Older subjects show no age-related decrease in cardiac baroreceptor sensitivity

Transcription

1 Age and Ageing 1999; 28: Older subjects show no age-related decrease in cardiac baroreceptor sensitivity SUZANNE L. DAWSON, THOMPSON G. ROBINSON, JANE H. YOUDE, ALISON MARTIN, MARTIN A. JAMES, PHILIP J. WESTON, RONNEY B. PANERAI 1,JOHN F. POTTER 1999, British Geriatrics Society Department of Medicine for the Elderly, University of Leicester, The Glenfield Hospital, Groby Road, Leicester LE3 9QP, UK 1 Department of Medical Physics, University of Leicester, Leicester Royal Infirmary, Infirmary Square, Leicester LE1 5WW, UK Address correspondence to: S. L. Dawson. Fax: (þ44) Abstract Objective: to examine the relationship between age, blood pressure and cardiac baroreceptor sensitivity derived from spectral analysis, the Valsalva manoeuvre and impulse response function. Methods: we studied 70 healthy normotensive volunteers who were free from disease and not taking medication with cardiovascular or autonomic effects. We measured beat-to-beat arterial blood pressure and used standard surface electrocardiography to record pulse interval under standardized conditions with subjects resting supine as well as during three Valsalva manoeuvres. We performed single, multiple and stepwise regression of patient characteristics against cardiac baroreceptor sensitivity results. Results: there is a non-linear decline in cardiac baroreceptor sensitivity with advancing age, increasing systolic blood pressure and heart rate values (except for the Valsalva-derived result), but little further decline after the fourth decade. Only age significantly influenced values derived using the Valsalva manoeuvre and impulse response analysis. Using spectral analysis, age, systolic and diastolic blood pressure and heart rate influenced cardiac baroreceptor sensitivity, age contributing to 50% of the variability. Age also influenced the relationship between pulse interval and blood pressure, possibly indicating more non-baroreceptor-mediated changes with advancing age. Conclusions: although age is the dominant factor influencing cardiac baroreceptor sensitivity in this normotensive population, there is little change in mean values after 40 years of age. The differences in the relationship between pulse interval and blood pressure with advancing age have implications for the calculation of cardiac baroreceptor sensitivity using spectral analysis. Keywords: age, blood pressure, cardiac baroreflex sensitivity, impulse response function, phase response, regression analysis Introduction The cardiac-baroreceptor reflex arc is one of the many physiological mechanisms involved in the short-term control of arterial blood pressure. Researchers have examined factors influencing cardiac baroreceptor sensitivity (BRS) [1 5]. Gribbin et al. [1] were the first to identify an age-related decline; others have confirmed this [2 5], but few subjects over 60 years have been investigated. This is partly because of the invasive nature of traditional methods of measuring beat-to-beat changes in arterial blood pressure. Increasing age is associated with a rise in blood pressure. There is debate as to which factor influences BRS more. Gribbin et al. [1] felt that age and blood pressure acted independently and that increasing values were inversely associated with BRS. Conversely, work from our department, using pharmacological and non-invasive methods of measuring BRS, found that age contributed only 7% of the variance in cardiac BRS compared with 27 45% for systolic blood pressure (SBP) in subjects over 60 years [4]. Spectral methods of calculating cardiac BRS have several advantages over conventional invasive pharmacological methods and give similar BRS values [6]. 347

