- Email: [email protected]
Semiquantitative dipyridamole myocardial stress perfusion imaging for cardiac risk assessment before noncardiac vascular surgery: A metaanalysis
Semiquantitative dipyridamole myocardial stress perfusion imaging for cardiac risk assessment before noncardiac vascular surgery: A metaanalysis Edward Etchells, MD, MSc,a,c Maureen Meade, MD, MSc (Epid),e George Tomlinson, PhD,b,d and Deborah Cook, MD, MSc (Epid),e Toronto and Hamilton, Ontario, Canada Background: Semiquantitative dipyridamole myocardial perfusion scintigraphy may provide better estimates of perioperative cardiac risk than nonquantitative scintigraphy. Objective: The purpose of this study was to conduct a metaanalysis of semiquantitive dipyridamole myocardial perfusion scintigraphy for the prediction of perioperative myocardial infarction and cardiac death in patients undergoing noncardiac vascular surgery. Methods: The data sources used were MEDLINE (from 1975 to 1999), citation lists, and correspondence with study authors. We included studies that evaluated preoperative semiquantitative dipyridamole myocardial perfusion scintigraphy in patients undergoing noncardiac vascular surgery. For each study, we calculated results on the basis of the proportion of myocardial segments with reversible perfusion defects. The complications of interest were cardiac death and nonfatal myocardial infarction. We calculated likelihood ratios (LRs) and areas under the receiver operating characteristic curves for individual studies and for the combined data. Data synthesis: We identified nine studies involving a total of 1179 patients with 82 cardiac complications (complication rate, 7.0%). Most studies were grade C in quality. Normal scans significantly reduced the likelihood of perioperative cardiac complications (LR, 0.42; 95% CI, 0.20 to 0.88). Fixed defects reduced the likelihood of complications, but the effect was not statistically significant (LR, 0.51; 95% CI, 0.24 to 1.1). Reversibility in less than 20% of myocardial segments did not change the likelihood of perioperative cardiac complications (LR, 1.3, 95% CI, 0.88 to 1.9). LRs for increasing extents of reversibility were: 20% to 29% reversibility (LR, 1.6; 95% CI, 1.0 to 2.6), 30% to 39% reversibility (LR, 2.9; 95% CI, 1.6 to 5.1), 40% to 49% reversibility (LR, 2.9; 95% CI, 1.4 to 6.2), and 50% or more reversibility (LR, 11; 95% CI, 5.8 to 20). The pooled area under the receiver operating characteristic curve was 0.78 (95% CI, 0.65 to 0.89). Heterogeneity was found among study results, but exclusion of heterogeneous studies did not significantly change the summary results. Conclusion: Reversible defects in less than 20% of myocardial segments do not significantly alter the risk of perioperative cardiac complications. Greater extents of reversibility on dipyridamole myocardial stress perfusion imaging increase the risk of perioperative complications after noncardiac vascular surgery, but the quality and amount of data regarding greater extents of reversibility are limited. (J Vasc Surg 2002;36:534-40.)
Perioperative myocardial infarction occurs in 2% to 15% of patients undergoing vascular surgery.1 One quarter to one third of perioperative myocardial infarctions are fatal.2,3 Accurate preoperative assessment of cardiac risk helps patients and physicians make informed decisions about surgery, helps physicians choose appropriate methods for perioperative care, and identifies patients who need more intensive perioperative monitoring.
From the Divisions of General Internal Medicine at Sunnybrook and Women’s College Health Sciences Centre,a and University Health Network,b Department of Medicine,c and the Department of Public Health Sciences,d University of Toronto; and the Departments of Medicine and Clinical Epidemiology and Biostatistics, McMaster University.e Competition of interest: nil. Correspondence: Dr Edward E. Etchells, Department of Medicine, Sunnybrook and Women’s College Health Sciences Centre, 2075 Bayview Ave, Rm C 4-10, Toronto ON M4N 3M5 Canada (e-mail: edward.etchells@ swchsc.on.ca). Copyright © 2002 by The Society for Vascular Surgery and The American Association for Vascular Surgery. 0741-5214/2002/$35.00 ⫹ 0 24/1/126563 doi:10.1067/mva.2002.126563
534
Practice guidelines for assessment of perioperative cardiac risk recommend a clinical assessment followed by selective use of noninvasive cardiac testing.4,5 The most extensively studied method of noninvasive cardiac testing is dipyridamole myocardial stress perfusion imaging. If a reversible myocardial perfusion defect is observed during dipyridamole imaging, then the patient is identified as high risk for perioperative myocardial infarction and cardiac death. This approach is based on the interpretation of myocardial stress perfusion imaging as showing either any reversibility or no reversibility of myocardial perfusion defects. However, a semiquantitative description of the extent of reversibility may result in better prediction of perioperative cardiac complications because the extent of reversibility is a powerful predictor of the extent of coronary artery disease on coronary angiography in patients with suspected coronary artery disease.6 The American College of Cardiology/American Heart Association guideline states that “the use of techniques to quantitate the extent of abnormality will probably improve the positive predictive nature of myocardial perfusion imaging.”4 The American College
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
of Physicians guideline states that “increased risk discrimination is obtained from semiquantitative interpretation.”5 Two published metaanalyses of dipyridamole myocardial stress perfusion imaging before vascular surgery have been published. One study did not address semiquantitative dipyridamole myocardial stress perfusion imaging.7 The other combined three studies8-10 (total of 433 patients) and found that patients with two or more reversible defects were more likely to have cardiac complications (including congestive heart failure and transient myocardial ischemia) than patients with one or more reversible defects.11 This analysis did not include two early studies that evaluated 305 patients and report semiquantitative analysis.12,13 Several additional studies of semiquantative dipyridamole myocardial stress perfusion imaging have been published since 1994. Our objective was to conduct a systematic review and metaanalysis of evidence about semiquantitative dipyridamole-thallium scintigraphy for prediction of nonfatal myocardial infarction and cardiac death after noncardiac vascular surgery. METHODS Assembly of studies We searched for published and unpublished data with computerized bibliographies, hand searching of relevant journals, and correspondence with study authors. In our MEDLINE search, we used MeSH headings dipyridamole (exploded) or thallium (exploded), limited to human and journal article, for the period from 1975 to January 1999. We also searched with MeSH heading sestamibi (exploded), but no additional articles were identified, so this search is not reported in detail. We considered reports in all languages. All article titles, abstracts, and subject headings were screened for potential relevance. We looked for evidence of preoperative dipyridamole myocardial stress perfusion imaging (including thallium-201, technetium-99m sestamibi, and rubidium-82, and planar, single photon emission computer tomography, and positron emission tomography) in the setting of noncardiac vascular surgery, including carotid endarterectomy, aortic procedures, renal artery bypass or repair, lower extremity arterial bypass or thromboendarterectomy, and diabetic amputations. We excluded articles indexed as a review article or a case report. We also excluded studies of renal transplantation and vascular access for hemodialysis. The first 300 citations were screened by two independent reviewers with excellent agreement regarding potential relevance (simple agreement, 99%; , 0.92), so the remaining citations were screened by a single reviewer. We also searched the citation lists of all selected studies and published guidelines to identify other potentially relevant studies. Study selection The full reports of all potentially relevant articles and abstracts were reviewed in detail for three additional selection criteria: 1, evidence of semiquantitative interpretation of scintigraphy results; 2, the reporting of perioperative
Etchells et al 535
myocardial infarction and cardiac death as study outcomes; and 3, follow-up for at least 1 month or for the duration of hospitalization. We did not consider unstable angina because it is not consistently reported, and we did not consider arrhythmias, pulmonary edema, and all-cause mortality because these are not as strongly related to reversible cardiac ischemia. When the same data were reported in more than one article, only the article with the most complete data set was included. A subset of potentially relevant articles was reviewed by two independent reviewers with excellent agreement regarding selection (simple agreement, 96%; , 0.90). Authors of studies that met most or all selection criteria were contacted by mail with a request to provide methodologic information and additional results. Evaluation of study methods (Table I) Two independent reviewers evaluated study methods according to critical appraisal principles of diagnostic test research.14 We resolved discrepancies with repeated review and discussion. We abstracted data on the prospective enrollment of consecutive participants, with blinded interpretation of scintigraphy and blinded assessment of perioperative cardiac complications. We assigned a quality grade on the basis of these features as previously described.15 Grade A studies involved the independent comparison of myocardial perfusion scintigraphy results with development of perioperative cardiac complications among at least 100 consecutive patients. Grade B studies met the criteria for grade A studies but had less than 100 patients. Grade C studies involved nonconsecutive patients or nonindependent comparisons between the scintigraphy and perioperative cardiac complications. Several features of perioperative cardiac studies were considered in our appraisals. First, routine use of dipyridamole myocardial stress perfusion imaging for all patients before noncardiac vascular surgery is not recommended.16,17 Accordingly, we determined whether study patients were selected on the basis of clinical characteristics indicating increased cardiac risk. Second, patients with reversible defects may have their surgery canceled or they may undergo coronary revascularization before noncardiac vascular surgery. In either situation, these patients are less likely to have a perioperative cardiac complication, thereby reducing the observed predictive value of a reversible defect. We determined whether dipyridamole myocardial stress perfusion imaging results led to cancellation of surgery or to preoperative coronary revascularization, and we abstracted data on the clinical and scintigraphic characteristics of such patients. Third, patients with reversible defect scans who undergo noncardiac vascular surgery may receive more intense postoperative electrocardiographic and cardiac enzyme monitoring. This surveillance bias can result in more diagnoses of postoperative myocardial infarction in patients with reversible defects, leading to spuriously high estimates of sensitivity. Therefore, we assessed the degree of blinding of caregivers (anesthetists, internists, and surgeons) to di-
