A New 12-Lead ECG Prognostic Score.

BACKGROUND: The Diagonal Line Rule (DLR) is an educational pattern recognition tool for the classification of 12-lead electrocardiograms (ECGs) as normal or abnormal. OBJECTIVE: To develop a score using the DLR for the risk stratification of abnormal ECGs. METHODS: Patients at the Palo Alto, VA, referred for a routine resting ECG between 1987 and 1999 were retrospectively considered. The initial ECGs of 43,798 patients were scored according to the DLR. The main outcome variable was cardiovascular mortality with 8 years of follow-up. RESULTS: ECG abnormalities identified by the DLR were combined with age, gender, and ST depression to develop an additive score. The score was validated and shown to stratify levels of cardiac risk with a 0.78 area under the receiver operating characteristic (ROC) curve for identifying those who had a cardiovascular death. CONCLUSION: The score effectively stratifies ECGs for prediction of risk of cardiovascular death.

Resting ST amplitude: prognosis and normal values in an ambulatory clinical population.

BACKGROUND: There is limited data describing ST segment amplitude in apparently healthy, asymptomatic populations. We analyzed ST amplitude in the standard resting electrocardiogram (ECG) in a large, multiethnic, stable, clinical population. METHODS: We evaluated computerized ST amplitude measurements from the resting ECGs of 29,281 ambulatory outpatients collected between 1987 and 1999 at the Palo Alto, VA. With the PR interval as the isoelectric line, both elevation criteria (≥0.1 mV, ≥0.15 mV, and ≥0.2 mV) and depression criteria (≤-0.05 mV or ≤-0.1 mV), were applied. Cox-Hazard survival analysis techniques were used to demonstrate in which leads ST amplitude displacement was associated with cardiovascular (CV) death. To create a cohort without ECG patterns clearly associated with disease, we excluded ECGs with inverted T waves, wide QRS, or diagnostic Q waves and coded the remaining "normal" ECGs for ST elevation and depression to determine a normal range. RESULTS: The only ST amplitudes that were significantly and independently associated with time to CV death when adjusted for age, gender, and ethnicity were ST depression in all of the lateral leads (I, V4 -V6 ). When isolated to the inferior leads, (II and AVF), no ST amplitude criteria were associated with CV death. Among the "normal ECG" subgroup the precordial leads exhibited the greatest median ST amplitudes and the most significant differences between the leads, genders and ethnicities. CONCLUSIONS: Significant differences in ST amplitude were present in the precordial leads according to gender and ethnicity. This was particularly apparent when amplitude threshold were set for comparisons. Our findings provide the normal range for ST amplitude that when exceeded, should raise clinical concern.

Prognostic value of the computerized ECG in Hispanics.

BACKGROUND: The prevalence and prognostic values of electrocardiogram (ECG) abnormalities in Hispanics have not been compared to other ethnicities in a large population. Despite a worse cardiovascular risk profile, the prevalence of cardiovascular disease is lower in Hispanics compared to non-Hispanics. HYPOTHESIS: We hypothesized that ECG abnormalities were less common in Hispanics and were not as strongly associated with cardiovascular mortality. METHODS: 45,563 ECGs ordered for usual clinical indications in a Veteran's hospital were available for analysis. 1,392 patients who died within one week of the ECG were excluded. Demographic characteristics were recorded and the population was followed for an average of 7.5 years using the California Death Index. The presence of baseline ECG characteristics were recorded and analyzed using the GE/Marquette computerized ECG system. Age, sex and heart rate adjusted Cox hazard ratio analyses were performed. RESULTS: Being Hispanic was associated with lower cardiovascular death, with a hazard ratio (HR) of 0.76 (95% CI 0.65-0.89). Findings such as atrial fibrillation, presence of Q-waves, left ventricular hypertrophy (LVH), upright T-waves in aortic valve replacement (aVR) and cardiac Infarction Injury Scores > 6 were significantly less prevalent in Hispanics than in non-Hispanics. These findings were similarly associated with increased cardiovascular mortality in both groups, each with a HR of approximately 2. CONCLUSION: The lower prevalence of ECG characteristics associated with coronary heart disease, atrial fibrillation and left ventricular hypertrophy support prior observations that cardiovascular disease is less prevalent in the Hispanic population. These findings, however, are similarly associated with increased mortality compared to non-Hispanics.

Exercise testing and training: clinical applications.

