Author + information
- Received January 19, 2009
- Revision received April 16, 2009
- Accepted April 26, 2009
- Published online July 21, 2009.
- Gianluca Pontone, MD⁎ (, )
- Daniele Andreini, MD,
- Antonio L. Bartorelli, MD,
- Sarah Cortinovis, MD,
- Saima Mushtaq, MD,
- Erika Bertella, MD,
- Andrea Annoni, MD,
- Alberto Formenti, MD,
- Enrica Nobili, MD,
- Daniela Trabattoni, MD,
- Piero Montorsi, MD,
- Giovanni Ballerini, MD,
- Piergiuseppe Agostoni, MD, PhD and
- Mauro Pepi, MD
- ↵⁎Reprint requests and correspondence:
Dr. Gianluca Pontone, Centro Cardiologico Monzino, IRCCS, Department of Cardiovascular Sciences, University of Milan, Via Carlo Parea 4, 20138 Milan, Italy
Objectives The aim of this study was to compare the diagnostic performance of multidetector computed tomography (MDCT) with prospective electrocardiogram (ECG) triggering versus retrospective ECG triggering.
Background MDCT allows the noninvasive visualization of the coronary arteries. However, radiation exposure is a reason for concern.
Methods One hundred eighty consecutive patients scheduled for invasive coronary angiography were enrolled in this study. Twenty patients were excluded due to contraindications to sustain MDCT. Of the 160 remaining patients, 80 were studied with MDCT with prospective ECG triggering (Group 1) and 80 with a retrospective ECG triggering (Group 2). The individual radiation dose exposure was estimated.
Results In nonstented segments, the evaluability of Groups 1 and 2 was 96% versus 97%, respectively (p = 0.05), the accuracy in segment-based model was 93% versus 96%, respectively (p < 0.05) including diagnostic segments and 91% versus 94%, respectively (p < 0.01) including all segments, whereas the accuracy in a patient-based model was 98% in both groups. In stented segments the evaluability in Groups 1 and 2 was 92% versus 94%, respectively, and the accuracy was 93% versus 92%, respectively, including diagnostic stented segments and 90% versus 89%, respectively, including all stented segments. Group 1 presented lower radiation dose compared with Group 2 (5.7 ± 1.5 mSv vs. 20.5 ± 4.3 mSv, p < 0.01).
Conclusions Prospective ECG-triggering computed tomography allows an accurate detection of coronary stenosis, despite a slight reduction of diagnostic performance, with a low radiation dose.
The 64-slice multidetector computed tomography (MDCT) allows noninvasive visualization of the coronary arteries (1) as an alternative imaging modality to invasive coronary angiography (ICA) in detection of coronary artery disease (CAD). However, radiation exposure is a reason for concern. Different strategies have been proposed to reduce the radiation dose, including the optimization of scan parameters (2), introduction of dual-source MDCT (3), and increase of slice number (4,5). More recently, prospective electrocardiogram (ECG) triggering has been rediscovered as an alternative to retrospective ECG triggering (6–10). However, there are no studies comparing MDCT with prospective ECG triggering versus MDCT with retrospective ECG triggering. Therefore, the aims of this study were to: 1) compare the evaluability and accuracy of MDCT with prospective ECG triggering versus retrospective ECG triggering in the detection of CAD; and 2) determine the reduction of radiation exposure achieved.
We enrolled 180 patients, on the basis of the availability of computed tomography slots, randomly drawn from a cohort of 1,675 consecutive patients scheduled for ICA for suspected CAD between January 2008 and June 2008. Exclusion criteria were contraindications to contrast agents, impaired renal function (creatinine clearance <60 ml/min), inability to sustain a 15-s breath hold, pregnancy, heart rate (HR) >65 beats/min despite beta-blockade treatment, cardiac arrhythmias, and previous coronary artery bypass graft surgery. A total of 14 patients were excluded due to inability to sustain breath (n = 2), impaired renal function (n = 5), and cardiac arrhythmias (n = 7). Patients were divided into 2 groups with a computer-generated randomized process and subsequently underwent MDCT-scan protocol with either prospective (n = 84, Group 1) or retrospective (n = 82, Group 2) ECG triggering (Fig. 1).All patients were studied by ICA within 3 days after MDCT. Written informed consent was obtained from all patients, and the study protocol was approved by the institutional ethical committee.
