Recommendations on how to measure resting heart rate

By P. Palatini , Italy
Department of Clinical and Experimental Medicine – University of Padova
Padua, ITALY

Resting heart rate is an easily measurable cardiovascular parameter, but is subject to high variability. Studies focusing on heart rate should take into account all possible sources of variability, including the resting period before measurement, environmental conditions, method of measurement (pulse palpation versus electrocardiogram), number of readings, duration of measurement, position of the body, and nature of the observer. To minimize the effects of these confounding factors, the measurement of this clinical variable should be strictly standardized. Exercise, alcohol, nicotine, and coffee should be avoided in the hours preceding measurement. Readings should preferably be taken by pulse palpation while the patient is comfortably seated in a chair with legs uncrossed. The room should be at a comfortable temperature and background noises should be avoided. The patient should refrain from talking during the procedure, and at least 5 minutes should elapse before the first reading is taken. Little is known about the predictive value of out-of-office heart rate measurement. Available data indicate that heart rate recorded with ambulatory monitoring or self-measured at home provides little or no additional prognostic information to heart rate measured in the clinic. However, heart rate recorded by self measurement does not involve any additional cost, thus for hypertensive subjects who undergo self blood pressure measurement, reporting of heart rate together with blood pressure data may provide useful information.

Medicographia. 2009;31:414-419 (see French abstract on page 419)

High heart rate has been shown to be an independent risk factor for all-cause and cardiovascular death in general population studies.1-3 Elevated heart rate has also been shown to provide important prognostic information in subjects with hypertension, diabetes, and coronary artery disease.3-5 Resting heart rate is an easily measurable cardiovascular parameter. However, it is subject to high variability, and this may have led to an underestimation of its impact on cardiovascular morbidity and mortality. Sources of variability include physical factors, psychic stimuli, environmental factors, body position, and methods of measurement.1,6 To minimize the effects of these confounding factors, the measurement of this clinical variable should be strictly standardized. In most published studies, information is missing on how heart rate was measured, even when heart rate was one of the major variables to measure.6 This prevents the reader from fully understanding the methodology of heart rate determination and from being able to compare the results coming from different laboratories. Studies focusing on heart rate should take into account all possible causes of variability, and in particular, the resting period before measurement, the position of the body, the environmental conditions, the method of heart rate recording, and the statistical approach to data analysis.2,6

The aim of this article is to focus upon the most important methodological problems in the assessment of heart rate, and to provide recommendations for a standardized method of measurement.

Office and out-of-office measurement

Most information on the prognostic significance of heart rate has been obtained from studies that used heart rate measured under resting conditions. However, as mentioned above, resting heart rate is subject to high variability, and may be influenced by the nature of the observer. In fact, Mancia et al showed that the actual visit to a doctor caused a heart rate increase of as high as 45 beats per minute (bpm), with a mean increase of 16 bpm.7 This suggests that heart rate measured out of the office, either with ambulatory measurement or self assessment, might be more representative of a subject’s usual heart rate. A study in 839 hypertensive subjects showed that the reproducibility of heart rate recorded twice 3 months apart was better for ambulatory than office measurement.8 Reproducibility of office heart rate was particularly poor when it was above the level of 85 bpm. The above data indicate that heart rate recorded with ambulatory monitoring has better reproducibility than office heart rate, and could, thus, have a stronger association with outcome than resting heart rate measured in the office. However, only a few studies have examined the association between ambulatory heart rate and total or cardiovascular mortality. Recent results from the Ohasama study showed that neither daytime nor nighttime heart rate predicted cardiovascular disease mortality, whereas both could predict noncardiovascular disease mortality.9 The lack of association between heart rate and cardiovascular death found in the Ohasama cohort might be due in part to a dilution effect caused by the preponderance of female participants within this cohort (>64%).

In a recent analysis of six study populations, 24-hour heart rate predicted total and noncardiovascular mortality, but not cardiovascular mortality or any of the combined fatal and nonfatal events.10 The above data are in agreement with previous results obtained in the Syst-Eur cohort,11 and indicate that heart rate measured with ambulatory recording is of little use for stratifying cardiovascular risk. Even less is known on the predictive value of self heart rate measurement. Data from the Ohasama study showed a 17% increase in the risk of mortality for a 5-bpm increase in home heart rate,12 whereas in the PAMELA study (Pressioni Arteriose Monitorate E Loro Associazioni), no association was found between home heart rate and mortality.13 However, heart rate recorded with self blood pressure measurement does not imply any additional cost, and thus it is the opinion of the experts that the reporting of heart rate together with blood pressure data in hypertensive patients who undergo self blood pressure measurement may provide useful information.2

Table. Procedures for heart rate measurement.

