Computer Methods and Programs in Biomedicine 50 (1996) 63-71
A computer program for comprehensive ST-segment depression/heart rate analysis of the exercise ECG test Rami Lehtinen* a, Henri V2nttinena, Ha.rri Sieviinenb, Jaakko Malmivuo” aRagnar
Granit Institute, bUKK-Institute for
Tarnpere University of Technology, Korkeakoulunkatu I, FIN-33720 Tampere, .Finland Health Promotion Research, Kaupinpuistonkatu 1, FIN-33520 Tampere, Finland
Received 14 February 1996; accepted 28 March 1996
AbStM!t The ST-segment depression/heart rate (ST/HR) analysishasbeenfound to improve the diagnosticaccuracyof the exercis,eECG test in detectingmyocardialischemia.Recently, three different continuousdiagnosticvariablesbased on the ST/HR analysishavebeenintroduced;the ST/HR slope,the ST/HR index and the ST/HR hysteresis. The latter utilisesboth the exerciseandrecoveryphases of the exerciseECGtest,whereasthe two formerarebasedon the exercise phaseonly. Thispresentarticle presentsa computerprogramwhichnot only calculatesthe abovethreediagnosticvariablesbut alsoplots the full diagramsof ST-segmentdepressionagainstheart rate during both exerciseand recovery phasesfor eachECG leadfrom given STlHR data. The programcan be usedin the exerciseECG diagnosisof daily clinicalpracticeprovided that the ST/HR data from the ECG measurement systemcan be linked to the program.At present,the mainpurposeof the programis to provide clinical andmedicalresearchers with a practicaltool for comprehensiveclinical evaluationand developmentof the ST/HR analysis.
Keywords:ExerciseEGG; Computeranalysis;Myocardial ischemia;ST/HR analysis;Software 1. Tntroduction Exercise ECG test remainsthe most widely used nonmvasive clinical examination for diagnosing myocardial ischemia and coronary artery disease. However, the diagnostic accuracy of the conventional end-exerciseST-segmentdepressionanalysis is limited to only about 70% in clinical populations [1,2]. Recently, various methods of ST-segment depression/heartrate (ST/HR) analysis such as the -+ Corresponding author, Tel.: +358 3 365 2524; fax: +358 3 365 2162; e-mail:
[email protected] 0169-2607/96/$15.00 PII:
ST/HR slope [3], the ST/HR index [4] and the ST/HR hysteresis [5] have been found to significantly improve the diagnostic accuracy of the exerciseECG test [3-l 11.Obviously, this improvement is due to a more physiologic nature of the ST/HR analysis; it interrelates a simple index of myocardial ischemia(ST-segmentdepression)to a simpleindex of myocardial oxygen demand (heart rate). To make clinicians confident about the value of the ST/HR analysis,a comprehensivevalidation and further development of this method are still needed. Unfortunately, most of the commercial IcomputerisedECG analysersdo not either include
0 1996 Elsevier Science Ireland Ltd. All rights reserved
SO l69-2607(96)01732-4
64
R. Lehtinen
et al. /Computer
Methods
and Programs
the ST/HR analysis or otherwise support the scientific clinical evaluation of the method. Therefore, the objective of this study is to facilitate further clinical evaluation of the ST/HR analysis by developing a computer program that performs a comprehensive ST/HR analysis of tlhe exercise ECG test. The program (1) calculates tlhe ST/HR hysteresis, ST/HR slope, ST/HR indelx, and (2) visualises the given ST/HR data during both the exercise and recovery phases for each lead of the exercise EGG lead system.