JOURNAL OF VASCULAR SURGERY September 2002
536 Etchells et al
Table I. Summary of studies included in metaanalysis MI/ Sample cardiac Study size deaths 8
237
17 (7%)
Avoided Patient MI Quality Blinding: Blinding: surgical Routine characteristics definition Scintigraphy grade Prospective Consecutive test outcome selection surveillance Diabetes 34% Angina 18% MI 30%
CK alone Thallium or with Planar and EKG SPECT Reinjection 9 101 10 (10%) Diabetes 100% EKG and Thallium Angina 46% CK Planar MI 39% 12 125 13 (10%) Diabetes 24% Any two: Thallium Angina 46% EKG, Planar CK, or pain 13 180 12 (7%) Diabetes 18% EKG and Thallium Angina 29% CK Planar MI 23% 26 72 5 (7%) Diabetes 26% CK with Rubidium Angina 31% pain or 82 PET MI 26% EKG 27 134* 12 (9%) Diabetes 13% EKG and Thallium CK SPECT Angina 23% Reinjection MI 29% 28 197 4 (2%) Diabetes 25% EKG and Sestamibi Angina 60% CK SPECT MI 23% 29 104 5 (5%) Diabetes 26% EKG and Thallium Angina 11% CK SPECT MI 7% 30 29 4 (14%) Diabetes 3% EKG and Sestamibi Angina 14% CK Planar Total 1179 82 (7.0%)
C
No
No
Yes
No
No
No
C
Yes
No
Yes
No
Yes
No
C
No
Yes
Yes
No
No
No
C
No
No
Yes
Yes
No
No
C
No
Yes
Yes
No
No
No
A
Yes
Yes
Yes
Yes
Yes
Yes†
C
No
Yes
Yes
No
No
No
C
No
Yes
Yes
Yes
No
No
B
Yes
Yes
Yes
Yes
Yes
Yes‡
*This study used explicit selection criteria (any two of: age ⬎70 years, history of myocardial infarction, history of congestive heart failure, history of angina, diabetes mellitus, hypertension with left ventricular hypertrophy, Q waves, or ischemic ST segment changes on baseline electrocardiogram). Of 514 patients considered for surgery, 134 had at least two of these risk factors, 23 surgeries were canceled for noncardiac reasons, 37 were canceled because of high cardiac risk, and 317 had less than two risk factors and underwent surgery without scanning. Cardiac events were observed in 11 of 314 patients (3.5%) with less than two risk factors who did not have scintigraphy. † Routine surveillance: electrocardiogram and cardiac enzymes every 24 hours for 2 days after surgery. ‡ Routine surveillance: electrocardiogram twice daily in intensive care unit (at least 24 hours after surgery) and cardiac enzymes twice daily for 3 days after surgery. MI definition: This column describes whether electrocardiographic changes, cardiac enzyme changes, or both are necessary for diagnosis of perioperative myocardial infarction. Some studies also included clinical criterion of prolonged typical ischemic chest pain. Blinding: test: Dipyridamole myocardial perfusion scintigraphy was interpreted without knowledge of patient’s perioperative outcome. Blinding: outcome: Assessment of cardiac complication was made without knowledge of patient’s preoperative dipyridamole myocardial perfusion scintigraphy. Avoided surgical selection: Surgery was not canceled and patients did not undergo revascularization before noncardiac surgery because of reversible defects noted on dipyridamole myocardial perfusion scintigraphy. If study did not avoid surgical selection, study would be less likely to show relationship between reversible defects and perioperative cardiac complications. Routine surveillance: Standard protocol was used for monitoring all patients after surgery, including electrocardiogram and cardiac enzyme measurement, regardless of preoperative scan result. MI, Myocardial infarction; CK, cardiac enzyme; EKG, electrocardiogram; SPECT, single photon emission computed tomography; PET, position emission tomography.
pyridamole myocardial stress perfusion imaging results and determined whether systematic protocols for postoperative electrocardiographic and cardiac enzyme monitoring were used. Finally, patients with reversible defects may receive additional perioperative treatments to reduce the likelihood of perioperative cardiac complications. If caregivers were not blinded to scan results, we abstracted information on the perioperative care for patients with and without reversible defects.
Data abstraction Two independent reviewers abstracted the data. We resolved discrepancies with repeated review and discussion. Studies quantified the extent of reversibility with division of the myocardial images into six to 24 segments and reporting of the number of segments with reversibility. To standardize results across studies, we calculated the percentage of myocardial segments showing reversible defects. For example, if the study used a six-segment
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
model, then reversibility in one segment would correspond to 17% of myocardium showing reversibility. We recorded study definitions of myocardial infarction, including electrocardiographic and cardiac enzyme criteria. Troponin I was not used in any studies. We did not record data on other cardiac complications, such as unstable angina, pulmonary edema, or arrhythmias. For each study, we identified patients who had coronary revascularization after scanning but before noncardiac vascular surgery. Whenever possible, we excluded these patients from our calculations of diagnostic accuracy. Inclusion of these patients would spuriously lower the specificity of an abnormal dipyridamole myocardial stress perfusion imaging result because perioperative cardiac complications are rare after successful myocardial revascularization. Data analysis Individual studies. We calculated nonparametric areas under the receiver operating characteristic (ROC) curve for each study with the method of Hanley and McNeil18 and compared the areas under each curve with z tests. We also calculated binormal estimates for the areas under the ROC curve, with ROCFIT software19 and a maximum likelihood method.20,21 A perfect test has an area under the ROC curve of 1. An area under the curve of 0.5 indicates a useless test that performs no better than a coin flip. Combining studies. To determine which studies could be meaningfully combined, we applied standard principles of metaanalysis.22 First, we tried to assemble studies with reasonably similar features, including patient population, type of surgery, type of imaging, and definition of postoperative myocardial infarction. Second, we assessed the methodologic quality of each study. Third, we determined similarity of results amongst the studies by comparing the areas under the ROC curves. With the BUGS23 software, we fitted a Bayesian hierarchic model24 to combine the binormal ROC curve estimates from individual studies. We then estimated summary likelihood ratios (LRs) for increasing extents of reversibility with the Bayesian hierarchic model. In advance, we defined seven scan results, including normal scans, fixed defects only, reversible defects in less than 20%, and reversible defects in 20% to 29%, 30% to 39%, 40% to 49%, and more than 50% of myocardial segments. The major analytic issue in estimating summary LRs was that individual studies used different intervals for reporting extents of reversibility. To calculate the desired LRs, we interpolated the results from each study to conform to the scan results listed previously. These interpolations had little effect on the shape or area of the original or summary ROC curves. Once the data had been interpolated to the intervals listed previously, we calculated stratum-specific LRs for each study, adding 0.5 to each cell to give the estimate better properties in the presence of small counts. We used the DerSimonian and Laird25 random effects method to obtain a summary LR for each stratum. The summary LRs
Etchells et al 537
were used to convert pretest probabilities of myocardial infarction to posttest probabilities for test outcomes at each level of reversibility. A LR of less than 1 reduces the likelihood of complications, and a LR of less than 0.4 means that the likelihood of complications is substantially reduced. A LR of greater than 1 increases the likelihood of complications, and a LR of 4 means that the likelihood of complications is substantially increased. RESULTS We screened 3160 MEDLINE citations, identified 88 potentially relevant articles, and selected 16 for inclusion in this metaanalysis. The 72 other potentially relevant articles were excluded for the following reasons: no semiquantitative analysis (n ⫽ 28), nonvascular surgery (n ⫽ 10), review article (n ⫽ 7), repeat publication (n ⫽ 7), no myocardial stress perfusion imaging (n ⫽ 6), no surgery (n ⫽ 4), heterogeneous outcomes (n ⫽ 3), sample size less than 10 patients (n ⫽ 2), and other reasons (n ⫽ 5).(A list of the excluded articles is available from the corresponding author). Of the 16 selected articles, we were able to obtain sufficient data to analyze nine studies.8,9,12,13,26-30 The remaining seven selected articles did not have enough information to be included in the metaanalysis, and we were unable to obtain additional information from the authors.17,31-36 Methodologic features of primary studies (Table I). One selected study had a quality grade A,27 another grade B,30 and the rest were grade C.8,9,12,13,26,28,29 All studies selected patients for myocardial stress perfusion imaging on the basis of a clinical suspicion of increased cardiac risk, but only one study27 explicitly described the selection criteria. Six studies were partly confounded by cancellation of surgery or revascularization because of abnormal scan results.8,12,13,26,28,29 In most cases, revascularization or cancellation occurred in patients with greater extents of reversibility. Caregivers (anesthetists, surgeons, and internists) were blinded to scintigraphy results in only two studies.27,30 The remaining studies provided little detail regarding additional treatments for patients with reversible defects. One study stated that patients with reversible defects tended to receive more invasive hemodynamic monitoring and intravenous nitrates,13 and another stated that optimization of antiischemic therapy and intraoperative hemodynamic monitoring was routinely recommended in patients with reversible defects.28 All studies used a similar protocol for infusing dipyridamole. The radiopharmaceutical was thallium-201 in six studies, 8,9,12,13,27,29 technetium-99m sestamibi in two studies,28,30 and rubidium-82 in one study.26 Five studies used single photon emission computed tomography,8,27,28,29 one used positron emission tomography,26 and three used planar images.9,12,13,30 The method of describing extent of reversibility was variable, with a range of five to 24 segments. Individual study results (Table II). The areas under the ROC curves were similar for the nonparametric and
JOURNAL OF VASCULAR SURGERY September 2002
538 Etchells et al
Table II. Areas under ROC curves for each study Study
Area under ROC curve (95% CI)
8* 9* 12 13 26† 27 28† 29 30* Combined Combined, excluding studies 26 and 28
0.76 (0.65 - 0.86) 0.82 (0.69 - 0.95) 0.96 (0.91 - 1.00) 0.86 (0.71 - 1.00) 0.46 (0.13 - 0.79) 0.84 (0.73 - 0.95) 0.47 (0.16 - 0.78) 0.88 (0.78 - 0.97) 0.73 (0.53 - 0.92) 0.78 (0.65 - 0.89) 0.82 (0.69 - 0.92)
*Differs from study 12, P ⬍ .05. † Differs from studies 9, 12, 13, 21, 23, P ⬍ .05.
