We sought to evaluate the diagnostic accuracy of noninvasive coronary angiography using 320-detector row computed tomography, which provides 16-cm craniocaudal coverage in 350 ms and can image the entire coronary tree in a single heartbeat, representing a significant advance from previous-generation scanners. We evaluated 63 consecutive patients who underwent 320-detector row computed tomography and invasive coronary angiography for the investigation of suspected coronary artery disease. Patients with known coronary artery disease were excluded. Computed tomographic (CT) studies were assessed by 2 independent observers blinded to results of invasive coronary angiography. A single observer unaware of CT results assessed invasive coronary angiographic images quantitatively. All available coronary segments were included in the analysis, regardless of size or image quality. Lesions with >50% diameter stenoses were considered significant. Mean heart rate was 63 ± 7 beats/min, with 6 patients (10%) in atrial fibrillation during image acquisition. Thirty-three patients (52%) and 70 of 973 segments (7%) had significant coronary stenoses on invasive coronary angiogram. Seventeen segments (2%) were nondiagnostic on computed tomogram and were assumed to contain significant stenoses on an “intention-to-diagnose” analysis. Sensitivity, specificity, and positive and negative predictive values of computed tomography for detecting significant stenoses were 94%, 87%, 88%, and 93%, respectively, by patient (n = 63), 89%, 95%, 82%, and 97%, respectively, by artery (n = 260), and 87%, 97%, 73%, and 99%, respectively, by segment (n = 973). In conclusion, noninvasive 320-detector row CT coronary angiography provides high diagnostic accuracy across all coronary segments, regardless of size, cardiac rhythm, or image quality.
The latest-generation 320-detector row computed tomographic (CT) system represents a significant advance from 64-detector row technologies. It enables 16-cm craniocaudal coverage in a rotation time of 350 ms, imaging the entire coronary tree in a single gantry rotation within a single heartbeat. Although these technical improvements allow axial volumetric imaging with single-beat image acquisition, uniform contrast enhancement, and an ability to study patients in atrial fibrillation (AF) with excellent image quality, limited data exist regarding implications for diagnostic accuracy. We therefore sought to determine the diagnostic accuracy of 320-detector row computed tomography in the detection of significant coronary artery stenoses across all coronary segments, regardless of size or cardiac rhythm, using an “intention-to-diagnose” analysis that accounted for all coronary segments.
Methods
We evaluated 63 eligible patients who underwent multidetector computed tomography and invasive coronary angiography (ICA) within 35 days with no known coronary artery disease (including percutaneous coronary intervention or coronary artery bypass surgery), impaired renal function (serum creatinine >110 μmol/L), contrast allergy, or pregnancy. Patients were not excluded due to arrhythmia. This group of 63 was selected after screening 1,048 consecutive patients referred to our tertiary hospital from October 2008 to April 2009 for investigation of suspected coronary artery disease with ICA (655 patients) or multidetector computed tomography (393 patients). Of the 655 patients undergoing initial ICA, 23 subsequently underwent multidetector computed tomography within 35 days, and 20 were enrolled (3 underwent subsequent multidetector computed tomography for evaluation of long-term total occlusions, 6 for suspected anomalous coronary circulation, and 11 as part of a workup for cardiac surgery), with 3 excluded due to previous coronary artery bypass surgery. Of the 393 patients undergoing initial multidetector computed tomography, 68 subsequently underwent early ICA because of clinical history and multidetector CT (MDCT) results. Forty-three of these patients were enrolled after excluding 11 with previous percutaneous coronary intervention and 14 with previous coronary artery bypass surgery. The local institutional ethics committee approved the study and all patients gave informed consent.
One hour before multidetector computed tomography, patients with a heart rate >60 or 70 beats/min received a single oral dose of metoprolol 50 or 100 mg, respectively. Patients with heart rates >65 beats/min at the time of scanning received additional intravenous metoprolol aiming to achieve a heart rate <65 beats/min. Nitroglycerin 400 μg sublingually (Nitrolingual Pumpspray, Sanofi-Aventis, Sydney, Australia) was administered 1 minute before contrast injection.
