The optimal time point to measure HRR has been a matter of much debate. Our results show that the prognostic value of HR increase is superior to HRR over a range of recovery times. Furthermore, we found that the correlation coefficient between HRR and HR increase was significant and increased systematically throughout recovery. None of the HRR measurements we evaluated were selected as predictors of either CV or all-cause mortality in our age-adjusted Cox models.
It is difficult to measure HR precisely at 1 or 2 min into recovery. Conceptually, a functional curve fit to the first 5-10 min of the recovery HR data offers the potential both to minimize the error associated with a single estimate and to leverage information that may be contained in the overall shape of the curve. Equation 1 models the HRR process and follows the general form of previous investigators (6). Figure 2 illustrates several important considerations that arise when fitting equation 1 to HRR data or when comparing clinical results from different investigators. Firstly, the start of recovery is often difficult to pinpoint. Some studies use a cool-down walk for the entry into recovery, but others immediately place patients in a supine position. Although the guidelines call for immediate return to the supine position, depending on patient mobility and test protocol, there may be a transition period of 10-20 s during which the patient is still active. This time interval is a large fraction of the typical 1-2 min postexercise interval commonly used in measuring HRR and undoubtedly introduces uncertainty into the score. For these reasons, the parameter t 0, the effective start of recovery, is an important addition to equation 1.
The second consideration is that the stable postexercise recovery rate, HRrec, is a dynamic variable strongly influenced by strenuous exercise and correlated with peak HR. It is commonly observed that HR recovers asymptotically to a value distinctly different from the initial resting HR before the test, often with differences as large as 20-40 bpm. Constraining the HRR curve to return to a pretest resting HR value will introduce systematic bias in the derived value of k. Allowing HRrec to vary in the curve fitting process avoids this bias.
The parameter k defines the normalized rate of decay of the exponential function that has been found in this study to describe robustly the observed HRR curves for the entire population. If the shape of the recovery curve were prognostic, independent of the HR increase scaling factor (HRpeak − HRrec), it should have been reflected in the derived values for k. Unfortunately, this was not the case. Neither of the parameter k or t 0 was found to have prognostic value in our study population, strongly suggesting that the prognostic information contained in equation 1 above is primarily present in the HR increase scaling term (HRpeak − HRrec) and not in the recovery term e - kt. As HRpeak − HRrec is highly correlated with HR increase, measures of HRR appear to contain little additional information beyond that already contained in HR increase.
These findings are consistent with several notable previous studies. In a study by Desai et al. (5) of HRR in normal subjects, subjects with CAD, and transplant patients, HRR was strongly correlated with HR increase, and the correlation increased with increasing time as HR returned to a resting value. In a supporting finding, Racine et al. (16) noted that when HRR was normalized by HR increase, forming the percent HRR decline, there was virtually no difference between healthy and HF patients; peak HR appeared to be the primary determinant of HRR. Similarly, in a 23-yr follow-up study of 5713 asymptomatic subjects by Jouven et al. (9), only HR increase was found to be statistically significant in the Cox analysis that also included resting HR and HRR. Although HRR is a strong univariate predictor of CV death, the correlation with HR increase and the superior prognostic value of the latter have resulted in limited enthusiasm for the prognostic utility of HRR in studies that have considered both parameters.
Many pharmacological blockade experiments have studied the combined effects of the dual branches of the autonomic system during recovery from maximal exercise. Early experiments using atropine and propranolol found that vagal reactivation strongly influenced HRR (8). Reassertion of vagal control in healthy individuals appeared to be most pronounced, relative to those suffering from HF, in the first 30-40 s after the end of exercise. Using atropine blockade and bicycle exercise, Kannankeril et al. (10) found that sympathetic drive is withdrawn in recovery and the maximum parasympathetic effect on HRR occurs in the first minute of recovery.
