|Home | About | Journals | Submit | Contact Us | Français|
To determine the usefulness of coronary computed tomography angiography (CTA) in patients with acute chest pain.
Triage of chest pain patients in the emergency department (ED) remains challenging.
Observational cohort study in chest pain patients with normal initial troponin and non-ischemic electrocardiogram. 64-slice coronary CTA was performed prior to admission to detect coronary plaque and stenosis (>50% luminal narrowing). Results were not disclosed. Endpoints were acute coronary syndrome (ACS) during index hospitalization and major adverse cardiac events (MACE) during 6- month follow-up.
Among 368 patients (mean age 53±12 years, 61% male) 31 had ACS (8%). By coronary CTA, 50% of these patients were free of CAD, 31% had nonobstructive disease, and 19% had inconclusive or positive CT for significant stenosis. Sensitivity and negative predictive value (NPV) for ACS were 100% (n=183/368; 95% confidence interval [CI]: 98 to 100%) and 100% (95%-CI: 0.89–1.00) with the absence of CAD; and 77% (95% CI: 59–90%) and 98% (n=300/368, 95%-CI: 95–99%) with significant stenosis by coronary CTA. Specificity of presence of plaque and stenosis for ACS were 54% (95%-CI: 0.49–0.60) and 87% (95%-CI: 0.83–0.90); respectively. Only one ACS occurred in the absence of calcified plaque. Both, the extent of coronary plaque and presence of stenosis predicted ACS independently and incrementally to TIMI risk score (AUC: 0.88, 0.82 vs. 0.63; respectively, all p<0.0001).
Fifty percent of patients with acute chest pain and low to intermediate likelihood of ACS are free of CAD by CT and have no ACS. Given the large number of such patients early coronary CTA may significantly improve patient management in the emergency department.
Patients who present with acute chest pain that is believed to be of ischemic origin but who have a normal initial biochemical markers for myocardial necrosis (Troponin, Creatinine kinase) and normal or non-diagnostic electrocardiograms (ECG) represent a major diagnostic challenge to emergency departments (1–7). As a result, most of these patients are admitted to the hospital for up to 24 hours and undergo serial ECG and troponin testing as well as a stress test to rule out acute coronary syndrome (ACS) at a cost in excess of $8 billion annually (8–10).
Coronary computed tomography angiography (CTA) is a rapid diagnostic test that has the unique ability to noninvasively and accurately detect significant coronary artery stenoses (11) and coronary atherosclerotic plaque (12,13). Several smaller studies suggest that coronary CTA may be helpful to facilitate early triage in patients with acute chest pain (14–17). However, the distribution of CT-angiographic findings of coronary artery disease such as plaque and stenosis and their association with ACS is not established. Such knowledge would provide the basis for the assessment of the clinical utility and the economic implications for using coronary CTA as an early triage tool. Thus, we conducted a prospective observational cohort study to assess the usefulness of coronary CTA in patients with acute chest pain who are being admitted with low to intermediate risk for ACS.
The patient population of the “Rule Out Myocardial Infarction using Computer Assisted Tomography” (ROMICAT) trial consisted of patients who had a chief complaint of acute chest pain lasting >5 minutes during the past 24 hours, normal initial troponin and an initial ECG without evidence of myocardial ischemia. In all patients, ED physicians had sufficient clinical suspicion for an ischemic origin of chest pain and admitted these patients to the hospital to rule out ACS. Notably, patients with a history of established CAD, defined as stent implantation or coronary artery bypass grafting were excluded. Detailed inclusion and exclusion criteria are provided in Table 1.
We screened patients who presented with a chief complaint of chest pain to the ED on weekdays from 7am to 7pm (May 2005 to May 2007). All eligible patients who agreed to participate underwent contrast-enhanced coronary CTA prior to admission to the hospital floor. All physicians, including those in the ED, who were involved in the standard clinical care of the patients, remained blinded to the result of coronary CTA. The institutional review board approved the study protocol and all patients provided written informed consent.
