Heart rate and mortality from cardiovascular causes: a 12 year follow-up study of menandwomenaged40 45years

Transcription

1 European Heart Journal (2008) 29, doi: /eurheartj/ehn435 CLINICAL RESEARCH Prevention and epidemiology Heart rate and mortality from cardiovascular causes: a 12 year follow-up study of menandwomenaged40 45years Aage Tverdal*, Vidar Hjellvik, and Randi Selmer The Norwegian Institute of Public Health, PO Box 4404 Nydalen, Oslo N-0403, Norway Received 27 March 2008; revised 4 September 2008; accepted 12 September 2008; online publish-ahead-of-print 27 September 2008 Aim To study the relationship between heart rate and (a) all deaths and (b) cardiovascular deaths in a large cohort of middle-aged Norwegian men and women.... Methods A prospective study of participants in cardiovascular surveys that were carried out in and covered men and results and women aged years in all counties except the capital, Oslo. In total, men and women aged years without cardiovascular history or diabetes accrued years of follow-up. There was a positive and graded association between heart rate and mortality from all causes, as well as between heart rate and deaths from cardiovascular disease (CVD), ischaemic heart disease, and stroke. However, these associations were greatly reduced when we adjusted for the main risk factors of disease. The hazard ratios for any death were reduced from 3.14 to 1.82 for men (95% CI, ) and from 2.14 to 1.37 for women (95% CI, ), when we compared 95 b.p.m. with,65 b.p.m. The corresponding figures for CVD were a reduction from 4.79 to 1.51 for men (95% CI, ) and from 2.68 to 0.78 for women (95% CI, ).... Conclusion In this cohort of middle-aged men and women, a crude association between heart rate and death from CVDs was greatly weakened when we adjusted for the main risk factors of disease. This suggests that an increased heart rate in middle age may be a marker of high cardiovascular risk, but is not an independent risk factor Keywords Introduction Cardiovascular disease Ischaemic heart disease Stroke Total cholesterol Blood pressure Smoking Mortality Risk factor Several studies have suggested that increased heart rate is a risk factor for both cardiovascular and non-cardiovascular mortality. Palatini and Julius 1 have quite recently reviewed the evidence and described different mechanisms by which a fast heart rate might increase the risk of cardiovascular disease (CVD). Most studies have been done in men where it is easier to reach sufficient power to detect an association. In studies which include both men and women the findings have largely been consistent, but some have found a more pronounced relationship between heart rate and CVD in women, 2 whereas others have reported the opposite. 3 Palatini 4 has reviewed the question of the different strength of association in men and women. A few studies have looked at how the relationship between heart rate and mortality varies according to age, but their findings were not consistent. 3,5,6 We have used a large cohort of men and women in a narrow age range, who have been followed up for death for an average of 13 years. This allowed us to study the relationship between heart rate and specific cardiovascular outcomes such as mortality from ischaemic heart disease (IHD) and stroke, as well as sudden death of unknown cause. We were able to do this separately for men and women. The major research question was whether increased heart rate contributes to increased risk of death from various cardiovascular events over and above the contribution of the major cardiovascular risk factors. * Corresponding author. Tel: þ , Fax: þ , aage.tverdal@fhi.no Published on behalf of the European Society of Cardiology. All rights reserved. & The Author For permissions please journals.permissions@oxfordjournals.org.

2 Heart rate and mortality from cardiovascular causes 2773 Methods Study population From 1985 until 1999, the Norwegian government conducted health surveys inviting all women and men who were years in the first year of the survey in each county. In addition, people aged years were invited in a few counties. As the screening lasted over two calendar years in some counties, the participants were years at time of screening. The aims of the study were: (i) monitoring the CVD risk in the middle aged, (ii) epidemiological research, (iii) prevention of CVD through a high-risk strategy (on the basis of screening findings, 10% of the high-risk participants were referred to the local primary health services for follow-up examinations). All counties were included except for the capital, Oslo, where a similar programme was run by a municipal hospital. The attendance rate decreased over time and varied from 88% in Hedmark in 1988 to 52% in Østfold in In total, individuals participated. Of these, reported of a history of heart infarction, angina pectoris, stroke, diabetes, or being treated for hypertension. These individuals were excluded from our analysis. A further 1385 individuals who did not answer all the questions about disease and use of anti-hypertensive drugs, 858 with no measurement of heart rate, and 14 individuals registered as emigrating or dying before the date of the screening were also excluded. This left us individuals for analysis with a mean follow-up of 12.6 years. Even in this young and healthy population, we had 90% power to detect a hazard ratio of 1.25 in men and of 1.75 in women under equal allocation into two groups using a two-sided a ¼ 0.05 and with IHD death as endpoint. 