The application of exercise in clinical cardiology continues to progress because of research findings. Advances have occurred in the applications, methodology and interpretation of exercise testing. Exercise training has been documented to have a place in the primary prevention of coronary heart disease. In regard to cardiac rehabilitation, both early ambulation and early discharge are safe and beneficial in patients with uncomplicated infarction, and a subsequent exercise program is at least as effective as other interventions. High intensity exercise training in the patient with heart disease may be necessary to cause changes in myocardial perfusion and performance, but it carries an increased risk.

Treadmill performance and cardiac function in selected patients with coronary heart disease.

To investigate the cardiac determinants of treadmill performance in patients able to exercise to volitional fatigue, 88 patients with coronary heart disease free of angina pectoris were tested. The exercise tests included supine bicycle radionuclide ventriculography, thallium scintigraphy and treadmill testing with expired gas analysis. The number of abnormal Q wave locations, ejection fraction, end-diastolic volume, cardiac output, exercise-induced ST segment depression and thallium scar and ischemia scores were the cardiac variables considered. Rest and exercise ejection fractions were highly correlated to thallium scar score (r = -0.72 to -0.75, p less than 0.001), but not to maximal oxygen consumption (r = 0.19 to 0.25, p less than 0.05). Fifty-five percent of the variability in predicting treadmill time or estimated maximal oxygen consumption was explained by treadmill test-induced change in heart rate (39%), thallium ischemia score (12%) and cardiac output at rest (4%). The change in heart rate induced by the treadmill test explained only 27% of the variability in measured maximal oxygen consumption. Myocardial damage predicted ejection fraction at rest and the ability to increase heart rate with treadmill exercise appeared as an essential component of exercise capacity. Exercise capacity was only minimally affected by asymptomatic ischemia and was relatively independent of ventricular function.

Does the rest electrocardiogram after myocardial infarction determine the predictive value of exercise-induced ST depression? A 2 year follow-up study in a veteran population.

The failure of exercise-induced ST segment depression to consistently predict prognosis in patients after myocardial infarction could be a result of population differences and the rest electrocardiogram (ECG). These hypotheses were tested by studying 198 veterans who survived a myocardial infarction, underwent a submaximal predischarge treadmill exercise test and were followed up for cardiac events for 2 years. During the 2 years, 29 deaths, 19 reinfarctions and 28 revascularization procedures were documented. The prevalence of death or reinfarction was two times higher in patients who had exercise-induced ST depression than in patients who did not. However, in the 55 patients without Q waves, the risk increased to 11 times for an abnormal ST response. These findings suggest that exercise-induced ST depression only predicts high risk in patients after myocardial infarction whose ECG at rest does not exhibit Q waves and that differences in the prevalence of rest ECG patterns are the most likely explanation for the failure of agreement among prior studies.

The effect of resting ST segment depression on the diagnostic characteristics of the exercise treadmill test.

OBJECTIVES: The aim of this study is to demonstrate the effect of resting ST segment depression on the diagnostic characteristics of the exercise treadmill test. BACKGROUND: Previous studies evaluating the effect of resting ST segment depression on the diagnostic characteristics of exercise treadmill test have been conducted on relatively small patient groups and based only on visual electrocardiogram (ECG) analysis. METHODS: A retrospective analysis of data collected prospectively was performed on consecutive patients referred for evaluation of chest pain. One thousand two hundred eighty-two patients without a prior myocardial infarction underwent standard exercise treadmill tests followed by coronary angiography, with coronary artery disease defined as a 50% narrowing in at least one major epicardial coronary artery. Sensitivity, specificity, predictive accuracy and area under the curve of the receiver operating characteristic (ROC) plots were calculated for patients with and without resting ST segment depression as determined by visual or computerized analysis of the baseline ECG. RESULTS: Sensitivity of the exercise treadmill test increased in 206 patients with resting ST segment depression determined by visual ECG analysis compared with patients without resting ST segment depression (77 +/- 7% vs. 45 +/- 4%) and specificity decreased (48 +/- 12% vs. 84 +/- 3%). With computerized analysis, sensitivity of the treadmill test increased in 349 patients with resting ST segment depression compared with patients without resting ST segment depression (71 +/- 6% vs. 42 +/- 4%) and specificity decreased (52 +/- 9% vs. 87 +/- 3%) (p < 0.0001 for all comparisons). There was no significant difference in the area under the curve of the ROC plots (0.66-0.69) or the predictive accuracy (62-68%) between the four subgroups. CONCLUSIONS: The diagnostic accuracy and high sensitivity of the exercise treadmill test in a large cohort of patients with resting ST segment depression and no prior myocardial infarction support the initial use of the test for diagnosis of coronary artery disease. The classification of resting ST segment depression by method of analysis (visual vs. computerized) did not affect the results.

