Author + information
- Received April 6, 2011
- Revision received October 11, 2011
- Accepted October 27, 2011
- Published online February 14, 2012.
- Kavitha M. Chinnaiyan, MD⁎,⁎ (, )
- Gilbert L. Raff, MD⁎,
- Tauqir Goraya, MD, PhD†,
- Karthikeyan Ananthasubramaniam, MD‡,
- Michael J. Gallagher, MD⁎,
- Aiden Abidov, MD, PhD§,
- Judith A. Boura, MS⁎,
- David Share, MD∥,¶ and
- Patricia A. Peyser, PhD¶
- ↵⁎Reprint requests and correspondence:
Dr. Kavitha M. Chinnaiyan, William Beaumont Hospital, 3601 West 13 Mile Road, Royal Oak, Michigan 48073
Objectives This study was conducted to evaluate the correlation between stress test results and coronary computed tomography angiography (CCTA) findings and comparative diagnostic performance of the 2 modalities in patients undergoing invasive coronary angiography (ICA).
Background Recent data suggest that only a third of patients undergoing ICA have obstructive coronary artery disease (CAD); accurate pre-ICA risk stratification is needed.
Methods At 47 centers participating in the ACIC (Advanced Cardiovascular Imaging Consortium) in Michigan, patients without known CAD who were undergoing CCTA within 3 months of a stress test were studied. Demographics, risk factors, symptoms, and stress test results were correlated with obstructive CAD (>50% stenosis) on CCTA and ICA.
Results Among 6,198 patients (age 56 ± 12 years, 48% men), >50% stenosis was seen in 1,158 (18.7%) on CCTA. Independent predictors included male sex (odds ratio [OR]: 2.37, 95% confidence interval [CI]: 1.83 to 3.06), current smoking (OR: 2.23, 95% CI: 1.57 to 3.17), older age (OR per 10-year increment: 2.14, 95% CI: 1.89 to 2.41), hypertension (OR: 1.8, 95% CI: 1.37 to 2.34), and typical angina (OR: 1.48, 95% CI: 1.03 to 2.12). Stress test results were not predictive. Among patients undergoing ICA (n = 621), there was a strong correlation of ICA with CCTA findings (OR: 9.09, 95% CI: 5.57 to 14.8, p < 0.001), but not stress results (OR: 0.79, 95% CI: 0.56 to 1.11, p = 0.17).
Conclusions Stress test findings did not predict obstructive CAD on CCTA, observed in <20% of patients in this large study group. The strong association of CCTA with ICA suggests that it may serve as an effective “gatekeeper” to invasive testing in patients needing adjudication of stress test results. (Advanced Cardiovascular Imaging Consortium: A Collaborative Quality Improvement Project [ACIC]; NCT00640068)
- Advanced Cardiovascular Imaging Consortium
- coronary computed tomography angiography
- stress tests
Coronary artery disease (CAD) results in more than half of all cardiovascular deaths in the United States and remains the leading cause of morbidity and mortality worldwide (1). With an increasing population at risk for CAD, healthcare systems are under financial pressure to deliver cost-effective diagnosis and risk stratification. Potential concerns about incorrect diagnosis including adverse clinical events and/or inappropriate testing or therapy have led to over-utilization of cardiovascular imaging and increased healthcare costs over the last decade (2,3).
As many as 10% of stress imaging studies are considered inconclusive, many leading to subsequent invasive coronary angiography (ICA) for definitive diagnosis, often with negative results (4,5). Recent data from the National Cardiovascular Data Registry of 400,000 patients demonstrated obstructive CAD in only 38% of those undergoing ICA, and positive stress tests were only moderately associated with obstructive CAD (odds ratio [OR]: 1.28, 95% confidence interval [CI]: 1.19 to 1.37) (6). Thus, strategies for risk stratification and patient selection for ICA need improvement.
