CAAS II system (Pie Medical Imaging, Maastricht, Netherlands). Automated distance calibration was used to determine pixel size. All analyses were performed during the ECG-gated end-diastolic frame. Angiographic views with the least foreshortening and yielding the best depiction of the stenoses were used. Edge detection correction was performed if required.

Three-dimensional quantitative coronary angiography was performed offline using 3D reconstruction software on the Leonardo workstation (IC3D, Siemens), which is derived from the Cardio-op B system (Paieon Medical, Rosh Ha'ayin, Israel). The contrast-filled non-tapered part of the guiding catheter was used to calibrate pixel size. The two best orthogonal angiographic views of the target lesion in the ECG-gated end-diastolic frame were used for 3D-QCA reconstructions. The site of minimum luminal diameter, and the proximal and distal coronary artery segments were manually identified on the first angiographic plane (Figure 1A and B), and repeated in a second image, at least 30° orthogonal to the first. Proximal and distal planes were derived automatically, and after the centre of the arterial lumen was manually identified (Figure 1C), the software automatically generated a 3D representation of the arterial lumen (Figure 1D).

Figure 1

Representative three-dimensional reconstruction of vessel geometry using two orthogonal angiographic planes. (A) Coronary angiogram of left anterior descending artery in left anterior oblique cranial view. (B) Initial definition of lesion, and non-stenotic segments proximal and distal to the lesion in first angiographic view. (C) Definition of lesion, proximal and distal sites in second orthogonal view. (D) Three-dimensional reconstruction of vessel lumen in the left anterior oblique cranial projection showing lesion length (L) and percentage cross-sectional area stenosis (CS). Yellow cross defines lesion site, cross denoted by ‘P' defines proximal site and cross denoted by ‘D' defines distal site.

Minimum luminal area (MLA), percentage area stenosis, minimum luminal diameter (MLD), percentage diameter stenosis and lesion length were measured using both 3D- and 2D-QCA. All measurements were performed twice and averaged by a single-experienced cardiologist blinded to the FFR results. Inter-observer error was determined by a second cardiologist. To compare correlation between FFR and QCA in eccentric vessels and non-eccentric vessels, eccentricity was determined angiographically as previously published.16

Measurement of fractional flow reserve

Fractional flow reserve was measured using institutional protocols in accordance with published techniques.17 In brief, a 6F angioplasty guiding catheter without side-holes was first used to engage the relevant coronary artery, and the pressure trace was compared before and after engagement to check for damping. The pressure reading obtained from the pressure-sensor guidewire (RADI, Uppsala, Sweden) was then equalized with that of the guiding catheter with the sensor placed at the tip of the catheter. The lesion was then crossed with the wire, and the pressure-sensor was placed at least 3 cm distal to the target lesion. Intracoronary nitroglycerine was administered (200 µg). Hyperaemia was induced using adenosine infusion (140 µg/kg/min) via the femoral vein or intracoronary bolus (72 µg in the left coronary system and 48 µg in the right coronary artery). Proximal arterial pressure and distal arterial pressure were measured during hyperaemia and the FFR was calculated by the formula: FFR = distal pressure/proximal pressure. The proximal arterial pressure trace was also checked to ensure that there was no damping during hyperaemia. In cases of ostial lesions or when there was damping of pressure due to guiding catheter engagement, intravenous adenosine infusion was used, and the proximal pressure was measured with the guiding catheter disengaged from the coronary artery. At the end of FFR measurement, the sensor was carefully pulled back, first across the lesion, and then to the tip of the guiding catheter to check for drift.

Statistical analysis

Results are expressed as mean ± standard deviation unless otherwise stated. Normality of the data was determined using the D′Agostino Pearson test and verified using histogram plots. Spearman′s correlation was performed for non-parametric data. Bland–Altman plots were used to compare measurements by 3D- and 2D-QCA. Paired t-tests were used to assess for significant differences between 3D- and 2D-QCA measurements. To obtain curves of best fit, curves of increasing complexity starting from a straight line were tested against one another using the F test. Receiver operating curve (ROC) analysis was used to analyse the accuracy of QCA in predicting functional significant stenosis. Paired comparisons between ROC curves were performed within the patient groups in accordance with previously published methods18 using MedCalc v. 11.1 (MedCalc Software, Mariakerke, Belgium). Other statistical analyses were performed using Graphpad Prism v. 5.01 (Graphpad Software, La Jolla, CA, USA) and SPSS v. 15 (SPSS, Chicago, IL, USA). A two-sided P value of <0.05 is considered significant.

