Improved Detection of Coronary Artery Disease by Novel Methods of Exercise ECG Analysis - A Multicenter Study
Novel computerized method for comprehensive analysis of exercise electrocardiogram (ECG) to support the doctor's decision making in diagnosing coronary artery disease (CAD) has been developed in Tampere University of Technology. The new method designated ST/HR hysteresis observes both the exercise and recovery phases of the exercise test and proportionates the ST-segment depression (index of myocardial ischemia) to the heart rate (index of myocardial oxygen demand). According to the clinical validation performed in co-operation with Tampere University Hospital with 347 patients, the ST/HR hysteresis improved the diagnostic accuracy of the exercise ECG in detection of CAD nearly up to 90% from the present 75%. In addition, the method has been shown to be more reproducible and less sensitive to the ST-segment depression measurement point and used ECG-leads.
In the diagnosis of CAD the exercise test is in a key position, since based on its positive result the patient is referred for further examinations, e.g., myocardial isotope imaging or coronary angiography. Exercise test is noninvasive, inexpensive and almost riskless. At present, every fourth exercise ECG is interpreted falsely. As a consequence, many patients needing treatment are dropped out from further examination, and on the other hand staff loading and expensive further examinations are performed for healthy subjects.
The primary working hypothesis of the project is that the diagnostic accuracy of the exercise ECG analysis in the detection of the presence, severity and jeopardy of CAD can be improved significantly, up to about 90% from the present 75%, by the novel methods of ST- segment/heart rate analysis.
Before the ST/HR hysteresis can be widely taken into a clinical practise, the results achieved have to be validated in a multicenter approach. For this purpose the exercise ECGs of 481 patients have been digitally stored in Turku University Central Hospital (TUCH) and the exercise ECGs of 675 patients stored in Helsinki University Central Hospital (HUCH). The myocardial isotope imaging in TUCH material and coronary angiography in HUCH material has been performed for every patient. These images have been interpreted considerably more detailed than in normal clinical practice enabling the validation and further development of the method. It is an objective to expand the multicenter study also to other countries of Europe and research collaboration with Lund University has already been agreed.
In addition, the computer program enabling the clinical use of the ST/HR hysteresis will be developed further especially the user-interface need to be built to meet the requirements of hospital routine. In connection to the multicenter study the possibility to develop the method further to detect also the severity and jeopardy of CAD will be studied. The further development will not only be based on statistical analysis of the patient materials but also the human body volume conduction model developed in Tampere University of technology will be applied in simulation of myocardial ischemia.
The socio-economic impact of the proposed project will be considerable. For example, if the recently published 15% improvement in specificity and 18% improvement in sensitivity obtained by using ST/HR hysteresis can be verified in the proposed multicenter study, the annual amount of false positive diagnosis will be estimated to reduce by more than 400 000 patients resulting in total savings of more than 200 million ECUs per year in health care costs in Europe. What is even more important, the amount of patients with CAD who are misdiagnosed (false negatives) could be reduced by more than 40 000 patients annually, improving the quality of life and decreasing markedly the risk of sudden death of these patients.
1. BACKGROUND
1.1 SUMMARY OF THE ISSUES INVOLVED
Coronary artery disease (CAD) is the major cause of premature, i.e., before the age of 60 years, death in Europe. The treatment of CAD requires considerable medical and economical resources. Therefore, it is important to improve the accuracy of the diagnosis in order to identify those subgroups of patients which are likely to benefit most from therapeutic interventions in terms of improved prognosis, working capacity and quality of life.
Generally, the first clinical test performed for a patient under CAD suspicion is the exercise ECG test. If the test result is positive the patient may be referred to exercise echocardiography or isotope SPECT myocardial imaging, or in case the need for surgical operation is obvious, directly to coronary angiography in order to identify the site(s) of the coronary artery stenosis.
A weak point of the present diagnostic procedure is the limited diagnostic accuracy of the conventional analysis of the exercise ECG test. Since the exercise ECG test is generally the first and most widely used clinical method for diagnosing CAD, any improvement in the exercise ECG analysis would have extremely high socio-economic impact. The purpose of the proposed project is to improve the diagnostic accuracy of the exercise ECG analysis.
1.2 CURRENT STATE OF KNOWLEDGE
According to the meta-analysis of Gianrossi et al. (1989) consisting of 147 published reports, the mean specificity and mean sensitivity of the exercise ECG test with conventional ST- segment analysis are only 77% and 68%, respectively. To improve the diagnostic accuracy of exercise ECG test, 7 different indexes of ST-segment/heart rate (ST/HR) analysis have been developed and preliminary evaluated in different centers at least in Canada, Finland, France, Hungary, Italy, Japan, Sweden, the Netherlands, United Kingdom, and USA.
