Concordance of Preoperative Clinical Risk With Angiographic Severity of Coronary Artery Disease in Patients Undergoing Vascular Surgery
Background Preoperative clinical indexes to stratify cardiac risk have not been validated angiographically. Our aims were to determine the concordance of clinical risk with severity of coronary stenosis and to develop and validate a preoperative clinical index to exclude the presence of significant coronary stenosis.
Methods and Results We carried out a prospective study of 878 consecutive patients (including the derivation and validation sets). “Severe” stenosis was defined as three-vessel (≥50% stenosis in each), two-vessel (≥50% stenosis in one when the other is ≥70% stenosis of the left anterior descending), or left main disease (≥50%); “critical” stenosis was three-vessel (≥70% stenosis in each) and/or left main stenosis ≥70%. A preoperative clinical index (diabetes mellitus, prior myocardial infarction, angina, age >70 years, congestive heart failure) was used to stratify patients. A gradient of risk for severe stenosis was seen with increasing numbers of clinical markers. The following prediction rules were developed: The absence of severe coronary stenoses can be predicted with a positive predictive value of 96% for patients who have no (1) history of diabetes, (2) prior angina, (3) previous myocardial infarction, or (4) history of congestive heart failure. The absence of critical coronary stenoses can be predicted with a positive predictive value of 94% for those who have no (1) prior angina, (2) previous myocardial infarction, or (3) history of congestive heart failure.
Conclusions By reliably identifying a large proportion of patients with a low likelihood of significant stenoses, these prediction rules can help to substantially reduce healthcare costs associated with preoperative cardiac risk assessment for noncardiac surgery.
Patients undergoing vascular surgery have a high incidence of both early and late postoperative cardiac complications.1 2 3 4 5 Previous studies have attempted to use a careful clinical evaluation to quantify the risk of postoperative cardiac events for patients undergoing noncardiac surgery.7 8 9 10 11 12 13 14 Such an assessment can have both prognostic and therapeutic implications.
These preoperative clinical algorithms include the Goldman8 9 and Modified Goldman10 11 indexes, both of which have been found to be useful in the preoperative assessment of the general surgical patient. However, several studies have suggested that these indexes underestimate the risk of cardiac complications in patients undergoing vascular surgery.15 16 This concern is exemplified by documentation of Hertzer and colleagues17 that patients undergoing vascular surgery often harbor severe but clinically unsuspected coronary disease.
In an attempt to estimate the risk of postoperative cardiac complications more accurately, Eagle and colleagues13 developed a preoperative clinical risk index for patients undergoing vascular surgery. By use of this clinical index, patients are stratified into low-, intermediate-, and high-clinical-risk groups for postoperative cardiac complications. Eagle and colleagues suggest that patients at intermediate clinical risk benefit the most from preoperative noninvasive testing, whereas low-risk patients do not need any further preoperative workup, and selected high-risk patients may be referred for coronary angiography if they are appropriate candidates for coronary revascularization. Long-term follow-up has revealed that these clinical markers are also useful for predicting late cardiac complications after vascular surgery.18
Although the clinical index proposed by Eagle and colleagues may provide a cost-effective approach to preoperative risk stratification before vascular surgery, the concordance of this clinical index with severity of coronary disease assessed by coronary angiography has never been fully characterized in an unselected population of patients undergoing elective vascular surgery. Thus, the specific aims of this study were (1) to determine the concordance of clinical risk assessed by a previously developed preoperative clinical index, with the severity of coronary stenosis as determined by angiography and (2) to derive and validate a clinical prediction rule to exclude the presence of “severe” multivessel disease or “critical” three-vessel and/or left main disease.
Based on the experience at the Cleveland Clinic from 1969 through 1978, during which period myocardial infarction accounted for 45% of operative deaths after resection of intact aortic aneurysms and 67% of those after elective aortic replacement for lower-extremity ischemia, Hertzer and colleagues17 19 20 surmised that coronary angiography was an important extension of the preoperative evaluation for a majority of patients scheduled for major vascular surgery. Although “routine” coronary angiography was not described as the permanent objective of this strategy in their original work, the investigators believed that it was the most reliable method of identifying particularly high-risk patients before vascular surgery. Furthermore, their series of 1000 patients predated the availability of pharmacological stress testing with thallium or echocardiographic imaging, and the only two choices for cardiac assessment at the time were stress ECG, with its recognized limitations, and coronary angiography.
The present study is based on data collected prospectively on the original cohort of patients, the study methods for which have been published previously.17 Of the original cohort of 1000 patients, 878 underwent coronary angiography within 6 months before elective vascular surgery, and these patients form the basis of our study. For the purpose of deriving a clinical prediction rule for excluding severe multivessel and critical three-vessel and/or left main disease, we identified a consecutive series of 439 patients, designating them the training (derivation) set, and the remaining 439 were designated the validation set.