Patient preparation for 64-MDCT coronary angiography
In all patients with resting HR >65 beats/min before MDCT, metoprolol was intravenously administered with a titration dose up to 15 mg to achieve a target HR ≤65 beats/min. Six patients were excluded from the study because target HR was not obtained (4 cases in Group 1 and 2 cases in Group 2) as shown in Figure 1. Before the scan, each patient performed a breath hold test to evaluate the heart rate variability (HRv) calculated as the SD from the average HR (10).
The MDCT exams were performed with a LightSpeed VCT XT Scanner (GE Healthcare, Milwaukee, Wisconsin). An initial unenhanced scan was performed for calcium score with the exception of patients with stent implantation. Then, in all patients MDCT was performed with the following parameters: slice configuration 64 × 0.625 mm, gantry rotation time 350 ms, tube voltage 120 KVp, effective tube current 700 mAs. The patients received a 90 ml bolus of contrast medium (Iomeron 400 mg/ml, Bracco, Milan, Italy) through an antecubital vein at an infusion rate of 5 ml/s, followed by 50 ml of saline solution. The scan was performed according to the bolus tracking technique.
In patients from Group 1, we applied a prospective ECG triggering (11) (SnapShot Pulse, GE Healthcare). The X-ray window (“padding”) varied with HRv. We used padding 0, corresponding to a window of 100-ms scanning time at only 1 distinct end-diastolic phase (i.e., 75% of R-R cycle) in patients without HR variability. Padding 100, which corresponds to a window of 200-ms scanning time at 2 distinct end-diastolic phases (i.e., 70% to 80% of R-R cycle), was used in patients with HR variability ≤2 beats/min, and padding 200, corresponding to a window of 400-ms scanning time at 4 distinct phases (i.e., 40% to 80% of R-R cycle), was used in patients with HR variability >2 beats/min (10).
In patients from Group 2, we used a retrospective ECG triggering (12). The X-ray window was selected according to the “ECG-pulsing technique” (13) with a maximum tube current of 700 mAs between 40% and 80% of the RR cardiac cycle.
MDCT image reconstruction and analysis
The coronary calcium score was assessed, and the overall Agatston score (14) was recorded for each patient. In Group 1, image MDCT datasets were analyzed with vessel analysis software (CardioQ3 Package, GE Healthcare). In Group 2, image reconstruction was retrospectively gated to the ECG including X-ray window between 40% and 80% of the RR cycle with a 0.4-mm increment. Image quality score was classified for each segment as excellent (no artifacts, unrestricted evaluation), good (minor artifacts, good diagnostic quality), adequate (moderate artifacts, acceptable for routine clinical diagnosis), or poor/not evaluable (severe artifacts impairing accurate evaluation) (15). The causes of impaired image quality were classified as artifacts related to nonrespect of breath hold, premature ventricular beats, blooming effect, motion artifacts related to HRv, presence of cardiac device, interference of cardiac veins, intramyocardial tract, and impaired signal/image noise ratio (16).
Coronary artery segments were classified according to the 15-segment American Heart Association classification (17). All segments with a diameter of at least 1.5 mm at their origin were included. Two independent and blinded readers classified each vessel segment on a post-processing workstation (Advantage Workstation version 4.2, GE Healthcare) for the presence of significant stenoses, defined as narrowing of the coronary lumen exceeding 50%. For any disagreement in data analysis between the 2 readers, consensus agreement was achieved.
Conventional ICA was performed by standard technique. The coronary arteries were classified with the American Heart Association Classification (17). The angiograms were analyzed with quantitative coronary angiography software (QantCor, QCA, Pie Medical Imaging, Maastricht, the Netherlands) by 2 interventional cardiologists blinded to MDCT data sets. The severity of coronary stenosis was quantified in 2 orthogonal planes, and a stenosis was classified as significant if the lumen diameter reduction was >50%.
Radiation dose parameters
For MDCT the dose-length product, defined as total radiation energy absorbed by patient's body, was measured in mGy × centimeters in each patient. The effective radiation dose (ED) was calculated as the product of dose-length product times a conversion coefficient for the chest (K = 0.017 mSv/mGy·cm) (18). For ICA, we calculated ED in men and women by multiplying the dose-area product by a conversion factor (K = 0.21 mSv/mGy·cm2) for lateral and postero-anterior radiation exposure in the chest area (19).