The results of the most recent methods of ST/HR analysis combining both the exercise and recovery phases are encouraging, however, no comprehensive multicenter study have not been performed which is an essential requirement before it is reasonable to take a new method into a clinical practice. Due do the methodological diversity it is even impossible to perform meta-analysis of these previous reports.
We have developed a novel computerised method of ST/HR analysis (Lehtinen et al., 1996a) combining the exercise and recovery phases, designated ST/HR hysteresis. In the study population of 347 patients, the method was significantly more accurate in detecting of the presence of CAD than the conventional end-exercise ST-segment depression (Lehtinen et al., 1996b). In this study, the ST/HR hysteresis improved the specificity of exercise ECG test by 15% and sensitivity by 18% compared to the conventional positive criterion of end- exercise ST depression 0.10mV. This size of improvement would certainly be clinically relevant and warrants for multicenter study. Furthermore we have shown that ST/HR hysteresis is more reproducible (Lehtinen et al. 1997a), less sensitive to ST-segment measurement point (Lehtinen et al. 1997b) and ECG-lead selection (Viik et al., 1997). In addition, we have performed modelling study investigating the linear relationship between ST-segment depression and heart rate by simulating myocardial ischemia with a thorax volume conductor model (Hyttinen et al. 1997).
1.3 SOCIO-ECONOMIC IMPACT
The exercise ECG test is the most widely used method for detection of CAD. At a rough estimate, approximately 4 million exercise ECG tests are performed each year in Europe. The prevalence of CAD among the exercise tested patients can be estimated to be approximately 10%. Thus with the conventional exercise ECG analysis having mean specificity of 77% and mean sensitivity of 68% (Gianrossi et al., 1989), more than 800 000 patients without CAD are annually misdiagnosed. Most of these patients are unnecessarily referred for expensive further examination or given unnecessary medication. If the average cost of these unnecessary procedures is estimated to be for example 500 ECU per patient, the result is that more than 400 million ECUs are spent for these unnecessary treatments annually. On the other hand, in each year more than 100 000 patients with CAD are diagnosed incorrectly, increasing the risk of sudden death of these patients.
For example, if the 15% improvement in specificity and 18% improvement in sensitivity obtained by the ST/HR hysteresis can be verified in the proposed multicenter study, the annual amount of false positive diagnosis could be reduced by more than 400 000 patients resulting in total savings of more than 200 million ECUs per each year in health care costs in Europe. What is even more important, the amount of incorrectly diagnosed patients with CAD could be reduced by more than 40 000 patients annually, improving the quality of life and decreasing markedly the risk of sudden death of these patients.
1.4 BENEFITS
The project will aim to validate in a European setting the diagnostic accuracy of new improved methods of ST/HR analysis in detection of CAD, which disease is the most frequent cause of premature death in Europe. If the study will succeed in its purpose, the contribution to European health and wealth will be considerable:
1.5 LIST OF THE MOST RELEVANT PUBLICATIONS OF THE GROUP
Lehtinen R, Sievänen H, Uusitalo A, Niemelä K, Turjanmaa V, Malmivuo J. Performance Characteristics of Various Exercise ECG Classifiers in Different Clinical Populations. Journal of Electrocardiology 1994 January;27(1):11-22. IF93 = 0.460
Hyttinen J, Viik J, Lehtinen R, Plonsey R, Malmivuo J. Computer Model Analysis of the Relation of ST-Segment and ST/HR Slope Response to the Constituents of the Ischemic Injury Source. Journal of Electrocardiology 1997 July; 30(3):161-174. IF95 = 0.174
Lehtinen R, Vänttinen H, Sievänen H, Malmivuo J. A computer program for comprehensive ST-segment depression/heart rate analysis of the exercise ECG test. Computer Methods and Programs in Biomedicine 1996 June;50(1):63-71. IF94 = 0.400
Lehtinen R, Sievänen H, Viik J, Turjanmaa V, Niemelä K, Malmivuo J. Accurate Detection of Coronary Artery Disease by Integrated Analysis of the ST-Segment Depression/Heart Rate Patterns During the Exercise and Recovery Phases of the Exercise Electrocardiography Test. American Journal of Cardiology 1996 November 1;78(9):1002-1006. IF94 = 2.253.
Lehtinen R, Sievänen H, Viik J, Vuori I, Malmivuo J. Reproducibility of the ST-Segment Depression/Heart Rate Analysis of the Exercise Electrocardiographic Test in Asymptomatic Middle-Aged Population. American Journal of Cardiology 1997 May 15;79(10):1414-1416. IF95 = 2.238.
