Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Original article 1279 Effect of altering conditions of the sequence method on barore¯ex sensitivity L. Ceri Daviesa , Darrel P. Francisa,b , Adam C. Scotta , Piotr Ponikowskic , Massimo Piepolia and Andrew J. S. Coatsa Background The sequence method is widely used as a simple, non-invasive measure of barore¯ex sensitivity (BRS). This technique, originally described in anaesthetized cats, has been transferred virtually unchanged to humans, without evidence that the optimal values in cats are the same as those in patients with cardiovascular disease. Objective To study the effect of altering the components of the sequence method on the measured BRS in patients with chronic heart failure (CHF) and in normal individuals. Methods Eighty patients with CHF [aged 62 6 12 years (mean 6 SD)] and 40 normal control individuals [aged 38 6 15 years (mean 6 SD)] underwent measurement of heart rate and non-invasive blood pressure. Altering only the shift between blood pressure and R±R interval and the required correlation coef®cient of the regression line had no effect on the value of BRS, but had a signi®cant effect on the number of valid sequences. Alteration of the blood pressure or R±R interval thresholds, however, affected not only the number of valid sequences, but also the value of BRS in both groups. In normal controls, agreement with the bolus phenylephrine method was improved by increasing the blood pressure threshold, although this led Introduction Attenuation of barore¯ex sensitivity (BRS) predicts poor outcome after myocardial infarction and in patients with chronic heart failure [1±3]. The ®rst measurements of BRS in humans [4] were performed by intravenous injections of a pressor agent that increases blood pressure (measured intra-arterially), which in turn reduces heart rate. Non-invasive techniques have subsequently been developed, but acceptance of these new techniques has been limited by their relatively poor agreement with the `gold-standard' invasive method [5]. The spontaneous, or sequence method, ®rst described in 1985 [6] is widely used as a measure of BRS. R±R interval and beat-to-beat systolic blood pressure data are scanned and sequences of three or more beats in which the blood pressure and R±R interval concomitantly increase (or decrease) by more than a threshold value are identi®ed. The BRS is de®ned as the slope of the regression line between the data points in these 0263-6352 & 2001 Lippincott Williams & Wilkins to a reduction in the number of valid sequences. In patients with CHF, agreement was optimized by decreasing both the blood pressure and R±R interval thresholds. This also had the effect of increasing the number of valid sequences. Conclusion Changes should be made to this technique, to optimize its validity in conscious humans, particularly when applied to patients with attenuated BRS. J Hypertens 19:1279±1287 & 2001 Lippincott Williams & Wilkins. Journal of Hypertension 2001, 19:1279±1287 Keywords: barore¯ex sensitivity, sequence method, chronic heart failure a National Heart & Lung Institute, Imperial College of Science, Technology and Medicine, b Department of Cardiology, St Mary's Hospital, London, UK and c Department of Cardiology, Military Hospital, Wroclaw, Poland. Sponsorship: L.C.D. is supported by the Robert Luff Fellowship, D.P.F. by the British Heart Foundation (FS 98005), M.P. by the Wellcome Foundation and A.J.S.C. by the Viscount Royston Trust. Correspondence and requests for reprints to Dr L. Ceri Davies, Heart Failure Unit, Royal Brompton Hospital, Sydney St, London SW3 6NP, UK. Tel: 44 207 352 8121 ext 2708; fax: 44 207 351 8510; e-mail: [email protected] Received 31 July 2000 Revised 28 December 2000 Accepted 15 February 2001 sequences. This technique was initially developed in anaesthetized cats [6,7]. However, the R±R interval and blood pressure thresholds, blood pressure to R±R interval shift and threshold correlation coef®cient of the regression line involve essentially arbitrary constants, which have been adopted virtually unchanged for the study of humans. There is no evidence that these are the optimal