binormal methods, so we report the nonparametric areas only. Two studies had significantly lower areas under the ROC curves than the remaining studies.26,28 Each study26,28 had a high rate of cancellation for patients with reversible defects (approximately 20% were canceled). One of these studies included results for seven patients with reversible defects who had myocardial revascularization before noncardiac vascular surgery.28 We were unable to obtain results of this study with these seven patients excluded. The individual and summary ROC curves can be viewed in the Fig (online only). The extent to which these outlying studies are “biased” estimates of the accuracy of dipyridamole myocardial stress perfusion imaging is impossible to determine. Therefore, we performed analyses with and without these two studies. Combined study results (Table III). The pooled area under the summary ROC curve was 0.78 (95% CI, 0.65 to 0.89), with the Bayesian hierarchic model to combine the binormal ROC curve area estimates. When the two outlying studies were excluded, the pooled area under the ROC curve was 0.82 (95% CI, 0.69 to 0.92). We obtained similar results when we combined the nonparametric ROC curve area estimates with the random effects model. Normal scans significantly reduced the likelihood of perioperative cardiac complications (summary LR, 0.42; 95% CI, 0.20 to 0.88). Fixed defects reduced the likelihood of complications, but the effect was not statistically significant (summary LR, 0.51; 95% CI, 0.24 to 1.1). Reversibility in less than 20% of myocardial segments did not change the likelihood of perioperative cardiac complications (LR, 1.3; 95% CI, 0.88 to 1.9). LRs for increasing extents of reversibility were: 20% to 29% reversibility (LR, 1.6; 95% CI, 1.0 to 2.6), 30% to 39% reversibility (LR, 2.9; 95% CI, 1.6 to 5.1), 40% to 49% reversibility (LR, 2.9; 95% CI, 1.4 to 6.2), and 50% or more reversibility (LR, 11; 95% CI, 5.8 to 20). When we excluded the two outlying studies, there were no statistically significant changes in the results. The only qualitative difference was that the LR for a normal scan was lower (0.18; 95% CI, 0.07 to 0.49). Table III shows the posttest probability for various scan results, assuming a pretest probability of cardiac complica-
tions of 7%. Table III also shows the distribution of various scan results: 60% of scans were normal or showed fixed defects, 17% of scans showed reversibility in less than 20% of myocardial segments, and 23% of scans showed reversibility in 20% or more of myocardial segments. Potentially relevant studies that could not be analyzed. Seven potentially relevant studies were not included in this analysis because of inadequate information regarding semiquantitative analysis. Four studies17,32,34,36 (n ⫽ 823, including 73 cardiac events) had results that would strengthen our conclusions. Two studies31,33 (n ⫽ 324, including 28 cardiac events) had results of uncertain effect on our conclusions. One study35 (n ⫽ 174, including four cardiac events) had results that would potentially weaken our conclusions. These studies are summarized in the Appendix (online only). DISCUSSION In this metaanalysis, our major finding is that patients undergoing preoperative semiquantitative dipyridamole myocardial perfusion scintigraphy who had reversible defects in up to 20% of myocardial segments did not have an increased risk of perioperative cardiac complications. We also found that patients with greater extents of reversible defects were at increasing risk of cardiac complications after noncardiac vascular surgery. Two major methodologic issues may affect our conclusions: the methodologic quality of the primary studies and publication bias. We believe that our major finding is unaffected by these methodologic issues, but it is possible that the predictive value of greater extents of reversibility has been overestimated. Most studies in this analysis had suboptimal methodologic quality. Ideally, we would like to base our recommendations on higher quality primary studies, but no such studies exist. Instead, we must examine how the methods of the individual studies may have affected the results of this analysis. First, caregivers were blinded to the results of myocardial stress perfusion imaging in only two studies. It is likely that additional treatments were given to patients with reversible perfusion defects, leading to fewer cardiac complications. However, the lack of blinding reflects what occurs in usual practice, so we conclude that patients with reversibility in less than 20% of myocardial segments, who receive appropriate perioperative care, have an average risk of perioperative cardiac complications. Second, unequal surveillance for postoperative myocardial infarction and unblinded interpretation of study outcomes would tend to overestimate the predictive value of myocardial stress perfusion imaging. However, our major finding is not attenuated by these methodologic limitations because we conclude that lesser extents of reversibility have limited predictive value. It is possible that the predictive value of greater extents of reversibility has been overestimated in our analysis. We made efforts to minimize the possibility of publication bias by contacting authors for additional unpublished
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
Etchells et al 539
Table III. Summary likelihood ratios and estimated posttest probability of perioperative cardiac complications for each scan result Extent of reversibility No defects Fixed defects only Reversibility ⬍20% Reversibility 20%-29% Reversibility 30%-39% Reversibility 40%-49% Reversibility ⱖ50%
Likelihood ratio (all studies; 95% CI)
Likelihood ratio (excluding studies 26 and 28; 95% CI)
Posttest probability of MI or cardiac death
Percentage of scans with this result
0.42 (0.20 - 0.88) 0.51 (0.24 - 1.1) 1.3 (0.88 - 1.9) 1.6 (1.0 - 2.6) 2.9 (1.6 - 5.1) 2.9 (1.4 - 6.2) 11 (5.8 - 20)
0.18 (0.07 - 0.49) 0.46 (0.17 - 1.2) 1.3 (0.86 - 2.0) 1.6 (0.96 - 2.6) 2.8 (1.4 - 5.4) 2.9 (1.3 - 6.4) 11 (5.6 - 20)
3.1% 3.7% 8.8% 11% 18% 18% 45%
30% 30% 17% 11% 6% 3% 3%
Assumption of pretest probability of 7%; final column shows percentage of scans with that particular result. MI, Myocardial infarction.
data. We obtained unpublished data for four of the studies included in this analysis. We also summarized studies that could not be included because of incomplete information, and most of these studies would likely have strengthened our main conclusion. We believe it is unlikely that we overlooked studies of any size with results showing a strong relationship between scintigraphy results and cardiac events. We believe that clinicians can use these results to refine their perioperative risk estimates. Reversible defects in up to 20% of myocardial segments did not significantly alter the likelihood of cardiac complications. For most patients with this scan result, it will be reasonable to proceed with surgery, after prescribing -blocker therapy37 where appropriate. Patients with greater extents of reversibility are at significantly increased risk. Delay or cancellation of surgery, invasive cardiac testing and revascularization, or -blocker therapy may be appropriate depending on the clinical circumstances. Only 23% of scans showed reversibility in more than 20% of myocardial segments, so the yield of preoperative scintigraphy is low even in selected patients. The extent of reversibility is not the only relevant information from preoperative dipyridamole myocardial stress perfusion imaging. The severity of reversibility within each defect6,38 and the location of the defect33 are also important considerations. Other findings, such as transient left ventricular cavitary dilatation38 or electrocardiographic changes during dipyridamole infusion,39 may also indicate increased risk of perioperative cardiac complications. These variables were not evaluated in this metaanalysis because they were not routinely reported in the primary studies. In summary, we conclude that reversible defects in less than 20% of myocardial segments do not significantly alter the risk of perioperative cardiac complications. Most of the studies were unblinded, so patients with reversible defects in less than 20% of myocardial segments may have received additional treatments that lowered their risk. Greater extents of reversibility on dipyridamole myocardial stress perfusion imaging increase the risk of perioperative complications after noncardiac vascular surgery, but the quality and amount of data regarding greater extents of reversibility are limited.