2 S. L. Dawson et al. Power spectral analysis involves the detection of rhythmicity in computer-derived tachograms of beatto-beat recordings of blood pressure and pulse interval (PI); various algorithms can then be used to assess the number, frequency and amplitude of the oscillatory components. However, this analysis requires the prior selection of frequency bands for blood pressure and PI. Previous studies favoured either the low-frequency band ( Hz) or the combined low- and highfrequency bands ( Hz) the a value [7 12]. Two other measures are important in spectral analysis: phase and coherence. Phase, in cardiac BRS settings, is the time relationship between changes in SBP and PI. A negative phase value implies that changes in SBP are leading changes in PI, reflecting a baroreceptor-mediated sequence. A positive phase implies PI leading SBP and reflects system noise or non-baroreceptor-mediated sequences. Coherence is a measure of input output coupling. It assesses the statistical significance between the dynamic change in arterial blood pressure and the resulting change in PI. It is similar to the correlation coefficient, with a range of 0 (no relation) to 1.0 (very strongly correlated). A value for squared coherence that is significantly <0.5 reflects too much signal noise or an output that is more dependent on variables other than blood pressure input [13, 14]. If there were to be a difference in phase with age and good coherence, this would give important information on baroreceptor activity with age. It may also help determine whether it is better to use low-, high- or combined frequency calculation of BRS. There are no previous reports of the influence of age on these factors (although there are reports of phase differences in the high-frequency bandwidth in patients with coronary heart disease [13]). Our aims were to assess individual and combined influences of age, blood pressure and other factors on the cardiac BRS measurements derived from spectral analysis, phase IV of the Valsalva manoeuvre and the newer impulse response function (IRF) technique [14]. IRF is similar to spectral analysis but examines data in the time domain. It reflects the dynamic change in an output (PI) in response to an input (SBP). It is thought to be less subject to problems of standardization, as no frequency bands need to be pre-selected, and it may lead to more information on BRS modulation. We have also examined the relation between blood pressure and PI in the various frequency bands used to calculate BRS by spectral methods to see if it was affected by age. Subjects We recruited from hospital staff, friends and relatives of patients attending outpatient departments, orthopaedic inpatients awaiting elective procedures and respondents to a newspaper advertisement calling for healthy volunteers. All subjects were independent in activities of daily living and free from cardiovascular disease (determined by history, examination, 12-lead electrocardiography and blood tests). They were not receiving drugs known to influence the cardiovascular or autonomic systems. We excluded those with a history of hypertension, diabetes mellitus or other diseases affecting the autonomic system, recent myocardial infarction, atrial fibrillation or cerebrovascular disease, and those with a mean of three clinic SBP readings >160 mmhg or diastolic blood pressure (DBP) phase V >95 mmhg. Methods All subjects were examined using a standardized protocol [15]. They were asked to attend in the morning after abstinence from caffeine, nicotine or alcohol-containing products for at least 12 h and at least 2 h after breakfast. The study was performed in a quiet laboratory kept at a constant temperature (20 24 C), with subdued lighting. Subjects wore loose comfortable clothing and were asked to micturate just before the study. Subjects lay supine on a couch, the head supported by two pillows. Standard surface electrocardiographic limb leads were attached for continuous monitoring. A Finapres cuff (Finapres NIBP Ohmeda System, CO, USA) [16 18] of appropriate size was placed around the right middle finger for beat-to-beat arterial blood pressure measurement. The arm was supported at atrial height on a specifically designed rest attached to the couch. Subjects were asked to breathe at a rate of 15 breaths per minute; if necessary, a metronome was used to aid this. After min when the readings had stabilized and subjects were accustomed to the setting, PI and blood pressure were recorded directly onto a dedicated microcomputer fitted with a 12-bit analogue digital converter with a sampling frequency of 200 samples/s per channel. The servo-adjust mechanism of the Finapres was disconnected and restarted between recording periods. We made three consecutive 5-min recordings while subjects lay quietly. They were asked not to talk. Each subject was then asked to sit up and with their arm supported at atrial height to perform three consecutive Valsalva manoeuvres (which had been practised beforehand) [19, 20]. They were asked to obtain a mouth pressure of 40 mmhg for 15 s. This pressure was visually displayed on a transducer to aid compliance. A constant bleed device was integral to the system. We recorded mouth pressure, heart rate and blood pressure changes throughout the procedure and for 1 min afterwards. When outputs had returned to baseline levels, the manoeuvre was repeated. Data analysis We analysed the data using specially designed software 348

