Heart, Victor F. Froelicher, Jonathan Mayers, fifth edition

240 E X E R C I S E A N D T H E H E A R T TABLE 7–14. Likelihood that signs and symptoms represent an acute coronary syndrome secondary to coronary artery disease (CAD) Feature High likelihood Intermediate likelihood Low likelihood absence of high- or any of the following absence of high-likelihood intermediate-likelihood features features: features and presence but may have: of any of the following History Chest or left arm pain or features: Probable ischemic symptoms in discomfort as chief symptom absence of any of the intermediate Examination reproducing prior documented Chest or left arm pain or likelihood characteristics ECG angina discomfort as chief Recent cocaine use Known history of CAD, symptom Cardiac including MI Chest discomfort reproduced by markers Transient MR, hypotension, Age >70 years palpation diaphoresis, pulmonary edema, Male sex or rales Diabetes mellitus T-wave flattening or inversion in New, or presumably new, Extracardiac vascular leads with dominant R waves transient ST-segment deviation (≥0.05 mV) or T-wave disease Normal ECG inversion (≥0.2 mV) with symptoms Fixed Q waves Normal Elevated cardiac TnI, TnT, or Abnormal ST segments or CK-MB T waves not documented to be new Normal From Braunwald E, Mark DB, Jones RH, et al: Unstable angina: Diagnosis and management. Rockville, MD, Agency for Health Care Policy and Research and the National Heart, Lung, and Blood Institute, US Public Health Service, US Department of Health and Human Services; 1994; AHCPR Publication No. 94-0602. 12-lead ECGs and biochemical cardiac marker the management of USA (see Table 7-14). They measurements to either exclude or confirm the reported a 46% reduction in the ultimate need for diagnosis of MI. Such a practice typically results hospital admission in intermediate-risk patients in a low percentage of admitted patients actually after a median stay of 9.2 hours in the chest pain being confirmed to have an MI. Given the inverse unit. Extension of the use of chest pain units to relationship between the percentage of patients intermediate-risk patients in an effort to reduce with a “rule-out MI evaluation” and the “MI miss inpatient costs is facilitated by making available rate,” the potential cost savings of a chest pain diagnostic testing modalities, such as treadmill unit varies depending on the practice pattern for testing and stress imaging (ECHO or NUC), 7 days the disposition of chest pain patients at individual a week.121 hospitals. Hospitals with a high admission rate of low-risk patients to “rule-out MI” (70% to 80%) Triage of Patients will experience the largest cost savings by imple- menting a chest pain unit approach but will have Patients with chest discomfort for whom a spe- the smallest impact on the number of missed MI cific diagnosis cannot be made after a review of patients. In contrast, hospitals with relatively low the history, physical examination, initial 12-lead admission rates of such patients (30% to 40%) ECG, and biochemical cardiac marker data should will experience greater improvements in the quality undergo a more definitive evaluation. Several cat- of care because fewer MI patients will be missed egories of patients should be considered according but will have a smaller impact on costs because of to the algorithm shown in Figure 7-18: the low baseline admission rate. • Patients with possible ACS (B3) are those Potential Expansion of the Use of Chest Pain who had a recent episode of chest discomfort Units for Intermediate-Risk Patients at rest, not entirely typical of ischemia, but are pain free when initially evaluated, have Farkouh et al120 extended the use of a chest pain a normal or unchanged ECG, and have no unit in a separate portion of the ED to include elevations of cardiac markers. patients at an intermediate risk of adverse clinical outcome based on the previously published Agency • Patients with a recent episode of typical for Health Care Policy and Research guidelines for ischemic discomfort that either is of new onset or severe or exhibits an accelerating

C H A P T E R 7 Diagnostic Application of Exercise Testing 241 TABLE 7–15. Short-term risk of death or nonfatal myocardial infarction (MI) in patients with unstable angina (ua)* Feature High risk at least one of the Intermediate risk no high-risk Low risk no high- or History following features must be feature but must have one of intermediate-risk feature but may have Character present: the following features: any of the following features: of pain Prior MI, peripheral or Accelerating tempo of cerebrovascular disease, or New-onset or progressive CCS Clinical ischemic symptoms in CABG, prior aspirin use Class III or IV angina in the past findings preceding 48 hr Prolonged (>20 min) rest angina, 2 weeks without prolonged (>20 min) Prolonged ongoing now resolved, with moderate or ECG (>20 min) rest pain high likelihood of CAD rest pain but with moderate Rest angina (<20 min) or or high likelihood of CAD (see Cardiac Pulmonary edema, most relieved with rest or sublingual Table 7-14) markers likely due to ischemia NTG Age >70 years Normal or unchanged ECG New or worsening MR during an episode of chest murmur T-wave inversions >0.2 mV discomfort S3 or new/worsening rales Pathological Q waves Hypotension, bradycardia, Normal tachycardia Slightly elevated (e.g., TnT Age >75 years >0.01 but <0.1 ng/mL) Angina at rest with transient ST-segment changes >0.05 mV Bundle branch block, new or presumed new Sustained ventricular tachycardia Elevated (e.g., TnT or TnI >0.1 ng/mL) *Estimation of the short-term risks of death and nonfatal cardiac ischemic events in UA is a complex multivariable problem that cannot be fully specified in a table such as this; therefore, this table is meant to offer general guidance and illustration rather than rigid algorithms. CABG, coronary artery bypass graft; CAD, coronary artery disease; CCS, Canadian Cardiovascular Society; NTG, nitroglycerin. From AHCPR Clinical Practice Guideline No. 10, Unstable Angina: Diagnosis and Management, May 1994. Braunwald E, Mark DB, Jones RH, et al: Unstable angina: Diagnosis and management. Rockville, MD, Agency for Health Care Policy and Research and the National Heart, Lung, and Blood Institute, US Public Health Service, US Department of Health and Human Services; 1994; AHCPR Publication No. 94-0602. pattern of previous stable angina (especially positive cardiac markers, or hemodynamic if it has occurred at rest or is within 2 weeks instability) (D2) should be admitted to the of a previously documented MI) should ini- hospital (H3). tially be considered to have a “definite ACS” (B4). However, such patients may be at a low Discharge from ED or Chest risk if their ECG obtained at presentation Pain Unit has no diagnostic abnormalities and the ini- tial serum cardiac markers (especially car- The initial assessment of whether a patient has USA diac-specific troponins) are normal (C2 and or non-ST elevation MI and which triage option D1). As indicated in the algorithm, patients is most suitable generally should be made imme- with either “possible ACS” (B3) or “definite diately on the patient’s arrival at a medical facility. ACS” (B4), but with nondiagnostic ECG and Rapid assessment of a patient’s candidacy for normal initial cardiac markers (D1), are can- additional observation can be accomplished based didates for additional observation in the ED on the status of the symptoms, ECG findings, or in a specialized area such as a chest pain and serum cardiac marker measurements. Patients unit (E1). In contrast, patients who present who experience recurrent ischemic discomfort, without ST-segment elevation but have fea- evolve abnormalities on a follow-up 12-lead tures indicative of active ischemia (ongoing ECG or cardiac marker measurements, or develop pain, ST-segment and/or T-wave changes,

242 E X E R C I S E A N D T H E H E A R T hemodynamic abnormalities, such as new or appointment with their primary care physician worsening congestive heart failure (CHF) (D2), as outpatients for further investigation into the should be admitted to the hospital (H3). cause of their symptoms (I1). They should be seen by a physician within 72 hours of discharge from Patients who are pain free, have either a normal the ED or chest pain unit. or nondiagnostic ECG or one that is unchanged from previous tracings, and have a normal set of Patients with possible ACS (B3) and those with a initial cardiac marker measurements are candi- definite ACS but a nondiagnostic ECG and normal dates for further evaluation to screen for nonis- biochemical cardiac markers when they are ini- chemic discomfort (B1) versus a low-risk ACS (D1). tially seen (D1) at institutions without a chest pain If the patient is low risk and does not experience unit (or equivalent facility), should be admitted to any further ischemic discomfort and a follow-up an inpatient unit. The inpatient unit to which 12-lead ECG and cardiac marker measurements such patients are to be admitted should have the after 6 to 8 hours of observation are normal (F1), same provisions for continuous ECG monitoring, the patient may be considered for an early stress availability of resuscitation equipment, and staffing test to provoke ischemia (G1). This test can be per- arrangements as described earlier for the design formed before the discharge and should be super- of chest pain units. vised by an experienced physician. Alternatively, the patient may be discharged and return for a Studies published since the guidelines include stress test as an outpatient within 3 days. The exact the excellent review from the Davis Group122 and nature of the stress test may vary depending on the the following. At Mayo Clinic a study was performed patient’s ability to exercise on either a treadmill or to assess the outcome of patients discharged with a bicycle and the local expertise in a given hospital diagnosis of chest pain of undetermined origin setting (e.g., availability of different testing modali- and to identify predisposing factors for further car- ties at different times of the day or different days diac events.123 Patient records from 1985 through of the week). Patients who are capable of exercise 1992 were reviewed for the occurrence of adverse and are free of confounding features on the baseline cardiac events and subsequent ED visits for recur- ECG, such as bundle branch block, LVH, or paced rent chest pain within 12 months of discharge. rhythms, can be evaluated with routine symptom- Associations between patient characteristics and limited conventional exercise stress testing. an adverse cardiac event were evaluated univari- Patients who are incapable of exercise or who have ately and summarized by using odds ratios (ORs). an uninterpretable baseline ECG should be consid- Long-term mortality was also determined. Among ered for pharmacological stress testing with either 1973 admitted ED patients with chest pain, 230 nuclear perfusion imaging or two-dimensional were given a diagnosis of chest pain of undeter- ECHO. Because left ventricular function is so inte- mined origin. Ten (4.4%) of 230 patients experi- grally related to prognosis and heavily affects ther- enced an adverse cardiac event. Factors significantly apeutic options, strong consideration should be associated with an adverse cardiac event included given to the assessment of left ventricular function an abnormal ECG on admission (OR 10×), pre-exist- with ECHO or radionuclide ventriculography in ing diabetes mellitus (OR 7×), and pre-existing CAD patients with documented ischemia. In sites at (OR 28×). Thirty-three (14%) patients returned which stress tests are not available, low-risk to the ED within 12 months of discharge; five patients may be discharged and the test scheduled patients were given a diagnosis of a cardiac condi- within 3 days. tion, and five were given a diagnosis of a gastroin- testinal condition. In long-term follow-up, 46 Patients who develop recurrent pain during patients died, with a mean time of 6.1 years from observation or in whom the follow-up studies (12- hospital discharge to death due to any cause and lead ECG, cardiac markers) show new abnormalities an estimated 5-year survival of 91.4%. They found (F2) should be admitted to the hospital (H3). that patients discharged from the hospital with a diagnosis of chest pain of undetermined origin, Because continuity of care is important in the those with an initial abnormal ECG, pre-existing overall management of patients with a chest pain diabetes, or pre-existing CAD are at higher risk of syndrome, the patient’s primary physician (if not a subsequent adverse cardiac event. In the absence involved in the care of the patient during the initial of such factors, cardiac outcome is excellent. episode) should be notified of the results of the evaluation and should receive a copy of the relevant The belief that chest pain relief with nitroglyc- test results. Patients with a noncardiac diagnosis erin indicates the presence of active CAD is com- and those with low risk or possible ACS with a neg- mon and often used in the ED.124 To define the ative stress test should be counseled to make an diagnostic and prognostic value of chest pain relief

C H A P T E R 7 Diagnostic Application of Exercise Testing 243 with nitroglycerin, a prospective observational From the Prince Charles Hospital, Australia cohort study was performed in an urban commu- comes a clinical audit of 630 consecutive patients nity teaching hospital. There were 459 consecu- who presented to the ED in 2001 with chest pain tive patients with chest pain admitted through the and intermediate-risk features.126 They applied ED who received nitroglycerin from emergency the Accelerated Chest Pain Assessment Protocol, services personnel or an ED nurse. Follow-up was as advocated by the “Management of unstable obtained by telephone contact at 4 months. Chest angina guidelines—2000” from the National Heart pain relief was defined as a decrease of at least 50% Foundation and the Cardiac Society of Australia in patients’ self-reported pain within 5 minutes of and New Zealand. Four hundred nine patients the initial dose of sublingual or spray nitroglycerin. (65%) were reclassified as low risk and discharged Active CAD was defined as any elevated serum at a mean of 14 hours after assessment in the enzyme level, coronary angiography demonstrat- chest pain unit. None had missed MIs, while three ing a 70% or greater stenosis, or a positive exer- (1%) had cardiac events by 6 months (all elective cise test result. Nitroglycerin relieved chest pain revascularization procedures). Another 110 patients in 39% of patients (181 of 459). In patients with (17%) were reclassified as high risk, and 21 (19%) of active CAD as the likely cause of their chest pain, these had cardiac events (mainly revascularizations) 35% (49 of 141) had chest pain relief with nitro- by 6 months. Patients who were unable to exercise glycerin. In contrast, in patients without CAD, or had nondiagnostic exercise tests (equivocal risk) 41% (113 of 275) had chest pain relief (P > 0.2). had an intermediate cardiac event rate (8%). Four-month clinical outcomes were similar in patients with or without chest pain relief with The Davis group has described their use of nitroglycerin (P > 0.2). These data suggest that, in immediate exercise testing to evaluate a large, het- a general population admitted for chest pain, erogeneous group of low-risk patients presenting relief of pain after nitroglycerin treatment does with chest pain.127 Patients presenting to the ED not predict active CAD and should not be used to with chest pain compatible with a cardiac origin guide diagnosis. and clinical evidence of low risk on initial assess- ment underwent immediate exercise treadmill test- A Spanish group reported 701 consecutive ing in our chest pain evaluation unit. Indicators of patients evaluated by clinical history (chest pain low clinical risk included no evidence of hemody- score and risk factors), ECG, troponin I, and early namic instability, arrhythmias, or ECG signs of (< 24 hours) exercise testing in low-risk patients ischemia. Serial measurements of cardiac injury (n = 165) in the ED.125 A composite endpoint (recur- markers were not obtained. Exercise testing was rent USA, acute MI, or cardiac death) was recorded performed to a sign- or symptom-limited endpoint during hospital stay or in ambulatory care settings in 1000 patients (520 men, 480 women; age range for patients discharged after early exercise testing 31 to 82 years) and was positive for ischemia in and occurred in 122 patients (17%). Multivariate 13%, negative in 64%, and nondiagnostic in 23% analysis identified the following predictors: chest of patients. There were no adverse effects of exer- pain score equaling 11 points or more (OR = 2×), cise testing, and all patients with a negative exer- age equal to or greater than 68 (OR 2×), insulin- cise test were discharged directly from the ED. At dependent diabetes mellitus (OR 2×), a history of 30-day follow-up there was no mortality in any coronary surgery (OR 3×), ST-segment depression of the three groups. Cardiac events in the three (OR 2×) and troponin I elevation (OR 1.6×). ST-seg- groups included: negative group, 1 non-Q-wave MI; ment depression produced a high endpoint increase positive group, 4 non-Q-wave MIs, and 12 myocar- (31% versus 13%). Troponin I elevation increased dial revascularizations; nondiagnostic group, 7 the risk in the subgroup without ST-segment myocardial revascularizations. depression (20% versus 11%) but did not further modify the risk in the subgroup with ST depression. SUMMARY OF THE DIAGNOSTIC Nevertheless, the negative ECG and troponin UTILIZATION OF EXERCISE I subgroup showed a non-negligible endpoint rate. TESTING Finally, no patient with a negative exercise test presented events compared to 7% of those with a It is appropriate to compare the newer diagnostic non-negative test (RR 2.5×). They concluded that modalities with the standard exercise test, since it ED evaluation of chest pain should not focus on a is a mature, established technology. The equip- single parameter; on the contrary, the clinical his- ment and personnel for performing it are readily tory, ECG, troponin, and early exercise testing available. Exercise testing equipment is relatively must be globally analyzed.

244 E X E R C I S E A N D T H E H E A R T inexpensive so that replacement or updating is study are included because of its excellent design. not a major limitation. The test can be performed To evaluate diagnostic characteristics, patients with in the doctor’s office and does not require injec- a prior MI should be excluded since the diagnosis tions or exposure to radiation. It can be an exten- of coronary disease is not an issue in them. sion of the medical history and physical exam, providing more than simply diagnostic informa- The ACC/AHA Guidelines for the diagnostic tion. Furthermore, it can determine the degree of use of the standard exercise test have stated that disability and impairment to quality of life as well it is appropriate for testing of adult male or female as be the first step in rehabilitation and altering a patients (including those with complete right major risk factor (physical inactivity). bundle branch block or with <1mm of resting ST depression) with an intermediate pretest proba- Some of the newer add-ons or substitutes for bility of CAD based on gender, age, and symptoms. the exercise test have the advantage of being able Table 7-13 indicates which patients were in this to localize ischemia as well as diagnose coronary probability level. disease when the baseline ECG negates ST analy- sis (1mm ST depression, LBBB, Wolff-Parkinson- Feinstein62,63 has promoted criteria for “method- White syndrome). In addition, nonexercise stress ological standards” for diagnostic tests. Their pur- techniques permit diagnostic assessment of pose was to improve patient care, reduce healthcare patients unable to exercise. Although the newer costs, improve the quality of diagnostic test infor- technologies appear to have better diagnostic mation, and to eliminate useless tests or testing characteristics, this is not always the case, partic- methodologies. The two most important criteria ularly when more than the ST segments from the to consider when evaluating such studies are exercise test are used in scores. limited challenge and work-up bias. Limited chal- lenge usually results in exaggerated values for sen- Test evaluation has been advanced and so we sitivity, specificity, predictive accuracy, and ROC are now in a better position to evaluate studies of curve area. Work-up bias results in shifting cut- test characteristics. A number of researchers have point performance further along the ROC curve, applied these guidelines along with meta-analysis and when removed shows that the exercise test has to come to consensus on the diagnostic charac- a high specificity in office practice. The eponyms teristics of the available tests for angiographically SnNout and SpPin help to remember the perfor- significant CAD. Table 7-11 presents some of the mance of a test with high values of either sensitivity results from meta-analysis and from multicenter or specificity. When a test has a very high sensitiv- studies. Since sensitivity and specificity are ity a Negative test rules out the diagnosis (SnNout); inversely related and altered by the chosen cut- when a test has a very high specificity, a Positive point for normal/abnormal, the predictive accu- test rules in the diagnosis (SpPin). The ACP Journal racy (percentage of patients correctly classified as Club has published an excellent roadmap for sys- normal and abnormal) is a convenient way to tematic reviews of diagnostic test evaluations.128 compare tests. For instance, while the sensitivity and specificity for ST-segment depression during In studies that took into account the number exercise testing and EBCT are nearly opposite, the of coronary arteries involved, all found increasing predictive accuracies of the tests are similar. This sensitivity of the test as more vessels were involved. means that altering their cutpoints (i.e., lowering The most false negatives have been found among the amount of ST-segment depression or raising patients with single-vessel disease, particularly the calcium score) would result in similar sensi- if the diseased vessel was not the left anterior tivities and specificities. Because predictive accu- descending artery. No matter what techniques are racy can be thought of as the number of used, there is a reciprocal relationship between individuals correctly classified out of 100 tested, sensitivity and specificity. The more specific a test simply subtracting predictive accuracies provides is (i.e., the more able it is to determine who is dis- an estimate of how many more patients are clas- ease-free), the less sensitive it is, and vice versa. sified by substituting one test for another. The values for adjusting the criterion can alter Predictive accuracy is affected by disease preva- sensitivity and specificity for the cutpoint used for lence, so comparisons are only valid in popula- abnormal. For instance, when the criterion for an tions with the same disease prevalence. abnormal exercise-induced ST-segment response is altered to 0.2 mV depression, making it more Although the nonexercise stress tests are very specific for CAD, the sensitivity of the test will be useful, the results shown below are probably better reduced by half. than their actual performance because of patient selection. The results of the CKG multicenter For patients subgrouped according to beta- blocker administration as initiated by their

C H A P T E R 7 Diagnostic Application of Exercise Testing 245 referring physician, no differences in test per- and the combination of leads V5 with II, because formance were found in a consecutive group of lead II had such a high false-positive rate. In males being evaluated for possible CAD. However, patients without prior MI and normal resting ECGs, the only way to maintain sensitivity with the ST depression in precordial lead V5 along with V4 standard exercise test in the beta-blocker group and V6 are reliable markers for CAD, and the who failed to reach target heart rate was to use a monitoring of inferior limb leads adds little addi- treadmill score or 0.5-mm ST depression as the tional diagnostic information, but ST elevation criterion for abnormal. Thus, in our most recent in these leads should not be ignored. In patients study of the effects of beta-blockade and heart rate with a normal resting ECG, exercise-induced ST- response, we found the sensitivity and predictive segment depression confined to the inferior leads accuracy of standard ST criteria for exercise- is of little value for the identification of coronary induced ST depression significantly decreased in disease. male patients taking beta-blockers and do not reach target heart rate. In those who fail to reach Once the diagnosis is made, how does this affect target heart rate and are not beta-blocked, sensi- outcomes? In Denmark, an observational study tivity and predictive accuracy were maintained. evaluated the association among different centers' Alhough perhaps optimal, for routine exercise referral practices for coronary angiography after testing it appears unnecessary for physicians to exercise testing, with 1- and 5-year outcomes.130 accept the risk of stopping beta-blockers before All 10 hospitals and six private consultants in testing when a patient is exhibiting possible Aarhus and Ringkjoebing counties (900,000 inhab- symptoms of ischemia. itants) were screened. They found that in 1996, 736 patients with an abnormal bicycle exercise The summary from the guidelines are well test were considered for referral for coronary inter- stated regarding testing women: concern about vention. As an immediate consequence of the exer- false-positive ST responses may be addressed by cise test, 61% of subjects were referred for cardiac careful assessment of posttest probability and catheterization. Centers were defined as exhibiting selective use of stress imaging test before proceed- low (<33%), intermediate (33%–66%) and high ing to angiography. Although the optimal strategy (>66%) referral patterns. A low compared with a for circumventing false-positive test results for the high referral fraction was associated with a simi- diagnosis of coronary disease in women remains to lar 5-year mortality and MI rate. The same was be defined, there is currently insufficient data to found for an intermediate, compared with a high justify routine stress imaging test as the initial fraction, referral center. Estimates were about the test in coronary disease in women. same after 1 year of follow-up with no major dif- ferences among centers in mortality or MI. Studies Studies considering non-ECG data consistently like this with a longer follow-up, including costs demonstrate that the multivariable equations out- and quality of life assessment, would be of great perform simple ST-diagnostic criteria. These equa- import to clinicians. tions generally provide a predictive accuracy of 80% (ROC area of 0.80). To obtain the best diag- Obviously the concept of ACS has altered the nostic characteristics with the exercise test clinical clinical milieu. The CNR Cardiology Research group and non-ECG test responses should be consid- in Italy reviewed the literature to see if evidence ered. Computerized ECG measurements and ECG still supports the use of exercise ECG as first scores are not superior to visual analysis but can choice of stress testing modality for ACS.131 They duplicate the results of expert readers. Multivariate concluded that a large body of evidence supports scores using computers to make the calculations the use of exercise ECG as a cost-effective tool for from logistic regression equations appear to sig- prognostic purposes and for quality of life assess- nificantly improve on test characteristics. We have ment following ACS. This is consistent with the validated our simple scores (Figs. 7-16 and 7-17) ACC/AHA guidelines. for both men and women at other institutions and have compared them to physicians. They should The guidelines state that patients who are be applied during every test along with the DTS pain-free, have either a normal or nondiagnostic score since they are easy to use and significantly ECG or one that is unchanged from previous improve the prediction of angiographic CAD. tracings, and have a normal set of initial cardiac marker measurements are candidates for further Miranda et al129 found exercise-induced ST-seg- evaluation to screen for nonischemic discom- ment depression in inferior limb leads to be a poor fort versus a low-risk ACS. If the patient is low marker for CAD in and of itself. Precordial lead V5 risk, does not experience any further ischemic alone consistently outperformed the inferior lead discomfort, and a follow-up 12-lead ECG and

