Heart rate reserve predicts cardiovascular death among physically unfit but otherwise healthy middle-aged men: a 35-year follow-up study

Original scientific paper Heart rate reserve predicts cardiovascular death among physically unfit but otherwise healthy middle-aged men: a 35-year follow-up study EU RO PEAN SOCIETY OF CARDIOLOGY European Journal of Preventive Cardiology 2016, Vol. 23(1) 59 66! The European Society of Cardiology 2014 Reprints and permissions: sagepub.co.uk/journalspermissions.nav DOI: 10.1177/2047487314553202 ejpc.sagepub.com Kristian Engeseth 1,2, Christian Hodnesdal 1,2, Irene Grundvold 1,2,3, Knut Liestøl 4, Knut Gjesdal 1,2, Gunnar Erikssen 2, Sverre E Kjeldsen 1,2, Jan E Erikssen 2, Johan Bodegard 1 and Per Torger Skretteberg 1,2 Abstract Background: Heart rate reserve (HRR) has been reported to be inversely associated with cardiovascular (CV) disease and death. The impact of physical fitness (PF) on this relationship has not, however, been described in detail. We investigated how different levels of PF influenced the association between HRR and CV death during a 35-year follow-up. Methods and results: HRR and PF were measured in 2014 apparently healthy, middle-aged men during a symptomlimited bicycle exercise test in 1972 75. The men were divided into tertiles (T1-T3) by age-adjusted HRR. Morbidity and mortality data were registered from hospital charts through 2007 and the Norwegian Cause of Death Registry. Adjusted Cox proportional hazard regression models were used to calculate risks. Incidence of CV death was 528 (26%) during median 30 years of follow-up. Men with the lowest HRR had 41% (HR 1.41 [1.14 1.75]) increased risk of CV death compared with the men with the highest. We found a significant interaction between age-adjusted PF and HRR. After stratifying the men by PF, results were statistically significant only among men with the lowest PF, where the men with lowest HRR had a 70% (HR 1.70 [1.12 2.67]) increased risk of CV death compared with the men with the highest. Conclusions: Low HRR was independently associated with increased risk of CV death in apparently healthy, middleaged men. The predictive impact of HRR on CV death risk was, however, confined to unfit men. Keywords Heart rate reserve, physical fitness, CVD risk prediction, cardiovascular death, coronary heart disease Received 11 June 2014; accepted 8 September 2014 Introduction The increment in heart rate (HR) from rest to peak during maximal exercise, termed heart rate reserve (HRR), predicts cardiovascular (CV) disease and death; hence, measurement of HRR could improve CV risk stratification. 1 We have previously shown that physical fitness (PF) is a significant predictor of CV disease and death, and that HRR and PF are correlated. 2,3 HRR and PF may, therefore, convey similar physiological information. The predictive abilities of HRR and PF on CV disease and death may not be independent of each other. Whether there is an interaction between HRR and PF in this context has, to our knowledge, not been studied before. The main aim of the present work was to investigate the possible association between HRR and risk of CV death among 2014 apparently healthy middle-aged men during long-term (35 years) follow-up. Secondly, we 1 Department of Cardiology, Oslo University Hospital, Ullevaal, Norway 2 Faculty of Medicine, University of Oslo, Norway 3 Centre for Clinical Heart Research, Oslo University Hospital, Norway 4 Department of Informatics, University of Oslo, Norway Corresponding author: Kristian Engeseth, Department of Cardiology, Oslo University Hospital, Ullevaal, PB 4956 Nydalen, 0424 Oslo, Norway. Email: engekr@gmail.com