After reference 2: Palatini P, Benetos A, Grassi G, et al. J Hypertens. 2006;24: 603-610. Copyright © 2006, Lippincott Williams, Wilkins, Inc.

Methods of measurement

_ Measurement of heart rate at rest
According to the Consensus Panel of the European Society of Hypertension,2 the following information should be provided in studies reporting heart rate data: (i) resting period before measurement; (ii) environmental conditions; (iii) method of measurement; (iv) number of measurements; (v) duration of measurement; (vi) body position; and (vii) nature of the observer. The Table shows the recommendations provided by the panel with regard to how heart rate should be measured.

_ Duration of measurement
To achieve a stable hemodynamic condition, the patient should rest for at least 5 minutes, although in subjects with a pronounced white-coat reaction, a longer period may be necessary. The duration of measurement ranges from 15 seconds to 1 minute in different studies.1,2,6 According to the aforementioned European consensus panel,2 30 seconds seem to be sufficient to obtain a reliable estimate of heart rate, because in most patients, 30 to 40 cardiac cycles can be averaged out. In subjects with a very low heart rate, a longer period may be necessary.

_ Number of measurements
It is know that heart rate tends to fall when measured repeatedly; this is because patients need time to adjust to the doctor’s office setting. This decrease is rather small, however, as shown by the results of the Hypertension and Ambulatory Recording VEnetia STudy (HARVEST) and the CArdiovascular STudy in the ELderly (CASTEL).14,15 In HARVEST, for instance, heart rate decreased by approximately 2 bpm over a period of 2 months.14 However, temporal changes in heart rate remarkably varied among individuals. It would seem desirable, therefore, to obtain an adequate number of measurements before making a diagnosis—especially in patients whose heart rate shows a substantial decline from the first to the second reading. However, in CASTEL, single baseline measurements all had great predictive value for cardiovascular mortality, and the mean of three measurements increased the predictive power only marginally.15 Thus, two measurements seem to be sufficient for a reliable estimate of resting heart rate in most patients. It should be kept in mind that the extent to which heart rate increases during the alarm reaction to the measurement largely depends on the nature of the observer, being higher if heart rate is measured by a doctor, intermediate if it is measured by a nurse, and lower if it is acquired with an automatic device in the absence of an observer.2,16 Thus, when comparing data from different laboratories, it is of paramount importance to know the nature of the observer.

Figure. Posture (A), resting period before measurement (B), and method of measurement (C) among publications detailing relevant information on the assessment of resting heart rate.

ECG, electrocardiogram.
After reference 6: Vogel CU, Wolpert C, Wehling M. Eur J Clin Pharmacol. 2004;60:461-466. Copyright © 2004, Springer-Verlag.

_ Body position
The choice of body position during heart rate measurement is also a source of controversy in the literature, where roughly half of epidemiological studies report using sitting heart rate, and the other half supine heart rate.1,2 However, in clinical studies, the supine position has been more frequently employed (Figure). There are no objective data to suggest that one position is better than the other. In the sitting position, one should expect a heart rate that is 1-2 bpm higher than in the lying position. In experimental studies requiring a long period of rest in order to achieve a stable hemodynamic condition, the supine position should be preferred. However, the sitting position is preferable for epidemiological studies or clinical routine assessment in which heart rate can be measured at the end of each blood pressure measurement. Heart rate may also provide useful clinical information when measured in the standing position, especially in elderly hypertensive patients who frequently have a drop in blood pressure while standing up, and in young hypertensive individuals who may be hyper-reactive to standing.17