CT imaging was performed using a 64-slice CT scanner (Sensation 64; Siemens Medical Solutions, Forchheim, Germany). In preparation for the scan, patients with a heart rate >60 bpm received an intravenous beta-blocker (metoprolol, 5–20 mg) unless their systolic blood pressure was <100 mmHg, or other contraindications were present. In addition, patients received 0.6 mg of sublingual nitroglycerin. All image acquisitions were performed during a single breath hold in inspiration.
Per standard protocol, a test bolus of 20 ml contrast agent was administered with a flow rate of 5 ml/s to determine the optimal timing of contrast injection. Coronary CTA data sets were acquired with 64 × 0.6 mm slice collimation, a gantry rotation time of 330 ms, tube voltage of 120 kV, and an effective tube current of 850 mAs using ECG-correlated tube current modulation when appropriate (18). Contrast agent (80–100 ml, Iodhexodol 320 mg/cm3, Visipaque, General Electrics Healthcare, Princeton, NJ, USA) was injected intravenously at a rate of 5 ml/s to ensure homogeneous enhancement of the entire coronary artery tree.
Axial images were reconstructed with a slice thickness of 0.75 mm and increment of 0.4 mm using a retrospectively ECG-gated half-scan algorithm with a temporal resolution of 165 ms. Images were initially reconstructed at 60%, 65%, 70%, and 35% of the cardiac cycle (19). Additional reconstructions were performed to minimize motion artifacts, if necessary. All reconstructions were transferred to an offline workstation for analysis (Leonardo, Siemens Medical Solutions, Forchheim, Germany).
Assessment of coronary CTA data sets for the presence of significant coronary stenosis and the presence of coronary atherosclerotic plaque was performed as a consensus reading by two experienced investigators (U.H./M.F. or F.B./M.D.S.) blinded to the subject’s clinical presentation and history using a modified 17-segment model of the coronary artery tree (20,21). If a consensus could not be reached, a third expert reader made the final diagnosis (S.A.). This method has been demonstrated to be highly reproducible (17).
The presence of any coronary atherosclerotic plaque per segment, whether calcified or non-calcified, was determined as described previously (12,13). Briefly, non-calcified plaque was defined as any discernible structure that could be assigned to the coronary artery wall and had CT attenuation below the contrast-enhanced coronary lumen but above the surrounding connective tissue/epicardial fat. Calcified plaque was defined as any structure with a CT attenuation of >130 HU that could be visually distinguished from the contrast-enhanced coronary lumen.
The presence of coronary artery stenosis was defined as a luminal obstruction >50% diameter in any coronary segment. If image quality did not permit definite exclusion of the presence of a significant stenosis (due to the presence of motion artifacts, calcification, or low contrast-to-noise ratio), the segment was classified as indeterminate. For calculation of diagnostic accuracy such cases were counted as positive.
ACS was defined as either an acute myocardial infarction (i.e. patients developed a positive troponin during serial testing (6 hour or 9 hours after ED presentation) or unstable angina pectoris (UAP) according to the AHA/ACC/ESC guidelines (22–24). UAP was defined as clinical symptoms suggestive of ACS (unstable pattern of chest pain - at rest, new onset, or crescendo angina), optimally with objective evidence of myocardial ischemia such as a positive stress test.
A standardized follow-up phone call was conducted six months after enrollment to determine the occurrence of MACE (death, myocardial infarction, and coronary revascularization). In addition, we retrieved medical record for all patients to verify all events potentially corresponding to a MACE such as a report of recurrent symptoms resulting in medical consultation, diagnostic testing, or hospital admissions were subsequently validated by review of medical records. Overall, this approach resulted in a follow-up completion rate of 92%. In the remaining 8% of patients, we assessed mortality using the online social security death index website.
To establish the diagnosis of ACS and MACE, an outcome panel of two experienced physicians with more than 10 years experience (one ED physician [J.T.N] and one cardiologist [S.S.]) reviewed patient data forms containing prospectively collected information on the history and nature of chest pain, risk factors and medical history, as well as medical records pertaining to the hospital admission. The outcome panel was blinded to the findings of coronary CT. Disagreement was resolved by consensus, which included an additional cardiologist (C.C.) (25).
