Relative Risk Forests for Exercise Heart Rate Recovery as a Predictor of Mortality
Journal article by Hemant Ishwaran, Eugene H. Blackstone, Claire E. Pothier, Michael S. Lauer; Journal of the American Statistical Association, Vol. 99, 2004
Journal Article Excerpt
Relative Risk Forests for Exercise Heart Rate Recovery as a Predictor of Mortality. by Hemant Ishwaran , Eugene H. Blackstone , Claire E. Pothier , Michael S. Lauer 1. INTRODUCTION Exercise stress testing is commonly used to assess patients with known or suspected coronary artery disease (Lauer 2001). During exercise testing, exercise capacity, heart rate changes, and changes on the electrocardiogram are recorded. Although exercise testing has classically been considered as a diagnostic test to identify patients likely to have important coronary artery disease, recent work has focused on its powerful prognostic value (Gibbons et al. 1997, 2002a,b; Lauer 2001; Williams, Fihn, and Gibbons 2001). During exercise, the heart rate rises due to regulatory effects of the autonomic nervous system. Autonomic nervous system function is determined by the balance of activity of its sympathetic and parasympathetic components: Sympathetic function results in increased heart rate and blood pressure; parasympathetic function results in their decrease. Heart rate rise during exercise is largely due to rapid withdrawal of parasympathetic tone (also known as vagal tone) as well as increased sympathetic tone (Hammond and Froelicher 1985; Arai et al. 1989). Failure of heart rate to rise appropriately during exercise, known as chronotropic incompetence (Lauer, Okin, Larson, Evans and Levy 1996; Lauer et al. 1999), is a predictor of all-cause mortality and coronary heart disease events (Lauer et al. 1996; 1999; Lauer 2001). Recently we have focused on fall in heart rate immediately after exercise, or heart rate recovery (Imai et al. 1994), as a potential predictor of morality (Cole, Blackstone, Pashkow, Snader and Lauer 1999; and Cole, Foody, Blackstone, and Lauer 2000; Nishime, Cole, Blackstone, Pashkow, and Lauer 2000; Shetler et al. 2001). Decrease in heart rate during the first minute after exercise is largely a function of reactivation of parasympathetic function (Imai et al. 1994). Depressed parasympathetic function is associated with increased risk of death in a wide spectrum of patients (Schwartz, La Rovere, and Vanoli 1992; La Rovere, Bigger, Marcus, Mortara and Schwartz 1998; La Rovere et al. 2001). Thus we hypothesize that attenuated heart rate recovery immediately after exercise is associated with increased risk of death (Cole et al. 1999). 1.1 Heart Rate Recovery as a Predictor of Death The hypothesis that heart rate recovery is an independent predictor of mortality has been tested and validated in a number of cohorts. Our initial effort involved over 2,000 patients (Cole et al. 1999) for which heart rate recovery was defined as the heart rate at peak exercise minus the heart rate measured 1 minute later. We defined a cut-off value for an abnormal heart rate recovery based on maximization of a log-rank statistic. Based on this, patients with heart rate recoveries less than or equal to 12 beats per minute (bpm) were found to be at substantially increased risk of death compared to those who had a normal heart rate recovery (greater than 12 bpm). Figure 1(a) provides confirmation of these findings in context to the dataset analyzed here. The figure records the score test statistic from a Cox proportional hazards model using different heart rate recovery threshold values. For each threshold value, the Cox model included a 0/1 dichotomized covariate for heart rate recovery (equal to 1 if heart rate recovery was less than or equal to the threshold value, otherwise 0). Included in the Cox model were additive terms for additional covariates identified as being important (see Remark 1 for details). The score test statistic has a flat maximum between 8 and 12 bpm and then drops off rapidly. The maximum seen at 12 bpm agrees with the findings found in Cole et al. (1999). Our later analysis also confirms 12 bpm as an optimal threshold for identifying patients at high risk. 1.2 Patient CohortNot only is heart rate recovery an independent predictor of mortality, it is also predictive of mortality after adjusting for several factors (Snader et al. 1997; Cole et al. 1999, 2000; Diaz, Brunken, Blackstone, Snader and Lauer 2001; Watanabe et al. 2001). Nonetheless, there are a number of important unanswered questions (Shetler et al. 2001). How does heart rate recovery interact with other clinical and physiological variables? What is its prognostic behavior when considered as a continuous variable? Standard methods provide some insight, but not detailed understanding. Consider, for example, Figure 1(b). Relative risk values are shown from a Cox regression analysis of heart rate recovery treated as a factor with 10 levels based on its deciles. The figure shows that relative risk decreases as ... |
End of free preview...
To continue reading this publication, you must have a Questia Subscription.
Try Us Today! Click Here
Questia provides the world's largest online library of scholarly books and journal articles, with integrated footnote and bibliography tools, highlighting, note taking and book marking. With a Questia subscription, you'll have access to the full text of more than 67,000 books and 1.5 million articles.
Already a subscriber? Login:
Sponsored Links
Free Newsletter
Get helpful how-to's, writing tips, search strategies, quizzes & more!
Search the Library