We thank the primary authors who supplied us with additional methodologic information and data from their studies. REFERENCES 1. Mangano DT. Perioperative cardiac morbidity. Anesthesiology 1990; 72:153-84. 2. Ashton CM, Petersen NJ, Wray NP, Kiefe CI, Dunn JK, Wu L, et al. The incidence of perioperative myocardial infarction in men undergoing noncardiac surgery. Ann Intern Med 1993;118:504-10. 3. Detsky AS, Abrams HB, Forbath N, Scott JG, Hilliard JR. Cardiac assessment for patients undergoing noncardiac surgery. A multifactorial clinical risk index. Arch Intern Med 1986;146:2131-4. 4. ACC/AHA Task Force Report. Committee on Perioperative Cardiovascular Evaluation for Noncardiac Surgery guidelines for perioperative cardiovascular evaluation for noncardiac surgery. J Am Coll Cardiol 1996;27:910-48. 5. Palda VA, Detsky AS. Perioperative assessment and management of risk from coronary artery disease. Ann Intern Med 1997;127:313-28. 6. Morise AP, Diamond GA, Detrano R, Bobbio M. Incremental value of exercise electrocardiography and thallium-201 testing in men and women for the presence and extent of coronary artery disease. Am Heart J 1995;130:267-76. 7. Mantha S, Roizen MF, Barnard J, Thisted RA, Ellis JE, Foss J. Relative effectiveness of four preoperative tests for predicting adverse cardiac outcomes after vascular surgery: a meta-analysis. Anesth Analg 1994; 79:422-33. 8. Bry JDL, Belkin M, O’Donnell TF, Mackey WC, Udelson JE, Schmid CH, et al. An assessment of the positive predictive value and costeffectiveness of dipyridamole myocardial scintigraphy in patients undergoing vascular surgery. J Vasc Surg 1994;19:112-24. 9. Lane SE, Lewis SM, Pippin JJ, Kosinski EJ, Campbell D, Nesto RW, et al. Predictive value of quantitative dipyridamole-thallium scintigraphy in assessing cardiovascular risk after vascular surgery in diabetes mellitus. Am J Cardiol 1989;64:1275-9. 10. McEnroe CS, O’Donnell TF Jr, Yeager A, Konstam M, Mackey WC. Comparison of ejection fraction and Goldman risk factor analysis to dipyridamole-thallium 201 studies in the evaluation of cardiac morbidity after aortic aneurysm surgery. J Vasc Surg 1990;11:497-504. 11. Shaw LJ, Eagle KA, Gersh BJ, Miller DD. Meta-analysis of intravenous dipyridamole-thallium-201 imaging (1985-1994) and dobutamine echocardiography (1991-1994) for risk stratification before vascular surgery. J Am Coll Cardiol 1996;27:787-98. 12. Lette J, Waters D, Lassonde J, Rene P, Picard M, Laureneau F, et al. Multivariate clinical models and quantitative dipyridamole-thallium imaging to predict cardiac mortality and death after vascular reconstruction. J Vasc Surg 1991;14:160-9. 13. Levinson JR, Boucher CA, Coley CM, Guiney TE, Strauss HW, Eagle KA. Usefulness of semiquantitative analysis of dipyridamolt-thallium-
540 Etchells et al
14.
15. 16.
17.
18.
19.
20.
21. 22.
23. 24. 25. 26.
27.
28.
201 redistribution for improving risk stratification before vascular surgery. Am J Cardiol 1990;66:406-10. Jaeschke R, Guyatt GH, Sackett DL. Users’ guide to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and how will they help me in caring for my patients? JAMA 1994;271:703-7. Holleman DR, Simel DL. Does the clinical examination predict airflow limitation? JAMA 1995;273:313-9. Mangano DT, Martin JL, Tubau JF, Browner WS, Hollenberg M, Krupski W, et al. Dipyridamole thallium-201 scintigrpahy as a preoperative screening test. A reexamination of its predictive potential. Circulation 1991;84:493-502. Baron J, Mundler O, Bertrand M, Vicaut E, Barre E, Godet G, et al. Dipyridamole-thallium scintigraphy and gated radionucleotide angiography to assess cardiac risk before abdominal aortic surgery. N Engl J Med 1994;330:663-9. Hanley JA, McNeil BJ. The meaning and use of the area under a receiving operator characteristic (ROC) curve. Radiology 1982;143: 29-36. Metz CE, Chen J-H, Wang P-L, Kronman HB. ROCFIT: a program for maximum likelihood estimation of a binormal ROC curve and its associated parameters from a set of categorical rating scale data. Chicago, Ill: Department of Radiology and The Franklin McLean Memorial Research Institute, University of Chicago; 1989. Dorfman DD, Alf E. Maximum likelihood estimation of parameters of signal detection theory and determination of confidence intervals: rating method data. J Math Psych 1969;6:487-96. Hellmich M, Abrams KR, Sutton AJ. Bayesian approaches to metaanalysis of ROC curves. Med Decis Making 1999;19:252-64. Irwig L, Tosteson A, Gatsonis C, Lau J, Colditz G, Chalmers TC, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994;120:667-76. Spiegelhalter DJ, Thomas A, Best NG. WinBUGS version 1.2 user manual. Cambridge, England: MRC Biostatistics Unit; 1999. Hellmich M, Abrams KR, Sutton AJ. Bayesian approaches to metaanalysis of ROC curves. Med Decis Making 1999;19:252-64. Dersimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177-88. Marwick TH, Shan K, Go RT, MacIntyre WJ, Lauer MS. Use of positron emission tomography for prediction of perioperative and late cardiac events before vascular surgery. Am Heart J 1995;130:1196-202. Vanzetto G, Machecourt J, Blendea D, Fagret D, Borrel E, Magne JL, et al. Additive value of thallium single-photon emission computed tomography myocardial imaging for prediction of perioperative events in clinically selected high cardiac risk patients having abdominal aortic surgery. Am J Cardiol 1996;77:143-8. Stratmann HG, Younis LT, Wittry MD, Amato M, Miller DD. Dipyridamole technetium-99m sestamibi myocardial tomography in patients evaluated for elective vascular surgery: prognostic value for perioperative and later cardiac events. Am Heart J 1996;131:923-9.
JOURNAL OF VASCULAR SURGERY September 2002
29. Huang Z, Komori S, Sawanobori T, Kohno I, Sano S, Ishihara T, et al. Dipyridamole thallium-201 single-photon emission computed tomography for prediction of perioperative cardiac events in patients with ateriosclerosis obliterans undergoing vascular surgery. Jpn Circ J 1998; 62:274-8. 30. Antalffy J, Bajnok L, Bertalan K, Varga J, Olvasto S, Fulop T Jr. Also vegtagi ermutetre kerulo betegek perioperativ kardialis kockazatanak felmerese dipyridamol myocardium szcintigrafiaval. Orvosi Hetilap 1995;136:703-7. 31. Brown KA, Rowen M. Extent of jeopardized viable myocardium determined by myocardial perfusion imaging best predicts perioperative cardiac events in patients undergoing noncardiac surgery. J Am Coll Cardiol 1993;21:325-30. 32. Kontos MC, Brath LK, Akosah KO, Mohanty PK. Cardiac complications in noncardiac surgery: relative value of resting two-dimensional echocardiography and dipyridamole thallium imaging. Am Heart J 1996;132:559-66. 33. Zarich SW, Cohen MC, Lane SE, Mittleman MA, Nesto RW, Hill T, et al. Routine perioperative dipyridamole 201T1 imaging in diabetic patients undergoing vascular surgery. Diabetes Care 1996;19:35560. 34. Fleisher LA, Rosenbaum SH, Nelson AH, Jain D, Wackers FJ, Zaret BL. Preoperative dipyridamole thallium imaging and ambulatory electrocardiographic monitoring as a predictor of perioperative cardiac events and long-term outcome. Anesthesiology 1995;83:90617. 35. Ombrellaro MP, Dieter Ra III, Freeman M, Stevens SL, Goldman MH. Role of dipyridamole myocardial scintigraphy in carotid artery surgery. J Am Coll Surg 1995;181:451-8. 36. Costello JM, Butcher RJ, Nassef LA, Youkey J, Menapace FJ. Dipyridamole thallium imaging: use in preoperative cardiac risk assessment in vascular surgery patients. Coron Artery Dis 1993;4:721-6. 37. Poldermans D, Boersma E, Bax JJ, Thomson IR, Van De Ven LLM, Blankensteijn JD, et al. The effect of bisoprolol on perioperative mortality and myocardial infarction in high-risk patients undergoing vascular surgery. N Engl J Med 1999;341:1789-94. 38. Lette J, Waters D, Cerino M, Picard M, Champagne P, Lapointe J. Preoperative coronary artery disease risk stratification based on dipyridamole imaging and a simple three-step three-segment model for patients undergoing noncardiac vascular surgery or major general surgery. Am J Cardiol 1992;69:1553-8. 39. L’Italien GJ, Paul SD, Hendel RC, Leppo JA, Cohen MC, Fleisher LA, et al. Development and validation of a Bayesian model for perioperative cardiac risk assessment in a cohort of 1,081 vascular surgical candidates. J Am Coll Cardiol 1996;27:779-86. Submitted Jan 9, 2002; accepted Apr 25, 2002.
Additional material for this article may be found online at www.mosby.com/jvs.