We prospectively collected data on demographics, risk factor profile, TIMI risk score, and clinical course in all patients. Presence of risk factors (i.e. hypertension, hypercholesterolemia, and diabetes mellitus) was established from actual measurements obtained during the hospitalization or related medication use. Medical records were reviewed to obtain results of all diagnostic tests performed during index hospitalization.
Demographics, traditional risk factors, clinical events, and prevalence of plaque and stenosis as detected by coronary CTA are presented as mean ± SD or medians and interquartile range for continuous variables, and percentages for categorical variables.
We determined the utility of coronary CT to guide triage decisions in the ED using two different analytic strategies. To determine the accuracy of coronary CTA, we calculated conventional measures of diagnostic accuracy (sensitivity, negative predictive value [NPV], specificity, and positive predictive value [PPV]) and test-positive and negative likelihood ratios with 95% confidence intervals based on a binomial distribution for (1) the absence of plaque and (2) the absence of significant stenosis for the detection of ACS. Chi-square test was used to compare proportions and measures of diagnostic accuracy between groups. To compare the extent of plaque between subjects with and without ACS, Wilcoxon’s Rank Sum test was applied. Further, we performed multivariate logistic regression modeling approach to examine the association between the extent of coronary atherosclerotic plaque and the presence of stenosis as detected by coronary CTA with the outcome of ACS. The crude models contained the presence of stenosis as a dichotomous variable, or the extent of plaque defined as the number of coronary segments with any plaque (1–17) as a continuous variable. We then tested whether the association between CT findings and ACS persisted after adjusted for age, gender and TIMI risk score. Model fit was assessed using c-statistics, which is equivalent to the area under the receiver-operating characteristics curve (AUC) (26). The asymptotic 95% confidence intervals for the AUCs were estimated using a nonparametric approach which is closely related to the jackknife technique as proposed by DeLong et al. (27). Also, we performed a two-sided asymptotic z-test to compare the AUC of the TIMI risk score and the CT finding, in which the standard error of the test statistics was derived from the asymptotic variance covariance (27).
The study was designed to assure a high precision of the estimates of diagnostic accuracy. We aimed to demonstrate lower bounds of 95% confidence intervals for NPV above 90%. Based on our initial experience and the literature we assumed the following: 12% prevalence of ACS, 90% sensitivity of significant coronary stenosis for ACS, sensitivity of 90% and specificity of 85% for coronary CTA to detect significant coronary artery stenoses. Thus, a sample size of 400 patients would assure tight (<10%) confidence interval of the NPV above 90%.
A two sided p-value of <0.05 was considered to indicate statistical significance. All analyses were performed using SAS (Version 9.1, SAS Institute Inc., Cary, NC, USA).
A total of 1869 patients with a primary complaint of chest pain lasting 5 >minutes were screened during the enrollment period. Exclusion criteria were present in 1270 patients (impaired renal function (n=454), history of CAD defined as previous stent placement or coronary bypass (n=231), ECG diagnostic for myocardial ischemia or positive initial biomarkers (n=209), arrhythmia (n=97), inability to pause metformin (n=68), who enrolled in different research study or were previously included in this study (n=63), history of allergy to iodine (n=58), inability to administer beta blocker due to asthma (n=37), clinically unstable (n=31), or lack of pregnancy testing (n=20)). In addition, we excluded 231 patients who were ineligible because of interference with standard clinical care (n=100), who refused participation (n=124), or did not complete the CT exam (n=7). Thus, the study cohort consisted of 368 patients (mean age: 52 years old, 40% female, Table 2).
Overall, 8.4% of patients (n=31/368) had ACS (myocardial infarction: n=8, UAP: n=23) while ACS was ruled out in the remaining 337 patients (91.6%). After a mean follow-up of 6.2±2.0 months, none of 337 subjects without ACS had suffered a MACE.