_ Measurement of heart rate during exercise
As many studies have consistently documented that measurement of heart rate during exercise and recovery is useful for identifying subjects at increased cardiovascular risk, careful monitoring of heart rate is mandatory whenever a subject undergoes exercise testing.18-20 In asymptomatic men and women, as well as in symptomatic referral populations, an inability of the heart to increase its rate appropriately during incremental exercise has been shown to be of prognostic value independent of traditional risk factors, other markers of risk derived from exercise testing, or thallium ischemia, and to also be independently predictive of death in patients with cardiac diseases.18,19 The chronotropic measurements most commonly used in the literature include absolute heart rate during submaximal exercise, absolute peak of heart rate achieved, change in heart rate from rest to peak exercise, one standard deviation of mean peak heart rate achieved, 80% to 85% age-predicted target heart rate, and the socalled “chronotropic index,” which takes into account age, physical fitness, resting heart rate, and the age-predicted maximum heart rate.21 According to a recent report by Savonen and colleagues,19 the slope of the increase in heart rate during exercise on a bicycle ergometer (maximal heart rate minus heart rate at 40% workload) has greater prognostic power than other exercise-derived chronotropic measurements. The 100%-40% heart rate slope depends chiefly on the response of the sympathetic nervous system, and does not include the early portion of the slope, which mainly reflects the withdraw al of the vagal tone, suggesting that the main factor mediating the association between chronotropic incompetence and mortality is a reduced ability to increase sympathetic activity. It has not been well established as to whether the prognostic information provided by heart rate assessment during exercise can provide complementary information to that provided by heart rate measured under resting conditions. However, whenever exercise is performed for diagnostic purposes with either a bicycle or a treadmill ergometer, measurement of heart rate throughout the exercise and recovery period is highly recommended.

Techniques for heart rate measurement

It is not known whether it is better to measure heart rate with an electrocardiogram or rather to take the pulse rate. The electrocardiogram would appear to be more accurate, but the number of cardiac cycles used for heart rate calculation is usually quite small. As mentioned, in most studies providing heart rate data, information on the method of measurement is missing.6 Whereas in epidemiological studies roughly 50% of the measurements are obtained by pulse palpation and 50% by electrocardiographic recording,1 in the majority of clinical studies, the latter technique has been used (Figure). Recently, electronic pulse meters were made available for measuring heart rate automatically from the finger, wrist, or chest.22 These devices allow measurements to be taken continuously, and can be used during exercise when manual measurement would be difficult or impossible.

_ Pulse palpation
Traditionally, heart rate has been measured by pulse palpation. The pulse rate is measured by counting the beats in a set period of time (from 15 to 60 seconds) and multiplying that number to get the number of bpm. This is still the method currently used by doctors and other healthcare professionals in daily routine. The pulse rate can be measured at any point on the body where an artery is close to the surface. Most common places are radial, carotid, brachial, and femoral arteries. If stroke volume is subject to high variability such as in cases of atrial fibrillation, some heart beats can be missed at pulse palpation, and heart rate should be measured directly from heart auscultation.

_ Electrocardiography
There are, however, techniques that allow heart rate to be measured more precisely and for longer periods of time. Electrocardiographic recording is the most precise method of heart rate measurement and is routinely carried out in many clinical settings, especially in critical care medicine. Whether electrocardiographic measurement may also be advantageous in epidemiological studies or in clinical routine is not known, however. The use of electrocardiography obviously implies greater financial costs, and it is not known whether increased measurement precision actually translates into more meaningful data. According to Erikssen et al, the two measurements are highly correlated (R>0.9) and provide similar information.23 For this reason, electrocardiographic measurement is not recommended for the measurement of resting heart rate, even in research.

_ Electronic devices
Electronic pulse meters consist of two parts, a transmitter placed over the artery and a receiver attached to a belt worn around the chest or a wrist watch receiver for display. A photo diode or a photo transistor can be used to detect pulse rate. The skin may be illuminated with visible (red) or infrared light emitting diodes using transmitted or reflected light for detection. Infrared sensors can easily be clipped to finger ends or ear lobes to detect the heart beat using plethysmographic technology.22 Because of frequent noise sources that may produce disturbance signals, valid pulse measurement requires extensive preprocessing of the raw signal. New systems combine analog and digital signal processing to suppress disturbance signals. A digital system is usually accurate to within 3-4 bpm. Simple heart rate monitors may only display the heart rate on the screen. More professional monitors are available that can be set to record time, calculate average and maximum heart rate for a given period, and sound an alarm when a person reaches or exceeds a predetermined target heart-rate zone. These devices are used mostly by athletes and sportspeople wishing to monitor their workouts in order to achieve their desired training benefit. More complex ambulatory devices can also record other biological signals such as breathing movements, nasal and oral flow, and blood oxygen saturation, which can be useful for monitoring sleep apnea episodes.24

