ACCF/ACR/AHA/NASCI/SAIP/SCAI/SCCT 2010 Expert Consensus Document on Coronary Computed Tomographic Angiography: A Report of the American College of Cardiology Foundation Task Force on Expert Consensus Documents FREE

Robert A. Harrington, MD, FACC, FAHA, Chair

Eric R. Bates, MD, FACC

Charles R. Bridges, MD, MPH, FACC, FAHA

Mark J. Eisenberg, MD, MPH, FACC, FAHA

Victor A. Ferrari, MD, FACC, FAHA

Mark A. Hlatky, MD, FACC, FAHA

Alice K. Jacobs, MD, FACC, FAHA

Sanjay Kaul, MD, MBBS, FACC

David J. Moliterno, MD, FACC

Debabrata Mukherjee, MD, FACC

Robert S. Rosenson, MD, FACC, FAHA

James H. Stein, MD, FACC, FAHA††

Howard H. Weitz, MD, FACC

Deborah J. Wesley, RN, BSN, CCA

This document was developed by the American College of Cardiology Foundation (ACCF) Task Force on Clinical Expert Consensus Documents (ECDs) and cosponsored by the American College of Radiology (ACR), American Heart Association (AHA), American Society of Nuclear Cardiology (ASNC), North American Society for Cardiovascular Imaging (NASCI), Society of Atherosclerosis Imaging and Prevention (SAIP), Society for Cardiovascular Angiography and Interventions (SCAI), and Society of Cardiovascular Computed Tomography (SCCT) to provide a perspective on the current state of computed tomographic angiography (CTA). ECDs are intended to inform practitioners and other interested parties of the opinion of the ACCF and document cosponsors concerning evolving areas of clinical practice and/or technologies that are widely available or new to the practice community. Topics are chosen for coverage because the evidence base, the experience with technology, and/or the clinical practice are not considered sufficiently well developed to be evaluated by the formal ACCF/AHA practice guidelines process. Often the topic is the subject of ongoing investigation. Thus, the reader should view the ECD as the best attempt of the ACCF and document cosponsors to inform and guide clinical practice in areas where rigorous evidence may not be available or the evidence to date is not widely accepted. When feasible, ECDs include indications or contraindications. Some topics covered by ECDs will be addressed subsequently by the ACCF/AHA Practice Guidelines Committee.

The task force makes every effort to avoid any actual or potential conflicts of interest that might arise as a result of an outside relationship or personal interest of a member of the writing panel. Specifically, all members of the writing panel are asked to provide disclosure statements of all such relationships that might be perceived as real or potential conflicts of interest to inform the writing effort. These statements are reviewed by the parent task force, reported orally to all members of the writing panel at the first meeting, and updated as changes occur. The relationships and industry information for writing committee members and peer reviewers are published in Appendix 1 and Appendix 2 of the document, respectively.

Robert A. Harrington, MD, FACC, FAHA Chair, ACCF Task Force on Clinical Expert Consensus Documents

The writing committee consisted of acknowledged experts in the field of CTA, as well as a liaison from the ACCF Task Force on Clinical ECDs, the oversight group for this document. In addition to 2 ACCF members, the writing committee included 2 representatives from the ACR and AHA and 1 representative from ASNC, NASCI, SAIP, SCAI, and SCCT. Representation by an outside organization does not necessarily imply endorsement.

At its first meeting, each member of the writing committee reported all relationships with industry and other entities relevant to this document topic. This information was updated, if applicable, at the beginning of all subsequent meetings and full committee conference calls. As noted in the Preamble, relevant relationships with industry and other entities of writing committee members are published in Appendix 1.

During the first meeting, the writing committee discussed the topics to be covered in the document and assigned lead authors for each section. Authors conducted literature searches and drafted their sections of the document outline. Over a series of meetings and conference calls, the writing committee reviewed each section, discussed document content, and ultimately arrived at consensus on a document that was sent for external peer review. Following peer review, the writing committee chair engaged authors to address reviewer comments and finalize the document for document approval by participating organizations. Of note, teleconferences were scheduled between the writing committee chair and members who were not present at the meetings to ensure consensus on the document.

This document was reviewed by 15 official representatives from the ACCF (2 representatives), ACR (2 representatives), AHA (2 representatives), ASNC (1 representative), NASCI (2 representatives), SAIP (2 representatives), SCAI (2 representatives), and SCCT (2 representatives), as well as 10 content reviewers, resulting in 518 peer review comments. See list of peer reviewers, affiliations for the review process, and corresponding relationships with industry and other entities in Appendix 2. Peer review comments were entered into a table and reviewed in detail by the writing committee chair. The chair engaged writing committee members to respond to the comments, and the document was revised to incorporate reviewer comments where deemed appropriate by the writing committee.

In addition, a member of the ACCF Task Force on Clinical ECDs served as lead reviewer for this document. This person conducted an independent review of the document at the time of peer review. Once the writing committee documented its response to reviewer comments and updated the manuscript, the lead reviewer assessed whether all peer review issues were handled adequately or whether there were gaps that required additional review. The lead reviewer reported to the task force chair that all comments were handled appropriately and recommended that the document go forward to the task force for final review and sign-off.

