Estimated age based on exercise stress testing performance outperforms chronological age in predicting mortality

Transcription

1 Full research paper Estimated age based on exercise stress testing performance outperforms chronological age in predicting mortality European Journal of Preventive Cardiology 0(00) 1 9! The European Society of Cardiology 2019 Article reuse guidelines: sagepub.com/journals-permissions DOI: / journals.sagepub.com/home/ejpc Serge C Harb, Paul C Cremer, Yuping Wu, Bo Xu, Leslie Cho, Venu Menon and Wael A Jaber Abstract Aims: We sought to estimate patients age based on their stress testing exercise performance (A-BEST), and evaluate whether A-BEST would be a better predictor of mortality when compared to chronological age. Methods: We included 126,356 consecutive patients referred for exercise (electrocardiography, echocardiography or myocardial perfusion imaging) stress testing at our institution from January 1st, 1991 to February 27th, Estimated age was computed based on exercise capacity (number of peak estimated metabolic equivalents of task), chronotropic reserve index and heart rate recovery, taking into account patient s gender and medications that affect heart rate. Uni and multivariable Cox models were used to determine the association of A-BEST with mortality. Improvement in predicting mortality using A-BEST compared to chronological age was evaluated with the use of net reclassification improvement and C statistic. Results: Mean age was years and 59% were men. At follow-up (mean duration was 8.7 years), 9929 (8%) died. After adjustment for clinical comorbidities, higher metabolic equivalents of task (adjusted hazard ratio (HR) for mortality 0.71, 95% confidence interval (CI) , P < 0.001) and higher chronotropic reserve index (adjusted HR for mortality 0.97, 95% CI , P ¼ ) were associated with improved survival, whereas abnormal heart rate recovery (adjusted HR for mortality 1.53, 95% CI , P < 0.001) and higher A-BEST (adjusted HR for mortality 1.05, 95% CI , P < 0.001) were associated with higher mortality. When comparing prediction models using A-BEST versus chronological age, a significant increase in the area under the curve was demonstrated if A-BESTwas used (0.82 vs. 0.79, P < 0.001). The overall net reclassification improvement was 0.30 (P < 0.001). Conclusion: Estimated age based on exercise stress testing performance is a better predictor of mortality when compared to chronological age. Keywords Exercise age, mortality, exercise stress testing Received 6 November 2018; accepted 6 January 2019 Introduction Stress testing is an important prognostic tool in the evaluation and management of patients with known or suspected heart disease. 1,2 While there are various modalities to assess response to stress (mainly electrocardiographic, echocardiographic or radionuclide), testing is typically performed by exercise or pharmacological stress, with the latter usually performed in patients unable to exercise. 3 Exercise stress testing is generally preferred because it provides a number of exercise variables that have been shown to be powerful prognostic predictors of an increased risk of adverse outcome. 4 7 For instance, exercise capacity (or workload achieved) was found to be the strongest predictor of mortality in both normal subjects and those with cardiovascular disease, with each 1 metabolic equivalents of task (MET) increase in exercise capacity Heart and Vascular Institute, Cleveland Clinic, USA Corresponding author: Serge C Harb, Heart and Vascular Institute, Cleveland Clinic, 9500 Euclid Avenue, J1-126, Cleveland, OH 44195, USA. harbs@ccf.org

