A description of the patient population in the present study was reported recently.6
A convenience sample of low- to intermediate-risk patients presenting to the emergency department with a chief complaint of chest discomfort and clinical suspicion for ACS, but who had normal initial troponin and an initial ECG without evidence of myocardial ischemia, was enrolled. ACS was defined as either an acute myocardial infarction or unstable angina pectoris during index hospitalization according to the AHA/ACC/ESC guidelines.7,8
Unstable angina pectoris was defined as clinical symptoms suggestive of ACS (typical chest discomfort or equivalent with an unstable pattern of chest pain, i.e. at rest, new onset, or crescendo angina) in the absence of elevated troponin, coinciding with appropriate objective evidence of myocardial ischemia in stress perfusion imaging or coronary angiography demonstrating a >50% coronary stenosis. To establish the diagnosis of ACS, an outcome panel of 2 experienced physicians blinded to CT findings reviewed patient data. Disagreement was resolved by consensus, which included an additional senior cardiologist. We collected information on cardiovascular risk profile and Thrombolysis in Myocardial Infarction (TIMI) risk score. For the purposes of the present analysis, we included all patients with a definitive significant stenosis (>50% in diameter) on coronary CTA by qualitative assessment. The Institutional Review Board approved the study.
All patients provided written informed consent. All patients were scanned using a 64-row CT scanner (Sensation 64; Siemens Medical Solutions, Forchheim, Germany). Intravenous beta-blocker (metoprolol, 5–20 mg) was administered in all patients with a heart rate >60 beats per minute. All patients received 0.6 mg of sublingual nitroglycerin. Contrast agent (Iodhexodol 320 g/cm3, Visipaque, General Electrics Healthcare, Princeton, NJ, USA) was injected at a rate of 5 mL/s. The scan parameters were: 64×0.6 mm detector configuration, 330 ms gantry rotation time, 120 kV tube potential, 850 mAs effective tube current-time product, and ECG-based tube current modulation. Axial images were reconstructed with a slice thickness of 0.75 mm using a retrospectively ECG-gated reconstruction.
Coronary CTA datasets were transferred to an offline workstation (Vitrea, Vital Images, Minnetonka, MN). An independent reader (>3 years cardiac CT experience) reviewed all studies and selected the series in cardiac phase with the best image quality in each subject for further analysis. The reader determined the culprit lesion in each patient with ACS using previously described rules.1
In patients with ACS and 1 significant lesion, this lesion was considered the culprit lesion. In patients with ACS and multiple significant lesions, we used information from diagnostic testing (ECG, nuclear perfusion imaging, invasive coronary angiography) to determine the culprit vessel. If there were >1 significant lesions in this vessel, the lesion with the most severe stenosis was analyzed. In patients without ACS, the lesion with the most severe stenosis by CTA was analyzed.
Two independent readers blinded to the demographic information as well as the presence of ACS evaluated the CT data set for characteristics of significant lesions. Each vessel containing a significant stenosis was analyzed in curved multiplanar reformat images in long axis and cross-sectional views. For each lesion, the proximal and distal reference were determined (). We used a semi-automated software (Vessel Probe, Vitrea, Vital Images, Minnetonka, MN) to segment lumen and outer vessel boundaries in all coronary artery cross-sections between the proximal and distal reference (). If necessary, the boundaries were manually adjusted. The lumen boundary was defined as the transition between high attenuation coronary lumen filled with the contrast agent and the intermediate attenuation of the non-calcified atherosclerotic plaque or the high attenuation of calcified plaque. The outer vessel boundary was defined as the transition between the intermediate attenuation on the non-calcified plaque or the high attenuation of the calcified plaque and the low attenuation of the perivascular fat.
Figure 1 Vessel Analysis. Panel A shows a series of vessel cross-sections created from a curved multiplanar reformatted image. Lumen (b lack arrowhead) and outer vessel (white arrowhead) boundaries were created with a semi-automated software. Lumen and outer vessel (more ...)