60 European Journal of Preventive Cardiology 23(1) tested to see if there was an interaction between HRR and PF in CV risk estimation; we also investigated possible associations between HRR and future CV death within subgroups of men according to their PF level. Finally, we tested possible associations between HRR, PF, and coronary heart disease (CHD). Methods Study population The Oslo Ischemia Study consists of 2014 apparently healthy men aged 40 59 years who were recruited from five governmental institutions in Oslo during the years 1972 75. 4 All men gave their informed consent before inclusion. None were treated with heart rate-lowering drugs. Further details about selection procedures and exclusion criteria have been presented elsewhere. 2,4,5 Examinations All participants underwent standardized clinical examinations, blood tests, chest x-ray, resting ECG, and a symptom-limited bicycle exercise ECG test at inclusion in 1972 75. Detailed family history of CHD was registered in questionnaires. Resting HR was measured after a standardized period of supine rest. All participants performed a standardized bicycle exercise ECG test and were examined by the same physician (J.E.). The initial workload was 6 minutes at 100 W, increased by 50 W every 6 minutes. The exercise tests were continued until a heart rate of at least 90% of the maximal predicted heart rate (MPHR) was reached unless specific symptoms or signs necessitated a premature termination. If an individual seemed physically fit despite his reaching 90% of MPHR þ10 bpm at the end of one load, he was encouraged to continue as long as possible on the next load, that is maximally for an additional 6 minutes on a higher load. 6 Exercise testing was repeated within 2 weeks in 130 of the participants and showed high reproducibility for heart rates and working capacity between the two tests, within 5% in 90% of the men, and within 10% in all of them. HR was measured every second minute throughout the test. Peak exercise HR was recorded just before termination of the test and age adjusted according to the age-dependent peak exercise HR decline found in our material. HRR was registered as the increment from resting HR to age-adjusted peak exercise HR, and age-adjusted HRR was used in all further statistical calculations. PF was defined as the total bicycle exercise work (Joules), calculated as the sum of work at all workloads divided by body weight (kg). Measurements of PF were age adjusted according to the age-dependent decline in PF found in our material, and age-adjusted PF was used in all further statistical calculations. Further details about HR and PF measurements have also been presented previously. 3 Follow-up Morbidity and mortality data were consecutively obtained from three clinical surveys in 1979 82 (Survey 2), 1989 90 (Survey 3), and 1994 95 (Survey 4), one questionnaire survey in 1987, and two nationwide searches of patient records from all Norwegian hospitals in 1995 96 and in 2005 08 with permission from relevant authorities. Mortality data was obtained from the Norwegian Cause of Death Registry. All morbidity and mortality data are complete up to 1 January 2007, and none of the participants were lost to follow-up. End points The main end point in the present study is CV death, consisting of fatal myocardial infarction, sudden cardiac death, stroke (cerebral infarction or hemorrhage), pulmonary embolism, or aortic disease. The secondary endpoint is CHD, consisting of angina pectoris, nonfatal myocardial infarction, or death from CHD. Angina pectoris was defined as typical angina symptoms and documented use of nitroglycerin. Most men with angina also had objective signs of exercise-induced ischemia during follow-up examinations. Acute myocardial infarctions and strokes, diagnosed in conjunction with major surgery, were not included. The interval between CV events had to exceed 6 months to count as separate events. Only the first event of a specific disease was counted. Statistical methods All statistical calculations were performed by SAS JMP 9 software. Kendall rank test was used to assess correlation (trend) between HRR-tertiles and clinical characteristics of participants. The risks of end points in HRR-tertiles were estimated by Kaplan-Meier plots and tested with log-rank tests. Cox proportionalhazard modeling was used when calculating hazard ratios. Significant variables in univariable analyses (p < 0.05) were entered into multivariable analysis. A succession of adjusted survival models were examined: univariable for baseline age; multivariable model 1, for the conventional CV risk factors (smoking, total cholesterol, and resting systolic blood pressure [SBP]); and multivariable model 2, for PF in addition to model 1. Statistical interactions between HRR and PF were tested by adding the interaction term of HRR and PF in the regression models. Finally, the men were