246 E X E R C I S E A N D T H E H E A R T cardiac marker measurements after 6 to 8 hours 15. Weiner DA, McCabe CH, Ryan TJ: Prognostic assessment of of observation are normal, the patient may be patients with coronary artery disease by exercise testing. Am Heart considered for an early exercise test to provoke J 1983;105:749-755. ischemia. This test can be performed before the discharge and should be supervised by an experi- 16. Weiner DA, McCabe CH, Ryan TJ: Identification of patients with left enced physician. Alternatively, the patient may main and three vessel coronary disease with clinical and exercise test be discharged and return for the test as an outpa- variables. Am J Cardiol 1980;46:21-27. tient within 3 days. 17. Mark DB, Hlatky MA, Lee KL, et al: Localizing coronary artery The most exciting new area in diagnosis is the obstructions with the exercise treadmill test. Ann Intern Med 1987; use of biomarkers such as BNP. The point of con- 106:53-55. tact analysis techniques available for these assays involves a hand-held battery-powered unit that uses 18. Fortuin NJ, Friesinger GC: Exercise-induced ST segment elevation: a replaceable cartridge. Finger stick blood samples Clinical, electrocardiographic and arteriographic studies in twelve are adequate for analyses and the results are avail- patients. Am J Med 1970;49:459-464. able immediately. If validated using appropriate study design (similar to QUEXTA), biomarkers 19. Hegge FN, Tuna N, Burchell HB: Coronary arteriographic findings could have a big impact on physician and patient in patients with axis shifts or ST segment elevations on exercise confidence in the standard office/clinic exercise testing. Am Heart J 1973;86:603-615. test. 20. Chahine RA, Raizner AE, Ishimori T: The clinical significance of REFERENCES exercise-induced ST-segment elevation. Circulation 1976;54: 209-213. 1. Snow V, Barry P, Fihn SD, et al: Evaluation of primary care patients with chronic stable angina: guidelines from the American College 21. Longhurst JC, Kraus WL: Exercise-induced ST elevation in patients of Physicians. Ann Intern Med 2004;141:57-64. without myocardial infarction. Circulation 1979;60:616-629. 2. Rosanski A, Diamond GA, Berman DS, et al: The declining speci- 22. Bruce RA, Fisher LD: Unusual prognostic significance of exercise- ficity of exercise radionuclide ventriculography. N Engl J Med 1983; induced ST elevation in coronary patients. J Electrocardiol 309:518-522. 1987;20(suppl):84-88. 3. Fagan TJ: Nomogram for Bayes theorem. N Engl J Med 1975; 23. De Feyter PJ, Majid PA, Van Eenige MJ, et al: Clinical significance of 293:257. exercise-induced ST segment elevation. Br Heart J 1981;46:84-92. 4. Wilson RF, Marcus ML, Christensen BV, et al: Accuracy of exercise 24. Kentala E, Luurela O: Response of R wave amplitude to posture electrocardiography in detecting physiologically significant coro- changes and to exercise. Ann Clin Res 1975;7:258-263. nary arterial lesions. Circulation 1991;83:412-421. 25. Bonoris PE, Greenberg PS, Christison GW, et al: Evaluation of R 5. Piek JJ, Boersma E, di Mario C, et al: For the DEBATE study wave amplitude changes versus ST segment depression in stress group. Angiographical and Doppler flow-derived parameters testing. Circulation 1978;57:904-910. for assessment of coronary lesion severity and its relation to the result of exercise electrocardiography. Eur Heart J 2000;21: 26. Eenige van MJ, De Feyter PJ, Jong JP, Roos, JP: Diagnostic incapacity 466-474. of exercise-induced QRS wave amplitude changes to detect coro- nary artery disease and left ventricular dysfunction. Eur Heart J 1982; 6. Pellinen TJ, Virtanen KS, Toivonen L, et al: Coronary collateral 3:9-16. circulation. Clin Cardiol 1991;14:111-118. 27. Myers J, Ahnve S, Froelicher V, Sullivan M: Spatial R wave ampli- 7. Blackburn H, Katigbak R: What electrocardiographic leads to take tude during exercise: Relation with left ventricular ischemia and after exercise? Am Heart J 1964;67:184-188. function. J Am Coll Cardiol 1985;6:603-608. 8. Miller TD, Desser KB, Lawson M: How many electrocardiographic 28. Goldschlager N, Selzer A, Cohn K: Treadmill stress tests as indica- leads are required for exercise treadmill tests? J Electrocardiol tors of presence and severity of coronary artery disease. Ann Intern 1987;20:131-137. Med 1976;85:277-286. 9. Mason RE, Likar I, Biern RO, Ross RS: Multiple-lead exercise electro- 29. Callaham PR, Thomas L, Ellestad MH: Prolonged ST-segment cardiography. Experience in 107 normal subjects and 67 patients depression following exercise predicts significant proximal left with angina pectoris, and comparison with coronary cinearteriog- coronary artery stenosis (abstract). Circulation 1987;76(suppl raphy in 84 patients. Circulation 1967;36:517-525. IV):IV-253. 10. Sketch MH, Nair CK, Esterbrooks DJ, Mohiuddin SM: Reliability of 30. Meyers DG, Bendon KA, Hankins JH, Stratbucker RA: The effect of single-lead and multiple-lead electrocardiography during and after baseline electrocardiographic abnormalities on the diagnostic exercise. Chest 1978;74:394-401. accuracy of exercise-induced ST-segment changes. Am Heart J 1990;119:272-276. 11. Simoons ML, Block P: Toward the optimal lead system and optimal criteria for exercise electrocardiography. Am J Cardiol 1981;47: 31. Tonkon MJ, Lee G, DeMaria AN, et al: Effects of digitalis on the 1366-1374. exercise electrocardiogram in normal adult subjects. Chest 1977;72:714-718. 12. Miranda CP, Liu J, Kadar A, Janosi A, et al: Usefulness of exercise- induced ST-segment depression in the inferior leads during exer- 32. Sketch MH, Moss AN, Butler ML, et al: Digoxin-induced positive cise testing as a marker for coronary artery disease. Am J Cardiol exercise tests: Their clinical and prognostic significance. Am J 1992;69:303-308. Cardiol 1981;48:655-659. 13. Viik J, Lehtinen R, Turjanmaa V, et al: Correct utilization of exer- 33. Sundqvist K, Atterhog JH, Jogestrand T: Effect of digoxin on the cise electrocardiographic leads in differentiation of men with coro- electrocardiogram at rest and during exercise in healthy subjects. nary artery disease from patients with a low likelihood of coronary Am J Cardiol 1986;57:661-665. artery disease using peak exercise ST-segment depression. Am J Cardiol 1998;81:964-969. 34. LeWinter M, Crawford M, O’Rourke R, Karliner J: The effects of oral propanolol, digoxin and combined therapy on the resting and exer- 14. Riff DP, Carleton RA: Effect of exercise on the atrial recovery wave. cise ECG. Am Heart J 1977;93:202-209. Am Heart J 1971;81:759-763. 35. Herbert WG, Dubach P, Lehmann KG, Froelicher VF: Effect of β- blockade on the interpretation of the exercise ECG: ST level versus delta ST/HR index. Am Heart J 1991;122:993-1000. 36. Gauri AJ, Raxwal VK, Roux L: Effects of chronotropic incompe- tence and beta-blocker use on the exercise treadmill test in men. Am Heart J 2001;142:136-141. 37. Cantwell JD, Murray PM, Thomas RJ: Current management of severe exercise-related cardiac events. Chest 1988;93:1264-1269. 38. Anastasiou-Nana MI, Anderson JL, Stewart JR: Occurrence of exercise- induced and spontaneous wide complex tachycardia during therapy with flecainide for complex ventricular arrhythmias: A probable proarrhythmic effect. Am Heart J 1987;113:1071-1077. 39. Whinnery JE, Froelicher VF, Stuart AJ: The electrocardiographic response to maximal treadmill exercise in asymptomatic men with left bundle branch block. Am Heart J 1997;94:316.

C H A P T E R 7 Diagnostic Application of Exercise Testing 247 40. Ibrahim NS, Selvester RS, Hagar JM, Ellestad MH: Detecting exercise- 66. Braitman LE; Davidoff F: Predicting clinical states in individual induced ischemia in left bundle branch block using the electrocar- patients. Ann Intern Med 1996;125:406-412. diogram. Am J Cardiol 1998;82:832-835. 67. Yamada H, Do D, Morise A, Froelicher V: Review of studies utilizing 41. Richter WS, Aurisch R, Munz DL: Septal myocardial perfusion in multi-variable analysis of clinical and exercise test data to predict complete left bundle branch block: Case report and review of the angiographic coronary artery disease. Prog Cardiovasc Dis 1997; literature. Nuklearmedizin 1998;37:146-150. 39:457-481. 42. Grady TA, Chiu AC, Snader CE, et al: Prognostic significance of 68. Concato J, Feinstein AR, Holford TR: The risk of determining risk exercise-induced left bundle-branch block. JAMA 1998;279:153-156. with multivariate models. Ann Intern Med 1993;118:201-210. 43. Munt B, Huckell VF, Boone J: Exercise-induced left bundle branch 69. Peterson MC, Holbrock JH, Hales DV, et al: Contributions of the block: A case report of false positive MIBI imaging and review of the history, physical examination, and laboratory investigation in mak- literature. Can J Cardiol 1997;13:517-521. ing medical diagnoses. West J Med 1992;156:163-165. 44. Whinnery JE, Froelicher VF, Stuart AJ: The electrocardiographic 70. Diamond GA: Future imperfect: The limitations of clinical predic- response to maximal treadmill exercise in asymptomatic men with tion models and the limits of clinical prediction. J Am Coll Cardiol right branch bundle block. Chest 1977;71:335. 1989;14:12A-22A. 45. Blackburn H: Canadian Colloquium on Computer-Assisted 71. Harrell FE, Jr, Lee KL, Mark DB: Multivariable prognostic models: Interpretation of Electrocardiograms. VI. Importance of the elec- Issues in developing models, evaluating assumptions and adequacy, trocardiogram in populations outside the hospital. Can Med Assoc and measuring and reducing errors. Stat Med 1996;15:361-387. J 1973;108:1262-1265. 72. Morise A, Diamond G: Estimating probability of coronary artery 46. Cullen K, Stenhouse NS, Wearne KL, Compston GN: Electro- disease (letter to the editor). 1993;22:340-341. cardiograms and 13 year cardiovascular mortality in Busselton study. Br Heart J 1982;47:209-212. 73. Morise AP, Bobbio M, Diamond G, et al: The effect of disease preva- lence adjustments on the accuracy of a logistic prediction model. 47. Aronow WS: Correlation of ischemic ST-segment depression on the Med Decis Making 1996;16:133-142. resting electrocardiogram with new cardiac event rates in 1,106 patients over 62 years of age. Am J Cardiol 1989;64:232-233. 74. Morise AP, Detrano R, Bobbio M, Diamond GA: Development and validation of a logistic regression - derived algorithm for esti- 48. Califf RM, Mark DB, Harrell FE, et al: Importance of clinical mea- mating the incremental probability of coronary artery disease sures of ischemia in the prognosis of patients with documented before and after exercise testing. J Am Coll Cardiol 1992;20: coronary artery disease. J Am Coll Cardiol 1988;11:20-26. 1187-1196. 49. Harris PJ, Harrell FE, Lee KL, et al: Survival in medically treated 75. Detrano R, Bobbio M, Olson H, et al: Computer probability esti- coronary artery disease. Circulation 1979;60:1259-1269. mates of angiographic coronary artery disease: Transportability and comparison with cardiologists’s estimates. Comput Biomed Res 50. Miranda CP, Lehmann KG, Froelicher VF: Correlation between 1992;25:468-485. resting ST segment depression, exercise testing, coronary angiog- raphy, and long-term prognosis. Am Heart J 1991;122:1617-1626. 76. Robert AR, Melin JA, Detry JM: Logistic discriminant analysis improves diagnostic accuracy of exercise testing for coronary 51. Kansal S, Roitman D, Sheffield LT: Stress testing with ST-segment artery disease in women. Circulation 1991;83:1202-1209. depression at rest. Circulation 1976;54:636-639. 77. Morise AP, Dalal JN, Duval RD: Value of a simple measure of estro- 52. Harris FJ, DeMaria AN, Lee G, et al: Value and limitations of exer- gen status for improving the diagnosis of CAD in women. Am J Med cise testing in detecting coronary disease with normal and abnor- 1993;94:491-496. mal resting electrocardiograms. Adv Cardiol 1978;22:11-15. 78. Harrell FE Jr, Lee KL, Mark DB: Multivariable prognostic models: 53. Roitman D, Jones WB, Sheffield LT: Comparison of submaximal Issues in developing models, evaluating assumptions and adequacy, exercise ECG test with coronary cineangiocardiogram. Ann Intern and measuring and reducing errors. Stat Med 1996;15:361-387. Med 1970;72:641-647. 79. Froelicher V, Shetler K, Ashley E: Better decisions through science: 54. Roger VL, Jacobsen SJ, Pellikka PA, et al: Gender differences in use exercise testing scores. Prog Cardiovasc Dis 2002;44:395-414. of stress testing and coronary heart disease mortality: A popula- tion-based study in Olmsted County, Minnesota. J Am Coll Cardiol 80. Swets JA, Dawes RM, Monahan J: Better decisions through science. 1998;32:345-352. Sci Am 2000;283:82-87. 55. James FW, Chung EK, (eds). Exercise ECG Test in children. In 81. Ashley E, Myers J, Froelicher V: Exercise testing scores as an exam- Exercise electrocardiography: A Practical Approach, 2nd ed. ple of better decisions through science. Med Sci Sports Exerc Baltimore, Williams & Wilkins, 1983, p. 132. 2002;34:1391-1398. 56. Lauer MS, Pashkow FJ, Snader CE, et al: Sex and diagnostic evalu- 82. Atwood JE, Do D, Froelicher V, et al: Can computerization of the ation of possible coronary artery disease after exercise treadmill exercise test replace the cardiologist? Am Heart J 1998;136: testing at one academic teaching center. Am Heart J 1997;134 543-552. (5 Pt 1):807-813. 83. Morise A: Comparison of the Diamond-Forrester method and a new 57. Lee DP, Fearon WF, Froelicher VF: Clinical utility of the exercise ECG score to estimate the pretest probability of coronary disease before in patients with diabetes and chest pain. Chest 2001;119: 1576-1581. exercise testing. Am Heart J 1999;138:740-745. 58. Lai S, Kaykha A, Yamazaki T, et al: Treadmill scores in elderly men. 84. Poses RM, Cebul RD, Collins M, Fager SS: The importance of dis- J Am Coll Cardiol 2004;43:606-615. ease prevalence in transporting clinical prediction rules. Ann Intern Med 1986;105:586-591. 59. Lavoie KL, Fleet RP, Lesperance F, et al: Are exercise stress tests appropriate for assessing myocardial ischemia in patients with 85. Do D, West JA, Morise A, Froelicher V: A consensus approach to major depressive disorder? Am Heart J 2004;148:621-627. diagnosing coronary artery disease based on clinical and exercise test data. Chest 1997;111:1742-1749. 60. Gianrossi R, Detrano R, Mulvihill D, et al: Exercise-induced ST depression in the diagnosis of coronary artery disease: A meta- 86. Raxwal V, Shetler K, Do D, Froelicher V: A simple treadmill score. analysis. Circulation 1989;80:87-98. Chest 2000;113:1933-1940. 61. Morise A, Diamond GA: Comparison of the sensitivity and specificity 87. Morise AP, Lauer MS, Froelicher VF: Development and validation of of exercise electrocardiography in biased and unbiased populations a simple exercise test score for use in women with symptoms of of men and women. Am Heart J 1995;130:741-747. suspected coronary artery disease. Am Heart J 2002;144:818-825. 62. Philbrick JT, Horwitz RI, Feinstein AR: Methodologic problems of 88. O’Rourke RA, Brundage BH, Froelicher VF, et al: American College exercise testing for coronary artery disease: Groups, analysis and bias. of Cardiology/American Heart Association Expert Consensus Am J Cardiol 1980;46:807-812. Document on electron-beam computed tomography for the diag- nosis and prognosis of coronary artery disease. J Am Coll Cardiol 63. Reid M, Lachs M, Feinstein A: Use of methodological standards in 2000;36:326-340. diagnostic test research. JAMA 1995;274:645-651. 89. Hlatky M, Bovinick E, Brundage B: Diagnostic accuracy of cardiol- 64. Guyatt GH: Readers’ guide for articles evaluating diagnostic tests: ogists compared with probability calculations using Bayes’ rule. What ACP Journal Club does for you and what you must do your- Am J Cardiol 1982;49:192-197. self. ACP Club 1991;115:A16. 90. Lipinski M, Do D, Froelicher V, et al: Comparison of exercise test 65. Ellestad MH, Savitz S, Bergdall D, Teske J: The false positive stress scores and physician estimation in determining disease probability. test. Multivariate analysis of 215 subjects with hemodynamic, Arch Intern Med 2001;161:2239-2244. angiographic and clinical data. Am J Cardiol 1977;40:681-687.

248 E X E R C I S E A N D T H E H E A R T 91. Lipinski M, Froelicher V, Atwood E, et al: Comparison of treadmill 112. Detrano R, Hsiai T, Wang S, et al: Prognostic value of coronary scores with physician estimates of diagnosis and prognosis in calcification and angiographic stenoses in patients undergoing patients with coronary artery disease. Am Heart J 2002;143: coronary angiography. J Am Coll Cardiol 1996;27:285-290. 650-658. 113. Budhoff MJ, Georgiou D, Brody A, et al: Ultrafast computed 92. Froelicher VF, Lehmann KG, Thomas R, et al: The electrocardio- tomography as a diagnostic modality in the detection of coronary graphic exercise test in a population with reduced workup bias: artery disease: A multicenter study. Circulation 1996;93:898-904. Diagnostic performance, computerized interpretation, and multi- variable prediction. Veterans Affairs Cooperative Study in Health 114. Gibbons RJ, Abrams J, Chatterjee K, et al: ACC/AHA 2002 guide- Services #016 (QUEXTA) Study Group. Quantitative Exercise Testing line update for the management of patients with chronic stable and Angiography. Ann Intern Med 1998;128(12 Pt 1):965-974. angina: A report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Committee 93. Ustin J, Umann T, Froelicher V: Data management: A better to Update the 1999 Guidelines for the Management of Patients approach. Physicians and Computers 1994;12:30-33. with Chronic Stable Angina). J Am Coll Cardiol 2003;41:159-168. 94. Froelicher V, Shiu P. Exercise test interpretation system. Physicians 115. Cannon CP, O’Gara PT: Critical pathways for acute coronary syn- and Computers 1996;14:40-44. dromes. In Cannon CP, (ed.): Management of Acute Coronary Syndromes. Totowa, NJ, Humana Press, 1999, pp 611-627. 95. Mark D, Hlatky M, Harrell F, et al: Exercise treadmill score for predicting prognosis in coronary artery disease. Ann Intern Med 116. Graff L, Joseph T, Andelman R, et al: American College of Emergency 1987;106:793-800. Physicians information paper: Chest pain units in emergency departments: A report from the Short-Term Observation Services 96. Reid M, Lachs M, Feinstein A: Use of methodological standards in Section. Am J Cardiol 1995;76:1036-1039. diagnostic test research. JAMA 1995;274:645-651. 117. Brillman J, Mathers-Dunbar L, Graff L, et al: Management of 97. Fletcher G, Balady G, Froelicher V, et al: AHA Medical/Scientific observation units. Ann Emerg Med 1995;25:823-830. Statement. Exercise Standards. Circulation 1995;91:580-615. 118. Graff LG, Dallara J, Ross MA, et al: Impact on the care of the emer- 98. Detrano R, Leatherman J, Salcedo EE, et al: Bayesian analysis ver- gency department chest pain patient from the Chest Pain Evaluation sus discriminant function analysis: Their relative utility in the Registry (CHEPER) study. Am J Cardiol 1997;80:563-568. diagnosis of coronary disease. Circulation 1986;73:970-977. 119. Gomez MA, Anderson JL, Karagounis LA, et al: An emergency 99. Morise AP, Duval RD: Comparison of three Bayesian methods to department-based protocol for rapidly ruling out myocardial estimate posttest probability in patients undergoing exercise ischemia reduces hospital time and expense: Results of a random- stress testing. Am J Cardiol 1989;64:1117-1122. ized study (ROMIO). J Am Coll Cardiol 1996;28:25-33. 100. Morise AP, Duval R, Detrano R, et al: Comparison of logistic regres- 120. Farkouh ME, Smars PA, Reeder GS, et al: Chest Pain Evaluation sion and Bayesian based algorithms to estimate posttest probability in the Emergency Room (CHEER) Investigators. A clinical trial of in patients with suspected CAD undergoing exercise ECG. a chest-pain observation unit for patients with unstable angina. Electrocardiol 1992;25:89-99. N Engl J Med 1998;339:1882-1888. 101. Shaw LJ, Peterson ED, Shaw LK, et al: Use of a prognostic tread- 121. Newby LK, Mark DB: The chest-pain unit-ready for prime time mill score in identifying diagnostic coronary disease subgroups. (editorial)? N Engl J Med 1998;339:1930-1932. Circulation 1998;98:1622-1630. 122. Amsterdam EA, Kirk JD, Diercks DB, et al: Early exercise testing 102. Fleischmann KE, Hunink MG, Kuntz KM, Douglas PS: Exercise in the management of low risk patients in chest pain centers. Prog echocardiography or exercise SPECT imaging? A meta-analysis of Cardiovasc Dis 2004;46:438-452. diagnostic test performance. JAMA 1998;280(10):913-920. 123. Prina LD, Decker WW, Weaver AL, et al: Outcome of patients with 103. Detrano R, Janosi A, Marcondes G, et al: Factors affecting sensi- a final diagnosis of chest pain of undetermined origin admitted tivity and specificity of a diagnostic test: The exercise thallium under the suspicion of acute coronary syndrome: A report from scintigram. Am J Med 1988;84:699-710. the Rochester Epidemiology Project. Ann Emerg Med 2004;43: 59-67. 104. Weiner DA: Accuracy of cardiokymography during exercise testing: Results of a multicenter study. J Am Coll Cardiol 1985;6: 124. Henrikson CA, Howell EE, Bush DE, et al: Chest pain relief by 502-509. nitroglycerin does not predict active coronary artery disease. Ann Intern Med 2003;139:979-986. 105. Gehring J, Koenig W, Donner M, et al: The diagnostic value of signal-averaged stress cardiokymography compared with exercise 125. Sanchis J, Bodi V, Llacer A, et al: Predictors of short-term out- electrocardiography. J Noninvasive Cardiol 1998;5:32-41. come in acute chest pain without ST-segment elevation. Int J Cardiol 2003;92:193-199. 106. Akdemir I, Aksoy N, Aksoy M, et al: Does exercise-induced severe ischaemia result in elevation of plasma troponin-T level in 126. Aroney CN, Dunlevie HL, Bett JH: Use of an accelerated chest pain patients with chronic coronary artery disease? Acta Cardiol 2002; assessment protocol in patients at intermediate risk of adverse 57:13-18. cardiac events. Med J Aust 2003;178:370-374. 107. Ashmaig ME, Starkey BJ, Ziada AM, et al: Changes in serum con- 127. Amsterdam EA, Kirk JD, Diercks DB, et al: Immediate exercise centrations of markers of myocardial injury following treadmill testing to evaluate low-risk patients presenting to the emergency exercise testing in patients with suspected ischaemic heart dis- department with chest pain. J Am Coll Cardiol 2002;40:251-256. ease. Med Sci Monit 2001;7:54-57. 128. Pai M, McCulloch M, Enanoria W, Colford JM, Jr: Systematic 108. Foote RS, Pearlman JD, Siegel AH, Yeo KT: Detection of exercise- reviews of diagnostic test evaluations: What’s behind the scenes? induced ischemia by changes in B-type natriuretic peptides. J Am ACP J Club 2004;141:A11-13. Coll Cardiol 2004;44:1980-1987. 129. Miranda CP, Liu J, Kadar A, et al: Usefulness of exercise-induced 109. Sabatine MS, Morrow DA, de Lemos JA, et al: TIMI Study Group. ST-segment depression in the inferior leads during exercise test- Acute changes in circulating natriuretic peptide levels in relation ing as a marker for coronary artery disease. Am J Cardiol to myocardial ischemia. J Am Coll Cardiol 2004;44:1988-1995. 1992;69:303-308. 110. Shemesh J, Apter S, Rozenman J, et al: Calcification of coronary 130. Niemann T, Labouriau R, Sorensen HT, et al: Prognostic impact of arteries: Detection and quantification with double-helix CT. different regional referral practices for interventional investiga- Radiology 1995;197:779-783. tion and coronary treatment after exercise testing: A population- based 5-year follow-up study. J Intern Med 2004;255:478-485. 111. Kajinami K, Seki H, Takekoshi N, Mabuchi H. Noninvasive predic- tion of coronary atherosclerosis by quantification of coronary 131. Bigi R, Cortigiani L, Desideri A: Exercise electrocardiography artery calcification using electron beam computed tomography: after acute coronary syndromes: still the first testing modality? Comparison with electrocardiographic and thallium exercise Clin Cardiol 2003;26:390-395. stress test results. J Am Coll Cardiol 1995;26:1209-1221.