All studies were performed on a 320-detector row system (Aquilion ONE, Toshiba Medical Systems, Tokyo, Japan). Studies were preceded by biplanar scout acquisition and did not include calcium scoring. A bolus of iopromide 75 ml (Ultravist 370, Bayer HealthCare, Berlin, Germany) was injected into an antecubital vein at a flow rate of 5 to 6 ml/s, followed by a 50-ml saline chaser. Scanning was triggered in the arterial phase using automated contrast bolus tracking, with a region of interest placed in the left ventricle manually triggered at a threshold of 180 HU. An axial scanning technique was used with slice collimation 0.5 mm. No concurrent table movement was required because the scanner provided 16-cm craniocaudal coverage. The actual number of detectors selected, hence, the volume coverage, was determined by cardiac size seen on biplanar scout images (16 cm using 320 detectors or 15 cm using 300 detectors). Exposure parameters included an x-ray tube potential of 100 to 135 kVp and effective tube current of 400 to 580 mA, based on vendor specifications and the patient’s body mass index. All scans were performed with either prospective electrocardiographic (ECG) triggering using 60% to 100% phase window or, in patients with an indication for evaluation of cardiac function, full-beat retrospective ECG triggering using tube current modulation. Mean effective radiation exposure was derived from the dose–length product multiplied by a conversion coefficient for the chest (κ = 0.014 mSv/mGy).
Regardless of rhythm, when images were acquired at heart rates <65 beats/min, scanning was completed within a single RR interval using a 180° segment, thus providing an effective temporal resolution of 175 ms and minimizing radiation exposure. For dose-modulated single-beat scans, a high milliampere window of 70% to 80% of the RR interval was used with the low modulated milliampere range set to 25% of the high milliampere setting. A switch to multisegment acquisition was automatically triggered by the system’s “arrhythmia rejection” software when heart rate was >65 beats/min, resulting in image acquisition over 2 or 3 RR intervals and thus improved temporal resolution at the expense of increased radiation exposure. Where dose modulation was used in multibeat studies, a high milliampere window of 40% to 80% was used with the low modulated milliampere range set to 25% of the high milliampere setting. Electrocardiographically gated datasets were reconstructed automatically to overlapping 0.5-mm slices in 0.25-mm intervals at 75% of the RR cycle length. Additional reconstruction windows were constructed after examination of initial datasets if motion or noise artifacts were present. A default “medium soft tissue” reconstruction kernel (FC04) was used.
Anonymized MDCT datasets were transferred to a postprocessing workstation (Vitrea FX 2.0, Vital Images, Minnetonka, Minnesota). Images were analyzed independently by 2 experienced readers unaware of clinical data and blinded to results of ICA using the 17-segment modified American Heart Association model. A third expert reader adjudicated where disagreement was present. All available coronary segments were assessed regardless of size, excluding those distal to total occlusions. Each lesion identified was examined using maximum intensity projections, multiplanar reconstructions, and axial slices along multiple longitudinal and transverse axes and visually analyzed for the presence or absence of significant stenosis (defined as maximal vessel diameter decrease >50%) in any plane.
Image quality was evaluated on a per-segment basis and classified on a 3-point scale: (1) excellent (absence of artifacts related to calcification, motion, or noise), (2) suboptimal but diagnostic (presence of artifacts but evaluation possible), and (3) nondiagnostic (severe image-degrading artifacts). In suboptimal and nondiagnostic segments, the predominant artifact was identified as calcification, motion, or poor signal-to-noise ratio. Arterial wall calcification was semiquantitatively assessed in all segments as (1) absent or affecting <10% of the vessel luminal cross-sectional area, (2) affecting 10% to 50% of the vessel luminal cross-sectional area, or (3) affecting >50% of the vessel luminal cross-sectional area.
ICA was performed according to standard techniques. Images were evaluated by an experienced cardiologist blinded to MDCT results, using the same 17-segment model as MDCT analysis. Quantitative analysis was performed on all lesions with visually estimated >20% stenoses with an automated edge detection system (QCA-CMS, Medis, Nuenen, The Netherlands) and, similar to multidetector computed tomography, classified as significant if maximal vessel diameter decrease by quantitative coronary angiography was >50% in any angiographic view.
Diagnostic accuracy of multidetector computed tomography to detect significant coronary stenoses was compared to ICA as the reference standard and presented as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio, and negative likelihood ratio. Precision of these parameters was expressed as 95% confidence intervals. Analysis was performed on an “intention-to-diagnose” basis, denoting that nondiagnostic coronary segments on multidetector computed tomogram due to any artifact were considered to contain significant stenoses (>50% luminal narrowing), reflecting real-world practice. Agreement between the 2 MDCT readers was calculated using Cohen kappa statistics.
