The Prognostic Importance of Abnormal Heart Rate Recovery and Chronotropic Response Among Exercise Treadmill Test Patients

Thomas M. Maddox, MD, MSc, FACC; Colleen Ross, MS; P. Michael Ho, MD, PhD, FACC; Frederick A. Masoudi, MD, MSPH, FACC; David Magid, MD, MPH; Stacie L. Daugherty, MD, MSPH; Pam Peterson, MD, MSPH, FACC; John S. Rumsfeld, MD, PhD, FACC


Am Heart J. 2008;156(4):736-744. 

In This Article


Study Population and Data Collection

We examined a consecutive, prospective cohort of patients referred for ETT between July 2001 and June 2004. All patients were enrolled in Kaiser Permanente of Colorado (KPCO; Denver, CO)—an integrated, nonprofit comanaged care organization that provides medical services to >430,000 members in the Denver, Colorado, metropolitan area.

During the study period, 9,569 patients underwent ETT. Of these, 34 (0.004%) patients had missing HRR or CR data and were excluded from analysis. In addition, only 16 (0.002%) patients had high-risk DTS (score <-10). Given this small number of high-risk patients and the likelihood that their clinical management would be unchanged by additional ETT information, we excluded these patients from our analysis and only evaluated those 9,519 patients with low (score ≥5) or intermediate-risk (score between 4 and - 10) DTS.

Before exercise testing, a structured history and medical record review were performed to document symptoms, medical history, medication use, cardiac risk factors, and prior cardiac events and procedures. Additional comorbidity data (eg, cerebrovascular and peripheral vascular disease) were obtained from the KPCO administrative databases. Symptom-limited ETT was performed according to standardized protocols, with the Bruce protocol used in 85% of tests. After achievement of peak exercise, all patients underwent a 1-minute "cool-down" period by walking on the treadmill at 1.0 mph. During each exercise stage and recovery stage, symptoms (eg, chest pain, shortness of breath, fatigue, dyspnea, and dizziness), blood pressure, heart rate, cardiac rhythm, and metabolic equivalents (METs) were recorded. The reasons for termination of exercise, including dyspnea, fatigue, chest pain, ischemic ST changes, marked elevation in blood pressure, or ventricular ectopy, were recorded. Achieving target heart rate alone was not used as a justification for terminating exercise. All stress tests were proctored by a clinician. Electrocardiograms were interpreted by computerized analysis, overread by the proctoring clinician, and recorded on data collection sheets at the time of testing. Abnormal ST-segment changes were defined by ≥1 mm downsloping or flat depressions at least 80 milliseconds after the J point. The DTS was calculated using the total exercise time or energy expenditure (minutes or METs, depending on exercise protocol) - (5 × the maximal ST-segment depression [in mm]) - (4 × the angina severity score [0 = no angina, 1 = nonlimiting angina, 2 = limiting angina]).[12] All clinical and exercise data were entered contemporaneously into an electronic database. For patients undergoing multiple treadmill tests during this period, only the first treadmill test was considered in the analyses.

Predictor Variables

Abnormal HRR was defined as a decrease of ≤12 beats/min from peak exercise heart rate at 1 minute to recovery.[20] Abnormal CR was defined as achievement of <80% of a patient's heart rate reserve (calculated as [220 - age] - resting heart rate) at peak exercise.[2] Because β-blocker medication can alter CR, and previous studies have indicated that achieving <62% of HR reserve was predictive of adverse events among patients taking β-blockers,[21] we used this cutoff to define abnormal CR among patients receiving β-blockers within 72 hours of the ETT (n = 1,722).

Our independent predictor variables of interest were normal HRR and CR, abnormal HRR only, abnormal CR only, and abnormal HRR and CR combined.

Outcome Variables

Our primary outcome of interest was combined all-cause mortality or hospitalization for nonfatal acute MI. Patients were observed for a median of 3.2 years and were censored once they experienced either event, disenrolled from the Kaiser Permanente system before the period of follow-up or at the completion of the study period. Secondary outcomes included all-cause mortality and hospitalization for nonfatal MI. All-cause mortality was determined from KPCO databases and validated by comparison with death certificates registered with the State of Colorado and the National Death Index. Hospitalizations for MI both inside and outside the KPCO system were identified by searching administrative and clinical records for the principal discharge diagnosis International Classification of Diseases, Ninth Revision (ICD-9) codes of 410.x for acute MI. Follow-up and vital status information was available after the exercise test on 99% of patients through October 31, 2005.

Statistical Analysis

After categorization of patients by HRR and CR, baseline characteristics (demographic factors, ETT indications, clinical history, ETT variables, and DTS) were compared using the χ2 for categorical variables and Wilcoxon rank sum scores for continuous variables. Outcomes by patient category were graphically represented by Kaplan-Meier curves and compared with the log-rank test. Independent associations between abnormal HRR and CR and outcomes were determined using multivariable Cox proportional hazards models. Patients with normal HRR and CR were the referent group in each analysis. Covariates entered into the multivariable models included demographic factors (age, sex), clinical factors (smoking status, history of CAD, cerebral vascular disease, peripheral vascular disease, cancer, chronic obstructive pulmonary disease, obstructive sleep apnea, diabetes mellitus, hypertension, lipid disorders, depression, congestive heart failure, renal failure), and DTS. Because β-blocker use at the time of ETT has the potential to affect HRR and CR, we also included use of any β-blocker within 72 hours of the ETT as a covariate. To assist with clinical interpretation of our findings, we stratified our cohort into low-risk or intermediate-risk DTS and repeated our multivariable analyses to measure the association between the HRR, CR, and outcomes. The Cox proportional hazards assumption was verified for all models by calculating and graphing Schoenfeld residuals by survival time.

To quantify the additional predictive power provided by HRR and CR to DTS, we compared the -2 log likelihood changes in Cox proportional hazard models and c-statistic changes in logistic regression models (using the same covariates as the Cox models) containing DTS without and with the inclusion of HRR and CR. Changes in -2 log likelihood values provide estimates of the specific contributions that HRR and CR parameters provide to the overall prediction power of the Cox proportional hazards models. Significance testing was performed on the -2 log likelihood value changes using a χ2 distribution with 3 degrees of freedom. Changes in c-statistics measure the ability of the overall multivariable model to accurately distinguish between patients who will experience adverse events and those who will not in logistic regression models.[22]

The study was approved by the Kaiser Permanente Colorado Institutional Review Board. All analyses were performed using the SAS statistical package version 9.1 (SAS Institute, Cary, NC).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.