7 Measurements A self-administered questionnaire was sent to the participants to be filled in at home. At the screening site, the questionnaire was checked for inconsistencies by a nurse, and omissions and logical inconsistencies were corrected according to a written protocol. The questionnaire included questions about current or previous CVD, diabetes, drug treatment for hypertension, physical activity in leisure time, smoking habits, and whether parents or siblings had suffered from heart infarction or stroke. Except for the years , the question about physical activity during leisure time had four alternatives classified as: sedentary, moderate, intermediate, or intensive. During ( participants, 22% of study population), we asked about serious physical activity (causing sweating or breathlessness) with alternatives 0,,1, 1 2, and 3þ h/week. We have dichotomized into sedentary vs. other and 0 h/week vs. other and used the term sedentary/not sedentary for both classifications. Heart rate and systolic and diastolic blood pressure were measured by a trained nurse using an automatic device (DINAMAP, Criticon, Tampa, USA). After 2 min rest, three recordings were made at 1 min intervals, and we have used the average of the second and third measurements, which are considered to be the most representative. The circumference was measured 10 cm above fossa cubiti, and one of the three cuffs was selected accordingly. A non-fasting blood sample was taken, and the serum was analysed for total cholesterol and triglycerides using an enzymatic method. Height and weight were measured to the nearest centimetre and half kilogram. Underwear, stockings, and trousers were allowed, but not shoes. Endpoints The individuals accrued person-years from the date of examination until the date of death (n ¼ 8951), emigration (n ¼ 2515), or 31 December We studied CVD [ICD 8 (1985) , ICD 9 ( ) , ICD 10 ( ) I00-I99], IHD (ICD , ICD 10 I20-I25), stroke (ICD , ICD 10 I60-I69), sudden death of unknown cause (ICD , 795, ICD ,798.2, ICD 10 R96), and all-cause mortality. The causes of death were taken from the National Cause of Death Register. Files could be linked because of the personal identification number allocated to each individual in Norway. The Norwegian Data Inspectorate and the Norwegian Directorate of Health gave permission to the linking. Statistical methods We performed analyses both with heart rate as a continuous and as a categorical variable. In the latter case, to cover the range in heart rate better, we used 65, 80, and 95 b.p.m. as group limits rather than the 25, 50, and 75% quantiles even though this resulted in an unequal distribution within groups. 8 The set of confounders were defined in advance and have all been shown to be related to CVD: total cholesterol, triglycerides, diastolic blood pressure, smoking, physical activity, and family history of myocardial infarction. Calendar year was included because cardiovascular mortality has decreased over time. Cholesterol, triglycerides, blood pressure, and calendar year were entered as continuous. We tested for differences in risk factors at different levels of heart rate using the F statistic from analysis of variance or the chi-square statistic from the analysis of contingency tables. Hazard ratios were estimated using the Cox proportional-hazards model. The different levels of heart rate were included as dummy variables with the lowest level (,65 b.p.m.) as the reference group. The proportional-hazards assumption was assessed by visual inspection of parallel lines in plots of ln ln(survival probability) vs. ln(time). The linearity assumption of the continuous variables was assessed by comparing the models having only a linear term with the models having both a linear and higher order polynominal terms. The interaction was tested using the log-likelihood ratio test from Cox proportionalhazards models with and without the interaction term. In these models, heart rate was entered as a continuous variable. We did separate analyses for subjects with high and low blood pressure, high and low cholesterol, and high and low body mass index, using the corresponding medians for defining the groups. We also stratified on smoking and physical activity. We obtained non-parametric estimates with 95% confidence bands of the log-hazard for heart rate by fitting Cox models with a P-spline function to the data (coxph and pspline functions in R 9 ). The log-hazard estimate from coxph was scaled, so the average log-hazard over all individuals was zero. From the log-hazard [denoted z(h)], we calculated approximate estimates of mortality per person-years as a function of heart rate (h) as ZðhÞ ¼ expfzðhþgn d Y 1 expfzðhþg 1 where N d and Y are the number of deaths and accrued person-years, respectively, and the overline denotes average. Confidence bands for Z(h) were calculated by replacing the first occurrence of z(h) in the above formula with the confidence bands for z(h). Results There was no significant association between heart rate and age, but all the other risk factors increased with increasing levels of heart rate (Table 1). Some risk factors increased distinctly. Diastolic blood pressure was roughly 13 mmhg higher in the highest than the lowest heart rate group, and the smoking prevalence varied