Results
Patient clinical characteristics are presented in Table 1 . Median interval between multidetector computed tomography and ICA was 2 days (interquartile range 1 to 6) and there were no clinical events between the 2 studies in any patient. Forty-five (71%) investigations were initially performed in patients with chest pain, 8 (13%) in preoperative patients awaiting valvular surgery to detect or exclude associated coronary stenoses, 6 (10%) after an equivocal exercise stress test, and 4 (6%) in patients with dyspnea.
| Characteristic | Value |
|---|---|
| Age (years), mean ± SD (range) | 63.2 ± 14 (31–85) |
| Men | 38 (60%) |
| Height (cm), mean ± SD | 169 ± 14 |
| Weight (kg), mean ± SD | 79.1 ± 15 |
| Body mass index (kg/m 2 ), mean ± SD (range) | 27.8 ± 5 (18–40) |
| Diabetes mellitus | 9 (14%) |
| Hypertension ⁎ | 33 (52%) |
| Hypercholesterolemia † | 37 (59%) |
| Current smoker | 11 (17%) |
| Obesity ‡ | 21 (33%) |
| Pretest likelihood, median (interquartile range) | 0.46 (0.13–0.67) |
⁎ Blood pressure >140/90 mm Hg or treatment for hypertension.
† Total cholesterol >180 mg/dl or treatment for hypercholesterolemia.
Fifty-seven patients (90%) were in sinus rhythm and 6 (10%) in AF during image acquisition. Forty-nine patients (78%) required β blockade, with 30 (48%) requiring oral metoprolol only (mean dose 54 ± 48 mg) and 19 (30%) requiring additional intravenous metoprolol (mean dose 7 ± 13 mg). Overall mean heart rate during image acquisition was 63 ± 7 beats/min (range 49 to 78), with patients in sinus rhythm and AF having mean heart rates 62 ± 7 beats/min (range 49 to 78) and 68 ± 8 beats/min (range 55 to 76), respectively. Overall, 41 patients had a single-beat acquisition (1 patient with AF), 20 had a 2-beat acquisition (4 with AF), and 2 had a 3-beat acquisition (1 with AF). Full-beat retrospective ECG triggering using tube current modulation was performed in 51 patients (81%), in whom cardiac function was also assessed. The remaining 12 patients (19%) underwent imaging with prospective ECG triggering. Overall median effective radiation dose was 10.6 mSv (interquartile range 8.0 to 13.8), comprising 5.4 mSv (interquartile range 4.4 to 6.3) for prospectively electrocardiographically triggered scans, 12.4 mSv (interquartile range 9.8 to 14.5) for retrospectively electrocardiographically triggered scans, 9.6 mSv (interquartile range 7.1 to 11.3) for single-beat acquisition scans, and 14.2 mSv (interquartile range 9.3 to 15.0) for multibeat acquisition scans. There were no complications associated with multidetector computed tomography, whereas there was 1 major complication (retroperitoneal hematoma requiring emergency surgery) associated with ICA.
ICA identified 70 significant stenoses involving 53 arteries in 33 patients (prevalence 52% per patient, 20% per vessel, and 7% per segment). Single-vessel disease was present in 16 of 63 patients (25%), 2-vessel disease in 14 of 63 (22%), and 3-vessel disease in 3 of 63 (5%). No significant left main stenosis was present.
In total 973 segments were available for analysis. Although 1,071 segments could potentially have been present (17 segments per patient), 98 segments were not visualized on invasive coronary angiogram (75 due to variations in coronary anatomy that were also not visualized with multidetector computed tomography, 15 due to inability to selectively engage the right coronary artery, and 8 due to proximal occlusion and poorly filled distal segments by collateral circulation). Sensitivity, specificity, PPV, and NPV of multidetector computed tomography for detecting significant stenoses per-segment were 87%, 97%, 73%, and 99%, respectively ( Table 2 ). The kappa value for interobserver variability for detection of significant stenoses was 0.71, indicating good agreement.