Appendix, online only. Main results of potentially relevant studies that could not be included Study
Sample size
MI/cardiac death
17
457
32
87
10 (11%)
34
172
16 (9%)
36
107
5 (5%)
31
231
19 (8.2%)
33
93
9 (9.7%)
35
174
4 (2%)
1321
105 (8%)
Total
42 (11%)*
Main result Thallium reversibility did not predict postoperative cardiac complication. Likelihood ratio for any reversible defect was 0.94 (95% CI, 0.60 – 1.5). Mean proportion of reversible defects was 7% in patients without events and 14% in patients with events. ‘Strongly positive’ scans had likelihood ratio 3.4 (95% CI, 1.5 – 6.9), and ‘weakly positive’ scans had likelihood ratio 1.0 (95% CI, 0.4 – 2.2). Likelihood ratio for any reversibility was 1.2 (95%, CI 0.4 – 1.9). Number of transient defects was predictive of outcome. Total number of defects (transient and fixed) was highly predictive of outcome. Likelihood ratio for reversibility in 1/12 segments was 3.3 (95% CI, 0.6 – 12).
Probable impact on conclusion of metaanalysis Strengthen
Strengthen
Strengthen
Strengthen Uncertain Uncertain
Weaken
*Study did not distinguish between cardiac and noncardiac deaths. MI, Myocardial infarction.
Online only. ROC curves for individual studies (dashed lines) and combined ROC curve (solid line). Two studies had significantly lower areas under ROC curve. Analysis was repeated after exclusion of these two studies.
Perioperative myocardial infarction occurs in 2% to 15% of patients undergoing vascular surgery.1 One quarter to one third of perioperative myocardial infarctions are fatal.2,3 Accurate preoperative assessment of cardiac risk helps patients and physicians make informed decisions about surgery, helps physicians choose appropriate methods for perioperative care, and identifies patients who need more intensive perioperative monitoring.
From the Divisions of General Internal Medicine at Sunnybrook and Women’s College Health Sciences Centre,a and University Health Network,b Department of Medicine,c and the Department of Public Health Sciences,d University of Toronto; and the Departments of Medicine and Clinical Epidemiology and Biostatistics, McMaster University.e Competition of interest: nil. Correspondence: Dr Edward E. Etchells, Department of Medicine, Sunnybrook and Women’s College Health Sciences Centre, 2075 Bayview Ave, Rm C 4-10, Toronto ON M4N 3M5 Canada (e-mail: edward.etchells@ swchsc.on.ca). Copyright © 2002 by The Society for Vascular Surgery and The American Association for Vascular Surgery. 0741-5214/2002/$35.00 ⫹ 0 24/1/126563 doi:10.1067/mva.2002.126563
534
Practice guidelines for assessment of perioperative cardiac risk recommend a clinical assessment followed by selective use of noninvasive cardiac testing.4,5 The most extensively studied method of noninvasive cardiac testing is dipyridamole myocardial stress perfusion imaging. If a reversible myocardial perfusion defect is observed during dipyridamole imaging, then the patient is identified as high risk for perioperative myocardial infarction and cardiac death. This approach is based on the interpretation of myocardial stress perfusion imaging as showing either any reversibility or no reversibility of myocardial perfusion defects. However, a semiquantitative description of the extent of reversibility may result in better prediction of perioperative cardiac complications because the extent of reversibility is a powerful predictor of the extent of coronary artery disease on coronary angiography in patients with suspected coronary artery disease.6 The American College of Cardiology/American Heart Association guideline states that “the use of techniques to quantitate the extent of abnormality will probably improve the positive predictive nature of myocardial perfusion imaging.”4 The American College
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
of Physicians guideline states that “increased risk discrimination is obtained from semiquantitative interpretation.”5 Two published metaanalyses of dipyridamole myocardial stress perfusion imaging before vascular surgery have been published. One study did not address semiquantitative dipyridamole myocardial stress perfusion imaging.7 The other combined three studies8-10 (total of 433 patients) and found that patients with two or more reversible defects were more likely to have cardiac complications (including congestive heart failure and transient myocardial ischemia) than patients with one or more reversible defects.11 This analysis did not include two early studies that evaluated 305 patients and report semiquantitative analysis.12,13 Several additional studies of semiquantative dipyridamole myocardial stress perfusion imaging have been published since 1994. Our objective was to conduct a systematic review and metaanalysis of evidence about semiquantitative dipyridamole-thallium scintigraphy for prediction of nonfatal myocardial infarction and cardiac death after noncardiac vascular surgery. METHODS Assembly of studies We searched for published and unpublished data with computerized bibliographies, hand searching of relevant journals, and correspondence with study authors. In our MEDLINE search, we used MeSH headings dipyridamole (exploded) or thallium (exploded), limited to human and journal article, for the period from 1975 to January 1999. We also searched with MeSH heading sestamibi (exploded), but no additional articles were identified, so this search is not reported in detail. We considered reports in all languages. All article titles, abstracts, and subject headings were screened for potential relevance. We looked for evidence of preoperative dipyridamole myocardial stress perfusion imaging (including thallium-201, technetium-99m sestamibi, and rubidium-82, and planar, single photon emission computer tomography, and positron emission tomography) in the setting of noncardiac vascular surgery, including carotid endarterectomy, aortic procedures, renal artery bypass or repair, lower extremity arterial bypass or thromboendarterectomy, and diabetic amputations. We excluded articles indexed as a review article or a case report. We also excluded studies of renal transplantation and vascular access for hemodialysis. The first 300 citations were screened by two independent reviewers with excellent agreement regarding potential relevance (simple agreement, 99%; , 0.92), so the remaining citations were screened by a single reviewer. We also searched the citation lists of all selected studies and published guidelines to identify other potentially relevant studies. Study selection The full reports of all potentially relevant articles and abstracts were reviewed in detail for three additional selection criteria: 1, evidence of semiquantitative interpretation of scintigraphy results; 2, the reporting of perioperative
Etchells et al 535
myocardial infarction and cardiac death as study outcomes; and 3, follow-up for at least 1 month or for the duration of hospitalization. We did not consider unstable angina because it is not consistently reported, and we did not consider arrhythmias, pulmonary edema, and all-cause mortality because these are not as strongly related to reversible cardiac ischemia. When the same data were reported in more than one article, only the article with the most complete data set was included. A subset of potentially relevant articles was reviewed by two independent reviewers with excellent agreement regarding selection (simple agreement, 96%; , 0.90). Authors of studies that met most or all selection criteria were contacted by mail with a request to provide methodologic information and additional results. Evaluation of study methods (Table I) Two independent reviewers evaluated study methods according to critical appraisal principles of diagnostic test research.14 We resolved discrepancies with repeated review and discussion. We abstracted data on the prospective enrollment of consecutive participants, with blinded interpretation of scintigraphy and blinded assessment of perioperative cardiac complications. We assigned a quality grade on the basis of these features as previously described.15 Grade A studies involved the independent comparison of myocardial perfusion scintigraphy results with development of perioperative cardiac complications among at least 100 consecutive patients. Grade B studies met the criteria for grade A studies but had less than 100 patients. Grade C studies involved nonconsecutive patients or nonindependent comparisons between the scintigraphy and perioperative cardiac complications. Several features of perioperative cardiac studies were considered in our appraisals. First, routine use of dipyridamole myocardial stress perfusion imaging for all patients before noncardiac vascular surgery is not recommended.16,17 Accordingly, we determined whether study patients were selected on the basis of clinical characteristics indicating increased cardiac risk. Second, patients with reversible defects may have their surgery canceled or they may undergo coronary revascularization before noncardiac vascular surgery. In either situation, these patients are less likely to have a perioperative cardiac complication, thereby reducing the observed predictive value of a reversible defect. We determined whether dipyridamole myocardial stress perfusion imaging results led to cancellation of surgery or to preoperative coronary revascularization, and we abstracted data on the clinical and scintigraphic characteristics of such patients. Third, patients with reversible defect scans who undergo noncardiac vascular surgery may receive more intense postoperative electrocardiographic and cardiac enzyme monitoring. This surveillance bias can result in more diagnoses of postoperative myocardial infarction in patients with reversible defects, leading to spuriously high estimates of sensitivity. Therefore, we assessed the degree of blinding of caregivers (anesthetists, internists, and surgeons) to di-
JOURNAL OF VASCULAR SURGERY September 2002
536 Etchells et al
Table I. Summary of studies included in metaanalysis MI/ Sample cardiac Study size deaths 8
237
17 (7%)
Avoided Patient MI Quality Blinding: Blinding: surgical Routine characteristics definition Scintigraphy grade Prospective Consecutive test outcome selection surveillance Diabetes 34% Angina 18% MI 30%
