|
| HOME | SUBSCRIPTIONS | CURRENT ISSUE | PAST ISSUES | CARDIOSOURCE | SEARCH | HELP | FEEDBACK |
|
|
|
|||||||||
|
|||||||||||
|
J Am Coll Cardiol, 2006; 48:1896-1910, doi:10.1016/j.jacc.2006.08.028
(Published online 25 September 2006). © 2006 by the American College of Cardiology Foundation |
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
,*
* Department of Radiology of the University Hospital of Caen, Caen, Normandy, France
Department of Cardiology of the University Hospital of Caen, Caen, Normandy, France
Department of Statistics of the University Hospital of Caen, Caen, Normandy, France
Hemodynamics and Cardiovascular Radiology Service, Policlinico San Donato, San Donato Milanese, Milan, Italy
|| Radiology Department, Thoraxcenter, Erasmus MC, Rotterdam, the Netherlands
¶ Cardiovascular Institute Middelheim, AZ Middelheim, Antwerp, Belgium.
Manuscript received July 5, 2006; revised manuscript received August 16, 2006, accepted August 26, 2006.
* Reprint requests and correspondence: Prof. Martial Hamon, Service des Maladies du Coeur et des Vaisseaux, Centre Hospitalier Universitaire de Caen, Avenue Côte de Nacre 14033 Caen, Normandy, France. (Email: hamon-m{at}chu-caen.fr).
 |
Abstract |
|---|
 Top Abstract MethodsResultsDiscussionConclusionsReferences |
|---|
BACKGROUND: Multislice spiral computed tomography has recently been proposed as an alternative to conventional coronary angiography (CA) for the diagnosis of CAD.
METHODS: Using Medline, we identified 29 studies (2,024 patients) evaluating CAD by means of both MSCT (
16 slices) and conventional CA before July 2006. After data extraction the analysis was performed according to a random-effects model.
RESULTS: The per-segment analysis pooled the results from 27 studies corresponding to a cumulative number of 22,798 segments. Among unassessable segments, 4.2% were excluded from the analysis and 6.4% were classified at the discretion of the investigators, underscoring the shortcomings of MSCT. With this major limitation, the per-segment sensitivity and specificity were 81% (95% confidence interval [CI] 72% to 89%) and 93% (95% CI 90% to 97%), respectively, with positive and negative likelihood ratios of 21.5 (95% CI 13.1 to 35.5) and 0.11 (95% CI 0.06 to 0.21), respectively, and positive and negative predictive values of 67.8% (95% CI 57.6% to 78.0%) and 96.5% (95% CI 94.7% to 98.3%), respectively. As expected, the per-patient analysis has shown an increased sensitivity of 96% (95% CI 94% to 98%) but a decreased specificity of 74% (95% CI 65% to 84%).
CONCLUSIONS: Multislice spiral computed tomography has shortcomings difficult to overcome in daily practice and, at the more clinically relevant per-patient analysis, continues to have moderate specificity in patients with high prevalence of CAD. Studies evaluating the diagnostic performance of the newest generation of MSCT, including patients with low to moderate CAD prevalence, will be critical in establishing the clinical role of this emerging technology as an alternative to CA.
| |||||||||||
|
Methods |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsReferences |
|---|
Study eligibility.
We included a study if: 1) it used MSCT as a diagnostic test for obstructive CAD, with >50% diameter stenosis selected as the cut-off criterion for significant CAD, using conventional invasive angiography as the reference standard; 2) it used the newest generation of MSCT (16 slices); and 3) it reported cases in absolute numbers of true positive (TP), false positive (FP), true negative (TN), and false negative (FN) results or presented sufficiently detailed data for deriving these figures. Studies were excluded if they were performed: 1) only in patients after coronary artery bypass graft surgery; 2) after percutaneous coronary intervention for long-term stent patency assessment; 3) in a subset of patients with prior heart transplant; or 4) with fewer than 30 enrolled patients.
Data extraction. The following information was extracted from each study: first author, year of publication, and journal; study population characteristics including sample size (number of subjects evaluated with both tests, number of patients excluded); number of patients with documented CAD; gender; mean age (and standard deviation); mean heart rate (and standard deviation); relative timing of the 2 imaging procedures and whether or not evaluation of one test was blind to the result of the other and to the clinical condition of the tested subject; technical characteristics of the MSCT, including type and brand of machine used; and rate of beta-blocker usage. Data were recorded separately, whenever available, at the level of segments, vessels, and subjects. Two investigators performed the data extraction independently. Discrepancies were solved by a third investigator and global consensus. The study quality conformed to the Quadas guidelines (3 ).