values. If the arbitrary constants used in the standard sequence method are not optimal, the results from the technique may have been unnecessarily poor. In order to optimize this method, by identifying optimum values for these constants, we assessed the effect on the BRS, derived by the sequence method, of altering the R±R interval and blood pressure thresholds, the blood pressure shift and the regression line correlation coef®cient, in healthy control individuals and patients with chronic heart failure, and compared them with other conventional methods of measuring BRS. 1280 Journal of Hypertension 2001, Vol 19 No 7 Methods Participants and measurements Eighty patients with chronic heart failure were recruited from a specialist clinic. They were diagnosed on the basis of clinical assessment (a history of dyspnoea and symptomatic exercise intolerance, with previous signs of pulmonary congestion or peripheral oedema), evidence of left ventricular dysfunction from radionucleide ventriculography or echocardiography, or both. Patients with atrial ®brillation, permanent pacemakers, more than two ectopic beats per minute or clinical instability within the preceding 3 months were not eligible for the study. Forty normal control individuals were also studied; they had no signi®cant past medical history, no abnormalities on examination, and were not taking regular medication. All participants gave informed consent and the study was approved by the local ethics committee. The individuals were studied between 1300 h and 1700 h, under standardized conditions, in a quiet room at a comfortable temperature. All were fasted for at least 2 h before testing and were not allowed to smoke or drink alcohol- or caffeine-containing beverages for 24 h before the study. On arrival to the investigation unit, the participants rested supine for 30 min and then underwent a 10 min recording of heart rate and non-invasive measurement of blood pressure. Twenty-six patients and 15 controls also underwent assessment of BRS by the bolus phenylephrine method [4]. Data collection Blood pressure was measured non-invasively by a Finapres device (model 2300; Ohmeda, Louisville, California, USA), with the cuffed ®nger resting comfortably at the level of the heart. The Finapres cuff was wrapped around the index ®nger of the left hand. The participants underwent several minutes of accustomization to the Finapres and the servo-adjust mechanism was turned off before recording. The electrocardiogram was acquired from the limb lead with the largest R wave (usually lead II). All data were sampled at 1000 Hz on a computer using an analog-to-digital converter (National Instruments, Newbury, Berkshire). The readings were saved onto ¯oppy disk and analysed off-line with custom software, which measured R±R intervals and beat-to-beat blood pressure. Barore¯ex sensitivity BRS by the sequence method The time series of R±R interval and systolic blood pressure (SBP) were scanned to identify the sequences in which R±R and SBP concurrently increased or decreased over three or more beats. The standard thresholds for SBP and R±R interval are 1 mmHg and 4 ms, respectively. In this study, we varied the threshold for SBP from 0 to 2 mmHg and the R±R interval threshold from 0 to 6 ms. In the standard version of the sequence method, there is a shift of 1 between the blood pressure pulse and the R±R interval, with the blood pressure pulse being plotted against the following R±R interval. In this study, we assessed a blood pressure shift from ÿ4 to 4 with respect to the R±R intervals (Fig. 1). The linear correlation between R±R and SBP was computed for each sequence. In the standard method, the sequence is only considered valid if the correlation coef®cient is . 