| Variable | Number | Prevalence of Disease | Sensitivity (95% CI) | Specificity (95% CI) | PPV (95% CI) | NPV (95% CI) | LR+ | LR− |
|---|---|---|---|---|---|---|---|---|
| Patient-based | 63 | 52% | 94% (78–99) | 87% (68–96) | 88% (72–96) | 93% (75–99) | 7.0 | 0.07 |
| Coronary artery-based | 260 | 20% | 89% (76–95) | 95% (91–98) | 82% (70–91) | 97% (93–99) | 18.4 | 0.12 |
| Left main | 63 | 0% | — | 100% (93–100) | — | 100% (93–100) | — | — |
| Left anterior descending | 63 | 37% | 91% (70–98) | 93% (79–98) | 88% (67–97) | 95% (81–99) | 12.2 | 0.09 |
| Left circumflex | 63 | 19% | 75% (43–93) | 94% (83–98) | 75% (43–93) | 94% (83–98) | 12.8 | 0.27 |
| Right coronary artery | 60 | 30% | 94% (71–100) | 93% (79–98) | 85% (61–96) | 98% (85–100) | 13.2 | 0.06 |
| Ramus intermedius | 11 | 0% | — | 91% (57–100) | — | 100% (66–100) | — | — |
| Segment-based | 973 | 7% | 87% (76–94) | 97% (96–98) | 73% (62–82) | 99% (98–100) | 34.2 | 0.13 |
From a total of 70 significant stenoses detected on invasive coronary angiogram, 61 were correctly identified with multidetector computed tomography ( Figure 1 ). Four lesions were not identified by multidetector computed tomography in the presence of poor signal-to-noise ratio (1 in a midright coronary artery, 1 in a distal left anterior descending coronary artery, 1 in a first diagonal branch, 1 in a first obtuse marginal branch), 2 in the presence of motion artifact (1 in a first diagonal branch, 1 in a posterior descending artery), and 1 in the presence of severe calcification in a first obtuse marginal branch. There was no artifact in 2 missed lesions (1 in an obtuse marginal branch, 1 in a posterior descending artery), although these were small-caliber vessels (1.5 and 2 mm, respectively). Twenty-three significant stenoses identified on multidetector computed tomogram were false-positive findings, all in the presence of suboptimal image quality. Nine were located in significantly calcified segments (2 in a proximal left anterior descending coronary artery, 2 in a midleft anterior descending coronary artery, 1 in a distal left anterior descending coronary artery, 1 in a proximal right coronary artery, 1 in a midright coronary artery, 1 in a first diagonal branch, 1 in a second diagonal branch), 7 in segments with poor signal-to-noise ratio (2 in a proximal left circumflex coronary artery, 2 in a distal left anterior descending coronary artery, 1 in a first diagonal branch, 1 in a second diagonal branch, 1 in a posterior descending artery), and 7 in nondiagnostic segments according to an “intention-to-diagnose” analysis.
Any vessel that contained ≥1 nondiagnostic segment was considered to contain a significant stenosis according to an intention-to-diagnose analysis, and thus no vessels were excluded. Sensitivity, specificity, PPV, and NPV of multidetector computed tomography for detecting significant stenoses per-vessel were 89%, 95%, 82%, and 97%, respectively ( Table 2 ). All left main coronary arteries were correctly identified as not containing significant stenoses, with specificity and NPV 100%. The left circumflex coronary artery had the lowest sensitivity of all vessels (75%), with 3 of 12 significant stenoses missed on multidetector computed tomogram in obtuse marginal branches, 1 in the presence of poor signal-to-noise ratio, 1 in the presence of severe calcification, and 1 in a small-caliber 1.5-mm vessel.
All patients were included in the analysis because any patient who had ≥1 nondiagnostic segment was considered to have significant coronary stenosis according to an intention-to-diagnose analysis. Multidetector computed tomography correctly identified significant stenoses in 31 of 33 patients (sensitivity 94%) and correctly excluded stenoses in 26 of 30 patients (specificity 87%).
Seventeen of 973 segments (2%) were nondiagnostic in 9 patients, meaning that a significant stenosis could not be confidently identified or excluded. This was due to severe calcification in 14 segments (4 in first obtuse marginal branches, 2 in a distal left anterior descending coronary artery, 2 in second diagonal branches, 1 in a proximal left anterior descending coronary artery, 1 in a midleft anterior descending coronary artery, 1 in a proximal right coronary artery, 1 in a midright coronary artery, 1 in a first diagonal branch, 1 in the ramus intermedius) and poor signal-to-noise ratio in 3 segments (1 in a midleft circumflex coronary artery, 1 in a distal left circumflex coronary artery, 1 in a posterior descending artery). Actual prevalence of significant stenoses in these nondiagnostic segments was 10 of 17 (59%) as identified by quantitative coronary angiography.
Overall, 956 of 973 segments (98%) were of diagnostic quality, with image quality rated as excellent in 667 of 973 segments (68%) and suboptimal but diagnostic in 289 of 973 segments (30%). Reasons for lower image quality were severe calcification in 60 of 306 (20%), motion artifact in 74 of 306 (24%), and poor signal-to-noise ratio in 172 of 306 (56%). There was a decrease in overall image quality from proximal to distal segments and side branches, with excellent image quality in 77% of proximal segments, 72% of midsegments, 66% of distal segments, and 62% of side branches. The most common artifacts were severe calcification ( Figure 2 ) and poor signal-to-noise ratio in proximal and midsegments and poor signal-to-noise ratio in distal segments and side branches ( Table 3 ).
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