Coronary computed tomography angiography (CCTA) has demonstrated excellent accuracy for defining the presence and severity of luminal coronary artery stenoses and can identify persons at risk for all-cause mortality (7,8). Thus, CCTA may be useful in determining the need for ICA, particularly in patients with inconclusive stress tests (9). The purpose of this study was to evaluate the relationship between clinical assessment, stress test results, and extent of CAD on CCTA in a large statewide registry. We also studied comparative diagnostic performance of these noninvasive modalities in patients undergoing subsequent ICA.
We studied patients prospectively enrolled in a statewide CCTA registry, the ACIC (Advanced Cardiovascular Imaging Consortium) (10). The ACIC is a collaborative quality improvement initiative sponsored by the Blue Cross/Blue Shield/Blue Care Network of Michigan and currently includes 47 hospitals and out-patient imaging centers. It is approved by institutional review boards at participating centers and includes a waiver of consent. Patients with no known CAD undergoing CCTA within 3 months after stress testing were included. The study population was divided into 3 groups according to stress test results, namely, normal, equivocal, or abnormal; and the groups were compared for demographics, symptoms, coronary risk factors, and extent and severity of CAD on CCTA.
Data collected for ACIC includes demographics, clinical risk factors, symptoms, results of prior testing, and medical history elicited by research coordinators using a structured patient interview as well as office and hospital medical records. Chest pain is classified as typical, atypical, or noncardiac on the basis of 3, 2, or 1 of the following, respectively: substernal chest discomfort, aggravation with exertion, and alleviation with rest and/or sublingual nitroglycerin (11). Additional symptoms such as dyspnea are noted. Pre-test CAD likelihood is classified as low (<10%), intermediate (10% to 90%), or high (>90%) on the basis of American College of Cardiology/American Heart Association criteria (11). For the purpose of this analysis, a modified Framingham Risk Score (FRS) was calculated with a moderate score (i.e., 1 point) imputed for a history of dyslipidemia or the use of lipid-lowering medications and hypertension or use of antihypertensive medications.
CCTA scan protocol and image interpretation
The CCTA was performed on various types of scanners available at each institution, ranging from 64- to 128-slice scanners, with scan techniques dictated by standard clinical protocols at each site, as described elsewhere (10). Interpretation of the CCTA was performed at each institution by cardiologists and/or radiologists with level II (or higher) training. Coronary stenoses were evaluated using a 16-segment model. For this analysis, the following categories were used: no stenosis, <50% stenosis, and >50% stenosis. Obstructive CAD was defined as coronary stenosis >50%. For distribution of obstructive CAD, major epicardial vessels (left main, proximal, and mid segments of the left anterior descending, left circumflex, and right coronary arteries, first and second diagonal and obtuse marginal branches) demonstrating >50% stenoses were evaluated.
Stress testing results
Stress tests included treadmill exercise testing (TMET), stress echocardiography (SE), and myocardial perfusion imaging (MPI) with single-photon emission computed tomography (SPECT). Referring physicians reported stress test results as normal, equivocal/nondiagnostic, or abnormal.
Invasive coronary angiography
The ICA results within 90 days of the CCTA were collected by chart review, and included location and degree of CAD severity as reported by the performing physician.
Categorical variables were examined using Pearson's chi-square or Fisher's exact tests, and are reported as counts and percent frequencies. Continuous variables were examined using nonparametric Kruskal-Wallis tests or analysis of variance, depending on the distribution of the data, and are reported as mean ± SD or medians where appropriate. A p value <0.05 was considered statistically significant. All tests were 2-sided.
Multivariable logistic-regression analysis was performed to identify factors associated with obstructive CAD from among baseline clinical variables. The Wald chi-square test was used to determine significant predictors of >50%. To understand the relative value of the factors in predicting obstructive CAD, 4 separate models were constructed: the first included the FRS alone (to predict risk in the absence of symptoms), and subsequently added other clinical factors (second model), followed by pre-test CAD likelihood (third model), and finally, stress test results (fourth model); in patients undergoing ICA, a fifth model was constructed by adding CCTA results. The predictive value of each model is represented by the C-statistic, the nonparametric estimate of the area under the receiver-operating characteristic curve. We tested whether the C-statistics for different models were significantly different from the first model that included the FRS alone using a nonparametric method (12). A p value <0.05 was considered to indicate statistical significance for all models. All statistical analyses were performed using the SAS software (version 9.1.3, SAS Institute, Cary, North Carolina).