Results

Baseline clinical and lesion characteristics

A total of 63 target lesions (46 LAD, 7 LCX, and 10 RCA) in 63 patients were studied. Baseline clinical characteristics are shown in Table 1. Mean target vessel diameter by 2D-QCA was 2.75 ± 0.49 mm (range: 1.62–4.13 mm), and by 3D-QCA was 2.69 ± 0.53 mm (range: 1.54–4.11 mm), and FFR was 0.74 ± 0.18. Baseline lesion characteristics and measurement error for both 3D- and 2D-QCA are shown in Table 2. Of the 63 lesions, 45 were of intermediate severity having percentage diameter stenosis by 2D-QCA between 40 and 70%. All patients had TIMI 3 grade flow.

View this table:
Table 1

Baseline clinical characteristics

Variable(n = 63)
Age [mean (SD), years]62.4 (10.2)
Male sex, n (%)43 (68.2)
Clinical history, n (%)
 Hypertension43 (68.3)
 Hypercholesterolemia44 (69.8)
 Diabetes28 (44.4)
 History of smoking29 (46.0)
Medications, n (%)
 Aspirin63 (100)
 Clopidogrel63 (100)
 Beta-blocker40 (63.5)
 ACE-I/ARB42 (66.7)
 Statin51 (81.0)
  • ACE-I, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker.

View this table:
Table 2

Baseline lesion characteristics and measurement error

Mean ± SDIntra-observer errorInter-observer error
3D-QCA MLA (mm2)1.87 ± 1.140.08 ± 0.040.13 ± 0.13
3D-QCA % area stenosis68.5 ± 13.11.8 ± 1.71.8 ± 1.0
3D-QCA MLD (mm)1.30 ± 0.420.07 ± 0.070.10 ± 0.10
3D-QCA % diameter stenosis51.3 ± 12.83.3 ± 3.43.8 ± 4.0
3D-QCA lesion length (mm)10.8 ± 3.91.0 ± 0.91.4 ± 0.9
2D-QCA MLA (mm2)2.06 ± 1.480.18 ± 0.190.28 ± 0.22
2D-QCA % area stenosis73.7 ± 13.43.4 ± 3.74.6 ± 3.7
2D-QCA MLD (mm)1.35 ± 0.470.06 ± 0.070.08 ± 0.09
2D-QCA % diameter stenosis50.8 ± 13.93.6 ± 3.45.3 ± 4.0
2D-QCA lesion length (mm)8.6 ± 3.10.86 ± 0.801.13 ± 0.29
  • 3D, three-dimensional; 2D, two-dimensional; QCA, quantitative coronary angiography; MLA, minimum luminal area; MLD, minimum luminal diameter.

Comparison between three- and two-dimensional quantitative coronary angiography

There was good correlation between 3D-QCA and 2D-QCA for all measurements of lesion severity (Figure 2A–E). Comparing 3D- vs. 2D-QCA for the different parameters measured yielded similar MLA (1.87 ± 1.14 vs. 2.06 ± 1.48 mm2, P = 0.615), significantly lower percentage area stenosis (68.5 ± 13.1 vs. 73.7 ± 13.4%, P < 0.001), similar MLD (1.30 ± 0.42 vs. 1.35 ± 0.47 mm, P = 0.250), similar percentage diameter stenosis (51.3 ± 12.8 vs. 50.8 ± 13.9%, P = 0.562) and greater lesion length (10.8 ± 3.9 vs. 8.6 ± 3.1 mm, P < 0.001). Bland–Altman plots demonstrating agreement between 3D- and 2D-QCA for measurements of lesion severity are shown (Figure 2F–I). Mean time for 3D-QCA was 2.2 ± 1.2 min compared with 1.4 ± 0.9 min for 2D-QCA (P = 0.037).

Figure 2

Correlation between 3D- and 2D-quantitative coronary angiography (QCA) measurements. (A) Minimum luminal area (MLA), (B) percentage area stenosis, (C) minimum luminal diameter (MLD), (D) percentage diameter stenosis and (E) lesion length. (FJ) Corresponding Bland–Altman plots. Broken lines denote the mean and 95% confidence limits.

Accuracy of three- and two-dimensional quantitative coronary angiography in predicting functionally significant stenosis

Receiver operating curve analyses of the predictive value of individual measurements of stenosis severity to predict FFR <0.75 and ≤0.80 are shown in Table 3. Overall, the C statistics tended to be slightly higher for 3D-QCA measurements compared with corresponding 2D-QCA measurements. However, paired comparisons of ROC curves of 3D- and 2D-QCA parameters showed no statistically significant differences between their predictive values. When the prediction of intermediate lesions was considered separately, C statistics for all 3D-QCA parameters remained significantly predictive of FFR <0.75, but 2D-QCA percentage area stenosis and 2D-QCA percentage diameter stenosis were no longer significantly predictive.