0.80. In this study, several thresholds of correlation coef®cient were compared: 0, 0.8, 0.95, 0.96, 0.97, 0.98, 0.99 and 1.0. The average value of the individual slopes of the valid sequences was taken as the BRS. BRS by the bolus phenylephrine method This was performed according to the standard modi®cation of the method of Smyth et al. [4], which uses phenylephrine instead of angiotensin and non-invasive blood pressure measurements instead of intra-arterial ones [8]. The procedure was carried out after informed consent and with appropriate resuscitation equipment and staff available, in 26 patients and 15 normal individuals. Phenylephrine was injected as a bolus at a starting dose of 2 ìg/kg, increasing by 50 ìg until the increase in SBP was at least 15 mmHg. The bolus Fig. 1 ECG 24 0 14 BP Diagram showing the shifts between blood pressure (BP) and R±R interval. ECG, electrocardiogram. Assessment of BRS by the sequence method Ceri Davies et al. 1281 injection was repeated at least three times at the adequate dose and the linear regression slope of R±R and SBP was calculated for each bolus. The ®nal result for each episode was computed as the mean of at least three values. BRS by the spectral analysis method BRS was calculated by the spectral analysis method as previously described. Brie¯y, power spectral analysis was performed on the same R±R interval and SBP data used for the sequence method, through the use of an autoregressive algorithm [9], with the model order selected according to the Akaike information criterion [10]. A model order between 12 and 18 was found to be appropriate in all cases. The component in the lowfrequency band (0.04±0.15 Hz) was considered. The áindex was computed as the square root of the ratio between R±R and systolic blood pressure spectral powers in the low-frequency band, in the presence of an adequate coherence (. 0.5) between the R±R interval and SBP as assessed by cross-spectral analysis [5,6]. Statistical analysis Numerical distributions are described by their mean SD. Comparisons between group means were carried out with Student's t-test and, in the case of multiple comparisons, analysis of variance (ANOVA). Results Patient characteristics The average age of the 80 patients was 62 12 years. Fifty-three had heart failure due to coronary artery disease, 25 had idiopathic dilated cardiomyopathy and two had heart failure due to mitral regurgitation. The mean ejection fraction was 0.30 0.14. Eleven patients were in New York Heart Association class I, 46 in class II, 23 in class III. Seventy-four of the patients were taking an angiotensin converting enzyme inhibitor or angiotensin II receptor antagonist, 70 were taking diuretics, and none was receiving a â-blocker. Sixty-two of the patients were men and 18 were women. The average age of the 40 normal individuals was 38 15 years. Twenty-®ve were men and 15 were women. Effect of changing blood pressure shift on measurement of BRS The effect of changing the blood pressure shift on the number of valid sequences is shown in Figure 2. For this analysis, the blood pressure threshold was set at 1 mmHg, the R±R interval threshold at 4 ms and the correlation coef®cient threshold at 0.8. In both groups, there was a signi®cantly larger proportion of valid sequences (number of valid sequences divided by total duration of recording multiplied by 100) when there was a zero shift between the blood pressure pulse and R±R interval (P , 0.0001 in controls and patients). The effect of different blood pressure shifts on the value of BRS obtained is shown in Figure 3. Again, the maximal value for BRS was obtained with a zero blood pressure shift, but there was no signi®cant difference in the values obtained within the two groups (P 0.9 in patients and P 0.1 in controls). Effect of changing sequence correlation threshold on measurement of BRS The effect of changing the threshold for the correlation coef®cient on the number of valid sequences is shown in Figure 4. The blood pressure and R±R interval thresholds remained unaltered from before, and the blood pressure shift was set to the conventional 1 beat. As the threshold correlation coef®cient was increased, there was a large decrease in the number of valid sequences in both patients and controls (P , 0.002 in both groups). The largest number of valid sequences Fig. 2 (b) 7 6 5 4 3 2 1 0 Proportion of valid sequences (%) Proportion of valid sequences (%) (a) 24 23 22 21 0 1 BP shift 2 3 4 16 14 12 10 8 6 4 2 0 24 23 22 21 0 1 BP shift 2 3 4 Effect of changing the extent and direction of blood pressure (BP) shift on the percentage of valid sequences in (a) patients with chronic failure and (b) normal controls. For example, a blood pressure shift of 1 means that the blood pressure series is shifted forward 1 beat before comparison with the R±R interval. 