Data synthesis and statistical analysis. Categoric variables from individual studies are presented as n/N (%) and continuous variables are presented as median values. Measures of diagnostic accuracy are reported as point estimates (with 95% confidence intervals [CI]). The main analysis was performed at the coronary artery segment level, because most studies focused on this level of information. Secondary analyses combined the available vessel-level data, considering 4 coronary arteries per patient (left main coronary artery, left anterior descending artery, circumflex artery, and right coronary artery) and patient-level data.
By means of TP, TN, FP, and FN rates we computed sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratios (4 ). Although sensitivity and specificity are well known as measures of diagnostic accuracy, their results may be influenced by the prevalence of disease in tested subjects. The positive likelihood ratio (the ratio between sensitivity and 1 – specificity) provides an estimate of the probability of a positive test in a patient with disease, and the negative likelihood ratio (the ratio between 1 – sensitivity and specificity) gives an estimate of the probability of a negative test among diseased subjects. Both likelihood ratios are roughly independent from prevalence rates, and there is consensus that a positive likelihood ratio of >10 and a negative likelihood ratio of <0.1 provide reliable evidence of satisfactory diagnostic performance (5 ). Finally, the information from both positive and negative likelihood ratios can be combined in a single parameter, the diagnostic odds ratio, which is computed as the ratio of positive to negative likelihood ratios and provides an estimate of how much greater the odds of having the disease are for the people with a positive test result than for the people with a negative test result. Although likelihood ratios are the recommended summary statistics for systematic reviews of diagnostic studies, predictive values may also be of interest for clinicians, even if these values vary widely in their dependence on disease prevalence. Such limitations of predictive values notwithstanding, these figures were also computed and reported as exploratory data in this review.
We computed all statistics for individual studies, and then combined them using a random-effects model, weighting each point estimate by the inverse of the sum of its variance and the between-study variance. Between-study statistical heterogeneity was also assessed using the Cochran Q chi-square tests. Because diagnostic parameters are by definition interdependent, independent weighting may sometimes give spurious results and provide biased estimates. Weighted symmetric summary receiver-operating characteristic plots, with pertinent areas under the curve, were computed using the Moses-Shapiro-Littenberg method to overcome this problem of interdependence (6,7 ).
Sources of clinical and statistical heterogeneity were explored by means of subgroup analyses and meta-regression (7,8 ). Although the findings of such analyses should be regarded mainly as hypothesis generating, statistical significance may suggest substantial changes in the diagnostic performance of the test under study as the covariate increases. Specifically, we performed stratified per-segment analyses according to publication year, sample size, number of interpretable segments, and 16 versus 64 slices.
Statistical computations were performed with SPSS 11.0 (SPSS, Chicago, Illinois) and Meta-DiSc (9 ), and significance testing was at the 2-tailed 0.05 level.
|
Results |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsReferences |
|---|
16-slice MSCT; 2) they looked only at grafts or stent patency or at atherosclerotic plaque assessment; 3) they had overlapping data; 4) they were in a language other than English; 5) it was impossible to find or calculate absolute figures from presented data; or 6) no systematic angiographic control was performed. Thus, we included 29 of these studies in the systematic review (10–38
).
|
|
per-segment analysis pooled results from 27 studies (Table 2
) (1,865 patients after exclusion of 32 patients because of unsuccessful MSCT, corresponding to 22,798 segments after exclusion of 4.2% of segments) and showed that, compared with invasive CA, MSCT had a sensitivity of 81% (95% CI 72% to 89%), a specificity of 93% (95% CI 90% to 97%), a positive likelihood ratio of 21.5 (95% CI 13.1 to 35.5), a negative likelihood ratio of 0.11 (95% CI 0.06 to 0.21), and a diagnostic odds ratio of 189.3 (95% CI 93.5 to 383.4). Overall, these summary estimates confirm the superior (93%) specificity of MSCT in patients undergoing work-up for suspected CAD, even if this occurs at the price of a moderate (81%) value for sensitivity. Indeed, the high specificity of MSCT translates into a positive predictive value (PPV) of 67.8% (95% CI 57.6% to 78.0%) and a negative predictive value (NPV) of 96.5% (95% CI 94.7% to 98.3%), assuming CAD prevalence similar to that found in the studies included in this review (median value of 63.5% and range of 8% to 100%).
