Office and out of office heart rate measurements: which clinical value?

Clinical value of heart rate measured in and outside a clinic setting A large number of epidemiologic studies have shown that elevated resting heart rate (HR) is associated with a higher risk of developing hypertension and atherosclerosis and is a potent predictor of cardiovascular morbidity and mortality [1–6]. In some of these studies, the predictive power of HR for mortality was equal to or even greater than that of total cholesterol, smoking or systolic blood pressure (BP) [2,5]. The predictive power of HR for cardiovascular mortality was evident not only in general populations but also among individuals affected by myocardial infarction (MI), chronic coronary heart disease, diabetes, arterial hypertension or heart failure [7]. On the background of this data, resting HR obtained in the clinical setting has become an accepted component of the cardiovascular risk profile assessment, and routine measurement of this variable has been recently included in screening strategies for the prevention of cardiovascular diseases [8]. In most epidemiologic studies, resting HR was assessed by one or a few measurements taken in the medical environment. However, it is known that HR is a highly variable physiologic parameter, which is subjected to the influence of many environmental stimuli, including those typical of a clinic setting [9]. Several studies have shown that the medical environment elicits a rise in BP, leading office measurements to overestimate the usual BP of patients because of a ‘white-coat’ effect [10,11]. To overcome this problem, strategies for assessing BP out of the office have been devised, including both home and ambulatory BP monitoring. A rise during the medical visit has also been observed for resting HR, the patients' reactions varying as a result of whether the measurement was carried out by a doctor or a nurse [10,11]. Thus, a few readings taken in a clinic environment might be poorly representative of the actual HR levels characterizing a given patient in his/her everyday life, leading to an underestimation of their real predictive power. A better estimate of patients' usual HR could be derived from measurements obtained outside the medical environment, as clearly demonstrated for BP values. Indeed, there is increasing evidence that BP measured at home or through 24-h ambulatory recordings is a better predictor of cardiovascular disease than isolated clinic BP readings [12]. This is partly attributed to the absence of a ‘white-coat’ phenomenon and to the superior reproducibility of out-of-clinic BP measurements as compared with the clinic ones. The same advantages could be attributed to out-of-clinic HR measurements. Indeed, a better reproducibility of ambulatory HR measurements than of clinic ones has also been reported. This could suggest that, as in the case of BP, HR measured over the 24 h is also a better prognostic indicator than conventional HR measurements obtained in the hospital or clinic setting [13]. Unfortunately, little data exist on the prognostic value of HR measured out of a clinic or office setting. The contribution of the Ohasama study on the background of previous observations The study by Hozawa et al.[14] published in this issue of the Journal provides new important information in this regard. In this study, the authors prospectively investigated the association of baseline ambulatory HR levels and modulation with subsequent mortality. This report was based on a 12-year longitudinal observation of individuals who had been participating in an ambulatory BP and HR measurement project in Ohasama, a rural town in Japan. The study population consisted of 1444 individuals of more than 40 years of age, without cardiovascular diseases or arrhythmias. The most common cause of death among the residents of this town was cancer, followed by cerebrovascular and heart diseases. In this general Japanese population, average daytime and night-time HRs proved to be independent predictors of noncardiovascular but not of cardiovascular mortality. This data is of interest, although not completely in line with the results of previous studies. In fact, the above finding is at variance with the results by Aronow et al. [15] who found a significant association between ambulatory HR and cardiovascular mortality in 1311 elderly patients. In this study, the risk of death increased by 14% for each 5 bpm increase in the mean 24-h HR [15]. However, it should be pointed out that in the study by Aronow et al., patients with heart disease were enrolled, and that HR was measured through beat-by-beat ECG Holter recordings. In the Systolic Hypertension in Europe (Syst-Eur) study [16], in which 24-h ambulatory BP and HR were recorded through discontinuous measurements in a subset of 807 elderly hypertensive patients, the positive relationship between clinic HR and the incidence of fatal end points found in the main study was confirmed in the ambulatory monitoring subgroup. However, in the Syst-Eur study, ambulatory HR was also associated with non-cardiovascular mortality but not with cardiovascular mortality. Conversely, no association of mean ambulatory HR with either cardiovascular or non-cardiovascular mortality was found in the Progetto Ipertensione Umbria Monitoraggio Ambulatoriale (PIUMA) cohort study or in the Pressioni Arteriose Monitorate E Loro Associazioni (PAMELA) population study [12,17]. No conclusive explanation has yet been provided for the lack of association between ambulatory HR and cardiovascular mortality, which is strikingly at variance from the association of clinic HR with either all-cause or