2 2 European Journal of Preventive Cardiology 0(00) conferring a 12% improvement in survival. 8 Its prognostic importance was also shown to be the same irrespective of age. 9 Abnormal heart rate recovery (AHRR) and chronotropic incompetence have also been extensively shown to provide important prognostic information. In fact, AHRR is associated with mortality, independent of workload and myocardial perfusion defects, 10 treadmill risk score, 11 and even after adjusting for left ventricular function and angiographic severity of coronary disease. 12 Similarly, chronotropic incompetence has been shown to predict death and coronary disease, 16 even after adjusting for myocardial ischemia. 17,18 While a number of exercise risk scores combining multiple exercise variables have been developed and validated as prognostic tools, we sought to estimate patients age based on their exercise stress testing performance (Age Based on Exercise Stress Testing [A- BEST]) and evaluate whether this A-BEST would be a better predictor of all-cause mortality when compared to chronological age (CA). The goal is to present both patients and healthcare providers with a more easy to understand and practical version of the risk estimate. This A-BEST could be used as a surrogate for physiological age, incorporating factors associated with a declining exercise performance on the treadmill. This concept has been proposed in several other studies, calculating vascular age-based carotid intimal medial thickness, 22 arterial age on the basis of coronary calcium score, 23 lung age, 24 fitness age 25 and bone age. 26 We propose an age estimate (A-BEST) based on readily available, prognostically validated, exercise variables in a large population of patients over a quarter of a century study period. To this end, we compared the performance of this A-BEST to CA in estimating all-cause mortality. Methods Cohort All consecutive patients referred for stress testing to our institution from January 1st, 1991 to February 27th, 2015 were initially considered. In case of repeat testing, only the first stress test was selected. The institutional review board approved the study and patient informed consent was waived. Stress testing modalities consisted of exercise electrocardiography (ECG; non-imaging) and exercise or pharmacological stress combined with imaging (stress echocardiography (echo) or stress radionuclide myocardial perfusion imaging (MPI)). We initially excluded all patients referred for pharmacological testing (echo or MPI; (n ¼ 38,828)). Also, patients who initially started with exercise testing, then were converted to pharmacological testing due to inability to reach heart rate, were excluded (n ¼ 467). We finally excluded patients in whom gender information (n ¼ 796) was missing. After all exclusions, a total of 126,356 patients constituted our final cohort (see flowchart Supplementary Figure 1). At the time of stress testing, patient demographics (age, gender, height and weight), vital signs (heart rate and blood pressure), medications (including negative chronotropic agents, beta-blocker (BB) and non-dihydropyridine calcium antagonist (ND-CA), use at time of testing) and comorbidities were prospectively documented. These comorbidities included history of coronary artery disease (CAD), diabetes mellitus (DM), hypertension, end-stage renal disease (ESRD), smoking, and body mass index (BMI). A history of CAD was defined as previous myocardial infarction or prior revascularization (percutaneous coronary intervention or coronary artery bypass grafting). Diabetes and hypertension were defined as self-reported history, or use of glucose-lowering and antihypertensive medications, respectively. We considered as a smoker a patient who was actively smoking or had smoked in the past, and defined ESRD as receiving dialysis. We calculated BMI as weight (kg)/height 2 (m 2 ). 27 Exercise stress testing Patients underwent symptom-limited treadmill testing according to standardized protocols. The specific protocol for each test was chosen by the exercise physiologist supervising the test, based on patient s reported activity, and the test was performed as recommended by the exercise testing guidelines. 3 Peak estimated METs were determined based on treadmill grade and speed at peak exercise. Peak estimated METs were calculated using the St James Take Heart Project formula for women (estimated METs ¼ 14.7 [0.13 age]) 28 and the Veterans Affairs cohort formula for men (estimated METs ¼ 18.7 [0.15 age]). 29 Percentages of estimated METs were then calculated as achieved (METs/estimated METs) 100. For heart rate recovery (HRR) calculation, the heart rate at 1 minute post exercise was subtracted from the peak heart rate. Recovery forms differed between modalities, with patients undergoing exercise ECG with no imaging and exercise MPI having a walking recovery, while those who had exercise echo having a supine recovery. Therefore AHRR was defined as 12 or fewer beats per minute for the former modalities and 18 or fewer beats per minute for the latter. 10,11,30 Finally, the chronotropic reserve index (CRI) was calculated as ([peak heart rate resting heart rate)]/ [age-predicted peak heart rate resting heart rate]). 31