1282 Journal of Hypertension 2001, Vol 19 No 7 Fig. 3 (a) (b) 25 10 BRS (ms/mmHg) BRS (ms/mmHg) 12 8 6 4 2 0 24 23 22 21 0 1 BP shift 2 3 4 20 15 10 5 0 24 23 22 21 0 1 BP shift 2 3 4 Effect of different blood pressure (BP) shifts on the resultant barore¯ex sensitivity (BRS) value in (a) patients with chronic heart failure and (b) normal controls. Fig. 4 4 3.5 3 2.5 2 1.5 1 0.5 0 (b) 10 Proportion of valid sequences (%) Proportion of valid sequences (%) (a) 0 0.8 0.9 0.95 0.96 0.97 0.98 0.99 1 Threshold correlation coefficient 8 6 4 2 0 0 0.8 0.9 0.95 0.96 0.97 0.98 0.99 1 Threshold correlation coefficient Effect of altering the required correlation coef®cient on the percentage of valid sequences in (a) patients with chronic heart failure and (b) healthy controls. occurred when the threshold coef®cient was 0, but there was no signi®cant difference between the number of sequences when the correlation coef®cient was 0.8 (P 0.9 for both groups). Changing the required correlation coef®cient had no effect on the value obtained for BRS (Fig. 5; P 0.9 for both groups). Effect of changing blood pressure and R±R interval thresholds on BRS The effect of changing the blood pressure and R±R interval thresholds on the number of valid sequences is shown in Figure 6, with a blood pressure shift of 0, and a correlation coef®cient threshold of 0.8. In patients with heart failure, increasing either the blood pressure threshold or R±R interval threshold resulted in a decrease in the number of valid sequences. However, in the healthy controls, an increase in R±R interval threshold (up to 6 ms) had very little effect on the number of valid sequences. The effect of different thresholds on the value of BRS is shown in Figure 7. BRS measurement was highly sensitive to changes in the selected thresholds for blood pressure and R±R interval. In patients (Fig. 7a), the average measured BRS increased from 6.2 to 12.5 ms/mmHg as the blood pressure threshold was decreased from 2 to 0 mmHg (with an R±R interval threshold of 4 ms). The average measured BRS increased from 7.0 to 10.6 ms/mmHg as the R±R interval threshold was increased from 0 to 6 ms (with a blood pressure threshold of 1 mmHg). In controls (Fig. 7b), there was a marked dependence of measured BRS on blood pressure threshold, with the mean BRS increasing from 15.2 to 29.4 ms/mmHg as the blood pressure threshold decreased from 2 to 0 mmHg (with an R±R interval threshold of 4 ms). However, there Assessment of BRS by the sequence method Ceri Davies et al. 1283 Fig. 5 (b) 30 10 BRS (ms/mmHg) BRS (ms/mmHg) (a) 12 8 6 4 2 0 0 0.8 0.9 0.95 0.96 0.97 0.98 0.99 Threshold correlation coefficient 25 20 15 10 5 0 1 0 0.8 0.9 0.95 0.96 0.97 0.98 0.99 Threshold correlation coefficient 1 Effect of changing the required correlation coef®cient on the value of barore¯ex sensitivity (BRS) in (a) patients with chronic heart failure and (b) normal controls. Fig. 6 10 18 (b) 9 8 7 6 5 4 3 0 2 1 0 0 0.8 BP threshold 1.5 (mmHg) 1 2 3 4 5 6 R–R nterval threshold (ms) Proportion of valid sequences (%) Proportion of valid sequences (%) (a) 16 14 12 10 8 6 0 0.5 4 2 0 0 1 2 3 4 5 R–R interval threshold (ms) 2 1 BP threshold 1.5 (mmHg) 6 Effect of changing blood pressure (BP) and R±R interval thresholds on the percentage of valid sequences in (a) patients with chronic failure and (b) normal controls. was little dependence of the measured BRS on R±R interval threshold. Comparison with other measures of BRS The BRS values obtained by changing the blood pressure and R±R interval thresholds were compared with BRS values obtained by the bolus phenylephrine and spectral analysis methods. The correlation coef®cients are shown in patients (Fig. 8) and normal individuals (Fig. 9). It can be seen that, in both groups, the agreement between the sequence method and the other methods is highly dependent on the choice of blood pressure and R±R interval thresholds. The standard