CHAPTER eight Prognostic Applications of Exercise Testing RATIONALE In addition to diagnostic information, the test gives practical and clinically valuable information There are two principal reasons for estimating prog- regarding exercise capacity and response to ther- nosis. The first is to provide accurate answers to apy. Patients with clinical data and exercise test patient’s questions regarding the probable outcome responses considered abnormal or associated with of their illness. Although discussion of prognosis is a high enough probability for cardiac events or for inherently delicate, and probability statements can having severe coronary artery disease (CAD) are fre- be misunderstood, most patients find this informa- quently evaluated further by coronary angiography. tion useful in planning their affairs regarding work, A study evaluating the appropriateness of the per- recreational activities, personal estate, and finances. formance of coronary angiography in clinical The second reason to determine prognosis is to practice considered angiography to be inappropri- identify those patients for whom interventions ate nearly a quarter of the time due to the failure might improve outcome. to obtain an exercise test.1 However, the indications for angiography have dramatically changed since Although improved prognosis equates with the eighties. Now, with drug-eluting stents, the risk increased quantity of life, quality of life issues must of reocclusion has been reduced to single digits. also be taken into account. In that regard, it is The possibility of putting patients at higher risk by apparent that in certain clinical settings, catheter or performing an intervention has drastically dropped. surgical interventions provide better therapy than Fewer patients with (CAD) are experiencing angina medication. However, these interventions, when pectoris than ever before, thanks to progress in misapplied, can have a negative impact on the interventional cardiology. quality of life (inconvenience, complications, and discomfort), as well as creating a financial burden As will be reviewed, numerous investigators to the individual and to society. have indicated that responses to exercise testing predict the severity of underlying coronary disease Part of the Basic Patient and prognosis. However, exercise testing cannot Evaluation predict angiographic findings or a poor prognosis with absolute certainty. In addition, only certain Patients with known or suspected coronary dis- groups of patients with specific CAD patterns are ease are usually evaluated initially after a careful conferred a survival benefit from coronary artery cardiac history and physical exam with an exercise bypass surgery (CABS).2 Because of the lack of cer- test. It can be performed safely and inexpensively tainty and the need for a “road map” for coronary and even accomplished in the physician’s office. intervention, coronary angiography is considered the “gold standard” for evaluating patients for the 249

250 E X E R C I S E A N D T H E H E A R T presence of coronary disease, and for determining grafting [CABG]) and outcome improved in which patients might benefit from interventional subsets. therapy. Angiography defines static anatomy, because it is performed at rest, but the addition of What exercise test responses are due to flow wires into the coronary arteries can quantitate myocardial ischemia or dysfunction? The exer- coronary blood flow.3 Techniques are even being cise responses due to ischemia include angina, explored to identify subcritical lesions likely to rup- ST-segment depression, and ST-segment eleva- ture or gather thrombus (i.e., labile plaque). This tion over ECG areas without Q waves. Predicting chapter begins with the pathophysiologic basis of the amount of ischemia (i.e., the amount of the exercise test responses and then follows with myocardium in jeopardy) is difficult. It appears to a discussion of the pertinent studies. be inversely related to the double product at the onset of signs/symptoms of ischemia. Pathophysiology The only response specifically associated with The basic pathophysiologic features of CAD LV dysfunction is ST elevation over Q waves. This that determine prognosis include the amount of carries an increased risk in patients with Q waves remaining myocardium (reflected by left ventric- and indicates that they have lower LV function ular [LV] function), the amount of myocardium in and possibly larger aneurysms as compared to jeopardy, and arrhythmic risk. Arrhythmias usu- those with Q waves without elevation.4 If ST ele- ally are the final event, whether primary or sec- vation decreases with subsequent exercise tests, ondary to ischemia or LV dysfunction. Arrhythmic such patients are thought to have a better progno- risk has become a very active area for research sis. Those with elevation over Q waves have poorer because of the data showing a marked improvement resting LV function than those without elevation.5 in survival in select patient groups with implanta- tion of defibrillators (ICD). Exercise test-induced The responses resulting from either ischemia or ventricular arrhythmias (ETIVA) indicate electri- LV dysfunction include chronotropic incompetence cal instability not necessarily related to exercise- or heart rate impairment,6 systolic blood pressure induced ischemia (except for the rare ST elevation (SBP) drops,7 and a poor exercise capacity.8 Their in a normal ECG that is very arrhythmogenic). combined association with ischemia and myocar- Although ETIVA can be due to LV dysfunction, most dial damage or dysfunction explains why they are likely in others it is due to a genetic propensity that so important in predicting prognosis. has yet to be characterized. Recent studies have highlighted the independent predictive power of Previous studies have shown that exercise capac- exercise test-induced arrhythmias. These findings ity correlates weakly with LV function in patients have been confusing, because they appear to be without signs or symptoms of right-sided failure.9 associated with long rather than short-term events, Exercise testing is not very helpful in identifying are relatively weak to other test responses, and patients with moderate LV dysfunction, which is a therapy is not readily available. Since primary part of the requirement for improved survival with arrhythmic deaths are relatively rare and ETIVA surgery. This is better recognized by a history of exert a modest independent risk largely after longer congestive heart failure (CHF), physical exam, rest- periods of follow-up, ETIVA are dealt with in ing ECG,10 echocardiogram, or nuclear imaging. Chapter 6 of this book. Who needs to undergo cardiac catheterization Myocardium in jeopardy refers to myocardium among patients with stable CAD? This is easy supplied by a coronary artery that has a critically to decide when symptoms cannot be controlled, but obstructive lesion or has a plaque that could rup- otherwise it is often difficult to decide who should be ture and/or form a thrombus (labile plaque). The considered for intervention to prolong life. Modern “dirty,” subcritical lesion does not cause ischemia statistical techniques applied to patient populations during stress; therefore, it cannot be detected by an followed for cardiac endpoints make it possible to exercise test. However, interventions other than identify low-risk patients, who do not need cardiac statins and acetylsalicylic acid (ASA) are not effective catheterization, and those at high risk, who may in improving outcomes. The obstructive lesion benefit when it is feasible to attempt intervention. should cause ischemia and be detectable with an exercise test. These lesions can be altered by inter- STATISTICAL METHODS USED FOR ventions (for example, percutaneous coronary SURVIVAL ANALYSIS intervention [PCI] and coronary artery bypass To answer these questions regarding patient deci- sions, follow-up studies must be performed and special statistical methods called survival analysis must be applied. Survival analysis consists of a

C H A P T E R 8 Prognostic Applications of Exercise Testing 251 group of uni- and multivariate mathematical tech- technique is helpful in illustrating findings and niques that consider person-time of exposure and facilitating drawing of survival curves, it discards use that to calculate hazard or risk. The key differ- valuable prognostic information and may weaken ence between survival analysis and other statistical the apparent prognostic significance of a continu- methods is censoring or removal from exposure. ous variable. Censoring is done at time of “lost to follow-up,” removal from risk (i.e., CABS, percutaneous trans- Endpoints and Censoring luminal coronary angioplasty), or termination of the study. The two most commonly used techniques The relative prognostic importance of the ischemic are Kaplan-Meier survival curves for univariate variables can be minimized by not censoring inter- analysis and the Cox hazard model for multivariable ventions for ischemia (i.e., removal of intervened analysis. Multivariable analysis is necessary because patients from observation when the intervention many of the variables interact. Univariately, vari- occurs in follow-up) because the intervention stops ables can be associated with death but the associa- patients from dying. Consideration of all-cause tion may be through other variables. For instance, mortality instead of cardiovascular (CV) mortality the use of digoxin associates with death through can have the same effect. This may explain why the CHF and exercise-induced ST elevation associates ischemic variables included in the Duke score that with death most often through the underlying clearly had diagnostic power12 do not predict all- Q waves. cause mortality. While all-cause mortality has advantages over CV mortality as an endpoint,13 the All variables should be explored by means of Duke score was generated using the endpoints of Kaplan-Meier survival curves for univariate com- infarction and CV death.14 Interventions such as parisons and the Cox model for multivariate analy- CABS or catheter procedures were censored in sis. A Cox proportional hazards model should be the Duke study (i.e., subjects were removed from used to determine the effect of a given indepen- the survival analysis when interventions occurred). dent variable on time to death. Many of the vari- Such censoring should increase the association of ables univariately predictive of death are likely to ischemic variables with outcome, by removing have overlapping prognostic significance; a multi- patients whose disease has been alleviated, and variate stepwise Cox regression analysis can be thereby would not be as likely to experience the employed. The Cox model assumes that the hazard outcome. Often researchers do not censor patients which equals the instantaneous death rate is given if they had a CV procedure during follow-up by the formula: because they do not have that information. From a previous study using a similar Veterans Affairs hi(t) = h(t) Ci, (VA) patient population with an annual all-cause mortality of 3%, our group found that 75% of where Ci = exp (B1X1i + B2X2i +…+ BpXpi) deaths were CV deaths, and that 6% of patients The model assumes that the hazard (h) of were censored in follow-up due to CABS.15 If the proportions are similar in our current population, death for patient i at time t (hi (t)) equals the haz- it would not be unreasonable to expect a bias ard of death for an “average patient” at the same against the predictive power of these variables. The time (h(t)) multiplied by a factor (Ci) that is a contradictory results regarding the prognostic function of the prognostic profile of patient i; this power of ischemic variables could also be due to is the proportional hazards assumption that gives the more effective methods of treatment currently the model its name. The proportional hazard coef- available for ischemic coronary disease compared ficient for patient i (Ci) is, in turn, a function of to LV dysfunction. the values for that patient of a set of prognostic fac- tors (X1i,…, Xpi), multiplied by a corresponding set The use of coronary interventions as end- of regression coefficients (B1,…, Bp) that measure points falsely strengthens the association of the strength of the association between the prog- ischemic variables with endpoints, because the nostic factor and outcomes of large numbers of ischemic responses clinically result in the inter- patients with the same disorder.11 Cox model also vention being performed. Although some investi- assumes that the effect of a prognostic factor on gators have justified their use by requiring a time outcome is linear. Variables of prognostic signifi- period to expire after the test before using the cance may be discrete or they may be continuous. intervention/procedure as an endpoint, this still Many studies analyze the strength of a continuous influences the associations. Another problem is prognostic factor by setting an arbitrary “cutpoint” that variables predicting infarction can be different and dividing the patients into subgroups with val- ues above and below the cutpoint. Although this

252 E X E R C I S E A N D T H E H E A R T than those predicting death, creating a situation In addition, specialized situations for predicting where one variable’s contrasting effects with prognosis, including silent ischemia, diabetes, and respect to two endpoints can cancel each other out. the elderly, will be targeted separately. All-Cause versus Cardiovascular Early Prognostic Studies Mortality In one of the first follow-up studies of treadmill Recent studies of prognosis have actually not been testing, Ellestad and Wan16 reported the predictive superior to the earlier studies that considered CV implications of maximal exercise testing in 2700 endpoints and removed patients who had inter- individuals followed from 6 months to 9 years. ventions from observation. This is because death ST depression and prior myocardial infarction were data is now relatively easy to obtain, whereas pre- both associated with subsequent higher mortality. viously investigators had to follow-up the patients The first study is from Duke by McNeer et al17 and contact them or review their records. Thus, and the other study is from the Coronary Artery prognostic studies were uncommon because of Surgery Study (CASS) data by Weiner et al.18 Both the expense of follow-up. CV mortality can be studies evaluated more than 1000 patients and had determined from death certificates. While death at least a 1 year follow-up. Those at high risk in the certificates have their limitations, in general, they Duke study had more than 1 mm of ST segment classify those with accidental, gastrointestinal, depression at less than 7 METs; the risk was even pulmonary, and cancer deaths, so that those higher if the maximal heart rate was less than remaining are most likely to have died of CV 120 beats per minute. Those at low risk did not causes. This endpoint is more appropriate for a have ST-segment depression, were able to exceed test for CV disease. Whereas all-cause mortality is 13 METs, or had a maximal heart rate of over 160. a more important endpoint for intervention stud- CASS patients at high risk either had markers of ies, CV mortality is more appropriate for evaluat- CHF or ST-segment depression at a low workload. ing a CV test (e.g., the exercise test). Identifying Patients able to exceed 13 METs were at low risk those at riskof death due to any cause does not regardless of their other responses. make it possible to identify those who might ben- efit from CV interventions, one of the two goals of The study by Podrid et al19 has placed some prognostication. doubt on the use of exercise testing to identify high-risk patients. They contend that the prevailing The patient groups that have been studied to view, “that patients with marked amounts of determine their prognosis using exercise testing ST depression have far advanced multivessel dis- include: (1) patients who have had myocardial ease and that CABS is the only way to improve infarction (Chapter 9), (2) patients with stable their outlook” is in error. In their select group of coronary heart disease (discussed in this chapter), patients with normal ventricular function, who (3) patients with CHF (Chapter 10), and (4) asymp- were referred because of profound ST-segment tomatic individuals (Chapter 11). Patients who depression, they did not find a bad prognosis. In have had myocardial infarction, those with CHF, 142 patients with CAD and severe ST-segment and asymptomatic individuals will be discussed in depression with a mean follow-up of 59 months, separate chapters. there was only 1.4% mortality and only 1.3% had CABS per year. This study points out that it is nec- PREDICTION OF HIGH RISK essary to consider multiple variables when predict- AND/OR POOR PROGNOSIS IN ing the risk of ischemic heart disease. A relatively PATIENTS WITH STABLE low-risk group can be found in any population CORONARY HEART DISEASE identified using one-risk predictor, by excluding patients with other risk predictors. This chapter will discuss prognostic studies using exercise testing in patients with stable coronary Dagenais et al20 analyzed the factors influencing heart disease predicting the following: the 5-year survival rate in 220 patients with at least 0.2 mV of ST-segment depression during exercise 1. Cardiovascular disease endpoints testing. They confirmed previous observations 2. Coronary angiographic findings that survival was inversely proportional to the exer- 3. Improved survival with CABS cise workload: all patients who achieved 10 METs survived and the patient survival rate declined in relation to exercise capacity.

C H A P T E R 8 Prognostic Applications of Exercise Testing 253 Bruce et al21 added to their analysis of the Maximal SBP, heart rate, and exercise capacity Seattle Heart Watch by applying noninvasive cri- were far less important. Cox’s regression analysis teria in a learning set for exercise-enhanced risk showed ejection fraction, age, number of diseased assessment for events resulting from coronary vessels, and resting ventricular arrhythmia, in that heart disease to a test series in a later population order, to be most predictive. From Bad Krozingen, sample. In this series, subsequent follow up in Gohlke et al25 followed 1034 patients with CAD 5308 men enrolled in the learning series of the specifically to answer the question, “Whether exer- Seattle Heart Watch from 1971 to 1974 were com- cise testing could provide additional prognostic pared with findings in 3065 men enrolled from information when angiographic information was 1975 to 1981. Of the 8373 men, 4105, or almost available?”. They found exercise workload, angina half, were classified before exercise testing as during the exercise test, and maximal heart rate asymptomatic healthy individuals. Another 1374 to independently predict risk of death. Exercise- men had hypertension and 2894 had prior clinical induced ST depression was the only independently manifestations of coronary heart disease includ- predictive risk factor in the subgroup with three- ing angina, myocardial infarction (MI), cardiac vessel disease and normal ventricular function. arrest, or cardiac death. Men in the same age and risk groups for each pretest clinical classification From the Italian Multicenter Study, Brunelli showed similar gradients of risk. Age-standardized et al26 reported their findings in 1083 patients event rate showed a reduction longitudinally in younger than 65 years of age, followed for a mean healthy men and in patients who underwent CABS. of 66 months. They found clinical markers to strat- It is important to realize that the majority of events ify risk and they found that coronary angiography occurred in men with only increased risk rather added prognostic information only in patients than high risk. The two exercise predictors of sur- with moderately severe disease. Q-wave presence vival were duration of exercise and the ST-segment and history of infarction were the most important response. clinical predictors (CHF was not considered). Exercise-induced ST depression was not consid- Follow-up Studies with Clinical, ered independently but rather was combined with Exercise Test, and Coronary angina and exercise capacity in order to create a Angiography marker associated with CV death. Since the pioneering studies from the University From the CASS, Weiner et al27 analyzed exercise of Alabama,22,23 numerous investigators have uti- test, coronary angiographic, and clinical variables lized clinical, exercise test, and catheterization data in 4083 patients to identify predictors of mortality to predict prognosis in patients with CAD. Implicit in medically treated patients with symptomatic in these studies has been the issue of which vari- CAD. This study was based on analysis of 16% of ables are predictive, and whether exercise testing the registry of patients with no previous CABS who and coronary angiography improve prediction were able to undergo a standard or modified Bruce sufficiently to merit their performance despite protocol within 1 month of their catheterization. their expense and risk. During the mean follow-up of 4 years, 212 patients, or 5%, died. This represents a very low annual Oberman et al22 found cardiac enlargement on mortality, and approximately 40% of the patients chest x-ray and a history of CHF to be the two most had a prior MI and 36% underwent CABS during predictive independent clinical variables and that a 3-year minimal follow-up. Standard clinical vari- angiography improved prediction of death. They ables, including chest pain, CHF, physical exam, did not consider exercise test results because of family history, risk factor index, drugs, and cardiac incomplete data, but they were the first investiga- catheterization findings were taken into account. tors to demonstrate the poorer prognosis found in Exercise test variables included limiting symp- those patients unable to perform the exercise test. toms, premature ventricular contractions (PVCs), peak heart rate, peak SBP, ST-segment response, From the Seattle Heart Watch, Hammermeister and final exercise stage. Thirty variables were ana- et al24 assessed 733 medically treated patients lyzed in 4000 patients. Regression analysis demon- by going stepwise, first through clinical markers strated that seven variables were independent and then the exercise test. CHF was the most predictors of survival. A high-risk subgroup (annual important clinical variable, and maximal double mortality about 5%) was identified consisting of product was the most important treadmill variable. patients with either CHF or ST-segment depression and less than 5 METs exercise capacity. When all 30 variables were analyzed jointly, the LV contraction

254 E X E R C I S E A N D T H E H E A R T pattern and the number of diseased vessels were Six steps were used to derive the prognostic the best predictors of survival. In a subgroup of treadmill score. First, the patient population 572 patients with three-vessel disease and good was randomly split into two groups: a training LV function, the probability of survival at 4 years sample of 1422 patients and a validation sample of ranged from 53%, for patients only able to achieve 1420 patients. Second, the Cox proportional haz- stage 1 or 2, to 100%, for patients able to perform ards regression model was used in the training sam- 10 METs. Thus, in patients with defined coronary ple to assess the strength of association between the lesions, clinical and exercise variables primarily primary study endpoint (death of CV cause) and relating to left ventricle function were helpful in treadmill responses. Treadmill responses were then assessing prognosis. The following are some of ranked using the likelihood ratio derived from the the univariate risk ratios generated by some of the Cox model. Third, the most important treadmill variables: age above 60 = 2.5x, prior MI = 2.4x, CHF response was entered into a Cox regression model, = 5x, cardiac enlargement = 9x, digoxin = 4x, less and the remaining responses were then entered in than stage 1 = 2x, more than 0.1 mV ST-segment order until the model represented the independent depression = 1.4x. The presence of CHF was the prognostic information available from the exercise most potent clinical predictor of survival when the test. Fourth, the regression coefficients from this clinical and exercise test variables were analyzed, regression model were used to form a linear tread- and has led us to consider CHF patients separately mill score. Fifth, the new score was tested to deter- when doing prognostic evaluation studies. mine if patients with different levels of scores had a survival pattern similar to that seen in the training The Duke Treadmill Score and Nomogram sample. Finally, the score was recalculated based on variables derived from the test results in Mark et al28 studied 2842 consecutive patients who all patients. Kaplan-Meier life table estimates underwent cardiac catheterization and exercise were used to generate cumulative survival curves. testing and whose data was entered into the Duke Subgroup rates were not calculated beyond the computerized medical information system. The point in follow-up when fewer than 15 patients median follow-up for the study population was remained at risk. All patients considered were 5 years and was 98% complete. All patients under- initially treated nonsurgically. In the 24% of the went a Bruce protocol exercise test and had standard patients who had CABS, the follow-up was mea- ECG measurements recorded. A treadmill angina sured to the time of surgery and then they were index was assigned a value of 0 if angina was absent, removed from observation. Seventy percent of the 1 if typical angina occurred during exercise, and 2 if study patients were men and the median age was angina was the reason the patient stopped exercis- 49 years. Two thirds had stable angina and one ing. Before the test, 54% of the patients had taken third had progressive anginal symptoms. A history propranolol and 11% had taken digoxin. ST mea- of MI was present in 29%, and 22% had pathologi- surements considered were sum of the largest net cal Q waves. At catheterization, the mean ejection ST depression and elevation, sum of the ST dis- fraction was 60%, and 27% had three-vessel or left placements in all 12 leads, the number of leads main CAD. showing ST displacement of 0.1 mV or more, the product of the number of leads showing ST dis- The largest net ST deviation recorded during placement, and the largest single ST displace- exercise in any one of the 12 leads proved to be the ment in any lead. To make the score apply to single most important variable for predicting prog- other treadmill protocols, it is necessary to con- nosis. After adjusting for maximum net ST devia- vert minutes in the Bruce protocol to METs with tion using the Cox model, only two other variables the equation: contained additional prognostic information: the treadmill angina index and exercise time. The METs = 1.3 (minutes) − 2.2 results did not change substantially when patients taking beta-blockers or digoxin were excluded. The or results also remained unchanged when patients treated surgically were excluded from the study. minutes in the Bruce protocol = (METs + A score was calculated as follows: 2.2) ÷ 1.3 exercise time − (5 × ST maximum net deviation) Patients with ST-segment elevation in ECG leads − (4 × angina index), with pathological Q waves were excluded because this ST-segment response has a different meaning. where exercise time is measured in minutes and ST deviation is measured in millimeters. Patients at

C H A P T E R 8 Prognostic Applications of Exercise Testing 255 high risk with a score of −11 or lower had a 5-year Other Prognostic Studies survival of 72%. Patients at moderate risk with a score of −10 to +4 had a 5-year survival of 91%, From the VA randomized trial of CABS, Peduzzi and patients with a low risk score of +5 or greater et al29 reported on the 7-year follow-up of the had a 5-year survival risk of 97%. When total car- 245 patients randomized to medical management diac events were considered, the high-risk group who had a baseline treadmill test. Univariately had a 5-year survival rate of 65%, the moderate and using Cox analysis, ST depression (≥2mm), risk group 86%, and the low risk group 93%. exercise-induced PVCs, and final heart rate The treadmill score contained prognostically greater than 140 beats per minute were signifi- important information even after the information cant predictors. Unfortunately, they did not cen- provided by clinical and catheterization data was sor on interventions. These results are in marked considered. The prognostic stratifying power of contrast to other studies and our findings, in the treadmill score was greatest in patients with which PVCs did not have an independent predic- three-vessel disease and lowest in those with one- tive power, and high heart rates were found to vessel disease. Patients at highest risk had the be protective rather than associated with risk. greatest potential to increase their survival dura- Peduzzi et al did not find a poor exercise capacity tion by having CABS. Patients with three-vessel to be predictive as well. These unusual results disease and a treadmill score of −11 or less had a might be explained by their failure to censor on 5-year survival rate of 67%. Patients with three- interventions. vessel disease and a risk of this magnitude appear to gain a survival advantage through surgery. Those In Buenos Aires, Lerman et al30 reported patients with three-vessel disease and a treadmill 190 patients with exercise test and coronary score equal to or greater than 7 had an excellent angiograms who were followed-up for 6 years. prognosis. This score has been implemented in a Their study began in 1978; patients had a high nomogram (Fig. 8-1). annual mortality and a low rate of interven- tions, yet exercise-induced ST depression failed to ■ FIGURE 8–1 The Duke treadmill nomogram with a case example.