CK alone Thallium or with Planar and EKG SPECT Reinjection 9 101 10 (10%) Diabetes 100% EKG and Thallium Angina 46% CK Planar MI 39% 12 125 13 (10%) Diabetes 24% Any two: Thallium Angina 46% EKG, Planar CK, or pain 13 180 12 (7%) Diabetes 18% EKG and Thallium Angina 29% CK Planar MI 23% 26 72 5 (7%) Diabetes 26% CK with Rubidium Angina 31% pain or 82 PET MI 26% EKG 27 134* 12 (9%) Diabetes 13% EKG and Thallium CK SPECT Angina 23% Reinjection MI 29% 28 197 4 (2%) Diabetes 25% EKG and Sestamibi Angina 60% CK SPECT MI 23% 29 104 5 (5%) Diabetes 26% EKG and Thallium Angina 11% CK SPECT MI 7% 30 29 4 (14%) Diabetes 3% EKG and Sestamibi Angina 14% CK Planar Total 1179 82 (7.0%)
C
No
No
Yes
No
No
No
C
Yes
No
Yes
No
Yes
No
C
No
Yes
Yes
No
No
No
C
No
No
Yes
Yes
No
No
C
No
Yes
Yes
No
No
No
A
Yes
Yes
Yes
Yes
Yes
Yes†
C
No
Yes
Yes
No
No
No
C
No
Yes
Yes
Yes
No
No
B
Yes
Yes
Yes
Yes
Yes
Yes‡
*This study used explicit selection criteria (any two of: age ⬎70 years, history of myocardial infarction, history of congestive heart failure, history of angina, diabetes mellitus, hypertension with left ventricular hypertrophy, Q waves, or ischemic ST segment changes on baseline electrocardiogram). Of 514 patients considered for surgery, 134 had at least two of these risk factors, 23 surgeries were canceled for noncardiac reasons, 37 were canceled because of high cardiac risk, and 317 had less than two risk factors and underwent surgery without scanning. Cardiac events were observed in 11 of 314 patients (3.5%) with less than two risk factors who did not have scintigraphy. † Routine surveillance: electrocardiogram and cardiac enzymes every 24 hours for 2 days after surgery. ‡ Routine surveillance: electrocardiogram twice daily in intensive care unit (at least 24 hours after surgery) and cardiac enzymes twice daily for 3 days after surgery. MI definition: This column describes whether electrocardiographic changes, cardiac enzyme changes, or both are necessary for diagnosis of perioperative myocardial infarction. Some studies also included clinical criterion of prolonged typical ischemic chest pain. Blinding: test: Dipyridamole myocardial perfusion scintigraphy was interpreted without knowledge of patient’s perioperative outcome. Blinding: outcome: Assessment of cardiac complication was made without knowledge of patient’s preoperative dipyridamole myocardial perfusion scintigraphy. Avoided surgical selection: Surgery was not canceled and patients did not undergo revascularization before noncardiac surgery because of reversible defects noted on dipyridamole myocardial perfusion scintigraphy. If study did not avoid surgical selection, study would be less likely to show relationship between reversible defects and perioperative cardiac complications. Routine surveillance: Standard protocol was used for monitoring all patients after surgery, including electrocardiogram and cardiac enzyme measurement, regardless of preoperative scan result. MI, Myocardial infarction; CK, cardiac enzyme; EKG, electrocardiogram; SPECT, single photon emission computed tomography; PET, position emission tomography.
pyridamole myocardial stress perfusion imaging results and determined whether systematic protocols for postoperative electrocardiographic and cardiac enzyme monitoring were used. Finally, patients with reversible defects may receive additional perioperative treatments to reduce the likelihood of perioperative cardiac complications. If caregivers were not blinded to scan results, we abstracted information on the perioperative care for patients with and without reversible defects.
Data abstraction Two independent reviewers abstracted the data. We resolved discrepancies with repeated review and discussion. Studies quantified the extent of reversibility with division of the myocardial images into six to 24 segments and reporting of the number of segments with reversibility. To standardize results across studies, we calculated the percentage of myocardial segments showing reversible defects. For example, if the study used a six-segment
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
model, then reversibility in one segment would correspond to 17% of myocardium showing reversibility. We recorded study definitions of myocardial infarction, including electrocardiographic and cardiac enzyme criteria. Troponin I was not used in any studies. We did not record data on other cardiac complications, such as unstable angina, pulmonary edema, or arrhythmias. For each study, we identified patients who had coronary revascularization after scanning but before noncardiac vascular surgery. Whenever possible, we excluded these patients from our calculations of diagnostic accuracy. Inclusion of these patients would spuriously lower the specificity of an abnormal dipyridamole myocardial stress perfusion imaging result because perioperative cardiac complications are rare after successful myocardial revascularization. Data analysis Individual studies. We calculated nonparametric areas under the receiver operating characteristic (ROC) curve for each study with the method of Hanley and McNeil18 and compared the areas under each curve with z tests. We also calculated binormal estimates for the areas under the ROC curve, with ROCFIT software19 and a maximum likelihood method.20,21 A perfect test has an area under the ROC curve of 1. An area under the curve of 0.5 indicates a useless test that performs no better than a coin flip. Combining studies. To determine which studies could be meaningfully combined, we applied standard principles of metaanalysis.22 First, we tried to assemble studies with reasonably similar features, including patient population, type of surgery, type of imaging, and definition of postoperative myocardial infarction. Second, we assessed the methodologic quality of each study. Third, we determined similarity of results amongst the studies by comparing the areas under the ROC curves. With the BUGS23 software, we fitted a Bayesian hierarchic model24 to combine the binormal ROC curve estimates from individual studies. We then estimated summary likelihood ratios (LRs) for increasing extents of reversibility with the Bayesian hierarchic model. In advance, we defined seven scan results, including normal scans, fixed defects only, reversible defects in less than 20%, and reversible defects in 20% to 29%, 30% to 39%, 40% to 49%, and more than 50% of myocardial segments. The major analytic issue in estimating summary LRs was that individual studies used different intervals for reporting extents of reversibility. To calculate the desired LRs, we interpolated the results from each study to conform to the scan results listed previously. These interpolations had little effect on the shape or area of the original or summary ROC curves. Once the data had been interpolated to the intervals listed previously, we calculated stratum-specific LRs for each study, adding 0.5 to each cell to give the estimate better properties in the presence of small counts. We used the DerSimonian and Laird25 random effects method to obtain a summary LR for each stratum. The summary LRs
Etchells et al 537
were used to convert pretest probabilities of myocardial infarction to posttest probabilities for test outcomes at each level of reversibility. A LR of less than 1 reduces the likelihood of complications, and a LR of less than 0.4 means that the likelihood of complications is substantially reduced. A LR of greater than 1 increases the likelihood of complications, and a LR of 4 means that the likelihood of complications is substantially increased. RESULTS We screened 3160 MEDLINE citations, identified 88 potentially relevant articles, and selected 16 for inclusion in this metaanalysis. The 72 other potentially relevant articles were excluded for the following reasons: no semiquantitative analysis (n ⫽ 28), nonvascular surgery (n ⫽ 10), review article (n ⫽ 7), repeat publication (n ⫽ 7), no myocardial stress perfusion imaging (n ⫽ 6), no surgery (n ⫽ 4), heterogeneous outcomes (n ⫽ 3), sample size less than 10 patients (n ⫽ 2), and other reasons (n ⫽ 5).(A list of the excluded articles is available from the corresponding author). Of the 16 selected articles, we were able to obtain sufficient data to analyze nine studies.8,9,12,13,26-30 The remaining seven selected articles did not have enough information to be included in the metaanalysis, and we were unable to obtain additional information from the authors.17,31-36 Methodologic features of primary studies (Table I). One selected study had a quality grade A,27 another grade B,30 and the rest were grade C.8,9,12,13,26,28,29 All studies selected patients for myocardial stress perfusion imaging on the basis of a clinical suspicion of increased cardiac risk, but only one study27 explicitly described the selection criteria. Six studies were partly confounded by cancellation of surgery or revascularization because of abnormal scan results.8,12,13,26,28,29 In most cases, revascularization or cancellation occurred in patients with greater extents of reversibility. Caregivers (anesthetists, surgeons, and internists) were blinded to scintigraphy results in only two studies.27,30 The remaining studies provided little detail regarding additional treatments for patients with reversible defects. One study stated that patients with reversible defects tended to receive more invasive hemodynamic monitoring and intravenous nitrates,13 and another stated that optimization of antiischemic therapy and intraoperative hemodynamic monitoring was routinely recommended in patients with reversible defects.28 All studies used a similar protocol for infusing dipyridamole. The radiopharmaceutical was thallium-201 in six studies, 8,9,12,13,27,29 technetium-99m sestamibi in two studies,28,30 and rubidium-82 in one study.26 Five studies used single photon emission computed tomography,8,27,28,29 one used positron emission tomography,26 and three used planar images.9,12,13,30 The method of describing extent of reversibility was variable, with a range of five to 24 segments. Individual study results (Table II). The areas under the ROC curves were similar for the nonparametric and
JOURNAL OF VASCULAR SURGERY September 2002
538 Etchells et al
Table II. Areas under ROC curves for each study Study
Area under ROC curve (95% CI)
8* 9* 12 13 26† 27 28† 29 30* Combined Combined, excluding studies 26 and 28
0.76 (0.65 - 0.86) 0.82 (0.69 - 0.95) 0.96 (0.91 - 1.00) 0.86 (0.71 - 1.00) 0.46 (0.13 - 0.79) 0.84 (0.73 - 0.95) 0.47 (0.16 - 0.78) 0.88 (0.78 - 0.97) 0.73 (0.53 - 0.92) 0.78 (0.65 - 0.89) 0.82 (0.69 - 0.92)
*Differs from study 12, P ⬍ .05. † Differs from studies 9, 12, 13, 21, 23, P ⬍ .05.