3 Harb et al. 3 Estimated A-BEST In order to estimate age based on exercise stress testing performance, we included in our prediction model METs, AHRR and CRI, exercise parameters that have been extensively shown to be prognostically important, 7 18,32,33 and which are readily available in the stress testing report. As negative chronotropic agent (BB and ND-CA) use does affect heart rate, a component of both HRR and CRI calculation, we also included BB and ND-CA use (at time of stress testing) in our model. Finally, a separate prediction model was performed for each gender to account for gender-related differences in the prognostic value of exercise testing. 34,35 Outcome The primary outcome was all-cause mortality and was determined from the Social Security Death Index (SSDI), 36 when available. It was supplemented by the institutional death index (chart documentation of patient s death), particularly for the time period following November 2011, when restrictions for SSDI access were implemented. The final censoring date was June 10th, Statistical analysis Numeric data are presented as mean SD. Categorical data are presented as n (%). Student s t-test or Wilcoxon rank sum test for continuous variables and 2 test or Fisher s exact test for categorical variables were used to examine the difference between groups. Hazard ratios (HRs) (and their corresponding 95% confidence intervals (CIs)) for the association of the covariates, including METs, AHRR, and CRI, with mortality were estimated using both univariable (unadjusted) and multivariable (adjusted) Cox models. Restricted cubic spine plots were used to determine the associations of exercise capacity (number of METs) and chronotropic reserve (CRI) with mortality, respectively. Kaplan Meier survival curves were used for the association of AHRR with death. A-BEST was estimated using a linear regression model based on METs, AHRR, CRI, BB and ND-CA use. In order to evaluate the prediction model performance using A-BEST versus CA, we performed area under the curve (AUC) comparison and category-free net reclassification improvement (NRI), as described by Pencina et al. 37 The risk of death was estimated from Cox models, adjusting for gender, CAD, DM, statin use, hypertension, smoking, ESRD, and BMI. All analyses were performed using R (Vienna, Austria) and two-sided P values less than 0.05 were considered statistically significant. Results A total of 126,356 patients underwent exercise stress testing. The modalities of stress testing were as follows: 29% exercise ECG without imaging, 50% exercise echo and 21% exercise MPI. The mean age of the cohort was 53.5 years (12.6 years), and 59% were men. At followup (mean duration 8.7 years), 9929 (8%) died. Table 1 presents the baseline characteristics, overall and by death status. Patients who died were significantly chronologically older ( years vs years, P < 0.001) and more likely to have comorbidities, specifically CAD (39.9% vs. 14.7%, P < 0.001), DM (24.6% vs. 10.6%, P < 0.001), hypertension (79.7% vs. 52.1%, P < 0.001) and ESRD (5.1% vs. 1.1%, P < 0.001). They were also more likely to be smokers (64.4% vs. 44.1%, P < 0.001) and less likely to be on statins (9.2% vs. 16.1%, P < 0.001). Their resting heart rate and systolic blood pressure (SBP) at the time of testing were significantly higher ( vs , P < for HR and vs , P < for SBP). Their exercise capacity, estimated by METs, and chronotropic competence, determined by CRI, were significantly worse ( vs , P < for METs, and vs , P < for CRI). The deceased patients were more likely to have AHRR (49.5% vs. 15.6%, P < 0.001). Univariable and multivariable