Recommendations before measurement

These recommendations are roughly the same as those used for blood pressure measurement. As mentioned, several physical or psychological factors can influence the assessment of heart rate during the office visit.1-3 In particular, exercise, alcohol, nicotine, and coffee can influence the heart rate and should be avoided in the hours preceding measurement. Readings should preferably be taken while the patient is seated in a chair. The patient should be comfortably seated, with legs uncrossed, and examined in a room with a comfortable temperature, avoiding any background noise such as telephones or beepers. The patient should refrain from talking during the procedure, and at least 5 minutes should elapse before the first reading is taken. Although heart rate should be measured in any subject under medical investigation, this hemodynamic variable is usually assessed in patients with hypertension or cardiac disease. Most of these patients are receiving pharmacological therapy, and the doctor should be aware that many cardiovascular drugs can either decrease or increase the heart rate.

The bradycardic effect is particularly pronounced for two classes of drugs, â-blockers and inhibitors of the so-called “funny” current in the sinus node pacemaker cells.25-27 The best known agent in this class, ivabradine, produces a dosedependent reduction in heart rate both at rest and during exercise, which is of the same magnitude as that seen with β-blockers.27 Nondihydropyridine calcium antagonists such as the phenylalkylamines and benzothiazepines also have a bradycardic action, although this is smaller than that seen with â-blockers or If current inhibitors.28 The centrally acting antihypertensive drugs guanfacine, clonidine, and methyldopa have a bradycardic effect of similar magnitude to that of the nondihydropyridine calcium antagonists.29 By contrast, the central I1-imidazoline receptor agonists moxonidine and rilmenidine have a negligible effect on heart rate.29 A slight bradycardic action has been described for the angiotensin II type 1 receptor blocking agents, but this might be due to an adaptation process that takes place during the observation period rather than to a true bradycardic effect.30,31 A cardiac slowing action is also often observed with some anti-arrhythmic drugs such as amiodarone or class Ic antiarrhythmic agents.

Some classes of antihypertensive agents can trigger a reflex increase in heart rate due to the sympathetic activation that occurs in response to the fall in blood pressure. Vasodilators such as minoxidil or hydralazine can produce remarkable increases in heart rate.28 The effect of nitrates on the arteriolar circulation is smaller, and a tachycardic effect of these drugs is less frequently observed. An increase in heart rate chiefly during acute administration has also been observed with α-blockers such as prazosin or doxazosin, or with urapidil, a drug combining peripheral α-adrenergic blockade with a central effect.32 However, the heart rate increase is somewhat less pronounced with urapidil than with pure á-blockers. An increase in heart rate can also be seen with the short-acting dihydropyridine calcium antagonists, but rarely with the longacting ones.28

Who should be considered at risk?

Under resting conditions, the adult human heart beats at about 70 to 75 bpm, and the heart rate tends to decrease with age. Women generally have a 3- to 7-bpm higher heart rate than men.33 The normal limits of resting heart rate are nominally between 60 and 100 bpm. However, the results of most epidemiologic studies indicate that this normality interval cannot be applied to heart rate when considering it as a cardiovascular risk factor.1-3 In fact in many studies, heart rates higher than 80-85 bpm have been shown to imply a considerable increase in risk. Conversely, heart rates lower than 60 bpm have been shown to confer a protective effect against cardiovascular disease. It is therefore evident that the 60- to 100-bpm range can no longer be considered as the normal interval for resting heart rate. Most epidemiological studies have shown that there is a considerable increase in risk with heart rates higher than 80-85 bpm,1-3 and thus the upper normal limit for this clinical variable could be— albeit arbitrarily—set at this level.34

Similar results have been obtained by our group using an objective method for identifying the partition level between normal and high heart rates.35 Using mixture analysis36 in several general or hypertensive populations, we found that the cutoff between people with a normal heart rate and those with a high heart rate was between 80 and 85 bpm.35 According to some authors, the upper limit of normal of a clinical variable should be defined as the level at which the benefits of therapeutic intervention exceed potential risks.37 β-Blockers have been found to be beneficial among patients with heart failure or myocardial infarction only if heart rate is higher than 80-85 bpm.25,38