Statistical analysis was performed with the SPSS version 13.0 software (SPSS Inc., Chicago, Illinois). Continuous variables were expressed as mean ± SD, and discrete variables were expressed as absolute number and percentage. A Student ttest was used to test differences of continuous variable between the 2 groups, and the chi-square test or Fisher exact test was used to study differences regarding categorical data. A p value <0.05 was considered statistically significant. Evaluability (number of segments evaluable/total number of coronary segments), sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV), and accuracy were calculated versus ICA for segments classified as evaluable and in a second analysis for all segments rating as positive the nondiagnostic segments. The 95% confidence interval for these parameters was calculated with the ratio estimator for variance. The diagnostic performance between the 2 groups was compared with the pairwise McNemar's test. The interobserver variability for the detection of significant disease on MDCT and ICA images was tested with Cohen's Kappa. The Spearman correlation and Bland-Altman analysis were performed to compare MDCT versus ICA results.
The 2 groups were homogeneous in terms of sex, age, and body mass index (Table 1).In Group 1, more patients were referred to ICA for chest pain or follow-up of known CAD, with fewer cases with a positive stress test. Moreover, Group 1 patients had a lower number of coronary segments with implanted stents compared with Group 2 patients but without significant differences in size and characteristics. Sixty-eight of 80 patients and 65 of 80 patients were receiving chronic beta-blockade therapy in Groups 1 and 2, respectively. The baseline HR was similar between the 2 groups. No significant differences were found in intravenous beta-blockade pre-treatment, HR during the scan, calcium score, or prevalence of significant CAD at ICA in a patient-based model between the 2 groups.
The baseline scan parameters were the same in the 2 groups (Table 2).According to the HRv, 3 (4%), 63 (80%), and 14 patients (17%) of Group 1 needed padding of 0, 100, and 200, respectively. The scan duration was longer in Group 1 than in Group 2.
The overall evaluability of nonstented segments (Table 3)was marginally better in Group 2 than in Group 1 (97% vs. 96%, p = 0.05), due to a significantly lower percentage of artifacts (3% vs. 8%, p < 0.01). Subanalysis of the artifacts showed a higher percentage of blooming artifacts and a lower number of HR-related motion artifacts in Group 2 versus Group 1. The evaluability of stented segments in Group 2 was slightly better but not significantly different than in Group 1 (94% vs. 92%) due to a lower number of artifacts (12% vs. 16%). In both groups, it was not possible to evaluate 8 stent segments.
MDCT image quality score
The image quality score of nonstented coronary arteries in Groups 1 and 2 was rated as excellent in 964 of 1,044 (92%) segments and 937 of 973 (95%) segments (p < 0.01), good in 18 (2%) segments and 5 (1%) segments (p < 0.01), adequate in 16 (2%) segments and 4 segments (1%) (p < 0.01), and poor in 46 segments (4%) and 27 segments (3%), respectively (p < 0.05).
Forty-four of 48 (92%) and 62 of 66 stents (93%) were rated as excellent in Groups 1 and 2, respectively (p = 0.63). The image quality of the remaining stents in both groups was classified as poor (Table 4).
MDCT accuracy in a segment-based model
In nonstented coronary segments, consensus between MDCT and ICA in classifying the coronary stenosis as significant was achieved in 195 of 998 segments and as angiographically normal in 737 of 998 segments. Overall, 27 lesions were underestimated by MDCT, and 39 segments were incorrectly graded as significantly stenotic. Including all diagnostic segments, the sensitivity, specificity, NPV, PPV, and accuracy were 88%, 95%, 96%, 83%, and 93%, respectively. Including all segments, the sensitivity, specificity, NPV, PPV, and accuracy were 89%, 91%, 96%, 75%, and 91%, respectively (Table 5).
Multidetector computed tomography correctly detected 11 significant in-stent restenoses (ISR) of 44 segments. One ISR was missed, and 2 significant ISR observed with MDCT were not confirmed at ICA. Including all diagnostic segments, the overall sensitivity, specificity, NPV, PPV, and accuracy were 92%, 94%, 97%, 85%, and 93%, respectively. Including all segments, the sensitivity, specificity, NPV, PPV, and accuracy were 93%, 88%, 97%, 76%, and 90%, respectively.
The Spearman correlation between the degree of stenosis measured by MDCT and ICA was good (r = 0.74 for nonstented and r = 0.83 for stented segments). Bland-Altman analysis demonstrated a slight overestimation of nonstented coronary artery stenosis and underestimation of in-stent percent stenosis (Fig. 2).
The Kappa value for detection of significant CAD was 0.88 for intraobserver agreement and 0.84 for interobserver agreement in nonstented segments and 0.78 and 0.76, respectively, in stented segments.