Between July 2007 and September 2010, 22,551 patients at 47 sites underwent CCTA (Fig. 1). The final study group included 6,198 patients without known CAD and with stress testing within preceding 3 months, representing 27.5% of all patients. Stress tests included MPI in 4,037 (65.1%), SE in 1,581 (25.55%), and TMET in 580 (9.45%).
Patient characteristics are presented in Table 1. Of the 6,198 patients, 1,548 (24.9%) had normal, 1,027 (16.6%) had equivocal, and 3,623 (58.5%) had abnormal stress tests. Patients with abnormal stress tests had higher mean FRS (p < 0.0001), higher mean body mass index (p < 0.0001), and higher frequencies of diabetes mellitus (p = 0.0002) and hypertension (p = 0.0003).
Symptoms and indications before CCTA
The majority of patients (5,118, 82.6%) had symptoms suggestive of ischemia; those with normal stress test results had the highest frequency of symptoms suggestive of CAD, whereas those with abnormal stress tests had the lowest (p < 0.0001) (Table 1). Indication for CCTA in the remaining (1,080, 17.4%) was cardiac risk factors.
Relationship between symptoms, stress test results, and CCTA findings
Obstructive CAD (>50% stenosis) was identified in 1,158 (18.7%) of all 6,198 patients in the study group. The remaining 5,040 (81.3%) had normal coronary arteries or nonobstructive CAD. Patients with normal stress tests had a higher frequency of normal coronary arteries, whereas patients with equivocal or abnormal stress tests had higher frequencies of >50% stenosis and 2-vessel or 3-vessel disease (p = 0.021) (Table 1, Fig. 2).
Of the entire group, 1,080 (17.45%) patients were asymptomatic, including 847 (78.4%) with abnormal, 131 (12.1%) with equivocal, and 102 (9.4%) with normal stress tests (Table 2). Asymptomatic patients were older, more often male, and had higher frequencies of hypertension and hyperlipidemia, with higher FRS compared with symptomatic patients. They also had a higher burden of CAD on CCTA (<50% as well as >50% stenoses) and lower frequency of normal coronary arteries (p < 0.0001). Normal coronary arteries or nonobstructive CAD were noted in 77.1% of asymptomatic patients and 82.2% of symptomatic patients (p < 0.0001).
Predictors of obstructive coronary artery disease
Independent predictors of obstructive CAD on CCTA included male sex, current smoking, older age (10-year increments), hypertension, and typical angina (Table 3). Stress test result was not independently associated with obstructive CAD after other factors were included in the model.
Incremental value of pre-CCTA work-up data
The results of 4 separate models for the prediction of >50% stenosis on CCTA are shown in Figure 3A. The first included only the FRS (C-statistic 0.73, 95% CI: 0.71 to 0.74). In the second model, risk factors not included in the FRS (history of peripheral vascular disease, cerebrovascular disease, and chronic obstructive lung disease) were added, with no increase in the C-statistic, which remained at 0.73 (95% CI: 0.71 to 0.74, p = 0.536). No significant increase was noted with the addition of pre-test CAD likelihood (C-statistic 0.73, 95% CI: 0.72 to 0.75, p = 0.061) in the third model. Finally, addition of stress test results in the fourth model had no effect on its predictive ability above and over that of the first model with FRS alone (C-statistic 0.73, 95% CI: 0.72 to 0.75, p = 0.776). Thus, FRS alone provided the same predictive ability as adding other clinical information or stress test results in this study group.