Based on this approach, the upper limit of normal should be set—at least in heart failure and post–myocardial infarction patients—at a heart rate of 80-85 bpm with a therapeutic target of 60 bpm or lower. The recent study BEAUTIFUL (mor- Bidity-mortality EvAlUaTion of the If inhibitor ivabradine in patients with coronary disease and left ventricULar dysfunction) added substantially to current knowledge concerning the threshold at which risk increases in coronary patients taking β-blockers.39 Patients in the placebo group (most of whom took β-blockers) with a baseline resting heart rate of 70 bpm or more had an increased risk for all cardiovascular outcomes after adjustment for other predictors of outcome.

These data indicate that special attention should be given to this high risk group with a resting heart rate above 70 bpm, who are at a high risk of cardiovascular events. _

1. Palatini P, Julius S. Heart rate and the cardiovascular risk. J Hypertens. 1997; 15:3-17.
2. Palatini P, Benetos A, Grassi G, et al. Identification and management of the hypertensive patient with elevated heart rate: statement of a European Society of Hypertension Consensus Meeting. J Hypertens. 2006;24:603-610.
3. Fox K, Borer JS, Camm AJ, et al. Resting heart rate in cardiovascular disease. J Am Coll Cardiol. 2007;50:823-830.
4. Kolloch R, Legler UF, Champion A, et al. Impact of resting heart rate on outcomes in hypertensive patients with coronary artery disease: findings from the INternational VErapamil-SR/trandolapril STudy (INVEST). Eur Heart J. 2008;29: 1327-1334.
5. Stettler C, Bearth A, Allemann S, et al. QTc interval and resting heart rate as long-term predictors of mortality in type 1 and type 2 diabetes mellitus: a 23-year follow-up. Diabetologia. 2007;50:186-194.
6. Vogel CU, Wolpert C, Wehling M. How to measure heart rate? Eur J Clin Pharmacol. 2004;60:461-466.
7. Mancia G, Bertinieri G, Grassi G, et al. Effects of blood pressure measurement by the doctor on patient’s blood pressure and heart rate. Lancet. 1983;2:695- 697.
8. Palatini P,Winnicki M, Santonastaso M, et al. Reproducibility of heart rate measured in the clinic and with 24-hour intermittent recorders. Am J Hypertens. 2000;13:92-98.
9. Hozawa A, Inoue R, Ohkubo T, et al. Predictive value of ambulatory heart rate in the Japanese general population: the Ohasama study. J Hypertens. 2008;26: 1571-1576.
10. Hansen TW, Thijs L, Boggia J, et al. Prognostic value of ambulatory heart rate re visited in 6928 subjects from 6 populations. Hypertension. 2008;52:229-235.
11. Palatini P, Thijs L, Staessen JA, et al. Predictive value of clinic and ambulatory heart rate for mortality in elderly subjects with systolic hypertension. Arch Intern Med. 2002;162:2313-2321.
12. Hozawa A, Ohkubo T, Kikuya M, et al. Prognostic value of home heart rate for cardiovascular mortality in the general population: the Ohasama study. Am J Hypertens. 2004;17:1005-1010.
13. Sega R, Facchetti R, Bombelli M, et al. Prognostic value of ambulatory and home blood pressures compared with office blood pressure in the general population: follow-up results from the Pressioni Arteriose Monitorate e Loro Associazioni (PAMELA) study. Circulation. 2005;111:1777-1783.
14. Palatini P, Dorigatti F, Zaetta V, et al. Heart rate as a predictor of development of sustained hypertension in subjects screened for stage 1 hypertension: the HARVEST Study. J Hypertens. 2006;24:1873-1880.
15. Palatini P, Casiglia E, Julius S, Pessina AC. High heart rate: a risk factor for cardiovascular death in elderly men. Arch Intern Med. 1999;159:585-592.
16. Mancia G, Parati G, Pomidossi G, Grassi G, Casadei R, Zanchetti A. Alerting reaction and rise in blood pressure during measurement by physician and nurse. Hypertension. 1987;9:209-215.