The average time to perform a coronary CTA (door to door time, including patient preparation) was 16±7 minutes. The mean actual scan time to obtain the coronary CTA data set was 14±2 seconds. Average time for the interpretation of CT images was 9±7 minutes (range: 3–29).
By coronary CTA, 50.3% (n=183/368) of these patients were free of CAD, 31.2% had plaque but no stenosis (n=117/368), and 18.5% had a positive CTA (34 were positive for stenosis and 34 rendered inconclusive assessment). The diagnostic test characteristics for ACS are shown in table 3.
Because none of the patients without plaque had ACS, sensitivity and NPV were excellent (100%). In contrast, specificity and PPV of the presence of coronary plaque was low to moderate, because many patients had plaque but no ACS (PPV: 17%, specificity: 54%). Notably, the specificity of the presence of coronary plaque for ACS was lower in older subjects due to high prevalence of plaque (21% vs. 59%, p<0.0001; for subjects ≥65 vs. <65 years of age, respectively). Similar findings were seen for the detection of MI (Table 3).
Among 185 subjects in whom any coronary plaque was detected, patients with ACS had a significantly more plaque (7.2±3.7 vs. 4.2±3.4 segments, p<0.0001) as compared to subjects without ACS. Similar results were seen for calcified plaque and non-calcified plaque (6.5±3.7 vs. 3.6±3.5 segments, p<0.0001; and 3.6±3.2 vs. 1.8±2.2 segments, p<0.0001, respectively). Among 14 subjects (4%) with exclusively non-calcified plaque, only one subject developed ACS (1/14 [7.1%]).
The absence of significant stenosis had excellent NPV of 98% but sensitivity was limited to 77%, as seven subjects in whom a stenosis was excluded by coronary CTA had ACS (Table 3). Characteristics of these subjects are detailed in table 4. Because a substantial fraction of patients with a positive CT developed ACS, PPV and specificity were reasonable and very good and higher than for the absence of plaque (PPV: 35%, specificity: 87%). Similar findings were seen for the detection of MI (Table 3).
The specificity of the presence of significant stenosis detected by coronary CTA for ACS was lower in older subjects (58% vs. 91%, p<0.0001; for subjects ≥65 vs. <65 years of age, respectively) in whom coronary calcification was more prevalent (84% vs. 39%; p<0.0001). The proportion of patients in whom a stenosis could not be definitely excluded in CT was significantly higher among subjects with ACS as compared to subjects without ACS (24/31 [77.4%] vs. 44/337 [13.1%], p<0.0001). Remarkably, 14/34 who had a significant stenosis detected by CT were not diagnosed with ACS. None of them had a MACE after 6 months.
In logistic regression analysis, each additional segment of plaque was associated with a 37% increase risk of having an ACS (OR: 1.37, 95%-CI: 1.25to 1.51]; p<0.001), while the presence of stenosis was associated with an over 20-fold increased risk of ACS (OR: 22.8, 95%-CI: 9.3 to 56.1; p<0.0001). These associations persisted after adjustment for age, gender, and TIMI risk score (OR: 1.28, 95%-CI: 1.14 to 1.43; p<0.0001 and OR: 11.69, 95%-CI: 4.4 to 31.0; p<0.0001 for the extent of plaque and the presence of stenosis, respectively). Area under the curve (AUC) in receiver operator curves for the prediction of ACS were higher for both the extent of plaque (AUC: 0.88, 95%-CI: 0.83–0.93) and the presence of stenosis (AUC: 0.82, 95%-CI: 0.74–0.89) as compared to TIMI risk score (AUC: 0.63, 95%-CI: 0.54–0.71; Figure 2).
On average, patients presented to the ED 6.7±7.4 hours (range: 0.08–24.0 hours) after the onset of chest pain. For most of these patients, the hospital course was characterized by obtaining serial troponin measurements and resting ECGs over the first 24 hours, and stress testing the following day. The average hospital length of stay was 40.5 ± 43.2 hours (range: 2.7 to 381.4 hours).