3 Age-related decline in cardiac baroreceptor sensitivity [21]. First, data were calibrated from the Finapres recording at the start of each session and then edited. If the Valsalva manoeuvre was being analysed, then the pressure transducer recording was also calibrated. The PI was marked from the Finapres trace: this is less affected by external stimuli (e.g. muscle activity during the Valsalva manoeuvre) than the surface electrocardiography trace. There is no difference between BRS values calculated using PI derived from the electrocardiographm or blood pressure trace [22]. Spectral analysis in the frequency domain using fast Fourier transformation (FFT) and then the Valsalva-derived BRS were calculated [7 12, 19, 20]. In each case, we took the mean of the three recordings. Signals were low-pass filtered with an eighth-order Butterworth digital filter with a cut-off frequency of 20 Hz. Beat-to-beat sequences of SBP and PI were extracted and were linearly interpolated in the event of an ectopic beat. We rejected recordings of more than four ectopics. We employed FFT which used 512 samples; beat-to-beat changes were interpolated using a third-order polynomial and then resampled with a 0.5-s interval. The power spectra were averaged over three readings and smoothed with a 13-point triangular window. We calculated the baroreflex sensitivity index,, from the mean of the square roots of the ratios of the spectral powers of SBP and PI in the low-frequency ( Hz) and high-frequency ( Hz) bandwidths. We calculated the Valsalva-derived BRS using the linear regression of SBP and PI during phase IV of the Valsalva manoeuvre. IRF was calculated from the same three recordings. A transfer function was computed by dividing the cross spectrum by the power spectrum of SBP. We applied a cosine-shaped low-pass filter with a cut-off frequency of 0.5 Hz in the frequency domain and obtained the IRF with an inverse FFT [14]. The highest value obtained is termed the IRF peak ; by using this value and averaging it and the two neighbouring values the IRF smooth is obtained. We used this latter value as our measure of IRF [14]. Statistical analysis We analysed data using the Minitab 10 release software program. These are presented as mean SD after assessing for normality, using Shapiro Wilk normality plots. Where necessary we carried out log transformation to achieve normality; the use of logarithmic transformation was an arbitrary decision. For each BRS measurement (FFT, Valsalva-derived, IRF), we performed single regression, multiple linear regression and forward stepwise regression analyses using the different subject characteristics as the determinant parameters. Statistical significance was set at the 5% level. The influence of age on phase and coherence was examined by dividing the study population into Table 1. Subject characteristics: demographics and cardiac baroreceptor sensitivity calculated by three different methods Variable Mean (SD) Range... Age, years 52 (18) Body mass index, kg/m (3.6) Clinic blood pressure, mmhg Systolic 131 (16) Diastolic 76 (9) Pulse, bpm 69 (11) Cardiac baroreceptor sensitivity, ms/mmhg FFT 13.7 (8.3) Valsalva 5.8 (3.4) IRF 9.0 (7.1) n ¼ 70 (32 male). FFT, baroreceptor sensitivity derived by power spectral analysis using fast Fourier transformation; Valsalva, baroreceptor sensitivity derived from phase IV of the Valsalva manoeuvre; IRF, baroreceptor sensitivity derived from the peak-smoothed impulse response function. two groups: <45 years and >50 years. This age division was made post-analysis since most decline in BRS was seen in the third and fourth decades and data on phase and coherence were missing in the few subjects aged years. We calculated values for each in the low-frequency ( Hz) and highfrequency ( Hz) bandwidths and compared them both graphically and by parametric tests. The study had local ethical committee approval and all subjects gave written informed consent. Results We studied 70 subjects (32 male) with a mean age of 52 years [range 22 82], SBP of 131 mmhg [range ] and DBP of 76 mmhg [range 57 92] (Table 1). Figure 1. Plot of fast Fourier transformation-derived baroreceptor sensitivity (BRS FFT) versus age. 349