in Biomedicine
50 (19%)
63-71
0.10 -
STMR hysteresis= -0.13 mV
2. TL? methods of ST/HR analysis The plot of ST-segment depression against heart rate throughout the exercise and during the first three minutes of postexercise recovery, i.e., the ST/HR diagram, provides the basis for the ST/H:R analysis. In the first studies concerning ST/FIR analysis, the §T/HR slope utilised only the exercise pbase ST/HR data pairs measured immediately before starting the exercise, at the end of the each exercise stage and at the end of exercise [3]. Whereas, the ST/HR index, a simplification of the STMR slope, needed only the ST/HR data pair measured immediately before starting the exercise and the end-exercise ST/HR data pair [4]. In contrast, the ST/HR hysteresis utilised both the exercise and recovery phase ST/HR data of the ST/IIR diagram without setting any limits to the sampling interval of the ST/HR data pairs [5]. The ST/HR hysteresis is considered the primary method in tlhis study, the ST/HR slope and ST/HR index remaining as supplementary methods only. The determination of the ST/HR hysteresis from a single lead ST/HR diagram is illustrated in Fig. 1. In this case the ST/HR diagram consists of t.he pairs of the ST-segment depression and heart rate measnred from lead V5 immediately before starting the exercise, at the end of each minute of exercise, at the end-exercise and at the end of each of the first three consecutive minutes of postexercise recovery. For determining the STlHR hysteresis, first a continuous piecewise linear function between ST-segment depression and heart rate for tbe exercise phase was obtained by connecting the consecutive ST/HR data pairs of the exercise phase (solid lines in Fig. 1). Similarly, a continuous
PATIENT-00 1 LeadV5
HR.ret
HR,,,
-0.40 60
100 14.0 Heart rate (beats/min)
180
Fig. 1. Determination of the ST/HR hysteresis. In this diagram the ST depression is plotted in an upward direction and negative values represent ST-segment elevation.
piecewise linear function for the postexercise recovery phase was constructed by (connecting the consecutive ST/HR data pairs of the first three minutes of recovery phase starting from the STAIR data pair at the end-exercise (dashed lines in Fig. 1). Then the difference between these two piecewise linear functions was integrated over the heart rate from the minimum heart rate of recovery (HR,) to the maximum heart rate (HR,). Finally the integrated net difference (i.e. the area of hysteresis in Fig. 1) was divided by th.e heart rate difference of the integration interval (AHR,,,) in order to normalise the ST/FIR hysteresis with respect to the postexercise heart rate decrement. Consequently, the dimension of the ST/HR hysteresis is millivolt representing the average difference between the ST-segment depressions during the recovery phase and the exercise phase. The ST/HR slope was calculated by linear regression analysis with least squares approach. Only if the correlation coefficient (r) between heart rate and ST-segment depression was statistically significant (P < 0.05) the ST/HR. slope was accepted. The steepest ST/HR slope in each lead was obtained by comparing the statistically significant
R. Lehtinen
et al. /Computer
Methods
slope of the final three ST/HR points if any with that obtained by progressively including preceding points at earlier stages of exercise as suggested by Elamiu et al. [3] and Kliglield et al. [6]. The ST/HP. slope calculation is illustrated in Fig. 12. The dimension of the ST/HR slope is PVibeats per mimlte. The ST/HR index was calculated as the gradient between the ST/HR pair measured immediately before starting the exercise and ST/HR pair measured at the end-exercise as suggested by Detrano et al. [4]. The ST/HR index determination is illustrated in the example given in Fig. 2. The d.imension of the ST/HR index is pV/beats per minute. 3. Program description 3.1. Input format of the ST/HR data The input STMR data from 1 to 20 ECG leads can be provided to the program as an ASCII-file with extension sth (Table 1). The tile should begin with a global header that conveys the information about the number of ECG leads from which the ST-segment depression is determined (line l), the duration of each exercise stage in seconds (line 2) the ST,IHR sampling interval in seconds both for the exercise phase (line 3) and for the recovery phase (line 4), and the names of the leads, one name per one line in order of appearance corresponding to the columns of the ST/HR data. The global header ends with two empty lines. After the global header follows the data of individual patient(s) for which the global header is common. Patient data begin with a four line patient header consisting identification code of the patient (line l), the total number of ST/HR samples (line 2), the ordinal number of the endexercise ST/HR sample (line 3), and one empty or comment line (line 4). From the global and patient header, the proIgram reads only the first number or character string of the line. Thus after the first string and a space delimiter the lines may contain unlimited amount of comments as denoted by asterisks in Table i e Subsequent lines after the patient header con-