Among the 31 patients with ACS, 20 patients underwent selective invasive coronary angiography, 19 of which revealed significant stenosis. One angiogram revealed a 30% stenosis that improved with intracoronary nitroglycerin, thus the patient’s NSTEMI was attributed to vasospasm. Two patients had exercise treadmill tests, both of which were positive. 13 patients had stress (exercise or adenosine) SPECT imaging, 11 of which were positive, one was negative at a submaximal heart rate, and one was indeterminate due to significant attenuation artifact. The one patient who had a negative submaximal SPECT was found to have significant stenosis on coronary CTA. The one patient who had an indeterminate SPECT also had an inconclusive coronary CTA.
Among the 337 (91.6%) patients in whom ACS was excluded, 13 patients underwent selective invasive coronary angiography, none of which revealed significant stenosis. 117 patients underwent exercise treadmill testing. Two ETTs were positive: one patient was positive but negative on subsequent invasive coronary angiography and negative on coronary CTA, one patient was positive but negative on subsequent SPECT and negative on CTA. 137 patients had stress (exercise or adenosine) SPECT imaging, three of which had ischemia on SPECT but no obstructive disease on invasive coronary angiography or coronary CTA.
In this blinded observational cohort study we demonstrate that 50% of patients who presented with acute chest pain to the ED and were at low to intermediate likelihood of ACS have no CAD by coronary CTA, a finding that has 100% NPV but limited PPV for the subsequent diagnosis of ACS and MACE. In addition, our results indicate that while the NPV remains excellent (98%), the exclusion of significant coronary stenosis by coronary CTA (>50%) has a limited sensitivity (77%) for the detection of ACS due to a number of false negative findings of lesions in small vessels. Both, plaque and stenosis by CT predict ACS independent of cardiovascular risk factors or TIMI risk score (AUC: 0.88, 0.82, and 0.63; all p <0.05; respectively). PPV of coronary CTA is limited in patients >65 years of age. Given the large number of patients with acute chest pain early coronary CTA may significantly improve patient management in the emergency department by aiding clinical decision making, specifically early discharge of subjects at low to intermediate likelihood of ACS without CAD.
A number of smaller studies have demonstrated that a negative CT, defined variably as the absence of coronary calcification, the absence of CAD, or the absence of non-significant stenosis has a high negative predictive value for ACS (14,15). In addition to confirming these findings, we are able to provide robust estimates of diagnostic accuracy with narrower confidence intervals (lower 95% confidence bound >85% for NPV) due to our larger sample size. Importantly, in contrast to previous publications we also demonstrate that the presence of significant stenosis (defined as >50% luminal narrowing) has also reasonable test characteristics for the detection of ACS, although expectedly this is not a perfect criterion for ED triage of patients with acute chest pain (sensitivity 77%, n=7/31). Possible explanations are rupture or thrombosis in subcritical (28,29) or microvascular disease (30) and limited accuracy of coronary CTA to detect stenosis in small caliber vessels (>2mm) (11). Our results suggest that the technique may be less efficient in elderly patients >65 years of age as the specificity of the plaque triage criterion is very significantly limited since most of these patients will have CAD (59% vs. 21% for patients >65 vs. <65 years of age).
In contrast to most published studies, the design of our study permits an unbiased assessment because coronary CTA was not part of standard care and thus patient management and subsequent patient outcomes were not affected by CTA (i.e. additional downstream testing such as coronary angiography due to suspected stenosis in CT). As a result, we are able to report the diagnostic accuracy of several CT-angiographic patterns of CAD, such as presence of both calcified and non-calcified plaque as well as coronary stenosis <50% for ACS. Interestingly, a significant stenosis by CT was detected in 14 patients whom were deemed to not have ACS based on clinical presentation, ECG, biomarkers, and a negative diagnostic test for ischemia. This suggests that CT is more sensitive in detecting significant luminal narrowing although the hemodynamic significance is unknown. Since none of these patients had a MACE over the following 6 months, the finding of significant CAD as detected by CT may be longstanding and coincidental with no relation to the patient’s acute clinical presentation in the ED. This finding warrants further research specifically related to the morphologic appearance of these lesions as compared to ACS lesions.