3 2774 A. Tverdal et al. Table 1 Baseline information in men and women Characteristics Heart rate (b.p.m.) P-equality... < Men n (with heart rate measurement) Age (years) Heart rate (b.p.m.), mean (SD) 58.4 (4.9) 71.9 (4.2) 85.7 (4.1) (7.7),0.001 Total cholesterol (mmol/l), mean (SD) 5.63 (1.03) 5.82 (1.08) 5.99 (1.13) 6.16 (1.19),0.001 Triglycerides (mmol/l), mean (SD) 1.82 (1.09) 2.15 (1.35) 2.41 (1.56) 2.62 (1.87),0.001 Diastolic blood pressure (mmhg), mean (SD) 76.4 (8.6) 79.9 (9.2) 83.7 (9.7) 88.9 (10.3),0.001 Systolic blood pressure (mmhg), mean (SD) (11.8) (12.6) (14.0) (16.8),0.001 Body mass index (kg/m 2 ), mean (SD) 25.3 (2.8) 25.7 (3.2) 26.0 (3.6) 26.3 (4.0),0.001 Smoking, % ,0.001 Sedentary, % ,0.001 Family history of CVD, % ,0.001 Women n (with heart rate measurement) Age (years) Heart rate (b.p.m.), mean (SD) 60.0 (3.9) 72.5 (4.2) 85.9 (4.1) (8.7),0.001 Total cholesterol (mmol/l), mean (SD) 5.25 (0.91) 5.38 (1.00) 5.54 (1.01) 5.70 (1.07),0.001 Triglycerides (mmol/l), mean (SD) 1.15 (0.62) 1.28 (0.74) 1.43 (0.88) 1.52 (0.97),0.001 Diastolic blood pressure (mmhg), mean (SD) 70.4 (8.5) 73.9 (9.1) 77.8 (9.5) 83.3 (10.4),0.001 Systolic blood pressure (mmhg), mean (SD) (11.7) (12.4) (13.8) (16.9),0.001 Body mass index (kg/m 2 ), mean (SD) 24.1 (3.3) 24.3 (3.7) 24.4 (4.1) 24.6 (4.5),0.001 Smoking, % ,0.001 Sedentary, % ,0.001 Family history of CVD, % ,0.001 by a factor of 2 between these groups. Mean heart rate in the study population was 72.6 b.p.m. for men and 77.1 b.p.m. for women. Shape of relationships For heart rates between 60 and 100 b.p.m., the relationship between heart rate and mortality was approximately linear on the log-scale for all four endpoints (Figure 1). Above 100 b.p.m., the mortality levelled off. The linearity assumption (on the log-scale) was rejected only for any death. Excluding persons with heart rate,60 or.100 b.p.m. gave only slightly higher hazard ratios than using the total study population. Cardiovascular death There was a strong trend for increased mortality with higher heart rate in both sexes. The hazard ratios between 95 and,65 b.p.m. were 4.79 for men and 2.68 for women (Table 2). They were substantially reduced to 1.51 and 0.78 after adjustment. There was still a significantly increased mortality in men with a higher heart rate, but only in the 95 b.p.m. group. The trend disappeared completely for women and, if anything, was reversed. In women, the mortality rates in the and b.p.m. groups were significantly lower than in the,65 b.p.m. group. Any death Mortality increased in a graded manner, more than threefold for men and more than twofold for women, from the lowest to the highest heart rate group (Table 2). After adjusting for other risk factors, the relationship weakened and the increase in mortality was less than twofold in men and,50% in women. There was no longer a significantly increased mortality in men and women with heart rate b.p.m. as compared with,65 b.p.m. Treating heart rate as a continuous variable, the increased risk associated with an increase of 10 b.p.m. was halved by adjusting for other risk factors, from 28 to 14% in men and from 20 to 9% in women. Death from ischaemic heart disease The pattern was quite similar to that for cardiovascular death for men (Table 2). After adjustment for risk factors, only men with 95 b.p.m. had a significantly higher mortality than the,65 b.p.m. group. For women, the relationship was stronger than for cardiovascular death, but after adjustment there was no significant trend. There was, however, a suggestion of increased mortality in the 95 b.p.m. group as compared with the b.p.m. group for women. There were 4.6 deaths in men for each death in women overall. An unadjusted increase in mortality of 42% per 10 b.p.m. was reduced to 12 and 10% in men and women, respectively, after adjustment for other risk factors.

4 Heart rate and mortality from cardiovascular causes 2775 Figure 1 Non-parametric estimates of mortality as a function of heart rate. Grey lines show 95% confidence bands. The y-axis is logtransformed. The distribution of heart rate for men and women is indicated in the lower right plot. Stroke death In men, a significant positive trend became non-significant after adjustment for other risk factors (Table 2). In women, a significant positive trend became significantly negative. Stroke deaths were almost equally frequent in men and women. There seemed to be an opposite trend in men and women after adjustment for risk factors, but the gender heart rate interaction was not significant (P ¼ 0.13). Of the 178 stroke deaths in women, 56.7% were diagnosed as subarachnoid haemorrhage, 21.9% as haemorrhage, 10.7% as ischaemia, and 10.7% as others (Transient cerebral ischemia, ill-defined, and late effects of cerebrovascular disease). The corresponding figures in men (164 stroke deaths) were 39.6, 34.1, 14.0, and 12.2%. Sudden death of unknown cause In men, there was no significant association between increased heart rate and sudden death by unknown cause (Table 2). Hazard ratios above 1.0 became,1.0 after adjustment. A significant positive trend among women disappeared after adjustment for other risk factors. Other than cardiovascular and sudden death In both sexes, there was a positive and graded association which was weakened, but still significant after adjustment (Table 2). The adjusted mortality in men doubled from the lowest to the highest heart rate group, and in women there was a 50% increase in mortality from the lowest to the highest heart rate group. The increase in mortality associated with an increase of 10 b.p.m. was 16 and 10% in men and in women, respectively. Adjustment for one confounder at a time Table 3 shows the reduction in the hazard ratio when adjusting individually for each of the three major risk factors smoking, diastolic blood pressure, and total cholesterol, and when adjusting for all three simultaneously. Diastolic blood pressure contributed most to the reduction for the cardiovascular deaths. The reduction was most pronounced for stroke deaths. Adjusting for triglycerides, physical activity, and family history in addition to the three major risk factors changed the hazard ratios very little (see Table 2).