and Programs
in Biomedicine
50 (1996)
63-71
65
tain the heart rates in beats/mm (column 1) and ST-segment depressions in millivolts (next columns) from the ST/HR sample immediately before starting the exercise to the ST/HR sample at three minutes of recovery, one line per one ST/HR sample. The file structure requires that the heart rate and ST-segment depression values of same line (i.e. same sample) should be delimited with a space or spaces. The heart rate values should be given as integer values and the ST-segment depressions as real values with two decimals. Of note, the negative values of ST-segment depression denote the ST-segment elevations. If the &-file contains more than one patient, there should be two empty lines between the data of two consecutive patients. An actual example of the ST/HR of one patient (PATIENT-001) together with the above described global header is given in Table 1. In this example, the Mason-Likar modification of the standard 1Zlead system has been used in a exercise test with four-minute exercise stages and the ST/HR sampling interval has been 60 s both in the exercise and recovery phases. Of note, the minus sign in lead aVR indicates that this lead is inverted. 3.2. Calculation of the diagnostic variables The STiHR hysteresis, ST/HER index and ST/HR slope are calculated in the STMR calculation module as described above. In addition, the end-exercise ST-segment depression and the STsegment depression at 3 min of recovery are determined in the module. The flow chart of this module is given in Fig. 3. The calculation of the ST/HR hysteresis and determination of recovery ST-segment depression requires that the ST/HR data exist up to three minutes of recovery. Furthermore, the calculation of ST/HR slope requires that patient has performed more than two work loads (i.e. exercise stages) in the exercise test, since a minimum of three ST/HR data pairs is required for acceptable linear regression analysis of ST/HR slope calculation. The program automatically checks the validity of these requirements from the patient header. If these requirements are not valid or the ST/HR slope is not acceptable a value of -99.99 is given for the diagnostic variable in question.
66
R. Lehtinen
et al. /Computer
Methocb
and Programs
in Biomedicine
SO (1996)
63-71
Table 1 The format of the ST/HR data input tile 12 240 60 6Q I II III -aVR aVL aVF Vl v2 V2 V4. V5 V6
*/The */The *IThe */The */The */The */The */The */The */The *IThe *Tithe *Bhe *flhe */The */The
PATIE:NT-001 21 18
*/The identification code of the patient */The total number of ST/HR samples (i.e. rows of ST/HR data) */The ordinal number of the end-exercise ST/HR sample
95 107 110 lug 107 117 123 127 128 142 149 150 153 160 163 166 166 1’76 167 147 138
0.00 0.00
-0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 0.00 -0.02 -0.02 -0.01 -0.02 -0.03 -0.02 0.00 -0.08 -0.10 -0.03
number of the ECG leads duration of each exercise stage (in seconds) ST/HR sampling interval for exercise phase (in seconds) ST/HR sampling interval for recovery phase (seconds) ECG lead of column 2 (column 1 contains heart rate data) ECG lead of column 3 ECG lead of column 4 ECG lead of column 5 ECG lead of column 6 ECG lead of column 7 ECG lead of column 8 ECG lead of column 9 ECG lead of column 10 ECG lead of column 11 ECG lead of column 12 ECG lead of column 13
-0.04
-0.03 -0.06 -0.05 -0.06 -0.05 -0.06 -0.05 -0.06 -0.05 -0.05 -0.10 -0.09 -0.07 -0.08 -0.09 -0.09 -0.01 -0.19 -0.23 -0.13
-0.03 -0.03 -0.05 -0.04 -0.06 -0.04 -0.05 -0.04 -0.05 -0.04 -0.05 -0.08 -0.07 -0.06 -0.07 -0.06 -0.08 -0.02 -0.11 -0.13 -0.09
-0.02 -0.01 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.03 -0.06 -0.05 -0.04 -0.05 -0.06 -0.06 -0.01 -0.13 -0.17 -0.08
0.01 0.01 0.02 0.02 0.03 0.02 0.02 0.01 0.02 0.02 0.02 0.03 0.03 0.02 0.02 0.01 0.03 0.01 0.02 0.02 0.03
-0.03 -0.03 -0.05 -0.05 -0.66 -0.04 -0.05 -0.05 -0.05 -0.04 -0.05 -0.09 -0.08 -0.06 -0.07 -0.07 -0.08 -0.01 -0.15 -0.18 -0.11
As, an example, the calculation of the ST/FIR hysteresis, ST/HR slope and ST/HR index for one patient (PATIENT-001) is illustrated in Figs. 1 and 2, in which the diagnostic variables are determined from the lead V5. The full ST/HR data of the patient is given in Table 1.