Engeseth et al. 61 stratified into three groups by PF before re-analysis of hazard ratios (univariable and multivariable model 1) between HRR-groups corresponding to the original HRR-tertiles. The same adjustment models, interaction analyses, and stratifications were used for all end points to obtain comparable results. Models were also tested after exclusion of men who exhibited ischemia during the exercise test by developing ST-depressions of more than 2 mm or chest pain. 7,8 Results Characteristics of participants Average age at baseline was 49.7 years, and 882 men (43.8%) were current smokers. Average resting HR was 61.4 (8.75) beats per minute (bpm) and average HRR was 101.4 (6.22) bpm. High HRR was correlated with high levels of age-adjusted physical fitness and inversely correlated with prevalence of smoking, SBP, and SBP at 100 W (SBP100W). There were no significant differences in age or serum cholesterol between the HRRtertiles. Levels of hemoglobin were significantly lower and BMI higher in increasing HRR-tertiles, but none of the groups stood out as anemic or overweight. Time to event was inversely correlated with HRR among the 528 men that died from CVD and the 754 men who developed CHD (Table 1). Crude incidence of CV death among the 2014 men was 528 (26%) and CHD 765 (38%) during follow-up of up to 35 years (median, 30 years) comprising 60,420 person years of follow-up until death or end of observation period after baseline (Table 3 (later) and online Supplementary Table S3). Predictors of CV death and statistical interaction HRR was a significant predictor of CV death and CHD, both in univariable and multivariable analysis (Table 2 and online Supplementary Table S1.1). PF was a significant predictor of CV death and CHD in univariable and multivariable analysis without HRR included (data not shown), but HRR cancelled out PF when both were entered together into the multivariate model. We found significant statistical interactions between HRR and PF in prediction of CV death (p ¼ 0.0120, 2 ¼ 6.3) and CHD (p ¼ 0.0054, 2 ¼ 7.8) (online Supplementary Table S2). After stratifying the men by PF, HRR was a significant predictor of CV death and CHD in men with low PF, but not among men with intermediate or high PF (Table 3 and online Supplementary Table S1.2). Survival and relative risks Survival free from CV death and CHD decreased with decreasing HRR among the 2014 men and in the stratum of 677 men with lowest PF, as shown in Figure 1 and online Supplementary Figure S1. There were no differences in survival free from CV death or CHD between the HRR-groups in the strata of men with intermediate or high PF (data not shown). Table 1. Characteristics of participants in tertiles (T1-T3) of age-adjusted heart rate reserve (n ¼ 2014). Clinical characteristics T1 T2 T3 p Range of age-adjusted heart rate reserve, bpm 37 to 93 94 to 107 108 to 153 Number of participants 679 672 663 Age, years (SD) 49.8 (5.5) 49.8 (5.3) 49.9 (5.7) 0.7589 Age adjusted heart rate reserve, bpm (SD) 86 (9) 102 (3) 116 (6) <0.0001 Age adjusted maximal heart rate, bpm (SD) 153 (13) 163 (8) 172 (8) <0.0001 Resting heart rate, bpm (SD) 67 (11) 61 (8) 56 (7) <0.0001 Age adjusted physical fitness, kj/kg (SD) 1.2 (0.5) 1.4 (0.5) 1.7 (0.6) <0.0001 Systolic blood pressure, mmhg (SD) 134 (20) 129 (17) 127 (16) <0.0001 Systolic blood pressure at 100 W, mmhg (SD) 185 (26) 181 (24) 179 (23) <0.0001 Total serum cholesterol, mmol/l (SD) 6.7 (1.2) 6.6 (1.1) 6.6 (1.2) 0.1491 Hemoglobin, g/dl (SD) 15.4 (1.0) 15.2 (1.1) 15.2 (1.1) <0.0001 BMI, kg/m 2 (SD) 24.9 (3.1) 24.5 (2.7) 24.3 (2.4) 0.0002 Observation time to CV death, years (SD) a 19.3 (8.0) 21.5 (7.5) 22.4 (7.4) 0.0002 Observation time to CHD, years (SD) b 13.0 (7.9) 14.4 (8.1) 16.0 (8.3) <0.0001 Values are mean (SD) or numbers, n (%). Heart rate reserve, maximal heart rate at exercise minus resting heart rate; physical fitness, total work (kj) divided by body weight; systolic blood pressure at 100 W, systolic blood pressure at 100 W workload during bicycle exercise test; p value for trend calculated by Kendall rank test. a Calculated among the 528 men who died from CVD only. b Calculated among the 754 men who developed CHD only.