Preoperative clinical data were recorded prospectively from patients' histories and medical records concurrent with hospitalization. All data points were collected and recorded before the hypothesis was generated by several of the investigators (S.D.P., K.A.E., K.M.K.), who were thus blinded to the preoperative clinical features of the patients as well as the severity of coronary disease as determined by coronary angiography. Investigators at the Cleveland Clinic Foundation (J.R.Y. and N.R.H.) were then contacted with a proposal to work together in subjecting their original database to further analysis.
Severity of coronary stenosis was based on visual qualitative estimation and conformed to the cardiac catheterization laboratory report generated for each patient. For the purpose of this study, we defined severe stenoses on coronary angiography as three-vessel disease (right coronary artery lesion ≥50%, left circumflex lesion ≥50%, and a left anterior descending lesion ≥50%); two-vessel disease (left anterior descending lesion ≥70% with right coronary artery lesion ≥50% or left anterior descending lesion ≥70% with left circumflex lesion ≥50%); or left main disease (stenosis ≥50%). We defined critical stenoses on coronary angiography as the presence of three-vessel disease with ≥70% stenosis in each vessel and/or left main disease with ≥70% stenosis. This is a new definition not previously used in prior publications using this database.
The preoperative clinical index consists of the following clinical markers13 : age >70 years, history of diabetes mellitus, history of angina, previous myocardial infarction, and previous congestive heart failure. Patients with none of these markers are classified as low risk, those with one or two markers are classified as intermediate risk, and those with more than two markers are high risk.
The proportions of patients with or without severe or critical coronary stenosis in each clinical risk stratum were compared by χ2 or Fisher's exact test when appropriate. Values of P≤.05 were considered significant, with a two-tailed value of α=.05.
The significance of the association between clinical markers composing the preoperative clinical index and either severe or critical coronary disease was determined by Fisher's exact test. In the training set of patients, two separate models were developed that used multiple logistic regression with forward stepwise selection to determine the independent predictors of severe and critical coronary stenosis, respectively (BMDP Statistical Software). Findings were considered significant at a two-sided value of P<.05 (two-tailed test). These independent predictors were then used to develop two prediction rules: (1) to exclude the presence of severe multivessel disease and (2) to exclude the presence of critical three-vessel and/or left main disease. After the clinical prediction rules were derived in the first 439 patients, they were tested in the validation set of 439 patients. Positive and negative predictive values were determined.
Our study cohort included 878 patients; 70% were men. The majority of patients (455, or 52%) underwent aortic surgery, 265 (30%) underwent carotid surgery, 128 (15%) underwent peripheral surgery, and 30 (3%) underwent other vascular procedures. Nearly one third of the patients in the study cohort had severe multivessel disease, 14% had underlying critical three-vessel disease and/or left main disease, and 3% had left main stenosis ≥70%. The distribution of clinical characteristics based on the degree of coronary stenoses is displayed in Table 1⇓. There were no significant differences in the proportion of elderly patients or women among patients with severe, critical, or left main stenoses. Similarly, the distribution of other clinical markers composing the preoperative clinical index did not differ according to the degree of coronary stenoses.
Concordance of Preoperative Clinical Index With Coronary Stenosis
There were 352 patients (40%) classified as low risk by the preoperative clinical index. Of the remaining patients, 456 (52%) were at intermediate risk, and 70 (8%) were at high risk. Fig 1⇓ illustrates the concordance of clinical risk and severity of coronary stenoses. Of the 352 patients at low clinical risk, only 17% had severe multivessel disease, compared with 55 of 70 patients (79%) in the high-clinical-risk group. This difference was highly significant (P<.000001). Similarly, only 17 patients (5%) in the low-risk group had critical three-vessel and/or left main disease, compared with 30 of 70 patients (43%) in the high-clinical-risk group (P<.000001). Thus, we found an increasing gradient of risk for severe multivessel or critical three-vessel and/or left main coronary stenosis with increasing clinical risk. In addition, whereas only 2% of patients at low clinical risk had left main stenosis ≥70%, 10% of the patients at high clinical risk had left main stenoses ≥70% on preoperative coronary angiography (P<.0045).
Table 2⇓ compares the demographics of the derivation and validation sets. There was a greater proportion of elderly patients in the validation set. However, there were no significant differences in the proportion of women or of those with prior angina, prior myocardial infarction, diabetes mellitus, or congestive heart failure between the training and validation sets. The univariate predictors of either severe or critical coronary stenoses included age ≥70 years, prior angina, prior myocardial infarction, history of congestive heart failure, and diabetes mellitus.