4 S. L. Dawson et al. Figure 2. Plot of logarithmically transformed fast Fourier transformation-derived baroreceptor sensitivity (BRS FFT) data versus age. The changes in cardiac BRS with age, as calculated using FFT, are illustrated in Figure 1. These indicate a non-linear decline in BRS with age. All three methods gave similar results, with little further decline after the fourth and fifth decades (only FFT data are shown graphically). Logarithmic transformation yields a more linear relation, with linear regression residuals that are normally distributed (Figure 2). All three BRS data sets were, therefore, logarithmically transformed; subject characteristics did not require transformation. There was a highly significant inverse relationship between the transformed BRS and age for all three measurements [P < 0:001; illustrated for FFT in Figure 2 (r ¼¹0:69)]. For further statistical analysis we used the transformed BRS values. The univariate correlates for the three methods of BRS calculation and age, BMI, gender, blood pressure and heart rate are shown in Table 2. The regression coefficients for age and clinic SBP were significant (P < 0:001) for all three methods of measuring BRS for example, logfft ¼ age (i.e. every year the BRS falls by 1.2% of the previous year s value) and logfft ¼ SBP (i.e. for every 1 mmhg increase in SBP BRS falls by 0.7% of the previous value). The only other measure which achieved statistical significance was heart rate with FFT and IRF. Multiple linear regression using the same independent variables as in the univariate analysis showed that age, clinic SBP, DBP and heart rate were significant predictors of FFT BRS variability (P < 0:05), but that only age was significant for Valsalva-derived (P < 0:05) and IRF-derived (P < 0:01) BRS variance. We placed these significant predictors for FFT variability in a stepwise linear regression model to examine their individual and cumulative contribution. Again, age was strongly predictive, contributing 50.3% of the variability. Although heart rate was a significant factor, only a further 4.2% was added to the predictability, and clinic SBP was rejected from this model. Single regression of age and SBP and age and DBP were highly significant (age ¼ ¹11:9 þ 0:486SBP, P < 0:001; age ¼ 11:4 þ 0:528DBP, P < 0:012), but neither age and heart rate nor heart rate and SBP were significantly related. The plots of phase versus frequency for the older (<50 years, n ¼ 42) and younger subjects (<45 years, n ¼ 20) are shown in Figure 3. The mean (SD) phase in the low- and high-frequency bands was ¹0.040 (0.11) and þ0.014 (0.14) radians for the older group, compared with ¹0.031 (0.08) and ¹0.011 (0.14) radians for the younger group. The low-frequency Table 2. Single regression of subject demographic indices against baroreceptor sensitivity values: results presented as y ¼ c þ mx Parameter (x) LogFFT ( y) LogVALS ( y) LogIRF ( y)... Age y ¼ 1:68 ¹ 0:0118x y ¼ 1:2 ¹ 0:0101x y ¼ 1:5 ¹ 0:013x [1.53, 1.83]; [¹0.014, ¹0.009] a [1.02, 1.34]; [¹0.013, ¹0.0069] [1.19, 1.82]; [¹0.02, ¹0.008] P < 0:001 P < 0:001 P < 0:001 Sex y ¼ 0:954 þ 0:0685x y ¼ 0:816 ¹ 0:0941x y ¼ 0:454 þ 0:209x P ¼ 0:334 P ¼ 0:192 P ¼ 0:087 BMI y ¼ 0:954 þ 0:0685x y ¼ 0:665 þ 0:0002x y ¼ 0:617 þ 0:0073x P ¼ 0:443 P ¼ 0:196 P ¼ 0:65 Clinic SBP y ¼ 2:0 ¹ 0:00711x y ¼ 1:34 ¹ 0:00501x y ¼ 1:89 ¹ 0:00826x [1.49, 2.51]; [¹0.011, ¹0.003] [0.786, 1.886]; [¹0.009, ¹ ] [1.03, 2.75]; [¹0.015, ¹0.0018] P < 0:001 P ¼ 0:018 P ¼ 0:014 Clinic DBP y ¼ 1:56 ¹ 0:00642x y ¼ 0:898 ¹ 0:00291x y ¼ 1:58 ¹ 0:0102x P ¼ 0:062 P ¼¹0:422 P ¼ 0:079 Pulse y ¼ 1:57 ¹ 0:00752x y ¼ 1:14 ¹ 0:00653x y ¼ 1:59 ¹ 0:0118x [1.102, 2.042]; [¹0.014, ¹0.0008] P ¼ 0:069 [0.12, 2.34]; [¹0.02, ¹0.0011] P < 0:001 P ¼ 0:033 BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure. a 95% confidence intervals for m and c respectively. 350