Use of clinical data in predicting improvement in exercise capacity after cardiac rehabilitation.

Fifty-nine men with coronary heart disease underwent 1 year of supervised aerobic exercise. They performed exercise tests for maximal oxygen uptake, ST segment analysis, thallium scintigraphy and radionuclide ventriculography before and after the year of exercise. A computerized data base that included clinical descriptors and exercise test results was retrospectively reviewed to determine whether initial features could predict the patient's response to the exercise intervention. Poor correlations were found between the initial measurements and change in maximal oxygen consumption and other indexes of training effect. Patients who initially were in the poorest state of fitness showed the most improvement with training. None of the initial features from the history and physical examination, treadmill study or radionuclide studies was a good predictor of a beneficial result from the exercise program. The usual measurements of work intensity during training were poor predictors of outcome. A significant decrease in the amount of ischemia measured by thallium perfusion scintigraphy was demonstrated after training.

Noninvasive testing in the evaluation of myocardial ischemia: agreement among tests.

The purpose of this study was to investigate the relations among four exercise-induced phenomena--angina, ST segment depression, decrease in ejection fraction and thallium perfusion defects--and to determine their impact on aerobic capacity. One hundred fifty-six men (mean age 52 +/- 8 years) with documented coronary heart disease were studied with radionuclide ventriculography during supine bicycle exercise, thallium scintigraphy and treadmill testing with computerized electrocardiography and maximal oxygen uptake. Of 624 administered tests, 243 results (39%) were considered to indicate ischemia. The average number of abnormal tests was 1.6 per patient and, when considered as continuous variables, their results correlated poorly. Correlations did not improve when adjusting for heart rate achieved or by eliminating patients with coronary artery bypass surgery or myocardial infarction. Statistical methods of comparing degree of interest agreement yielded surprisingly weak relations among the four tests of ischemia. Treadmill performance was markedly impaired by angina, but much less impaired by other indicators of ischemia. It is concluded that the usual test responses implying ischemia have weak agreement when uniformly applied to patients with known coronary artery disease.

Nomogram based on metabolic equivalents and age for assessing aerobic exercise capacity in men.

OBJECTIVES: The goal of this study was to create a nomogram, based on maximal exercise capacity (in metabolic equivalents [METs]) and age, for assessing a patient's ability to perform dynamic exercise to quantify the level of physical disability or relative capacity for physical activity. BACKGROUND: Providing an estimation of exercise capacity relative to age is clinically useful. Such an estimate can be derived from measured or estimated maximal oxygen uptake (in METs) from treadmill exercise testing and age. It is an effective means of communicating to patients their cardiopulmonary status, encouraging improvement in exercise capacity and quantifying disability. METHODS: Exercise test results of 1,388 male patients (mean age 57 years, range 21 to 89) free of apparent heart disease who were referred for exercise testing for clinical reasons were retrospectively reviewed. This referral group as well as subgroups of active (n = 346) and sedentary (n = 253) patients were analyzed to determine norms for age and for age by decades for exercise test responses, including METs, maximal heart rate and maximal systolic blood pressure. Regression equations were calculated from this information, and a nomogram for calculating degree of exercise capacity from age and MET level achieved by a patient was created. A similar analysis was performed in a separate group of 244 apparently healthy, normal male volunteers (mean age 45 +/- 14 years, range 18 to 72) who underwent exercise testing with direct measurement of expired gases. RESULTS: Equations for predicted METs for age were derived for the entire clinical referral group (METs = 18.0-0.15[Age]) and for the subgroups of active (METs = 18.7-0.15[Age]) and sedentary (METs = 16.6-0.16[Age]) patients. All results achieved statistical significance, with p values < 0.001. In the volunteer group of normal men who performed exercise testing with ventilatory gas exchange, the decline in maximal heart rate and METs with age was not as steep as in the referral group. Although the normal group confirmed nomograms published previously among similar subjects, the equations derived from the patients differed from those previously reported; in contrast to previous studies using healthy volunteers, the equations and nomograms for the referral group are more appropriate for patients typically referred for testing in a hospital or office-based internal medicine practice. CONCLUSIONS: Norms for METs based on age are presented as well as population-specific nomograms that enable physicians to assess patients' exercise capacity relative to their age group.