The writing committee formally signed off on the final document, as well as the relationships with industry that would be published with the document. The ACCF Task Force on Clinical ECDs also reviewed and formally approved the document to be sent for organizational approval.

The final version of the document, along with the peer review comments and responses to comments were circulated to the ACCF Board of Trustees for review and approval. The document was approved in November 2009. The document was then sent to the governing boards of the ACR, AHA, ASNC, NASCI, SAIP, SCAI, and SCCT for endorsement consideration, along with the peer review comments/responses for their respective official peer reviewers. ACCF, ACR, AHA, NASCI, SAIP, SCAI, and SCCT formally endorsed this document. This document will be considered current until the ACCF Task Force on Clinical ECDs revises or withdraws it from publication.

This document presents an expert consensus overview of the current and emerging clinical uses of coronary CTA in patients with suspected or known coronary artery disease (CAD). Since the evidence base for this technology is not felt to be sufficiently mature to support a clinical practice guideline at present, this ECD offers an alternative vehicle in which the state of the art of coronary CTA can be described without the requirement to provide explicit recommendations accompanied by formal ratings of the quality of available evidence.

The intention of this document is to summarize the strengths and weaknesses of current clinical uses of coronary CTA as reflected in the published peer-reviewed literature and as interpreted by the writing committee. The document is not intended primarily as either a comprehensive literature review or as an instruction guide for those interested in performing or interpreting coronary computed tomography (CT) angiograms. The document also does not offer specific statements rating the appropriateness of various potential clinical uses of coronary CTA, as this has been dealt with in the ACCF/ACR/SCCT/SCMR/ASNC/NASCI/SCAI/SIR 2006 Appropriateness Criteria for Cardiac Computed Tomography and Cardiac Magnetic Resonance Imaging (1). Finally, this document does not address the evaluation of coronary calcium using CT, except as it pertains to CTA studies in patients with suspected or known CAD, since this topic has also been covered in the ACCF/AHA 2007 Clinical Expert Consensus Document on Coronary Artery Calcium Scoring by Computed Tomography in Global Cardiovascular Risk Assessment and in Evaluation of Patients With Chest Pain (1a).

Advances in CT imaging technology, including the introduction of multidetector row systems with electrocardiographic gating, have made imaging of the heart and the coronary arteries feasible. The potential to obtain information noninvasively comparable to that provided by invasive coronary angiography has been the major driving force behind the rapid growth and dissemination of cardiac CT imaging. In the future, the ability of CTA to provide information not currently available from invasive angiography may provide the basis for a major shift in how patients with atherosclerotic cardiovascular disease are classified and managed. Currently, cardiac CTA can provide information about coronary anatomy and left ventricular (LV) function that can be used in the evaluation of patients with suspected or known CAD.

The technology for performing coronary CT angiograms is evolving at a rate that often outpaces research evaluating its incremental benefits. Multidetector CT technology prior to 64-channel or “slice” systems should now be considered inadequate for cardiac imaging (except for studies limited to assessing coronary calcium). The incremental value of recently introduced CT hardware with 128-, 256-, and 320-channel systems over 64-channel systems has not yet been determined. As with any diagnostic technology, coronary CTA has technical limitations with which users should be familiar, and proper patient selection and preparation are important to maximize the diagnostic accuracy of the test. Most cardiac CTA examinations result in a large 4-dimensional (4D) dataset of the heart obtained over the entire cardiac cycle. Physicians who interpret these examinations must be able to analyze the image data interactively on a dedicated workstation and combine knowledge of the patient with expertise in coronary anatomy, coronary pathophysiology, and CT image analysis techniques and limitations. In addition, integration of coronary CTA data into clinical practice requires that the results be evaluated in terms of what was known diagnostically and prognostically before the test was performed and, thus, what incremental information the test provides. The ability of a test such as coronary CTA to provide incremental diagnostic information that alters management (as contrasted with increasing diagnostic certainty alone) is heavily dependent both on the pretest probability and on the alternative diagnostic strategies considered.

The published literature on the diagnostic accuracy of 64-channel coronary CTA compared with invasive coronary angiography as of June 2009 consists of 3 multicenter cohort studies along with over 45 single-center studies, many of the latter involving fewer than 100 patients. This literature reflects careful selection of study subjects and test interpretation by expert readers, typically with exclusion of patients who would be expected to have lower quality studies, such as those with irregular heart rates (e.g., atrial fibrillation), obesity, or inability to comply with instructions for breath holding. In addition, because the cohorts for these studies were assembled from patients referred for invasive coronary angiography, they do not necessarily reflect, in terms of obstructive CAD prevalence or clinical presentation, the population to which coronary CTA is most likely to be applied in clinical practice. Accepting these caveats, some consistent conclusions emerge from this literature that may be useful in clinical decision making. In these studies, overall sensitivity and specificity on a per-patient basis are both high, and the number of indeterminate studies due to inability to image important coronary segments in the select cohorts represented is less than 5%. In most circumstances, a negative coronary CT angiogram rules out significant obstructive coronary disease with a very high degree of confidence, based on the post-test probabilities obtained in cohorts with a wide range of pretest probabilities. However, post-test probabilities following a positive coronary CT angiogram are more variable, due in part to the tendency to overestimate disease severity, particularly in smaller and more distal coronary segments or in segments with artifacts caused by calcification in the arterial walls. At present, data on the prognostic value of coronary CTA using 64-channel or greater systems remain quite limited. Furthermore, no large-scale studies have yet made a direct comparison of long-term outcomes following conventional diagnostic imaging strategies versus strategies involving coronary CTA.