In this study, the complicating factors of varying HR increase and elevated postexercise stable recovery rates, correlated both with fitness and survival, have been minimized as described above. The normalized HRR slopes, as shown in both Figures 3B and 4, show different slopes at about 1 min. The curves associated with patients who subsequently died of CV causes show declines in the recovery rates, relative to those of survivors, over the time interval 50-70 s. These HRRS differences may reflect the accumulative effects during recovery of weakened vagal response in the CV mortality population.
HRRS and prognosis.
HRRS50-70 was tested in a multivariate Cox survival analysis model and was found to be a significant predictor of CV mortality, independent of all tested variables including DTS and HR increase. It was not found to be a predictor of all-cause mortality, which we believe reflects the fact that this variable is specifically a measure of CV risk; the non-CV mortality population represents about 62% of all-cause mortality. Although the additional discriminative accuracy provided by HRRS50-70 was limited in our hierarchical modeling analysis, Cook (4) has demonstrated that discriminative accuracy may asymptotically approach limits less than one as model parsimony decreases. Thus, this approach to evaluating additional discriminative accuracy provided by a variable may underestimate the prognostic importance of variables that have been shown to significantly augment risk stratification. Furthermore, the conditional probability that the model containing HRRS50-70 was the best fit of all models considered overwhelmingly demonstrated the preference for this parameter over the established HRR at 1 and 2 min for prognostication. Kaplan-Meier survival analyses for combinations of the three parameters HR increase, DTS, and HRRS50-70 demonstrated that the hazard ratio increased from 8.8 (95% CI, 2.5-31) when only HR increase and DTS were abnormal (n = 195, CV deaths = 10), to 22.4 (95% CI, 7.9-63) when all three scores were abnormal (n = 150, CV deaths = 20), a factor of 2.5 increase. This result, along with the improvement in predictive power and discriminative accuracy for CV mortality when HRRS50-70 was included with other exercise and clinical parameters, suggests the new parameter can add value in clinical risk stratification.
Gibbons (7) provided a critique on the many diverse HRR studies, noting the significant potential for HRR to play a valued role in risk stratification similar to the DTS but also noting that inconsistencies in the ways HRR is measured, and the confounding problems of diverse cohort selection and censoring, have limited clinical adoption. This study has found that HRR is a disguised measure of HR increase, already known to be prognostic for CV mortality (11). Nonetheless, in a closer examination of the shape of the recovery curve, through quantification of HRR slope, we found a significant difference in HRRS50-70 between survivors and subjects who died of CV causes that is also independent of both peak HR and HR increase.
The possibility exists that HRRS50-70 is strongly influenced by the difference in exertion levels between patients with CV mortality and healthy survivors. However, peak METs achieved and the Borg scale were found to have low correlations with HRRS50-70 (all < 0.1), arguing against the hypothesis that exertion strongly influences HRRS50-70.
Another consideration is the influence of beta-adrenergic receptor antagonists on HRRS50-70. Although beta-adrenergic receptor antagonists limit the sinus node to a lower peak HR, the use does not appear to influence the normalized HRR slope significantly. Beta-adrenergic receptor antagonists use was included as one of the parameters in the Cox models and was not found to be a significant predictor of CV mortality (P ∼ 0.1). This is consistent with previous studies that have examined the impact of beta-adrenergic receptor antagonists on the prognostic value of patients exhibiting abnormal HRR (16,17).
Our study has several important limitations. The study cohort is overwhelmingly male with an average age of 57 ± 12 yr, limiting the generalization of the results, particularly to women and younger populations. The study is retrospective, and future prospective studies of more diverse populations are warranted before these results are applied. The time of maximum HRRS difference is most likely related to the exercise protocol. We suggest that the cause of the HRRS difference may be tied to the accumulative effects of weakened vagal response in the CV population, but blockade experiments are necessary to establish a true cause and effect relationship.