62 European Journal of Preventive Cardiology 23(1) Table 2. Possible predictors of CV death ranked by 2 in multivariable analysis (n ¼ 2014). Univariable analysis Multivariable analysis Possible predictors HR 95% CI p 2 HR 95% CI p 2 Age (1SD) 2.1 1.84 2.20 <0.0001 238.9 1.98 1.80 2.18 <0.0001 205.8 Smoker (yes/no) 1.69 1.42 2.01 <0.0001 36.2 1.61 1.36 1.93 <0.0001 28.8 Systolic blood pressure (1SD) 1.45 1.34 1.57 <0.0001 33.1 1.23 1.11 1.31 <0.0001 20.1 Total serum cholesterol (1SD) 1.30 1.20 1.40 <0.0001 24.4 1.18 1.09 1.28 0.0001 14.8 Family history of CHD 1.53 1.26 1.85 <0.0001 17.9 1.41 1.16 1.70 0.0007 11.4 HRR (1SD) 0.77 0.71 0.84 <0.0001 32.6 0.86 0.78 0.95 0.0022 9.4 PF (1SD) a 0.74 0.65 0.82 <0.0001 30.4 0.91 0.81 1.02 0.1162 2.5 Hazard ratios per 1 SD increase in parameter (parameter units as in Table 1). Possible predictors are measured at baseline. Heart rate reserve, maximal heart rate at exercise minus resting heart rate; physical fitness, total work (kj) divided by body weight. CV death, death from myocardial infarction, sudden cardiac death, stroke, pulmonary embolism or aortic disease; multivariable analysis, adjusted for baseline age, smoking status, total cholesterol, family history of CHD, HRR, PF, and systolic blood pressure. a Significant in univariable and multivariable analysis, but not significant when entered into the multivariable model along with age-adjusted heart rate reserve. Table 3. Age-adjusted heart rate reserve as CV death-predictor in men with low, intermediate or high level of age adjusted physical fitness (n ¼ 2014). Fitness level and adjustments HR 95% CI p 2 Low PF, range 0.19 to 1.11 kj/kg, n ¼ 677 Age adjusted 0.77 0.68 0.89 0.0002 13.9 Multivariable 0.82 0.71 0.95 0.0006 11.7 Intermediate PF, range 1.12 to 1.37 kj/kg, n ¼ 672 Age adjusted 0.85 0.72 1.01 0.0694 3.3 Multivariable 0.92 0.78 1.10 0.3427 0.9 High PF, range 1.68 to 5.71 kj/kg, n ¼ 665 Age adjusted 0.96 0.78 1.18 0.6896 0.2 Multivariable 0.98 0.79 1.21 0.8585 0.0 Relative risks of CV death and CHD increased with decreasing HRR-tertiles (Table 4 and online Supplementary Table S3). In the lowest HRRtertile, T1, the risks of CV death and CHD, respectively, were 68% and 42% higher than in the highest HRR-tertile, T3, after adjustment for age and 41% and 25% after further adjustment for systolic blood pressure, smoking status, family history of CHD, and cholesterol. Similar increases in risks of CV death and CHD, respectively, were shown among men with lowest PF. Risks in the low compared with the high HRR-group were 77% and 47% higher, after adjustment for age, and 70% and 39% (borderline significant p ¼ 0.0560) higher after multiple adjustments. However, among the men with intermediate or high PF, there were no trends towards differences in risks of CV death or CHD between the HRR-groups. Sensitivity analyses Coronary ischemia was detected in 91 men during the baseline exercise test. Relative risks of CV death among the HRR-tertiles and in the HRR-groups after stratification by PF remained unchanged after exclusion of these 91 men (Table 4). There was, however, only a borderline significant trend towards an unfavorable risk of CHD in the low HRR-tertile, and no significant trends towards CHD risk differences in the HRRgroups after stratification by PF. All results presented were calculated after individual age adjustment of HRR and PF. When unadjusted HRR and PF were used in the same regression models, the estimated effects were slightly stronger (data not shown). Several other possible predictors of CV death, including fasting blood glucose, hemoglobin, systolic blood pressure at 100 W workload, 1 second forced

