The CAST RR Interval Sub-Study Database

This database is described in

Stein PK, Domitrovich PP, Kleiger RE, Schechtman KB, Rottman JN. Clinical and demographic determinants of heart rate variability in patients post myocardial infarction: insights from the Cardiac Arrhythmia Suppression Trial (CAST). Clin Cardiol 23(3):187-94; 2000 (Mar)

Please cite this publication when referencing this material, and also include the standard citation for PhysioNet:

Goldberger AL, Amaral LAN, Glass L, Hausdorff JM, Ivanov PCh, Mark RG, Mietus JE, Moody GB, Peng C-K, Stanley HE. PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals. Circulation 101(23):e215-e220 [Circulation Electronic Pages;]; 2000 (June 13).


The Cardiac Arrhythmia Suppression Trial (CAST) was a landmark NHLBI-sponsored study designed to test the hypothesis that the suppression of asymptomatic or mildly symptomatic ventricular premature complexes (PVCs) in survivors of myocardial infarction (MI) would decrease the number of deaths from ventricular arrhythmias and improve survival. Enrollment required an acute MI within the preceding 2 years and 6 or more PVCs per hour during a pre-treatment (qualifying) long-term ECG (Holter) recording. Those subjects enrolled within 90 days of the index MI were required to have left ventricular ejection fractions less than or equal to 55%, while those enrolled after this 90 day window were required to have an ejection fraction less than or equal to 40%. CAST enrolled 3,549 patients in all.

After initial qualification, patients in CAST were randomly assigned to receive encainide, flecainide, moricizine (antiarrhythmic drugs) or a placebo. Patients who had significant suppression of PVCs with a particular agent were then continued on that agent or on placebo. In April of 1989, the Data and Safety Monitoring Board recommended that the encainide and flecainide arms of the study be discontinued because of excessive mortality in the drug arms of the trial primarily due to arrhythmia, acute MI with shock, or chronic congestive heart failure.

CAST was then followed by CAST-II, which involved continuation of the moricizine arm of the CAST and placebo. CAST-II was divided into two blinded, randomized phases: an early, 14-day exposure phase that evaluated the risk of starting treatment with moricizine after MI and a long-term phase that evaluated the effect of moricizine on survival after MI in patients whose PVCs were either adequately suppressed by moricizine or only partially suppressed. CAST-II was stopped early because long-term treatment with moricizine after an MI was associated with a trend to excess mortality as compared with no treatment or placebo.

The CAST-II study authors concluded that as with the other antiarrhythmic agents used in CAST (flecainide and encainide), the use of moricizine to suppress asymptomatic or mildly symptomatic ventricular premature complexes after MI was not only ineffective but also harmful.

In order to investigate further the relationship between survival and changes in heart rate variability in response to anti-arrhythmic treatment, Phyllis K. Stein, Ph.D., of Washington University, St. Louis, obtained pre-treatment (qualifying) and on-therapy (suppression) long-term ECG recordings from participants in CAST (including CAST II) from the CAST data coordinating center. She selected for study a subset of the CAST patients, consisting of those who

These criteria were satisfied by 734 patients, of whom 69 died during the study period. The CAST RR Interval Sub-Study Database consists of RR interval time series from the pre-treatment and on-therapy recordings from these patients. The database also includes data from 75 additional subjects for whom only one acceptable recording was available but who satisfied the other selection criteria. A small number of subjects included in this database were excluded from Dr. Stein’s study because of the occurrence of atrial fibrillation or paced rhythm.

All recordings were scanned by an experienced research arrhythmia analyst, using standard Holter analysis procedures. Beat annotation files were edited in a second pass to identify improperly measured RR intervals to be excluded from further analysis. A limited amount of additional information has been recorded in a header (.hea) file for each record.

