The study population included 176 patients consecutively referred to our outpatient cardiology clinic for assessment of CAD between January 2010 and November 2011. Inclusion criteria were age >40 years, symptoms compatible with CAD and ≥2 cardiovascular risk factors or positive or inconclusive treadmill test. Exclusion criteria were clinical instability, known CAD, significant valvular heart disease, atrial fibrillation, pregnancy, creatinine clearance <60 ml/min and standard contraindications for CMR, adenosine and contrast media. Figure 1 summarizes the study design and exclusion criteria.

Figure 1.

Study flow-chart and reasons for exclusions. CAD: coronary artery disease; CMR-Perf: cardiac magnetic resonance perfusion imaging; FFR: fractional flow reserve; ICA: invasive coronary angiography.

(0.41MB).

The final population consisted of 101 individuals with an intermediate to high pretest probability (Table 1). The local research ethics committee approved the study protocol and written informed consent was obtained from all patients.

CTA and CMR perfusion imaging studies were performed in the week previous to ICA. Patients were instructed to refrain from smoking, consuming coffee or tea, or taking theophyllines, beta-blockers, calcium antagonists or nitrates in the previous 24 hours. Readers were blinded to the clinical information and other test results. The decision for coronary revascularization was the responsibility of the patients’ attending clinicians, who were not bound by standard FFR criteria.

The CTA protocol used a 64-slice SOMATOM Sensation 64 scanner (Siemens Medical Solutions, Forchheim, Germany) as part of a previously described comprehensive stress-rest protocol22 optimized to study stress CTP.

The study began with a prospective ECG-gated acquisition without contrast (120 kV, 190 mAs) for coronary calcium scoring, which was used to guide subsequent acquisitions. Infusion of adenosine at 140 μg/kg/min was then initiated. After at least 3 min, the first pass of iodinated contrast (Ultravist, iopromide 370 mg/ml, Bayer Healthcare, Leverkusen, Germany) injected at 4.5 ml/s, was registered by retrospective acquisition with tube current modulation (maximum power 60–65% of the R-R interval) and bolus-tracking (150 HU threshold, delay 4 s.) The collimation achieved was 64 mm×0.6 mm. The full tube current was adjusted to body mass index (600 mAs for 20–30 kg/m2, 800 for >30 kg/m2 and 400 for <20 kg/m2).

Immediately after this acquisition, adenosine infusion was suspended. If heart rate exceeded 65 bpm (44 patients), intravenous boluses of fractionated metoprolol (5–20 mg) were administered for a target rate of ≤60 bpm. Five minutes before CTA all patients were given 0.05 mg of sublingual nitroglycerin. Ten minutes after the stress acquisition, a new prospective ECG-gated acquisition was performed (65% of the R-R interval, 100 kV, 110 mAs). The timing and amount of contrast used were similar to the stress acquisition.

Although the study was optimized for CTP, as this was a concomitant target of our research center, only coronary angiography data were analyzed for the purposes of this study.22 Acquisitions both under stress and at rest were reconstructed using a soft kernel (Siemens B25f) with 0.6 mm slice thickness. The compiled data were blinded and sent to a postprocessing station (Aquarius, TeraRecon Inc., CA, USA) for interpretation by two cardiologists experienced in cardiovascular CT (NF and NB). A modified American Heart Association 17-segment model was used for segmentation of the coronary arteries.26 Each coronary segment was graded as normal, non-significant stenosis (<50%), 50–70% stenosis, ≥70% stenosis/occlusion or uninterpretable. The effective radiation dose was calculated by multiplying the thoracic radiation factor (k=0.014 mSv mGy−1 cm−1) by the dose-length product obtained for each acquisition.

The CMR protocol was performed using a 1.5 T Siemens MAGNETOM Symphony TIM scanner (Siemens, Erlangen, Germany) with a 12-channel head coil (six anterior coils plus six spine coils in the patient table).21 No premedication was administered.