Engeseth et al. 63 (a) 1.0 Survival from CV death 0.9 0.8 0.7 0.6 (b) 1.0 Survival from CV death 0.9 0.8 0.7 0.6 expiratory volume and body mass index were introduced into the multivariable analysis, in addition to the model described earlier, to evaluate the potential impact on results. The results were only altered marginally, and none of the additional adjustments changed trends in risk differences between HRR-tertiles significantly. No significant association was found between HRR and cerebrovascular disease, defined as fatal/nonfatal stroke (cerebral infarction or hemorrhage) or transient ischemic attack (data not shown). Discussion T1, ahrr 37-93 bpm, n=679 T2, ahrr 94-107 bpm, n=672 T3, ahrr 108-153 bpm, n=663 0 5 10 15 20 25 30 35 Observation time, years Log-Rank; Chi Square 14.35, P-value 0.0008 T1, ahrr 37-93 bpm, n=369 T2, ahrr 94-107 bpm, n=202 T3, ahrr 108-153 bpm, n=106 0 5 10 15 20 25 30 35 Observation time, years Log-Rank; Chi Square 6.65, P-value 0.0359 Figure 1. Survival free from cardiovascular (CV) death. (a) Survival from CV death by HRR-tertiles among all men, n ¼ 2014. Kaplan-Meier curves showing survival (%) free from CV death in tertiles (T1-T3) of age-adjusted heart rate reserve (HRR) among 2014 apparently healthy men. (b) Survival from CV death by HRR-groups among men with low PF, n ¼ 685. Kaplan-Meier curves showing survival (%) free from CV death in groups (T1-T3) of ahrr among 685 apparently healthy men with low age-adjusted physical fitness (PF). We investigated a possible association between HRR during maximal exercise and long-term (35 years) risk of CV death and the impact of PF on this association among 2014 apparently healthy middle-aged men followed for more than 60,000 person-years. We confirmed that HRR was a significant long-term predictor of CV death after adjustment for classic CV risk factors. The new finding was a significant interaction between HRR and PF on CV death risk assessment, and after stratifying the participating men by low, intermediate, or high PF, the predictive abilities of HRR were confined to the group of men with low PF. Similar findings were detected when separately assessing CHD risk. Potential pathophysiological mechanisms The ability to increase cardiac output during exertion depends on the ability to increase stroke volume, HR, or both. During maximal aerobic exercise in healthy humans, maximal oxygen uptake increases about 4-fold (2.2-fold increase in HR, a 0.3-fold increase in stroke volume, and a 1.5-fold increase in arteriovenous oxygen difference). 1 Hence, individuals with a subnormal ability to increase HR must compensate by increasing stroke volume and oxygen extraction to meet increasing oxygen demand of tissues on exertion. It seems plausible that those with good physical fitness despite low HRR must have high ability to increase arteriovenous oxygen difference and/or stroke volume and are in better CV health than those unable to compensate. On the other hand, individuals who exhibit both low PF and low HRR must either have very low HRR, impaired ability to increase stroke volume and/ or arteriovenous oxygen difference, or inability to increase all of these parameters, which have all been shown to be independent CV disease and death risk factors in a large number of other studies and are common in patients with CHD and heart failure. 1,9 12 Ellestad and Wan found that patients without detectable ischemia but with low HRR had a 4-fold greater incidence of CHD than did those with normal HRR during merely 4 years of follow-up. 13 We found a significant trend towards shorter observation time to CHD among men with low HRR than among men with intermediate or high HRR. However, when excluding 91 men with detected ischemia during exercise testing, there were only borderline significant CHD risk differences among the HRR-tertiles and no risk differences among the corresponding HRR-groups in any PF stratum, suggesting that preexisting CHD among men with low HRR could be the main cause of CHD risk differences observed in our study population. On the other hand, the borderline significant risk differences could reflect insufficient statistical power in our CHD risk analyses. We also found a trend towards shorter observation time to CV death among men with low HRR than among men with intermediate or high HRR. When excluding the 91 men with detectable ischemia at baseline, the risks of CV death among the HRR groups remained unchanged, suggesting that