This database has several important limitations:

  1. Sixty percent of the tapes were recorded on reel-to-reel recorders or other recorders that lacked a calibrated timing signal. Such recordings are likely to contain frequency-domain artifacts, the result of variations in the speed of tape recording or playback (wow and flutter), which, if unrecognized, can confound analysis of heart rate variability. The header (.hea) file for each record indicates if a calibrated timing signal was recorded.
  2. The purpose of the analysis was to identify N-N intervals accurately. Manually inserted beats were always labeled as Q (unclassified). Great effort was made to identify beats with mismeasured onset points; these beats were also labeled as Q.
  3. Attention was given primarily to N-N intervals of durations that fall between limits indicated in the .hea file for each record. Little effort was made to correct N-N intervals with unusually short or long durations; beats following such intervals were labeled as Q. Therefore, although beats identified as V or A are almost certainly premature ventricular or supraventricular beats, respectively, their incidences are probably underestimated.
  4. The original ECG signals are not available. Accordingly, it is not possible to verify the recorded intervals, to identify or correct errors in beat labels or location, or to study features of the ECG other than inter-beat intervals.
  5. The RR interval time series contain small numbers of anomalously long intervals that typically result from errors in beat detection during the standard Holter analysis. For example, if a single beat is missed, an interval that is about twice as long as the correct intervals appears in the RR interval time series. These long intervals appear as noise in the RR interval time series. It is possible that rare sinus pauses may also account for a few of these long intervals, but this cannot be verified since the original ECG signals are unavailable.

This significant collection of time series, comprising in aggregate about 150 million RR intervals, has been contributed to PhysioNet by Dr. Stein, with the consent of the NHLBI. Limited access to additional data from the original CAST and CAST II studies is available to qualified investigators on application to the NHLBI; see the CAST page on the NHLBI’s web site.

Description of the database

The database is divided into three groups of records (e, f, and m), corresponding to the three medications that the subjects received. Within each group, the data from a single subject are represented by a pair of records (baseline and on-therapy; note that the database includes a few subjects that are represented by only one record each).

Record names are of the form dnnnx. The first character, d, indicates which of the three medications under study was to be given to the subject:

The medication code is followed by a three-digit number, nnn. These first four characters of the record name, dnnn, form the subject identifier. The final character of the record name indicates if the subject was receiving the study medication:

Thus, for example, record f004b is from a subject receiving Flecainide, and the corresponding baseline data for the same subject is in record f004a.

The .atr files are annotation files that contain a label for each QRS complex, indicating its type and time of occurrence. The RR interval time series can be obtained by reading these files using software such as ann2rr; see the RR Intervals, Heart Rate, and HRV Howto for further information.

The .hea files contain additional information about the recordings that may be of interest, including the subject’s age range (in 5-year quantiles) and gender, a note about the type of recording (indicating if it is flutter-compensated), the time of day at which the recording began, and a recording date. Note that for purposes of deidentification, the true recording date is not provided. For each subject, a randomly chosen offset has been applied to the baseline and on-therapy record dates (the same offset is applied to both dates, so that the interval between the recordings is correct). The fictitious recording dates given are close enough to the true dates that it should still be possible to study, for example, seasonal variations of variables of interest in this database.

A list in plain text form of the names of the 1543 records in this database can be found in the RECORDS file.

Links to datasets

Selected references

Many additional publications that discuss the CAST and CAST II studies are listed here.

Icon  Name                    Last modified      Size  Description
[DIR] Parent Directory - [   ] ANNOTATORS 02-Jul-2004 12:22 31 list of annotators [   ] DOI 21-Sep-2015 13:00 19 [   ] MD5SUMS 12-Jul-2005 22:02 413 [   ] RECORDS 02-Jul-2004 12:22 12K list of record names [   ] SHA1SUMS 12-Jul-2005 22:02 527 [   ] SHA256SUMS 18-Sep-2007 12:40 642 [DIR] e/ 08-Oct-2015 09:11 - [DIR] f/ 08-Oct-2015 09:11 - [DIR] m/ 08-Oct-2015 09:10 -
Apache/2.2.22 (Debian) mod_auth_tkt/2.1.0 mod_auth_pgsql/2.0.3 mod_ssl/2.2.22 OpenSSL/1.0.1t Server at Port 80