Heart rate recovery: validation and methodologic issues.

OBJECTIVES: The goal of this study was to validate the prognostic value of the drop in heart rate (HR) after exercise, compare it to other test responses, evaluate its diagnostic value and clarify some of the methodologic issues surrounding its use. BACKGROUND: Studies have highlighted the value of a new prognostic feature of the treadmill test-rate of recovery of HR after exercise. These studies have had differing as well as controversial results and did not consider diagnostic test characteristics. METHODS: All patients were referred for evaluation of chest pain at two university-affiliated Veterans Affairs Medical Centers who underwent treadmill tests and coronary angiography between 1987 and 1999 as predicted after a mean seven years of follow-up. All-cause mortality was the end point for follow-up, and coronary angiography was the diagnostic gold standard. RESULTS: There were 2,193 male patients who had treadmill tests and coronary angiography. Heart rate recovery at 2 min after exercise outperformed other time points in prediction of death; a decrease of <22 beats/min had a hazard ratio of 2.6 (2.4 to 2.8 95% confidence interval). This new measurement was ranked similarly to traditional variables including age and metabolic equivalents but failed to have diagnostic power for discriminating those who had angiographic disease. CONCLUSIONS: Heart rate at 1 or 2 min of recovery has been validated as a prognostic measurement and should be recorded as part of all treadmill tests. This new measurement does not replace, but is supplemental to, established scores.

Exercise-induced silent ischemia: age, diabetes mellitus, previous myocardial infarction and prognosis.

The purposes of this study were 1) to determine the prognosis of silent ischemia in an unselected group of patients referred for exercise testing, and 2) to assess whether age or the presence of myocardial infarction or diabetes mellitus influences the prevalence of silent myocardial ischemia during exercise testing. The design was retrospective, with a 2 year mean follow-up period. The study group consisted of 1,747 predominantly male in-patients and outpatients referred for exercise testing at a 1,200 bed Veterans Administration hospital. The main result was that the mortality rate was significantly greater (p = 0.02) among patients with abnormal ST segment depression than in patients without ST depression. The presence or absence of angina pectoris during exercise testing was not significantly related to death. The prevalence of silent ischemia was not significantly different among patients categorized according to myocardial infarction or diabetes mellitus status, but was directly related to age. It is concluded that, in patients with an ischemic ST response to exercise testing, the presence or absence of angina pectoris during the test does not alter the risk of death. The prevalence of silent ischemia during exercise testing is not statistically different among patients with recent, past or no myocardial infarction or with insulin-dependent or noninsulin-dependent diabetes mellitus.

Effect of beta-adrenergic blockade on exercise performance in patients with chronic atrial fibrillation.

Beta-adrenergic blocking agents are commonly used in combination with digitalis to control excessive heart rate during exercise in patients with chronic atrial fibrillation. However, little is known about the effect of beta-adrenergic blockade on exercise capacity in these patients. Accordingly, a randomized, double-blind, cross-over placebo-controlled study was performed to assess the efficacy of celiprolol, a new cardioselective beta-blocker with partial intrinsic sympathomimetic activity, on exercise performance in nine men with chronic atrial fibrillation. All but one patient was receiving maintenance digitalis during the study. Heart rate, blood pressure and gas exchange variables were measured at rest and during treadmill exercise testing while the patients were receiving maintenance celiprolol or placebo. Significant reductions in heart rate and systolic blood pressure compared with control values were observed at submaximal exercise, at the gas exchange anaerobic threshold and at maximal exertion while the patients were taking celiprolol. However, oxygen uptake at the gas exchange anaerobic threshold during celiprolol therapy was 12.3 versus 14.0 ml oxygen/kg per min during placebo administration (a 12% difference, p less than 0.01). Similarly, oxygen uptake at maximal exertion during celiprolol therapy was 17.6 versus 21.0 ml/kg per min during placebo administration (a 16% difference, p less than 0.01). Treadmill time was also reduced during the celiprolol phase compared with placebo (11.3 versus 10.3 minutes; a 19% difference, p less than 0.01). These results indicate that in patients with atrial fibrillation the major beneficial effects of beta-adrenergic blockade--reduced submaximal and maximal exercise heart rate and blood pressure--must be weighed against the decrease in exercise capacity.