choice of thresholds (1 mmHg and 4 ms) did not result in the maximum agreement for BRS between the sequence method and the phenylephrine method or á-index, in either groups. Discussion The development of risk strati®cation techniques for patients after myocardial infarction and for patients with chronic heart failure offers the potential to identify those at greatest risk, so that expensive (or risk-laden) diagnostic and therapeutic options can be targeted at those with most to gain. The arterial barore¯ex has long been recognized to be attenuated in these patients [1± 3], and the degree of this attenuation has been con®rmed as an important risk strati®er, independent of 1284 Journal of Hypertension 2001, Vol 19 No 7 Fig. 7 (a) (b) 16 32 28 BRS (ms/mmHg) BRS (ms/mmHg) 14 12 10 8 0 0.5 6 4 0 1 2 3 2 4 24 20 0 0.5 16 1 BP threshold 1.5 (mmHg) 12 0 1 2 3 4 5 6 R–R interval threshold (ms) 5 6 R–R interval threshold (ms) 2 1 BP threshold 1.5 (mmHg) Effect of different blood pressure (BP) and R±R interval thresholds on measured barore¯ex sensitivity (BRS) in (a) patients with chronic heart failure and (b) normal controls. Fig. 8 (b) Seq vs α-index Patients 0.65 0.55 0 0.45 0.5 0.35 1 0.25 0 1 1.5 2 3 4 R–R interval threshold (ms) 5 BP threshold (ms/mmHg) 2 6 Agreement between methods (correlation coefficient) Agreement between methods (correlation coefficient) (a) Seq vs Phe Patients 0.65 0.55 0.45 0 0.5 0.35 1 0.25 0 1 1.5 2 3 4 5 BP threshold (ms/mmHg) 2 6 R–R interval threshold (ms) Effect of different blood pressure (BP) and R±R interval thresholds on the agreement between barore¯ex sensitivity (BRS) obtained by the sequence method (Seq) and (a) the bolus phenylephrine method (Phe) and (b) the spectral á-index method, in patients with chronic heart failure. conventional clinical markers in patients after acute myocardial infarction [11]. The gold-standard method of measuring BRS, the bolus phenylephrine method, is invasive, which has limited its widespread use in clinical practice. The sequence, or spontaneous, method of measuring BRS is more widely used because it is non-invasive and, unlike the spectral á-index method, does not require the use of complicated and opaque Assessment of BRS by the sequence method Ceri Davies et al. 1285 Fig. 9 (b) Seq vs α-index Controls (a) Seq vs Phe Controls 0.3 0.2 0.1 6 3 R–R interval threshold (ms) Agreement between methods (correlation coefficient) 0.4 0 0 0 0.5 1.5 1 BP threshold (mmHg) 2 0.7 0.6 0.5 6 R–R interval threshold (ms) Agreement between methods (correlation coefficient) 0.8 0.5 0.4 3 0 0 0.5 1 1.5 2 BP threshold (mmHg) Effect of different blood pressure (BP) and R±R interval thresholds on the correlation between barore¯ex sensitivity (BRS) obtained by the sequence method (Seq) and (a) the bolus phenylephrine method (Phe) and (b) the spectral á-index method, in healthy controls. mathematics. However, the conventional sequence method approach has a relatively high test failure rate in patients with attenuated BRS [12], and has been reported to correlate only poorly with the bolus phenylephrine method [5]. In this study, we have shown that altering the conditions for acceptance of valid sequences can have profound effects, not only on the number of valid sequences, but also on the value of BRS and hence its correlation with other measures of BRS. The individual components of the sequence method that we assessed in this study were the shift of blood pressure with respect to R±R interval, the threshold correlation coef®cient of the regression line between blood pressure and R±R interval, and the blood pressure and R±R interval thresholds. Blood pressure shift In the original description of the sequence method [6], the shift between blood pressure and R±R interval was set at 1 beat, with the R±R intervals being paired with the systolic pressure peak occurring within the preceding R±R interval (Fig. 1). However, previous work with bolus injections of phenylephrine [13] showed that, when blood pressure is transiently disturbed, the temporal relationship between the R±R interval sequence and the systolic blood pressure sequence depended on the heart rate, with a