As with invasive coronary angiography, the results of coronary CTA are often not concordant with stress single-photon emission computed tomography (SPECT) myocardial perfusion imaging (MPI). The differences in the parameters measured by MPI (“function” or “physiology”) and CTA (“anatomy”) must be considered when making patient management decisions with these studies. Of note, a normal MPI does not exclude the presence of coronary atherosclerosis although it does signify a very low risk of future major adverse events over the short to intermediate term. Conversely, coronary CTA allows detection of some coronary atherosclerotic plaques that are not hemodynamically significant. The optimal management of such disease has not been established. Neither test can presently identify with any reasonable clinical probability nonobstructive coronary plaques that might rupture in the future and cause acute myocardial infarction (MI). Invasive coronary angiography has a similar limitation.

Studies comparing coronary CTA with fractional flow reserve (FFR) measured as part of invasive coronary angiographic studies complement the MPI comparisons described in the preceding text by showing that coronary CTA anatomic data do not provide very accurate insights into the probability that specific lesions will produce clinically significant ischemia. Similar observations have been made about the relationship of FFR data and the anatomic information provided by invasive coronary angiography.

In the context of the emergency department evaluation of patients with acute chest discomfort, currently available data suggest that coronary CTA may be useful in the evaluation of patients presenting with an acute coronary syndrome (ACS) who do not have either acute electrocardiogram (ECG) changes or positive cardiac markers. However, existing data are limited, and large multicenter trials comparing CTA with conventional evaluation strategies are needed to help define the role of this technology in this category of patients.

Coronary CTA imaging of patients with prior coronary bypass surgery yields very accurate information about the state of the bypass grafts but less accurate information about the native arteries distal to the bypasses and the ungrafted arteries. Because chest pain after bypass surgery might be associated with disease progression in either a graft or a native coronary artery, the difficulty of accurately assessing the native vessels is an important limitation for the clinical use of coronary CTA in the post-bypass patient.

Coronary stents pose some significant technical challenges for coronary CTA, since the metal in the stents may create several types of artifacts in the images. Special algorithms are now routinely used that may reduce some of these artifacts during image reconstruction. The literature suggests that in patients who have large diameter stents, good image quality, and whose clinical presentation suggests low-to-intermediate probability for restenosis, 64-channel coronary CTA can be used to rule out severe in-stent restenosis. There are no studies that directly compare a coronary CTA strategy with an invasive coronary angiography strategy in patients with coronary stents, and such data will be required to understand the efficiencies and tradeoffs of these 2 strategies in this population.

The literature on the assessment of LV function using cardiac CTA in patients with suspected or known CAD is much smaller than that for diagnostic coronary imaging. One likely reason is that echocardiography already provides a readily available, noninvasive means of assessing ventricular function and wall motion and does so without exposing patients to ionizing radiation or iodinated contrast agents.

Available comparisons with cardiovascular magnetic resonance (CMR) suggest that CTA estimation of LV ejection fraction is accurate over a wide range of values. Accuracy may, however, be reduced at higher heart rates due to difficulties in capturing end-systolic and end-diastolic phases accurately. Use of some newer strategies to reduce the radiation dose of coronary CTA studies, such as sequential scanning, will eliminate the ability to assess LV function with the same study.

The writing committee considered several emerging applications where empirical data were deemed insufficient to support development of a consensus. Imaging of noncalcified coronary plaques may in the future become a useful application for coronary CTA, but it has no role in current practice since there are insufficient data to assess its clinical utility. CTA assessment of total atherosclerotic burden and potential plaque vulnerability similarly will require substantial additional technical development and clinical investigation to define their potential value in patient management.

The writing committee identified 3 areas without consensus: the interpretation of incidental noncardiac findings on the CT examination, the use of coronary CTA in asymptomatic subjects, and the “triple rule-out” examination of patients with acute chest pain in the emergency department.

Use of coronary CTA raises 2 important safety issues: 1) the amount of radiation absorbed by the body tissues; and 2) the exposure to iodinated contrast agents that have the potential to produce allergic reactions and acute renal injury. Median effective radiation dose (which is a calculated rather than empirically measured quantity) for coronary CTA with current technology was 12 mSv in a cross-sectional international study of 50 sites (both academic and community) assessed in 2007. Individual sites in this study varied from a median of 5 to 30 mSv. In a 15-hospital imaging registry in Michigan in 2007, prospective use of a set of best practice radiation dose reduction recommendations resulted in a reduction in the average scan effective radiation dose from 21 mSv to 10 mSv with no reduction in image quality.

Several preliminary economic studies using claims data and/or modeling have examined the use of coronary CTA in the diagnostic evaluation of suspected coronary disease and in the evaluation of acute chest pain in the emergency department. Within the limits imposed by the data available, these studies suggest that a diagnostic strategy using coronary CTA may potentially reduce both the time spent in the diagnostic process and the overall costs of clinical evaluation in selected populations, particularly in lower-risk subjects who otherwise would have been subjected to more expensive and possibly less accurate testing strategies. However, longer-term empirical studies will be required to establish the full economic impact of this technology in contemporary practice.