5 2776 A. Tverdal et al. Table 2 Mortality from various causes of death by heart rate (hazard ratios) Cause of death Heart rate (b.p.m.) HR per 10 b.p.m.... (95% CI) Deaths, hazard ratio < (HR) (95% CI) Men Person-years Any death HR a Ref 1.33 ( ) 1.85 ( ) 3.14 ( ) 1.28 ( ) HR b Ref 1.05 ( ) 1.21 ( ) 1.82 ( ) 1.14 ( ) CVD HR a Ref 1.50 ( ) 2.35 ( ) 4.79 ( ) 1.40 ( ) HR b Ref 0.94 ( ) 1.01 ( ) 1.51 ( ) 1.10 ( ) IHD HR a Ref 1.47 ( ) 2.44 ( ) 5.06 ( ) 1.42 ( ) HR b Ref 0.91 ( ) 1.05 ( ) 1.61 ( ) 1.12 ( ) Stroke HR a Ref 1.31 ( ) 1.99 ( ) 4.09 ( ) 1.35 ( ) HR b Ref 0.84 ( ) 0.88 ( ) 1.32 ( ) 1.06 ( ) Sudden death HR a Ref 1.30 ( ) 1.31 ( ) 1.44 ( ) 1.05 ( ) HR b Ref 0.98 ( ) 0.81 ( ) 0.78 ( ) 0.89 ( ) Other causes HR a Ref 1.29 ( ) 1.73 ( ) 2.73 ( ) 1.24 ( ) HR b Ref 1.10 ( ) 1.32 ( ) 1.98 ( ) 1.16 ( ) Women Person-years Any death HR a Ref 1.23 ( ) 1.74 ( ) 2.14 ( ) 1.20 ( ) HR b Ref 1.06 ( ) 1.28 ( ) 1.37 ( ) 1.09 ( ) CVD HR a Ref 1.02 ( ) 1.58 ( ) 2.68 ( ) 1.30 ( ) HR b Ref 0.67 ( ) 0.70 ( ) 0.78 ( ) 1.00 ( ) IHD HR a Ref 1.28 ( ) 2.72 ( ) 4.42 ( ) 1.42 ( ) HR b Ref 0.79 ( ) 1.11 ( ) 1.22 ( ) 1.10 ( ) Stroke HR a Ref 0.99 ( ) 1.23 ( ) 1.96 ( ) 1.18 ( ) HR b Ref 0.63 ( ) 0.52 ( ) 0.50 ( ) 0.87 ( ) Sudden death HR a Ref c Ref c 1.00 ( ) 3.48 ( ) 1.50 ( ) HR b Ref c Ref c 0.55 ( ) 1.03 ( ) 1.10 ( ) Other causes HR a Ref 1.24 ( ) 1.75 ( ) 2.03 ( ) 1.18 ( ) HR b Ref 1.12 ( ) 1.40 ( ) 1.48 ( ) 1.10 ( ) a Adjusted for calendar year. b Adjusted for calendar year, total cholesterol, triglycerides, diastolic blood pressure, smoking, physical activity, and family history. c Reference group:,80 b.p.m. Subgroup analyses: ischaemic heart disease Crude mortality rates The mortality increased with increasing heart rate in all subgroups (Table 4). The mortality in smoking men with a heart rate,65 b.p.m. was comparable with the mortality in non-smoking men inthehighestheartrategroup(table 4). Smoking women had a higher mortality than non-smoking women, regardless of heart rate level. The largest mortality difference between the outer heart rate groups was found for smokers. This applies to both men and women.

6 Heart rate and mortality from cardiovascular causes 2777 Table 3 Hazard ratios with 95% CI estimated by Cox proportional-hazards regression Adjusted for calendar HR per 10 b.p.m. (95% CI) year and:... Men Women... Any death None 1.28 ( ) 1.20 ( ) Diastolic blood pressure 1.23 ( ) 1.15 ( ) (mmhg) Smoking (yes, no) 1.22 ( ) 1.16 ( ) Total cholesterol (mmol/l) 1.26 ( ) 1.19 ( ) All three 1.14 ( ) 1.10 ( )... Cardiovascular disease None 1.40 ( ) 1.30 ( ) Diastolic blood pressure 1.23 ( ) 1.09 ( ) (mmhg) Smoking (yes, no) 1.31 ( ) 1.22 ( ) Total cholesterol (mmol/l) 1.34 ( ) 1.29 ( ) All three 1.10 ( ) 1.01 ( )... Ischaemic heart disease None 1.42 ( ) 1.42 ( ) Diastolic blood pressure 1.26 ( ) 1.20 ( ) (mmhg) Smoking (yes, no) 1.32 ( ) 1.33 ( ) Total cholesterol (mmol/l) 1.34 ( ) 1.41 ( ) All three 1.11 ( ) 1.10 ( )... Stroke None 1.35 ( ) 1.18 ( ) Diastolic blood pressure 1.14 ( ) 0.98 ( ) (mmhg) Smoking (yes, no) 1.28 ( ) 1.10 ( ) Total cholesterol (mmol/l) 1.35 ( ) 1.18 ( ) All three 1.06 ( ) 0.90 ( ) Adjusted relative mortality Among men, there was a clearly weaker association between heart rate and IHD mortality in the high blood pressure group than in the low blood pressure group (Table 5). The confidence intervals for the hazard ratio per 10 b.p.m. did not overlap. There was also a suggestion of a stronger relationship between heart rate and mortality at low than at high levels of cholesterol and body mass index. For smoking, on the other hand, the strongest association was for the high-risk group (smokers), with non-overlapping confidence interval for the hazard ratios. Among women, there was also a weaker relationship between heart rate and death from IHD in the low blood pressure groups. However, there were fewer deaths overall and no significant trends in any of the subgroups except for the high cholesterol group where the hazard ratio per 10 b.p.m. is borderline significant. Discussion This study revealed evidence of a positive and graded association between heart rate and mortality from cardiovascular and non-cardiovascular causes and between heart rate and death from any cause. After adjustment for other risk factors, the relationship between heart rate and risk of death from CVD was much weaker. However, a statistically significant effect of high heart rate remained for cardiovascular and IHD mortality among men. High heart rate was associated with an unfavourable pattern of risk factors. The association with lipids is not readily explained. In a subpopulation within this study, of men and women, we have information on the type of fat used on bread and on educational level. Those with higher heart rate levels had less education and used less favourable types of fat. In this subpopulation, adjustment for education, marital status, type of fat, physical activity, and smoking reduced the difference between the mean cholesterol levels in the highest and lowest heart rate group by 7%. The difference