64 European Journal of Preventive Cardiology 23(1) Table 4. Relative risks of cardiovascular death in tertiles of age-adjusted heart rate reserve (n ¼ 2014/1923). Relative risks of cardiovascular death in all participants End points and adjustments T1 (37 to 93 bpm) n ¼ 679 T2 (94 107 bpm) n ¼ 672 T3 (108 to 153 bpm) n ¼ 663 CV death, n (%) 198 (29.1) 179 (26.6) 143 (21.6) Unadjusted 1.45 (1.19 1.80) 1.10 (0.89 1.37) 1 Adjusted for age 1.68 (1.36 2.07) 1.21 (0.97 1.50) 1 Multiple adjusted 1.41 (1.14 1.75) 1.16 (0.93 1.44) 1 Multiple adjusted* 1.41 (1.13 1.76) 1.09 (0.87 1.37) 1 Hazard ratios of cardiovascular death in three strata of PF Men with low PF T1 (0.19 1.11 kj/kg) n ¼ 369 T2 n ¼ 202 T3 n ¼ 106 CV death, n (%) 122 (33.1) 56 (27.7) 25 (23.6) Unadjusted 1.60 (1.06 2.51) 1.18 (0.75 1.93) 1 Adjusted for age 1.77 (1.17 2.78) 1.33 (0.84 2.19) 1 Multiple adjusted 1.70 (1.12 2.67) 1.38 (0.87 2.26) 1 Multiple adjusted a 1.66 (1.09 2.66) 1.25 (0.78 2.08) 1 Men with intermediate PF (1.12 1.37 kj/kg) T1 n ¼ 205 T2 n ¼ 236 T3 n ¼ 231 CV death, n (%) 57 (27.8) 57 (24.2) 66 (28.6) Unadjusted 1.06 (0.74 1.50 0.88 (0.61 1.26) 1 Adjusted for age 1.25 (0.87 1.78) 0.99 (0.69 1.42) 1 Multiple adjusted 1.08 (0.74 1.55) 0.97 (0.68 1.39) 1 Multiple adjusted a 1.10 (0.76 1.59) 0.93 (0.64 1.34) 1 Men with high PF (1.68 5.71 kj/kg) T1 n ¼ 111 T2 n ¼ 220 T3 n ¼ 334 CV death, n (%) 25 (22.5) 52 (23.6) 68 (20.4) Unadjusted 1.14 (0.71 1.78) 1.17 (0.82 1.68) 1 Adjusted for age 1.31 (0.81 2.05) 1.21 (0.84 1.73) 1 Multiple adjusted 1.14 (0.70 1.80) 1.14 (0.79 1.65) 1 Multiple adjusted a 1.10 (0.67 1.76) 1.08 (0.74 1.60) 1 n indicates number of men; CV death, death from myocardial infarction, sudden cardiac death, stroke, aortic disease and pulmonary embolism; end points given as crude incidence (%); relative risks are given in hazard ratios (95% confidence interval) compared with T3; HRR: heart rate reserve; PF: physical fitness; multiple adjusted: adjusted for baseline age, smoking status, total cholesterol, family history of CHD, and systolic blood pressure. a 91 men with detected ischemia at exercise test excluded. pathophysiological pathways other than preexisting CHD cause the CV death risk differences observed. Resting HR is modifiable by endurance training and is, to a large degree, age independent. A low resting HR is associated with a high level of physical fitness (PF), and we have previously shown that high PF is associated with low CV death risk. 2,3 Still, there is no association between resting HR and CV death in our material (data not shown). Maximal HR, which is not modifiable by endurance training and largely age dependent, is significantly associated with CV death risk in our material, but we see no significant interaction between age-adjusted maximal HR and PF as CV death predictors (data not shown). Despite this, HRR and PF interact to a highly significant degree as CV death predictors, and HRR predicts CV death and CHD only in men with low PF. HRR reflects the complex interaction between the autonomic nervous system (ANS) and the cardiovascular system during exercise. Resting heart rate is mainly determined by parasympathetic activity, whereas increase of heart rate during exercise is determined by vagal withdrawal and sympathetic activation. Thus, a low HRR could be caused by a preexisting imbalance in the ANS or an inadequate heart response to an adequate ANS input. Jouven et al. found that subjects with low HRR had increased risk of sudden-chd death, but not of non-sudden CHD death, in a 25 year