The diameters at the site of the maximum stenosis and at the proximal and distal references were measured (). The degree of stenosis was calculated as the ratio of the difference between the diameter at the maximum stenosis and the mean of the diameters at the proximal and distal references divided by the mean of the diameters at the proximal and distal references and expressed as percentage. The remodeling index was calculated as the outer vessel area at the site of the maximum stenosis divided by the mean of the outer vessel areas at the proximal and distal references. Positive remodeling was defined as a remodeling index of >1.05.1,9,10
The plaque length was measured as the length of the coronary segment between the proximal and distal reference in the curved multiplanar reformat images. The stenosis length was measured as the length of the visually significantly narrowed segment in the curved multiplanar reformat images.
The plaque volume was automatically calculated as the volume of all voxels segmented between the luminal and outer vessel boundaries in curved multiplanar reformatted images. The proximal and distal references were used as the proximal and distal ends of the plaques. We reported the total volume of plaque and the volumes of plaques in the range of <30 HU, <90 HU, 90–150 HU, and >150 HU.2,4,11–13
The composition of the plaque was assessed visually and categorized as calcified and non-calcified plaques as described previously ().14
In the presence of coronary calcium, calcified plaques were further characterized as spotty calcium (discrete calcified nodules clearly surrounded by non-calcified plaque, <3 mm in diameter) and heavy calcium (confluent calcium occupying most of the vessel wall) ().2
Plaque Composition. Coronary plaques were classified as non-calcified (Panel A), with spotty calcium (Panel B) and heavy calcium (Panel C)
Continuous variables are reported as mean (standard deviation) or median (interquartile range) and categorical variables as percentage (counts). Wilcoxon signed-rank test, and Fishers exact test were used to assess for differences within continuous and categorical variables. We calculated Pearson’s correlation coefficient to assess the agreement of the measurements between 2 observers.
In univariate analysis, characteristics of the stenosis and plaque were individually evaluated for association with ACS. We generated 3 scores (Score A, B and C) based on the characteristics of the stenosis and plaque. Score A consisted of the 2 features previously shown to be associated with culprit lesions in ACS: positive remodeling index and spotty calcium.1–3
For additional lesion scores continuous variables associated with the presence of the ACS at a significance level of <0.10 in univariate analysis were added according to their AUC (area under the receiver-operating characteristics curve) values in univariate analysis in the descending order. The variables were dichotomized by using the cut-points with the best discriminatory capacity (highest AUC) for ACS. The presence of each plaque feature increases the score by 1 point.
The discriminatory capacity of the scores for the prediction of ACS was assessed using c-statistics.15
The asymptotic 95% confidence intervals (95%CI) for the AUC were estimated using a nonparametric approach which is closely related to the jackknife technique.16
Logistic regression analysis was performed to calculate odds ratios with 95%CI for the diagnosis of ACS by each score. To assess the internal validity of ACS prediction using the model scores and to adjust for over-fitting/optimism, bootstrap resampling procedures were used. One hundred bootstrap samples were drawn with replacement from the original data set. The difference of AUC from the original dataset and from the bootstrap sample (mean AUC) represented an estimate of the optimism in the apparent performance. The optimism was subtracted from the apparent performance to estimate the internally validated performance and adjust for over-fitting/optimism.
Finally, we used 2×2 tables to calculate the diagnostic accuracy (sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) of single score values for the presence of ACS and provided binomial 95%CI. Pre-test probability of ACS was defined as the prevalence of ACS within the cohort (21/34). The post-test probability of ACS for all lesion scores was determined using Baye’s theorem (pretest odds×positive likelihood ratio=post-test odds of ACS). All performed tests were 2-sided and a p-value <0.05 was considered as statistically significant. All analyses were performed with SAS (version 9.2, SAS Institute Inc., Cary, NC USA).