5 Age-related decline in cardiac baroreceptor sensitivity patients. We have not considered the role of sleep, exercise, mental stress and posture, all of which can have important effects on BRS values. We did control for the influence of respiration. IRF is a new technique [17], which does not require prior frequency band selection or the incorporation of phase and coherence into its calculation. We were able to examine subjects over both a wide range of ages (60 years) and SBP and DBP (<60/35 mmhg respectively). All subjects had blood pressures <160/95 mmhg and therefore within the normotensive range as defined by the British Hypertensive Society guidelines. Subjects with higher blood pressure values were excluded, as hypertension can independently alter cardiac BRS sensitivity [1, 4]. We have not examined the influence of different physical activity levels, but elderly people who undertake endurance training have less reduction of the high-frequency components of the BRS than sedentary individuals [23]. Figure 3. Plot of phase relation between systolic blood pressure and pulse interval versus frequency for young (<45 years) and old (>50 years) subjects. Results are smoothed with a sixth-order polynomial. A similar phase relation is seen in both groups in the low-frequency band ( Hz), while there is a negative phase relation in the young group and positive phase relation in the older group in the highfrequency band ( Hz). results did not differ between age groups, whereas the high-frequency results did (P < 0:001). The mean coherence values (SD) for the older subjects for lowfrequency and high-frequency were 0.51 (0.05) and 0.42 (0.04) respectively and for the younger group 0.67 (0.06) and 0.38 (0.02); i.e. the coherence for both groups in the high-frequency bandwidth is <0.5. Discussion We present three different methods for measuring BRS, and the influence of population demographics on the values we obtained. We have demonstrated an inverse relationship between BRS and age as well as SBP. Cardiac BRS (as assessed by FFT and phase IV of the Valsalva manoeuvre) shows a good correlation with the traditional pharmacological and neck chamber methods in subjects of all ages [6 12]. These new methods are simple to perform and more acceptable to BRS and age We have demonstrated an inverse relationship between age and BRS, with little change in mean values after the fourth decade. Our findings concur with those of Veerman et al. (FFT-derived BRS) [3] and Ebert et al. (pharmacological and neck chamber methods) [2]. Veerman et al. showed a decline in the low-frequency ( Hz) and high-frequency ( Hz) components of the cardiac BRS in their elderly (70 90 years) compared to their adult (20 40 years) population, but not their young (10 15 years) and adult populations. This is in keeping with our findings of a greater decline in the third and fourth decades. The decline in cardiac BRS with age may represent an alteration in the balance between the sympathetic and parasympathetic nervous system, with vagal activity increasing while b-adrenoreceptor activity is reduced [3, 24]. However, the main fall in BRS appears to occur before any of these changes are clinically evident and certainly before atherosclerotic changes in the major vessels are apparent. Consequently, other mechanisms should be considered. The potential role of prostacyclin, free radicals and nitrous oxide on the endothelium and endothelial damage at points of shear forces are being examined [25, 26]. There is no age-related change in the phase relation between SBP and PI in the low-frequency range, as opposed to the high-frequency range where phase is positive in the elderly group and negative in the younger group. In this study coherence was low (<0.5) in the high-frequency bandwidth for both age groups and so it is difficult to infer whether this phase difference represents signal noise, influence of another modulating system (such as the cerebral vasomotor regulating centre) or an age-related change towards non-baroreceptor-mediated control of blood pressure. 351