binormal methods, so we report the nonparametric areas only. Two studies had significantly lower areas under the ROC curves than the remaining studies.26,28 Each study26,28 had a high rate of cancellation for patients with reversible defects (approximately 20% were canceled). One of these studies included results for seven patients with reversible defects who had myocardial revascularization before noncardiac vascular surgery.28 We were unable to obtain results of this study with these seven patients excluded. The individual and summary ROC curves can be viewed in the Fig (online only). The extent to which these outlying studies are “biased” estimates of the accuracy of dipyridamole myocardial stress perfusion imaging is impossible to determine. Therefore, we performed analyses with and without these two studies. Combined study results (Table III). The pooled area under the summary ROC curve was 0.78 (95% CI, 0.65 to 0.89), with the Bayesian hierarchic model to combine the binormal ROC curve area estimates. When the two outlying studies were excluded, the pooled area under the ROC curve was 0.82 (95% CI, 0.69 to 0.92). We obtained similar results when we combined the nonparametric ROC curve area estimates with the random effects model. Normal scans significantly reduced the likelihood of perioperative cardiac complications (summary LR, 0.42; 95% CI, 0.20 to 0.88). Fixed defects reduced the likelihood of complications, but the effect was not statistically significant (summary LR, 0.51; 95% CI, 0.24 to 1.1). Reversibility in less than 20% of myocardial segments did not change the likelihood of perioperative cardiac complications (LR, 1.3; 95% CI, 0.88 to 1.9). LRs for increasing extents of reversibility were: 20% to 29% reversibility (LR, 1.6; 95% CI, 1.0 to 2.6), 30% to 39% reversibility (LR, 2.9; 95% CI, 1.6 to 5.1), 40% to 49% reversibility (LR, 2.9; 95% CI, 1.4 to 6.2), and 50% or more reversibility (LR, 11; 95% CI, 5.8 to 20). When we excluded the two outlying studies, there were no statistically significant changes in the results. The only qualitative difference was that the LR for a normal scan was lower (0.18; 95% CI, 0.07 to 0.49). Table III shows the posttest probability for various scan results, assuming a pretest probability of cardiac complica-
tions of 7%. Table III also shows the distribution of various scan results: 60% of scans were normal or showed fixed defects, 17% of scans showed reversibility in less than 20% of myocardial segments, and 23% of scans showed reversibility in 20% or more of myocardial segments. Potentially relevant studies that could not be analyzed. Seven potentially relevant studies were not included in this analysis because of inadequate information regarding semiquantitative analysis. Four studies17,32,34,36 (n ⫽ 823, including 73 cardiac events) had results that would strengthen our conclusions. Two studies31,33 (n ⫽ 324, including 28 cardiac events) had results of uncertain effect on our conclusions. One study35 (n ⫽ 174, including four cardiac events) had results that would potentially weaken our conclusions. These studies are summarized in the Appendix (online only). DISCUSSION In this metaanalysis, our major finding is that patients undergoing preoperative semiquantitative dipyridamole myocardial perfusion scintigraphy who had reversible defects in up to 20% of myocardial segments did not have an increased risk of perioperative cardiac complications. We also found that patients with greater extents of reversible defects were at increasing risk of cardiac complications after noncardiac vascular surgery. Two major methodologic issues may affect our conclusions: the methodologic quality of the primary studies and publication bias. We believe that our major finding is unaffected by these methodologic issues, but it is possible that the predictive value of greater extents of reversibility has been overestimated. Most studies in this analysis had suboptimal methodologic quality. Ideally, we would like to base our recommendations on higher quality primary studies, but no such studies exist. Instead, we must examine how the methods of the individual studies may have affected the results of this analysis. First, caregivers were blinded to the results of myocardial stress perfusion imaging in only two studies. It is likely that additional treatments were given to patients with reversible perfusion defects, leading to fewer cardiac complications. However, the lack of blinding reflects what occurs in usual practice, so we conclude that patients with reversibility in less than 20% of myocardial segments, who receive appropriate perioperative care, have an average risk of perioperative cardiac complications. Second, unequal surveillance for postoperative myocardial infarction and unblinded interpretation of study outcomes would tend to overestimate the predictive value of myocardial stress perfusion imaging. However, our major finding is not attenuated by these methodologic limitations because we conclude that lesser extents of reversibility have limited predictive value. It is possible that the predictive value of greater extents of reversibility has been overestimated in our analysis. We made efforts to minimize the possibility of publication bias by contacting authors for additional unpublished
JOURNAL OF VASCULAR SURGERY Volume 36, Number 3
Etchells et al 539
Table III. Summary likelihood ratios and estimated posttest probability of perioperative cardiac complications for each scan result Extent of reversibility No defects Fixed defects only Reversibility ⬍20% Reversibility 20%-29% Reversibility 30%-39% Reversibility 40%-49% Reversibility ⱖ50%
Likelihood ratio (all studies; 95% CI)
Likelihood ratio (excluding studies 26 and 28; 95% CI)
Posttest probability of MI or cardiac death
Percentage of scans with this result
0.42 (0.20 - 0.88) 0.51 (0.24 - 1.1) 1.3 (0.88 - 1.9) 1.6 (1.0 - 2.6) 2.9 (1.6 - 5.1) 2.9 (1.4 - 6.2) 11 (5.8 - 20)
0.18 (0.07 - 0.49) 0.46 (0.17 - 1.2) 1.3 (0.86 - 2.0) 1.6 (0.96 - 2.6) 2.8 (1.4 - 5.4) 2.9 (1.3 - 6.4) 11 (5.6 - 20)
3.1% 3.7% 8.8% 11% 18% 18% 45%
30% 30% 17% 11% 6% 3% 3%
Assumption of pretest probability of 7%; final column shows percentage of scans with that particular result. MI, Myocardial infarction.
data. We obtained unpublished data for four of the studies included in this analysis. We also summarized studies that could not be included because of incomplete information, and most of these studies would likely have strengthened our main conclusion. We believe it is unlikely that we overlooked studies of any size with results showing a strong relationship between scintigraphy results and cardiac events. We believe that clinicians can use these results to refine their perioperative risk estimates. Reversible defects in up to 20% of myocardial segments did not significantly alter the likelihood of cardiac complications. For most patients with this scan result, it will be reasonable to proceed with surgery, after prescribing -blocker therapy37 where appropriate. Patients with greater extents of reversibility are at significantly increased risk. Delay or cancellation of surgery, invasive cardiac testing and revascularization, or -blocker therapy may be appropriate depending on the clinical circumstances. Only 23% of scans showed reversibility in more than 20% of myocardial segments, so the yield of preoperative scintigraphy is low even in selected patients. The extent of reversibility is not the only relevant information from preoperative dipyridamole myocardial stress perfusion imaging. The severity of reversibility within each defect6,38 and the location of the defect33 are also important considerations. Other findings, such as transient left ventricular cavitary dilatation38 or electrocardiographic changes during dipyridamole infusion,39 may also indicate increased risk of perioperative cardiac complications. These variables were not evaluated in this metaanalysis because they were not routinely reported in the primary studies. In summary, we conclude that reversible defects in less than 20% of myocardial segments do not significantly alter the risk of perioperative cardiac complications. Most of the studies were unblinded, so patients with reversible defects in less than 20% of myocardial segments may have received additional treatments that lowered their risk. Greater extents of reversibility on dipyridamole myocardial stress perfusion imaging increase the risk of perioperative complications after noncardiac vascular surgery, but the quality and amount of data regarding greater extents of reversibility are limited.
We thank the primary authors who supplied us with additional methodologic information and data from their studies. REFERENCES 1. Mangano DT. Perioperative cardiac morbidity. Anesthesiology 1990; 72:153-84. 2. Ashton CM, Petersen NJ, Wray NP, Kiefe CI, Dunn JK, Wu L, et al. The incidence of perioperative myocardial infarction in men undergoing noncardiac surgery. Ann Intern Med 1993;118:504-10. 3. Detsky AS, Abrams HB, Forbath N, Scott JG, Hilliard JR. Cardiac assessment for patients undergoing noncardiac surgery. A multifactorial clinical risk index. Arch Intern Med 1986;146:2131-4. 4. ACC/AHA Task Force Report. Committee on Perioperative Cardiovascular Evaluation for Noncardiac Surgery guidelines for perioperative cardiovascular evaluation for noncardiac surgery. J Am Coll Cardiol 1996;27:910-48. 5. Palda VA, Detsky AS. Perioperative assessment and management of risk from coronary artery disease. Ann Intern Med 1997;127:313-28. 6. Morise AP, Diamond GA, Detrano R, Bobbio M. Incremental value of exercise electrocardiography and thallium-201 testing in men and women for the presence and extent of coronary artery disease. Am Heart J 1995;130:267-76. 7. Mantha S, Roizen MF, Barnard J, Thisted RA, Ellis JE, Foss J. Relative effectiveness of four preoperative tests for predicting adverse cardiac outcomes after vascular surgery: a meta-analysis. Anesth Analg 1994; 79:422-33. 8. Bry JDL, Belkin M, O’Donnell TF, Mackey WC, Udelson JE, Schmid CH, et al. An assessment of the positive predictive value and costeffectiveness of dipyridamole myocardial scintigraphy in patients undergoing vascular surgery. J Vasc Surg 1994;19:112-24. 9. Lane SE, Lewis SM, Pippin JJ, Kosinski EJ, Campbell D, Nesto RW, et al. Predictive value of quantitative dipyridamole-thallium scintigraphy in assessing cardiovascular risk after vascular surgery in diabetes mellitus. Am J Cardiol 1989;64:1275-9. 10. McEnroe CS, O’Donnell TF Jr, Yeager A, Konstam M, Mackey WC. Comparison of ejection fraction and Goldman risk factor analysis to dipyridamole-thallium 201 studies in the evaluation of cardiac morbidity after aortic aneurysm surgery. J Vasc Surg 1990;11:497-504. 11. Shaw LJ, Eagle KA, Gersh BJ, Miller DD. Meta-analysis of intravenous dipyridamole-thallium-201 imaging (1985-1994) and dobutamine echocardiography (1991-1994) for risk stratification before vascular surgery. J Am Coll Cardiol 1996;27:787-98. 12. Lette J, Waters D, Lassonde J, Rene P, Picard M, Laureneau F, et al. Multivariate clinical models and quantitative dipyridamole-thallium imaging to predict cardiac mortality and death after vascular reconstruction. J Vasc Surg 1991;14:160-9. 13. Levinson JR, Boucher CA, Coley CM, Guiney TE, Strauss HW, Eagle KA. Usefulness of semiquantitative analysis of dipyridamolt-thallium-
540 Etchells et al
14.