View this table:
Table 3

Area under receiver operating curve (C statistic) for prediction of reduced fractional flow reserve using three- and two-dimensional quantitative coronary angiography measurements of stenosis severity

Overall cohort (n = 63)Intermediate lesions (n = 45)
3D-QCA2D-QCA3D-QCA2D-QCA
C statisticP-valueC statisticP-valueC statisticP-valueC statisticP-value
Prediction of FFR <0.75
MLA (mm2)0.86<0.0010.77<0.0010.810.0010.690.038
% Area stenosis0.85<0.0010.76<0.0010.800.0010.670.067
MLD (mm)0.83<0.0010.80<0.0010.760.0040.720.019
% Diameter stenosis0.77<0.0010.78<0.0010.720.0200.680.053
Prediction of FFR <0.80
MLA (mm2)0.79<0.0010.730.0020.700.0270.620.172
% Area stenosis0.750.0010.740.0010.680.0390.660.050
MLD (mm)0.720.0030.76<0.0010.620.1520.630.152
% Diameter stenosis0.630.0500.740.0010.580.3810.670.048
  • 3D, three-dimensional; 2D, two-dimensional; QCA, quantitative coronary angiography; MLA, minimum luminal area; MLD, minimum luminal diameter.

The C statistics for all measurements of lesion severity by 3D- and 2D-QCA was lower when FFR ≤0.80 was used as a cut-off compared with a cut-off of FFR <0.75. There was also a further decline in C statistic for prediction of FFR ≤0.80 for lesions of intermediate severity (Table 3). For intermediate lesions and an FFR cut-off of ≤0.80, the C statistics of most 3D- and 2D-QCA parameters were of borderline or non-significant predictive value, with the exception of 3D-QCA MLA, which appeared to remain significant for all cut-offs and for all lesions studied.

To predict FFR <0.75, the most accurate 3D-QCA measurement was MLA (cut-off <1.60 mm2: sensitivity of 80.0%, specificity 79.2%) (Figure 3A), and the most accurate 2D-QCA measurement was MLD (cut–off <1.25 mm: sensitivity 75.0%, specificity 66.7%) (Figure 3B). There was no statistically significant difference between these curves.

Figure 3

Receiver operating curves of quantitative coronary angiography (QCA) to predict fractional flow reserve<0.75. (A) Three-dimensional QCA minimum luminal area (MLA) in predicting fractional flow reserve of <0.75 and (B) Two-dimensional QCA minimum luminal diameter (MLD) prediction of fractional flow reserve <0.75. AUC, area under curve.

Correlation between fractional flow reserve and quantitative coronary angiography

As shown in Table 4, all 3D- and 2D-QCA measurements of stenosis severity correlated with FFR. However, the correlation coefficients were only modest. Of all 3D-QCA measurements, MLA correlated best with FFR (R = 0.63, P < 0.001). Of all 2D-QCA measurements, MLD best correlated with FFR (R = 0.58, P < 0.001).

View this table:
Table 4

Correlation between fractional flow reserve and angiographic measurements of stenosis severity by three- and two-dimensional quantitative coronary angiography

Spearman's RhoP-valueSpearman's RhoP-value
3D-QCA MLA (mm2)0.63<0.0012D-QCA MLA (mm2)0.49<0.001
3D-QCA % area stenosis−0.59<0.0012D-QCA % Area stenosis−0.52<0.001
3D-QCA MLD (mm)0.55<0.0012D-QCA MLD (mm)0.58<0.001
3D-QCA % diameter stenosis−0.390.0022D-QCA % diameter Stenosis−0.52<0.001
  • 3D, three-dimensional; 2D, two-dimensional; QCA, quantitative coronary angiography; MLA, minimum luminal area; MLD, minimum luminal diameter.

The relationship between FFR and apparent stenosis severity was found to be curvilinear for both 3D-QCA MLA (Figure 4A) and 2D-QCA MLD (Figure 4B). The goodness of fit (R2) for the two curves were similar (R2 = 0.50 for 3D-QCA MLA and R2 = 0.53 for 2D-QCA MLD). The 95% confidence intervals for the correlation between FFR and QCA measurements widened as FFR cut-off increased and QCA diameter increased. Thus, if FFR = 0.75, the mean and 95% confidence intervals for QCA MLA by 3D-QCA was 1.28 ± 0.30 mm2, whereas the mean and 95% confidence intervals for FFR = 0.80 was 1.92 ± 0.88 mm2 (Figure 4A). A similar increase in 95% confidence interval was observed for 2D-QCA MLD (Figure 4B) as FFR cut-off increased.