This document focuses on the perspective of clinicians caring for patients with suspected or known CAD in evaluating the potential current uses for cardiac CTA. Therefore, the use of cardiac CTA for other primary clinical questions, such as the diagnosis of pulmonary embolism, pulmonary parenchymal disease, pericardial disease, cardiac masses, arrhythmogenic right ventricular dysplasia, thoracic aortic disease, and congenital heart disease will not be directly addressed. Such disorders, of course, are relevant to the subject matter of this report when they are identified by the cardiac CT angiogram as a possible cause of the patient's symptoms. This report does consider cardiac CT angiographic estimation of LV ejection fraction and evaluation of regional wall-motion abnormalities because these findings may help refine the assessment of the severity and clinical relevance of CAD. Detection of coronary calcium by CT has been addressed in the ACCF/AHA 2007 Clinical Expert Consensus Document on Coronary Artery Calcium Scoring by CT in Global Cardiovascular Risk Assessment and in Evaluation of Patients With Chest Pain (1a), and therefore will not be considered here except where assessment of coronary calcification is relevant to the performance and interpretation of coronary CTA. Information provided by coronary CTA that is relevant to the patient with suspected or known CAD is considered to the extent made possible by the available published evidence. The writing committee felt that abstracts and oral presentations were not sufficiently reliable sources to be used in the construction of this document.

Noninvasive coronary imaging requires a system capable of acquiring motion-free, high spatial resolution images within less than 20 seconds, while patients are holding their breath. Current generation 64-channel multidetector row computed tomography (MDCT) fulfills these requirements reasonably well (2). This section will briefly review selected technical and interpretive issues specifically relevant to the performance of MDCT coronary imaging. Readers of the literature should not be confused by the fact that several equivalent terms are used to refer to this technology, including multidetector CT, multidetector row CT, multislice CT, and multichannel CT.

Appropriate patient selection and preparation are major preimaging determinants of image quality. Key aspects of the imaging process include heart rate and rhythm control, the proper timing of the scan relative to the introduction of the intravenous contrast bolus into the circulation, and minimization of patient motion. Interactive image reconstruction techniques are critical to proper diagnostic interpretation but cannot remedy deficiencies in collection of raw radiographic data. The determinants of patient radiation dose and the trade-offs between radiation dose and image quality are discussed in (14), Safety Considerations.

Image quality of coronary CTA is improved by achieving a slow, regular heart rate, excluding very obese patients, selecting patients able to cooperate with instructions to be motionless and to hold their breath during imaging, and by assessing the presence and distribution of coronary calcification. All of these are evident from an initial patient evaluation except coronary calcification, which is typically assessed during the precontrast scans taken at the start of imaging. At present, there is no firm consensus on the extent of coronary calcification that precludes a technically adequate coronary CT angiogram. Innovations in the scanning process currently under investigation may reduce the importance of this issue in the future.

Patient preparation steps include achieving intravenous access, typically in an antecubital vein suitable for contrast administration at a flow rate of 4 to 6 mL/s, and administering preprocedure beta blockade when needed to achieve the desired heart rate and rhythm. Administration of sublingual nitroglycerin can be used to enhance coronary vasodilatation at the time of imaging. Rehearsal of the breath hold with the patient improves compliance, serves to decrease patient anxiety, and may lessen motion artifact as a result. The rehearsal of breathing instructions can also be used as an opportunity to identify any unusual effects that might occur to heart rate and regularity from breath holding in individual patients.

CT is an excellent method of creating high-resolution, volumetric images of body structures that can be held relatively stationary. In such situations, current generation CT systems can resolve very small, submillimeter, abnormalities. Movement of the target organ creates the need for high-temporal resolution to reduce motion-related blurring artifacts. Two kinds of motion, respiratory and cardiac, must be controlled during CT imaging of the coronary arteries. Careful patient selection and preimaging coaching can control respiratory motion via a voluntary breath hold. Breath-hold times on 64-channel systems for a cardiac CTA range from 10 to 15 seconds (may be shorter on systems with 128 channels or higher) and are well within the capability of most patients, even those with respiratory compromise. Strategies to “control” cardiac motion rely on a combination of pharmacology and technology. The coronary arteries move in a complex pattern through space during each cardiac cycle. Each coronary artery moves at a different velocity and in a different pattern from the others, and even the individual segments of each coronary do not move uniformly (3). Because coronary artery velocity and acceleration during the cardiac cycle increase with higher heart rates, preimaging heart rate control with beta blockers is commonly used to slow coronary motion and is an important part of patient preparation (4). In 1 multicenter international study of 1965 patients undergoing coronary CTA, 12% of subjects were on daily beta blockers prior to study and an additional 46% received beta blockers in preparation for their scan (5). As the heart rate decreases, the phase of relative cardiac quiescence in mid- to late-diastole (at approximately 60% to 75% of the R to R interval) widens. With a sufficiently slow heart rate, typically between 50 to 65 beats per minute, ECG gating can be used to select (retrospectively) the portion of the cardiac cycle for image reconstruction where the motion of each coronary segment is at a minimum. Due to different patterns of motion during the cardiac cycle, the optimal images for defining details of the right coronary artery may occur in a different phase of the cardiac cycle from the optimal images of the left coronary artery.