Comparison of the ramp versus standard exercise protocols.

To compare the hemodynamic and gas exchange responses of ramp treadmill and cycle ergometer tests with standard exercise protocols used clinically, 10 patients with chronic heart failure, 10 with coronary artery disease who were asymptomatic during exercise, 11 with coronary artery disease who were limited by angina during exercise and 10 age-matched normal subjects performed maximal exercise using six different exercise protocols. Gas exchange data were collected continuously during each of the following protocols, performed on separate days in randomized order: Bruce, Balke and an individualized ramp treadmill; 25 W/stage, 50 W/stage and an individualized ramp cycle ergometer test. Maximal oxygen uptake was 16% greater on the treadmill protocols combined (21.4 +/- 8 ml/kg per min) versus the cycle ergometer protocols combined (18.1 +/- 7 ml/kg per min) (p less than 0.01), although no differences were observed in maximal heart rate (131 +/- 24 versus 126 +/- 24 beats/min for the treadmill and cycle ergometer protocols, respectively). No major differences were observed in maximal heart rate or maximal oxygen uptake among the various treadmill protocols or among the various cycle ergometer protocols. The ratio of oxygen uptake to work rate, expressed as a slope, was highest for the ramp tests (slope +/- SEE ml/kg per min = 0.80 +/- 2.5 and 0.78 +/- 1.7 for ramp treadmill and ramp cycle ergometer, respectively). The slopes were poorest for the tests with the largest increments in work (0.62 +/- 4.0 and 0.59 +/- 2.8 for the Bruce treadmill and 50 W/stage cycle ergometer, respectively).(ABSTRACT TRUNCATED AT 250 WORDS)

Ventricular tachycardia during routine treadmill testing. Risk and prognosis.

Exercise-induced ventricular tachycardia during exercise testing is considered to increase risk during testing. Moreover, exercise-induced ventricular tachycardia has been considered to confer a poor prognosis although this has not been specifically studied. On a retrospective review of 3351 patients who had undergone routine clinical exercise testing between September 1984 and June 1989, we identified 55 patients with exercise-induced ventricular tachycardia. The mean follow-up was 26 months (range, 2 to 58 months). Fifty patients had nonsustained ventricular tachycardia during exercise testing and one of these patients died due to congestive heart failure during the follow-up period. Five patients had sustained ventricular tachycardia during exercise testing and one died suddenly 7 months after the test. Ventricular tachycardia was reproduced in only two of the 29 patients who underwent repeated exercise testing. Ventricular tachycardia during routine clinical exercise testing occurred rarely (prevalence of 1.5%) and was not associated with complications during testing. The total mortality in the exercise-induced ventricular tachycardia group (3.6%) was not significantly different from the mortality in the entire population (5.1%). Nonsustained ventricular tachycardia occurring during clinical exercise testing is not an independent marker of a poor prognosis.

Methodologic problems in exercise testing research. Are we solving them?

To evaluate the comparative effects of methodologic factors on the reported accuracies of two standard exercise tests, 56 publications comparing the exercise thallium scintigram with the coronary angiogram were analyzed for conformation to five methodologic standards. Analyzed were adequate definition of study group, avoidance of a limited challenge group, avoidance of workup bias, and blinded analysis of the coronary angiogram and myocardial scintigram. Study group characteristics and technical factors were also reviewed. Better conformation with methodologic standards was found than has been reported previously for treadmill exercise testing. Furthermore, study group characteristics and technical factors were better predictors of sensitivity and specificity than were methodologic deficiencies. Only workup bias and test blinding were significantly associated with test accuracy. The percentage of patients with previous myocardial infarction had the highest correlation and was independently and directly related to sensitivity and inversely related to specificity.

Angiographic patterns and severe coronary artery disease. Exercise test correlates.

In a Veterans Affairs Medical Center, we studied 607 male patients to determine whether patterns and severity of coronary artery disease could be predicted by means of standard clinical and exercise test data. We found significant differences in clinical, hemodynamic, and electrocardiographic measurements among patients with progressively increasing disease severity determined by angiography. Left main disease produced responses significantly different from those of three-vessel disease only when accompanied by a 70% or greater narrowing of the right coronary artery. Discriminant function analysis revealed that the maximum amount of horizontal or downsloping ST depression in exercise and/or recovery was the most powerful predictor of disease severity, with 2-mm ST depression yielding a sensitivity of 55% and a specificity of 80% for prediction of severe coronary artery disease (three-vessel disease plus left main disease). Patients with increasingly severe disease also demonstrated a greater frequency of abnormal hemodynamic responses to exercise.