|
|
|
|
|
|
Per-vessel meta-analysis. Per-vessel results were reported in, and thus pooled from, only 8 studies, shown in Table 2 (2,726 coronary arteries); therefore, meta-analysis should be viewed in light of the risk of small-study and publication bias. Per-vessel analysis provided the following results: 82% (95% CI 80% to 85%) sensitivity, 91% (95% CI 90% to 92%) specificity, 11.8 (95% CI 6.7 to 20.6) positive likelihood ratios, 0.08 (95% CI 0.02 to 0.32) negative likelihood ratio, and 146.5 (95% CI 31.9 to 671.2) diagnostic odds ratio, with 0.97 symmetric area under the curve. Pooled PPV for per-vessel analysis was 81% (95% CI 78% to 83%), whereas NPV was 92% (95% CI 91% to 93%). Heterogeneity was significant (p < 0.001) for all of the aforementioned diagnostic performance measures.
Per-patient meta-analysis.
Table 2
shows that somewhat incomplete reporting was available for per-patient results from 22 studies (1,616 patients), so the risk of small-study and publication bias should not be dismissed. The study of Probst et al. (22
) was excluded from the pooled analysis because a 0 value for FP, FN, and TN meant that the variance is infinite and confidence intervals could not be computed. As shown in Figures 7 to 11
, pooled per-patient analysis—including 21 studies, corresponding to 1,570 patients—provided the following results: 96% (95% CI 94% to 98%) sensitivity, 74% (95% CI 65% to 84%) specificity, 5.4 (95% CI 3.4 to 8.3) positive likelihood ratio, 0.05 (95% CI 0.03 to 0.09) negative likelihood ratio, and 133.05 (95% CI 57.3 to 308.9) diagnostic odds ratio. Pooled PPV for per-patient analysis was 83% (95% CI 76% to 90%) and pooled NPV was 94% (95% CI 89% to 99%). Heterogeneity was again significant (p < 0.001) for all of the aforementioned diagnostic performance measures.
|
|
|
|
|
|
|
|
Finally, we performed meta-regression analyses exploring the impact of sample size and publication year on the diagnostic performance of MSCT. Whereas the latter did not disclose significant results, we found a significant interaction between changes in sample size and diagnostic odds ratios in the individual studies (p < 0.05), suggesting that smaller studies were more likely to provide higher diagnostic odds ratios. Although it is possible that these results were due to differences in design, setting, and technique, they may also be indirect evidence of small study bias.
|
Discussion |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsReferences |
|---|
16 slices) to assess the extent and severity of coronary stenoses compared with conventional CA. The main analysis was performed at the segment level, because most studies focused on that level of information. Our meta-analysis confirms the relatively high (93%) specificity of MSCT, which occurs at the price of a moderate (81%) value for sensitivity. This high specificity of MSCT-CA may be translated in the future into a clinically useful tool in assessable coronary segments; a high NPV of 96.5% would suggest exclusion of CAD in selected subjects considered for conventional CA. However, these optimistic figures were provided in highly selected patients and after exclusion from the analysis of 4.2% of scanned segments, with an additional 6.4% of unassessable segments classified at the discretion of the investigators, and favoring the positive perception of MSCT as a valuable diagnostic tool. In addition to post hoc exclusion of some segments showing poor image quality, such as calcium-blooming artifacts, the great majority of studies pre-specified post-acquisition exclusion criteria based on reference vessel size. Keeping in mind this major restriction, the present meta-analysis demonstrates that MSCT-CA provides, for the per-segment analysis, across a wide range of diverse clinical centers, consistent diagnostic performance in a highly selected study population. In our secondary analyses, the results are largely consistent and show the anticipated increase in sensitivity and decrease in specificity as one moves from segments to vessels to subjects. This consistent trend in the diagnostic value is also observed with the NPV of 96.5% at segment level dropping to 94% at patient level and the diagnostic odds ratio of 189.3 (95% CI 93.5 to 383.4) at segment level dropping to 133.05 (95% CI 57.3 to 308.9) at patient level. This finding is likely to reflect the fact that studies on MSCT have primarily focused on patients with a high prevalence of CAD, in whom an increase in false-positive observations (related mainly to heavy calcification) can be anticipated, whereas true negative findings are expected to remain low and essentially unaffected as the analysis moves from per-segment to per-patient. However, this major drawback in reporting data in manuscripts and reporting some unassessable segments as negative findings favors the positive perception of MSCT when reading reports and in the present meta-analysis. In real life and in daily practice, the accuracy of MSCT is certainly decreased by the fact that uninterpretable segments must be considered as positive to avoid misdiagnosis for individuals, especially when these unassessable segments are proximal and masked by heavy calcifications. It is noteworthy that the only multicenter study following this rule in the present meta-analysis found the lowest per-segment specificity, 65%, leading to a specificity of only 55% at patient level (27 ).