256 E X E R C I S E A N D T H E H E A R T predict prognosis. Maximal SBP of less than the model. A simple score based on one item of 130 mmHg was the strongest predictor. clinical information (history of CHF), a resting electrocardiogram finding (ST depression), and Wyns et al31 evaluated the independent prog- an exercise test response (exertional hypotension) nostic information provided by exercise testing by stratified our patients for 4 years after testing calculating the survival rates with the life table from 75% with a low risk (annual cardiac mortal- method in 372 men referred for coronary angiog- ity of 1%), to 17% with a moderate risk (annual raphy. A previous MI was noted in 146 and 248 mortality of 7%), and 1% with a high risk (annual had typical angina. During a mean follow-up of cardiac mortality of 12%; hazards ratio of 20, 95% 29 months, 32 patients died and 27 patients had confidence interval 6 to 70×). Three quarters of nonfatal events. Typical angina pectoris and/or an those usually undergoing cardiac catheterization old MI and an abnormal exercise test (angina and/or could be identified by simple noninvasive vari- ST-segment shifts) had significant prognostic value. ables as being at such low risk that invasive inter- In patients with an MI and/or angina, the 5-year vention is unlikely to improve prognosis. cumulative survival rate was 76% if the exercise test was abnormal versus 94% if it was normal. Also from the Duke database, Califf et al34 Cox regression analysis was performed, and by uni- applied clinical measures of ischemia (exercise test variate analysis, the age and the maximal workload results not considered) to predict infarct-free sur- were the only noninvasive predictive variables vival in 5896 patients with angina and angiograph- for survival or cardiac events. Exercise capacity ically significant (>75% lesions) CAD. The Cox provided prognostic information that was not regression model chose the following variables in available either from the history or from cardiac descending order: more than 1 mm of resting catheterization. ST depression or T-wave inversion, frequency of angina, unstable angina, typical angina, and dura- A second study from Saint-Luc Hospital in tion of symptoms. An angina score was derived Brussels excluded patients with a prior MI.32 From from the Cox coefficients, and when entered into 1978 to 1985, 470 consecutive male patients with a model with catheterization data, the following complaints of chest pain underwent a maximal variables were chosen in descending order predict- exercise perfusion test and coronary angiography. ing survival: ejection fraction, number of diseased During follow-up (1 to 8 years) 32 patients died of vessels, left main stenosis, angina score, age, and CV causes and 30 had a nonfatal MI. The average sex. This score helped predict prognosis even when annual CV death rate was 2%. Of historical vari- the catheterization data was considered. ables, only age was chosen as significant multi- variately, while angina and pretest likelihood were Detre et al35 developed a multivariate risk func- chosen univariately. A maximal exercise test score tion from the 508 patients randomized to medical based on maximal heart rate, ST60 at maximal exer- treatment in the VA randomized study of CABS. cise, angina during the test, maximal workload, The variables, in order of importance, were ST- and ST slope was chosen in multivariate analysis. segment depression on resting ECG, history of This combined score is similar to the ischemic MI, history of hypertension, and New York Heart index used in the Italian Study which made a test Association functional classification III or IV. abnormal if one or more of the following occurred: Applying the risk function to medical and surgical angina, ST depression, or poor exercise capacity. patients of the 1972−1974 cohort yielded a 5-year probability of dying for each patient. Investigation In a VA Medical Center, 588 male patients who of treatment effects in approximate terciles, underwent exercise testing and cardiac catheteri- obtained by collapsing the probability distribution zation were followed-up to determine whether CV into low-, middle-, and high-risk groups, showed mortality could be predicted by clinical and exer- that surgery was beneficial for patients in the high- cise test data.33 Over a mean follow-up period of risk tercile even after removal of patients with 4 years, there were 39 CV deaths and 45 nonfatal left main CAD (17% surgical versus 34% medical MIs. The Cox proportional hazard model demon- mortality at 5 years). This finding was accentu- strated the following characteristics to have a ated when patients in the 10 hospitals with the significant independent hazards ratio: history of lowest operative mortality (3.3%) were compared. CHF (relative risk = 4×), ST depression on the rest- Mortality results in the low-risk tercile favored ing ECG (3×), and a drop of SBP below rest dur- medical treatment (medical versus surgical mor- ing exercise (5×). Exercise-induced ST depression tality, 7% versus 17%). The risk function predicted was not associated with either death or nonfatal mortality not only for the VA medical group, but MI. From cardiac catheterization, only the ejec- also for an independent symptomatic coronary tion fraction added independent information to

C H A P T E R 8 Prognostic Applications of Exercise Testing 257 heart disease population from the University of improve prediction sufficiently to merit their per- Alabama angiographic registry. formance despite their expense and risk. A careful literature search has yielded the nine studies in WHY DO PROGNOSTIC STUDIES Table 8-1 for comparison. All used multivariate FAIL TO AGREE? survival analysis techniques and the variables chosen are listed in order of predictive power. Some These nine studies have utilized clinical, exercise investigators combined variables, and others did test, and catheterization data to predict prognosis not consider key variables or excluded patients with in patients with CAD. Implicit in these studies has certain clinical features (e.g., CHF, those receiving been the issue of which variables are predictive, and digoxin). Nevertheless, two of the nine found a whether exercise testing and coronary angiography history of CHF, two found exercise SBP, and one found resting ST depression to be associated with TA B L E 8 – 1 . Population descriptors including clinical variables and results from exercise testing and coronary angiography in the follow-up studies of multivariate prediction of cardiac events Descriptors LB VAMC LB VAMC VA CABS CASS DUKE (No cath) Clinical 1984–1990 1984–1990 1970–1974 1974–1979 1969–1981 Years entered 2546 588 245 4083 2842 Population size 59 59 (mean) 51 (mean) 50 49 (median) Age 100 100% 100% 80% 70% Males (%) 5% 8% 9% 8% 4% Congestive heart failure 23% 45% 54% 40% 29% Myocardial infarction 21% 37% 38% 22% 22% Q waves (at least one) 8% 8% NA 14% 11% Digoxin 22% 35% 14% 40% 54% Beta-blockers 21% 52% 100% 50% 47% Typical angina Exercise Test 22% 58% 72% 44% 35% % with 1 mm ST-segment depression 4% 35% 66% 80% 50% % angina 137 124 125 138 134 Maximal heart rate (beats/min) 175 159 156 171 160 Maximal systolic blood pressure (mmHg) 8.4 6.6 5.7 NA 7 METs 5% 12% 19% 12% 6% Premature ventricular contractions NA 14% 55% 23% 22% Cardiac Catheterization Findings NA 7% 13% 7% 5% Three-vessel disease (%) NA 26% 0% 34% 40% Left main artery disease (%) NA 60 (mean) 57% 60 (median) No significant lesion (%) NA 70% 50% 70% 75% Ejection fraction Significant lesion criteria 5 5 7 5 5 Follow-up 2% 20% 24% 36% 24% Years 1.5% 2.7% NA 1.0% 1.6% Coronary artery bypass surgery 2.8% 3.5% 4.0% 1.6% 1.8% Annual cardiovascular mortality Annual total mortality CHF/digoxin CHF E-I PVCs Independent Predictors of METs SBP drop MHR >140 Mortality by Priority Max SBP Resting ST E-I ST dep depression >2 mm CHF E-I ST E-I ST Treadmill stage depression depression E-I ST depression Angina index Treadmill time Continued

258 E X E R C I S E A N D T H E H E A R T TA B L E 8 – 1 . Population descriptors including clinical variables and results from exercise testing and coronary angiography in the follow-up studies of multivariate prediction of cardiac events—cont’d Italian Belgian Belgian (No MI) German Seattle Buenos Aires 1976–1979 1972–1977 1978–1985 1975–1978 1971–1974 1972–1982 1083 372 470 1238 733 180 49 (mean) 48 52 50 (mean) 52 (mean) 51 (mean) 90% 100% 100% 90% 80% 96% Excluded 1% Excluded 13% Excluded 42% 39% Excluded >50% 40% 64% 37% 39% Excluded 50% 45% 0% Excluded 8% 18% 71% 95% 0% 65% 42% 67% 75% 95% 86% 60% 60% 27% 54% 56% 128 130 49% 44% 61% 145 151 171 148 140 118 160 5.2 5.4 NA 186 182 6.5 21% 15% 9 8 5 18% 44% 5% 2% NA 12% 8 5% 34% 26% 33% 0% 26% 8% 8% 0% 39% 60 18% 22% 0% 60 75% 75% NA 65 60 70% 6 5.5 50% 50% 50% 3.5 9% 15% 5 5 5 4.6% 1.5% 28% 29% 2.6% 2.0% 1.8% 2.0% 2.4% 3.1% Max SBP <130 2.4% ST elevation Q-wave Age Exercise capacity CHF Prior MI Age Max exercise Angina Max double <4 METs Exercise capacity score product (−2 to +2) MHR Inappropriate Effort ischemia (MHR, ST60, AP, Max SBP dyspnea Exercise capacity watts, ST slope) Angina Resting ST dep AP, angina pectoris; CABS, coronary artery bypass surgery; CASS, Coronary Artery Surgery Study; CHF, congestive heart failure; Dep, depression; E-I, exercise-induced; LB, Long Beach; METs, metabolic equivalents; MHR, maximal heart rate; MI, myocardial infarction; PVCs, premature ven- tricular contractions; SBP, systolic blood pressure; VA, Veterans Affairs; VAMC, Veterans Affairs Medical Center. death, as we did. However, in contrast to the Long characteristics in Table 8-1. Other than the differ- Beach VAMC study three found exercise-induced ST ences detailed previously, the Duke population depression and six of the nine found poor exercise and the VA CABS study patients appeared to be capacity to be predictive of death. Unfortunately, more “ischemic;” no obvious population, method- for comparison sake, the Duke study did not have ological, or test characteristics explain the different maximal SBP collected for consideration. The results. All studies had to deal with interventions choice of variables in the Cox hazard models from these studies is tabulated in Table 8-2. Age is not TA B L E 8 – 2 . Meta-analysis of prognosis in chosen by most of the studies because of the nar- stable coronary artery disease studies requiring an row age range for patients submitted to cardiac exercise test and coronary angiography catheterization. Exertional hypotension has pre- viously been examined in our population and in Poor exercise capacity 6 of 9 studies the other studies reviewed. Note, however, that Congestive heart failure 3 of 9 studies this is the first time it was chosen by a Cox model ST-segment depression rather than just observed univariately. 2 of 9 studies Resting 3 of 9 studies Because of the differences in the variables Exercise 3 of 9 studies chosen to have independent predictive power in Exercise systolic blood pressure the reported studies, we have presented their key

C H A P T E R 8 Prognostic Applications of Exercise Testing 259 that alter the natural history but each censored Given this etiological milieu, associating clinical on them as we did, except for the earlier VA CABS and test markers with death as an outcome study. The first explanation that comes to mind becomes quite difficult. Other ischemic events (i.e., for the failure of ST depression to predict prognosis unstable angina, hospitalization for chest pain) in six of the nine studies might be that the clini- were too “soft” for consideration. In addition, inter- cal process was highly effective in selecting high- ventions, even if only considered an endpoint if risk patients with exercise-induced ST depression they occur months after testing, are clearly related for interventions. However, all patients were cen- to the test response (i.e., patients are submitted sored at the time of their CABS or PTCA and the for interventions because of abnormal tests). Since same variables were chosen when the patients who nonfatal MI most likely is an ischemic event, received these interventions during follow-up were infarct-free survival is another way of including excluded. Also, in the five comparable studies that more ischemic endpoints, but we had similar did not find ST-segment depression to be predictive, results when this was considered the endpoint this did not appear to be related to surgical inter- in the Cox model. Differences in populations may vention rates. have a higher proportion of one or the other type of mortality (pump failure versus ischemia). This Ischemic exercise test variables are clearly may explain why ischemic variables are more pre- related to ischemic events during follow-up dictive in one population and “myocardial dam- (i.e., nonfatal MI, CABS, PTCA). This is logical but of age” variables more predictive in another. One little help in clinical decision-making, since the cli- could argue that our population included a major- nician has no trouble in justifying these procedures ity of patients who died secondary to CHF. However, for patients whose symptoms accelerate after ade- the same results were obtained after removing quate medical management, given the established patients who either carried that diagnosis or were symptomatic benefit from interventions. The taking digoxin at time of entry into our study; problem lies in justifying intervention to improve most of the other studies had a similar proportion survival for patients whose symptoms are satisfacto- of such patients. rily managed medically. Our study demonstrates that simple clinical indicators can stratify these patients Work-Up Bias with stable CAD into high- or low-risk groups. Surprisingly, exercise-induced ischemic variables All of the above studies selected patients by (ST depression), commonly thought by physicians requiring that they also underwent coronary to identify high risk, did not do so in five of the angiography. To evaluate the effect of this selection nine comparable studies. process, the Duke group repeated their analysis in an outpatient population that did not undergo car- Spectrum of Cardiac Death diac catheterization.14 The same variables were cho- sen in their Cox model and the same equation was The following discussion of endpoints will derived. Similarly, we analyzed 2546 male patients attempt to explain why the studies available who underwent noninvasive evaluation for CAD, for making clinical prediction rules do not agree. including exercise testing. Over a mean follow-up Cardiac death occurs in a spectrum between period of 2.8 years, there were 119 CV deaths and patients with myocardial damage who die of CHF 44 nonfatal MIs. The Cox proportional hazard (or pump failure) and those with normal ventricles model demonstrated the following characteristics in whom ischemia precipitates death (Fig. 8-2). to have a significant independent hazards ratio: The clinical and test markers would thus be history of CHF and/or taking digoxin, exercise- expected to be quite different for patients who die induced ST depression, exercise capacity in METs, at the extremes of this spectrum. Whereas mark- and the response of SBP during exercise. A simple ers of myocardial damage (history of CHF, Q waves) score based on these four factors stratified patients track the former, markers of ischemia (angina, from low risk (annual cardiac mortality of less than ST-segment depression) better track the latter. 1%) to high risk (annual cardiac mortality of 7%). Arrhythmias, poor exercise capacity, and exer- tional hypotension are associated with both. The first Duke study used “in-patients,” all of Further complicating prediction algorithms, “dam- whom had a catheterization, while the later report age” markers predict short-term deaths, while only included outpatients evaluated prior to the “ischemic” markers predict deaths occurring 2 or decision for cardiac catheterization. Their score, more years later. based on treadmill time, exercise-induced ST

260 E X E R C I S E A N D T H E H E A R T ■ FIGURE 8–2 The spectrum of cardiovascular death. Cardiovascular mortality for individuals and for whole populations occurs in varying degrees as a result of ischemia and left ventricular dysfunction. This range in pathophysiologic characteristics has two conse- quences: (1) variations in study population and design can produce different results compared with prognostic studies and (2) applying noninvasive variables for prognosis studies, and knowledge of local population characteristics, has the potential for greater precision in prediction. depression, and angina score during the test, per- group with a higher annual cardiac mortality formed as well for prognostication as it did in the (2.6% versus 1.5%). This second study included a first report. Therefore, “work-up” bias did not affect population with a near normal age-adjusted exer- their prognostication model. We have attempted cise capacity, while the first study population had the same type of validation in this study. In con- an average age-adjusted exercise capacity 75% of trast to the Duke group, we included exercise SBP normal. Age, as a variable, is not chosen by most of and clinical data in our model. Although history the studies, including ours, because of the narrow of CHF or digoxin was the most powerful variable age range for patients referred for evaluation of in both of our VA studies, surprisingly, different CAD and its relationship to other variables. exercise test variables were chosen. The model from our first VA study in patients selected for LONG-BEACH VA TREADMILL catheterization only chose exertional hypotension, SCORE while the model from this second VA study (only noninvasive clinical evaluation) found exercise- Using stepwise selection, the Cox model was induced ST depression, exercise SBP, and exercise allowed to build on each variable group to arrive capacity to have predictive power. at the final model that chose history of CHF or digoxin, the change in SBP score, METs, and The “work-up” bias inherent in choosing exercise-induced ST depression.36 A score was patients for cardiac catheterization in our first study resulted in a sicker, older, more disabled

C H A P T E R 8 Prognostic Applications of Exercise Testing 261 then formed using the coefficients from the Cox Figure 8-4 illustrates the ROC curves for the model as follows: Duke score and the VA score predicting CV deaths in the total group (n = 3134). The area under the 5 × (CHF/dig [yes = 1, no = 0]) + (exercise- VA score curve (0.76) was significantly greater induced ST depression in millimeters) + (change than the area under the Duke score curve (0.68). Similar results were also obtained in the popula- in SBP score) − (METs) tion presented in this study (n = 2546). These scores are summarized in Table 8-3. Because of its This resulted in a likelihood ratio statistic of reproducibility, its applicability in women, and its 68 with 4 degrees of freedom. Three groups functionality as a diagnostic score, the DTS is were formed using the score: <−2 (low risk), −2 highly recommended. to +2 (moderate risk), and >2 (high risk). The Kaplan-Meier survival curves are illustrated in Unfortunately, one of the only studies to also Figure 8-3. This score enabled identification of compare these two scores was carried out on a low-risk group (80% of the population), with patients who had prior MI and the follow-up was an annual mortality of less than 1% over the first only for 6 months. The GISSI investigators com- 3 years after their exercise test. In addition, a pared the performance of the DTS and the Veteran moderate-risk group (14% of the population), with Affairs Medical Center Score (VAMCS) in predict- a 4% annual mortality, and a high-risk group (6% ing 6-month death in GISSI-2 study survivors of of the population), with a 7% annual mortality acute MI treated with thrombolytic agents to a over the 3 years after their exercise test, were simple predictive scoring system developed from identified. the same database.37 Patients of the GISSI-2 study (n = 6251) performed a symptom-limited exercise In addition, the Duke treadmill score (DTS) test 1 month after MI. They calculated for each was calculated for each of our VA patients. The patient the DTS, the VAMCS, and the new GISSI treadmill angina index was modified because we score. All three scores were able to stratify risk did not have angina coded as the reason for stop- into three groups: a low risk of less than 1%, a ping, but was coded as 0 for not present, and 1 as moderate risk of 2%, and a high risk of 5%. They occurring during the test; we used METs instead concluded that exercise test-derived prognostic of minutes of exercise scores in a population of survivors of acute MI treated with thrombolytic drugs could stratify DTMS = METs − 5 × [mm ST depression during risk. exercise] − 4 × [treadmill angina index] 1.00 0.75 Survival probability 0.50 0.25 0.00 < –2 2 3 4 0 –2 to +2 >2 1 ■ FIGURE 8–3 Years The Kaplan-Meier survival curve for the Veterans Affairs (VA) prognostic score. VA TM score = 5 (CHF [0,1]) + EI ST depression (millimeters) + change SBP score (0–5) – METs

262 E X E R C I S E A N D T H E H E A R T ■ FIGURE 8–4 ROC curve for the Veterans Affairs (VA) score and the Duke score for predicting cardiovascular death. However, Morise and Jalisi38 compared our new guidelines and better than the DTS. These results simple scores for predicting angiographic CAD to extended to diabetics, inpatients, women, and the DTS for predicting all-cause mortality. They patients on beta-blockers. It would really be bene- utilized 4640 patients without known coronary ficial to see these same analyses performed predict- disease, who underwent exercise testing, to evaluate ing infarct-free survival with censoring for cardiac symptoms of suspected coronary disease between interventions. 1995 and 2001. Overall mortality was 3.0% with 3 years of follow-up. All three scores stratified PROGNOSIS IN “ALL-COMERS” patients into low-, intermediate-, and high-risk TO THE EXERCISE LAB groups. No differences were seen when patients were evaluated as subgroups according to gender, Previous prognostic studies focused on specific diabetes, beta-blockers, or inpatient status. Low- subsets of patients, so we decided to analyze all risk patients defined by the DTS had consistently patients referred for evaluation at our exercise lab higher mortality and absolute number of deaths between 1987 and 2000.39 There were 6213 males compared with low-risk patients using other scores. (mean age 59± 11 years) who had standard exercise In addition, the DTS had less incremental ECG treadmill tests over the study period with a stratifying value than the new exercise scores. mean 6-year follow-up. There were no complica- He concluded that simple pretest and exercise tions of testing in this clinically referred popula- scores risk-stratified patients with suspected tion, 78% of whom were referred for chest pain, coronary disease in accordance with published risk factors, or signs or symptoms of ischemic heart disease. Overlapping thirds had typical angina or TA B L E 8 – 3 . The two major treadmill history of MI. Of the patients, 579 had prior CABS prognostic scores and 522 had a history of CHF. Indications for testing were in accordance with published guide- Duke score = METs − 5 × (mm E-I ST depression) − 4 × lines. Twenty percent had died over the follow-up, (TM AP index) giving rise to an average annual mortality of 2.6%. VA score = 5 × (CHF/Dig) + mm E-I ST depression + Cox hazard function chose the following variables change in SBP score − METs in rank order as independently and significantly associated with time to death: METs less than 5, Treadmill angina pectoris (TM AP) score: 0 if no angina; age greater than 65, history of CHF, and history 1 if angina occurred during test; 2 if angina was the of MI. A score based on simply adding these vari- reason for stopping. ables classified patients into low-, medium-, and Change in systolic blood pressure (SBP) score: from 0 for rise greater than 40 mmHg to 5 for drop below rest.

C H A P T E R 8 Prognostic Applications of Exercise Testing 263 high-risk groups. The high-risk group (score of 3 1.0 or more) had a hazard ratio of 5 (4.7 to 5.3, 95% 0 CI) and a 5-year mortality of 31% (Fig. 8-5). 1 When CV mortality was available, we repeated 0.9 2 these analyses. Two additional treadmill responses appeared as independently significant: exercise- 0.8 3 induced ST depression and arrhythmias (Fig. 8-6). Survival SPECIFIC PROGNOSTIC ISSUES 0.7 Predicting Prognosis in Women ≥4 Clinical presentation, performance in diagnostic Score = (yes=1, no=0) tests, and prevalence of CAD is different between men and women presenting with chest pain. To 0.6 demonstrate the value of exercise testing in [CHF history] + [MI history] + [METs<5] + women, Duke University researchers analyzed [age>65] + [exercise-induced ST depression] + data from 976 women referred for evaluation of [exercise-induced ventricular ectopy] chest pain and who underwent exercise treadmill test and cardiac catheterization.40 Women and 0.5 1.0 2.0 3.0 4.0 5.0 6.0 7.0 men differed significantly in DTS, disease preva- 0 lence (32% versus 72% significant CAD), and 2-year mortality (1.9% for the study women com- Follow-up years pared with 4.9% for the men). Mortality increased for high-risk DTS groups in both genders. Two- ■ FIGURE 8–6 year mortality for women was 1.0%, 2.2%, and The “All-comers” treadmill score for predicting cardiovascu- 3.6%, respectively for low-, moderate-, and high- lar mortality. risk groups; For men it was 1.7%, 5.8%, and 16.6%, respectively. Because of the differences in disease Within the Women’s Ischemia Syndrome prevalence, women had better survival at all val- Evaluation study, Morise et al41 evaluated 563 ues of the DTS. In addition, the DTS actually women undergoing coronary angiography for performed better in women than in men for suspected myocardial ischemia. The prevalence excluding disease, with fewer low-risk women of angiographically significant CAD was 26%. having mild or severe disease. Overall, 189 women underwent treadmill testing. Prognostic endpoints included death, MI, stroke, Kaplan-Meier survival curves for the “all-comers” prognostic and revascularization. The simple scores and the score with all-cause mortality as the endpoint. DTS score stratified women into three probability groups according to the prevalence of coronary 1.000 disease—pretest: low 20/164 (12%), intermediate 53/245 (22%), and high 75/154 (49%); exercise 0.750 a test: low 11/83 (13%), intermediate 22/74 (30%), and high 17/32 (53%). However, the Duke score Survival b did not stratify as well. When pretest and exercise scores were considered together, the best strati- 0.500 fication with the exercise test score was in the intermediate pretest group. The Duke score did not 0.250 Score = (1= yes, 0 = no) c stratify this group at all. Pretest and exercise test d scores also stratified women according to prognos- METs < 5 + age > 65 + history tic endpoints. The exercise test score is most useful of CHF + history of MI or Q wave in women with an intermediate pretest score, con- sistent with American College of Cardiology/ a = 0, b =1, c = 2, d = more than 2 10.5 14.0 American Heart Association guidelines. 0 Predicting Prognosis in 0 3.5 7.0 Patients with Resting Follow-up years ST Depression ■ FIGURE 8–5 Kwok et al42 demonstrated that DTS can effec- The “All-comers” treadmill score for predicting all-cause tively risk-stratify patients with ST-T abnormalities mortality. on resting the ECG. When patients with ST-T

264 E X E R C I S E A N D T H E H E A R T abnormalities were classified into risk groups protocol. The tests were stopped when 85% to according to DTS, there were significant overall dif- 90% of peak heart rate was achieved, and there ferences among the risk groups for all outcome was no cool-down walk. HRR was measured endpoints. The 7-year event-free survival was 94%, at 2 minutes of recovery. HRR continued to 88%, and 69% for the low-, intermediate-, and be a strong predictor of all-cause mortality; high-risk groups, respectively. More patients with patients with an abnormal value had a mortality ST-T changes were classified as high-risk (5% ver- rate of 10% while patients with a normal value sus 2%) and their 7-year survival was lower than had a mortality rate of 4% at 12 years of that of the control population high-risk patients follow-up. (76% versus 93%). In order to further elucidate the power of HRR Should Heart Rate in Recovery in distinct populations, these same investigators be Included in Prognostic then published another study using patients Scores? referred for standard treadmill testing.45 Using the same methods as the original study, the inves- Recent studies have highlighted the prognostic tigators found similar results, although notably value of heart rate recovery (HRR), or the drop the cut-off value for an abnormal test was differ- in heart rate, after an exercise test. Although ear- ent. Patients with abnormal HRR had 8% mortal- lier physiological studies suggested a rapid HRR ity at 5.2 years, whereas patients with normal response to exercise to be a marker of physical fit- HRR had only 2% mortality. Neither this nor the ness, only recently has its prognostic value been previous study censored for CABG or PCI, and this reported. The rate of heart rate return to baseline study had 8% patients with CABG enrolled along postexercise is theorized to be due to high vagal with 75% asymptomatic individuals. The investi- tone associated with fitness and good health. gators also compared the prognostic ability of While the prognostic value of HRR has recently heart rate recovery to that of the DTS. Whereas been highlighted, its relative value compared with the ischemic components of the Duke score did other treadmill responses and its diagnostic value not have prognostic power, METs did, because the remains uncertain. Table 8-4 shows comparisons DTS produced similar survival curves to HRR. of the HRR studies across a number of important Patients with abnormal DTS and HRR survival parameters. In the first study, Cole et al43 looked were even further compromised. at 2428 adults who had been referred for exercise nuclear perfusion scans. Cole et al found that We attempted to validate the use of HRR for using a drop of 12 beats per minute or less as prognosis in a male veteran population.46 The the definition of an abnormal response exhibited mortality rate in our study was higher than in a relative risk of 4.0 for death; the group with a previous studies of HRR. Using similar statistical value lower than 12 having a mortality of 19%, and analysis, we found that HRR of less than 22 beats the group with a value higher than 12 having a per minute at 2-minute recovery identified a high- mortality of 5% over the 6-year period. The study risk group of patients. We also found that employed the symptom-limited Bruce protocol beta-blockers had no significant impact on the prog- with a 2-minute cool-down walk, and HRR was nostic value of HRR. Through multivariate analysis, measured at 1 minute after peak exercise. Patients we evaluated the power of several other clinical on beta-blockers were included in the study, and and treadmill variables to see how they compared in these patients no difference was seen in the to HRR in their ability to predict poor outcome. ability of the test to discriminate between low- Similar to Cole et al,43 we found that low and high-risk groups. In this study, the investi- MET capacity was the most powerful predictor of gators used all-cause mortality and performed mortality. survival analysis both with and without censoring of interventions (CABG and PCI), and found no A distinct advantage over previous studies is difference in results. that we selected a group who underwent coronary angiography. This made it possible to evaluate the The same investigators then studied a different diagnostic ability of HRR. Surprisingly, HRR was patient population.44 Asymptomatic patients not selected among the standard variables to be enrolled in the Lipid Research Clinics Prevalence included in a logistic model and its ROC curve did Study underwent exercise testing using a Bruce not indicate any discriminatory value. Therefore, although, HRR has been validated as an important prognostic variable, it did not help in diagnosing coronary disease in this study. Because in general