shift of 0 being optimal for individuals with heart rates of 75 beats/min or less and 1 for those with faster heart rates. Others have used the shift that produced the largest number of valid sequences (in most cases a shift of 0) [14,15]. In our study, altering the shift between ÿ4 and 4 had a profound in¯uence on the number of valid sequences obtained both in patients with heart failure and in normal controls. In both groups, the shift that gave the greatest number of valid sequences was 0. In our study, the mean heart rate of the patients with heart failure was 76 14 beats/min and the mean heart rate of the normal controls was 66 10 beats/min. However, the value for BRS was unaffected by the blood pressure shift in both groups, as described previously [15]. In the control group, the value of measured BRS peaks at blood pressure shifts of zero and 1, which may represent `true' barore¯ex sequences. The valid sequences obtained at the other blood pressure shifts correspond to the relative change between R±R interval and blood pressure from other modulatory systems. In patients with CHF, BRS is attenuated and so other cardiovascular re¯exes (such as the chemore¯ex) play a more prominent role in blood pressure and R±R interval control. As a result, the relationship between the changes in blood pressure and R±R interval is less dependent on their shift. It would seem optimal, therefore, to use the blood pressure shift that produces the 1286 Journal of Hypertension 2001, Vol 19 No 7 largest number of sequences, as this maximizes the opportunity to measure the BRS successfully, and does not bias the overall result. Correlation coef®cient threshold Studies investigating BRS by the sequence method have used different threshold correlation coef®cients to select valid sequences of beats, ranging from no speci®ed threshold correlation coef®cient [16] to a correlation coef®cient threshold of more than 0.95 [17]. In this study, it can be seen that an increase in the threshold correlation coef®cient results in a small decrease in the number of valid sequences, but has no effect on the value of BRS obtained. In fact, the number of sequences obtained with a threshold of 0.8 was not signi®cantly different from that obtained with no correlation coef®cient requirement. This is because the initial selection of sequences of three successive points with blood pressure and R±R interval moving consistently in the same direction results in sequences that have very high correlation coef®cients (nearly all greater than 0.90). Therefore, excluding sequences with lower correlation coef®cients has little impact on the number of valid sequences. We suggest, therefore, that it would be optimal to place no condition at all on the correlation coef®cient, as this maximizes the number of valid sequences without introducing bias into the BRS value. Blood pressure and R±R interval thresholds The original description of the sequence method in cats included sequences in which the blood pressure increased or decreased by 1 mmHg and the pulse interval either lengthened or shortened [6]. In a fuller description of the technique, the same authors proposed thresholds of 1 mmHg in blood pressure and 4 ms in R±R interval [7]. These thresholds are now very widely used in clinical research, although alternative values have ranged from 6 ms [18] down to no threshold [14]. The ®ndings in our study are that altering the blood pressure and R±R interval thresholds has a profound effect, not only on the number of valid sequences, but also the measured value of BRS. In the healthy control group, alteration of the R±R interval threshold had no effect on the number of valid sequences, the value of BRS or the correlation between either the bolus phenylephrine or spectral analysis BRS. As these individuals have a normal BRS, there is a large change in R±R interval for a small change in blood pressure. If, for example, an individual has a BRS of 15 ms/mmHg, an increase in blood pressure of 1 mmHg (required to attain the blood pressure threshold) will, on average, lead to a lengthening of R±R interval by 15 ms, easily exceeding the R±R interval threshold. In persons