Viik J, Lehtinen R, Turjanmaa V, Niemelä K, Malmivuo J. The effect of lead selection on traditional and heart rate-adjusted ST segment analysis in the detection of coronary artery disease during exercise testing. American Heart Journal 1997 September 1;134(3):488-494. IF95 = 1.290.
Lehtinen R, Sievänen H, Turjanmaa V, Niemelä K, Malmivuo J. Effect of ST-segment measurement point on performance of exercise ECG analysis. International Journal of Cardiology 1997;61(3):239-245. IF95 = 0.418.
Viik J, Lehtinen R, Turjanmaa V, Niemelä K, Malmivuo J. Correct utilization of exercise ECG in discrimination of men with coronary artery disease from patients with low likelihood of coronary artery disease using peak exercise ST-segment depression. American Journal of Cardiology 1998 April;81 (in press).
2 OBJECTIVES AND METHODS
The hypothesis of the study: The diagnostic accuracy of the exercise ECG analysis in the detection of the presence and severity of coronary artery disease and myocardial ischemia can be improved significantly, up to about 90% from the present 75%, by the new methods of ST- segment/heart rate analysis.
Table I. The objectives of the project | |
I | To clinically evaluate the validity of the improved diagnostic accuracy of the ST/HR hysteresis in detection of the presence and severity of CAD and determine the diagnostic criteria to be used in clinical practice |
II | To work out a recommendation for harmonised clinical use of ST/HR hysteresis with respect to the diagnostic criteria, signal processing, ECG-leads, and ST-segment measurement point |
III | To provide basis for clinical implementation of ST/HR hysteresis in commercial exercise analysers accessible to clinical users |
IV | To establish a CD-ROM based European exercise ECG data library |
2.2 METHODOLOGY TO BE APPLIED
2.2.1 Study population
The exercise ECGs of 481 patients have been digitally stored in Turku University Central Hospital (TUCH) and the exercise ECGs of 675 patients stored in Diakonissalaitos in Helsinki. The myocardial isotope imaging in TUCH material and coronary angiography in Diakonissalaitos material has been performed for every patient. These images have been interpreted considerably more detailed than in normal clinical practice enabling the validation and further development of the method.
The repeated exercise ECG tests performed for normal subjects as a part of follow-up studies in UKK Institute has been stored for us for later analysis. This study population includes 100 male subjects (35-55 years old) who have all performed treadmill exercise ECG test twice and 150 female subjects (55-65 years old) who all have performed bicycle exercise ECG test three times. The clinical status of these subject have been carefully screened in connection to each exercise test and the likelihood of CAD can be considered to be very low. These study populations enables the studies on reproducibility of the ST/HR hysteresis.
2.2.2 Methods of ST/HR analysis
The methods of ST/HR analysis includes the ST/HR hysteresis, the Stress-recovery index (SRI), the quantitative HR-recovery loop, MUSTA, HR-recovery loop, ST/HR index, and ST/HR slope. These diagnostic variables will be determined by existing computer programs. First the ST- segment depression and heart rate data pairs will be extracted from the raw ECG throughout the exercise and recovery phase of the exercise test. Then the ST/HR data will be visualised as ST/HR diagrams and the seven different diagnostic variables will be calculated as described in the literature (Lehtinen et al, 1996b; Bigi et al, 1994; Herpin et al, 1996; Siev nen et al., 1991, Okin et al, 1989; Detrano et al, 1986, Elamin et al, 1980). The ST/HR diagrams determined from lead V5 in four different cases are presented in Figure 2.1. For an example of these methods, the determination of the ST/HR hysteresis is graphically presented in this Figure. The ST/HR slope will be calculated by linear regression analysis with least squares approach. If the correlation coefficient (r) will be statistically significant (p < 0.05) the ST/HR slope is accepted. The ST-segment depression and heart rate data pairs will be measured at starting the exercise when the patient is on the bicycle, at the end of each exercise stage and at peak exercise. The steepest ST/HR slope in each lead will be obtained by comparing the statistically significant slope of the final three points with that obtained by progressively including further points at earlier levels of exercise as suggested by Elamin et al. (1980) and Kligfield et al (1989).
The ST/HR index will be calculated as the gradient between the ST/HR pairs at the start-exercise and at the end-exercise as suggested by Detrano et al.(1986).