The temporal resolution of a CT scan, or the ability to resolve separate points in time, is determined in CT acquisition by the time required to acquire the data for reconstruction of a single transverse section or “slice.” Thus, the speed of gantry rotation (the gantry contains both the X-ray source and the detector array and is rotated around the patient during imaging) is one of the primary determinants of the temporal resolution of the MDCT scan. The minimum gantry rotation time on current generation scanners (the time required to complete a 360° rotation) is between 280 and 400 ms, depending on the manufacturer and model. Tremendous centrifugal forces are created by the need to spin the imaging components inside the gantry around the patient and significant further increases in rotation times are limited by the ability of current mechanical components to withstand such forces. Thus, alternative methods have been employed to further improve temporal resolution. The routine use of half-scan reconstruction results in an effective temporal resolution of approximately one half the time required for the CT gantry to complete a single 360° rotation or approximately 140 to 200 ms (6). Other methods include the use of partial scan reconstructions from multiple adjacent cardiac cycles to improve effective temporal resolution. In 2007, a “dual-source” CT scanner was introduced that contained 2 X-ray sources and 2 sets of detectors offset 90° from each other in the CT gantry (7). This configuration is able to achieve an additional improvement in temporal resolution (to approximately 83 ms) by combining the data from the 2 detectors using just 90° of gantry rotation as opposed to the required 180° of gantry rotation needed with a single-source system (8). However, half-scan and partial-scan reconstructions may decrease spatial resolution due to misregistration artifacts. For reference, conventional invasive angiography using 30 frames per second has a temporal resolution of approximately 33 ms (9).

Spatial resolution of a CT scan is defined in terms of the in-plane or x-y axis resolution and the through-plane or z-axis resolution. The x-y spatial resolution of a CT scan is the smallest distance between 2 high-contrast objects that still allows recognizing the objects as separate. Modifiable parameters that can affect in-plane resolution include the reconstruction algorithm that translates the projection data into planar images, the reconstructed field of view and the image matrix size (typically 512 × 512 pixels). The principal limit on the z-axis, or “slice” resolution (along the patient's long axis), lies in the detector array geometry. Within the detector array are rows of array elements, which are typically 0.4 to 0.6 mm in size along the z-axis. Thus, a “64-detector row CT” generally has 64 rows of detectors in its detector array. The width of the X-ray beam is collimated (i.e., physically limited) in relation to the width of the detector array, which can vary among different CT systems from 20 to 160 mm along the z-axis.

During data acquisition, the CT system records the “raw” scan data and converts it to X-ray attenuation Hounsfield units (HUs). This file of raw projection data is used to reconstruct axial images, most commonly using a filtered back-projection algorithm (a standard algorithm for reconstructing CT images). Each image is reconstructed into a 512 × 512 matrix for display. If the reconstructed image has, for example, a field of view of 260 mm, the pixels in the resulting image would have a nominal size of 0.5 mm × 0.5 mm (i.e., 260 mm/512=0.5 mm). With detector elements measuring 0.6 mm along the z-axis (see the preceding text), this example would result in each volume data element, or voxel, measuring 0.5 mm × 0.5 mm × 0.6 mm in the x, y, and z dimensions, respectively. These 3-dimensional (3D) “voxels” have the desirable property of being “near-isotropic”: each voxel of the dataset has nearly the same size in all 3 dimensions. What this means practically for physicians is that the data can be displayed on a workstation in any plane or orientation without sacrificing spatial resolution. This capability is critical for cardiac and coronary imaging and allows visualization of the heart in the axial planes acquired as well as the short axis, vertical long axis (2-chamber), and horizontal long axis (4-chamber), all from the same acquisition. For coronary imaging, the near isotropic datasets provide views of each coronary artery segment along both its long axis and short axis (i.e., cross section).

For a coronary luminal diameter of about 3 mm, a cross section reconstructed from a CT scan with cubic voxel dimensions of 0.5 mm per edge will display the diameter of the lumen using about 6 voxels. Because disease cannot be resolved at the subvoxel level, the voxel size relative to the object being imaged defines the limits of quantitative resolution. Thus, grading of coronary lesions with coronary CTA can be done at the ordinal level, but full quantification remains problematic (10). For reference, invasive coronary angiography has a spatial resolution of about 0.16 mm (9). Thus, a 3-mm coronary artery lumen would be displayed using about 18 pixels, providing the opportunity for much more accurate quantification of disease affecting the coronary artery lumen.

The number of longitudinal detector rows/data channels that can independently measure X-ray attenuation simultaneously determines the volumetric coverage of the CT scanner, or the amount of the cardiac volume (which in adults is about 12 cm in the axial dimension) that is imaged with each CT gantry rotation. Using current generation 64-channel scanners, routine submillimeter imaging can be performed with scan durations of 10 to 20 seconds and longitudinal coverage of 20 to 40 mm of cardiac anatomy per gantry rotation. However, a 64-channel CT system that involves 32 detector rows and 2 focal spot positions (32 × 2=64 data channels) does not have the same volumetric coverage as a system with 64 detector rows. To cover the entire heart in the most common mode of scanning, multiple 360° gantry rotations gated to the cardiac cycle are used as a motorized table moves the patient through the CT scanner. Thus, the X-ray beam traces a continuous helical (spiral) path around the section of the patient's body being imaged.