6 S. L. Dawson et al. Therefore, although our non-invasive calculation of BRS using power spectral analysis (FFT) derives the BRS value from spectral powers for SBP and PI in the low- and high-frequency bandwidths, we cannot state whether only low-frequency bandwidth values should be used since high-frequency results may be subject to age-related change. This needs further investigation. BRS and blood pressure We have also shown an inverse relationship between BRS and blood pressure, especially SBP, even in this normotensive range. When age is accounted for this relationship decreases, age being the dominant factor. Gribbin et al. [1] found a decline in BRS using pharmacological methods in all age groups as blood pressure increased, the largest decline being in the younger groups. Our department has also found a reduction in FFT- and IRF-derived BRS values with elevated blood pressure [14]. Increasing blood pressure levels may cause this decline in BRS because of the increased incidence of atherosclerosis of the large vessels leading to reduced arterial compliance particularly at the carotid bifurcation and the aortic arch, where the afferent receptors of the BRS arc are located. However, some authors have found a difference in vascular compliance between subjects with isolated systolic and those with combined hypertension, but no difference in BRS [27]. BRS and gender Huikuri et al. [28] found that menopausal women had a lower Valsalva-derived BRS than men, and Sato et al. [29] described changes in power spectral analysis of heart rate variability throughout the menstrual cycle. We have not shown any clinically important differences with gender. However, a hormonal factor may influence BRS values, especially as the higher values seen in men than women in the same older age group studied by Huikuri et al. disappear with the use of hormone replacement therapy [29], underlining the possible role of the endothelium in changes in BRS with age. These changes in BRS with advancing age and blood pressure may help to explain the increased incidence of conditions characterized by disturbed arterial blood pressure control with advancing age (e.g. orthostatic hypotension, post-prandial hypotension and possibly cerebrovascular ischaemia and stroke [2, 24]). In conclusion, age and increasing SBP levels are associated with a non-linear decline in cardiac BRS whether derived from power spectral, Valsalva manoeuvre or impulse response analyses. Age appears to be the strongest predictor of BRS variability. The greatest decline occurs in the third and fourth decades, with little change after 40 years of age. In older age groups (>60 years), blood pressure still exerts an effect on BRS, higher blood pressure values being associated with a decline in BRS [4]. When considering predictors of cardiac BRS, age seems to be the strongest predictor for the younger group and SBP for the older group, but this hypothesis needs further testing. We should consider the influence of these patient demographic factors on BRS values, if we are to use them as prognostic indicators [30, 31] and derive a set of ageadjusted values for clinical reference. Key points Increasing age and systolic blood pressure levels are inversely related to cardiac baroreceptor sensitivity. There is little further decline in baroreceptor sensitivity after the fourth decade. Age is the dominant factor in decline in cardiac baroreceptor sensitivity in normotensive subjects. There are age-associated differences in the relationship between pulse interval and blood pressure change in different frequency bands, which must be considered in the non-invasive calculation of cardiac baroreceptor sensitivity in older people. Acknowledgement S.L.D. is funded by a grant from The Stroke Association of the United Kingdom. References 1. Gribbin B, Pickering TG, Sleight P, Peto R. Effect of age and high blood pressure on baroreflex sensitivity in man. Circulation Res 1971; 29: Ebert TJ, Morgan BJ, Barney JA, Denahan T, Smith JJ. Effects of aging on baroreflex regulation of sympathetic activity in humans. Am J Physiol 1992; 263 (Heart Circ Physiol 32): H Veerman DP, Imholz BPM, Wieling W, Karemaker JM, van Montfrans GA. Effects of aging on blood pressure variability in resting conditions. Hypertension 1994; 24: James MA, Robinson TG, Panerai RB, Potter JF. Arterial baroreceptor-cardiac reflex sensitivity in the elderly. Hypertension 1996; 28: James MA, Panerai RB, Potter JF. Spectral and sequence analysis in the assessment of arterial baroreflex sensitivity in the elderly (Abstract). J Hypertens 1996; 12: James MA, Panerai RB, Potter JF. Sequence and spectral analysis in the assessment of arterial baroreceptor cardiac reflex sensitivity in the elderly. A comparison with established pharmacological methods. Clin Sci 1998; 94: Parati G, Saul JP, Di Rienzo M, Mancia G. Spectral analysis of blood pressure and heart rate variability in evaluating cardiovascular regulation. A critical appraisal. Hypertension 1995; 25: Robbe HWJ, Mulder LJM, Ruddel H, Langewitz WA, Veldman JBP, Mulder G. Assessment of baroreceptor reflex sensitivity by means of spectral analysis. Hypertension 1987; 10:

7 Age-related decline in cardiac baroreceptor sensitivity 9. Smith SA, Stallard TJ, Littler WA. Estimation of sinoaortic baroreceptor heart rate reflex sensitivity and latency in man: a new microcomputer assisted method of analysis. Cardiovasc Res 1986; 20: Malliani A, Pagani M, Lombardi F, Cerutti S. Cardiovascular neural regulation explored in the frequency domain. Circulation 1991; 84: Hughson RL, Quintin L, Annat G, Yamamoto Y, Gharib C. Spontaneous baroreflex by sequence and power spectral methods in humans. Clin Physiol 1993; 13: Parlow J, Viale JP, Annat G, Hughson R, Quintin L. Spontaneous cardiac baroreflex in humans. Comparison with drug-induced responses. Hypertension 1995; 25: Airaksinen KEJ, Tahvanainen KUO, Kuusela TA et al. Cross spectral analysis in assessment of baroreflex gain in patients with coronary heart disease. Ann Non-invasive Electrocardiol 1997; 2: Panerai RB, James MA, Potter JF. Impulse response analysis of baroreceptor sensitivity. Am J Physiol 1997; 272: H Dawson SL, Robinson TG, Youde JH et al. The reproducibility of cardiac baroreceptor activity assessed non-invasively by spectral and sequence techniques. J Clin Autonom Res 1997; 7: Imholz BPM, Van Montfrans GA, Settels JJ, Van der Hoeven GMA, Karemaker JM, Wieling W. Continuous non-invasive blood pressure monitoring: reliability of Finapres device during the Valsalva manoeuvre. Cardiovasc Res 1988; 22: Lal SKL, Henderson RJ, Cejnar M, Hart MG, Hunyor SN. Physiological influences on continuous finger and simultaneous intra-arterial blood pressure. Hypertension 1995; 26: Stokes DN, Clutton-Brock T, Patil C, Thompson JM, Hutton P. Comparison of invasive and non-invasive measurement of continuous arterial pressure using the Finapres. Br J Anaesthesia 1991; 67: Smith SA, Stallard TJ, Salih MM, Littler WA. Can sinoaortic baroreceptor heart rate reflex sensitivity be determined from phase IV of the Valsalva manoeuvre? Cardiovasc Res 1987; 21: Palmero HA, Caeiro TF, Iosa DJ, Bas J. Baroreceptor reflex sensitivity index derived from phase 4 of the Valsalva manoeuvre. Hypertension 1981; 3II: Panerai RB, James MA, Potter JF, Fan L, Evans DH. Baroreceptor sensitivity in human subjects: sequence or spectral analysis. Computers Cardiol 1995; Dawson SL, Panerai RB, Potter JF. ECG or Finapres derived pulse interval for calculation of cardiac BRS? Blood Pressure Monitoring 1998; 3: Bowman AJ, Clayton RH, Murray A, Reed JW, Subhan MF, Ford GA. Baroreflex function in sedentary and endurance-trained elderly people. Age Ageing 1997; 26: Mancia G, Di Renzo M, Grassi G, Parati G. Blood pressure variability and reflex control in the elderly. Aging Clin Exp Res 1995; 7: Chapleau MW, Cunningham JT, Sullivan MJ, Wachtel RE, Abboud FM. Structural versus functional modulation of the arterial baroreflex. Hypertension 1995; 26: Majmudar NG, Robson SC, Francis R, Ford GA. Changes in nitric oxide activity associated with the menopause in the human forearm (Abstract). Age Ageing 1997; 27 (suppl. 1): Sumimoto T, Mukai M, Murakami E, Hiwada K. Haemodynamic characteristics in elderly patients with isolated systolic hypertension. J Hum Hypertens 1990; 4: Huikuri HV, Pikkujamsa SM, Airaksinen J et al. Sex-related differences in autonomic modulation of heart rate in middle-aged subjects. Circulation 1996; 94: Sato N, Miyake S, Akatsu J, Kumashiro M. Power spectral analysis of heart rate variability in healthy young women during the normal menstrual cycle. Psychosomatic Med 1995; 57: Farrell TG, Odemuyiwa O, Bashir Y et al. Prognostic value of baroreflex sensitivity testing after acute myocardial infarction. Br Heart J 1992; 67: Hirschl M, Kundi M, Blazek G. Five year follow-up of patients after thromboendarterectomy of the internal carotid artery. Stroke 1996; 27: Received 5 February; accepted in revised form 10 December