cardiovascular death or both observed in a large number of studies [1–7]. This might be related to the different nature of HR values measured in the clinic and outside the medical environment, as suggested by the results of previous studies [9]. In the HARVEST study, clinic HR was correlated with clinic BP and the rough measure of the white-coat effect represented by the difference between clinic BP and mean 24-h BP. This may indicate that changes in clinic HR contribute to the BP reaction to the doctor's visit [9,18]. Similar results were obtained in a Belgian population, in which clinic HR appeared to be related to clinic BP and the white-coat effect [9]. On the contrary, in both studies, clinic HR did not predict the level of 24-h BP, and HR recorded in ambulatory conditions was unrelated to ambulatory BP and the white-coat effect. Similar results were obtained in the Tecumseh study in which clinic and home BPs were measured [9]. In this study, clinic HR was related to clinic BP but showed little or no relationship with home BP [9]. Overall, these data indicate that HR and BP are related to one another when measured in the clinic under the somewhat stressful condition represented by the doctor's visit, whereas clinic HR is not predictive of BP measured outside the office or clinic. This supports the view that clinic HR may reflect the sympathetic activation associated with the alerting reaction induced by consultation in the office environment. Given that clinic HR has been shown to be a strong predictor of subsequent cardiovascular events, the common assumption that the haemodynamic reaction caused by a doctor's visit is a totally harmless phenomenon might therefore be erroneous. Moreover, average ambulatory HR is less influenced by any individual environmental stimulus and broadly reflects the overall HR ‘burden’ exerted on the cardiovascular system over the 24 h, including the effects of night sleep. However, it has to be pointed out that 24-h ambulatory HR is affected by a number of confounders related to the different daily life behaviours, as exemplified by the degree of physical activity habits that may greatly differ from individual to individual. On such a background, the better relationship of clinic HR rather than of ambulatory HR with cardiovascular complications might thus reflect its greater ability to more specifically reflect an underlying sympathetic overactivity. In light of these considerations, the strong relationship between HR measured at home and cardiovascular mortality previously found by the authors of the present study in the Ohasama population is difficult to explain. In 1780 participants, Hozawa et al. found a 17% increase in the risk of mortality for a 5 bpm increase in home HR [18]. Whether ambulatory and home measurements can be regarded as mutually interchangeable, or whether they provide complementary, non-superimposable information is still a matter of debate [19]. The ambulatory and home settings are different in as much that ambulatory monitoring provides measurements in many different conditions, at home, at work and during sleep, whereas self-measured HR provides measurements in the sitting posture in a relaxed environment usually at well identified times of the day, most commonly in the morning and the evening [19]. Indeed, when ambulatory and home BP values were assessed in the same study [20], the correlation between the two sets of BP values was not found to be very high (0.70 or less). The same applied to the surrogate measure of the white-coat effect calculated by subtracting either value from office BP readings [20]. In the PAMELA study, it was shown that when considering ambulatory and home BP values together, each one provided additional prognostic information on top of the other [12]. The results of the Ohasama study suggest that this may also be the case for ambulatory versus home HR, but more data are needed to clarify this issue. The association between high ambulatory HR and noncardiovascular mortality found in the Ohasama and the Syst-Eur studies may be due to the occurrence of a chronic illness unrecognized at the time of baseline evaluation [14,16]. In such a case, an elevated HR would be an indicator of poor physical health or loss of myocardial contractile reserve. An impaired left ventricular contractility and pump function may be an early clinical finding in individuals who are still asymptomatic many years before the development of overt congestive heart failure [9]. This is more likely to occur in older individuals such as those enrolled in the two studies mentioned above (mean age, 61.2 years in the Ohasama study and 70.2 years in the Syst-Eur study) [14,16]. To rule out this possible pathogenetic connection, in epidemiologic analyses, deaths occurring during early years after baseline assessment should be eliminated [2,5]. However, this analysis was not possible in the Syst-Eur study [16] because the mean follow-up (2 years) was too short and was not performed by the authors of the present study either [14]. In the Syst-Eur study [16], ambulatory and out-of-clinic HR was a significant predictor of cancer mortality. To explain the association between HR and cancer mortality, several hypotheses can be put forward. Lifestyle factors such as smoking and alcohol are known risk factors for cancer and may affect HR as well [21]. Extensive previous literature reported the existence of an association of either stress or specific personality types or both with cancer [22]. However, in some studies, the positive relationship between HR and cancer mortality remained significant after taking into account other risk factors such as tobacco, alcohol consumption and physical activity [23]. Another possibility is that at the cellular and molecular levels, common elements such as growth factors could be associated with increased stress on the arterial wall and the evolution of cancer. Previous results suggest common mechanisms between hypertension and atherosclerosis on the one hand and hypertension and tumour formation on the other [23,24]. In agreement with previous results by Verdecchia et al.