Some institutions are now also using 128-channel scanners, and both 256- and 320-channel scanners have been introduced (11). The latter configurations offer the potential to image the entire heart during a single heartbeat (12- 13). While this sounds like a theoretically attractive next step in CT technology, substantial technical challenges are imposed by the creation of CT scanners that require the use of a cone X-ray beam as wide as 16 cm in the z-axis direction. Although volumetric coverage is increased with these new scanners, increasing the number of detector rows does not by itself improve spatial or temporal resolution above that provided by 64-channel scanners. The benefits and limitations of these newly introduced CT scanners will not be known until formal analyses of image quality, diagnostic accuracy, radiation dose, and clinical performance are evaluated in appropriate large multicenter studies.

Coronary CTA examinations are typically performed using nonionic intravenous contrast medium with high iodine concentration (greater than or equal to 300 to 350 mg I/mL) to assure adequate opacification of the coronary artery lumen and sufficient contrast with the arterial wall. This contrast injection is followed by injection of normal saline to “push” contrast through the venous capacitance of the upper extremity and the right heart structures (14). The contrast injection should result in a high-level plateau of arterial opacification (greater than 300 to 350 HUs) during CT image acquisition. Several different methods, including the use of a test bolus and automated bolus tracking, are available to ensure that the period of maximum concentration of intravenously administered contrast material in the coronary arteries is properly synchronized with the period of scan acquisition. If the contrast bolus arrives either too early or too late, coronary image quality will be diminished, and diagnostic information may be lost. Adult coronary CTA requires the use of a high injection rate, typically 4 to 6 mL/s, with the duration of injection and the volume of intravenous contrast agent prescribed based on the structures to be imaged and the specifics of the CT systems used to acquire the exam.

Image reconstruction is the process of converting raw CT attenuation data into axial (i.e., transverse) sections. Although much of the process involves the use of proprietary mathematical algorithms developed by each manufacturer, some elements are under the control of the technician working with the study. Decisions that must be made during this initial processing of the data include use of options for noise reduction and correction for any evident blurring or motion artifact. In order to find the part of the cardiac cycle that best captures motion-free images of the right and left coronary arteries, multiple phases of the cardiac cycle may need to be reconstructed and examined. Decisions about this are operator specific, with some choosing to create up to 20 reconstructions at 5% increments of the R to R interval (from 5% to 95%). Thus, a coronary CT angiogram may result in 350 to over 5000 transverse sections available for physician examination, with most coronary CT angiograms falling into a range of 1500 to 3000 images.

The final phase of the CT angiogram study is the creation of 2-dimensional (2D) reformatted images and 3D volume-rendered images from the transverse reconstructions. The approach to interpretation of a CT coronary angiogram varies by operator, but some general principles can be described. For most experts, the source transverse sections supplemented with oblique reformations are the primary tools for interactive interpretation of coronary CTA examinations (15). In addition, interpretation of the transverse sections provides a general understanding of the anatomic relationships of the heart and coronary arteries with surrounding structures such as the great vessels, nonvascular mediastinal structures, lungs, and pleura. Comparisons of abnormalities detected on reformatted or 3D-rendered images with the source transverse sections may help to minimize errors in interpretation related to postprocessing artifacts.

Multiplanar reconstruction images can be oriented along any plane within the imaged volume, making it possible to view the long and short axis of the coronary artery segments and the cardiac chambers. Curved planar reformation images can be created manually or by using vessel centerline tracking algorithms to display the course of a coronary artery. Curved planar reformation images display the coronary artery as if it were stretched along a hypothetical straight line. Distortion is a concern with these reconstructions, particularly if vessel tortuosity creates difficulties for the computer's vessel-tracking calculations. Branch points can similarly be problem areas. The maximum intensity projection is a visualization technique that combines data from a user defined “slab” (i.e., multiple adjacent “slices”) to produce a single summary image that displays the maximum intensity along each projection through the slab from the perspective of the display (or viewer). This allows the course of a contrast-filled coronary artery to be viewed as if one could see through the slab instead of only being able to see the portion of the artery on the surface of the slab. To look at the full length of a coronary artery, a “sliding slab” technique may be used that allows the operator to move the slab along the entire course of the artery interactively (16).

Volume rendering provides a 3D reconstruction that can be useful for displaying large amounts of data in a single view. The technique requires removing structures through editing or by setting levels of opacity (windows) for display. Volume rendering can be valuable for understanding the distribution of coronary arterial supply to the underlying myocardium and the position and course of coronary bypass grafts but is not considered reliable for detecting and grading coronary stenoses. Neither volume-rendering images nor maximum-intensity projection images are sufficient by themselves for assessing the distribution and severity of coronary atherosclerosis.

As workstation capabilities improve, more complex reconstructions become possible, potentially reducing the amount of physician time required for each study. At present, reconstruction is highly operator dependent. The extent to which variations among operators may influence the quality of diagnostic information provided has not been empirically tested. In addition, there are no universally accepted conventions or standards for the display of cardiac or coronary images, in contrast with echocardiography and nuclear cardiology. The SCCT has recently published a consensus document covering the interpretation and reporting of coronary CTA studies (17). The complexity of the physician–computer interaction poses substantial challenges to those desiring to assess the performance of this technology, since it may be difficult to assess whether specific aspects of this interaction vary across centers and practices, and if so, whether the differences improve or impair diagnostic performance.

There are 2 basic diagnostic approaches to symptomatic patients with chest pain, loosely referred to as “anatomic” and “functional.” Anatomic tests, such as coronary CTA and invasive coronary angiography, provide direct radiographic visualization of the structural features of the coronary artery lumen. Invasive coronary angiography creates a 2D coronary “lumenogram.” By moving the fluoroscopy unit, or by using a biplane system, multiple projections of the lumen of each coronary artery can be obtained. To estimate coronary artery stenosis severity from this technique, one must compare any evident narrowing of the luminal outline with presumably normal adjacent segments to allow estimation of a “percent diameter stenosis.” This visual grading process, which is still the standard clinical method of interpretation used in catheterization laboratories around the world, has a high degree of intra- and interobserver variability (18- 20). Computer-assisted interpretation, which could serve to reduce at least some of this variability, has not yet been accepted into routine clinical practice.

Invasive coronary angiography is considered the “reference standard” for diagnostic coronary testing, despite the foregoing limitations, for several reasons. First, until the advent of 16-channel CT coronary angiography, it was the only method of directly visualizing the lumens of coronary arteries that was suitable for routine clinical use. Second, the assessment of luminal stenosis severity on coronary angiography, typically summarized in a very simple 1-, 2-, or 3-vessel obstructive disease ranking, has been repeatedly demonstrated to be one of the most important prognostic factors in patients with coronary disease (21- 22). Finally, the results of invasive coronary angiography have formed the basis for revascularization treatment selection decisions for almost 40 years. Thus, invasive coronary angiography is the reference standard in coronary assessment primarily because of the extensive evidence documenting its value in patient management and secondarily because of its higher spatial and temporal resolution compared with alternative coronary imaging options.

The extensive evidence base relating invasive angiography results to prognosis and patient management cannot necessarily be extrapolated to the findings of coronary CTA. However, it is worth noting that invasive coronary angiography itself has undergone major changes in imaging methodology, evolving from an analog film image intensifier system to digital image generation using flat-panel detectors. No empirical studies have yet examined whether this change in technology, which has had a significant impact on the fundamental imaging characteristics as well as radiation exposure, has altered the relationships between test results and patient outcomes.

Coronary CTA provides information about the coronary lumen that approximates the information available from invasive coronary angiography. In addition, it provides information about the presence of nonobstructive plaque in the vessel walls. Invasive coronary angiography is subject to uncertainties about whether the reference segment itself is diseased with plaque and whether the luminal narrowing is concentric or eccentric (19). Coronary CTA is able to image the plaque that is external to the lumen and display its relationship with the lumen. As with invasive coronary angiography, visual grading of coronary segment narrowing by ranges of stenosis is the current standard of practice and has been shown to provide useful clinical information relative to invasive coronary angiography (10,23). Quantitative coronary CTA has been used in some research applications but is not currently a routine part of clinical interpretation (24- 26). In a recent multicenter study, visual and quantitative assessments of stenosis severity by coronary CTA were quite similar (23).

Functional tests assess the ability of coronary arteries (including their collateral vessels) to provide a sufficient blood supply to the myocardium both at rest and during exercise or pharmacological stress. The detection of myocardial ischemia using this approach relies on measuring parameters such as LV blood flow/perfusion patterns or LV function and wall-motion patterns that reflect the impact of reduced blood supply and its consequences. Functional testing data therefore reflect both the severity and consequences of obstructive CAD and are prognostically incremental to anatomic imaging in several important clinical settings (27- 28). The apparent dissociation between anatomic imaging results and functional test results can be attributed to several issues (Table 1). Thus, despite the detailed anatomic information it provides, CTA may not eliminate the need for assessing the functional significance of lesions of intermediate or indeterminate severity.

LV function, as reflected by the ejection fraction, is the single most important prognostic parameter in patients with established CAD. In addition, LV size and regional wall-motion data can influence decisions about appropriate therapies. Several methods provide quantitative evaluation of LV function, including transthoracic echocardiography (TTE), gated SPECT, radionuclide angiography, invasive left ventriculography, and cardiovascular magnetic resonance. LV assessment by CTA is based on use of retrospective gating with reconstruction of up to 20 phases of the cardiac cycle including end-systole and end-diastole. Many of the desired LV functional calculations can be automated using the workstation software, although some operator interaction with manual correction is often required. Clinical use of these CT-derived data requires proper understanding of features unique to coronary CTA compared to other more familiar methods of assessing LV structure and function measures. Values for LV volume, LV ejection fraction, and LV mass for cardiac CTA have recently been reported from a series of 103 apparently healthy adults free of hypertension and obesity (mean age 51 years) (34).

The temporal resolution of current-generation 64-channel multidetector scanners, reviewed briefly in (7), Coronary CT Angiography: Brief Overview of the Technology, is less than that of echocardiography and invasive LV angiography. Cardiovascular magnetic resonance can generate images with higher average temporal resolution secondary to acquiring data over multiple cardiac cycles. Limited temporal resolution of CTA is primarily relevant at higher heart rates and with the use of single-source MDCT scanners because fewer discrete time points of the cardiac cycle can be properly reconstructed, and may produce inaccuracies in LV parameter measures due to improper identification of end-diastole and end-systole. At heart rates between 55 to 65 beats per minute, however, current 64-channel CTA provides sufficient cine frame rates to provide LV function information with accuracy comparable to other noninvasive and invasive modalities.

Clinicians caring for a patient with suspected or known CAD typically consider 3 types of questions. First, is coronary disease present in this patient, and if present, what is its current extent? This is a diagnostic question and effectively addresses the likelihood of certain findings if a reference standard test was performed. As discussed in the preceding text, invasive coronary angiography is the current reference standard diagnostic test for defining the presence and severity of obstructive CAD based on luminal stenosis. However, this status is based more on demonstrated value in defining prognosis and choosing treatment than on documented ability to provide accurate and reproducible assessment of the extent and severity of coronary atherosclerosis. Other technologies that can image the diseased vessel wall, such as intravascular ultrasound (IVUS), CMR, and optical coherence tomography, may actually be a more appropriate reference standard for some aspects of CTA's diagnostic performance given the ability of CTA to image the vessel wall in addition to the lumen.

Second, is this patient likely to suffer a major fatal or nonfatal cardiovascular event in the foreseeable future? This is a prognostic question and addresses the ability of CTA to help stratify risk.

Third, will CTA help clinicians alter management in ways that lead to reduced risk of major adverse clinical events? This is a therapeutic question and addresses the ability of the information derived from the test to help clinicians alter patient outcome.

Research studies evaluating novel diagnostic tests should consider using study designs that minimize biases and maximize generalizability. These designs often include the following features: 1) selection of the study patients consecutively or at random from the target population at multiple centers; 2) performance of both the new test of interest and the reference standard test (e.g., coronary CTA and invasive coronary angiography) in all patients in random order; 3) interpretation of both tests by multiple readers who are completely blinded to any clinical information including the results of other tests and who reflect the spectrum of readers likely to interpret the test in clinical practice; 4) assessment of intra- and interobserver variability for both studies. These methodological ideals have rarely been achieved in practice for any noninvasive imaging test, due to logistics, funding, and other barriers. As a consequence, a number of important biases may distort measured diagnostic performance. For example, most studies of the diagnostic accuracy of CTA have focused on patients who were already referred for invasive coronary angiography (24,35). While this study design is appropriate if CTA will be used as a direct replacement for invasive angiography, it is not ideal if the study population is substantially different from the one in which the test is most likely to be used clinically. Although recognition of such potential biases is an important part of the due diligence involved in vetting any new test for clinical practice, it is also important to recognize that virtually all the tests already accepted as a part of routine clinical practice, including stress nuclear and stress echo tests, had similar bias problems in their initial assessment and reported diagnostic performance (36). A few of the more important bias problems that occur regularly in the diagnostic testing literature are summarized in (Table 2).

What clinicians most want to know from the use of tests for diagnostic purposes are the post-test probabilities: “given the observed test result, what is the new (revised) probability my patient does/does not have disease?” These probabilities are often referred to as “predictive values,” but this latter term has been a source of confusion to many in that it implies that these probabilities are fixed performance characteristics of diagnostic tests. Post-test probability, on the other hand, clearly indicates an estimate that is a revision of an earlier estimate (the pretest probability). To calculate these probabilities, one can employ Bayes' formula for simple cases or logistic regression models for more complex cases. Most of the predictive value/post-test probabilities reported in the coronary CTA literature are calculated from the study sample using 2 × 2 tables of sensitivity/specificity versus obstructive CAD present/absent. Because these estimates are valid for the study population from which they were derived, they may not be relevant to other patient populations. The critical factor to remember is that post-test probabilities may vary importantly according to pretest probability, and a given reported “predictive-value” figure does not apply across all possible pretest probabilities.

Likelihood ratios and receiver-operator characteristic (ROC) curves provide 2 useful and complementary ways of summarizing diagnostic test accuracy. Neither is dependent on disease prevalence per se, although both are affected by changes in the distribution of the severity of disease in the population being tested. A likelihood ratio is the likelihood of a given test result in a patient with disease relative to the same test result in a patient without disease (37). For a positive test, the likelihood ratio is calculated as (sensitivity/[1 − specificity]), and higher values indicate that the test in question is more accurate at identifying patients with disease, particularly if the value is 10 or greater. For a negative test, the likelihood ratio is calculated as ([1 − sensitivity]/specificity), and values less than 0.1 indicate a test particularly accurate at ruling out disease.

ROC curves display in graphical form the relationship between the true positive rate of a test (its sensitivity) and its false positive rate (1 − specificity) because the definition of a “positive” test is varied. Calculation of the area under the ROC curve provides a useful numeric summary measure that ranges from 1.0 (a perfect test) to 0.5 (a completely noninformative test). Statistical comparison of the ROC areas for 2 or more tests assessed in the same study population may be used to identify the more accurate test providing that the curves are of similar shape.