After acquisition of scout images in the orthogonal planes, three short-axis slices (basal, mid and apical) were planned. After 3 min of intravenous adenosine infusion (140 μg/kg/min) to achieve maximal hyperemia, a bolus of 0.07 mmol/kg gadobutrol (Gadovist, Bayer HealthCare Pharmaceuticals, Berlin, Germany) was injected at a rate of 4 ml/s with a power injector (Medrad Europe, Maastricht, The Netherlands). The first pass of gadolinium contrast in these slices was recorded over 50 cycles by a gradient echo sequence with a shared single saturation pre-pulse and during an end-expiratory breath hold. Typical sequence parameters were: TE 1.18 ms, TR 192 ms, TI 110 ms, flip angle 12°, slice thickness 10 mm; FOV 290–460 mm, matrix 192 mm×128 mm, spatial in-plane resolution 1.5–2.4 mm and bandwidth 789 Hz per pixel. Any occurrence of symptomatic, hemodynamic or arrhythmic events during adenosine infusion was recorded. Shortly after the end of stress image acquisition, adenosine infusion was stopped.

For volumetric and functional assessment, long- and short-axis cine images were obtained using a steady-state free precession sequence with retrospective gating during an end-expiratory breath hold. Ten minutes after the first-pass gadolinium assessment, another acquisition was performed at rest with the same stress parameters. To achieve the full dose of 0.2 mmol/kg of gadolinium, the remaining 0.06 mmol/kg were then administered.

Ten minutes after the last contrast injection, delayed-enhanced images were acquired in breath-hold with a 2D phase-sensitive inversion-recovery sequence.

Tests were also blinded and evaluated by two independent readers with Society for Cardiovascular Magnetic Resonance (SCMR) level III training (AC and AS). When needed, a third reader, also with SCMR level III training (NB), acted to reach a consensus. Perfusion defects were defined as transmural or subendocardium-limited areas of lower signal intensity compared with adjacent healthy myocardium, lasting at least 10 frames. The stress and rest acquisitions were viewed simultaneously to better identify artefacts. In circumferential perfusion defects, the perfusion gradient from epicardium to endocardium was also assessed. Areas of hypoperfusion were matched to myocardial segments according to the classic 17-segment model, excluding the apex as not being displayed in CMR perfusion imaging.27 Each of the 16 segments was classified using a 4-point scoring system (from no defect to transmural defect). Late gadolinium enhancement was analyzed simultaneously, distinguishing inducible ischemia from scar. Isolated changes in myocardial kinetics or scarring without further ischemia were not considered as ischemia. Image quality and the interpreters’ degree of trust were classified into four grades from poor to excellent and from very unsure to very confident, respectively.

ICA and FFR assessment were performed according to standard techniques. Readers were unaware of CTA and CMR perfusion imaging data. When coronary segments with visual stenosis >40% were observed, a pressure wire (Certus, St. Jude Medical, MN, USA) was used to determine the corresponding FFR under steady-state hyperemia with intravenous adenosine infusion (140 μg/kg/min over 3–6 minutes) using the RadiAnalyzer system (St. Jude Medical, MN, USA). Significant stenosis was defined as FFR <0.80, occlusion or subocclusion, or ≥50% left main stenosis.

For vessel-based analysis, each myocardial segment with a perfusion defect on CMR perfusion imaging was matched to a major vascular territory: left anterior descending (LAD), circumflex (Cx) or right coronary artery (RCA). This association was carried out after checking ICA for the coronary anatomical pattern.

To assess the potential of integrated noninvasive anatomical and functional assessment, similar to the principle of the use of FFR in ICA, significant lesions were defined as >50% stenosis on CTA matched with a perfusion defect on CMR perfusion imaging in the corresponding segments. Less than 50% stenosis on CTA without perfusion defects on CMR perfusion imaging, and vice versa, were considered negative (see Figure 2).