Figure 4

Regression curves of best fit (solid curves) and 95% confidence intervals (broken curves). (A) Fractional flow reserve (FFR) vs. Three-dimensional quantitative coronary angiography (3D QCA) minimum luminal area. X-axis intercepts (and 95% confidence intervals) corresponding to FFR: 0.75 = 1.28 ± 0.30 and FFR 0.80 = 1.92 ± 0.88 mm2 (arrows). (B) Fractional flow reserve vs. Two-dimensional QCA minimum luminal diameter (MLD). X-axis intercepts (and 95% confidence intervals) corresponding to FFR: 0.75 = 1.21 ± 0.16 and FFR 0.80 = 1.45 ± 0.28 mm (arrows).

There was a trend towards better correlation between QCA and FFR in LAD lesions compared with non-LAD lesions (see Supplementary material online, Table S1), and in smaller vessels compared with larger vessels (see Supplementary material online, Table S2). Within the small number of eccentric lesions, 3D-QCA correlated with FFR, but 2D-QCA did not (see Supplementary material online, Table S3).

Discussion

The accurate assessment of lesion severity is crucial for the appropriate targeting of vessels for revascularization procedures.8 The present study represents the first direct comparison between 3D- and 2D-QCA in predicting functionally significant stenosis. We identify that 3D-QCA tends to be more accurate than 2D-QCA in predicting functionally significant stenosis in lesions of intermediate severity. We demonstrate for the first time that the relationship between FFR and lesion severity is curvilinear, and that the prediction of functional severity by QCA is very dependent on the severity of the stenosis and the FFR cut-off used to identify functional significance.

Limitations of two-dimensional quantitative coronary angiography

Two-dimensional quantitative coronary angiography has several limitations, the major ones including limited spatial and temporal resolution, the inability to account for extraluminal abnormalities, and the inability to properly account for vessel tortuosity and lesion asymmetry.19 Although IVUS3 and optical coherence tomography20 are more accurate in depicting vessel anatomy, 2D-QCA remains the most widely used technique of assessing anatomical lesion severity because of its ease of use.

Utility of three-dimensional quantitative coronary angiography

Three-dimensional quantitative coronary angiography is relatively new technology that allows the fusion of two or more angiographic views to enable 3D reconstruction of the coronary artery of interest.1 Although 3D-QCA is still ‘luminography' and cannot overcome the limitations involved in angiography image resolution, it may be able to better account for vessel tortuosity and lesion asymmetry in the assessment of coronary stenosis.1,2,4,1113

Three-dimensional quantitative coronary angiography can be quickly obtained from existing coronary angiographic images during cardiac catheterization or offline, and only requires that two or more orthogonal, non overlapped images of the target lesion are taken.12 Commercially available software is now accessible that can be incorporated into existing standard catheterization laboratory workstations.1

Quantitative coronary angiography measurements which best correlate with, and predict, fractional flow reserve

Paired comparisons between the ROC curves for 3D-QCA and 2D-QCA yielded no statistically significant differences. However, there was a trend towards higher C statistics for all measurements by 3D-QCA compared with 2D-QCA for predicting functionally significant stenosis, and this was most apparent in lesions of intermediate severity, where 2D-QCA has been known to have limitations.21

Of all 3D- and 2D- measurements, the 3D-QCA MLA was the single most useful predictor of functional significance, across varying cut-offs and varying lesion severity. This can be explained by the fact that of all simple measurements of lesion severity by QCA, the physiological restriction of flow is most directly related to the stenosis area,22 and 2D-derived measurements of area are prone to error.4,11 The accuracy reported in our paper is lower than that previously reported by one study, which assessed the ability of 3D-QCA to predict FFR <0.75 (reported sensitivity 88.9% and specificity 87.5% for percentage area stenosis cut-off 57%) in lesions of intermediate severity by visual assessment.15 However, 3D-QCA measurement revealed that most of the lesions in this previous study were mild with percentage diameter stenosis 35.9 ± 11.2% and only 9 of 41 lesions had FFR <0.75, and this probably caused the increased accuracy reported.