associations with increased mortality All the exercise parameters included in the estimated age (A-BEST) were significantly associated with mortality. In particular, lower exercise capacity, as reflected by decreasing METs, was associated with increased mortality in a quasi-linear fashion (Figure 1(a)). Similarly, lower chronotropic reserve (decreasing CRI) was also associated with increased mortality (Figure 1(b)). Kaplan Meier survival analysis showed significantly higher mortality in patients with AHRR compared to those with normal HRR (Figure 1(c)). All these associations remained significant on multivariable analysis (Table 2), after adjusting for age, gender, and comorbidities. Higher METs and CRI remained significantly protective (HR 0.71, 95% CI , P < for METs and HR 0.97, 95% CI , P ¼ for CRI), while AHRR remained significantly associated with an increased risk of death (HR 1.53, 95% CI ,

4 4 European Journal of Preventive Cardiology 0(00) Table 1. Baseline characteristics (clinical and medications) and exercise data for the overall cohort and by death status. Variable Overall (n ¼ 126,356) Alive (n ¼ 116,427) Dead (n ¼ 9929) P value Clinical Age, mean SD, years <0.001 Men, no. (%) 74,724 (59.1) 67,715 (58.2) 7009 (70.6) <0.001 CAD, no. (%) 21,123 (16.7) 17,162 (14.7) 3961 (39.9) <0.001 DM, no. (%) 14,806 (11.7) 12,364 (10.6) 2442 (24.6) <0.001 Hypertension, no. (%) 68,550 (54.3) 60,641 (52.1) 7909 (79.7) <0.001 Smoker, no. (%) 57,790 (45.7) 51,400 (44.1) 6390 (64.4) <0.001 ESRD, no. (%) 1451 (1.4) 993 (1.1) 458 (5.1) <0.001 BMI, mean SD, Kg/m <0.001 Medications Statins, no. (%) 19,313 (15.6) 18,413 (16.1) 900 (9.2) <0.001 BB, no. (%) 30,568 (24.2) 27,003 (23.2) 3565 (35.9) <0.001 ND-CA, no. (%) 6301 (5.8) 5366 (5.4) 935 (10.1) <0.001 Aspirin, no. (%) 41,919 (33.2) 37,795 (32.5) 4124 (41.5) <0.001 ACE/ARB, no. (%) 30,814 (24.4) 26,576 (22.8) 4238 (42.7) <0.001 Insulin, no. (%) 3855 (3.1) 2932 (2.5) 923 (9.3) <0.001 Exercise data Resting SBP, mean SD, mmhg <0.001 Resting HR, mean SD, beats/min <0.001 Peak SBP, mean SD, mmhg <0.001 Peak HR, mean SD, beats/min <0.001 Peak METs, mean SD <0.001 AHRR, a no. (%) 23,046 (18.2) 18,131 (15.6) 4915 (49.5) <0.001 CRI, b mean SD <0.001 A-BEST, mean SD <0.001 CAD: coronary artery disease; DM: diabetes mellitus; ESRD: end-stage renal disease; BMI: body mass index (calculated as weight in kilograms divided by height in meters squared); BB: beta-blocker; ND-CA: non-dihydropyridine calcium antagonist; SBP: systolic blood pressure; HR: heart rate; METs: estimated metabolic equivalents; ACE/ARB: angiotensin-converting enzyme inhibitor/angiotensin receptor blocker; AHRR: abnormal heart rate recovery; CRI: chronotropic reserve index. a Calculated as peak exercise heart rate minus heart rate at 1 minute after exercise. b Calculated as ([peak heart rate resting heart rate)]/[age-predicted peak heart rate resting heart rate]). P < 0.001). Of note, statin use was protective (HR 0.58, 95% CI , P < 0.001). Higher BMI was also protective (0.77, 95% CI , P < 0.001). Estimated A-BEST Age was estimated, for each gender separately, based on METs, CRI, AHRR, and BB and ND-CA use at time of testing. For men, the estimated age was predicted as follows: A-BEST ¼ METs þ CRI þ AHRR þ BB þ ND-CA. For women, the derived formula was: A-BEST ¼ METs þ CRI þ AHRR þ BB þ ND-CA. When comparing the estimated age to the CA, 55% of men and 57% of women between the ages of 50 and 60 years had their estimated age younger than their CA. Univariable and multivariable associations of A-BEST with mortality On univariable analysis, A-BEST was significantly predictive of all-cause mortality (HR 1.33, 95% CI , P < 0.001) and remained significantly associated even after adjusting for gender, CAD, DM, hypertension, statin use, smoking, ESRD, and BMI (HR 1.05, 95% CI , P < 0.001). Comparison of A-BEST with CA When evaluating model performance using A-BEST versus CA, the overall NRI was 0.30, highly statistically