In this cohort, simple measurements of HRR appear to be disguised measures of HR increase. The HRR slope index HRRS50-70 is independent of and complementary to HR increase as well as DTS and may aid in quantifying vagal tone. Patients with abnormal HRRS50-70 and abnormal DTS and/or HR increase are at substantially increased risk for CV mortality.
Conflict of interest disclosures: David Hadley, Cardiac Science (V.P. Research, Salary & Equity). Cardiac Science products include the diagnostic Quinton and Burdick product lines used in exercise stress testing. Hadley has recently retired from Cardiac Science but continues to hold company equity and provides occasional consulting services.
Salary support for Dr Hadley was provided by Cardiac Science, manufacturer of the Burdick exercise stress testing system used in this study for data acquisition. The methods discussed in this article are not a component of any Cardiac Science product, and this study does not constitute an endorsement of any product or service.
1. Burnham KP, Anderson DR. Model Selection and Inference: A Practical Information-Theoretic Approach
, New York, New York, USA: Springer-Verlag, 1998, p. 353.
2. Cole CR, Blackstone EH, Pashkow FJ, Snader CE, Lauer MS. Heart-rate recovery immediately after exercise
as a predictor of mortality
. N Engl J Med
3. Cole CR, Foody JM, Blackstone EH, Lauer MS. Heart rate recovery after submaximal exercise
testing as a predictor of mortality
in cardiovascularly healthy cohort. Ann Intern Med
4. Cook NR. Use and misuse of the receiver operating characteristic curve in risk prediction. Circulation
5. Desai M, De la Pena-Almaguer E, Mannting F. Abnormal heart rate recovery after exercise
as a reflection of an abnormal chronotropic response. Am J Cardiol
6. Franklin JN. Well-posed stochastic extension of ill-posed linear problems. J Math Anal Appl
7. Gibbons R. Commentary: abnormal heart-rate recovery after exercise
. The Lancet
8. Imai, K, Sato H, Hori M, et al. Vagally mediated heart rate recovery after exercise
is accelerated in athletes but blunted in patients with chronic heart failure. J Am Coll Cardiol
9. Jouven X, Empana J-P, Schwartz PJ, Desnos M, Courbon D, Ducimetiere P. Heart-Rate profile during exercise
as a predictor of sudden death. N Engl J Med
10. Kannankeril PJ, Le FK, Kadish AH, Goldberger JJ. Parasympathetic effects on heart rate recovery after exercise
. J Investig Med
11. Lauer MS, Okin PM, Larson MG, Evans JC, D Levy. Impaired heart rate response to graded exercise
: prognostic implications of chronotropic incompetence in the Framingham Heart Study. Circulation
12. Mark DB, Shaw L, Harrell FE, et al. Prognostic value of a treadmill exercise
score in outpatients with suspected coronary artery disease. N Engl J Med
13. Myers J, Bader D, Madhaven R, Froelicher V. Validation of a specific activity questionnaire to estimate exercise
tolerance in patients referred for exercise
testing. Am Heart J
14. Nishime EO, Cole CR, Blackstone EH, Pashkow FJ. Heart rate recovery and treadmill exercise
score as predictors of mortality
in patients referred for exercise
15. Nissinen, SI, Makikallio TH, Seppanen T, et al. Heart rate recovery after exercise
as a predictor of mortality
among survivors of acute myocardial infarction. Am J Cardiol
16. Racine, NR, Blanchet M, Ducharme A, et al. Decreased heart rate recovery after exercise
in patients with congestive heart failure: effect of β-blocker therapy. J Cardiac Failure
17. Shetler K, Marcus R, Froelicher VF, Vora S, et al. Heart rate recovery: validation and methodologic issues. J Am Coll Cardiol
18. Wiggins RA. The general linear inverse problem: implications of surface waves and free oscillations for earth structure. Rev Geophys and Space Phys
Keywords:©2008The American College of Sports Medicine
AUTONOMIC NERVOUS SYSTEM; EXERCISE; MORTALITY; PROGNOSIS