-0.05 -0.06 -0.05 -0.04 -0.05 -0.04 -0.03 -0.01 0.00 0.01 0.02 0.03 0.04 0.03 0.04 0.04 0.05 0.02 0.10 0.07 0.04
-0.07 -0.07 -0.07 -0.05 -0.06 -0.04 -0.04 -0.04 -0.05 0.00 -0.03 -0.02 -0.03 -0.02 -0.02 -0.05 -0.04 -0.02 -0.08 -0.14 -0.06
-0.07 -0.08 -0.10 -0.07 -0.08 -0.06 -0.08 -0.08 -0.10 -0.05 -0.08 -0.08 -0.10 -0.08 -0.09 -0.1 I -0.12 -0.07 -0.22 -0.27 -0.13
-0.05 -0.06 -0.08 -0.06 -0.07 -0.05 -0.08 -0.09 -0.13 -0.07 -0.11 -0.11 -0.13 -0.12 -0.12 -0.14 -0.14 -0.08 -0.28 -0.32 -0.13
-0.03 -0.04 -0.06 -0.05 -0.05 -0.05 -0.07 -0.08 -0.09 -0.06 -0.09 -0.09 -0.11 -0.09 -0.08 -0.13 -0.13 -0.06 -0.26 -0.27 -0.11
-0.03 -0.03 -0.05 -0.04 -0.03 -0.03 -0.05 -0.06 -0.07 -0.04 -0.06 -0.07 -0.08 -0.06 -0.06 -0.08 -0.09 -0.01 -0.19 -0.22 -0.10
3.3. Visualisation of the ST/HR analysis The ST/HR visualisation module plots the full ST/I-IR diagrams for each EC6 lead complemented by the ST/HR hysteresis, ST/HR slope and ST/HR index values. The flow chart of the
R. Lehtinen
et al. /Computer
Methods
and Programs
in Biomedicine
50 (1996)
63-71
61
OT
-(),%O lLeadV5 60
I 140 100 Heart rate (beatdmin)
i 180
Fig. 2. Determination of the ST/HR slope (SLO) and the ST/HR index @ND). The number of ST/HR data pointsin the ST/FIR slope calculation is given in the brackets. In this specific example any acceptable (P < 0.05) ST/HR slope was not achieved.
ST/HR visualisation module is given in Fig. 4 and the graphical interface of the program in Fig. 5. As an input, the ST/HR visualisation module requires four different ASCII-files; one with extension sth containing the ST/HR data pairs, one with extension ~JW containing the calculated values of the ST/HR hysteresis, one with extension slo containing the calculated values of the ST/HR slope, and one with extension ind containing the calculated values of the ST/HR index. The three latter files are the output files from the ST/HR calculation module of the program (Fig. 3). When the desired ST/HR data file is opened the application searches for the data files of ST/HR hysteresis, ST/HR slope, and ST/HR index which have the same name as the open main file (e.g. samplel2.sth - samplel2.hys). Then the program reads only the headers of the &z-file and displays the patients and leads that the given file contains’. Now the user can select a patient by doubleclicking the patient code with a mouse. When the code is selected only the corresponding data section is searched from the files and read into the memory. When a new patient is selected the old data is replaced by the new one, thus minimising the amount of memory needed. However, if the. &z-file is large (> 1 MB) the search time for a patient data located at the end of the file mav.I
and STrec values to files Yhys and *.str
1 files *.ind and *.ste
0
END
Fig. 3. Flow chart of the ST/HR calculalion module.
increase up to several minutes in the case of a low plerformance computer. Therefore, it is advised to divide the patients in several files if the s&file is larger than 1 MB. All functions in the program can be used with a mouse or keyboard driven menus or by using shortcut keys solely. The application has also
68
R. Lehtinen
et al. / Computer
Methods
and Programs
in Biomedicine
50 (1996)
63-7~’
built-in on-line help option which provides assistance on usage and properties of the given software. 3.4. Output options of the program
j
content
Ij
Display
data
I/
Print
1
.._....._..........
patient data -
ilisplay parmlkad
----l----Y
Print
print
_..... r-+2
all
Print selected leads
r-
c
Print leads
Print selected lead(s)
I
-L-
.
Determine
_....