5 Harb et al. 5 (a) (b) 1.5 Hazard Ratio for Death Hazard Ratio for Death Number of METs Achieved Calculated CRI (c) 100 Event-Free Survival, % Log-rank p < Normal HRR Abnormal HRR Study Year No. at Risk: Normal HRR Abnormal HRR Figure 1. Association of exercise capacity (number of metabolic equivalents of task (METs) (a)) and chronotropic competence (chronotropic reserve index (b)) with all-cause mortality (cubic spline plots). Data points show the hazard ratio of death. Dashed lines are 95% confidence intervals. (c) The Kaplan Meier survival curves for abnormal heart rate recovery (AHRR) compared to normal heart rate recovery (HRR). Data points indicate median. significant (P < 0.001, Table 3), and there remained a significant increase in AUC if A-BEST was used (0.82 vs. 0.79, P < 0.001). Figure 2 compares the AUC for both ages (A-BEST and CA) in the overall population and by gender. Even after exclusion of patients with known CAD (N ¼ 21,123), A-BEST remained significantly predictive of death (univariable HR 1.32, 95% CI , P < 0.001). This association also remained significant after adjusting for gender, DM, hypertension, statin use, smoking, ESRD, and BMI (adjusted HR 1.28, 95% CI , P < 0.001). Similarly, the overall NRI remained statistically significant (NRI ¼ 0.21, P < 0.001) and there remained a significant increase in AUC if A-BEST is used (0.81 vs. 0.77, P < 0.001). Analysis of derivation and validation cohorts In order to validate further the A-BEST, we repeated the analysis splitting our patient population into a derivation cohort (60% of total, n ¼ 75,813 patients) and a validation (40% of total, n ¼ 50,543) cohorts. Supplementary Table 1 compares the characteristics of both cohorts. In the derivation and validation cohorts, A-BEST had a comparable performance. When using A-BEST compared to CA, the overall NRI was 0.5 (P < 0.001) (Supplementary Tables 2 and 3), and there remained a significant increase in AUC (0.82 vs. 0.79, P < 0.001, in the derivation cohort, Supplementary Figure 2; 0.83 vs. 0.79, P < 0.001, in the validation cohort, Supplementary Figure 3). In addition, the model was

6 6 European Journal of Preventive Cardiology 0(00) Table 2. Univariable and multivariable hazard ratios for the association of various variables with mortality. Univariable Multivariable HR with 95% CI P value HR with 95% CI P value Age 2.4 ( ) < ( ) <0.001 Men 1.62 ( ) < ( ) <0.001 CAD 3.29 ( ) < ( ) <0.001 DM 2.65 ( ) < ( ) <0.001 Hypertension 3.54 ( ) < ( ) <0.001 Statin 0.8 ( ) < ( ) <0.001 ESRD 5.11 ( ) < ( ) <0.001 Smoker 2.12 ( ) < ( ) <0.001 BMI 0.9 ( ) < ( ) <0.001 Peak METs 0.31 ( ) < ( ) <0.001 CRI 0.48 ( ) < ( ) AHRR 4.68 ( ) < ( ) <0.001 A-BEST* 1.33 ( ) < * ( ) <0.001 CAD: coronary artery disease; DM: diabetes mellitus; ESRD: end-stage renal disease; BMI: body mass index (calculated as weight in kilograms divided by height in meters squared); METs: estimated metabolic equivalents; CRI: chronotropic reserve index; AHRR: abnormal heart rate recovery. *For A-BEST, the multivariable analysis included gender, CAD, DM, hypertension, statin use, smoking, ESRD, and BMI but did not include age, peak METs, CRI, or AHRR (the individual components of A-BEST). Table 3. Discriminant analysis (comparing estimated age (A-BEST) and chronologic age (CA)). Discrimination analysis % (95% CI) P value AUC Model with A-BEST 0.82 ( ) <0.001 Model with CA 0.79 ( ) NRI 0.3 <0.001 A-BEST: estimated age based on exercise stress testing performance; CA: chronological age; AUC: area under the curve; NRI: net reclassification index. P values compare the prediction model performance using exercise age versus chronological age. Both models were adjusted for gender, body mass index, statins, hypertension, diabetes mellitus, smoking, coronary artery disease and end-stage renal disease. Risk of mortality was estimated from Cox models. well calibrated in the validation cohort (Supplementary Figure 4). Discussion To our knowledge, these findings are derived from the largest cohort of exercise stress tests reported in the literature. Our present study demonstrates that estimated physiological age based on readily available exercise stress testing variables performs better in predicting all-cause mortality compared to CA. Increasing age is among the most robust and consistent risk factors for adverse outcomes 38,39 and A-BEST showed superior