The trend toward increasing sensitivity and specificity in moving from 16-slice to 64-slice CT supports active research in the new generation of CT scanners. From this perspective, it is noteworthy that some studies (32,33 ), based on the newest generation of 64-slice CT, reported results from almost all evaluated segments, irrespective of vessel size and image quality.
Indeed, cardiac imaging using CT is a technically demanding task. Not only is high spatial resolution required for imaging small structures such as the coronary arteries, but high temporal resolution must also be achieved for motion-free imaging of the heart, given heart rates that may range from 50 to more than 100 beats/min (39 ). To overcome and minimize the effects of this issue, most investigators use beta-blockers, as found in our systematic review and as previously suggested by some reports, which demonstrated a variation in diagnostic performance according to the patient’s heart rate (40 ). Respiratory motion must also be eliminated for cardiac imaging, so scanning must be performed in a single breath hold. All of these technical issues are critical for optimal acquisition, and even with the remarkable progress of successive generations of scanners providing faster imaging at progressively higher spatial and temporal resolutions, only highly selected patients can be explored by MSCT for CAD assessment.
Clinical implications and cautionary notes. It should be emphasized that the observed per-segment high specificity has been obtained in patients selected to undergo CA and who are, therefore, presenting with reasonably high probability of CAD. The median prevalence of CAD was as high as 63.5% (95% CI 8% to 100%) among the included studies. Whether the performance of MSCT in terms of negative predictive value can be reproduced in patients at lower prevalence of CAD remains to be assessed. Moreover, patients with a high probability of CAD were selected based on several parameters, including regular and controlled heart rate, renal function, breath-hold capacity, hemodynamic status, and, often, previous coronary instrumentation. This highlights the limited external validity of the present findings and calls for studies evaluating the diagnostic performance of MSCT in a less highly selected patient population before its application in the clinical setting can be suggested as an alternative to CA.
The relatively high radiation dose received with MSCT, compared with CA, should also be acknowledged. The effective radiation dose varies based on the presence of the tube current modulation, ranging from 5.4 to 16.3 mSv for 16-slice CT and from 10 to 21.4 mSv for 64-slice CT in the papers where this information was provided. The effective radiation dose for invasive CA is known to be in the range of 2 to 5 mSv (41,42 ).
It is also important to acknowledge that if the MSCT detection of coronary calcification results frequently in a false positive classification when CA fails to identify significant luminal narrowing in that segment, from a clinical point of view this finding adds valuable prognostic information. Indeed the presence of coronary calcification has been associated with future coronary events and can justify aggressive risk factor reduction therapy (43–45 ).
Study limitations. As mentioned in the preceding, substantial statistical heterogeneity has been documented, casting a shadow of caution on the results and interpretation of these estimates of comprehensive, pooled effects. The well known tendency toward publication bias favoring studies with positive and encouraging results also complicates comprehensive evaluation. In the present meta-analysis, data abstraction and quality assessment were done by independent reviewers and, in the case of any divergences, resolution was made by consensus. Therefore, the interoperator agreement could not be quantitatively assessed. We should also acknowledge that not all reports provided complete data concerning subject and vessel levels. More rigorous reporting of future clinical research on coronary artery imaging technologies should be encouraged.