Absolute measurements vs. percentage measurements of stenosis severity

For 3D-QCA in particular, absolute measurements of lesion severity (MLA and MLD) tended to have higher overall C statistics in predicting functionally significant stenosis than percentage measurements. Further, the predictive ability of 2D-QCA percentage area stenosis was not significant, and the predictive ability of 2D-QCA percentage diameter stenosis was of borderline significance in intermediate lesions for predicting FFR <0.75 (Table 3). While not conclusive, these results support previous reports comparing FFR to 2D-QCA9 and to IVUS,23 which indicate that absolute measurements are more reliable correlates of FFR than are percentage measurements. One study that assessed 3D-QCA in lesions judged to be of intermediate severity by visual assessment found that percentage area stenosis was more discriminative of FFR <0.75 compared with MLA.15 However, lesions investigated in this study appeared to be substantially milder compared with the lesions in our patients.

Previous investigators have suggested that although percentage measurements take into account both minimal diameter and normal reference diameter, the fact that two measurements are incorporated may cause an amplification of errors leading to weaker functional depiction of the narrowing.9 Also, percentage measurements require the accurate determination of ‘normal' vessel diameter which can be inaccurate with angiography.19

Relationship between quantitative coronary angiography and fractional flow reserve

We demonstrated a curvilinear relationship between measurements of lesion stenosis with FFR (Figure 4), whereas two previous studies reported a linear relationship between pre-angioplasty FFR and 2D-QCA percentage diameter stenosis and MLD.9,10 It was not specified in both these studies whether progressively complex curves were compared as we have, and scatter plots of pre-angioplasty FFR vs. both percentage diameter stenosis and MLD for both studies appear similar to ours.9,10

The curvilinear relationship has important implications for prediction of FFR by QCA, as the 95% confidence interval in QCA measurements compatible with a given FFR widens as the FFR cut-off increases from 0.75 to 0.80 (Figure 4). The C statistics indicate that prediction of FFR is less accurate for lesions of intermediate severity than the cohort as a whole. In addition, they demonstrate that change of FFR cut-off from 0.75 to 0.80 decreases the predictive accuracy of QCA. Overall, these results indicate that the lesion severity and FFR cut-off used will affect the prediction of FFR by coronary angiography.

It is to be expected that the goodness of fit of the regression curve would be similarly modest for the best 3D-QCA (MLA) and 2D-QCA (MLD) parameters (Figure 4) because actual FFR calculation would incorporate complex 3D vessel lumen geometry, blood flow, and the amount of myocardium perfused by the vessel. For example, angiographically significant stenoses in vessels with low blood flow or small area of myocardium perfused may have non-significant FFR.

There was a trend towards better correlation between QCA and FFR in LAD lesions compared with non-LAD lesions, and in smaller vessels compared with larger vessels. Lesion eccentricity particularly adversely affected 2D-QCA correlations with FFR. In all subgroup analyses, 3D-QCA MLA remained the best overall predictor of FFR. These differences between lesion subgroups results may have implications for the selection of patients in future studies comparing QCA and FFR.

Study limitations

There are several limitations to our study. Firstly, because the FFR is dependent on the size and viability of the distal myocardial bed, this study was limited to one of the three major epicardial coronary arteries (LAD, LCX, and RCA), and the conclusions cannot be extended to branch vessels or vessels of smaller calibre. Secondly, left main coronary stenoses were excluded due to the inherent differences between left main stenosis severity and functional significance.7 Thirdly, due to the fact that microvascular perturbances from myocardial infarction cause discordance between the FFR and angiographic stenosis severity,24 lesions with known infarction in the target vessel territory were excluded. Fourthly, serial stenoses are known to complicate the measurement of FFR,25,26 and therefore were excluded from this study. Lastly, we tested a single 3D-QCA system in this study, which used a manual catheter calibration method to determine pixel size. As such, newer 3D-QCA systems which have automated pixel size determination or other improvements may enhance the performance of 3D-QCA in predicting significant FFR.1

Conclusions

The accuracy of QCA in predicting functionally significant FFR is limited and is dependent on FFR cut-off used and lesion severity. Three-dimensional quantitative coronary angiography shows a non-significant trend towards more accurate prediction of FFR than 2D-QCA, with 3D-QCA MLA being the single best predictor of FFR. If FFR or IVUS are not available, or are contraindicated, 3D-QCA may be particularly helpful for quantifying coronary stenoses visually assessed as being of intermediate severity. However, 3D-QCA should not replace FFR in the functional evaluation of coronary stenosis severity.

Funding

This study is supported by the National Health and Medical Research Council of Australia [Postgraduate Medical Scholarship to A.S.C.Y.].

Conflict of interest: none declared.

Acknowledgements

The authors would like to acknowledge Ms Joan Kong for her help in preparing the angiographic images, Ms Teena West for her help with statistical analyses, and the cardiac catheterization laboratory staff at Concord and Royal Prince Alfred Hospitals for their assistance in performing the studies.

References

View Abstract