For nonstented coronary artery evaluation, the presence of significant stenoses was correctly detected in 238 segments and correctly excluded in 669 segments. Seventeen significant lesions were missed, and 22 stenoses rated as significant by MDCT were not confirmed by ICA. Including all diagnostic segments, the sensitivity, specificity, NPV, PPV, and accuracy were 93%, 97%, 98%, 92%, and 96%, respectively, whereas including all segments they were 94%, 94%, 98%, 86%, and 94%, respectively.
For stented-segments evaluation, of a total of 62 evaluable stents, 8 ISR were correctly detected, 3 were missed, and 2 were not confirmed by ICA. Including all diagnostic segments, the sensitivity was 73%, specificity 96%, NPV 94%, PPV 80%, and the overall accuracy 92%. Including all segments, the sensitivity was 77%, specificity 92%, NPV 94%, PPV 71%, and the overall accuracy 89%.
The Spearman correlation between MDCT and ICA was excellent for nonstented coronary arteries (r = 0.82) and for stented segments (r = 0.86). Bland-Altman analysis showed a very slight discrepancy between the percent stenosis detected with MDCT compared with ICA (Fig. 3).
The Kappa value for detection of significant CAD was 0.86 for intraobserver agreement and 0.82 for interobserver agreement in nonstented segments and 0.75 and 0.74 in stented segments.
Comparison of MDCT diagnostic accuracy in a segment-based model
Regarding nonstented coronary artery evaluation, the sensitivity, specificity, PPV, and accuracy were significantly higher in Group 2 than in Group 1, either including only diagnostic segments or all segments. In stented segment evaluation, Group 1 showed a significantly better sensitivity compared with Group 2, whereas no difference was found with regard to overall accuracy.
MDCT diagnostic accuracy in a patient-based model
In Group 1, 72 of 73 patients with significant stenosis in at least 1 coronary segment at ICA were correctly identified by MDCT. One patient was missed by MDCT. Moreover, significant CAD was correctly ruled out by MDCT in 6 patients, and in only 1 case, a stenosis was diagnosed as significant by MDCT and was found to be <50% at ICA. Regardless of the inclusion of diagnostic segments or all segments in Group 1, MDCT demonstrated a patient-based sensitivity of 99%, specificity of 86%, NPV of 86%, PPV of 99%, and accuracy of 98%. In Group 2, sensitivity was 99%, specificity 89%, NPV 86%, PPV 99%, and accuracy 98%, regardless of the inclusion of diagnostic segments or all segments. Therefore, no significant difference was found between the 2 groups.
Comparison of radiation dose parameters
Radiation dose exposure in each group is listed in Table 6.A 72% reduction (from 20.5 to 5.7 mSv) of the ED was observed with prospective ECG triggering. Moreover, when HRv allowed the use of padding 0, the ED was further reduced to an average dose as low as 3.8 mSv. The average ED of ICA was similar to that of MDCT with prospective ECG gating (6.3 ± 1.5 mSv vs. 5.7 ± 1.5 mSv, p = NS), whereas MDCT with retrospective ECG triggering showed a significantly higher ED.
Figure 4shows a good image quality of multiple restenoses of right coronary artery stent despite a very low effective radiation dose (2.5 mSv).
In this study we compared MDCT with retrospective versus prospective ECG triggering and observed a high evaluability and diagnostic accuracy in detecting significant CAD despite low-dose radiation. Indeed, the overall evaluability was 96% for nonstented coronary arteries and 92% for coronary stent evaluation. Moreover, the image quality score was excellent in 92% of nonstented and stented coronary segments. Specifically, despite a comparable overall calcium score between the 2 groups and a higher number of artifacts in Group 1, the computed tomography with prospective ECG triggering seems to present a lower number of artifacts due to blooming effect. Indeed, Stolzman et al. (8) demonstrated a high diagnostic performance of prospective ECG triggering despite the presence of high calcium score. Moreover, despite a similar HR between the 2 groups, the higher number of motion artifacts due to HRv suggests a significant impact of HR on performance of prospective ECG triggering. This might represent a relevant limitation of this technique for clinical application.