between triglyceride levels was reduced by 6% in men and 11% in women. This suggests that only a minor fraction of the lipid contrast across heart rate levels could be attributed to differences in our indicators of life style and socioeconomic status. The relationship between body mass index and heart rate was quite weak. Linear regression analysis gave an increase of 0.37 b.p.m. per 1 kg/m 2 increase in body mass index in men and 0.12 b.p.m. in women. Restricting the analyses to the normotensives with systolic and diastolic blood pressure,140 and 90 mmhg, respectively, halved the coefficient for men and the coefficient for women became slightly negative. Thus, there was none or minor influence of body mass index on heart rate in the absence of raised blood pressure. Cigarette smoking acutely raises the heart rate, and the raised level remains for min after the cigarette is put out. 10,11 Many smokers may well have smoked within this time interval prior to the screening. Furthermore, cigarette smoking is reported to have an unfavourable effect on arterial stiffness which in the longer run may influence the heart rate. 10,11 According to a review article by Palatini, 4 heart rate may be a marker of the sympathetic nervous system activity. 12 It is known that long-term sympathetic over-activity can produce a state of insulin resistance. 13,14 A recent study on non-diabetic subjects concluded that hypertriglyceridemia, low HDL-cholesterol, and hyperuricemia seemed to be particularly linked with both hyperinsulinemia and insulin resistance. 15 On the other hand, the relationship between heart rate and other causes of death points towards additional underlying factors other than the metabolic syndrome. Only 43 men and 26 women in our study had diabetes mellitus mentioned on the death certificate. Many studies have concluded that heart rate is an independent risk factor for CVD. 3,5,16,17 In some of the studies, there were only minor changes in the hazard ratio estimates when adjusting for covariates, 16,17 whereas other studies found lower ratios after risk adjustment. 5 In our study, a significant effect of heart rate on ischaemic death remained after adjustment among males. On the other hand, there was no significant effect on stroke after adjustment for men, but for women there was an inverse effect with hazard ratios for the group being significantly smaller than 1.0. Palatini 4 has discussed whether heart rate is a cardiovascular risk factor in men only, because many studies have reported a much weaker association in women. In our

7 2778 A. Tverdal et al. Table 4 Mortality from ischaemic heart disease by heart rate among subpopulations Per person-years (deaths) Heart rate (b.p.m.)... < Men Systolic blood pressure (mmhg),median (132.5) 14.1 (49) 22.9 (119) 42.4 (78) (28) Median 30.2 (81) 40.7 (219) 61.6 (181) (114) Diastolic blood pressure (mmhg),median (79.5) 15.3 (60) 21.0 (105) 34.6 (52) (23) Median 31.2 (70) 41.8 (233) 63.2 (207) (119) Total cholesterol (mmol/l),median (5.74) 11.7 (41) 16.6 (86) 27.8 (57) 78.4 (35) Median 33.6 (89) 46.7 (252) 74.2 (202) (106) Body mass index (kg/m 2 ),Median (25.3) 14.8 (50) 29.2 (154) 47.4 (105) (62) Median 28.8 (80) 34.7 (184) 60.0 (153) (80) Smoking Yes 37.1 (57) 55.0 (246) 82.0 (215) (119) No 15.9 (73) 15.2 (92) 20.6 (44) 43.9 (23) Physical activity Sedentary 43.0 (34) 34.4 (77) 76.6 (94) (41) Not sedentary 17.9 (96) 31.3 (261) 46.5 (165) (101) Women Systolic blood pressure (mmhg),median (122.5) 2.8 (6) 3.6 (23) 6.6 (19) 14.0 (6) Median 4.7 (6) 6.1 (34) 13.2 (59) 19.3 (35) Diastolic blood pressure (mmhg),median (74.5) 2.6 (6) 3.6 (22) 4.8 (12) 10.6 (4) Median 5.3 (6) 6.0 (35) 13.7 (66) 19.9 (37) Total cholesterol (mmol/l),median (5.34) 2.6 (5) 2.6 (16) 4.3 (14) 9.5 (8) Median 4.7 (7) 6.8 (40) 15.5 (63) 23.7 (33) Body mass index (kg/m 2 ),Median (23.6) 3.3 (6) 4.7 (29) 10.0 (37) 13.5 (15) Median 3.6 (6) 4.6 (27) 11.1 (40) 23.2 (26) Smoking Yes 8.4 (8) 10.1 (45) 18.3 (65) 30.1 (35) No 1.6 (4) 1.6 (12) 3.5 (13) 5.6 (6) Physical activity Sedentary 0 (0) 6.0 (15) 16.4 (29) 19.9 (11) Not sedentary 4.2 (12) 4.4 (42) 8.8 (49) 17.8 (30) study population, there were five times more ischaemic than stroke deaths in men whereas in women there was a one-to-one ratio. As stroke deaths make up a larger proportion of cardiovascular deaths in women than in men, this accounts for the somewhat weaker relationship between heart rate and cardiovascular mortality among women. Stroke deaths also make up a different proportion of cardiovascular deaths at different ages. Therefore, one would expect the heart rate-cardiovascular death relationship to vary with age at baseline, and also with length of follow-up. This needs to be tested in a population of wider age distribution. Quite recently a state-of the art paper was published on the issue of resting heart rate in CVD. 18 The conclusion was that