TA B L E 8 – 4 . Previous published prognostic studies relating to the decrease of heart rate after exercise Study Population Sample Exclusion Years Test Minutes of Mortality Sensitivity/ Other Beta- size criteria f/u protocol/ recovery/ (all-cause) specificity variables blocker Cole* (% recovery cutpoint for death studied status N Engl J women) status 213 (9%) (cutpoint = Med 1 min/12 bpm 12 bpm) METs, male Used by 12% Referral for 2428 CABG, 6 Bruce, with 56%/77% sex, age, of study angiography, 2-min cool (cutpoint = perfusion population exercise (37) CHF/digoxin down 8 bpm) defects on No use, LBBB symptom- 33%/90% scintigraphy, association perfusion; 9% limited 54%/69% chronotropic with incompetence abnormal test with known CAD 49%/81% No comparison Excluded Cole Ann Participants in 5234 Beta-blockers, 12 Bruce, without 2 min/42 bpm 325 (6.2%), 33%/87% METs, maximal Heart rate C H A P T E R 8 Prognostic Applications of Exercise Testing 265 Intern Lipid Research (39) 36% felt to be HR, recovery not Med Clinics other cardiac cool down cardiovascular 35%/83% Duke treadmill predictive of Prevalence Study, follow-up score; TM AP death in beta- Nishime asymptomatic medications, 85% age 100% score and EI- blocker JAMA ST depression group h/o predicted 312 (3%) not prognostic Data not N/A cardiovascular heart-rate available Used by 34% disease Age, METs, of the study history of population Referral for 9454 CHF, LBBB, 5.2 Bruce, with 2- 1 min/12 bpm typical angina; heart rate ETT; 8% prior (22) TM AP score recovery CABG, 75% digoxin, min cool down and EI-ST equally depression not predictive screening valvular heart symptom- prognostic asymptomatic, 9% prior MI disease limited Watanabe Referral for 5785 CHF, valve 3 N/A 1 min/18 bpm 190 (3.5%) Circulation ETT-ECHO (37) disease, Afib, no cool down 2 min/22 bpm 413 (19%) pacer Shetler Referral for 2193 6.8 Ramp without J Am Coll Standard ETT; (all CABG, cool down Cardiol men) angiography, symptom- 42% with prior LBBB, pacer limited MI

266 E X E R C I S E A N D T H E H E A R T these studies did not censor on events or consider In a cohort study, the Cleveland clinic group event-free survival, heart rate recovery may well examined 2333 male diabetics who underwent a just be a surrogate for physical fitness or activity treadmill test at the Cooper Clinic.49 Hazard level predicting outcome along with medical ther- ratios for CV and all-cause death were adjusted for apy. Our conclusion is that these studies support age, METs, resting heart rate, fasting blood glu- the health benefits of a lifestyle of physical activity cose, body mass index, smoking habit, alcohol con- rather than the addition of HRR to scores designed sumption, lipids, and history of CV disease. During to help direct patients to appropriate therapies. 15 years follow-up, there were 142 deaths that However, it is still an important addition to every were considered CV disease related and 287 total exercise test performed. deaths. Compared with men in the highest quar- tile of HRR, the adjusted hazard ratio was 1.5 to 2 The following studies addressed the issues of for CV death. the ability of HRR to predict CV disease or events rather than all-cause mortality, and its performance Why is Cardiac Catheterization in women and diabetics. Considered to be Superior for Predicting Prognosis? Because no prior study considered the associa- tion of HRR after exercise with the incidence Enthusiasm for cardiac catheterization led to an of coronary heart disease and CV disease, 2967 acceptance of invasive measurements as superior Framingham study subjects (1400 men, mean age to clinical variables for prognostication in patients 43 years) free of CV disease were analyzed.47 During with CAD. Although clinical variables were men- 15 years follow-up, 214 subjects experienced a coro- tioned in the early studies, often, key ones were nary heart disease event, 312 developed a CV disease not considered nor were they considered together event, and 167 died. In multivariable models, con- or defined as accurately as they are today. It was tinuous HRR indexes were not associated with the assumed that laboratory methods and images were incidence of coronary heart disease or CV disease more accurate and precise than simple clinical events, or with all-cause mortality. However, in data. In addition, the importance of clinical data models evaluating quintile-based cutpoints, the could have been underestimated because of the top quintile of HRR (greatest decline in heart nonavailability of modern survival analysis tech- rate) at 1-minute after exercise was associated niques. A further consideration is that the decline with half the coronary heart disease and CV disease in mortality by vessel score recently noted is not as the bottom quintile, but not all-cause mortality. actually due to disease treatment but by patient selection (e.g., excluding patients with CHF A total of 2994 asymptomatic women without because of a better recognition of it). CV disease, 30 to 80 years of age, performed a near- maximal Bruce-protocol treadmill test as part of On the basis of clinical and exercise test data, the Lipid Research Clinics Prevalence Study (1972– patients with signs and symptoms of coronary heart 1976).48 They were followed for 20 years with CV disease can be classified into low- and high-risk cat- and all-cause mortality as the endpoints. There egories. The latter clearly should be considered were 427 (14%) deaths, of which 147 were due to for cardiac catheterization, while the former should CV causes. Low exercise capacity, low HRR, and not not, unless their symptoms dictate otherwise. The achieving target heart rate were independently problem lies in justifying intervention to improve associated with increased all-cause and CV mor- survival for patients whose symptoms are satisfacto- tality. There was no increased CV death risk for rily managed medically. Simple clinical indicators exercise-induced ST-segment depression, but there can stratify these patients with stable CAD into was an age-adjusted 20% increase for every MET high- or low-risk groups. Cardiac catheterization decrement and a 36% increase in CV mortality per is not needed to do so in the majority of such 10 beats per minute decrement in HRR. After patients. Clinical judgment must be applied to adjusting for multiple other risk factors, women decide whether intervention is likely to improve who were below the median for both exercise capac- survival in our high-risk patients. ity and HRR had a 3.5-fold increased risk of CV death. Among women with low-risk Framingham With the number of excellent outcome studies scores, those with below median levels of both that have been completed but with divergent results METs and HRR had significantly increased risk to predict outcome, one must conclude that compared with women who had above median patient population as well as selection has a great levels of these two exercise variables, 45 and four impact upon results. The Duke and VA predictive CV deaths per 10,000 person-years, respectively (hazard ratio of 13).

C H A P T E R 8 Prognostic Applications of Exercise Testing 267 equations appear to be the best and represent the imaging (SPECT) and subtle differences could be “state of the art” in prognostication. plotted and scored. In recent years, ventriculo- grams based on the imaged wall, as apposed to the Does the Addition of Imaging blood in the chambers (as with RNV), could be Techniques Improve constructed. Because of the technical limitations Prognostication? of thallium (i.e., source and lifespan), it has largely been replaced by chemical compounds called isoni- Skeptics may say that the standard exercise test triles that can be tagged with technetium, which must be augmented by either radionuclide imag- has many practical advantages over thallium as an ing or echocardiographic imaging to optimize the imaging agent. The isonitriles are trapped in the ability of the exercise test to predict prognosis. microcirculation permitting imaging of the heart Therefore, a brief review of these technologies is with a scintillation camera. The differences in tech- appropriate. Naturally there has been an evolution nology over the years and the differences in in the technologies available since their inception, expertise at different facilities can complicate the which makes any conclusions regarding their appli- comparisons of the results and actual application cation somewhat of a “moving target.” Nuclear of this technology. techniques were first employed so we will start with them. Echocardiography Nuclear Ventricular Function Assessment The impact of the echocardiogram on cardiology has been impressive. This imaging technique One of the first techniques added to exercise test- comes second only to contrast ventriculography via ing was radionuclear ventriculography (RNV). This cardiac catheterization for measuring ventricular involved the intravenous injection of technetium- volumes, wall motion, and ejection fraction. With tagged red blood cells. Using ECG gating of images Doppler added, regurgitant flows can be estimated obtained from a scintillation camera, images of the as well. Echocardiographers were quick to add this blood circulating within the LV chamber could be imaging modality to exercise, with most studies obtained. While regurgitant blood flow from valvu- showing that supine, post-treadmill assessments lar lesions could not be identified, ejection fraction were adequate, and the more difficult imaging dur- and ventricular volumes could be estimated. The ing exercise was not necessary. resting values could be compared to those obtained during supine exercise and criteria were estab- Available Prognostic Studies lished for abnormal. The most common criteria involved a drop in ejection fraction. The following is a summary of the available liter- ature addressing the added value of these avail- Nuclear Perfusion Imaging able add-ons to the exercise test. Although initially popular, these blood volume Radionuclide Ventriculography. Simari et al50 techniques have come to be surpassed by perfusion evaluated the ability of supine exercise ECG and techniques. The first agent used was thallium, an RNV to predict subsequent cardiac events in 265 isotopic analog of potassium that is taken up at patients with a normal resting ECG and who were variable rates by metabolically active tissue. When not receiving digoxin, but who had undergone taken up at rest, images of metabolically active cardiac catheterization. The Cox model chose ST muscle, such as the heart, are possible. With the depression, exercise heart rate, and patient gen- nuclear camera placed over the heart after intra- der as equivalent to the RNV data. They concluded venous injection of this isotope, images were ini- that exercise RNV was not justified for use over tially viewed using an x-ray film. The normal standard exercise variables. This conclusion was complete donut-shaped images gathered in multi- countered by Lee et al51 in his study of 571 patients ple views would be broken by “cold” spots where a in which RNV provided more prognostic informa- scar was present. Defects viewed after exercise tion than clinical variables. could be due to either scar or ischemia. Follow-up imaging confirmed that the “cold” spots were due Nuclear Perfusion. An elegant review by Brown53 to ischemia if they filled in later. As computer-imag- summarized the available knowledge regarding ing techniques were developed, three-dimensional the prognostic value of nuclear perfusion imag- ing as of 1991. The two most comparable and

268 E X E R C I S E A N D T H E H E A R T comprehensive studies up to that point were Gibbon’s group57 at Mayo Clinic sought to deter- those by Kaul et al53 and Melin et al54, who con- mine whether their previously validated clinical cluded that change in heart rate and other exer- score could identify patients with a low-risk DTS cise test variables were superior to perfusion who had a higher risk of adverse events and, there- defects for prognostication. fore, in whom myocardial perfusion imaging would be valuable for risk stratification. They studied The group at Cedars-Sinai identified 5183 con- 1461 symptomatic patients with low-risk DTS secutive patients who underwent stress/rest nuclear (≥5) who underwent myocardial perfusion imag- perfusion and were followed-up for the occurrence ing. The score was derived by assigning one point to of cardiac death or MI.55 Over a mean follow-up of each of the following variables: typical angina, his- 2 years, 119 cardiac deaths and 158 MIs occurred tory of MI, diabetes, insulin use, male gender, and (3.0% cardiac death rate, 2.3% MI rate). Patients each decade of age over 40 years. A score cut-off with normal scans were at low risk (≤0.5% per equaling or greater than 5 or less than 5 was used year), and rates of both outcomes increased sig- to categorize patients as high risk (n = 303 [21%]) nificantly with worsening scan abnormalities. or low risk (n = 1158 [79%]). Perfusion scans were Patients who underwent exercise testing and had categorized as low, intermediate, or high risk on mildly abnormal scans had low rates of cardiac the basis of a score. High-risk scans were more death but higher rates of MI (0.7% versus 2.6% common in patients with a high-risk score. The per year). After adjustment for prescan informa- score and perfusion score were significant inde- tion, scan results provided incremental prognostic pendent predictors of cardiac death. However, in value toward the prediction of cardiac death. patients with a low score, 7-year cardiac survival Myocardial nuclear perfusion yielded incremental was excellent, regardless of the nuclear perfusion prognostic information toward the identification result (99% for normal and 99% even for severely of cardiac death. Patients with mildly abnormal abnormal scans). In contrast, patients with a high scans after exercise stress are at low risk for car- score had a lower 7-year survival rate (92%), which diac death, but intermediate risk for nonfatal MI, varied with the nuclear perfusion score. They con- and thus may benefit from a noninvasive strategy cluded that in symptomatic patients with low-risk and may not require invasive management. DTS and low clinical risk, myocardial perfusion imaging is of limited prognostic value. In patients A prognostic study was based on 3400 consecu- with low-risk DTS and high clinical risk, annual tive adults undergoing exercise nuclear perfusion cardiac mortality (>1%) is not low, and myocardial testing at the Cleveland Clinic Foundation between perfusion imaging has independent prognostic September 1990 and December 1993; none had value. previous invasive procedures, heart failure, or valve disease.56 Estimated METs, classified by age and French researchers assessed the long-term gender, and perfusion defects, expressed as a stress prognostic value of the ECG and nuclear perfusion extent score on a 12-segment scale, were analyzed exercise tests in a large population of patients with to determine their relative prognostic importance low to intermediate risk of cardiac events.58 They during 2 years of follow-up. Of 3400 patients, 108 followed 1137 patients (857 men, mean age 55 (3.2%) died during follow-up; 32 deaths were years), referred for typical (62%) or atypical (22%) identified as cardiac related. On univariate analy- chest pain, or suspected silent ischemia (16%) for sis, estimated METs was a strong predictor of 6 years. Overall mortality was higher after death, with 62 (57%) deaths occurring in patients strongly abnormal (ST depression >2 mm, or >1 achieving less than 6 METs. On multivariable mm for a workload ≤5 METs) (2.4% per year) analysis, the strongest independent predictors of or nondiagnostic (1.6% per year) bike test than all-cause mortality were low METs (relative risk after normal (0.85% per year) or abnormal (1.4% [RR] of 4×) and age. The presence of perfusion per year) bike test, and after abnormal nuclear defects was a less powerful predictor of death. perfusion than after normal perfusion (1.6% per Cardiac mortality was predicted by decreased year versus 0.7% per year). The major cardiac event METs (RR of 4.4×) and by stress extent score (RR rate (cardiac death or MI) was 0.9%, 1.6%, 2%, and of 1.4×). In this clinically low-risk group, METs 2% per year after negative, positive, strongly posi- was a strong and overwhelmingly important inde- tive, and nondiagnostic exercise test and 0.6%, pendent predictor of all-cause mortality among 1.4%, and 2% per year in patients with 0, 1 to 2, patients undergoing exercise perfusion testing. and equal or greater than 3 abnormal segments The extent of myocardial perfusion defects was on nuclear perfusion, respectively. In multivariate of comparable importance for the prediction of analysis, nuclear perfusion had modest incremental cardiac mortality.

C H A P T E R 8 Prognostic Applications of Exercise Testing 269 prognostic value over clinical and exercise test by cardiac conditions. Nonfatal MI occurred in data for predicting overall mortality and major 19 patients (3%), and 89 patients (14%) underwent cardiac events, and the predictive value of nuclear late coronary revascularization. An abnormal scan perfusion was maintained over 6 years. and SSS were independent predictors of cardiac death (hazard ratio 3.5) and provided incremental A retrospective study was performed in 388 information over clinical and exercise test data. consecutive patients comparing the prognostic utility of perfusion imaging and exercise ECG Echocardiography. At the Mayo Clinic the outcome in patients with an exercise capacity exceeding of 1325 patients after normal exercise echocardio- 7 METs.59 Follow-up was performed at 1.5 years graphy were examined to identify potential pre- for adverse events (n = 40), including revascular- dictors of subsequent cardiac events.62 Endpoints ization, MI, and cardiac death. Of the patients, were overall and cardiac event-free survival. Cardiac 19 had revascularization related to the testing or events were defined as cardiac death, nonfatal MI, their condition and were not included in further and coronary revascularization. Patient charac- analysis; 17 (12%) with abnormal scan and 4 (2%) teristics were analyzed in relation of time to first with normal scan had adverse cardiac events. In cardiac event in a univariate and multivariate man- Cox proportional hazards analysis, an abnormal ner to determine which, if any, were associated scan had a hazard of eight times, while neither with an increased hazard of subsequent cardiac the ECG nor DTS had similar hazards. These con- events. Overall survival of the study group was trary results are probably due to a biased selection significantly better than that of an age- and gen- of patients for scanning and the exclusion of der-matched group obtained from life tables. The patients due to test results leading to interventions. cardiac event-free survival rates at 1, 2, and 3 years were 99.2%, 97.8%, and 97.4%, respectively. The The Cleveland Clinic group evaluated nuclear cardiac event rate per person-year of follow-up perfusion for prediction of all-cause mortality was 0.9%. Subgroups with an intermediate or high when considered along with METs and HRR.60 pretest probability of having CAD also had low They followed 7163 consecutive adults referred for cardiac event rates. Multivariate predictors of sub- symptom-limited exercise nuclear perfusion (mean sequent cardiac events were angina during exercise age 60, 25% women) for 7 years; 855 deaths were testing (RR 4×), low exercise capacity (RR 3×), found to have occurred. Using information the- and echocardiographic LV hypertrophy (RR 2.6×). ory, they identified a probable best model relating The outcome after normal exercise echocardiog- nuclear findings to outcome in order to calculate raphy was good even in those with an intermedi- a prognostic nuclear score. Intermediate- and high- ate or high pretest probability of having CAD. In a risk prognostic nuclear scores were noted in 28% subsequent study, these investigators examined and 10% of patients. Compared with those who the outcomes of 1874 patients with known or sus- had low-risk scans, patients with an intermediate- pected CAD (mean age 64, 64% men) who had good risk score were at increased risk for death (14% exercise capacity (≥5 METs for women, 7 METs for versus 9%, hazard ratio 1.7), while those with high- men) but abnormal exercise echocardiograms and risk scores were at greater risk (24%, hazard ratio analyzed the association between clinical, exercise, 3). Impaired exercise capacity and decreased HRR and echocardiographic variables and subsequent provided additional prognostic information to the cardiac events.63 Multivariate predictors of time to nuclear perfusion score. cardiac death or nonfatal MI were diabetes mellitus (RR 2×), history of MI (RR 2.4×), and an increase To assess the incremental value of exercise or no change in LV end-systolic size in response to nuclear perfusion imaging for the prediction of car- exercise (RR 1.6×). Using echocardiographic vari- diac events in patients with known or suspected ables that were of incremental prognostic value, CAD, follow-up was performed in 648 patients.61 they were only able to further stratify the cardiac Ten patients underwent early coronary revascu- risk from 1.6% for patients who had a decrease larization, and seven lost to follow up were in LV end-systolic size in response to exercise (n = excluded. Endpoints were cardiac death, nonfatal 1330) to 1.2% for patients with normal exercise infarction, and late (>60 days) coronary revascular- ECHO (n = 868). ization. An abnormal study was defined as the pres- ence of fixed and/or reversible perfusion defects. A Marwick et al64 collected clinical and exercise summed stress score (SSS) was derived to estimate testing data on 5375 patients (aged 54 years, 60% the extent and severity of perfusion defects. An men) referred for exercise echocardiography and abnormal scan was detected in 344 patients followed them up for 6 years; 649 patients died. (54%). During a mean follow-up period of 4 years, 56 patients (9%) died; 22 of the deaths were caused

270 E X E R C I S E A N D T H E H E A R T The Duke score classified 59% of patients as low COMPARISON OF PREDICTION risk, 39% as intermediate risk, and 2% as high risk. Resting LV dysfunction was present in 27% EQUATIONS WITH and the exercise echocardiogram was abnormal in 47%. Those with normal exercise echocardio- CARDIOLOGISTS grams had a mortality of 1% per year. Ischemia was an independent predictor of mortality. In To study the accuracy with which long-term prog- sequential Cox models, the predictive power of nosis can be predicted in patients with CAD, prog- clinical data was strengthened by adding the Duke nostic predictions from a data-based multivariable score, resting LV function, and the results of exer- statistical model were compared with predictions cise echocardiography. Exercise echocardiography from senior cardiologists.66 Test samples of was able to substratify patients with intermediate- 100 patients each were selected from a large series risk Duke scores into groups with a yearly mortal- of medically treated patients with significant coro- ity of 2% to 7%. They concluded that a normal nary disease. Using detailed case summaries, five exercise echocardiogram confers a low risk of senior cardiologists each predicted 1- and 3-year death and that the ECHO add-on was particularly survival and infarct-free survival probabilities for useful in patients with intermediate-risk DTS. 100 patients. Fifty patients appeared in multiple However, they did not consider CV death or MI as samples for assessing interphysician variability. endpoints. Cox regression models, developed using patients not in the test samples, predicted corresponding Elhendy et al65 studied 5679 patients (aged 62; outcome probabilities for each test patient. Overall, 3231 men) who were followed for a mean of 3 years model predictions correlated better with actual after treadmill echocardiography. Patients were patient outcomes than did the doctors’ predictions. randomly divided into a modeling and a training For 3-year survival, rank correlations were 0.61 group. The modeling group underwent multivari- (model) and 0.49 (doctors). For 3-year infarct-free ate analysis to define independent predictors of survival predictions, correlations with outcome mortality. Three hundred bootstrap resamplings were 0.48 (model) and 0.29 (doctors). Comparisons were performed to determine parameter coeffi- by individual doctor revealed Cox model 3-year cients. Patients were divided into five risk cate- survival predictions were better than those of four gories according to their composite score. The of five doctors (model predictions added signifi- validation group comprised patients for whom the cant [p < 0.05] prognostic information to the doc- risk model was applied. Patients were divided into tor's predictions, whereas the converse was not five risk categories based on data obtained from the true). For infarct-free survival, the Cox model was modeling group. During follow-up, 315 patients superior to all five doctors. In cases where multiple died (151 in the modeling group). Independent doctors made predictions, the interphysician vari- predictors of mortality were exercise wall motion ability was substantial. In CAD, statistical models score index (first), workload (second), male gender developed from carefully collected data can provide (third), and age (fourth). Application of the compos- prognostic predictions that are more accurate ite score in the validation group resulted in an effec- than predictions of experienced clinicians, made tive stratification of patients for mortality and from detailed case summaries. cardiac events. However, it is important to note that exercise capacity and basic demographics provide A computer algorithm for estimating probabil- the majority of the prognostic information. ities of any significant coronary obstruction and triple-vessel/left main obstructions was derived, Summary. What do these findings mean to the cli- validated, and compared with the assessments of nician? First, it should be noted that all studies cardiac clinician angiographers.67 The algorithm have population-specific attributes that may be dif- performed at least as well as the clinicians when ficult to define. Nevertheless, if the aim is to predict the latter knew the identity of the patients whose infarct-free survival, the DTS is preferred because angiograms they had decided to perform. The cli- censoring was performed and infarct-free survival nicians were more accurate when they did not was predicted. All of the findings strengthen the know the identity of the subjects but worked from importance of exercise capacity—a reflection of tabulated objective data. Referral and value-induced the integrity of the cardiopulmonary system and a bias may affect physician judgment in assessing dis- marker of a physically active lifestyle—as an impor- ease probability. Application of computer aids tant predictor of survival along with, or in spite of, or consultation with cardiologists not directly modern medical treatment. involved with patient management may assist in more rational assessments and decision-making.