[17], in the Ohasama study, the individuals who subsequently died had a blunted reduction of ambulatory HR from day to night [14]. In the study by Verdecchia et al.[17], for each 10% less reduction in the HR from day to night, the relative risk of mortality was 1.30 [95% confidence interval (CI), 1.02–1.65]. Similarly, in the Ohasama study [14] a 10% increase in the day–night dip was associated with a reduction in mortality (hazard ratio, 0.85; 95% CI, 0.72–0.99), which was similar for cardiovascular and non-cardiovascular death. However, when night-time HR was included in the regression, the day–night dip was no longer accepted by the model indicating that night-time HR was the most important predictor of mortality, rather than the day–night difference in its values. During sleep, a physiological decline in sympathetic activity and an increase in vagal tone occur in normal conditions. Autonomic dysfunction, which is common in the old age, can alter the diurnal HR rhythm. This condition has been reported to be associated with an increased rate of cardiovascular events [25] and can account for the relationship between night-time HR, or its changes between day and night, and mortality found in the Ohasama as well as other studies [14,17]. Limitations of the study by Hozawa et al. The results of the study by Hozawa et al.[14], notwithstanding their potential interest, must be interpreted in the context of their limitations. First, as the authors themselves acknowledge, the lack of clinic HR measurement prevented an exploration of respective roles of clinic and ambulatory HRs in predicting fatal events. Results from the Syst-Eur study indicate that mean ambulatory HR does not provide prognostic information on top of clinic HR [16], whereas the results by Verdecchia et al.[17] suggest that this is the case for the day–night HR dip. The Ohasama dataset would allow, however, for a comparison between ambulatory and home HR values. Previous results from Ohasama suggest that home HR is a better predictor of cardiovascular mortality than ambulatory HR [18], and a direct comparison between the HR levels recorded at home and in ambulatory conditions over 24 h could help in clarifying this issue. Second, in the Ohasama study, individuals with low HR were much older and were more frequently treated than the rest of the cohort [14]. The authors of the present study tried to account for the effect of several confounders in the regression analyses making use of statistical corrections, but it is possible that the between-group imbalance in several relevant factors might not have been completely corrected simply through statistical adjustments. In addition, the confounding effect of important variables such as physical activity could not be tested. Third, patients with coronary heart disease or heart failure are frequently treated with beta-blockers that are known to affect the HR levels and their day–night changes, which, in turn, might affect outcomes. The improvement in the association between ambulatory HR and mortality after elimination of treated patients from the regression suggests that this may also be the case in the Ohasama study [14]. In fact, no relationship between ambulatory HR and mortality was observed in the treated arm of the Syst-Eur study [16], further indicating that pharmacological treatment can obscure the relationship between baseline HR and outcome. As a concluding comment, it has to be emphasized that the Ohasama dataset would allow for additional important analyses to be carried out. As an example, the predictive power of HR for cardiovascular and total mortality has been found to be much greater in male than in female subjects, and in several studies, no independent association of HR with outcome was observed among women [2,7,9]. The lack of association between HR and cardiovascular death found in the Ohasama cohort might be partly due to a dilution effect caused by the preponderance of female participants within this cohort (>64%). A sex-specific analysis could clarify whether an association between HR and cardiovascular mortality was present at least within the male subgroup. Moreover, an issue that is still controversial is whether changes in HR during the follow-up may influence prognosis. According to recent results from the International Verapamil-SR/trandolapril study [26], the decline in HR caused by the treatment with either atenolol or verapamil improved patients' outcome. Thus, it would be interesting to know whether spontaneous or drug-induced changes in HR during follow-up could affect prognosis among the Ohasama participants. Finally, a unique advantage of ambulatory HR monitoring that should not be disregarded is the possibility it offers to quantify not only HR mean values but also HR variability parameters, which in several studies have been shown to have a clinical value. The assessment of HR variability, often accompanied by the concomitant assessment of BP variability, is indeed increasingly used to evaluate the dynamic features of cardiovascular control mechanisms and assess cardiovascular risk more precisely. In spite of the discontinuous nature of ambulatory HR measurements obtained in the Ohasama study, some