FIGURE 2.1. Plots of ST-segment depression against HR during both the exercise and first three minutes of postexercise recovery (i.e., ST/HR diagrams) illustrating the computerized determination of the ST/HR hysteresis from a single lead. In these diagrams, ST depression is plotted in an upward direction on the vertical axis, and the negative values represent ST elevation. Typical ST/HR diagrams are shown for a clinically normal subject (upper left panel) with an unambiguously negative hysteresis in ST depression against HR (i.e., clockwise hysteresis loop in recovery) and for a patient with coronary artery disease (upper right panel) with an unambiguously positive hysteresis in ST depression against HR (i.e., counterclockwise hysteresis loop in recovery). In the lower panels, more complex ST/HR diagrams are shown for a normal patient (lower left panel) and for a patient with coronary artery disease (lower right panel), in which cases the direction of the hysteresis in ST depression against HR was changed during the postexercise recovery phase. In graphical sense, the ST/HR hysteresis was obtained by subtracting the area of negative hysteresis from the area of positive hysteresis and dividing the resulted net area by the HRrec. HRexe = maximum HR; HRrec = the minimum heart rate of recovery; HRrec = HRexe-HRrec; 3v-CAD = three-vessel coronary artery disease.
2.2.3 Exercise ECG test
All subjects are tested on a bicycle ergometer. The exercise protocols are individualised to some extent depending on patient's physical condition. The protocol of each center are harmonised as well as possible.
The ECG recordings are made with a commercial ECG recording system supplemented by a system (hardware, software, PC and writing CD-ROM station) that enables the digital storage of raw exercise ECG signal. The ECG lead system used in the exercise test will be the Mason- Likar modification of the standard 12-lead system. Exercise tests will be sign- or symptom- limited maximal tests using recommended criteria for termination.
2.2.4 Coronary angiography
The selective coronary angiography will be performed with the Judkins technique. Each coronary artery will be imaged in multiple views in all cases. The degree of stenosis will be defined as the greatest percentage reduction of luminal diameter in any view compared with the nearest normal segment without knowledge of the exercise ECG data. Coronary artery disease will be considered significant when 50% or more luminal narrowing is observed at least in one major coronary artery: in the left main, in the left anterior descending, in the left circumflex or in the right coronary artery. Each member of the expert panel will interpret the angiograms separately and fill the especially designed data collection forms. Forms will be collected and analysed and the cases of disagreement will be solved by the panel consensus meeting.
2.2.5 Isotope SPECT myocardial imaging
The isotope SPECT myocardial perfusion imaging will follow the standard clinical routine. Abnormalities in the myocardial perfusion will be identified by the presence of abnormal distribution images. Distribution images are determined by computer analysis of the results obtained from the single photon emission computed tomography (SPECT).
2.2.6 Statistical methods
The data of each center will be analysed both separately and pooled. The performance characteristics of the exercise ECG variables in the detection of the presence of CAD or myocardial ischemia and multivessel CAD will be compared by analysing the receiver operating characteristic (ROC) curves. The area under the ROC curve represents the overall diagnostic performance; i.e., the probability that a random pair of patients with and without CAD will be correctly diagnosed (Hanley & McNeil, 1982). The differences between the area under the receiver operating characteristic curves of the ST/HR hysteresis and those of the other diagnostic variables will be compared using a computer program (Vida, 1993) for nonparametric analysis of correlated receiver operating characteristic curves (DeLong et al.1988).
The primary objective of the project is to compare the diagnostic performance characteristics of the ST/HR hysteresis and the conventional diagnostic variable, the end-exercise ST depression. Based on the previous study (Lehtinen et al., 1996b), the area under the ROC curves of the end-exercise ST depression is estimated to be 0.775 and that of the ST/HR hysteresis to be 0.875. With these estimates, to achieve the 95% statistical power in comparing the ROC areas with alpha-level of 0.05 (one-tailed), at least 209 patients with disease and 209 patients without disease will be needed to be collected at each center (Hanley et al., 1982).
The reproducibility of the exercise ECG variables between the repeated measurements will be analysed as suggested by Bland and Altman (1986). Definition of reproducibility is 1.96 times the standard deviation of the difference between the repeated measurements.
3. RESEARCHERS AND RESOURCES
Researchers:
Collaborators and managers:
Research will be performed in Tampere University of Techology. Clinical centers include Tampere University Hospital, Turku University Hospital, Diakonissalaitos of Helsinki and UKK Institute. In these centers, the data is collected, computer program will be tested and medical expertise achieved. International collaboration will be done with Lund University with the ECG signal processing group lead by assoc. prof. Olle Pahlm, MD and assoc. prof. Leif Sörnmo, PhD.