C H A P T E R 8 Prognostic Applications of Exercise Testing 271 With these two papers as background, we used sample of 2342 patients referred to Duke since our database to compare exercise test scores and 1983. The model also accurately estimated the ST measurements with a physician’s estimation prevalence of severe CAD in a large series of patients of the probability of the presence and severity of reported in the literature. Hubbard et al71 per- angiographically significant CAD and the risk of formed a similar study from the Mayo Clinic. Five death.68,69 A clinical exercise test was performed variables were found to be predictive of severe dis- and an angiographic database was used to print ease: age, gender, diabetes, typical angina, and his- patient summaries and treadmill reports. The clin- tory of prior MI. An international cross-validation ical and treadmill test reports were sent to expert study was performed by Detrano et al72 conclud- cardiologists and to two other groups, including ing that use of their algorithm could avert at least randomly selected cardiologists and internists. 10 angiograms on patients with less severe dis- They classified the patients summarized in the ease for every missed case of severe disease. These reports as having a high-, low-, or intermediate studies demonstrate that the clinician’s initial probability for the presence of any and also severe evaluation even without a treadmill test can iden- angiographically significant disease using a tify patients at high or low risk of anatomically numerical probability from 0% to 100%. The severe CAD. These important studies emphasize Social Security Death Index was used to determine that cost-conscious quality care can be accom- survival status of the patients. Of the patients, 26% plished by consideration of simple clinical vari- had severe angiographically significant disease, ables to identify patients at higher risk for severe and the annual mortality rate for the population CAD who are most likely to benefit from further was 2%. Forty-five expert cardiologists returned evaluation. estimates on 473 patients, 37 randomly chosen practicing cardiologists returned estimates on 202 Using Exercise Test Responses patients, 29 randomly chosen practicing internists returned estimates on 162 patients, 13 academic Studies have tried to predict left main disease cardiologists returned estimates on 145 patients, using exercise testing.17,73-74 Different criteria have and 27 academic internists returned estimates on been used with varying results. Predictive value 272 patients. When probability estimates for pres- here refers to the percentage of those with the ence and severity of angiographically significant abnormal criteria that actually had left main dis- disease were compared, in general, the treadmill ease. Naturally, most of the “false positives” actu- scores were superior to physicians’ and ST analysis ally had CAD but in a less severe form. Sensitivity at predicting severe angiographic disease. When here refers to the percentage of those with left prognosis was estimated, treadmill prognostic main disease only that are detected. These criteria scores did as well as expert cardiologists and better have been refined over time and the last study by than most other physician groups. Weiner et al18 using the CASS data deserves fur- ther mention. Weiner et al18 defined a markedly PREDICTING SEVERE positive exercise test in a study of 436 consecutive ANGIOGRAPHICALLY patients referred for suspected or known CAD SIGNIFICANT CAD who were able to undergo both exercise testing and coronary angiography. All patients underwent Using Clinical Variables treadmill testing using the Bruce protocol, and 12-lead ECG were obtained during exercise. A Pryor et al70 examined clinical characteristics pre- lesion of the left main coronary artery was consid- dictive of severe disease in 6435 consecutive ered significant if it had greater than 50% diameter symptomatic patients referred for suspected CAD narrowing and this criterion was 70% in other between 1969 and 1983. Eleven of 23 characteris- vessels. Fifty-five patients were excluded because tics were important for estimating the likelihood of LV hypertrophy, digoxin therapy, left bundle of severe angiographically significant disease. branch block, and for the attainment of less than These included chest pain type, previous MI, age, 85% maximal predictive heart rate. Of these sex, duration of chest pain symptoms, risk factors, 55, two had left main CAD, and four had three- carotid bruit, and chest pain frequency. A model vessel disease, therefore the predictive value of using these characteristics accurately estimated being excluded was about 10%. Four patient the likelihood of severe disease in an independent groups were defined by angiographic findings: (1) 35 with left main disease, (2) 89 with three-vessel

272 E X E R C I S E A N D T H E H E A R T disease without left main disease, (3) 188 patients highly predictive and reasonably sensitive for left with either one-or two-vessel disease, and (4) main or three-vessel coronary disease. The ques- 124 patients with no significant coronary disease. tion still remains of how to identify those with Of the 35 patients with left main disease, most had abnormal resting ejection fractions, those that disease of other coronary arteries and nearly half will benefit the most with prolonged survival after had three-vessel disease. Exercise test responses CABS. Perhaps those with a normal resting ECG that were considered included the amount of will not need surgery for increased longevity ST-segment depression, configuration, onset, and because of the associated high probability of normal duration, and the number of leads in which it ventricular function. occurred. Hemodynamic responses included treadmill time, SBP, and maximal heart rate. Other Blumenthal et al75 validated the ability of a measurements included angina, PVCs, and abnor- strongly positive exercise test to predict left main mal R-wave response in lead V5. coronary disease even in patients with minimal or no angina. The criteria for a markedly positive Nearly all patients with left main disease had at test included: (1) early ST-segment depression, least 0.l mV of ST depression and 91% had 0.2 mV (2) 0.2 mV or more of depression, (3) downsloping or more of ST-segment depression. Patients with ST depression, (4) exercise-induced hypotension, left main disease as a group were distinguished (5) prolonged ST changes after the test, and from patients with three-vessel disease by an early (6) multiple areas of ST depression. onset and longer persistence of ST-segment depres- sion, as well as by a greater number of leads in While Lee et al76 included many clinical and which the depression occurred. A fall in SBP exercise test variables, only three variables were occurred in 23% of the patients with left main found to help predict left main disease: angina disease versus 17% of those with triple-vessel dis- type, age, and the amount of exercise-induced ST- ease and 6% of those with single- or double-vessel segment depression. Using a Bayesian approach, disease. As an indicator of either left main or three- the pretest likelihood of left main disease was best vessel disease, a fall in SBP had a predictive value determined by the type of angina and age. In spite of 66% and a sensitivity of 19%. The criterion of of the many clinical markers considered, such as 0.3 mV or more of ST-segment depression occurred unstable angina, history of MI, and others, only in 44% of such patients and had only a slightly age and the angina type were found best to predict lower predictive value (64%). Combined analysis of pretest probability of disease. The only exercise test variables (i.e., a markedly abnormal response, test variable that was found to then improve the Table 8-5) disclosed that the development of 0.2 mV post-test probability was the amount of ST-segment or more of downsloping ST-segment depression depression. There is a low pretest probability of beginning at 4 METs, persisting for at least 6 min- left main disease in 40-year-old men with atypical utes into recovery, and involving at least five ECG angina and a high pretest probability of left main leads had the greatest sensitivity (74%) and pre- disease in older men with typical angina. Given dictive value (32%) for left main coronary disease. a pretest probability of 50%, for example, the This abnormal pattern identified either left main post-test probability could range from 20% to 75% or three-vessel disease with a sensitivity of 49%, a according to the degree of ST-segment depression. specificity of 92%, and a predictive value of 74%. The problem with using the amount of depres- It appears that individual clinical or exercise test sion as the sole predictor is that in many exercise variables are unable to detect left main coronary labs, an exercise test is stopped at 2 mm of ST disease because of their low sensitivity or predictive depression for safety reasons or because of severe value. However, a combination of the amount, pat- angina. In addition, some physicians stop the test at tern, and duration of ST-segment response was an age-predicted maximal heart rate. Surprisingly, exercise-induced hypotension and exercise dura- tion did not impact on post-test probability in their analysis. TA B L E 8 – 5 . The markedly positive responses Meta-Analysis of Studies identified by Weiner Predicting Angiographic Severity >0.2 mV downsloping ST-segment depression To evaluate the variability in the reported accu- Involving five or more leads racy of the exercise ECG for predicting severe coro- Occurring at less than 5 METs nary disease, Detrano et al77 applied meta-analysis Prolonged late into recovery

C H A P T E R 8 Prognostic Applications of Exercise Testing 273 to 60 consecutively published reports comparing numerical score, while a logistic regression pro- exercise-induced ST depression with coronary vides an actual probability; this, however, may vary angiographic findings. The 60 reports included from one population to another. 62 distinct study groups comprising 12,030 patients who underwent both tests. Both technical and Logistic regression results in an equation that methodologic factors were analyzed. Wide variabil- takes the form: ity in sensitivity and specificity was found (mean sensitivity 86% [range 40% to 100%]; mean speci- Probability = 1 / (1 + e − (a + bx + cy … )) ficity 53% [range 17% to 100%]) for left main or triple-vessel disease. All three variables found to where a is the intercept; b and c are coefficients; be significantly and independently related to test x and y are variable values such as 0 or 1 for gender, performance were methodological. Exclusion of diabetes, or chest pain; and there is a continuous patients with right bundle branch block or who value for age or heart rate. were receiving digoxin improved the prediction of triple vessel or left main CAD and comparison with Studies Using Multivariate Techniques a “better” exercise test decreased test performance. to Predict Severe Angiographically Significant CAD Hartz et al78 compiled results from the litera- ture on the use of the exercise test to identify Since 1979, 13 studies reported combining the patients with severe CAD. Pooled estimates of sen- patient’s medical history, symptoms of chest pain, sitivity and specificity were derived for the ability hemodynamic data, exercise capacity, and exer- of the exercise test to identify three-vessel or left cise test responses to calculate the probability of main CAD. One millimeter criteria averaged a severe angiographic CAD.72,80-89 The results are sum- sensitivity of 75% and a specificity of 66% while marized in Table 8-6. Of the 13 studies, 9 excluded two millimeters criteria averaged a sensitivity of patients with previous CABS or prior PCI and in 52% and a specificity of 86%. There was great vari- the remaining 4 studies, exclusions were unclear. ability among the studies examined in the esti- The definition of significant percentage stenosis mated sensitivity and specificity for severe CAD for angiographically significant disease ranged that could not be explained by their analysis. from 50% to 70%. The percentage of patients with one-, two-, and three-vessel disease was described in Multivariable Equations and 10 of the 13 studies. The definition of severe disease Scores to Predict Severe Angiographically Significant CAD TA B L E 8 – 6 . A summary of the results from the 13 studies (14 equations) predicting disease severity The most common statistical methods employed include Bayesian statistics, logistic regression, Clinical variables Significant predictors and discriminant function analysis. The Bayesian approach, which considers pretest clinical variables, Gender 7/9 78% is a logical method in clinical practice and helps one Chest pain symptoms 8/11 73% decide which tests are appropriate. However, it Diabetes mellitus 6/10 60% appears that logistic regression or discriminant Age 8/14 57% function analysis permits a more robust predic- Abnormal resting ECG 4/8 50% tion of disease. Elevated cholesterol 4/10 40% Family history of CAD 1/4 25% Multivariable analysis is a statistical tech- Smoking history 2/8 25% nique that seeks to separate subjects into different Hypertension 1/6 17% groups on the basis of measured variables.79 Clinical investigators have commonly used two Exercise test variables 11/14 79% types of analysis: discriminate function and logistic 6/8 75% regression analysis. Logistic regression has been ST-segment depression 4/7 57% preferred since it models the relationship to a sig- ST-segment slope 5/11 45% moid curve (which often is the mathematical rela- Double product 4/13 31% tionship between a risk variable and an outcome) Delta systolic BP 4/13 31% and its output is between 0 and 1 (i.e., from 0% to Exercise capacity 1/10 10% 100% probability of the predicted outcome). Thus, Exercise induced angina 0/4 the output of a discriminate function is a unitless Maximal HR 0% Maximal systolic BP

274 E X E R C I S E A N D T H E H E A R T or disease extent (multivessel versus three-vessel resting ECG abnormalities and diabetes were the or left main artery disease) also differed. In 5 of only other variables chosen in more than half the the 13 studies disease extent was defined as multi- studies. In contrast, the most consistent clinical vessel disease (i.e., more than one vessel involved). variables chosen for diagnosis were: age, gender, In the remaining 8 studies, it was defined as chest pain type, and hypercholesterolemia. ST three-vessel or left main disease and in one of them depression and slope were frequently chosen for as only left main artery disease and in another the severity, but METs and heart rate were less consis- impact of disease in the right CAD on left main tently chosen than for diagnosis. Double product disease was considered. The prevalence of severe and delta SBP were chosen as independent predic- disease ranged from 16% to 48% in the studies tors in more than half of the studies predicting defining disease extent as multivessel disease and severity. from 10% to 28% in the studies using the more strict criterion of three-vessel or left main disease. History of Myocardial Infarction as a Clinical Predictor. Although it makes little sense to consider Not all of the publications of the reviewed stud- a history of MI or Q-wave evidence for MI in stud- ies included the equations derived from the multi- ies dealing with diagnosis, there is some justifica- variable analyses they performed. These equations tion for considering them in studies dealing with are critical to the validation of their findings.90 disease severity. This variable has been defined The actual equations developed in the studies in numerous ways, including patient history (or were available for only 4 of the 13 studies predict- chart review) or by review of resting ECG for Q ing disease extent or severity. waves. One coding scheme called for a 1 if by his- tory only, 2 if diagnostic Q waves were present, Some notable results were obtained in 1 of the and 3 if both history and Q waves were present. 13 studies that did not produce a score because The amount of LV damage (considered to be the discriminate function analysis was utilized. Ribisl result of severe coronary disease) has been esti- et al91 studied 607 male patients to determine mated in some studies by a Q-wave score or by whether patterns and severity of CAD could be summing the number of diagnostic Q waves. This predicted using standard clinical and exercise test variable was a significant predictor in 2 of the data. Left main disease produced responses signif- 8 studies that considered it. Due to the inaccuracy icantly different from three-vessel disease only of historical data alone, emphasis should also be when accompanied by a 70% or greater narrowing given to objective measures such as diagnostic of the right coronary artery. The maximum amount criteria for Q waves for a prior infarction. of horizontal or downsloping ST depression in exer- cise and/or recovery was the most powerful pre- Consensus or Agreement to dictor of disease severity, with 2-mm ST depression Improve Prediction yielding a sensitivity of 55% and specificity of 80% for prediction of severe CAD (three-vessel Only two of the studies (Detrano et al72 and plus left main disease). Patients with increasingly Morise et al85) have published equations that have severe disease also demonstrated a greater fre- been validated in large patient samples. Although quency of abnormal hemodynamic responses to validated, the equations from these studies must exercise. It appears that the exercise test will best be calibrated before they can be applied clinically. distinguish left main or left main equivalent dis- For example, a score can be discriminating but ease only when there is significant disease in the provide an estimated probability that is higher or right coronary artery (i.e., similar to three-vessel lower than the actual probability. The scores can disease). Otherwise, the exercise responses are sim- be calibrated by adjusting them according to dis- ilar to patients with two-vessel disease. The exercise ease prevalence; most clinical sites, however, do not test did not function worse in patients selected for know their disease prevalence and even if known, beta-blocker administration and that standard it could change from month to month. ST analysis outperforms the ST/HR index in either situation.92 At the National Aeronautics and Space Administration (NASA), trajectories of spacecraft Chosen Predictors. Surprisingly, some of the are often determined by agreement between three variables chosen for predicting severe disease are or more equations calculating the vehicle path. different than those for predicting disease pres- With this in mind, we developed an agreement ence for diagnosis. While gender and chest pain method to classify patients into high, no agree- were chosen to be significant in more than half of ment, or low-risk groups for probability of severe the severity studies, age was less important and

C H A P T E R 8 Prognostic Applications of Exercise Testing 275 disease by requiring agreement in all three equa- survival after CABS. Their data came from 2000 men tions (Detrano, Morise and ours [Long Beach and with coronary heart disease enrolled in the Seattle Palo Alto]).93 This approach adjusts the calibration Heart Watch who had a symptom-limited maxi- and makes the equations applicable in clinical mal treadmill test; these subjects received usual populations with varying prevalence of CAD. community care, which resulted in 16% of them having CABS in nonrandomized fashion. The We demonstrated that using simple clinical diagnosis of coronary heart disease was based and exercise test variables could improve the stan- on a history of angina, MI, or cardiac arrest. dard application of ECG criteria for predicting Cardiomegaly was determined by physical and severe CAD. By setting probability thresholds for chest x-ray examinations. The patients were divided severe disease at less than 20% and greater than into three groups. One group had only myocardial 40% for the three prediction equations, the agree- ischemia manifested by exercise test-induced nor- ment approach divided the test set into popula- mal ST-segment elevation or depression and/or tions with low risk, no agreement, and high risk for angina. The second group could have myocardial severe CAD. Since the patients in the no-agreement ischemia, but had to have “LV dysfunction” mani- group would be sent for further testing and would fested by at least two of the following: cardiomegaly, eventually be correctly classified, the sensitivity of less than 4 METs exercise capacity, and less than the agreement approach was 89% and the speci- 130 mmHg maximal SBP. A third group had none ficity was 96%. The agreement approach appeared of the above. Comparisons were then made within to be unaffected by disease prevalence, missing each group between the operated and unoperated data, variable definitions, or even by angiographic patients and surprisingly little difference was found. criterion. Cost analysis of the competing strate- However, life table analysis showed a significantly gies revealed that the agreement approach com- higher survival rate of 94% at 4 years among the pares favorably with other tests of equivalent operated patients, as compared with the 68% sur- predictive value, such as nuclear perfusion imag- vival of the unoperated patients in the group with ing, reducing costs by 28%, or $504, per patient LV dysfunction. If the 4.6% death rate due to sur- in the test set. gery in those with “ischemia” only was reduced, perhaps the patients who were operated on in that Requiring diagnosis of severe coronary disease group would have had a significantly improved to be dependent on agreement between these three survival as well. Thus, patients with cardiomegaly, equations has made them likely to function in all less than 5 MET exercise capacity and/or a maxi- clinical populations. Excellent predictive charac- mal SBP of less than 130 mmHg would have a bet- teristics can be obtained using simple clinical data ter outcome if treated with surgery. Two or more entered into a computer. Cost analysis suggests of the above parameters present the highest risk that the agreement approach is an efficient method and the greater differential for improved survival for the evaluation of populations with varying with bypass. In this group, 4-year survival would prevalence of CAD, limiting the use of more expen- be 94% for those who had surgery versus 67% for sive noninvasive and invasive testing to patients those who received medical management (in those with a higher probability of left main or three-ves- who had two or more of the above factors). In the sel CAD. This approach provides a strategy for European surgery trial,95 patients who had an assisting the practitioner in deciding when fur- exercise test response of 1.5 mm of ST-segment ther evaluation is appropriate or interventions depression had improved survival with surgery. indicated. This also extended to those with baseline ST seg- ment depression and those with claudication. PREDICTING IMPROVED SURVIVAL WITH CORONARY ARTERY BYPASS From the CASS study group,96 in more than SURGERY 5000 nonrandomized patients, although there were definite differences between the surgical and non- Which exercise test variables indicate those surgical groups, this could be accounted for by patients who would have an improved prognosis if stratification in subsets. The surgical benefit they underwent CABS? The limitation of the regarding mortality was greatest in the 789 patients available studies is that the patients were not with 1-mm ST-segment depression at less than randomized to surgery according to their exer- 5 METs. Among the 398 patients with triple-vessel cise test results and the analysis is retrospective. disease with this exercise test response, the 7-year survival was 50% in those medically managed Bruce et al94 demonstrated noninvasive screen- versus 81% in those who underwent CABS. ing criteria for patients who had improved 4-year

276 E X E R C I S E A N D T H E H E A R T There was no difference in mortality in patients appears, in the elderly, data is limited. To deter- able to exceed 10 METs exercise capacity. From mine the prognostic value of the treadmill test in the VA CABS randomized trial, Hultgren et al97 the elderly, researchers from the Mayo Clinic and reported a 79% survival rate with CABS versus the Olmsted Medical Group compared the prog- 42% for medical management in patients with two nostic value of the test in patients less than 65 and or more of the following: 2 mm or more of ST older than 65 years of age.99 Elderly (n = 514) and depression, heart rate of 140 or greater at 6 METs, younger (n = 2593) patients who underwent and/or exercise-induced PVCs. The results from treadmill testing between 1987 and 1989 were iden- those four studies are summarized in Table 8-7. tified retrospectively and followed-up for 6 years. Compared to younger patients, elderly patients SPECIALIZED SITUATIONS FOR had more comorbid conditions, a higher preva- PREDICTING POOR PROGNOSIS lence of abnormal ST depression (28% versus 9%) AND/OR SEVERE CAD and achieved lower workloads (6 versus 11 METs). A poor exercise capacity and angina during the exer- • In the elderly cise test were associated with future cardiac events. • In diabetic patients and those with silent Exercise-induced ST depression did not carry sig- nificant value in the elderly and was associated ischemia with future cardiac events only in younger patients. An increase of 1 MET in the workload was associ- The Elderly ated with a 14% decrease in risk for a cardiac event in younger patients and with an 18% risk reduc- The decline in function that accompanies aging tion among the elderly. After adjustment for clin- is a consequence of age-related decrements in CV, ical factors, there was a strong inverse association pulmonary, and musculoskeletal structure. between exercise capacity and outcome. METs was Ultimately, these result in impaired physical the only treadmill exercise-testing variable that function in the elderly.98 Whereas the DTS was val- provided prognostic information for mortality and idated in patients in the age range when CAD first cardiac events. In the elderly, exercise capacity was also inversely associated with the likelihood of TA B L E 8 – 7 . Studies evaluating exercise test nursing home placement. Spin et al100 also demon- responses indicate improved survival with coronary strated the strong association between METs esti- artery bypass surgery mated from exercise testing and all-cause mortality in the elderly. Study Markers of improved survival with coronary artery bypass surgery Kwok et al101 found that the DTS could not pre- dict death, MI, and cardiac interventions in patients Seattle heart • Cardiomegaly 75 years or older. Lai et al102 considered both death watch • Less than 5 METs exercise and angiographic endpoints and found age-specific scores to be necessary in the elderly. Given this European surgery capacity last study, we entered the DTS and age into the trial • Maximal systolic blood Cox analysis and found them to have similar coeffi- cients but opposite sign so that a new score equa- Coronary artery pressure less than 130 tion was expressed as DTS minus age. Thus, age surgery study • ST-segment depression at rest was as strong a prognostic predictor as the DTS in (CASS) • 1.5 mm of ST-segment our population. A score of DTS minus age pro- vided a significant improvement in area under the Veterans affairs depression with exercise curve compared to DTS alone in the whole popu- coronary artery • Claudication lation and the subset of younger subjects, but bypass surgery • 1 mm of ST-segment there was no improvement in the elderly. study depression at less than Why do the exercise test variables other than 5 METs METs not provide prognostic information in those • No difference if 10 METs over 75 years of age? Possibly it is due to the many exceeded competing causes of mortality in the elderly com- Two or more of the following: pared to younger subjects, who are more likely to • 2 mm of ST-segment die of one cause. It is also possible that the elderly depression are survivors who, for instance, have coronary dis- • Heart rate less than 140 at ease but have extensive collaterals that protect 6 METs • Exercise-induced premature ventricular contractions

C H A P T E R 8 Prognostic Applications of Exercise Testing 277 them from death, but not ischemia. Reduced exer- was independently associated with CV mortality cise capacity in the elderly is partially explained by only in those 45 to 55 years of age. The failure of the high prevalence of coexisting medical problems, DTS to have prognostic value in our population such as deconditioning, muscle weakness, orthope- remains a mystery since in the very same popula- dic problems, neurological problems, and periph- tion it is one of the important predictors for the eral vascular disease. Elderly patients are also more presence of angiographic disease.103 Results of this likely to have a nondiagnostic exercise ECG because study are provided in Table 8-8. of the greater prevalence of resting ECG abnormal- ities. These factors could confound the association Our study considered a large number of patients between exercise test responses and outcomes. who underwent treadmill testing for clinical indica- tions in a general hospital or clinic setting. Patients To further study this issue, we classified our with prior MI and/or coronary artery revasculariza- patients into subsets based on age. METs were tion were excluded from the study, leaving chosen by the Cox hazard model most consistently 3745 male veterans. Exercise testing variables in the age groups using either endpoint. Even were analyzed within the age groups to evaluate when age was added to the DTS, prediction of death the effect of age and the choice of outcome, CV, or did not improve in those over 70 years of age all-cause death. Our results show the importance of because of the nonlinear relationship between age, age and the endpoint used in the Cox hazard analy- the exercise test variables, and time to death. The ses to develop prognostic scores. We also showed most important age cutpoints for clinically impor- that age has a nonlinear relationship to the tant differences in exercise test predictors appeared variables, and outcomes such as adding age to to be 70 and 75 years of age. In the patients 70 to scores does not improve prediction in the elderly. 75 years of age, METs was the only variable pre- This is most likely because other clinical predic- dictive of all-cause mortality and exercise-induced tors (comorbidities, psychosocial factors, and sub- ST depression was the only predictor of CV death; clinical conditions) overpower the treadmill in the patients older than 75 years of age, none of responses in the elderly even in a population such the exercise test responses were predictive of either as ours, in which patients with recognized heart death outcome (Table 8-8). None of the treadmill disease were removed. Both age and the outcome variables were selected as a predictor of outcome selected as an endpoint affect the exercise test in those 45 years old or younger. This is probably responses chosen for scores to predict prognosis. due to the small number of deaths and our lack of Differences in age of the subjects tested and/or data regarding cardiac interventions during follow- the outcome selected as the endpoint can explain up. Exercise-induced ST depression was signifi- the differences in the studies using exercise testing cantly more prevalent in those who died, but it to predict prognosis. TA B L E 8 – 8 . Results of cox-hazard model with cardiovascular mortality as the endpoint for age groupings illustrating how the predictive power of the treadmill responses change with age Age <45 45-55 55-65 65-75 >75 Regression coefficient/ N 619 987 1081 717 174 Hazard ratio CV death 8 35 72 77 22 95% Cl hazard METs −0.22 −0.18 −0.13 −0.13 NS 0.81 0.83 0.87 0.87 Regression coefficient Exercise-induced 0.67–0.97 0.75–0.93 0.81–0.94 0.80–0.95 NS Hazard ratio ST depression NS 0.61 NS NS 95% CI hazard 1.85 Duke Angina NS 1.39–2.46 NS NS NS 1st in Cox score NS NS NS NS 2nd in Cox Max SBP NS 3rd in Cox 0.014 NS NS NS Max HR NS 1.01 1.00–1.03 Resting ST NS depression NS NS NS 0.81 NS 2.27 1.26–4.07 Modified from Yamazaki T, Myers J, Froelicher VF. Effect of age and end point on the prognostic value of the exercise test. Chest 2004;125:1920-1928.