8 Heart rate and mortality from cardiovascular causes 2779 Table 5 Mortality from ischaemic heart disease by heart rate among subpopulations Hazard ratio a Heart rate (b.p.m.) Hazard ratio a (HR) (95% CI)... per 10 b.p.m. (95% CI) < Men Systolic blood pressure (mmhg),median (132.5) Ref 1.04 ( ) 1.39 ( ) 2.80 ( ) 1.28 ( ) Median Ref 0.83 ( ) 0.88 ( ) 1.30 ( ) 1.07 ( ) Diastolic blood pressure (mmhg),median (79.5) Ref 0.98 ( ) 1.30 ( ) 4.17 ( ) 1.31 ( ) Median Ref 0.92 ( ) 1.04 ( ) 1.45 ( ) 1.09 ( ) Total cholesterol (mmol/l),median (5.74) Ref 0.94 ( ) 1.13 ( ) 2.60 ( ) 1.20 ( ) Median Ref 0.89 ( ) 0.99 ( ) 1.37 ( ) 1.10 ( ) Body mass index (kg/m 2 ),Median (25.3) Ref 1.15 ( ) 1.24 ( ) 2.10 ( ) 1.16 ( ) Median Ref 0.75 ( ) 0.87 ( ) 1.22 ( ) 1.08 ( ) Smoking Yes Ref 1.16 ( ) 1.37 ( ) 2.14 ( ) 1.18 ( ) No Ref 0.69 ( ) 0.68 ( ) 0.96 ( ) 0.96 ( ) Physical activity Sedentary Ref 0.54 ( ) 0.85 ( ) 0.97 ( ) 1.13 ( ) Not sedentary Ref 1.07 ( ) 1.06 ( ) 1.90 ( ) 1.11 ( ) Women Systolic blood pressure (mmhg),median (122.5) Ref 0.81 ( ) 1.08 ( ) 1.88 ( ) 1.23 ( ) Median Ref 0.82 ( ) 1.15 ( ) 1.13 ( ) 1.06 ( ) Diastolic blood pressure (mmhg),median (74.5) Ref 0.98 ( ) 0.98 ( ) 2.10 ( ) 1.16 ( ) Median Ref 0.82 ( ) 1.36 ( ) 1.34 ( ) 1.11 ( ) Total cholesterol (mmol/l),median (5.34) Ref 0.60 ( ) 0.66 ( ) 0.92 ( ) 0.99 ( ) Median Ref 0.92 ( ) 1.37 ( ) 1.41 ( ) 1.13 ( ) Body mass index (kg/m 2 ),Median (23.6) Ref 0.84 ( ) 1.05 ( ) 0.83 ( ) 1.10 ( ) Median Ref 0.72 ( ) 1.05 ( ) 1.33 ( ) 1.13 ( ) Smoking Yes Ref 0.85 ( ) 1.13 ( ) 1.17 ( ) 1.08 ( ) No Ref 0.66 ( ) 0.98 ( ) 1.22 ( ) 1.07 ( ) Physical activity Sedentary Ref b Ref b 2.20 ( ) 1.69 ( ) 1.20 ( ) Not sedentary Ref 0.64 ( ) 0.78 ( ) 1.00 ( ) 1.06 ( ) a Adjusted for calendar year, total cholesterol, triglycerides, diastolic blood pressure (or systolic in diastolic subgroups), smoking, physical activity, and family history. b Reference group:,80 b.p.m. heart rate is a risk factor for cardiovascular mortality independent of currently accepted risk factors and other potentially confounding demographic and physiological characteristics. The paper specifically referred to two recent large follow-up studies. The first one 19 included men with no clinically detectable CVD and the other 20 included men and women with suspected or proven coronary artery disease. The study population in the first one is comparable with the male part of our study population,

9 2780 A. Tverdal et al. and a particularly strong relationship was found with sudden death from myocardial infarction. 19 Unfortunately, the ICD classification does not distinguish between sudden (,1 h after symptom onset) and non-sudden death from myocardial infarction. In our male and female study population, 66 and 79%, respectively, of the deaths from IHD were classified as acute myocardial infarction that includes first-time cases specified as acute or with a stated duration of 28 days or less from onset. The relationship between heart rate and IHD mortality might be due to a relationship between heart rate and incidence or between heart rate and case fatality. The study of Jouven et al. 19 points to the contribution of both relationships. Few studies have reported on both incidence and mortality. In the NHANES I study, both the incidence and the mortality of coronary heart disease were used as endpoints. 5 There was no consistent pattern across gender and age. Our study differs from others in that the relationship between heart rate and mortality from IHD is stronger in our study, when measured as unadjusted hazard ratios. This is probably because of a relatively young study population. It has been shown that the higher the risk in absolute terms the lower the risk in relative terms. Furthermore, few, if any, other studies have reported precise risk estimates at levels as high as 95 b.p.m. Another particular feature of our study is the distinct change in the relationships after adjustment for known risk factors. High levels of heart rate were associated with clearly higher levels of the major risk factors. The correlation coefficient between heart rate and diastolic blood pressure was 0.35 and 0.37 in men and women, respectively. This may be compared with the correlation coefficients in eight studies reported in the review article by Palatini. 4 The coefficients for men were somewhat lower than in our study whereas the coefficients for women were distinctly lower. A study from Tromsø which was carried out by the same personnel and with the same equipment as our study, found comparable estimates to those in our study, 0.30 in men and 0.33 in women. 21 It would be expected that the higher the correlation with blood pressure the greater the decrease in risk estimates when adjustment is made for blood pressure. We cannot rule out the idea that the relatively high correlation coefficients in our study are partly due to the measurement environment. The participants may, for instance, have been anticipating the venipuncture that would succeed the blood pressure and heart rate measurements at the screening site, which may have led to raised levels of blood pressure as well as heart rate. The substantial reduction in hazard ratios after adjustment indicates residual confounding. A recent huge meta-analysis concluded that the ratio total/hdl-cholesterol was the strongest predictor of mortality from IHD, more than twice as informative as total cholesterol. 