Figure 2.

Algorithm for classification of stenosis by integration of CTA and CMR perfusion imaging data. CMR: cardiac magnetic resonance; CMR-Perf: cardiac magnetic resonance perfusion imaging; CTA: computed tomography angiography.

(0.25MB).

Follow-up was completed for all patients, with a mean period of 2.9±0.6 years (0.3–3.8 years). The follow-up was conducted by telephone and/or by letter. Data were collected regarding current health status and cardiovascular events, including death, acute coronary syndrome, elective and urgent coronary revascularization, stroke or unplanned hospitalization for heart failure or other cardiovascular causes.

The diagnostic performance of CTA and CMR perfusion imaging, alone and integrated, for the detection of functionally significant CAD was compared using FFR as the reference standard. No segments interpretable by CTA were classified as positive for disease in isolated analysis, but when integrated they were considered positive or negative according to the perfusion data. All continuous variables were expressed as mean ± standard deviation, while categorical variables were expressed as a percentage. The McNemar test was used to calculate the differences between proportions (sensitivity, specificity and accuracy) obtained from paired observations. Cohen's kappa was used to assess intermodality and intra- and interobserver variability. Receiver operating characteristic (ROC) curve analysis was used to compare the diagnostic accuracy and prediction of revascularization of CTA and CMR perfusion imaging, alone or integrated. Areas under the curve (AUCs) were compared using DeLong's method.28

Statistical analyses were performed using SPSS version 18.0 (SPSS, Chicago, IL, USA) and MedCalc for Windows, version 12.7.7 (MedCalc Software, Ostend, Belgium). A p value <0.05 was considered statistically significant.

Image quality was rated poor in nine patients, moderate in 39, good in 52 and excellent in one, and readers were unsure in 49 cases, confident in 47 and very confident in five. Mean radiation exposure in the CTA protocol, including coronary calcium scoring, was 5.47±0.96 mSv. Thirty-three (33%) patients had at least one non-interpretable segment, often due to the presence of extensive calcification. Among patients who had fully interpretable scans, 10 had no atherosclerotic disease, 25 had mild disease (<50% stenosis) and 33 had ≥50% stenosis. When non-interpretable segments were recorded as having significant stenosis, 65 patients were classified as having significant CAD.

Image quality was rated poor in two patients, moderate in 20, good in 57 and excellent in 22, and readers were unsure in the diagnosis of eight patients, confident in 60 and very confident in 33. Forty-six patients (70 out of 303 vascular territories) showed perfusion defects suggestive of ischemia during vasodilator stress. Thirty were within the LAD territory, 15 in the Cx territory and 25 in the RCA territory. Sixteen patients had late gadolinium enhancement compatible with ischemic etiology. Intra- and inter-observer variability was good, with kappa values of 0.71 and 0.57, respectively.

Mean left ventricular ejection fraction was 65.4±10.8%.

Fifty-four patients with >40% visually estimated stenosis were considered for FFR assessment. Coronary arteries with mild (≤40%) disease (29 vessels) or without plaque (179 vessels) did not undergo FFR measurement. Also, vessels with luminal diameter <2 mm (10 vessels) were excluded. Nineteen totally occluded vessels and 11 vessels with subocclusion did not undergo FFR measurement and were regarded as positive. Additionally, FFR measurement was not performed in nine vessels with long lesions and severely calcified and tortuous or decreased flow after intracoronary nitrate injection (TIMI flow ≤2). It was also not measured in five patients with left main disease because the potential risk was perceived as excessive. Lesions in which FFR could not be measured for the above reasons were considered positive for comparative purposes. In total, 36 non-occluded vessels (27 patients) were assessed by FFR.

Using this approach, 72 arteries were classified as positive, with a diagnosis of CAD in 44 patients (43.6% of the population): 24 with 1-vessel disease, 12 with 2-vessel disease and eight with 3-vessel disease.