Engeseth et al. 65 follow-up study of 5713 asymptomatic men. This and other studies indicate that autonomic imbalance, for example as revealed by low HRR, might be associated with proneness to develop lethal arrhythmias. 14 16 Despite this, we found that HRR did not predict CV death or CHD among men with intermediate or high PF. One way to interpret this difference among the PF strata is that autonomic imbalance caused by subclinical heart disease is more likely to impact PF than autonomic imbalance of other origin. It could also reflect that CV-protective effects of PF above a certain level cancel out unfavorable effects of autonomic imbalance or the effect of protective confounders associated with high PF, such as increased glucose metabolism, healthy diet, low alcohol consumption, and regular sleep, which have not been adjusted for in our statistical analyses. Clinical relevance Results of the present study add to the growing amount of evidence that autonomic imbalance, for example as revealed by measurement of HRR during exercise testing, is an important cardiovascular risk predictor, and that the added use of PF amplifies the predictive value of HRR on severe heart disease. Heart rate response to maximal exercise is easily measured during exercise testing and allows self-assessment during endurance training using commonly available HR and exercise monitoring equipment. We suggest that the combination of HRR and PF can be used to better target individuals at high CV risk and in greater need of primary preventive treatment. We have previously shown that added use of exercise data to classic risk factors increases the precision of CV risk models. 17 It is possible that use of HRR and PF could improve the accuracy of CV risk prediction, for example by re-classifying individuals from medium to high CV risk. Strengths The present study is prospective in design, and all data sets are complete with none lost to follow-up. We have no work-up bias, and all event data are on the basis of complete hospital records and cause-specific death records. All men were healthy and free of drugs, and the study group has not interfered with treatment in case of disease. The reproducibility of our exercise data was verified by a re-examination of 130 participants within 2 weeks of the initial examination and, as described earlier, reproducibility for heart rates and working capacity between the two tests was within 5 % in 90% of the men and within 10% in all of them. PF, as defined in the present study (total exercise work divided by body weight), has been shown to be highly correlated with maximum oxygen uptake, which is the most accepted measure of PF. 18,19 Limitations A recent study indicates that predicted maximal HR underestimates true maximal HR in middle-aged and older persons. We cannot rule out that this and intravariability in exercise capacity and responses in some participants could have influenced the results. 20,21 Our cohort consists of middle-aged white men who were healthy and employed full-time at inclusion and our findings cannot necessarily be generalized to individuals of other ages, ethnicities, or genders. Neither can our findings be applied in the setting of significant comorbidity and/or chronic drug use. Conclusions Low HRR was independently associated with increased risk of CV death in apparently healthy, middle-aged men. The predictive impact of HRR on CV death risk was, however, confined to unfit men. Funding This work was supported by a 3-year PhD scholarship from The South Eastern Norway Health Authority. The Oslo Ischemia study group received an unrestricted grant from AstraZeneca in 2010 and an unrestricted grant from Pronova Biopharma Norge AS in 2011. Conflict of interest Dr. Skretteberg received travel support from AstraZeneca and Pronova Biopharma Norge AS. Dr. Kjeldsen received honoraria for lectures from AstraZeneca, Bayer, MSD, Medtronic, Novartis, and Takeda and for consultancy agreements with Bayer, Medtronic, Serodus, and Takeda. Dr. Bodegard holds a full-time position as an epidemiologist with AstraZeneca. The other authors have no conflicts to report. References 1. Brubaker PH and Kitzman DW. Chronotropic incompetence: Causes, consequences, and management. Circulation 2011; 123: 1010 1020. 