information on HR variability could be usefully provided. Importance of heart rate variability analysis Both experimental and clinical studies have shown that HR variability analysis may offer a unique insight into autonomic cardiovascular modulation in health and disease [27–31], both in a laboratory environment and under daily life conditions [32,33]. The widespread utilization of this approach has been greatly facilitated by technological progress in the methods for non-invasive beat-by-beat recording of cardiovascular signals [34,35]. It has also been facilitated by the development of powerful data analysis procedures. Among the latter, spectral analysis techniques may offer a detailed quantification of the various frequency components, characterizing BP and HR variability, which have been reported to reflect the patterns of autonomic cardiovascular regulation both in normal and in pathological conditions [29]. An additional and more complex approach is based on the simultaneous quantification of fluctuations in several biological signals and the evaluation of their mutual interactions, sometimes with the help of mathematical models [29]. A typical example of these advanced methods is represented by the techniques, developed both in the frequency and in the time domain, for the assessment of the interactions between BP and HR fluctuations. These techniques allow the quantification of ‘spontaneous’ sensitivity of baroreflex control of HR, with no need for external interventions such as injection of vasoactive drugs or application of a neck chamber device, and are now increasingly being used both in pathophysiological studies and in the diagnostic and prognostic approach to a number of diseases [36–41]. A number of studies have indeed demonstrated that parameters quantifying HR variability are powerful predictors of cardiovascular events in cardiac patients, namely after an acute MI or stroke and in chronic heart failure [42–44]. The past two decades have provided a clear evidence for the tight and causal relationship existing between arrhythmic mortality and the autonomic nervous system, particularly with imbalances characterized either by a decrease in vagal tone or an increase in sympathetic activity or both. The possibility to use the analysis of baroreflex sensitivity as a means to quantify sympathovagal modulation at the sinus node level and the critical role played by abnormalities in the arterial baroreflex function in arrhythmia genesis have made this methodology a useful tool for diagnostic and prognostic evaluation of cardiovascular disease. Data obtained by either time and frequency domain methods for assessing spontaneous baroreflex sensitivity have already provided a large body of information on daily life baroreflex function in healthy individuals and its derangement in disease. This information consists of detailed evidence that in a number of pathological conditions, not only the average spontaneous baroreflex sensitivity computed over a given recording period is reduced, in agreement with the results of laboratory studies, but the baroreflex impairment typical of these conditions also involves its dynamic modulation of cardiac This is exemplified by the finding that the physiological increase in baroreflex sensitivity found in healthy individuals is blunted in conditions known to be affected by baroreflex impairment such as and arterial hypertension or In pathological conditions characterized by an autonomic dysfunction, spontaneous baroreflex sensitivity is clearly This is the case in patients with autonomic failure and patients with in baroreflex sensitivity is impaired both during and the various and this impairment is associated with the degree of arterial information on the actual prognostic value of spontaneous baroreflex sensitivity assessment is also In fact, baroreflex sensitivity was predictive of mortality in patients with recent stroke Whether a of information obtained with of spontaneous baroreflex sensitivity and modulation of baroreflex sensitivity might further the clinical evaluation of such patients an The must from longitudinal studies. These studies should also further whether laboratory may be in specific clinical conditions, such as in elderly patients with diabetes, often characterized by the of an number of baroreflex in the recordings due to the decrease in HR control with or in patients with and which the HR variability and baroreflex sensitivity analysis This issue is highly because of and pathophysiological Finally, whether HR variability and baroreflex sensitivity analysis might a prognostic value also in individuals without evidence of cardiovascular disease at a population level also to be investigated in longitudinal outcome studies. The present on the clinical value of HR and of its variability measured out of the clinic is The lack of association between HR recorded with ambulatory techniques and cardiovascular mortality found by Hozawa et al.[14] in the present study to indicate that the mean ambulatory HR level and its diurnal profile are of little value for cardiovascular risk and patient Conversely, the clinical of resting HR measured in the clinic has been demonstrated in a large number of studies and the prognostic provided by this clinical variable does not to be improved by the ambulatory However, only studies in which clinic HR, home HR and ambulatory HR are be to clarify the respective role of these in predicting the cardiovascular This is the case also for the predictive power of beat-by-beat HR variability and spontaneous baroreflex sensitivity parameters, for which the evidence to be further by additional outcome are no of