In the follow-up only three major cardiovascular adverse events (three patients) occurred: one acute coronary syndrome with concomitant urgent revascularization and two deaths (one from cardiovascular cause). Thirty-six patients (35.6% of the sample) were referred for coronary revascularization, 23 (63.8%) by percutaneous coronary angioplasty and 13 (36.1%) by surgery. Two revascularized patients had no functionally significant disease by FFR and CMR perfusion imaging, and the decision was based on purely anatomical findings.

Diagnostic assessment by CTA had excellent sensitivity and NPV (100%) for functionally significant CAD. As expected, the specificity and PPV were comparatively low (61% and 67%, respectively).

Of the 46 patients with perfusion defects suggestive of ischemia on CMR perfusion imaging, 39 presented functionally significant disease in the invasive assessment. Of the 55 patients with negative CMR perfusion imaging exams, 50 were true negatives. CMR perfusion imaging had good diagnostic performance, with sensitivity of 89% and specificity of 88%, and with PPV of 85% and NPV of 91%.

When CTA and CMR perfusion imaging were integrated (CT+CMRint), sensitivity and NPV remained high (89% and 92%, respectively) compared with CMR perfusion imaging alone. Specificity and PPV increased considerably (95% and 93%, respectively). However, the differences in specificity were not statistically significant (p=0.1250). Four patients without disease on invasive assessment and with perfusion defects on CMR perfusion imaging were correctly reclassified as not having >50% stenosis on CTA, while 19 of 22 patients without significant disease by an invasive approach and at least one vessel with >50% stenosis on CTA were correctly reclassified as not having significant functional CAD since they had no compatible perfusion defects on CMR perfusion imaging.

Diagnostic accuracy for detection of CAD with functional significance by FFR as reference was 78% for CTA, 88% for CMR perfusion imaging and 92% for CT+CMRint. When C-statistics were calculated to compare the diagnostic accuracy of the various methods, the integrated protocol showed statistically significant superiority to predict functionally significant CAD by FFR (AUC=0.917, 95% CI: 0.85–0.96) compared with either CTA (AUC=0.807, 95% CI: 0.72–0.88, p=0.0057) or CMR perfusion imaging (AUC=0.882, 95% CI: 0.80–0.94, p=0.0398) alone (Figure 3).

Figure 3.

Receiver operating characteristic curve analysis of diagnostic accuracy of the different protocols with fractional flow reserve ≤0.80 as reference. CMR-Perf: cardiac magnetic resonance perfusion imaging; CT+CMRint: integrated CTA and CMR perfusion imaging; CTA: computed tomography angiography.

(0.12MB).

Patient-based performances are summarized in Tables 2 and 3.

A total of 303 vessels (101 patients) were analyzed. CTA had greater sensitivity and NPV than CMR perfusion imaging in the detection of functionally significant CAD (95% vs. 79% and 97% vs. 93% respectively). With regard to specificity and PPV, CMR perfusion imaging performed better (93% vs. 67% and 79% vs. 48%, respectively). CMR perfusion imaging tended to perform better (AUC=0.87, 95% CI: 0.82–0.90) than CTA (AUC=0.81, 95% CI: 0.79–0.88, p=0.0811) and the integrated protocol (AUC=0.84, 95% CI: 0.75–0.85, p=0.3510), but without statistical significance.

Vessel-based performances are summarized in Tables 2 and 4.

ROC curves were also constructed to test the ability of the various modalities to predict the occurrence of coronary revascularization during follow-up. The integrated protocol (AUC=0.824, 95% CI: 0.70–0.87) had superior performance compared to CTA (AUC=0.769, 95% CI: 0.68–0.85, p=0.5979) and CMR perfusion imaging (AUC=0.794, CI 95%: 0.70–0.87, p=0.0405) alone, and was statistically significant for the latter.

However, as surgical revascularization was not triggered by FFR targets, anatomical assessment had more influence compared to functional data.