Our data also demonstrate that coronary CTA can risk stratify patients with acute chest pain and intermediate likelihood of ACS independent of cardiovascular risk factors, TIMI risk score. While such an analysis is familiar from observational trials using nuclear perfusion imaging at rest in the pre-troponin era (31,32) our results suggest that coronary CTA is superior to nuclear perfusion imaging (OR: 3.83 95%-CI: 2.36 to 6.21 for nuclear imaging vs. OR: 8.65 95%-CI: 3.69 to 20.26 for coronary CTA) (33). This information may guide assessment of the level of care necessary for these patients and moreover, may improve risk assessment and prevention efforts in patients without ACS who are found to have coronary atherosclerosis. Several studies now support that the presence and extent of CAD is also a powerful predictor of future cardiovascular events (34,35).
It appears that there is broad agreement that coronary CTA may improve management of patients with acute chest pain (14,15,17). However, our results demonstrate that the strength of coronary CTA is the high negative predictive value for ACS. Half of all patients in our population had no CAD as detected by coronary CTA. In these patients alternative diagnostic tests such as exercise stress testing or stress nuclear perfusion imaging were positive in up to 20% of cases reflecting that they have limited specificity as compared to CT (33,36–38). Because none of these patients had ACS they may be directly discharged from the ED without further diagnostic testing or hospital admission. Overall, our results may provide the rationale to establish recommendations for the actual clinical use of cardiac CT in populations with a low or intermediate likelihood of ACS, a population in whom diagnostic imaging tests have been generally recommended (22,39).
One of the major limitations of coronary CT is the associated radiation exposure (40). Reduction of radiation exposure using ECG tube modulation or prospective ECG triggering (41) will greatly facilitate acceptance in clinical practice.
Our study has several limitations. This is a single center study with enrollment limited to weekday day time hours. However, subjects presenting outside enrollment hours were not significantly different from enrolled subjects with respect to age and gender, which are the strongest predictors of the prevalence of CAD. Due to exclusion of patients with known CAD and renal impairment, the very elderly are underrepresented in this study. In a real world clinical scenario many of these patients may be eligible for coronary CTA. Also, we used a CT scanner system from a single manufacturer. However, differences in the accuracy of coronary CTA for the detection plaque and stenosis are marginal between vendors (11). In addition, coronary CTA exams in our study were performed by a dedicated research team and interpreted by readers with a high level of expertise in the field, having at least two years experience with coronary CTA with more than 800 studies interpreted. Thus, it is possible that our results, including the high reproducibility of CT readings and the small number of inconclusive examinations, may be replicated only in centers with similar levels of expertise.
In conclusion, both plaque and stenosis by CT predict ACS independent of cardiovascular risk factors or TIMI risk score. In this study, fifty percent of patients with acute chest pain and low to intermediate likelihood of ACS are free of CAD by CT and have no ACS. Given the large number of such patients early coronary CTA may significantly improve patient management in the emergency department.
We gratefully acknowledge the enthusiastic support in patient enrollment of the team of faculty, residents, nursing and administrative staff of the Emergency Department Services of the Massachusetts General Hospital.
SOURCES OF FUNDING
This work was supported by the NIH (R01 HL080053), and in part supported by Siemens Medical Solutions and General Electrics Healthcare. Drs. Rogers, Truong, Shapiro, and Moloo were supported by the National Institutes of Health grant T32HL076136.
Dr. Hoffman has received research grants from GE and Siemens.
Dr. Rogers has no financial conflicts of interest to report. However, Dr. Rogers acknowledges salary support from NIH Institutional Training Grant T32HL076136.
Dr. Truong acknowledges support from National Institutes of Health grant T32HL076136.
Dr. Abbara has received research funding from Bracco and consulting honoraria (minor) from Ezem, Siemens, Partners Imaging, Magellan Health, Perceptive Informatics.
Dr. Achenbach has received grant support from Siemens and Bayer Schering Pharma.
Dr. Nagurney is funded by Biosite for a biomarker research study.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.