Relationship between stress tests and CCTA with ICA
Of the 6,198 patients, 621 (10%) underwent ICA within 90 days. Among them, 230 (59.3%) of 388 with abnormal stress tests and 150 (64.3%) of 233 patients with normal or equivocal stress tests had >50% stenosis on ICA (Table 4). In this population, of all stress test modalities, abnormal TMET had the highest sensitivity (69.4%), positive predictive value (PPV) (62.5%), and negative predictive value (NPV) (47.6%), whereas SE had the highest specificity (48.9%). The large majority (294 of 388, 76%) (Table 4) of stress tests were MPI scans. Overall sensitivity, specificity, PPV, and NPV of abnormal stress tests were 60.4%, 34.2%, 59.3%, and 35.2%, respectively. The sensitivity, specificity, PPV, and NPV for >50% stenosis on CCTA compared with ICA (n = 506) were 93.7%, 37.9%, 70.6%, and 79.1%, respectively. The association between findings on CCTA and ICA was significant (OR: 9.085, 95% CI: 5.57 to 14.81, p < 0.001), whereas that between stress tests and ICA was not (OR: 0.79, 95% CI: 0.56 to 1.11, p = 0.1707).
An additional multivariable analysis was performed to evaluate the ability of available data to predict >50% stenosis on ICA (Fig. 3B). The first model included only the FRS (C-statistic 0.59, 95% CI: 0.55 to 0.64). In the second, addition of clinical risk factors not included in the FRS (history of peripheral vascular disease, cerebrovascular disease, and chronic obstructive lung disease) did not change the C-statistic (0.60, 95% CI: 0.55 to 0.65, p = 0.489). No further change was noted by addition of pre-test CAD likelihood or stress test results in the third model (C-statistic 0.60, 95% CI: 0.56 to 0.65, p = 0.375) and fourth model (C-statistic 0.62, 95% CI: 0.57 to 0.67, p = 0.168). However, addition of CCTA results in the fifth model had the greatest effect on its predictive ability, and it had a significantly higher C-statistic (0.69, 95% CI: 0.64 to 0.74, p < 0.0001) compared with the model with FRS alone.
Thus, degree of stenosis on CCTA was the strongest additional predictor of >50% on ICA beyond FRS, with stress test results having no incremental value over clinical assessment.
This large study encompassing a wide variety of institutions that included large academic centers, community hospitals, and free-standing imaging centers suggests that in low-to-intermediate risk patients with prior stress tests and no known CAD, CCTA is being used judiciously and may aid in reducing rates of normal or nonobstructive findings on ICA. In the entire study group, stress test results did not accurately predict CAD on ICA whereas CCTA did. These findings suggest that physicians selected patients appropriately for a second noninvasive test, and the chosen test, CCTA, effectively performed its gatekeeper function. The large majority of the study population, 81.3%, had normal or nonobstructive CAD on CCTA with limited incremental value of stress tests to predict obstructive disease. Additionally, in patients undergoing ICA, the accuracy for detection of >50% stenosis was higher for CCTA than for stress test results. These findings support prior data indicating that because CCTA performs well in ruling out significant CAD, it could be used effectively to determine the need for ICA. This is of particular importance for patients with abnormal or equivocal stress tests who may otherwise be referred to ICA or may undergo prolonged medical therapy under conditions of diagnostic uncertainty. Ruling out obstructive CAD in >80% of patients in this study group suggests that CCTA may obviate the need for ICA in such patients (9).
