Defining a set of patient attributes that predict exercise performance outcome following an exercise training program in a population of coronary artery disease patients
Files
TR Number
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The purpose of this study was to evaluate the utility of baseline clinical and graded exercise test (GXT) variables in predicting exercise training outcome in cardiac rehabilitation patients. Data were extracted from the records of 60 cardiac patients who had participated in a community-based exercise program for 5-9 months. Two separate markers of exercise tolerance were used to evaluate training effect: 1) rate-pressure product at a submaximal reference workload of 5 METs (RPPSMETs), and 2) estimated peak METs (pkMETs). These two markers of exercise tolerance were found to have a low correlation (r = -0.29).
Patients were classified into one of three possible outcome categories for each marker of exercise tolerance, i.e., improvement, no change, or decline. Thresholds for classifying patients into improvement or decline groups were ± 10% for the RPPSMETs marker and ± 1 MET for the pkKMETs marker. Outcome classifications using the RPPSMETs marker were as follows: improvement, 37 (62%); no change, 13 (22%); decline, 10 (17%). Use of the pkKMETs marker to classify patients into outcome groups yielded: improvement, 45 (75%); no change, 15 (25%); with no patients classified in the decline group. Multiple logistic regression was used to identify patient attributes predictive of improvement and decline for each exercise tolerance marker.
Baseline variables were found to yield a model highly predictive of improvement in RPPSMETs (correct classification rate = 87%; sensitivity = 92%; specificity = 78%). The best single predictor of improvement outcome was high baseline RPPS5MET values. A model could not be generated to successfully predict decline in exercise tolerance. Baseline variables selected for prediction of improvement outcome, as defined by pkMETs marker, yielded a model with limited utility due to low specificity (correct classification rate = 87%; sensitivity = 96%; specificity = 60%).