17. Vriz O, Soon G, Lu H, Weder AB, Canali C, Palatini P. Does orthostatic testing have any role in the evaluation of the young subject with mild hypertension?: an insight from the HARVEST study. Am J Hypertens. 1997;10:546-551.
18. Lauer M, Froelicher ES, Williams M, Kligfield P. Exercise testing in asymptomatic adults: a statement for professionals from the American Heart Association Council on Clinical Cardiology, Subcommittee on Exercise, Cardiac Rehabilitation, and Prevention. Circulation. 2005;112:771-776.
19. Savonen KP, Lakka TA, Laukkanen JA, et al. Heart rate response during exercise test and cardiovascular mortality in middle-aged men. Eur Heart J. 2006;27: 582-588.
20. Cole CR, Blackstone EH, Pashkow FJ, Snader CE, Lauer MS. Heart-rate recovery immediately after exercise as a predictor of mortality. N Engl J Med. 1999; 341:1351-1357.
21. Palatini P. Exercise testing in asymptomatic subjects: from diagnostic test to prognostic tool? Eur Heart J. 2008;29:1803-1806.
22. Wilhelm FH, Roth WT, Sackner MA. The lifeShirt. An advanced system for ambulatory measurement of respiratory and cardiac function. Behav Modif. 2003; 27:671-691.
23. Erikssen J, Rodahl K. Resting heart rate in apparently healthy middle-aged men. Eur J Appl Physiol. 1979;42:61-69.
24. Tiihonen P, Pääkkönen A, Mervaala E, Hukkanen T, Töyräs J. Design, construction and evaluation of an ambulatory device for screening of sleep apnea. Med Biol Eng Comput. 2009;47:59-66.
25. Kjekshus JK. Importance of heart rate in determining beta-blocker efficacy in acute and long-term acute myocardial infarction intervention trials. Am J Cardiol. 1986;57:43F-49F.
26. Bangalore S, Sawhney S, Messerli FH. Relation of beta-blocker-induced heart rate lowering and cardioprotection in hypertension. J Am Coll Cardiol. 2008; 52:1482-1489.
27. Di Francesco D, Camm JA. Heart rate lowering by specific and selective I(f) current inhibition with ivabradine: a new therapeutic perspective in cardiovascular disease. Drugs. 2004;64:1757-1765.
28. Palatini P, Benetos A, Julius S. Impact of increased heart rate on clinical outcomes in hypertension: implications for antihypertensive drug therapy. Drugs. 2006;66:133-144.
29. Van Zwieten PA. Centrally acting antihypertensives: a renaissance of interest. Mechanisms and haemodynamics. J Hypertens. 1997;15(suppl 1):S3-S8.
30. Kaul CL, Ramarao P. Renin release and the sympathetic nervous system. Drugs Today (Barc). 2000;36:699-713.
31. Palatini P, Mugellini A, Spagnuolo V, et al. Comparison of the effects on 24-h ambulatory blood pressure of valsartan and amlodipine, alone or in combination with a low-dose diuretic, in elderly patients with isolated systolic hypertension (Val-syst Study). Blood Press Monit. 2004;9:91-97.
32. Grassi G, Seravalle G, Mancia G, Zanchetti A. Hemodynamic and neural effects of urapidil and prazosin in essential hypertensives. Blood Press Suppl. 1994; 4:13-18.
33. Palatini P. Heart rate as a cardiovascular risk factor: do women differ from men? Ann Med. 2001;33:213-221.
34. Palatini P. Need for a revision of the normal limits of resting heart rate. Hypertension. 1999;33:622-625.
35. Palatini P, Casiglia E, Pauletto P, Staessen J, Kaciroti N, Julius S. Relationship of tachycardia with high blood pressure and metabolic abnormalities. A study with mixture analysis in three populations. Hypertension. 1997;30:1267-1273.
36. Schork NJ, Schork MA. Skewness and mixture of normal distributions. Comm Stat Theoret Methods. 1988;17:3951-3969.
37. Pickering TG. Modern definitions and clinical expressions of hypertension. In: Laragh JH, Brenner BM, eds. Hypertension. New York, NY: Raven Press; 1995: 17-21.
38. Cucherat M. Quantitative relationship between resting heart rate reduction and magnitude of clinical benefits in post-myocardial infarction: a meta-regression of randomized clinical trials. Eur Heart J. 2007;28:3012-3019.
39. Fox K, Ford I, Steg PG, Tendera M, Robertson M, Ferrari R. 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.