with a normal BRS, therefore, it is not necessary to include an R±R interval threshold. The situation is, however, very different in those with attenuated BRS, such as the patients in this study who had CHF. In this group, there is a much smaller increase in R±R interval for a given increase in blood pressure and therefore alterations in the R±R interval threshold will have an impact on the number of valid sequences. This is clearly seen in Figure 6a where, as both thresholds increase, the proportion of valid sequences decreases. The measured value of BRS itself increases as the blood pressure threshold decreases and the R±R interval threshold increases. Measured BRS is effectively the change in R±R interval divided by the blood pressure change. Therefore, if the R±R interval threshold is high and the blood pressure threshold is low, the validity thresholds will select sequences which yield a high measured BRS. These sequences are rare in patients with a low BRS and may represent artefact, rather than true barore¯ex phenomena. Agreement between sequence and other methods of measuring BRS In the healthy controls, the correlation coef®cients increased with increasing blood pressure threshold, although there was a slight decrease in correlation coef®cient with the phenylephrine method at the greatest blood pressure threshold. The agreement between the different methods could therefore be optimized by increasing the blood pressure threshold above the standard of 1 mmHg, although this would result in a decrease in the number of sequences. Doing this would remove the small, artefactual, sequences that are not barore¯ex-mediated. In patients with attenuated BRS, the method with the largest body of evidence for its prognostic usefulness is the bolus phenylephrine technique. Agreement between non-invasive measures of BRS and the phenylephrine technique is poor and this has led some authors to conclude that `the results obtained by means of non-invasive BRS assessments should not be used in clinical practice as an alternative to those obtained by the phenylephrine method' [5]. However, against this point of view, several arguments can be offered. First, the reproducibility of the bolus phenylephrine method in patients with low BRS is far from excellent [12], which places an upper limit on how closely it can agree with any other method [19]. Second, it is likely that the phenylephrine method and sequence method do not assess the same part of the barore¯ex arc and would therefore not be expected to agree. The changes in blood pressure are greater and sustained for longer with the phenylephrine method and may therefore explore the saturation point of the baroreceptors. In addition, the choice of an increase in blood pressure of 15 mmHg and a blood pressure shift of 1 beat are as arbitrary as the choices of threshold in Assessment of BRS by the sequence method Ceri Davies et al. 1287 the sequence method, and phenylephrine may have direct pharmacological actions, unrelated to its effect on the baroreceptors. However, it can be seen that altering the thresholds in the sequence method also has an effect on agreement with the á-index BRS, which is calculated from the same blood pressure and R±R interval data, although this is more profound in the normal controls. Study limitations The patients with heart failure and healthy controls were not age-matched and it is possible that normal elderly patients with a depressed BRS will show a pattern similar to that in the heart failure group. Conclusion Modi®cation of the arbitrary thresholds of the sequence method of BRS assessment in patients with heart failure and normal controls has a profound effect on the number of valid sequences obtained and on the value of BRS. Altering the blood pressure shift, whilst having no effect on the measured value of BRS, can dramatically increase the number of valid sequences. In both groups, a blood pressure shift of 0 gave the greatest number of sequences. Varying the threshold correlation coef®cient of the regression lines affected the number of valid sequences, but had no effect on the measured value of BRS. Finally, we found that altering the