|
Conclusions |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsReferences |
|---|
|
Acknowledgments |
|---|
|
References |
|---|
|
TopAbstractMethodsResultsDiscussionConclusionsReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  E. J. Heffernan, J. D. Dodd, and D. E. Malone Cardiac Multidetector CT: Technical and Diagnostic Evaluation with Evidence-based Practice Techniques Radiology, August 1, 2008; 248(2): 366 - 377. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
W. P. Shuman, K. R. Branch, J. M. May, L. M. Mitsumori, D. W. Lockhart, T. J. Dubinsky, B. H. Warren, and J. H. Caldwell Prospective versus Retrospective ECG Gating for 64-Detector CT of the Coronary Arteries: Comparison of Image Quality and Patient Radiation Dose Radiology, August 1, 2008; 248(2): 431 - 437. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Cheng, A. Gutstein, A. Wolak, Y. Suzuki, D. Dey, H. Gransar, L. E.J. Thomson, S. W. Hayes, J. D. Friedman, and D. S. Berman Moving Beyond Binary Grading of Coronary Arterial Stenoses on Coronary Computed Tomographic Angiography: Insights for the Imager and Referring Clinician J. Am. Coll. Cardiol. Img., July 1, 2008; 1(4): 460 - 471. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W T Roberts, J J Bax, and L C Davies Cardiac CT and CT coronary angiography: technology and application Heart, June 1, 2008; 94(6): 781 - 792. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Van Brabandt Recommendations or mere prose? Eur. Heart J., June 1, 2008; 29(11): 1473 - 1473. [Full Text] [PDF]
|
|
|
|
|
|
M. Hamon, O. Lepage, P. Malagutti, J. W. Riddell, R. Morello, D. Agostini, and M. Hamon Diagnostic Performance of 16- and 64-Section Spiral CT for Coronary Artery Bypass Graft Assessment: Meta-Analysis Radiology, June 1, 2008; 247(3): 679 - 686. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. J Schunemann, A. D Oxman, J. Brozek, P. Glasziou, R. Jaeschke, G. E Vist, J. W Williams Jr, R. Kunz, J. Craig, V. M Montori, et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies BMJ, May 17, 2008; 336(7653): 1106 - 1110. [Full Text] [PDF]
|
|
|
|
 |
|
 I. Gottlieb and J. A.C. Lima Screening High-Risk Patients With Computed Tomography Angiography Circulation, March 11, 2008; 117(10): 1318 - 1332. [Full Text] [PDF]
|
|
|
|
|
|
S. Achenbach, U. Ropers, A. Kuettner, K. Anders, T. Pflederer, S. Komatsu, W. Bautz, W. G. Daniel, and D. Ropers Randomized Comparison of 64-Slice Single- and Dual-Source Computed Tomography Coronary Angiography for the Detection of Coronary Artery Disease J. Am. Coll. Cardiol. Img., March 1, 2008; 1(2): 177 - 186. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. A.C. Lima Another Step Forward in CT Angiography J. Am. Coll. Cardiol. Img., March 1, 2008; 1(2): 187 - 189. [Full Text] [PDF]
|
|
|
|
|
|
S. Schroeder, S. Achenbach, F. Bengel, C. Burgstahler, F. Cademartiri, P. de Feyter, R. George, P. Kaufmann, A. F. Kopp, J. Knuuti, et al. Cardiac computed tomography: indications, applications, limitations, and training requirements: Report of a Writing Group deployed by the Working Group Nuclear Cardiology and Cardiac CT of the European Society of Cardiology and the European Council of Nuclear Cardiology Eur. Heart J., February 2, 2008; 29(4): 531 - 556. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
L. F. Tops, S. C. Krishnan, J. D. Schuijf, M. J. Schalij, and J. J. Bax Noncoronary Applications of Cardiac Multidetector Row Computed Tomography J. Am. Coll. Cardiol. Img., January 1, 2008; 1(1): 94 - 106. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Dewey, B. Hamm, H-P. Dubel, and W. Rutsch Multislice computed tomography coronary angiography for triaging patients with positive radionuclide myocardial perfusion imaging QJM, January 1, 2008; 101(1): 65 - 67. [Full Text] [PDF]
|
|
|
|
 |
|
 I. Decramer, P. K. Vanhoenacker, G. Sarno, L. Van Hoe, O. Bladt, W. Wijns, and P. M. Parizel Effects of Sublingual Nitroglycerin on Coronary Lumen Diameter and Number of Visualized Septal Branches on 64-MDCT Angiography Am. J. Roentgenol., January 1, 2008; 190(1): 219 - 225. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G Pundziute, J D Schuijf, J W Jukema, J M van Werkhoven, E Boersma, A de Roos, E E van der Wall, and J J Bax Gender influence on the diagnostic accuracy of 64-slice multislice computed tomography coronary angiography for detection of obstructive coronary artery disease Heart, January 1, 2008; 94(1): 48 - 52. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Abdulla, S. Z. Abildstrom, O. Gotzsche, E. Christensen, L. Kober, and C. Torp-Pedersen 64-multislice detector computed tomography coronary angiography as potential alternative to conventional coronary angiography: a systematic review and meta-analysis Eur. Heart J., December 2, 2007; 28(24): 3042 - 3050. [Abstract] [Full Text] [PDF]
|
|