discrimination and net reclassification improvement compared to CA. In addition, A-BEST remained significantly associated with mortality, even after adjustment for major cardiovascular risk factors (DM, ESRD, hypertension, smoking, CAD and obesity). The exercise variables included in A-BEST computation were also individually significantly associated with mortality, even after adjusting for age and traditional cardiovascular risk factors. In particular, higher exercise capacity, as reflected by the increasing number of METs, was associated with 29% lower risk of death, irrespective of age, gender, and comorbidities. This is concordant with previously published literature showing the powerful predictive value of exercise capacity in patients with and without cardiovascular disease, 8 irrespective of age, 9 and gender. 21 In addition, AHRR was associated with a 50% higher risk of death. This has been postulated to be related to abnormal vagal tone 40 and has been previously validated and advocated to be an integral part of the stress testing report. 14 Similarly, chronotropic incompetence, as reflected by lower CRI, was also shown here to be associated with increased mortality, as previously demonstrated. 16 The main advantage of using A-BEST, an estimated age based on exercise performance, is to provide a reliable and convenient transformation of exercise variables into a scale more easily appreciated by both patients and healthcare providers. This measure provides the patient with a more easily understandable risk estimate. For example, it may be more relevant to tell a 45-year-old patient who achieves an A-BEST

7 Harb et al. 7 (a) Sensitivity Estimated age Chronologic age AUC P Value 0.80 < Specificity (b) 1.0 (c) Sensitivity Sensitivity Estimated age Chronologic age AUC P Value 0.79 < Estimated age Chronologic age AUC P Value 0.81 < Specificity Specificity Figure 2. Area under the curve (AUC) comparison between estimated age (A-BEST) and chronological age in the overall (a) study cohort and in both male (b) and female (c) populations. of 55 that their exercise stress testing performance is more consistent with an age of 55 years. Conversely, a patient with CA of 65 years old and an A-BEST of 48 years has an actuarially better survival than his perceived age cohort. Such an approach was preferred by patients, as shown in a study exploring patients perceptions of cholesterol and cardiovascular risk in which a strategy of translating cholesterol results into riskadjusted age was preferred compared to the absolute results or visual representations of 10-year Framingham risk. 41 The goal is to motivate the patient to adopt behavioral changes that would increase their exercise performance, which would translate into improved survival. The value and positive impact of similar strategies have previously been demonstrated. For instance, telling smokers their lung age based on spirometric assessment of their lung function significantly increased the likelihood of quitting smoking. 42 It remains to be proved that an increase or a decrease in A-BEST in an individual patient will lead to a change in mortality risk. However, given that most of the parameters included in the derivation of A-BEST are dynamic and not static (as opposed to CA), patients can be motivated to improve their A-BEST. Limitations Our study has notable limitations. First, although this study was conducted in a large cohort of patients over a 25-year period, it is a single center study, and its findings need to be confirmed in other populations. Second, the outcome assessed was all-cause mortality instead of cardiac death. However, all-cause mortality is an unbiased endpoint and may therefore be preferred. Third, we only included certain exercise variables in computing exercise age. These variables were selected based on multiple prior studies showing their prognostic utility. Other clinical, electrocardiographic, and imaging (in case of exercise echo or MPI) were not taken into consideration. Nonetheless, the goal of this study was to develop a simple age estimate based specifically on exercise parameters that would be readily and consistently available.

8 8 European Journal of Preventive Cardiology 0(00) Conclusions In a large population of patients over a quarter of a century period, estimated physiological age (A-BEST), based on exercise stress testing parameters (namely exercise capacity, chronotropic reserve and HRR) and taking into account patient s gender and use of medications that affect heart rate, was significantly associated with all-cause mortality, and outperformed CA in predicting all-cause death. We strongly believe this estimated age provides both patients and healthcare providers a convenient and simple transformation of exercise variables into a more easily understandable and refined risk estimate, with the premise that this would lead to positive behavioral changes. Acknowledgments The authors would like to acknowledge Ms Theresa Guy for her help with the stress database. Author contribution SCH, PCC, VM and WAJ contributed to the conception and design of the work. SCH and WAJ contributed to the acquisition of the data. SCH, PCC, YW and WAJ contributed to the analysis of the data. SCH, PCC, BX, LC, VM and WAJ contributed to the interpretation of the data. SCH drafted the manuscript. PCC, BX, LC, VM and WAJ critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work ensuring integrity and accuracy. Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding The author(s) received no financial support for the research, authorship, and/or publication of this article. References 1. Harb SC and Marwick TH. Prognostic value of stress imaging after revascularization: a systematic review of stress echocardiography and stress nuclear imaging. Am Heart J 2014; 167: Goraya TY, Jacobsen SJ, Pellikka PA, et al. Prognostic value of treadmill exercise testing in elderly persons. Ann Intern Med 2000; 132: Gibbons RJ, Balady GJ, Bricker JT, et al. ACC/AHA 2002 Guideline update for exercise testing: summary article. A report of the American College of Cardiology/ American Heart Association Task Force on Practice Guidelines (Committee to Update the 1997 Exercise Testing Guidelines). J Am Coll Cardiol 2002; 40: McNeer JF, Margolis JR, Lee KL, et al. The role of the exercise test in the evaluation of patients for ischemic heart disease. Circulation 1978; 57: Weiner DA, Ryan TJ, McCabe CH, et al. Prognostic importance of a clinical profile and exercise test in medically treated patients with coronary artery disease. JAm Coll Cardiol 1984; 3: Peduzzi P, Hultgren H, Thomsen J, et al. Veterans Administration Cooperative Study of medical versus surgical treatment for stable angina progress report. Section 8. Prognostic value of baseline exercise tests. Prog Cardiovasc Dis 1986; 28: Zafrir B. The prognostic value of exercise testing: exercise capacity, hemodynamic response, and cardio-metabolic risk factors. Eur J Prev Cardiol 2017; 24: Myers J, Prakash M, Froelicher V, et al. Exercise capacity and mortality among men referred for exercise testing. N Engl J Med 2002; 346: Spin JM, Prakash M, Froelicher VF, et al. The prognostic value of exercise testing in elderly men. Am J Med 2002; 112: Cole CR, Blackstone EH, Pashkow FJ, et al. Heart-rate recovery immediately after exercise as a predictor of mortality. N Engl J Med 1999; 341: Nishime EO, Cole CR, Blackstone EH, et al. Heart rate recovery and treadmill exercise score as predictors of mortality in patients referred for exercise ECG. JAMA 2000; 284: Vivekananthan DP, Blackstone EH, Pothier CE, et al. Heart rate recovery after exercise is a predictor of mortality, independent of the angiographic severity of coronary disease. J Am Coll Cardiol 2003; 42: Watanabe J, Thamilarasan M, Blackstone EH, et al. Heart rate recovery immediately after treadmill exercise and left ventricular systolic dysfunction as predictors of mortality the case of stress echocardiography. Circulation 2001; 104: Shetler K, Marcus R, Froelicher VF, et al. Heart rate recovery: validation and methodologic issues. J Am Coll Cardiol 2001; 38: Cole CR, Foody JM, Blackstone EH, et al. Heart rate recovery after submaximal exercise testing as a predictor of mortality in a cardiovascularly healthy cohort. Ann Intern Med 2000; 132: Lauer MS, Okin PM, Larson MG, et al. Impaired heart rate response to graded exercise. Prognostic implications of chronotropic incompetence in the Framingham Heart Study. Circulation 1996; 93: Lauer MS, Mehta R, Pashkow FJ, et al. Association of chronotropic incompetence with echocardiographic ischemia and prognosis. J Am Coll Cardiol 1998; 32: Lauer MS, Francis GS, Okin PM, et al. Impaired chronotropic response to exercise stress testing as a predictor of mortality. JAMA 1999; 281: Mark DB, Hlatky MA, Harrell Jr FE, et al. Exercise treadmill score for predicting prognosis in coronary artery disease. Ann Intern Med 1987; 106: Lauer MS, Pothier CE, Magid DJ, et al. An externally validated model for predicting long-term survival after