4. EXPLOITATION PLANS
4.1 DELIVERABLES FOR EXPLOITATION
The results of the proposed project will improve significantly the scientific and technical basis for further possible product development. On completion of the project, there will be the following main deliverables which deserve special attention for exploitation:
REFERENCES
Bigi R, Maffi M, Occhi G, Bolognese L, Pozzoni L: Improvement of identification of multivessel disease after acute myocardial infarction following stress-recovery analysis of ST depression in heart rate domain during exercise. Eur Heart J 15:1240-1246, 1994.
Bland JM, Altman, DG: Statistical methods for assessing agreement between two methods of clinical measurement. The Lancet 8, 307-310, 1996.
DeLong ER, DeLong DM, Clarke-Pearson DL: Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 44: 837-845, 1988.
Detrano R, Salcedo E, Passalacqua M, Friis R: Exercise electrocardiographic variables: a critical appraisal. J Am Coll Cardiol 8:4, 836-847, 1986.
Elamin M, Mary D, Smith D, Linden R: Prediction of severity of coronary heart disease using slope of submaximal ST-segment/heart rate relationship. Cardiovasc Res 14:14, 681-691, 1980.
Gianrossi R, Detrano R, Mulvihill et al.: Exercise-induced ST depression in the diagnosis of coronary artery disease: A meta-analysis. Circulation 80:1 87-98, 1989.
Hanley JA, McNeil BJ: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 143:1, 29-36, 1982.
Herpin D, Ferrandis J, Borderon P et al.: Comparison of the diagnostic accuracy of different methods of measurement heart rate - adjusted ST-segment depression during exercise testing for identification of coronary artery disease. Am J Cardiol 76:1147-1151, 1995.
Herpin D, Ferrandis J, Couderq C et al: Usefulness of a quantitative analysis of the recovery phase patterns of the ST- segment depression in the diagnosis of CAD. Am J Med 101:592-598, 1996.
Hyttinen J, Viik J, Lehtinen R, Plonsey R, Malmivuo J. Computer model analysis of the relation of ST-segment and ST/HR slope response to the constituents of the ischemic injury source. Journal of Electrocardiology 30(3):161-174, 1997.
Lehtinen R, Sievänen H, Uusitalo A, Niemelä K, Turjanmaa V, and Malmivuo J: Performance Characteristics of Various Exercise ECG Classifiers in Different Clinical Populations. J Electrocardiol 27:1, 11-22, 1994
Lehtinen R, Vänttinen H, Sievänen H, Malmivuo J. A computer program for comprehensive ST-segment depression/heart rate analysis of the exercise ECG test. Computer Methods and Programs in Biomedicine 50(1):63-71, 1996a
Lehtinen R, Sievänen H, Viik J, Turjanmaa V, Niemelä K, Malmivuo J. Accurate detection of coronary artery disease by integrated analysis of the ST-segment depression/heart rate patterns during the exercise and recovery phases of the exercise electrocardiography test. American Journal of Cardiology 78(9):1002-1006, 1996b.
Lehtinen R, Sievänen H, Viik J, Vuori I, Malmivuo J. Reproducibility of the ST-segment depression/heart rate analysis of the exercise electrocardiographic test in asymptomatic middle-aged population. American Journal of Cardiology 79(10):1414-1416, 1997a
Lehtinen R, Sievänen H, Turjanmaa V, Niemelä K, Malmivuo J. Effect of ST-segment measurement point on performance of exercise ECG analysis. International Journal of Cardiology 61(3):239-245, 1997b.
Okin PM, Ameisen O, Kligfield P: Recovery-phase patterns of ST segment depression in the heart rate domain. Identification of coronary artery disease by the rate-recovery loop. Circulation 80:3, 533-541, 1989.
Sievänen H, Karhumäki L, Vuori I, Malmivuo J: Improved diagnostic performance of the exercise ECG test by computerized multivariate ST-segment/heart rate analysis. J Electrocardiol 24:2, 129-143, 1991.
Sievänen H, Karhumäki L, Vuori I, Malmivuo J: Compartmental multivariate analysis of the exercise ECGs for accurate detection of myocardial ischaemia. Med Biol Eng Comput 32 (Suppl.): S3-S8, 1994.
Vida S: A computer program for non-parametric receiver operating characteristic analysis. Comput Meth Prog Bio 40: 95-101, 1993.
Viik J, Lehtinen R, Turjanmaa V, Niemel K, Malmivuo J. The effect of lead selection on traditional and heart rate-adjusted ST segment analysis in the detection of coronary artery disease during exercise testing. American Heart Journal 134(3):488-494, 1997.
Last updated 1998-03-10