278 E X E R C I S E A N D T H E H E A R T Exercise myocardial perfusion was evaluated in Preliminary studies led to the hypothesis that elderly patients with interpretable exercise ECG “silent” myocardial ischemia had a worse prognosis tests by considering clinical, ECG, scan, and fol- than angina pectoris because patients with it do low-up data for 626 outpatients aged 65 years or not have an intact “warning system.” However, in older with interpretable ECGs between 1992 and studies of patients referred for diagnostic purposes 1996.104 Follow-up was for 4 years. After exclusion or with stable coronary syndromes, silent myocar- of the 27 patients who underwent revasculariza- dial ischemia detected by exercise testing has tion within 90 days, there were 361 men and been associated with either a lesser or similar prog- 217 women with a mean age of 70. By univariate nosis compared to patients with angina pectoris. analysis, numerous variables (including male Because exercise testing has advantages over ambu- gender, age, rest ECG, poor exercise capacity, latory monitoring with regard to the leads moni- peak heart rate, and exercise ST-segment depres- tored, chest pain description, and fidelity of the sion) predicted death or MI. By multivariable recording apparatus, confirmation of these findings modeling, only increasing patient age, male sex, would help resolve the controversy over the relative poor exercise capacity, and the number of ischemic prognostic impact of silent myocardial ischemia. scan segments were predictive of subsequent death Exercise testing studies give us one means of eval- or MI. uating the risk of silent ischemia. Unfortunately these exercise test studies do not evaluate patients In Diabetics and Those with with true silent ischemia. The patients are being Silent Ischemia tested because of some symptoms, usually angina, although they may not have angina at the time These two situations are discussed together of their test. However, patients with true silent because of the widespread belief that silent ischemia are rare. Therefore, the following data ischemia is more common in diabetics. An open from exercise test studies gives us a good idea of mind should be taken in this regard, however, how the usual patients seen in clinical practice since the basis of evidence for this belief is weak. with silent ischemia, at least in some occasions, are likely to perform. Silent Ischemia during Exercise Testing Ellestad and Wan16 reported the predictive The interest in silent ischemia (i.e., ST depression implications of maximal exercise testing in 2700 without anginal symptoms) has come about individuals followed for 6 months to 9 years. ST because of five clinical observations: (1) the depression and prior MI were both associated with increased risk of coronary events when screening subsequent higher mortality. From the CASS reg- asymptomatic men, (2) the frequency of painless istry of patients who underwent coronary angiog- ST-segment depression during exercise testing in raphy and exercise testing and were followed up patients with coronary heart disease, (3) episodes for 7 years, the significance of ischemic ST segment of painless ST-segment depression noted during depression without associated chest pain during Holter ambulatory monitoring, (4) the clinical exercise testing was studied.105 Of the 2982 patients, impression that silent ischemia is more common those with proven CAD were grouped according to in diabetics, and (5) the apparent high risk of whether they had at least 1 mm of ST-segment painless ST-segment depression in patients with depression or anginal chest pain during exercise unstable ischemic syndromes. Potential dangers of testing: 424 had ischemic ST depression without silent ischemia include asymptomatic progression angina, 232 had angina but no ischemic ST depres- to sudden death and myocardial fibrosis (leading sion, 456 had both ischemic ST depression and to CHF) due to lack of a warning mechanism. angina, and 471 had neither ischemic ST depres- sion nor angina. The 7-year survival rates were As for many other clinical syndromes, dividing similar for patients in all groups (77%), except for silent ischemia into subsets can be very helpful. patients without ST depression or angina, who The types of silent ischemia described by Cohn did better (88%). Among silent ischemia patients, are particularly useful: survival was related to severity of CAD. The 7-year survival rate was significantly worse than that in • Type I—occurring in asymptomatic, appar- a separate group of 282 patients with ischemic ST ently healthy individuals depression but without angina during exercise testing who had no CAD (95% survival). This • Type II—occurring in patients after an MI study demonstrated that in patients with silent myocardial ischemia during exercise testing the • Type III—occurring in patients with known CAD

C H A P T E R 8 Prognostic Applications of Exercise Testing 279 extent of CAD and the 7-year survival rate were sim- pectoris (difference not significant), and 98% and ilar to those of patients with angina during exercise 99%, respectively, for the control patients. Among testing. Prognosis was determined primarily by patients with silent ischemia the probability of the severity of CAD. remaining free of MI and sudden death at 7 years was related to the severity of CAD and presence of At Duke, Marks et al106 evaluated the clinical cor- LV dysfunction, and ranged from 90% for patients relates and long-term prognostic significance of with one-vessel CAD and preserved LV function to silent ischemia during exercise. They analyzed 1698 38% for patients with three-vessel CAD and abnor- consecutive symptomatic patients with CAD who mal LV function. Thus, patients with either silent had both treadmill testing and cardiac catheteriza- or symptomatic ischemia during exercise testing tion. These patients were classified into three have a similar risk of developing an acute MI or groups; group 1 included patients with no exercise sudden death, except in the three-vessel CAD sub- ST deviation (n = 856), group 2 included patients group, where the risk is greater in silent ischemia. with painless exercise ST deviation (n = 242), and group 3 included patients with both angina and Callaham et al110 performed a study to deter- ST-segment deviation during exercise (n = 600). mine the effect of silent ischemia on prognosis in Patients with exercise angina had a history of patients undergoing exercise testing. In addition, a longer and more aggressive anginal course we took the opportunity to demonstrate if differ- (with a greater frequency of angina, with nocturnal ences in the prevalence of silent ischemia and its episodes and/or progressive symptom pattern) and impact on the prognosis of patients with silent more severe CAD (almost two thirds had three- ischemia could be explained by age or by their MI vessel disease). The 5-year survival rate among and diabetes mellitus status. The design was retro- the patients with painless ST deviation was simi- spective with a 2-year mean follow-up. The patient lar to that of patients with ST deviation (86% population was inpatient and outpatient referrals and 88%, respectively) and was significantly better for exercise testing at a 1000-bed VA hospital. than that of patients with both symptoms and Exercise test responses were analyzed separately ST deviation (5-year survival rate 73% in patients for the four subgroups: angina plus ST depression, with exercise-limiting angina). Similar trends silent ischemia, angina only, and no ischemia. were obtained in subgroups defined by the amount Mean maximal heart rate, maximal SBP, and max- of CAD present. In the total study group of imal MET level attained were significantly higher 1698 patients, silent ischemia on the treadmill for patients with silent ischemia than patients was not a benign finding (average annual mortal- with angina plus ST depression. Mean maximal ity rate 2.8%) but, compared with symptomatic ST segment depression was significantly greater ischemia, did indicate a subgroup of patients with among patients with angina plus ST depression CAD who had a less aggressive anginal course, than patients with silent ischemia. The prevalence less CAD, and a better prognosis. Other smaller of silent ischemia increased with age, while angiographic studies agree with this finding,107,108 the prevalence of angina plus ST depression did which may reflect the bluntness of the tool which not. There was a 7% prevalence of silent ischemia is exercise-induced ST depression. It is possible among patients less than 50 years of age, 17% preva- that those patients with no pain had less severe lence in patients aged 50 to 59 years, 20% preva- disease despite similar levels of ST depression. lence in patients aged 60 to 69 years, and 36% prevalence for patients aged 70 or greater. Among To evaluate whether patients with angio- 326 patients undergoing cardiac catheterization, graphic evidence of CAD with silent myocardial the mean number of vessels diseased (two) and LV ischemia during exercise testing are at increased ejection fraction (58%) were not significantly dif- risk for developing a subsequent acute MI or sud- ferent according to ischemia status. During 2-year den death, another analysis from the CASS reg- follow-up, 71 patients died, 68 patients underwent istry was performed.109 The study involved 424 CABS, 51 patients underwent PCI as their sole patients with silent ischemia who were compared revascularization procedure, and 13 patients under- with another 456 patients with CAD who had both went both CABS and PCI. Patients in the angina ischemic ST depression and angina pectoris dur- plus ST depression and silent ischemia groups ing exercise testing, and with 1019 control had significantly higher overall 2-year mortality patients without CAD. The probability of remain- than patients without ST-segment depression. ing free of a subsequent acute MI or sudden death Overall mortality in patients with angina and ST at 7 years was 80% and 91%, respectively, for depression and patients with silent ischemia was patients with silent ischemia; 82% and 93%, respec- not significantly different. We recently repeated tively, for patients with ST depression and angina

280 E X E R C I S E A N D T H E H E A R T these analyses in a larger data set of veterans, ischemia is associated with a similar prognosis as including those from Palo Alto VA, with a longer ST depression associated with angina pectoris. follow-up (Fig. 8-7). These findings demonstrate that silent ischemia occurring with treadmill testing does not confer Prior MI and Silent Ischemia. We investigated an increased risk for death relative to patients whether prior MI influenced silent ischemia and experiencing angina. Thus, therapy should not be prognosis. Patients who had recently suffered an guided by the false hypothesis that patients with MI (within 2 weeks), and patients who had suf- silent ischemia are at higher risk for death than fered an MI in the past (>2 weeks) were grouped those with angina and ST depression. Recently we separately. No significant difference was seen repeated these analyses in diabetics in a larger in the prevalence of silent ischemia angina plus veteran population, with a longer follow-up ST depression among the three groups. Prognosis (Figure 8-8). was significantly worse among patients with a recent MI, particularly when ischemic ST segment Dagenais et al20 reported 6-year cumulative depression was present. survival in 298 moderately treated patients with exercise-induced ST-segment depression equal or Diabetes and Silent Ischemia. Ninety-three greater than 2 mm. In those with silent myocar- insulin-dependent and 87 non-insulin-dependent dial ischemia, survival was 85%, while it was signif- patients with diabetes mellitus were tested. Of icantly lower (80%) in those with angina pectoris. those with ischemic ST-segment depression, Patients with silent myocardial ischemia reached 64% of insulin-dependent and 61% of non- a greater heart rate and higher MET level than those insulin-dependent diabetic patients had silent with painful ischemia. Cumulative survival was very myocardial ischemia. The prevalence of silent much related to the MET level achieved. Those ischemia among the nondiabetic patients (60%) who reached 10 METs had very few deaths, while and diabetic patients (62%) was not significantly those with less than 5 METs had approximately a different. Mortality was significantly greater among 50% survival. In a small study of less than 100 dia- patients with abnormal ST-segment depression betics from the CASS registry, contrary results compared with those without ST segment depres- were reported.111 These data suggested that, among sion. The presence or absence of angina pectoris patients with diabetes and CAD, silent myocardial during exercise testing was not significantly related ischemia during exercise testing adversely affects to death. The prevalence of silent ischemia is not survival, and that CABS improves the survival of statistically different during exercise testing in diabetic patients with silent myocardial ischemia patients with recent MI, remote MI, or no history and three-vessel CAD. of MI, or those with insulin–dependent or non- insulin-dependent diabetes mellitus. Thus, silent The Cedars group studied 1271 consecutively registered patients with diabetes and 5862 patients without diabetes with known or suspected 1.0 1.0 0.9 0.8 0.9 Diabetics Survival Survival 0.8 0.7 No ST dep, no angina (n = 3,817; 369 CV deaths, 0.8%/year) 0.7 No ST dep, no angina (n=360; 33 CV deaths, 1.0%/year) No ST dep, with angina (n = 475; 39 CV deaths, 0.9%/year) No ST dep, with angina (n=57; 3 CV deaths) With ST dep, no angina (n = 827; 118 CV deaths, 1.6%/year) With ST dep, no angina (n=92; 17 CV deaths, 2.5%/year) 0.6 With ST dep, with angina (n = 462; 80 CV deaths, 2.0%/year) 0.6 With ST dep, with angina (n=58; 10 CV deaths, 3.3%/year) Log-rank p <0.001 0.5 Log-rank p < 0.001 0 2.0 4.0 6.0 8.0 0.5 0 Follow-up years 2.0 4.0 6.0 8.0 ■ FIGURE 8–7 Follow-up years Silent ischemia Kaplan-Meier curves for the general population. ■ FIGURE 8–8 Silent ischemia Kaplan-Meier curves in diabetic patients.

C H A P T E R 8 Prognostic Applications of Exercise Testing 281 CAD undergoing nuclear perfusion myocardial (n = 213). Both underwent coronary angiography perfusion imaging with exercise or adenosine and were compared with each other with respect pharmacologic testing.112 Patients were followed- to various exercise and angiographic parameters. up for at least 1 year except for the 6% lost to fol- Patients with exercise-induced silent ischemia low-up. Over the follow-up period, patients with exercised longer, reached a higher peak exercise diabetes had significantly higher rates of “hard” heart rate, and a higher peak exercise rate pres- events (cardiac death or nonfatal MI) (4.3% versus sure product than patients with exercise-induced 2.3% per year) and higher total event rates (hard angina pectoris. In the latter group, more patients events and late revascularization) (9.0% versus showed exercise-induced ST-segment depression 5.3% per year) compared with rates among greater than 2 mm. The group of patients with patients without diabetes. Cox proportional haz- silent ischemia encompassed more individuals ards analysis revealed that nuclear testing added with normal coronary arteries. More patients with incremental value over clinical and historical exercise- induced angina pectoris had three-vessel variables among patients with diabetes. The event disease. The exclusion of patients with normal rates rose significantly as a function of summed coronary arteries (23% in those with silent stress score and summed difference score among ischemia group and 6% in those with exercise- both patients with diabetes and patients without induced angina) had no influence on the level of diabetes. The patients with diabetes with normal significance for peak heart rate, mean exercise scans had relatively low hard event rates (1% to duration, and exercise duration greater than 10 2% per year), those with mildly abnormal scans minutes. As in most other studies, exercise- had intermediate hard event rates (3% to 4% per induced silent myocardial ischemia was associ- year), and those with moderately to severely ated with better exercise performance and less abnormal scans had relatively high hard event extensive coronary disease than in exercise- rates (>7% per year). induced angina pectoris. Giri et al113 followed-up patients with symptoms Miranda et al114 performed a retrospective of CAD, who were undergoing nuclear perfusion analysis of 416 male veterans referred for exercise imaging from five centers for 2.5 years, for the sub- testing and selected for cardiac catheterization. sequent occurrence of cardiac death, MI, and revas- We found that exercise-induced ST depression was cularization. Perfusion scan results were a better marker for CAD than exercise test-induced categorized as normal or abnormal (fixed or angina and that symptomatic ischemia (ST depres- ischemic defects). Of 4755 patients, 929 (20%) were sion plus angina) was a better indicator of severe diabetic. Patients with diabetes, despite an angiographically significant CAD than silent increased revascularization rate, had twice as many ischemia. As part of the Program on the Surgical cardiac events (8.6%; 39 deaths and 41 MIs) com- Control of the Hyperlipidemias (POSCH), subjects pared to the nondiabetics (5%; 69 deaths and with hyperlipidemia who had one healed MI were 103 MIs). Abnormal perfusion was an independent studied and followed-up for 9 years.115 Of the 417 predictor of cardiac death and MI in both popula- control subjects, 279 had a treadmill test result tions. Diabetics with ischemic defects had an that was definitely positive or negative. There was increased number of cardiac events, with the high- no difference in survival between subjects with a est MI rates (17%) observed with three-vessel positive or negative test result with or without ischemia. Similarly, a multivessel fixed defect was angina and as regards to blood lipids, type of MI associated with the highest rate of cardiac death (Q or non-Q wave), and LV function. (14%) among diabetics. In multivariable Cox analysis, both ischemic and fixed defects independ- The angiographic studies of silent ischemia ently predicted cardiac death alone or cardiac reviewed by Miranda et al114 are summarized in death/MI. Table 8-9. In this review encompassing almost 6000 patients, a consistent finding was that patients with Angiographic Studies of Silent Ischemia symptomatic ischemia had a higher prevalence of severe angiographically significant CAD than did Visser et al108 from the Netherlands studied 280 patients with silent ischemia. patients with anginal complaints, without prior MI and with an abnormal exercise test. They were Comparison of Treadmill Testing to divided into two groups: one with (n = 67) with Ambulatory Monitoring exercise-induced silent ischemia (n = 67) and the other with exercise-induced angina pectoris In one of only a few studies comparing the prog- nostic value of the treadmill test and ambulatory

TA B L E 8 – 9 . Studies of silent ischemia during exercise testing with angiographic correlation 282 E X E R C I S E A N D T H E H E A R T NOISCH APO SI STAP MVD 3V/LM 3V/LM MVD 3V/LM Investigator No. MVD (%) 3V/LM (%) (%) (%) (%) patients Exclusions (%) 62 — MVD (%) (%) 82 — 50 5 77 — 71 23 Amsterdam 92 No CAD, coronary stenoses <70%, normal ET — — 64 18 Deligonul 390 No CAD, coronary 49 10 45 — 75 — 50 — 45 — Erikssen 103 stenoses <50%, digoxin, —— 84 55/7 LBBB, LM, LVH, failed — — 85 56/5 Falcone 473 PTCA — 27 — 37 88 60/12 Coronary stenoses <50%, 79 48/12 Mark 1698 previously known CAD, 16 6 13 9 51 30 \"other\" heart disease, 50 20 Miranda 200 HTN, DM, malignancy, musculoskeletal disorders, any other advanced disease No CAD, coronary stenoses <50%, digoxin, LBBB, CHF, valvular disease, variant angina, no exercise ST depression No CAD, coronary stenoses <75%, LBBB, exercise ST elevation, USA, valvular disease, congenital heart disease, cardiac surgery Digoxin, coronary stenoses <75%, LVH, LBBB, CABG/PCI, women, prior MI, resting ST depression

Ouyang 60 Coronary stenoses <70%, — — — — 74 29/11 81 54/6 Stern 480 no exercise ST depression — — — — 66 33 72 36 No CAD, coronary Visser 280 stenoses <70%, no exercise — — — — 38 13 74 38 ST depression, digoxin, Weiner 1583 \"baseline ECG changes,\" 42 13 55 23 63 29 74 38 Weiner 302 valvular disease, 21 — 67 — 51 — 94 — cardiomyopathy Total: — No CAD, coronary — — — —— — —— C H A P T E R 8 Prognostic Applications of Exercise Testing 283 5877 stenosis <50%, no exercise 36 13 46 17 63 31 76 41 ST depression, LVH, LBBB, CHF, valvular disease, cardiomyopathy, prior MI, congenital heart disease No CAD, coronary stenoses <70% Coronary stenoses <70%, digoxin, LBBB, LVH, valvular disease, patients without exercise angina or ST depression that did not achieve 85% of submaximal heart rate Means: APO, Angina pectoris only during exercise test; CABG, coronary artery bypass graft; CAD, coronary artery disease; DM, diabetes mellitus; ET, exercise test; HTN, hypertension; LBBB, left bundle branch block; LM, left main coronary artery disease is >50% narrowing; LVH, left ventricular hypertrophy; MI, myocardial infarction; MVD, multivessel disease (two-, three-vessel, or LM CAD); NOISCH, normal exercise test; PTCA, percutaneous transluminal coronary angioplasty; SI, ST-segment depression only during exercise test; STAP, ST-segment depression and angina pectoris during exercise test; 3V/LM, three-vessel/left main coronary artery disease; USA, unstable angina pectoris.

284 E X E R C I S E A N D T H E H E A R T monitoring, Tzivoni et al116 followed 224 low-risk after exercise testing. There were 31 patients with patients with prior MI for a mean of 28 months angina and ST-segment depression. The angina (range 12 to 58 months). Seventy-four patients pectoris disappeared at 3 minutes while the ST-seg- developed ischemic changes during daily activity, ment depression disappeared at 6 minutes in recov- of which 44 (60%) were silent, 14 (19%) were symp- ery. There was no change in this with nitroglycerin tomatic, and 16 (21%) were both. All 74 patients and the persistence was longer in more elderly peo- had ischemic responses to treadmill testing, but in ple. They found that silent myocardial ischemia addition, of the 150 patients with ischemic changes persists after disappearance of exercise-induced on Holter, 44 did show ischemia with the tread- angina pectoris. mill test. The incidence of cardiac events (i.e., car- diac death, nonfatal MI, development of unstable The most important study is that of Hedblad angina, CABG, or angioplasty), was significantly et al121 performed in “men born in 1914 in greater in patients with positive Holter and tread- Malmo.” It essentially shows that pretest proba- mill tests (38/74 = 51%) compared to abnormal bility (i.e., chest pain symptoms) affects ambula- exercise tests but negative Holter (9/44 = 20%). As tory monitoring for screening as it does for might be expected, the group with the least car- exercise testing. diac events had a normal Holter and exercise test (9/106). Interestingly, of the 74 patients with Silent Ischemia More Prevalent in Diabetics? ischemic events on the Holter, there was no corre- lation between symptoms (silent or symptomatic) Not all studies have shown a difference in the and prognosis. prevalence of silent myocardial ischemia between diabetics and the general population. In a land- Two representative studies comparing exercise mark Danish study,122 the prevalence of ischemia testing to ambulatory monitoring that make an was compared in a random sample of 120 users of interesting comparison are those of Mody et al117 insulin and 120 users of oral hypoglycemic agents and Mulcahy et al.118 Mody et al studied 97 patients aged 40 to 75 years. The observed prevalence who were not on antianginal medications. Sixty- of silent ischemia on treadmill or Holter testing three patients had no ischemia (poor sensitivity), in diabetics was 13.5% and was no different in 22 patients had 1 to 60 minutes of ischemia dur- matched controls. No association was found ing 24 hours, and 12 patients exceeded 60 minutes between silent ischemia and gender or diabetes for 24 hours. There was no correlation with exer- type. Although hypertension was highly predictive cise duration or time to ST-segment depression. of silent ischemia in the diabetic subjects, other However, prolonged ischemia on Holter correlated variables did not have a predictive value. This with the angiographically severe CAD. Conversely, finding is hard to explain. Scandinavian popula- Mulcahy et al118 found that in patients whose tions have previously been noted to have increased exercise test was negative, or who do not develop CV disease prevalence123 and it is possible that a ST-segment depression before 5 METs, rarely had high level of baseline disease in the nondiabetic silent myocardial ischemia during ambulatory population masks the population differences seen monitoring. They found that ST-segment depres- in other studies. sion occurs at a lower heart rate with ambulatory monitoring than with exercise testing, but the Data from the Asymptomatic Cardiac Ischemia results were highly correlated. Stern et al119 found Pilot124 revealed that asymptomatic ST-segment that in 544 patients (299 with abnormal angiograms depression during Holter monitoring was 94% in and 241 had prior MI, all of whom had abnormal diabetics and 88% in nondiabetics. In addition, treadmill tests) 47% had silent myocardial the time to onset of 1-mm ST-segment depression ischemia while 53% had chest pain. The age, prior and time to onset of angina were similar in both MI, medications, number of diseased vessels, heart groups. Unlike the previous study, however, entry rate, blood pressure, and maximal ST depression into the Asymptomatic Cardiac Ischemia Pilot were similar in both groups. At 1 mm of ST- required a cardiac event, so the disease was not segment depression, patients with silent ischemia consistently silent. had a higher heart rate and exercise level, reached a higher double product, and had a faster recovery Making sense of these seemingly contradictory postexercise. However, if the ST-segment depres- findings is not straightforward. Silent myocardial sion exceeded 2 mm, there were no differences ischemia has been found to be associated with the between the two groups. Flugelman et al120 stud- same, lower, and higher risk as nonsilent ischemia. ied painless persistent ST-segment depression It has also been found to occur with the same or higher frequency in diabetics. What is clear, how- ever, is that, whether silent or not, ischemia during

C H A P T E R 8 Prognostic Applications of Exercise Testing 285 treadmill testing in the general population predicts equal to the percentage in symptomatic diabetic and increased risk for death. However, in general, these higher than in asymptomatic nondiabetic (46%) study patients presented for testing with chest and symptomatic nondiabetic (44%) patients. The pain symptoms. That is, their ischemia was ‘silent’ breakdown of high-risk scans followed a similar only with respect to the exercise test itself. This pattern in the four patient subsets, roughly about inconsistent appearance of pain could represent a 20%. Patients with diabetes had more ECG and different process in the general population (e.g., a scan evidence for silent MI versus those without different severity of disease) than in the diabetic diabetes. The finding that approximately one in population, where autonomic dysfunction is known five of these individuals has a high-risk scan sug- to be present. This idea is supported by Weiner’s gests a potentially more widespread application studies, in which patients with silent ischemia and of screening stress nuclear perfusion in asymp- either three-vessel disease or diabetes had a tomatic diabetic patients to identify those with poorer outcome. severe CAD. Screening Studies in Diabetics THE ACC/AHA GUIDELINES FOR THE PROGNOSTIC USE OF THE The inherent uncertainty of assessing the preva- STANDARD EXERCISE TEST lence and risk of silent CAD retrospectively has led some investigators to assess the impact of truly The task force to establish guidelines for the use silent disease by carrying out prospective screening of exercise testing has met and produced guide- studies. We found only four studies which took a lines in 1986, 1997, and 2002. The following is a truly asymptomatic diabetic population and synopsis of these evidence-based guidelines. screened for CV disease. Koistinen125 found that 29% of diabetics and 5% of control patients had Indications for exercise testing to assess risk ischemic results in one or more noninvasive tests and prognosis in patients with symptoms or a whilst Gerson et al126 found a quarter of 110 prior history of CAD: asymptomatic, insulin-requiring, diabetic patients had abnormal ST depression or an inadequate Class I (Definitely Appropriate). Conditions for heart rate response. Janand-Delenne et al127 found which there is evidence and/or general agreement 16% of noninvasive tests to be positive in 203 that the standard exercise test is useful and helpful patients screened for 1 year with exercise ECG to assess risk and prognosis in patients with symp- and perfusion scans followed-up with coronary toms or a prior history of CAD. angiography. Angiographically significant (>50% stenosis) disease was found in 9.3%. Finally, May • Patients undergoing initial evaluation with et al122 found the prevalence of silent ischemia to be suspected or known CAD. Specific exceptions 13.5% in a randomly chosen diabetic population. are noted below in Class IIb. It seems likely then that the prevalence of • Patients with suspected or known CAD pre- silent CV disease in diabetic populations is high, viously evaluated with significant change in and probably of the order 10% to 30%, compared clinical status. with a control rate of around 5%. However, the diagnostic tools available to assess coronary dis- Class IIb (Maybe Appropriate). Conditions for ease in the general population need to be re-eval- which there is conflicting evidence and/or a diver- uated for the diabetic population, in light of the gence of opinion that the standard exercise test is significantly higher preprobability of disease, and useful and helpful to assess risk and prognosis in the potential confounding factor of silent ischemia. patients with symptoms or a prior history of CAD but the usefulness/efficacy is less well established. The Mayo clinic group examined nuclear per- fusion exercise imaging in asymptomatic diabetic • Patients who demonstrate the following patients. The results of stress nuclear perfusion in ECG abnormalities: patients without prior MI or coronary revascular- ization were compared in asymptomatic diabetics • Pre-excitation (Wolff-Parkinson-White) syn- (n = 1738) versus symptomatic diabetic patients drome (n = 2998), asymptomatic nondiabetic patients (n = 6215), and symptomatic nondiabetic patients (n = • Electronically paced ventricular rhythm 16,214).128 Abnormal scans were present in 59% of asymptomatic diabetic patients, approximately • More than 1 mm of resting ST depression • Complete left bundle branch block