Scaling
no autoscale
nutoscale all I selected leads
Fig. 4. Flow
Use user
selected
chart of the ST/HI7
The ST/HR calculation module produces five output files. All these ASCII-files have the same name as the respective sth-file, only the extensions are different. The output files with extension hys contain the ST/HR hysteresis values, the files with extension slo contain the ST/HR slope values, the files with extension ind contain the ST/HR index values, the files with extension sre contain the endexercise ST-segment depression values and the files with extension str contain the values of STsegment depression at the three minutes of recovery. In all these output files the data of an individual patient are given in one row. This row starts with a patient identification colde Followed by the value determined from the firs,t lead, the second lead, etc. These values are delimited with spaces providing thus an easy importation of these output files to commercial database programs or statistical analysis packages. An (example of the output file with extension hys is given in Table 2. The program can display either the ST/HR diagrams of all leads of an individual patient simultaneously or a selected set of leads or a single lead (Fig. 5). The ST-segment millivolts (mV) is given in the vertic heart rate in beats per minute (bp zontal axis. The divisions of the vertical and horizontal grids are 0.10 mV and 10 bp:m, respectively. Lead names and the values of the ST/HR hysteresis, ST/HR slope and ST/‘HR index are displayed in a box at the left bottom corner of each diagram. Similarly, the symbols and colors in use are displayed in a box at the right bottom comer of the upper left ST/HR diagram. The identification code of a patient is displayed above the upper left ST/HR diagram. An automatic or a user defined scaling of STsegment depression and heart rate can be applied. In the single-lead-display the aut80matic scaling determines the axis ranges in a way that the hysteresis curve is as large as, possible. If the all/selected-leads-display mode is used the
leads
visualisation
module.
R. Lehtinen
et al. /Computer
Methods
and Programs
in Riomedicine
Fig, 5. IJser-interface of the program. In this example, the all-leads-display of PATIENT-001 SLO = ST/HR slope; IND = ST/HR index.
autoscaling determines the axis ranges so that all leads heavethe same ranges, scales, and zero levels, however, not allowing any curve to go out of the range. This option would provide a convenient visual comparison of different leads from an individual patient. The user defined axis ranges are applied if autoscale option is not selected. Table 2 The format
50 (1996)
69
63-71
is selected. HYS = ST/HR hysteresis;
In principle, the number of leads visualised by this program is not limited, but the display is optimised for 12 leads. If there are more than 12 leads on the display at the same time, the all-leadsdisplay will appear indistinct. The user, however, can take a closer look to a single lead by doubleclicking the lead name in the list of leads or cor-
of the ST/HR hysteresis output-file
PATIENT-
ID
I
II
111
-aVE
aVL
aVF
Vl
v2
v3
V4
V5
V6
PATIENTPATIENTPATIENTPATIENT-
001 002 003 004
-0.06 0.02 -0.01 0.01
-0.11 0.03 -0.04 -0.02
-0.05 0.01 -0.04 -0.03
-0.09 0.02 -0.03 -0.01
0.00 0.00 0.02 0.02
-0.08 0.02 -0.04 -0.03
0.04 -0.01 -0.05 -0.01
-0.07 0.05 -0.06 -0.03
-0.13 0.06 -0.07 -0.03
-0.14 0.05 -0.04 -0.01
-0.13 0.04 -0.03 0.02
-0.11 0.03 -0.03 0.01
70
R. Lehtinen
et al. /Computer
Methos&
and Pro,yrams
responding ST/HR diagram on the all/selectedleads-display. If the user wishes to study some specific set of leads the given set can be selected/unselected from the all-leads-display by clicking the corresponding diagram. Selected diagrams become highlighted. These selections are valid until the all-leads-display mode is selected again. The user may compare the selected leads from different patients just by selecting a new patient for display when this option is used. All display modes can be zoomed in and out. The aspect ratio can be maintained or the image can be zoomed horizontally and vertically independently. Relative vertical and horizontal scales are displayed at the bottom right-hand sifde corner of the display. The: graphical representation of ST/HR da.ta shown on the display can also be printed or copied to clipiboard as a line drawing. From the clipboard the fignre can be pasted into other Windows applications. There are several possibilities to print figures even in colors if the printer supports color printing. The printing dialog has options for printing all leads, selected lead(s) and all patients in tlhe selected sth-file. The scaling of the printing can also be selected; figures may either fill all the printable area or be based on a predefined scale of 10 mm/O.10 mV or a user defined scale. Landscaipe oriemation of the paper is preferable, because tihe aspect ratio of the printout corresponds to the aspect ratio of the view on the display. 