The most common stress test used in this study population was MPI (65.1%). False-positive and equivocal/inconclusive MPI scans are known to be the “Achilles heel” of SPECT imaging; fewer than half of all patients with abnormal CCTA have a corresponding perfusion abnormality on SPECT imaging (13,14). Discrepancy between functional and anatomic imaging is well-described in prior studies of ICA as well (6,15). In the present study, these disparities were replicated: only 19.7% of patients with abnormal stress tests had a coronary stenosis >50% on CCTA, and only 59.3% of patients with abnormal stress tests had >50% stenosis on ICA, reiterating the utility of anatomic imaging in these situations (9,16). The accuracy of the various stress testing modalities in this population was lower than previously reported. It is likely that artifacts (i.e., attenuation artifacts on SPECT or poor imaging windows on SE) may have resulted in lower accuracy on imaging stress tests. The lower sensitivities of imaging studies could also be attributed to using >50% (and not >70%) stenosis on ICA as the reference marker. The TMET displayed the highest sensitivity (69%) of all stress modalities, nearly identical to that reported in the literature (17). The lower specificity of all modalities is likely due to the lower risk population included in this study. Thus, the figures presented here demonstrate the exceptionally poor diagnostic performance of all stress tests in this specific group of patients referred for CCTA compared with prior studies.
Patients with normal stress test results were more often symptomatic. Persistence of symptoms with normal stress tests is an “appropriate” indication for CCTA (19), and the presence of persistent symptoms in the majority (93%) of patients with normal stress tests in this study suggests appropriate referral. Our finding that the strongest predictors of obstructive CAD were clinical risk factors also supports the referring physicians' decisions to order CCTA in spite of normal stress findings. Moreover, although abnormal stress test is considered an “uncertain” indication (19), our data demonstrating normal coronaries in the large majority with abnormal stress tests suggest that CCTA is excellent for further evaluation when a false positive stress test result is suspected on clinical grounds.
Most important, the location and degree of abnormality on stress testing was not available. Thus, no concrete conclusions could be reached regarding physiologic and anatomic correlation; any coronary stenosis >50% was considered significant without knowledge of subtended myocardial perfusion data. Additionally, the performance of the FRS may be inaccurate because of imputation of values for lipid levels and blood pressure. However, this method has been used previously in large studies based on registry data (6). Moreover, the analysis of noninvasive test accuracy compared with ICA has several limitations. There was a significant referral bias because CCTA and ICA were not performed in a prospectively pre-specified manner; all studies were performed on clinical grounds. Moreover, the lack of data on reasons for stress imaging studies to be nondiagnostic were not available and may have contributed to their lower accuracy. Furthermore, the large majority of patients in this study group were of low risk, raising the question of inappropriate stress test selection. Additionally, the results are subject to validation bias as not all patients undergo ICA to confirm “true negative” scans. However, such bias is unavoidable in such registries. Last, it is highly likely that patients with “severely” abnormal stress tests were referred directly to ICA, resulting in the discordance between stress test and CCTA findings among the remaining patients included in this study. The demonstration of a correlation between the modalities was not the primary outcome of this study. Moreover, only ICAs performed within the institution as the index CCTA were captured and analyzed as a representative sample. While data on the impact of CCTA on downstream cardiac outcomes, cost effectiveness, and long-term clinical outcomes are urgently needed, these queries were beyond the scope of this study. Further investigation of this question in large-scale prospective trials is warranted. This study lacked the rigors of a randomized controlled study, but represents what was observed in a real-world clinical setting.
Despite the study limitations, our data provide strong evidence that CCTA provides important information in patients with prior stress tests. Previous studies have demonstrated the value of CCTA with prior stress results, effectively substituting for, or serving as a “gatekeeper” to, invasive testing (18,19). Nearly 10% of stress tests performed annually are interpreted as equivocal or inconclusive, and discordance between clinical presentation and stress test results increases the proportion in whom the test data are considered incomplete (9). Although adjudicating stress test findings is not a class I indication for ICA by the American College of Cardiology/American Heart Association guidelines, it is common in clinical practice. In this circumstance, the low diagnostic yield of elective ICA is not justified by increased costs and by the small but definite risk of complications (6). As a preferred strategy, CCTA may be utilized for further risk stratification and prudent patient selection for ICA, with overall lowered costs and radiation exposure (20,21).