blood pressure and R±R interval thresholds away from their standard values affects not only the measured values of BRS, but also their agreement with other BRS measurement techniques. The clinical implications of these ®ndings are that the optimal thresholds vary, depending on whether the individual has normal or attenuated BRS, that a single threshold may not be applicable to both groups, and that values of BRS from different centres or studies, obtained using different threshold criteria, may not be directly comparable. References 1 2 3 4 5 6 Osterziel K, Hanlein D, Willenbrock R, Eichhorn C, Luft F, Dietz R. Barore¯ex sensitivity and cardiovascular mortality in patients with mild to moderate heart failure. Br Heart J 1995; 73:517±522. Farrell T, Odemuyiwa O, Bashir Y, Cripps T, Malik M, Ward D, Camm J. Prognostic value of barore¯ex sensitivity testing after acute myocardial infarction. Br Heart J 1992; 67:129±137. Mortara A, La Rovere M, Pinna G, Prpa A, Maestri R, Febo O, et al. Arterial barore¯ex modulation of heart rate in chronic heart failure. Clinical and haemodynamic correlates and prognostic implications. Circulation 1997; 96:3450±3458. Smyth H, Sleight P, Pickering G. Re¯ex regulation of arterial pressure during sleep in man. A quantitative method of assessing barore¯ex sensitivity. Circ Res 1969; 24:109±121. Pitzalis M, Mastropasqua F, Passantino A, Massari F, Ligurgo L, Forleo C, et al. Comparison between noninvasive indices of baroreceptor sensitivity and the phenylephrine method in post-myocardial infarction patients. Circulation 1998; 97:1362±1367. Bertinieri G, di Rienzo M, Cavallazzi A, Ferrari A, Pedotti A, Mancia G. A new approach to analysis of the arterial barore¯ex. J Hypertension 1985; 3:S79±S81. 7 8 9 10 11 12 13 14 15 16 17 18 19 Bertinieri G, Di Rienzo M, Cavallazzi A, Ferrari A, Pedotti A, Mancia G. Evaluation of baroreceptor re¯ex by blood pressure monitoring in anesthetized cats. Am J Physiol 1988; 254:H377±H383. Piepoli M, Coats A, Adamopoulos S, Bernadi L, Feng Y, Conway J, Sleight P. Persistent peripheral vasodilataion and sympathetic activity in hypotension after maximal exercise. J Appl Physiol 1993; 75: 1807±1814. Pagani M, Somers V, Furlan R, Dell'Orto S, Conway J, Baselli G, et al. Changes in autonomic regulation induced by physical training in mild hypertension. Hypertension 1988; 12:600±610. Sakando Y, Ishiguro M, Kibagawe G. Aikake information criterion statistics. In: Hingham M (editor): Dordrecht: D. Reidel Publishing; 1986 La Rovere M, Bigger T, Marcus F, Mortara A, Schwartz P. Barore¯ex sensitivity and heart-rate variability in prediction of total cardiac mortality after myocardial infarction. Lancet 1998; 351:478±484. Davies LC, Francis DP, Jurak P, Kara T, Piepoli M, Coats AJS. Reproducibility of methods for assessing barore¯ex sensitivity in normal controls and in patients with chronic heart failure. Clin Sci 1999; 97:515±522. Pickering T, Davies J. Estimation of the conduction time of the baroreceptor-cardiac re¯ex in man. Cardiovasc Res 1973; 7:213±219. Kardos A, Rudas L, Simon J, Gingl Z, Csanady M. Effect of postural changes on arterial barore¯ex sensitivity assessed by the spontaneous method and Valsalva manoeuvre in healthy subjects. Clin Auton Res 1997; 7:143±148. Blaber A, Yamamoto Y, Hughson R. Methodology of spontaneous barore¯ex relationship assessed by surrogate data analysis. Am J Physiol 1995; 268:H1682±H1687. Barlow J, Viale J-P, Annat G, Hughson R, Quintin L. Spontaneous cardiac barore¯ex in humans. Comparison with drug-induced responses. Hypertension 1995; 25:1058±1068. Herpin D, Ragot S. Mid- and long-term reproducibility of noninvasive measurements of spontaneous arterial barore¯ex sensitivity in healthy volunteers. Am J Hypertens 1997; 10:790±797. Parati G, di Rienzo M, Bertinieri G, Pomidossi G, Casadei R, Groppelli A, et al. Evaluation of the baroreceptor-heart rate re¯ex by 24-hour intraarterial blood pressure monitoring in humans. Hypertension 1988; 12:214±222. Francis DP, Coats AJS, Gibson D. How high can a correlation coef®cient be? Effects of limited reproducibility of common cardiological measures. Int J Cardiol 1999; 69:185±189.