2. Sandvik L, Erikssen J, Thaulow E, et al. Physical fitness as a predictor of death among healthy, middle-aged Norwegian men. N Engl J Med 1993; 328: 533 537. 3. Sandvik L, Erikssen J, Ellestad M, et al. Heart rate increase and maximal heart rate during exercise as predictors of cardiovascular death: A 16-year follow-up study of 1960 healthy men. Coron Artery Dis 1995; 6: 667 679. 4. Erikssen J, Enge I, Forfang K and Storstein O. False positive diagnostic tests and coronary angiographic findings in

66 European Journal of Preventive Cardiology 23(1) 105 presumably healthy males. Circulation 1976; 54: 371 376. 5. Mundal R, Kjeldsen SE, Sandvik L, et al. Exercise blood pressure predicts cardiovascular death in middle-aged men. Hypertension 1994; 24: 56 62. 6. Bodegard J, Erikssen G, Bjornholt JV, et al. Reasons for terminating an exercise test provide independent prognostic information: 2014 apparently healthy men followed for 26 years. Eur Heart J 2005; 26: 1394 1401. 7. Hodnesdal C, Prestegaard E, Erikssen G, et al. Rapidly upsloping ST-segment on exercise ECG: a marker of reduced coronary heart disease mortality risk. Eur J Prev Cardiol 2013; 20(4): 541 548. 8. Bodegaard J, Erikssen G, Bjørnholt JV, et al. Symptomlimited exercise testing, ST-depressions and long-term coronary heart disease mortality in apparently healthy men followed for 26 years. Eur J Cardiovasc Prev Rehabil 2004; 11(4): 320 327. 9. Tardif JC. Heart rate as a treatable cardiovascular risk factor. Brit Med Bul 2009; 90: 71 84. 10. Fox K, Ford I, Steg PG, et al. Heart rate as a prognostic risk factor in patients with coronary artery disease and left-ventricular systolic dysfunction (beautiful): A subgroup analysis of a randomised controlled trial. Lancet 2008; 372: 817 821. 11. Orso F, Baldasseroni S and Maggioni AP. Heart rate in coronary syndromes and heart failure. Prog Cardiovasc Dis 2009; 52: 38 45. 12. Kannel WB. New perspectives on cardiovascular risk factors. Am Heart J 1987; 114: 213 219. 13. Ellestad MH and Wan MKC. Predictive implications of stress-testing follow-up of 2700 subjects after maximal treadmill stress testing. Circulation 1975; 51: 363 369. 14. Schwartz PJ, La Rovere MT and Vanoli E. Autonomic nervous system and sudden cardiac death: Experimental basis and clinical observations for post-myocardial risk stratification. Circulation 1988; 78: 969 979. 15. Jouven X, Empana JP, Schwartz PJ, et al. Heart rate profile during exercise as a predictor of sudden death. N Engl J Med 2005; 352: 1951 1958. 16. La Rovere MT, Bigger JT Jr, Marcus FI, et al. Baroreflex sensitivity and heart rate variability in prediction of total cardiac mortality after myocardial infarction. Lancet 1998; 351: 478 484. 17. Erikssen G, Bodegard J, Bjørnholt JV, et al. Exercise testing of healthy men in a new perspective: from diagnosis to prognosis. Eur Heart J 2004; 25: 978 986. 18. A strand PO. Textbook of work physiology; physiological bases of exercise. New York: McGraw-Hill, 1986. 19. Bonjer FH. Measurement of working capacity by assessment of the aerobic capacity in a single session. Fed Proc 1966; 25: 1363 1365. 20. Bruce RA. Methods of exercise testing. Step test, bicycle, treadmill, isometrics. Am J Cardiol 1974; 33(6): 715 720. 21. Nes BM, Janszky I, Wisløf U, et al. Age-predicted maximal heart rate in healthy subjects: The HUNT fitness study. Scand J Med Sci Sports 2013; 23(6): 697 704.