As concerns the detection of coronary artery stenosis, a high diagnostic accuracy was found in nonstented (91% to 93%) and stented (90% to 93%) segments. In particular, Group 2 had a slightly better evaluability and accuracy than Group 1 in nonstented segments. In contrast, the stent evaluation showed significantly better sensitivity in detecting ISR with prospective compared with retrospective ECG triggering. These findings are likely related to technical issues. During prospective ECG triggering the table remains stationary while the X-ray tube rotates around the patient and is advanced for the subsequent scan only when data acquisition is completed, preventing overlapping among the slices. This reduces the blooming effect, which might be relevant with stents or calcified plaques, and enhances the sensitivity to artifacts related to HRv (7). Moreover, the incidence of artifacts is generally higher in prospective ECG triggering. This is due to the fact that, when a “100” padding is used with this technique, no more than 2 cardiac phases are available for image reconstruction, whereas 4 cardiac phases are always available for image reconstruction with the retrospective ECG triggering modality. These factors have a differential impact on the performance of the 2 techniques. Indeed, the blooming effect was more frequently observed with the retrospective ECG triggering, whereas the HRv artifacts occurred more often in the prospective ECG triggering.
Nevertheless, in a patient-based model, which is more useful from the clinical standpoint, the accuracy was very high and exactly the same in the 2 groups (98%), regardless of the ECG-triggering technique used and the inclusion of only diagnostic segments or all segments. Therefore, the technique using prospective ECG triggering showed a high diagnostic performance for the assessment of CAD including all coronary segments and coronary stents. These findings are of clinical importance, because MDCT use is increasing for the evaluation of stent patency and assessment of the presence of CAD.
The effective radiation dose exposure was low with prospective ECG triggering (5.7 ± 1.5 mSv). It should be noted, however, that in our study population, the radiation dose in Group 1 was slightly higher than that reported in the published data, likely because individual adaptation of effective tube current and kilovoltage was not used (2). However, the prospective ECG triggering allowed a significant reduction of ED dose up to 72%. Notably, these radiation doses are lower than those of other noninvasive diagnostic tests used in cardiology, including nuclear perfusion scans (18). Indeed, depending on the technique of cardiac studies, a patient could typically receive a radiation dose up to 29 mSv (20). At these doses, there are concerns with regard to radiation-induced carcinogenesis (21). To put these radiation levels in context, it has been estimated that a coronary MDCT angiogram with an effective dose of 10 mSv has a risk of inducing a fatal cancer in 1 in over 2,000 cases (21). However, this is an age-averaged value that overestimates actual radiogenic risk in the older patients who most likely need to be investigated for CAD. Indeed, the lifetime attributable risk of cancer associated with radiation exposure is strictly dependent on patient sex and age. For example, the risk is significantly higher (0.70%) for young women in their 20s than for elderly men in their 80s (0.075%) (21).
First, we included a relatively small number of patients with coronary stents. Therefore, further studies are needed to confirm our preliminary results on the noninvasive assessment of coronary stents with this technique. Second, our study population had a high pre-test likelihood of CAD, and the accuracy of MDCT in these cases has been demonstrated to be significantly lower than in patients with low and intermediate pre-test likelihood of CAD (22).
Cardiac MDCT with prospective ECG triggering can reduce the radiation exposure with a slight reduction of evaluability and accuracy of noninvasive imaging of coronary arteries and stents.
- Abbreviations and Acronyms
- coronary artery disease
- effective radiation dose
- heart rate
- heart rate variability
- invasive coronary angiography
- in-stent restenosis
- multidetector computed tomography
- negative predictive value
- positive predictive value
- Received January 19, 2009.
- Revision received April 16, 2009.
- Accepted April 26, 2009.
- American College of Cardiology Foundation
- Schroeder S.,
- Achenbach S.,
- Bengel F.,
- et al.,
- Working Group Nuclear Cardiology and Cardiac CT; European Society of Cardiology; European Council of Nuclear Cardiology
- Wang M.,
- Qi H.T.,
- Wang X.M.,
- Wang T.,
- Chen J.H.,
- Liu C.
- Husmann L.,
- Valenta I.,
- Gaemperli O.,
- et al.
- Herzog B.A.,
- Husmann L.,
- Burkhard N.,
- et al.
- Scheffel H.,
- Alkadhi H.,
- Leschka S.,
- et al.
- Agatston A.S.,
- Janowitz W.R.,
- Hildner F.J.,
- et al.
- Leschka S.,
- Alkadhi H.,
- Plass A.,
- et al.
- Einstein A.J.,
- Moser K.W.,
- Thompson R.C.,
- Cerqueira M.D.,
- Henzlova M.J.
- Leung K.C.,
- Martin C.J.
- Coles D.R.,
- Smail M.A.,
- Negus I.S.,
- et al.