22 We had data on HDL-cholesterol only for the individuals screened between 1995 and 1999 (29% of the study population), and only 8% of the ischaemic deaths occurred within this young sub-population. Therefore, HDL-cholesterol was not included in the analysis. We also used a crude measure of physical activity and we did not have data on physical fitness. Among emerging risk factors are C-reactive protein, lipoprotein(a), fibrinogen, and homocysteine, but there is little data on their prognostic value in addition to the major risk factors. 23 Williams et al. 24 found that heart rate was strongly related to concentrations of lipoprotein sub-fractions in sedentary men. This lends support to their hypothesis of lipoprotein-induced relationships between coronary heart disease and heart rate. As mentioned, we had sparse information on HDL and it may be that adjustment for the ratio total/hdl-cholesterol would have weakened the heart rate IHD relationship further towards the null, although adjusting for total/hdl-cholesterol in the subpopulation for which HDL-cholesterol was measured did not give strong indications in that direction [replacing total cholesterol with the ratio total/hdl-cholesterol in the set of confounders only led to a reduction from 1.21 ( ) to 1.20 in the adjusted hazard ratio for the 95 group for ischaemic death in men]. The measurement of risk factors is encumbered by errors caused by equipment and intra-individual variation. If these errors are random and if the relationship of an exposure with an outcome is graded, then random errors will result in underestimation of the exposure effect. The association with true resting heart rate is stronger than the associations based on a single measurement. The correlation between repeat measurements taken 3 years apart in a small sample (4750 subjects) was 0.64 for heart rate and 0.70 for blood pressure. So, the effect of regression dilution is only slightly stronger for heart rate than for blood pressure. The average relative change in heart rate and diastolic blood pressure during the 3 years was,1% for both sexes. The percentage of smokers decreased from 55 to 52 for men and from 42 to 41 for women. All in all, the changes during the first part of follow-up seemed to be small for this subgroup. This study reveals a positive and graded association between heart rate and non-cardiovascular mortality. Cancer dominated with 1608 (49% of non-cardiovascular deaths) and 2207 (73% of non-cardiovascular deaths) in men and women, respectively. Second most frequent was external causes with 840 and 329 deaths. The relationship between heart rate and specific noncardiovascular causes is, however, not the focus of this study, and has not been pursued. The strength of this study is its size and the complete follow-up to death or emigration within the time period considered. Even though the survey extended over 15 years, the same procedures, the same equipment, the same laboratory, and the same personnel were used throughout. The narrow age range means that this important confounder is, to a large extent, handled in the design. On the other hand, we cannot study whether the relationship between heart rate and mortality varies with age. Limitations The endpoints in this study were taken directly from the official death certificates that are sent to Statistics Norway. The certificates are checked for completeness so that the coding of the causes of death is possible according to the International Classification of Deaths. During the follow-up period, only 10 15% of all deaths in Norway have been autopsied. In this study, 43% of the cardiovascular deaths took place in a hospital or an institution, whereas 53% took place outside hospital/institution. For the remaining 4%, the place of death is not recorded. No external validation of the diagnoses on the death certificate has been done. Despite standardized procedures, measurement errors do occur. In this study, these are most likely non-differential and random. Random measurement errors of heart rate would yield

10 Heart rate and mortality from cardiovascular causes 2781 estimates biased towards the null. On the other hand, random errors of the confounders could implicate suboptimal adjustment of the hazard ratios. The overall consequence is hard to predict. There are only hard endpoints in this study. The observed relationships between heart rate and mortality might be a consequence of a relation both to morbidity and to case fatality. A relationship to morbidity is well documented in other studies. However, we are not aware of any documentation regarding heart rate and case fatality. One should also keep in mind that 10% of those with the highest levels of the major risk factors were referred to local primary health services for follow-up examination. A larger percentage of subjects with high than with low heart rate were thereby recommended