Severe Coronary Stenosis
Table 3⇓ displays the multivariate predictors of severe coronary disease, along with their relative risks and 95% CIs. These independent predictors were then used to develop our clinical prediction rule for excluding severe coronary stenoses, as shown in Fig 2⇓. A patient is not likely to have severe multivessel disease if the patient has (1) no history of diabetes mellitus, (2) no prior angina, (3) no previous myocardial infarction, and (4) no history of congestive heart failure. However, if the answer to any of the four questions in the prediction rule is “yes,” then the patient is likely to have severe coronary disease (Fig 2⇓).
This simple algorithm, composed of four easily determined clinical markers, predicted that 278 patients would not have severe three-vessel disease in the derivation set (Table 4⇓). Among the predicted 278 patients, 266 (96%) were classified correctly (positive predictive value, 0.96). In addition, of the 161 patients who were not predicted by this algorithm to be free of severe coronary disease, 133 were actually not free of disease (negative predictive value, 0.83).
When applied to the validation set of patients, this prediction rule performed extremely well in identifying those without severe multivessel disease. Of the 267 patients in the validation set who were predicted to have an absence of severe coronary disease, 257 (96%) were actually demonstrated to have an absence of significant stenoses on coronary angiography (positive predictive value, 0.96). Furthermore, of the 172 patients who were not predicted by this algorithm to be free of severe coronary disease, 131 were actually not free of disease (negative predictive value, 0.76).
Critical Coronary Stenosis
Table 3⇑ also displays the multivariate predictors of critical coronary stenosis, along with their relative risks and 95% CIs. These independent predictors were used to develop our clinical prediction rule for excluding critical coronary stenosis (Fig 2⇑). A patient is not likely to have critical multivessel disease if the patient has (1) no prior angina, (2) no previous myocardial infarction, and (3) no history of congestive heart failure. However, if the answer to any of the three questions in the prediction rule is yes, then the prediction rule for severe coronary disease should be applied (Fig 2⇑), with the understanding that the exact severity of the stenosis is impossible to determine accurately without performing coronary angiography.
When this algorithm was applied to the derivation set of 439 patients, it predicted that 280 patients would not have critical coronary stenosis (Table 5⇓). Of the 280 predicted patients, 264 (94%) did not have critical coronary disease (positive predictive value, 0.94). In addition, of the 159 patients in whom the severity of stenosis could not be predicted, 52 had critical coronary stenosis (negative predictive value, 0.33).
This prediction rule was then tested in the validation set of patients. In this second group of patients, 267 were predicted to be free of critical coronary stenosis, of whom 251 (94%) were correctly classified (positive predictive value, 0.94). In addition, of the 172 patients in whom the severity of stenosis could not be predicted, 35 had critical coronary stenosis (negative predictive value, 0.22). The low negative predictive value for the prediction rule for excluding critical coronary stenosis is of minor importance, since patients classified into a group in which the severity of stenosis cannot be predicted would most likely undergo further preoperative testing.
We and others have previously suggested that preoperative clinical risk assessment is important for predicting both early and late postoperative cardiac complications in patients undergoing noncardiac surgery.12 13 14 18 21 This study provides an angiographic validation for the value of preoperative clinical risk stratification in patients undergoing vascular surgery. In a consecutive series of vascular surgery patients, all of whom underwent coronary angiography without selection bias, we have demonstrated for the first time that an increasing gradient in the number of clinical risk factors is associated with an increasing gradient in the risk for severe or critical coronary stenosis. Thus, this model emphasizes application of the Bayesian perspective to preoperative risk assessment. That is, the low-clinical-risk group also has a low pretest probability of severe or critical coronary disease. Conversely, the intermediate- and high-clinical-risk groups have a far higher likelihood of severe or critical coronary disease. We have argued that noninvasive testing is particularly useful for the intermediate-risk group, in which anatomy is most unpredictable. Finally, these data add to the contention that a very-high-clinical-risk group of patients exists, with such a high probability of severe or critical coronary disease that a negative stress test may raise concern about a false-negative study rather than providing confidence that advanced coronary disease is absent.
We also derived and validated a clinical prediction rule for excluding the presence of severe or critical coronary stenosis. On the basis of the absence of simple clinical markers, we were able to exclude the presence of severe or critical coronary disease with a high degree of accuracy (positive predictive values, 94% to 96%) in a large proportion of patients. This provides further evidence that critical and potentially life-threatening coronary disease is very unlikely in patients harboring none of the aforementioned clinical markers. According to this simple clinical prediction rule, low-clinical-risk patients can be easily identified with a great deal of confidence and can be allowed to proceed to surgery without preoperative noninvasive or invasive testing, particularly those patients with excellent functional capacity. Thus, when used as an initial screening test, this clinical rule can limit the number of patients who need to undergo expensive tests for further cardiac risk stratification.
This study looked at clinical markers found to be physiological correlates of postoperative and late cardiac ischemic complications.12 13 18 We used these markers to derive and validate a simple clinical rule that can help the clinician to make decisions regarding the need for coronary evaluation before vascular surgery. Since several large studies5 17 20 21 22 have shown that preoperative coronary revascularization reduces the long-term risk of cardiac complications after vascular surgery, the decision to proceed with coronary angiography is an important one to make before preoperative cardiac clearance of patients.
However, since the risk of perioperative cardiac complications after coronary artery bypass graft surgery is higher in patients with peripheral vascular disease than in those without it,5 20 23 the use of prophylactic coronary revascularization for improving long-term prognosis may be justified only for critical coronary stenosis. The prediction rule developed and validated in this study can exclude patients with critical coronary stenosis with a very high degree of certainty. These patients do not need any further preoperative testing. Only those who are classified into the group in which coronary stenosis is unpredictable (by the prediction rule for critical stenosis) should be considered for further risk assessment. Such an approach could result in a substantial reduction in healthcare costs by eliminating the costs of noninvasive screening and subsequent invasive diagnostic or therapeutic procedures in low-risk individuals.
This study supplants our earlier recommendations regarding noninvasive cardiac assessment for patients with any one or two of the five previously described preoperative risk factors, because we have now demonstrated that age ≥70 years alone or diabetes mellitus alone is a less compelling reason for pursuing a preoperative cardiac workup. It appears that if the patient has never manifested coronary heart disease, in the form of either heart failure, a history or an ECG suggestive of a previous myocardial infarction, or a history consistent with angina or atypical angina, the likelihood of critical coronary disease (and, we suspect, also the likelihood of a catastrophic perioperative event) is extremely low. For patients with all of these markers, consideration of coronary angiography outright would seem appropriate in selected instances. For patients with one or two factors, the decision either to proceed with stress testing or not depends on the age of the patient, his or her functional status, and the relative risk of the operation being considered. In a standard preoperative assessment, additional variables, including features of the preoperative ECG,4 5 creatinine levels,26 cardiomegaly on the chest radiograph,5 etc, may be predictive of poor outcome, as detailed in other reports.
Our results apply only to patients undergoing elective vascular surgery and not to those with an emergent need for surgery. In addition, the severity of stenosis on coronary angiography was based on visual estimation. Our model also excludes left ventricular ejection fraction. However, since congestive heart failure is a powerful predictor of severe coronary stenosis, this is probably a good clinical surrogate for impaired left ventricular function. Furthermore, although this was a retrospective analysis, we believe that information bias was minimal, because the hypothesis was developed and tested after the prospective collection of all the clinical and angiographic data.
Presented in part at the 67th Scientific Sessions of the American Heart Association, Dallas, Tex, November 14-17, 1994.
- Received August 16, 1995.
- Revision received March 28, 1996.
- Accepted April 11, 1996.
- Copyright © 1996 by American Heart Association
Jamieson WRE, Janusz MT, Miyagishima RT, Gerein AN. Influence of ischemic heart disease on early and late mortality after surgery for peripheral occlusive vascular disease. Circulation. 1982;66(suppl I):I-92-I-96.
American Society of Anesthesiologists, Inc. New classification of physical status. Anesthesiology. 1963;24:111.
Dripps RD, Lamont A, Eckenhoff JE. The role of anesthesia in surgical mortality. JAMA. 1961;178:261-266.
Eagle KA, Coley CM, Newell JB, Brewster D, Darling R, Strauss H, Guiney T, Boucher C. Combining clinical and thallium data optimizes preoperative assessment of cardiac risk before major vascular surgery. Ann Intern Med. 1989;110:859-866.
Lette J, Waters D, Bernier H, Champagne P, Lassonde J, Picard M, Cerino M, Nattel S, Boucher Y, Heyen F, Dube S. Preoperative and long-term cardiac risk assessment: predictive value of 23 clinical descriptors, 7 multivariate scoring systems, and quantitative dipyridamole imaging in 360 patients. Ann Surg. 1992;216:192-204.
Gerson MC, Hurst JM, Hertzberg VS, Doogan PA, Cochran MB, Lim SP, McCall N, Adolph RJ. Cardiac prognosis in noncardiac geriatric surgery. Ann Intern Med. 1985;103:832-837.
Paul SD, Coley CM, Field TS, Boucher CA, Brewster DC, Cambria RP, Abbott WM, Eagle KA. Predicting long-term cardiac complications after vascular surgery: importance of preoperative clinical features, thallium data and perioperative complications. Circulation. 1991;84(suppl II):II-22. Abstract.
Birkmeyer JD, O'Connor GT, Quinton HB, Ricci MA, Morton JR, Leavitt BJ, Charlesworth DC, Hernandez F, McDaniel MD, for the Northern New England Cardiovascular Disease Study Group. The effect of peripheral vascular disease on in-hospital mortality with coronary artery bypass grafting. Circulation. 1994;90(suppl I):I-95. Abstract.