286 E X E R C I S E A N D T H E H E A R T • Patients with a stable clinical course who CV endpoints and removed patients from observa- undergo periodic monitoring to guide man- tion who had interventions. This is because death agement data is now relatively easy to obtain, whereas previ- ously investigators had to follow the patients and Class III (Not Appropriate). Conditions for which contact them or review their records. CV mortal- there is evidence and/or general agreement that ity can be determined from death certificates. While the standard exercise test is not useful and help- death certificates have their limitations, in general, ful to assess risk and prognosis in patients with they classify those with accidental, gastrointesti- symptoms or a prior history of CAD and in some nal, pulmonary, and cancer deaths so that those cases may be harmful. remaining are most likely to have died of CV causes. This endpoint is more appropriate for a • Patients with severe comorbidity likely to test for CV disease. Whereas all-cause mortality is a limit life expectancy and/or candidacy for more important endpoint for intervention studies, revascularization CV mortality is more appropriate for evaluating a CV test (i.e., the exercise test). Identifying those The evidence supporting these guidelines has at risk of death due to any cause does not make been presented in this chapter. it possible to identify those who might benefit from CV interventions, one of the two goals of SUMMARY prognostication. The two principal reasons for estimating prognosis Rather than the differences perhaps it is better are to provide accurate answers to patient’s ques- to stress the consistencies. Considering simple tions regarding the probable outcome of their clinical variables can assess risk. A good exercise illness and to identify those patients in whom capacity, no evidence or history of CHF or ventric- interventions might improve outcome. There is a ular damage (Q waves, history of CHF), no ST lack of consistency in the available studies because depression, or having only one of these clinical find- patients die along a pathophysiologic spectrum ings are associated with a very low risk. These ranging from those that die due to CHF with little patients are low risk in exercise programs and need myocardium remaining to those that die from an not be considered for interventions to prolong their ischemic-related event with ample myocardium life. High-risk patients can be identified by group- remaining. Clinical and exercise test variables ings of two or more clinical markers. Exertional most likely associated with CHF deaths (CHF hypotension is particularly ominous. Identification markers) include a history or symptoms of CHF, of high risk implies that in exercise training pro- prior MI, Q waves, and other indicators of LV dys- grams such patients should have lower goals and function. Variables most likely associated with should be monitored. Such patients should also be ischemic deaths (ischemic markers) are angina, considered for coronary interventions to improve ST depression at rest, and exercise ST depres- their longevity. Furthermore, with each drop in sion. Some variables can be associated with either METs there is a 10% to 20% increase in mortality, extremes of the type of CV death; these include so simple exercise capacity has consistent impor- exercise capacity, maximal heart rate, and maximal tance in all patient groups. SBP that may explain why they are reported most consistently in the available studies. A problem The mathematical models for determining prog- exists that ischemic deaths occur later in follow- nosis are usually more complex than those used for up and are more likely to occur in those lost to identifying angiographically severe disease. follow–up, whereas CHF deaths are more likely Diagnostic testing can utilize multivariate discrim- to occur early (within 2 years) and are more likely inant function analysis to determine the probabil- to be classified. Work-up bias probably explains ity of the presence or absence of angiographically why exercise-induced ST depression fails to be a pre- severe disease. Prognostic testing must utilize sur- dictor in most of the angiographic studies. Ischemic vival analysis, which includes censoring for patients markers are associated with a later and lesser risk, with uneven follow-up due to “lost to follow-up” or whereas CHF or LV dysfunction markers are associ- other cardiac events (e.g., CABS, PCI) and must ated with a sooner and greater risk of death (these account for time-person units of exposure. Survival concepts are illustrated in Fig. 8-2). curves must be developed and the Cox proportional hazards model is often preferred. We have pro- Recent studies of prognosis have actually not posed the rules in Table 8-10 to assess prognos- been superior to the earlier studies that considered tic studies. The newest kid on the block: heart rate recovery has yet to be validated with the

C H A P T E R 8 Prognostic Applications of Exercise Testing 287 TA B L E 8 – 1 0 . Proposed criteria for studies assessing prognostic value of clinical and exercise test variables 1. Study population: inclusion criteria such as catheterization should be specified. Prevalences of congestive heart failure, congestive heart failure-associated conditions (prior myocardial infarction, Q waves on resting ECG), and angina should be stated. 2. Avoidance of “work-up bias”: limited study populations, such as patients referred for catheterization, should be avoided, or validation studies in different populations or bootstrapping techniques should be used. 3. Exercise testing procedures: protocols used and criteria for abnormal values should be well described. 4. Clinical and exercise test variables: variables must be clearly defined and entered into the statistical analysis separately. 5. Study endpoints: cardiovascular death and nonfatal myocardial infarction should be used. 6. Avoidance of “overfitting the data”: the ratio of events to the number of variables studied should be at least 10 to ensure enough “hard” outcomes per given variable studied. 7. Follow-up: length and completeness should be documented. 8. Treatment of interventions: coronary artery bypass surgery and percutaneous transluminal coronary angioplasty should not be used as endpoints. 9. Censoring: patients should be censored on interventions (coronary artery bypass surgery or percutaneous transluminal coronary angioplasty) and on “lost to follow-up.” 10. Relationship between censored events and studied variables: it should be determined whether censoring is random or correlated with specific clinical and exercise test markers. 11. Multivariate survival analysis techniques: Cox proportional hazard model or discriminate analysis should be used. 12. Concordance with the hierarchical nature of clinical data acquisition: variables should be entered into multivariate analysis in an order similar to clinical practice (i.e., clinical parameters followed by exercise test variables and then inva- sive test variables). 13. Interactions between variables: associations between variables (i.e., digoxin use and congestive heart failure or ST elevation over Q waves) should be noted and treated appropriately. 14. Avoidance of test-review bias: investigators should be blinded to patient characteristics and results of other diagnostic and prognostic tests. more appropriate endpoint of CV mortality. In 2. Yusuf S, Zucker D, Peduzzi P, et al: Effect of coronary artery bypass fact, it appears to predict non-CV death better than graft surgery on survival: Overview of 10-year results from ran- CV death. domised trials by the Coronary Artery Bypass Graft Surgery Trialists Collaboration. Lancet 1994;344:563-570. From this perspective, it is obvious that there is much information supporting the use of exer- 3. Marcus ML, Wilson FR, White CW: Methods of measurement of cise testing as the first noninvasive step after the myocardial blood flow in patients: A critical review. Circulation history, physical exam, and resting ECG in the 1987;76:245-251. prognostic evaluation of CAD patients. It accom- plishes both of the purposes of prognostic testing: 4. Stone PH, Turi ZG, Muller JE, et al: Prognostic significance of the to provide information regarding the patient’s treadmill exercise test performance 6 months after myocardial status and to help make recommendations for infarction. J Am Coll Cardiol 1986;8:1007-1017. optimal management. The exercise test results help us make reasonable decisions for selection of 5. Haines DE, Beller GA, Watson DD, et al: Exercise-induced ST seg- patients who should undergo coronary angiogra- ment evaluation 2 weeks after uncomplicated myocardial infarc- phy. Since the exercise test can be performed in tion: Contributing factors and prognostic significance. J Am Coll the doctor’s office and provides valuable informa- Cardiol 1987;9:996-1003. tion for clinical management in regard to activity levels, response to therapy, and disability, the 6. Hammond HK, Kelly TL, Froelicher VF: Radionuclide imaging cor- exercise test is the reasonable first choice for relatives of heart rate impairment during maximal exercise testing. prognostic assessment. This assessment should J Am Coll Cardiol 1983;2:826-833. always include calculation of the estimated annual mortality using the DTS though its 7. Dubach P, Froelicher VF, Klein J, et al: Exercise-induced hypoten- ischemic elements have less power in the elderly. sion in a veteran population — criteria, causes, and prognosis. Circulation 1988;78:1380-1387. REFERENCES 8. Morris CK, Ueshima K, Kawaguchi T, et al: The prognostic value of 1. Chassin MR, Kosecoff J, Solomon DH, Brook RH: How coronary exercise capacity: A review of the literature. Am Heart J angiography is used. Clinical determinants of appropriateness. 1991;122:1423-1430. JAMA 1987;258:2543-2547. 9. McKirnan MD, Sullivan M, Jensen D, Froelicher VF: Treadmill per- formance and cardiac function in selected patients with coronary heart disease. J Am Coll Cardiol 1984;3:253-261. 10. Bounous EP, Califf RM, Harrell FE, et al: Prognostic value of the sim- plified Selvester QRS score in patients with coronary artery disease. J Am Coll Cardiol 1988;11:35-41. 11. Hlatky MA, Califf RM, Harrell FE, et al: Clinical judgment and ther- apeutic decision making. J Am Coll Cardiol 1990;15:1-14. 12. Shaw LJ, Peterson ED, Shaw LK, et al: Use of a prognostic tread- mill score in identifying diagnostic coronary disease subgroups. Circulation 1998;98:1622-1630. 13. Lauer MS, Blackstone E, Young J, Topol E: Cause of death in clinical research: Time for a reassessment? J Am Coll Cardiol 1999;34:618-620. 14. Mark DB, Shaw L, Harrell FE, Jr, et al: Prognostic value of a tread- mill exercise score in outpatients with suspected coronary artery disease. N Engl J Med 1991;325:849-853.

288 E X E R C I S E A N D T H E H E A R T 15. Froelicher VF, Morrow K, Brown M, et al: Prediction of artheroscle- 38. Morise AP, Jalisi F: Evaluation of pretest and exercise test scores to rotic cardiovascular death in men using a prognostic score. Am J assess all-cause mortality in unselected patients presenting for Cardiol 1994;73:133-138. exercise testing with symptoms of suspected coronary artery dis- ease. J Am Coll Cardiol 2003;42:842-850. 16. Ellestad M, Wan M: Prediction implications of stress testing. Circulation 1975;51:363-369. 39. Prakash M, Froelicher V, Myers J, et al: Clinical and exercise test predictors of all-cause mortality: Results from over 6,000 consecu- 17. McNeer JF, Margolis JR, Lee KL, et al: The role of the exercise test tive referred male patients. Chest 2001(in press). in the evaluation of patients for ischemic heart disease. Circulation 1978;57:64-70. 40. Alexander K, Shaw L, Delong E, et al. Value of exercise treadmill testing in women. J Am Coll Cardiol 1998;32:1657-1664. 18. Weiner DA, McCabe CH, Ryan TJ: Identification of patients with left main and three vessel coronary disease with clinical and exercise 41. Morise AP, Olson MB, Merz CN, et al: Validation of the accuracy test variables. Am J Cardiol 1980;46:21-27. of pretest and exercise test scores in women with a low prevalence of coronary disease: The NHLBI-sponsored Women’s Ischemia 19. Podrid PJ, Graboys T, Lown B: Prognosis of medically treated patients Syndrome Evaluation (WISE) study. Am Heart J 2004;147: with coronary artery disease with profound ST segment depression 1085-1092. during exercise testing. N Engl J Med 1981;305:1111-1116. 42. Kwok JM, Miller TD, Christian TF, et al: Prognostic value of a tread- 20. Dagenais GR, Rouleau JR, Christen A, Fabia J: Survival of patients mill exercise score in symptomatic patients with nonspecific ST-T with a strongly positive exercise electrocardiogram. Circulation abnormalities on resting ECG. JAMA 1999;282:1047-1053. 1982;65:452-456. 43. Cole CR, Blackstone EH, Pashkow FJ, et al: Heart-rate recovery 21. Bruce RA, Fisher LD, Hossack KF: Validation of exercise-enhanced immediately after exercise as a predictor of mortality. N Engl J risk assessment of coronary heart disease events: Longitudinal Med: 1999;341:1351-1357. changes in incidence in Seattle community practice. J Am Coll Cardiol 1985;5:875-881. 44. Cole CR, Foody JM, Blackstone EH, Lauer MS: Heart rate recovery after submaximal exercise testing as a predictor of mortality in 22. Oberman A, Jones WB, Riley CP, et al: Natural history of coronary a cardiovascularly healthy cohort. Ann Intern Med 2000;132: 552-555. artery disease. Bull NY Acad Med 1972;48:1109-1125. 45. Nishime EO, Cole CR, Blackstone EH, et al: Heart rate recovery and 23. Reeves TJ, Oberman A, Jones WB, Sheffield LT: Natural history of treadmill exercise score as predictors of mortality in patients angina pectoris. Am J Cardiol 1974;33:423-430. referred for exercise ECG. JAMA 2000;284:1392-1398. 24. Hammermeister KE, DeRouen TA, Dodge HT: Variables predictive of 46. Shetler K, Marcus R, Froelicher VF, et al: Heart rate recovery: survival in patients with coronary disease. Selection by univariate Validation and methodologic issues. J Am Coll Cardiol and multivariate analyses from the clinical, electrocardiographic, 2001;38:1980-1987. exercise, arteriographic, and quantitative angiographic evaluation. Circulation 1979;59:421-430. 47. Morshedi-Meibodi A, Larson MG, Levy D, et al: Heart rate recovery after treadmill exercise testing and risk of cardiovascular disease events 25. Gohlke H, Samek L, Betz P, Roskamm H: Exercise testing provides (The Framingham Heart Study). Am J Cardiol 2002;90:848-852. additional prognostic information in angiographically defined sub- groups of patients with coronary artery disease. Circulation 1983; 48. Mora S, Redberg RF, Cui Y, et al: Ability of exercise testing to pre- 68:979-985. dict cardiovascular and all-cause death in asymptomatic women: A 20-year follow-up of the lipid research clinics prevalence study. 26. Brunelli C, Cristofani R, L’Abbate A, for the ODI Study Group: JAMA 2003;290:1600-1607. Long-term survival in medically treated patients with ischemic heart disease and prognostic importance of clinical and electrocar- 49. Cheng YJ, Lauer MS, Earnest CP, et al: Heart rate recovery follow- diographic data. (The Italian CNR Multicenter Prospective Study ing maximal exercise testing as a predictor of cardiovascular dis- OD1). Eur Heart J 1989;10:292-303. ease and all-cause mortality in men with diabetes. Diabetes Care 2003;26:2052-2057. 27. Weiner DA, Ryan T, McCabe CH: Prognostic importance of a clini- cal profile and exercise test in medically treated patients with coro- 50. Simari RD, Miller TD, Zinsmeister AR, Gibbons RJ: Capabilities of nary artery disease. J Am Coll Cardiol 1984;3:772-779. supine exercise electrocardiography versus exercise radionuclide angiography in predicting coronary events. Am J Cardiol 1991; 28. Mark DB, Hlatky MA, Harrell FE, et al: Exercise treadmill score for 67:573-577. predicting prognosis in coronary artery disease. Ann Intern Med 1987;106:793-800. 51. Lee KL, Pryor DB, Pieper KS, et al: Prognostic value of radionu- clide angiography in medically treated patients with coronary 29. Peduzzi P, Hultgren H, Thomsen J, Angell W: Prognostic value of artery disease. A comparison with clinical and catheterization vari- baseline exercise tests. Prog Cardiovasc Dis 1986;28:285-292. ables. Circulation 1990;82:1705-1717. 30. Lerman J, Svetlize H, Capris T, Perosio A: Follow-up of patients 52. Brown KA: Prognostic value of thallium-201 myocardial perfusion after exercise test and catheterization. Medicina (Buenos Aires) imaging. A diagnostic tool comes of age. Circulation 1991;83:363-381. 1986;46:201-211. 53. Kaul S, Lilly DR, Gasho JA, et al: Prognostic utility of the exercise 31. Wyns W, Musschaert-Beauthier E, Van Domburg R, et al: Progostic thallium-201 test in ambulatory patients with chest pain. Circulation value of symptom limited exercise testing in men with a high 1988;77:745-748. prevalence of coronary artery disease. Eur Heart J 1985;6:939-945. 54. Melin JA, Robert A, Luwaert R, et al: Additional prognostic value of 32. Detry JM, Luwaert J, Melin J, et al: Non-invasive data provide exercise testing and thallium-201 scintigraphy in catheterized independent prognostic information in patients with chest pain patients without previous myocardial infarction. Int J Cardiol without previous myocardial infarction: findings in male patients 1990;27:235-243. who have had cardiac catheterization. Eur Heart J 1988;9:418-426. 55. Hachamovitch R, Berman DS, Shaw LJ, et al: Incremental prognostic 33. Morris CK, Morrow K, Froelicher VF, et al: Prediction of cardiovas- value of myocardial perfusion single photon emission computed cular death by means of clinical and exercise test variables in tomography for the prediction of cardiac death: Differential stratifica- patients selected for cardiac catheterization. Am Heart J 1993; tion for risk of cardiac death and myocardial infarction. Circulation 125:1717-1726. 1998;97:535-543. 34. Califf RM, Mark DB, Harrell FE, et al: Importance of clinical mea- 56. Snader CE, Marwick TH, Pashkow FJ, et al: Importance of esti- sures of ischemia in the prognosis of patients with documented mated functional capacity as a predictor of all-cause mortality coronary artery disease. J Am Coll Cardiol 1988;11:20-26. among patients referred for exercise thallium single-photon emis- sion computed tomography: Report of 3,400 patients from a single 35. Detre K, Peduzzi P, Murphy M, et al: Effect of bypass surgery on center. J Am Coll Cardiol 1997;30:641-648. survival in patients in low- and high-risk subgroups delineated by the use of simple clinical variables. Circulation 1981;63:1329-1338. 57. Poornima IG, Miller TD, Christian TF, et al: Utility of myocardial perfusion imaging in patients with low-risk treadmill scores. J Am 36. Morrow K, Morris CK, Froelicher VF, Hideg A: Prediction of cardio- Coll Cardiol 2004;43:194-199. vascular death in men undergoing noninvasive evaluation for CAD. Ann Intern Med 1993;118:689-695. 58. Vanzetto G, Ormezzano O, Fagret D, et al: Long-term additive prog- nostic value of thallium-201 myocardial perfusion imaging over clin- 37. Villella M, Villella A, Santoro L, et al: Ergometric score systems ical and exercise stress test in low to intermediate risk patients: Study after myocardial infarction: Prognostic performance of the Duke in 1137 patients with 6-year follow-up. Circulation 1999;100: 1521- Treadmill Score, Veterans Administration Medical Center Score, 1527. and of a novel score system, GISSI-2 Index, in a cohort of survivors of acute myocardial infarction. Am Heart J 2003;145:475-483.

C H A P T E R 8 Prognostic Applications of Exercise Testing 289 59. Chatziioannou SN, Moore WH, Ford PV, et al: Prognostic value of 81. Fisher L, Kennedy J, Chaitman B, et al: Diagnostic quantification myocardial perfusion imaging in patients with high exercise toler- of CASS (Coronary Artery Surgery Study) clinical and exercise ance. Circulation 1999;99:867-872. test results in determining presence and extent of coronary artery disease. Circulation 1981;63:987-1000. 60. Diaz LA, Brunken RC, Blackstone EH, et al: Independent contribu- tion of myocardial perfusion defects to exercise capacity and heart 82. McCarthy D, Sciacca R, Blood D, Cannon P: Discriminant func- rate recovery for prediction of all-cause mortality in patients with tion analysis using thallium 201 scintiscans and exercise stress known or suspected coronary heart disease. J Am Coll Cardiol test variables to predict the presence and extent of coronary artery 2001;37:1558-1564. disease. Am J Cardiol 1982;49:1917-1926. 61. Schinkel AF, Elhendy A, van Domburg RT, et al: Incremental value 83. Hung J, Chaitman BR, Lam J, et al: A logistic regression analysis of exercise technetium-99m tetrofosmin myocardial perfusion sin- of multiple noninvasive tests for the prediction of the presence gle-photon emission computed tomography for the prediction of and extent of coronary artery disease in men. Am Heart J 1985; cardiac events. Am J Cardiol 2003;91:408-411. 110:460-469. 62. McCully RB, Roger VL, Mahoney DW, et al: Outcome after normal 84. Christian T, Miller T, Bailey K, Gibbons R: Exercise tomographic exercise echocardiography and predictors of subsequent cardiac thallium-201 imaging in patients with severe coronary artery events: Follow-up of 1,325 patients. J Am Coll Cardiol 1998;31: disease and normal electrocardiograms. Ann Intern Med 1994; 144-149. 121:825-832. 63. McCully RB, Roger VL, Mahoney DW, et al: Outcome after abnor- 85. Morise A, Bobbio M, Detrano R, Duval R: Incremental evaluation mal exercise echocardiography for patients with good exercise of exercise capacity as an independent predictor of coronary artery capacity: Prognostic importance of the extent and severity of exer- disease presence and extent. Am Heart J 1994;127:32-38. cise-related left ventricular dysfunction. J Am Coll Cardiol 2002;39:1345-1352. 86. Morise A, Diamond G, Detrano R, Bobbio M: Incremental value of exercise electrocardiography and thallium-201 testing in men and 64. Marwick TH, Case C, Vasey C, et al: Prediction of mortality by exer- women for the presence and extent of coronary artery disease. cise echocardiography: A strategy for combination with the Duke Am Heart J 1995;130:267-276. treadmill score. Circulation 2001;103:2566-2571. 87. Moussa I, Rodriguez M, Froning J, Froelicher VF: Prediction of 65. Elhendy A, Mahoney DW, McCully RB, et al: Use of a scoring model severe coronary artery disease using computerized ECG measure- combining clinical, exercise test, and echocardiographic data to ments and discriminant function analysis. J Electrocardiol predict mortality in patients with known or suspected coronary 1992;25: 49-58. artery disease. Am J Cardiol 2004;93:1223-1228. 88. Christian TF, Miller TD, Bailley KR, Gibbons RJ: Noninvasive 66. Lee KL, Pryor DB, Harrell FE, Jr, et al: Predicting outcome in coro- identification of severe coronary artery disease using exercise nary disease. Statistical models versus expert clinicians. Am J Med tomographic thallium-201 imaging. Am J Cardiol 1992;70: 1986;80:553-560. 14-20. 67. Detrano R, Bobbio M, Olson H, et al: Computer probability esti- 89. Hung J, Chaitman BR, Lam J, et al: Noninvasive diagnostic test mates of angiographic coronary artery disease: Transportability and choices for the evaluation of coronary artery disease in women: comparison with cardiologists’ estimates. Comput Biomed Res A multivariate comparison of cardiac fluoroscopy, exercise 1992;25:468-485. electrocardiography and exercise thallium myocardial perfusion scintigraphy. J Am Coll Cardiol 1984;4:8-16. 68. Lipinski M, Do D, Froelicher V, et al: Comparison of exercise test scores and physician estimation in determining disease probability. 90. Diamond GA: Future imperfect: The limitations of clinical predic- Arch Intern Med 2001;161:2239-2244. tion models and the limits of clinical prediction. J Am Coll Cardiol 1989;14:12A-22A. 69. Lipinski M, Froelicher V, Atwood E, et al: Comparison of treadmill scores with physician estimates of diagnosis and prognosis in 91. Ribisl PM, Morris CK, Kawaguchi T, et al: Angiographic patterns patients with coronary artery disease. Am Heart J 2002;143: and severe coronary artery disease. Arch Intern Med 650-658. 1992;152:1618-1624. 70. Pryor DB, Shaw L, Harrell FE, et al: Estimating the likelihood of 92. Herbert WG, Lehmann KG, Dubach P, et al: Effect of beta block- severe coronary artery disease. Am J Med 1991;90:553-562. ade on the exercise ECG: ST level versus delta ST/HR index. Am Heart J 1991;122:993-1000. 71. Hubbard BL, Gibbons RJ, Lapeyre AC, 3rd, et al: Identification of severe coronary artery disease using simple clinical parameters. 93. Do D, West JA, Morise A, Froelicher VF: Agreement predicting Arch Intern Med 1992;152:309-312. severe angiographic coronary artery disease using clinical and exercise test data. Am Heart J 1997;134:672-679. 72. Detrano R, Janosi A, Steinbrunn W, et al: Algorithm to predict triple-vessel/left main coronary artery disease in patients without 94. Bruce RA, Hossack KF, DeRouen TA, Hofer V: Enhanced risk myocardial infarction. An international cross validation. assessment for primary coronary heart disease events by maximal Circulation 1991;83(suppl):III89-96. exercise testing: 10 years’ experience of Seattle Heart Watch. J Am Coll Cardiol 1983;2:565-573. 73. Cheitlin MD, Davia JE, de Castro CM, et al: Correlation of “critical” left coronary artery lesions with positive submaximal 95. European Cooperative Group: Long-term results of prospective exercise tests in patients with chest pain. Am Heart J 1975;89: randomized study of coronary artery bypass surgery in stable 305-310. angina pectoris. Lancet 1982;2:1173-1180. 74. Goldschlager N, Selzer A, Cohn K: Treadmill stress tests as indica- 96. Weiner DA, Ryan TJ, McCabe CH, et al: The role of exercise test- tors of presence and severity of coronary artery disease. Ann Intern ing in identifying patients with improved survival after coronary Med 1976;85:277-286. artery bypass surgery. J Am Coll Cardiol 1986;8:741-748. 75. Blumenthal DS, Weiss JL, Mellits ED, Gerstenblith G: The predic- 97. Hultgren HN, Peduzzi P, Detre K, Takaro T: The 5 year effect of tive value of a strongly positive stress test in patients with minimal bypass surgery on relief of angina and exercise performance. symptoms. Am J Med 1981;70:1005-1010. Circulation 1985;72:V79-V83. 76. Lee TH, EF Cook, Goldman L: Prospective evaluation of a clinical 98. Brechue W, Pollock M: Exercise training for coronary artery and exercise-test model for the prediction of left main coronary disease in the elderly. Clin Geriatr Med 1996;12:207-229. artery disease. Med Decis Making 1986;6:136-144. 99. Goraya T, Jacobsen S, Pellikka P, et al: Prognostic value of tread- 77. Detrano R, Gianrossi R, Mulvihill D, et al: Exercise-induced ST seg- mill exercise testing in elderly persons. Ann Intern Med 2000; ment depression in the diagnosis of multivessel coronary disease: A 132:862-870. meta analysis. J Am Coll Cardiol 1989;14:1501-1508. 100. Spin JM, Prakash M, Froelicher VF, et al: The prognostic value of 78. Hartz A, Gammaitoni C, Young M: Quantitative analysis of the exercise testing in elderly men. Am J Med 2002;112:453-459. exercise tolerance test for determining the severity of coronary artery disease. Int J Cardiol 1989;24:63-71. 101. Kwok JM, Miller TD, Hodge DO, et al: Prognostic value of the Duke treadmill score in the elderly. J Am Coll Cardiol 2002; 79. Concato J, Feinstein AR, Holford TR: The risk of determining risk 39:1475-1481. with multivariate models. Ann Intern Med 1993;118:201-210. 102. Lai S, Kaykha A, Yamazaki T, et al: Treadmill scores in elderly 80. Cohn K, Kamm D, Feteih N, et al: Use of treadmill score to quan- men. J Am Coll Cardiol 2004;43:606-615. tify ischemic response and predict extent of coronary disease. Circulation 1979;59:286-296. 103. Raxwal V, Shetler K, Morise A, et al: Simple treadmill score to diagnose coronary disease. Chest 2001;119:1933-1940.