4. Couclusion The described program not only calculates the three diagnostic variables of ST/HR analysis but also displays or prints the full diagrams of STsegment depression against heart rate during both the exercise and recovery phases for each ECG lead. These properties certainly facilitate the cornprehensive clinical evaluation of the ST/HR analysis both by visualising the complex pattern of the ST/HR diagram and providing the three continuous diagnostic variables for comparison. Furthermore, the program can be used in the exercise ECG diagnosis of daily clinical practice provided that the ST/WR data from the ECG measurement system can be linked to the program. At present,
in Biomedicine
50 (1996)
63-71
the main purpose of the program is to provide clinical and medical researchers with a practical tool for comprehensive clinical evaluation and development of the ST/HR analysis. 5. Hardware and software specifications The only requirement for the program Microsoft Windows version 3.1 or later. 6. Availability
is
of tbe program
The program written in C++ for Windows (Borland 4.0) is available from the corresponding author. More detailed documentation will also be provided upon request. Acknowledgements The financial support of the Academy of Finland, Alfred Kordelin Foundation, Finnish Cultural Foundation (Fund of Pirkanmaa), Finnish Cultural Foundation, Ragnar Granit Foundation, Tampere Science Foundation, and Wihuri Foundation is gratefully acknowledged. References 111R. Gianrossi, R. Detrano, D. MuPvihiJl, K. Lehmann, P.
Dubach, A. Colombo et al., Exercise-induced ST depression in the diagnosis of coronary artery disease: A metaanalysis, Circulation 80 (1989) 87-98. 121 R. Detrano, Variability in the accuracy of the exercise ST-segment in predicting the coronary angiogram: How good can we be? J. Electrocardiol. 24 (1991) 54-61. 131 M. Elamin, D. Mary, D. Smith and R. Linden, Prediction of severity of coronary heart disease using slope of submaximal ST-segment/heart rate relationskip, Cardiovasc. Res. 14 (1980) 681-691. 141 R. Detrano, E. Salcedo, M. Passalacqua and R. Friis, Exercise electrocardiographic variables: a critical appraisal, J. Am. Coll. Cardiol. 8 (1986) 836-847. 151 R. Lehtinen, H. Sievlnen, J. Viik, ‘V. Turjanmaa, K. Niemell and J. Malmivuo, Accurate detection of Coronary Artery Disease by integrated analysis of the STsegment depression/heart rate patterns during both the exercise ECG test, Am. J. Cardiol. 78 (1996) in press. [61 P. Kligfield, 0. Ameisen and P.M. Okin, Heart rate adjustment of ST segment depression for improved detection of coronary artery disease, Circulation 79 (1989) 245-255. [71 M. Schiariti, M. Ciavolella, P.E. Puddu, C. Giannitti, D. Scali, N. Schad et al., ST/HR slope and improved exer-
R. Lehtinen
et al. /Computer
Methods
cise:ECG detection of myocardial ischemia in patients with suspected coronary artery disease, J. Electrocardiol. 24 (1991) 307-314. [S] H. Sieviinen, L. Karhumiiki, I. Vuori and J. Mahnivuo, Improved diagnostic performance of the exercise ECG test by computerized multivariate ST-segmentiheart rate analysis, J. Electrocardiol. 24 (1991) 129-143. [9] H. Sievanen, L. Karhumaki, I. Vuori and J. Mahnivuo, Compartmental multivariate analysis of the exercise ECGs for accurate detection of myocardial ischaemia, l&d. Biol. Eng. Comput. 32 (1994) S3-S8.
and Programs
in Biomedicine
50 (1996)
63-71
71
[lo] R. Lehtinen, H. Sieviinen, A. Uusitalo, K. Niemell, V. Turjanmaa and J. Mahnivuo, Performance characteristics of various exercise electrocardiographic classifiers in different clinical populations, J. Electrocardiol. 27 (1994) 11-22. (111 P.M. Okin and P. Kligfield, Heart rate adjustment of ST segment depression and performance of the exercise electrocardiogram: A critical evaluation, J. Am. Coll. Cardiol. 25 (1995) 1726-1735.