to follow-up examination. If this closer contact with the health services resulted in a greater reduction in risk factors, this would have biased our risk estimates for CVD mortality towards the null. Conclusions To summarize, we found a positive and graded relationship between heart rate and mortality from CVD, IHD, stroke, and death from any cause both in men and women, which was greatly decreased after adjustment for major cardiovascular risk factors. This suggests that heart rate in middle age is a marker for cardiovascular death, possibly through its relationship with other risk factors. On the other hand, there was evidence of an effect of high heart rate on IHD mortality in male subgroups over and above the effect of major cardiovascular risk factors. Finally, the association with CVD mortality was weaker in women than in men. This can be explained by varying associations with different cardiovascular endpoints. In brief, increased heart rate in middle aged men may be a marker of increased risk of death from IHD. Conflict of interest: none declared. References 1. Palatini P, Julius S. Elevated heart rate: a major risk factor for cardiovascular disease. Clin Exp Hypertens 2004;26: King DE, Everett CJ, Mainous AG III, Liszka HA. Long-term prognostic value of resting heart rate in subjects with prehypertension. Am J Hypertens 2006;19: Kannel WB, Kannel C, Paffenbarger RS Jr, Cupples LA. Heart rate and cardiovascular mortality: the Framingham Study. Am Heart J 1987;113: Palatini P. Heart rate as a cardiovascular risk factor: do women differ from men? Ann Med 2001;33: Gillum RF, Makuc DM, Feldman JJ. Pulse rate, coronary heart disease, and death: the NHANES I epidemiologic follow-up study. Am Heart J 1991;121: Greenland P, Daviglus ML, Dyer AR, Liu K, Huang C-F, Goldberger JJ, Stamler J. Resting heart rate is a risk factor for cardiovascular and noncardiovascular mortality: the Chicago Heart Association Detection Project in Industry. Am J Epidemiol 1999;149: Therneau TM, Grambsch PM. Modeling Survival Data. Extending the Cox Model. New York: Springer; p Cox DR. Note on grouping. J Am Stat Assoc 1957;280: R Development Core Team. R: a Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; ISBN Caro CG, Parker KH, Lever MJ, Fish PJ. Effect of cigarette smoking on the pattern of arterial blood flow: possible insights into mechanisms underlying the development of arteriosclerosis. Lancet 1987;330: Kim JW, Park CG, Hong SJ, Park SM, Rha SW, Seo HS, Oh DJ, Rho YM. Acute and chronic effects of cigarette smoking on arterial stiffness. Blood Pressure 2005;14: Grassi G, Vailati S, Bertinierei G, Tverdal A, Hjellvik V, Selmer R. Heart rate as a marker of sympathetic activity. J Hypertens 1998;16: Jamerson KA, Julius S, Gudbrandsson T, Andersson O, Brant DO. Reflex sympathetic activation induces acute insulin resistance in the human forearm. Hypertension 1993;21: Deibert DC, DeFronzo RA. Epinephrine-induced insulin resistance in man. J Clin Invest 1980;65: Bonora E, Capaldo B, Perin PC, Del Prato S, De Mattia G, Frittitta L, Frontoni S, Leonetti F, Luzi L, Marchesini G, Marini MA, Natali A, Paolissi G, Piatti PM, Pujia A, Solini A, Vettor R, Bonadonna RC, on behalf of the Group of Italian Scientists of Insulin Resistance (GISIR). Hyperinsulinemia and insulin resistance are independently associated with plasma lipids, uric acid, and blood pressure in non-diabetic subjects. The GISIR database. Nutr Metab Cardiovasc Dis 2007; doi: / j.numecd Benetos A, Rudnichi A, Thomas F, Safar M, Guize L. Influence of heart rate on mortality in a French population: role of age, gender, and blood pressure. Hypertension 1999;33: Seccareccia F, Pannozzo F, Dima F, Minoprio A, Menditto A, Noce CL, Giampaoli S. Heart rate as a predictor of mortality: the MATISS project. Am J Public Health 2001;91: Fox K, Borer JS, Camm AJ, Danchin N, Ferrari R, Sendon JLL, Steg PG, Tardif J-C, Tavazzi L, Tendera M, for the Heart Rate Working Group. Resting heart rate in cardiovascular disease. J Am Coll Cardiol 2007;50: Jouven X, Empana JP, Schwartz PJ, Desnos M, Courbon D, Ducimetiere P. Heartrate profile during exercise as a predictor of sudden death. N Engl J Med 2005; 352: Diaz A, Bourassa MG, Guertin MC, Tardif JC. Long-term prognostic value of resting heart rate in patients with suspected or proven coronary artery disease. Eur Heart J 2005;26: Bønaa KH, Arnesen E. Association between heart rate and atherogenic blood lipid fractions in a population. Circulation 1992;86: Prospective Studies Collaboration. Blood cholesterol and vascular mortality by age, sex, and blood pressure: a meta-analysis of individual data from 61 prospective studies with vascular deaths. Lancet 2007;370: Hackam DG, Anand SS. Emerging risk factors for atherosclerotic vascular disease. A critical review of the evidence. J Am Med Assoc 2003;290: Williams PT, Haskell WL, Vranizan KM, Blair SN, Krauss RM, Superko HR, Albers JJ, Frey-Hewitt B, Wood PD. Associations of resting heart rate with concentrations of lipoprotein subfractions in sedentary men. Circulation 1985;71: