Dr. Sampson is currently a Fellow in Cardiovascular Diseases in the Cardiovascular Medicine Division at Vanderbilt University, Nashville, TN. He received his medical degree from the University of Ibadan, Nigeria. Dr. Sampson received his Masters of Public Health and Business Administration from the University of Medicine and Dentistry-Robert Wood Johnson Medical School and Rutgers University, New Brunswick, NJ. He trained in Internal Medicine at the Brooklyn Hospital Center, Cornell University, NY. Following his residency, he received the Bowen Brooks Fellowship award of the New York Academy of Medicine for advanced studies at the University of Oxford, where he studied clinical epidemiology and received a Masters in Evidence-Based Health Care. He then completed clinical and research fellowships in Noninvasive Cardiology, and Nuclear Cardiology, PET-CT, and CT angiography at the Brigham and Women's Hospital, Harvard Medical School, Boston, MA. He is board-certified in Internal Medicine and Nuclear Cardiology. Dr. Sampson plans a career in academic medicine and intends to conduct translational/clinical research with a focus on the applications of molecular and metabolic imaging.

Since the advent of multislice computed tomography (MSCT), remarkable technical advances have occurred. Arguably, 4-slice scanners have become obsolete, and 16-slice systems may soon be considered antiquated with the arrival of faster 64-slice technology. Improvements in spatial and temporal resolution have accompanied this rapid progress in technology, thus allowing for better visualization of coronary anatomy. This systematic review evaluates state-of-the-art cardiac MSCT in the diagnosis of coronary artery disease and discusses its performance relative to competing modalities.

The early detection and treatment of coronary artery disease (CAD) remains paramount, given its morbidity, mortality, and economic conse-quences. ¹ Conventional invasive angiog- raphy is the gold standard for the definitive delineation of coronary anatomy. However, new imaging modalities are now available for the evaluation of CAD, and they are increasingly being used for the diagnostic and therapeutic management of patients. These modalities include cardiac magnetic resonance imaging (MRI), intravascular ultrasonography (IVUS), and cardiac computed tomography (CT), all of which continue to evolve technologically. The invasive nature of conventional angiography and IVUS make cardiac MRI and CT particularly attractive as noninvasive options for CAD evaluation. In the past few years, we have also witnessed the rapid evolution of cardiac positron emission tomography (PET) instrumentation, which now offers hybrid scanners that integrate PET with CT (PET/CT), resulting in higher sensitivity ² for CAD detection. Catalyzed by the emergence of portable rubidium-82 generators, ^3-6 the proliferation of PET/CT instruments has been rapid, accounting for approximately 80% of the new PET units installed in 2003. ⁷

Since the advent of multislice computed tomography (MSCT) in 1998, we have witnessed remarkable technologic advancement. Thus, 4slice scanners have arguably become obsolete, and 16-slice systems may soon be considered antiquated with the arrival of faster, newer- generation 64-slice technology; interestingly, 128- and 256-slice scanners beck-on from the horizon. This rapid progress in technology with associated improvement in spatial and temporal resolution now allows for better visualization of native and non-native coronary anatomy. Thus, numerous studies have assessed the diagnostic accuracy of MSCT in the detection of CAD. However, in the arena of CAD evaluation and treatment where competing diagnostic modalities exist, the exact role of cardiac of CT in the evaluation of patients with known or unknown CAD awaits clear definition. In the context of a systematic review of the diagnostic accuracy of cardiac MSCT for CAD evaluation, this article provides a succinct discussion of its performance relative to other diagnostic tests for CAD.

Methods

Selection criteria

Studies that addressed the diagnostic accuracy of MSCT in the detection of CAD were included in this review. Only studies that utilized either 16- or 64-slice CT technology were selected. Studies had to have enrolled consecutive patients with or without known CAD and explicitly reported quantitative results in the form of sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Studies that focused exclusively on a highly selective group of patients, cardiac CT methodology, plaque characterization or morphology, and calcium scoring were excluded.

Search strategy

The Cochrane Library, MEDLINE, and EMBASE databases were searched for articles focusing on cardiac CT that were published between 2002 and 2006. The search terms used were "coronary artery disease," or "coronary arteriosclerosis," and "tomography, X-ray," or "tomography, X-ray computed" or "tomography, spiral computed." A second, more specific, search using common phrases found in retrieved articles ("cardiac CT," "MSCT," "coronary artery disease," "coronary artery stenoses," "diagnostic accuracy," or "coronary heart disease") was subsequently performed. Only full-text papers written in the English language were identified. An author search was carried out as well as hand search of relevant references. All articles were assessed from titles and abstracts for relevance. In cases of uncertainty of relevance, full text of the article was used to assess relevance.

Data extraction and study appraisal

In addition to data on sensitivity, sensitivity, PPV, and NPV, the following data was extracted from each study: study design, subjects, definition of CAD, and coronary model used in assessment. On the latter, for studies that adopted multiple models for analyses (eg, segment-, vessel-, or patient-based), the default data extracted was based on the fraction of available segments evaluated. Given the variations in study design, setting, and patient population, a quantitative synthesis of data was not deemed appropriate for the purpose of this review.

Results

The initial search returned >4000 papers. The more specific second search returned 213 papers. A total of 28 studies met the eligibility criteria and are summarized according to study design (Tables 1 and 2). ^8-35 Of these, 24 (86%) studies were prospectively conducted (Table 2). The most frequent coronary model utilized in analyses was the segment-based model (79% of the studies). With the exception of 1 study, the detection of significant CAD was defined by the degree of lumen narrowing with a threshold of >50% as the most frequently used criterion. A total of 24 (86%) studies used 16-slice CT systems; of these, 67% reported >90% assessibility of available coronary territories. Of the 4 studies that utilized 64-slice CT technology, 3 succeeded in evaluating 100% of the coronary segments. The details of the studies and results on diagnostic accuracy are summarized in Tables 1 and 2. Overall, the unweighted mean sensitivity, specificity, PPV, and NPV derived from the 28 studies are 90% (range 70% to 100%), 96% (range 86% to 100%), 81% (range 58% to 99%), and 96% (range 76% to 100%), respectively.

Discussion

The recent progress in CT technology has led to improved diagnostic outcome. In a recent meta-analysis, Schuijf et al ³⁶ reported a weighted mean sensitivity and specificity to detect CAD of 80% (range 66% to 90%) and 90% (range 71% to 99%), respectively, for studies that utilized 4-slice CT systems. Of note, the weighted mean assessibility was 78% (range 68% to 100%). However, the reported corresponding values for the newer-generation 16-slice CT technology were sensitivity of 88% (range 70% to 98%), specificity of 96% (range 86% to 98%), and assessibility of 96% (range 83% to 100%). The NPVs derived from these studies are consistently high, with an overall value of 97%. The findings of this current review are similar but also provide preliminary evidence of the performance of 64-slice systems; 3 of the 4 studies that utilized these systems reported 100% assessibility.

Understandably, in addition to improved spatial resolution, faster speed and shorter acquisition time translate to decreased patient breath-hold time, which increases assessibility by reducing motion artifacts secondary to breathing. These technical improvements in combination with better testing protocols have broadened the imaging and testing goals of using cardiac CT. The feasibility of assessing stent restenosis ^10,37 and bypass grafts ^13,23 has been documented. In their assessment of restenosis in 51 patients who had previous stent implantation, Cademartiri et al ¹⁰ reported sensitivity, specificity, PPV, and NPV of 83.3%, 98.5%, 83.3%, and 97.3%, respectively. In a recent study of graft assessment, Burgstahler et al ²³ analyzed 43 bypass grafts and reported a sensitivity of 100%, a specificity 93%, a PPV of 89%, and an NPV of 100%; of note, 41 (95%) of the 43 were analyzable. In a pooled analysis of 7 and 5 studies of graft occlusion and stenosis, respectively, Cademartiri et al ³⁸ reported weighted mean sensitivities of 88% and 84%, and specificities of 98% and 95%, respectively. Because of artifacts-beam hardening and volume averaging-the assessment of coronary stents remains challenging; in this setting, the use of special higher-convolution filters is helpful, as these improve the visualization of stent struts as well as lumen. ³⁸ The solutions to some other limitations have emerged; the use of beta-block-ers to achieve optimum heart rate and electrocardiographic (ECG) editing for mild heart rhythm irregularities now allows for broader patient application with improved results. In a study of 120 patients, Cademartiri et al ¹¹ reported sensitivity and specificity of 92% and 96% for patients with low heart rate (52 ± 4 bpm), and 90% and 92% for patients with higher heart rate (63 ± 5 bpm) ( P <0.05). In another study, the use of ECG editing led to a reduction in the proportion of nonassessible segments from 17% to 2% ²⁵ ; similarly, the sensitivity, specificity, PPV, and NPV of cardiac CT for the detection of significant stenoses before and after ECG editing were 63% and 92%; 97% and 96%; 87% and 87%; 91% and 97%, respectively ( P <0.05). ²⁵ In the same vein, although age and obesity have been implicated in suboptimal image quality, some studies report no significant compromise in diagnostic accuracy for CAD in these patient subgroups ^8,39 ; future experience with the newer-generation 64-slice scanners may provide further clarification. Overall, noninvasive angiography with MSCT has undergone significant developmental milestones such that the acquisition of images of diagnostic quality is now routinely possible in a broad array of patients with known or suspected CAD. The consistently high NPV of cardiac MSCT renders it excellent in ruling out CAD. However, in the management of these patients, the evolution of its role may hinge on an understanding of its performance in relation to competing diagnostic modalities such as cardiac MRI.

Cardiac CT and other tests for CAD

Well-established risk predictors, such as left ventricular (LV) volumes and LV ejection fraction (LVEF), provide added information in the comprehensive evaluation for CAD. Cardiac MRI is the reference standard for the determination of these parameters. However, cardiac CT has been shown to provide estimates of LVEF and LV volumes comparable with values derived from cardiac MRI and echocardiography. ^40,41 Thus, the routine determination of these parameters during the evaluation of the coronary anatomy with MSCT will provide added value for prognostication and risk stratification. Similar agreement exists between assessments of resting segmental wall motion abnormality-a marker for underlying CAD-by MSCT and echocardiography. ⁴² A recent meta-analysis of the comparative diagnostic performance of MRI and MSCT for noninvasive coronary angiography indicated that MSCT has a higher accuracy to detect or exclude significant CAD. ³⁶ Thus, in addition to shorter image acquisition time, cardiac CT is superior to MRI in the detection of CAD while providing comparable information on LVEF and LV volumes. However, cardiac MRI can provide multiple details (eg, rest and stress myocardial perfusion) in combination with ventricular function, mass, and wall motion. Furthermore, MRI provides better assessment of the morphology and characteristics of plaque, although the implications of these data are unknown. It is important to note that MRI does not require contrast injection or radiation exposure, unlike cardiac CT, the effective radiation exposure with ECG-controlled dose modulation is in the range of 7 to11 mSv. ^43,44

Although limited by its invasiveness and confinement to conventional angiography, IVUS is excellent for the detection and characterization of plaque. Its spatial resolution of 100 µm far exceeds that of MSCT (>500µm) or MRI (>500 µm). This excellent spatial resolution allows for the reliable characterization of atherosclerotic plaque, the detection of intimal hyperplasia, and intraluminal thrombus. In the setting of conventional coronary angiography, the role of IVUS includes the assessment of complications, the guidance of angioplasty, and the evaluation of left main disease, stent restenosis, and hazy or unusual lesions. Some studies have shown that the use of IVUS to guide angioplasty may improve cardiovascular outcomes, ^45,46 while others provide conflicting results. ⁴⁷ Further im- provements in resolution may lead to future IVUS applications for the detection of early atherosclerosis, with an emphasis on the determination of plaque vulnerability by visualization of its thin fibrous cap. However, the confinement to invasive coronary angiography limits the application of this modality to a wider patient population as opposed to cardiac MSCT, in which a major advantage is the potential to provide definitive evaluation of subjects with low- and intermediate-risk profiles--the predominant patient population.

The unique one-stop-shop promise of dual-modality PET/CT systems

The understanding that CAD is a disease spectrum should negate an all-or-none paradigm. Therefore, the traditional binary classification or set criterion based on luminal diameter that is visually or quantitatively estimated on CT or conventional angiography may be misleading, as this focuses only on anatomy. The classic pathologic postulate for the development and progression of CAD ⁴⁸ is now widely accepted in the wake of clinical evidence from studies of IVUS. ^49-51 More commonly, luminal stenosis occurs after a period of intimal plaque accumulation and subsequent exhaustion of the positive remodeling process. Therefore, before the appreciation of obstructive luminal stenosis on angiography, plaque accumulation and/or endothelial dysfunction may have physiologic consequences detectable by cardiac PET/CT-a modality that as-sesses perfusion and function rather than anatomy.

Lesions that are not deemed critical on CT angiography may indeed be hemodynamically significant; likewise, the converse is true. The fact that cardiac CT can be performed on the same dual-modality PET/CT systems presents a one-stop- shop promise whereby anatomic information (ie, coronary artery narrowing with or without calcium scores) can be obtained along with the corresponding physiologic information needed to guide patient management decisions. Preliminary experience with this model of evaluation suggests the complementary role of CT and PET for optimizing posttest management decisions in patients with suspected CAD.

Di Carli et al ⁵² studied 79 consecutive patients (mean age 56 ± 11 years, 51% male) with low-intermediate pretest likelihood of CAD, without prior myocardial infarction or revascularization, who underwent stress rubidium-82 myocardial perfusion PET imaging and cardiac CT (16- or 64-slice CT) for diagnostic purposes, a normal CT angiogram was reported to have an NPV of 96% (per vessel analysis) and 91% (per patient analysis) for identifying obstructive CAD. However, among the 63 patients with normal stress PET, only 33 (52%) had normal coronary arteries on cardiac CT, whereas the remaining 30 patients had varying evidence of coronary atherosclerosis on cardiac CT angiogram. Thus, cardiac CT was a poor discriminator of those patients with objective evidence of stress-induced ischemia on cardiac PET evaluation. Conversely, a normal stress PET was a relatively poor discriminator of patients without evidence of non-flow-limiting (subclinical) coronary atherosclerosis. ⁵²

Conclusion

At its current level of development, cardiac CT has emerged as a modality of high diagnostic performance in noninvasive coronary angiography. It has a very high negative predictive value, which makes it suitable for the definitive evaluation of low- to intermediate-risk populations. The detection of plaques in such patient populations could provide the impetus for aggressive medical therapy with appropriate cardiac protective agents (eg, statins). The advent of dual-modality systems potentially allows for the combined evaluation of anatomy and physiology. The ability for excellent evaluation of bypass grafts and increasingly reasonable visualization of stents further broadens the potential applications of cardiac CT. With further developments in CT instrumentation, its role in plaque evaluation may become crucial in clinical cardiology. As technical advances continue, it appears that radiation exposure may become the major drawback of cardiac CT angiography in comparison with other noninvasive modalities; therefore, further efforts at improving CT instrumentation should include a major focus on the reduction of radiation exposure.