Detection of latent coronary stenosis in conscious dogs: regional functional and electrocardiographic responses to isoprenaline.

A conscious animal model was developed in which coronary stenosis could be produced while regional myocardial function and local surface electrocardiograms were measured. Responses to isoprenaline stress in the presence of mild (latent) coronary stenosis were then examined. In the absence of coronary stenosis, isoprenaline produced increases in regional function and no change in the surface VCG; at higher doses, increases in the endocardial ST segments occurred. After partial coronary stenosis, which produced no apparent regional dysfunction or electrocardiographic changes, isoprenaline infusion for 3 min (0.02 microgram . kg-1 . min-1) rapidly produced decreases in percentage wall thickening (average 17 +/- 4%, mean +/- SE, P < 0.01) and increases in the mean sum of VCG ST segments by 0.23 +/- 0.06 mV (P < 0.05). 1 min after stopping isoprenaline, most dogs showed further significant deterioration of both measures of ischaemia, but by 5 min there was no significant mean change from control. We conclude that in the presence of latent partial coronary stenosis, stress due to mild sympathomimetic stimulation alone can rapidly induce regional myocardial ischaemia. Deterioration of regional myocardial contractile function during such stress can provide as sensitive means of detecting latent coronary obstruction.

Application of meta-analysis using an electronic spread sheet to exercise testing in patients after myocardial infarction.

Decision analysis is being applied to medical practice in order to achieve cost efficacy in health care delivery. Critical to this process is establishing the diagnostic and prognostic accuracy of medical tests and the effectiveness of interventions. Meta-analysis is an approach that applies statistical methods to groups of studies in order to extract consensus results. Electronic spreadsheets facilitate meta-analysis with their ability to store, sort, graph, and mathematically manipulate both the methodologic approaches and clinical findings of seemingly disparate studies. As an example, this application is demonstrated with an analysis of studies that were performed to evaluate the prognostic value of exercise testing in patients recovering from a myocardial infarction. The following conclusions were reached: (1) patients excluded from exercise testing have the highest mortality; (2) only subsets of patients have been tested resulting in highly selected patient samples that make findings difficult to generalize; (3) of the five exercise test responses, only an abnormal systolic blood pressure response and a poor exercise capacity predicted risk more frequently than by chance; (4) submaximal or predischarge testing has greater predictive power than postdischarge or maximal testing; and (5) exercise-induced ST segment depression only appears to be predictive of increased risk in patients with inferior-posterior myocardial infarctions. This approach to combining studies is important since even careful analysis of a single study cannot elucidate all of the complex interactions and selective biases that have occurred. However, comparison of many heterogeneous studies is at best an arduous and time-consuming task. This approach to using electronic spreadsheets to collate and analyze multiple studies facilitates recognition of the population characteristics, clinical factors, and methodologic considerations that affect outcome and allows the quick inclusion of additional studies for re-analysis and interpretation.

Factors affecting sensitivity and specificity of a diagnostic test: the exercise thallium scintigram.

Technical and methodological factors might affect the reported accuracies of diagnostic tests. To assess their influence on the accuracy of exercise thallium scintigraphy, the medical literature (1977 to 1986) was non-selectively searched and meta-analysis was applied to the 56 publications thus retrieved. These were analyzed for year of publication, sex and mean age of patients, percentage of patients with angina pectoris, percentage of patients with prior myocardial infarction, percentage of patients taking beta-blocking medications, and for angiographic referral (workup) bias, blinding of tests, and technical factors. The percentage of patients with myocardial infarction had the highest correlation with sensitivity (0.45, p = 0.0007). Only the inclusion of subjects with prior infarction and the percentage of men in the study group were independently and significantly (p less than 0.05) related to test sensitivity. Both the presence of workup bias and publication year adversely affected specificity (p less than 0.05). Of these two factors, publication year had the strongest association by stepwise linear regression. This analysis suggests that the reported sensitivity of thallium scintigraphy is higher and the specificity lower than that expected in clinical practice because of the presence of workup bias and the inappropriate inclusion of post-infarct patients.