A commonly encountered clinical scenario is that of asymptomatic patients with abnormal stress tests. Although routine screening for asymptomatic CAD is not recommended by clinical guidelines (22), patients with cardiac risk factors often undergo stress testing, particularly before surgery (23) or before beginning a vigorous exercise program (21). Stress tests in this population have a very low reported PPV, and such patients are often referred to ICA for definitive diagnosis (24). Our data suggest that CCTA may be an effective diagnostic choice in such patients to adjudicate abnormal or equivocal stress test findings.
It is notable that asymptomatic patients in the present study had a higher rate of obstructive CAD compared with symptomatic patients. However, asymptomatic patients also had higher frequencies of abnormal stress tests compared with symptomatic patients, suggesting that this discrepancy led referring physicians to choose CCTA for further evaluation. In asymptomatic subjects, CCTA has not been demonstrated to be of value when used as a screening tool (25). However, asymptomatic patients in this study differed significantly from the low-to-intermediate risk, self-referred subjects in the study by Choi et al. (25). Although the appropriateness of upstream stress testing in such patients remains questionable, our results suggest that the clinical suspicion of the referring physicians was high enough to warrant the downstream CCTA. An important additional observation was the demonstration of nonobstructive CAD (34%) and obstructive CAD (15%) in patients with normal stress tests. These findings suggest that regardless of symptoms or type and results of stress tests, CCTA provides incremental information that may aid in further risk stratification and determining optimal therapy.
In this large, statewide registry, current practice patterns suggest that in patients with prior stress tests, CCTA is being used appropriately, has high diagnostic accuracy, and may aid in effective utilization of ICA in patients with abnormal findings. In patients with low-to-intermediate CAD risk needing adjudication of stress test results, CCTA may be used to identify patients for whom ICA may provide meaningful information as well as those for whom it may be unnecessary, and serve to support a “gatekeeping” approach to invasive testing.
This study was sponsored and funded by the Blue Cross/Blue Shield/Blue Care Network of Michigan (BCBSM), Southfield, Michigan. Dr. Raff receives grant support from Siemens and Bayer. Dr. Ananthasubramaniam is on the Speaker's Bureau and advisory board for Astellas Pharma, Inc.; on the Speaker's Bureau of and a consultant to Lantheus Medical Imaging; and receives grant support from GE Healthcare, Astellas Pharma Global Development, Trovis Pharmaceuticals, and GlaxoSmithKline. Dr. Abidov is on the Speaker's Bureau of and is a consultant to Astellas Pharma, Inc.; and receives grant support from Sarver Heart Center, Tucson, Arizona. Dr. Share is an employee of BCBSM. All other authors have reported they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- coronary artery disease
- coronary computed tomography angiography
- confidence interval
- Framingham Risk Score
- invasive coronary angiography
- myocardial perfusion imaging
- odds ratio
- stress echocardiography
- single-photon emission computed tomography
- treadmill exercise testing
- Received April 6, 2011.
- Revision received October 11, 2011.
- Accepted October 27, 2011.
- American College of Cardiology Foundation
- Lloyd-Jones D.,
- Adams R.,
- Carnethon M.,
- et al.
- Mowatt G.,
- Cook J.A.,
- Hillis G.S.,
- et al.
- Meijboom W.B.,
- Meijs M.F.,
- Schuijf J.D.,
- et al.
- Abidov A.,
- Gallagher M.J.,
- Chinnaiyan K.M.,
- Mehta L.S.,
- Wegner J.H.,
- Raff G.L.
- Gibbons R.J.,
- Abrams J.,
- Chatterjee K.,
- et al.
- Schuijf J.D.,
- Wijns W.,
- Jukema J.W.,
- et al.
- Gianrossi R.,
- Detrano R.,
- Mulvihill D.,
- et al.
- Greenland P.,
- Alpert J.S.,
- Beller G.A.,
- et al.
- Hashimoto J.,
- Suzuki T.,
- Nakahara T.,
- Kosuda S.,
- Kubo A.
- Newman R.J.,
- Darrow M.,
- Cummings D.M.,
- et al.
- Choi E.K.,
- Choi S.I.,
- Rivera J.J.,
- et al.