15. 16.
17.
18.
19.
20.
21. 22.
23. 24. 25. 26.
27.
28.
201 redistribution for improving risk stratification before vascular surgery. Am J Cardiol 1990;66:406-10. Jaeschke R, Guyatt GH, Sackett DL. Users’ guide to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and how will they help me in caring for my patients? JAMA 1994;271:703-7. Holleman DR, Simel DL. Does the clinical examination predict airflow limitation? JAMA 1995;273:313-9. Mangano DT, Martin JL, Tubau JF, Browner WS, Hollenberg M, Krupski W, et al. Dipyridamole thallium-201 scintigrpahy as a preoperative screening test. A reexamination of its predictive potential. Circulation 1991;84:493-502. Baron J, Mundler O, Bertrand M, Vicaut E, Barre E, Godet G, et al. Dipyridamole-thallium scintigraphy and gated radionucleotide angiography to assess cardiac risk before abdominal aortic surgery. N Engl J Med 1994;330:663-9. Hanley JA, McNeil BJ. The meaning and use of the area under a receiving operator characteristic (ROC) curve. Radiology 1982;143: 29-36. Metz CE, Chen J-H, Wang P-L, Kronman HB. ROCFIT: a program for maximum likelihood estimation of a binormal ROC curve and its associated parameters from a set of categorical rating scale data. Chicago, Ill: Department of Radiology and The Franklin McLean Memorial Research Institute, University of Chicago; 1989. Dorfman DD, Alf E. Maximum likelihood estimation of parameters of signal detection theory and determination of confidence intervals: rating method data. J Math Psych 1969;6:487-96. Hellmich M, Abrams KR, Sutton AJ. Bayesian approaches to metaanalysis of ROC curves. Med Decis Making 1999;19:252-64. Irwig L, Tosteson A, Gatsonis C, Lau J, Colditz G, Chalmers TC, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994;120:667-76. Spiegelhalter DJ, Thomas A, Best NG. WinBUGS version 1.2 user manual. Cambridge, England: MRC Biostatistics Unit; 1999. Hellmich M, Abrams KR, Sutton AJ. Bayesian approaches to metaanalysis of ROC curves. Med Decis Making 1999;19:252-64. Dersimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986; 7:177-88. Marwick TH, Shan K, Go RT, MacIntyre WJ, Lauer MS. Use of positron emission tomography for prediction of perioperative and late cardiac events before vascular surgery. Am Heart J 1995;130:1196-202. Vanzetto G, Machecourt J, Blendea D, Fagret D, Borrel E, Magne JL, et al. Additive value of thallium single-photon emission computed tomography myocardial imaging for prediction of perioperative events in clinically selected high cardiac risk patients having abdominal aortic surgery. Am J Cardiol 1996;77:143-8. Stratmann HG, Younis LT, Wittry MD, Amato M, Miller DD. Dipyridamole technetium-99m sestamibi myocardial tomography in patients evaluated for elective vascular surgery: prognostic value for perioperative and later cardiac events. Am Heart J 1996;131:923-9.
JOURNAL OF VASCULAR SURGERY September 2002
29. Huang Z, Komori S, Sawanobori T, Kohno I, Sano S, Ishihara T, et al. Dipyridamole thallium-201 single-photon emission computed tomography for prediction of perioperative cardiac events in patients with ateriosclerosis obliterans undergoing vascular surgery. Jpn Circ J 1998; 62:274-8. 30. Antalffy J, Bajnok L, Bertalan K, Varga J, Olvasto S, Fulop T Jr. Also vegtagi ermutetre kerulo betegek perioperativ kardialis kockazatanak felmerese dipyridamol myocardium szcintigrafiaval. Orvosi Hetilap 1995;136:703-7. 31. Brown KA, Rowen M. Extent of jeopardized viable myocardium determined by myocardial perfusion imaging best predicts perioperative cardiac events in patients undergoing noncardiac surgery. J Am Coll Cardiol 1993;21:325-30. 32. Kontos MC, Brath LK, Akosah KO, Mohanty PK. Cardiac complications in noncardiac surgery: relative value of resting two-dimensional echocardiography and dipyridamole thallium imaging. Am Heart J 1996;132:559-66. 33. Zarich SW, Cohen MC, Lane SE, Mittleman MA, Nesto RW, Hill T, et al. Routine perioperative dipyridamole 201T1 imaging in diabetic patients undergoing vascular surgery. Diabetes Care 1996;19:35560. 34. Fleisher LA, Rosenbaum SH, Nelson AH, Jain D, Wackers FJ, Zaret BL. Preoperative dipyridamole thallium imaging and ambulatory electrocardiographic monitoring as a predictor of perioperative cardiac events and long-term outcome. Anesthesiology 1995;83:90617. 35. Ombrellaro MP, Dieter Ra III, Freeman M, Stevens SL, Goldman MH. Role of dipyridamole myocardial scintigraphy in carotid artery surgery. J Am Coll Surg 1995;181:451-8. 36. Costello JM, Butcher RJ, Nassef LA, Youkey J, Menapace FJ. Dipyridamole thallium imaging: use in preoperative cardiac risk assessment in vascular surgery patients. Coron Artery Dis 1993;4:721-6. 37. Poldermans D, Boersma E, Bax JJ, Thomson IR, Van De Ven LLM, Blankensteijn JD, et al. The effect of bisoprolol on perioperative mortality and myocardial infarction in high-risk patients undergoing vascular surgery. N Engl J Med 1999;341:1789-94. 38. Lette J, Waters D, Cerino M, Picard M, Champagne P, Lapointe J. Preoperative coronary artery disease risk stratification based on dipyridamole imaging and a simple three-step three-segment model for patients undergoing noncardiac vascular surgery or major general surgery. Am J Cardiol 1992;69:1553-8. 39. L’Italien GJ, Paul SD, Hendel RC, Leppo JA, Cohen MC, Fleisher LA, et al. Development and validation of a Bayesian model for perioperative cardiac risk assessment in a cohort of 1,081 vascular surgical candidates. J Am Coll Cardiol 1996;27:779-86. Submitted Jan 9, 2002; accepted Apr 25, 2002.
Additional material for this article may be found online at www.mosby.com/jvs.
Appendix, online only. Main results of potentially relevant studies that could not be included Study
Sample size
MI/cardiac death
17
457
32
87
10 (11%)
34
172
16 (9%)
36
107
5 (5%)
31
231
19 (8.2%)
33
93
9 (9.7%)
35
174
4 (2%)
1321
105 (8%)
Total
42 (11%)*
Main result Thallium reversibility did not predict postoperative cardiac complication. Likelihood ratio for any reversible defect was 0.94 (95% CI, 0.60 – 1.5). Mean proportion of reversible defects was 7% in patients without events and 14% in patients with events. ‘Strongly positive’ scans had likelihood ratio 3.4 (95% CI, 1.5 – 6.9), and ‘weakly positive’ scans had likelihood ratio 1.0 (95% CI, 0.4 – 2.2). Likelihood ratio for any reversibility was 1.2 (95%, CI 0.4 – 1.9). Number of transient defects was predictive of outcome. Total number of defects (transient and fixed) was highly predictive of outcome. Likelihood ratio for reversibility in 1/12 segments was 3.3 (95% CI, 0.6 – 12).
Probable impact on conclusion of metaanalysis Strengthen
Strengthen
Strengthen
Strengthen Uncertain Uncertain
Weaken
*Study did not distinguish between cardiac and noncardiac deaths. MI, Myocardial infarction.
Online only. ROC curves for individual studies (dashed lines) and combined ROC curve (solid line). Two studies had significantly lower areas under ROC curve. Analysis was repeated after exclusion of these two studies.