9 Harb et al. 9 exercise treadmill testing in patients with suspected coronary artery disease and a normal electrocardiogram. Ann Intern Med 2007; 147: Cremer PC, Wu Y, Ahmed HM, et al. Use of sex-specific clinical and exercise risk scores to identify patients at increased risk for all-cause mortality. JAMA Cardiol 2017; 2: Stein JH, Fraizer MC, Aeschlimann SE, et al. Vascular age: integrating carotid intima-media thickness measurements with global coronary risk assessment. Clin Cardiol 2004; 27: McClelland RL, Nasir K, Budoff M, et al. Arterial age as a function of coronary artery calcium (from the Multi- Ethnic Study of Atherosclerosis [MESA]). Am J Cardiol 2009; 103: Morris JF and Temple W. Spirometric lung age estimation for motivating smoking cessation. Prev Med 1985; 14: Nes BM, Janszky I, Vatten LJ, et al. Estimating VO 2 peak from a nonexercise prediction model: the HUNT Study, Norway. Med Sci Sports Exerc 2011; 43: Greulich WW and Pyle SI. Radiographic atlas of skeletal development of the hand and wrist, 2nd ed. Stanford, CA: Stanford University Press, Lloyd-Jones DM, Hong Y, Labarthe D, et al. Defining and setting national goals for cardiovascular health promotion and disease reduction: the American Heart Association s strategic Impact Goal through 2020 and beyond. Circulation 2010; 121: Gulati M, Pandey DK, Arnsdorf MF, et al. Exercise capacity and the risk of death in women: the St James Women Take Heart Project. Circulation 2003; 108: Morris CK, Myers J, Froelicher VF, et al. Nomogram based on metabolic equivalents and age for assessing aerobic exercise capacity in men. J Am Coll Cardiol 1993; 22: Watanabe J, Thamilarasan M, Blackstone EH, et al. Heart rate recovery immediately after treadmill exercise and left ventricular systolic dysfunction as predictors of mortality: the case of stress echocardiography. Circulation 2001; 104: Brubaker PH and Kitzman DW. Chronotropic incompetence: causes, consequences, and management. Circulation 2011; 123: Chrysohoou C, Skoumas J, Georgiopoulos G, et al. Exercise capacity and haemodynamic response among 12,327 individuals with cardio-metabolic risk factors undergoing treadmill exercise. Eur J Prev Cardiol 2017; 24: Mandini S, Grazzi G, Mazzoni G, et al. A moderate 1-km treadmill walk predicts mortality in men with mid-range left ventricular dysfunction. Eur J Prev Cardiol 2017; 24: Daugherty SL, Magid DJ, Kikla JR, et al. Gender differences in the prognostic value of exercise treadmill test characteristics. Am Heart J 2011; 161: Mosca L, Barrett-Connor E and Wenger NK. Sex/gender differences in cardiovascular disease prevention what a difference a decade makes. Circulation 2011; 124: Newman TB and Brown AN. Use of commercial record linkage software and vital statistics to identify patient deaths. J Am Med Inform Assoc 1997; 4: Pencina MJ, D Agostino Sr RB, D Agostino Jr RB, et al. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008; 27: ; discussion Savji N, Rockman CB, Skolnick AH, et al. Association between advanced age and vascular disease in different arterial territories a population database of over 3.6 million subjects. J Am Coll Cardiol 2013; 61: Kannel WB and Vasan RS. Is age really a non-modifiable cardiovascular risk factor? Am J Cardiol 2009; 104: Schwartz PJ, La Rovere MT and Vanoli E. Autonomic nervous system and sudden cardiac death. Experimental basis and clinical observations for post-myocardial infarction risk stratification. Circulation 1992; 85(1 Suppl.): I77 I Goldman RE, Parker DR, Eaton CB, et al. Patients perceptions of cholesterol, cardiovascular disease risk, and risk communication strategies. Ann Fam Med 2006; 4: Parkes G, Greenhalgh T, Griffin M, et al. Effect on smoking quit rate of telling patients their lung age: the Step2quit randomised controlled trial. BMJ 2008; 336: 598B 600B.