Loss of lagresponse curvilinearity of indices of heart rate variability in congestive heart failure
 Tushar P Thakre^{1, 2} and
 Michael L Smith^{1}Email author
DOI: 10.1186/14712261627
© Thakre and Smith; licensee BioMed Central Ltd. 2006
Received: 15 February 2006
Accepted: 12 June 2006
Published: 12 June 2006
Abstract
Background
Heart rate variability (HRV) is known to be impaired in patients with congestive heart failure (CHF). Timedomain analysis of ECG signals traditionally relies heavily on linear indices of an essentially nonlinear phenomenon. Poincaré plots are commonly used to study nonlinear behavior of physiologic signals. Lagged Poincaré plots incorporate autocovariance information and analysis of Poincaré plots for various lags can provide interesting insights into the autonomic control of the heart.
Methods
Using Poincaré plot analysis, we assessed whether the relation of the lag between heart beats and HRV is altered in CHF. We studied the influence of lag on estimates of Poincaré plot indices for various lengths of beat sequence in a public domain data set (PhysioNet) of 29 subjects with CHF and 54 subjects with normal sinus rhythm.
Results
A curvilinear association was observed between lag and Poincaré plot indices (SD1, SD2, SDLD and SD1/SD2 ratio) in normal subjects even for a small sequence of 50 beats (p value for quadratic term 3 × 10^{5}, 0.002, 3.5 × 10^{5} and 0.0003, respectively). This curvilinearity was lost in patients with CHF even after exploring sequences up to 50,000 beats (p values for quadratic term > 0.5).
Conclusion
Since lagged Poincaré plots incorporate autocovariance information, these analyses provide insights into the autonomic control of heart rate that is influenced by the nonlinearity of the signal. The differences in lagresponse in CHF patients and normal subjects exist even in the face of the treatment received by the CHF patients.
Background
Poincaré plot is an intuitive and commonly used method to assess complex nonlinear behavior in the study of physiological signals [1–6]. In the assessment of heart rate variability (HRV), Poincaré plots are constructed by plotting duplets of successive RR intervals [1, 2, 4–6], with an implicit assumption that the next RR interval is significantly determined by the current one. This assumption lends itself to further generalization of Poincaré plots by plotting mlagged plots where m represents the distance (in number of beats) between the duplet beats, that is, the 'lag' of the second beat from the first [2]. It has been observed in the context of the short term variability that the current RR interval can influence up to approximately eight subsequent RR intervals [2]. Therefore, a series of lagged Poincaré plots can potentially provide more information about the behavior of Poincaré plot indices in health and disease than the conventional 1lagged plot does [2].
Heart rate variability analysis provides a noninvasive means to assess the autonomic status of the heart [6–8]. Under normal conditions, the feedback elements characterized by vagal and sympathetic activation of the heart combined with the cardiac automaticity determine the HRV [7]. In various clinical conditions in which the sympathovagal balance is disturbed, such as after an episode of myocardial infarction, in diabetic autonomic neuropathy and in congestive heart failure, HRV is usually reduced [7–10]. In the context of congestive heart failure (CHF), the decrease in HRV has also been observed to correlate with disease severity [11–15]. However, two issues relating to the strategies employed for HRV analysis deserve closer scrutiny. First, the majority of methods of quantifying HRV (including conventional Poincaré plots) use successive RR interval duplets only, with the implicit assumption that the current beat is influenced by the immediately preceding beat. However, it has been reported that a heart beat influences not only the beat immediately following it, but also up to 6–10 beats downstream [2], possibly as a consequence of respiratory sinus arrhythmia. Thus, an analysis hinging on the use of only successive RR interval duplets will likely underestimate the role of the autocovariance function of RR intervals i.e., the ability of heart beats to influence a train of succeeding beats. Second, the Poincaré plot indices relating to shortterm and longterm variability in RR intervals do not capture the nonlinear disposition of HRV [16]. The autocovariance function of RR intervals captures the additional aspects of HRV (e.g. nonlinearity) that can be masked by the strong correlation between successive beats if 1lagged plots are used. Indeed, Brennan et al [16] argue that lagged Poincaré plots can fully describe the autocovariance as well as the power spectrum of HRV. Our proposed analysis uses lagged Poincaré plots to overcome the limitations of the present practice of timedomain analysis of HRV.
Therefore, we hypothesized that the lagresponse patterns of linear and nonlinear Poincaré plot indices would be different in a diseased heart as compared to a normal heart. To test our hypothesis, we compared longterm ECG recordings of a group of CHF subjects with those of normal subjects. We also explored whether linear versus nonlinear indices of HRV behaved differentially with respect to the lag. Finally, we compared the lagresponses of CHF patients and normal subjects in light of the fact that heart rate variability may have been potentially restored by the pharmacologic therapy in the CHF patients. Thus, we suggest an alternative analytic strategy that has the potential to provide unique insights into the pathophysiological basis of CHF.
Methods
Study subjects
We used the PhysioNet internet resource which is a large repository of various physiologic signals including electrocardiograms recorded by a 24hour Holter monitor [17, 18]. From this repository, we chose the data sets which included information on interbeat RR intervals. The chf2db database included records on 29 individuals with congestive heart failure. These data come from two previous trials of longterm digoxin [19] and carvedilol [20] therapy. These subjects included eight men and 2 women (gender was unknown in the remaining 21 subjects) aged 34 – 79 years. There were 4 subjects belonging to the NYHA class I, 8 belonged to NYHA class II and 17 to NYHA class III. For comparison, we used a set of ECG recordings from 54 healthy subjects with normal sinus rhythm. These subjects included 30 men (aged 28.5 – 76 years) and 24 women (aged 58 – 73 years). Complete details of the study subjects are provided in the Supplementary Table 1 (see additional file 1).
Extraction of NN intervals
The databases in the PhysioNet repository contain 24hour Holter ECG recordings in two data formats; unaudited beatwise annotations sampled at a frequency of 128 samples per second and the corresponding header identification information. The WFDB software package provided by the PhysioNet resource was used to extract beatwise RR intervals. We used the "ann2rr" command to extract a sequence of up to 50,000 normal beats (approximately 12 hours of ECG recording) to estimate the normaltonormal (NN, by specifying the N option in the command) intervals in seconds.
Study variables
For our analyses, we included only those indices that mathematically depend on the lag. As Mean RR and SDRR are theoretically independent of lag, we did not include these indices for studying the lagresponse. However, mean deltas and SDSD can be appropriately generalized to measure the lag response. For instance, we estimated deltas as the difference between the mlagged RR intervals. Thus, we were able to estimate mlagged deltas and their standard deviations (SDLD). In the special case of 1lagged beats, the mean deltas and SDLD are respectively equal to mean successive delta and SDSD. We, then, estimated the SD1 and SD2 for each of the mlagged beat sequence. Thus, we included the following four indices in our analysis: SDLD, SD1, SD2 and SD1/SD2 ratio.
Length of beat sequence used for analysis
Controversy exists with regard to the period of time for which ECG recordings should be monitored to best capture the HRV dynamics. While some researchers maintain that recordings less than 18 hours are insufficient [15], others have observed that shortterm RR interval recordings are as reliable and accurate as longterm recordings in analyzing HRV [22]. Therefore, we studied the lagresponse of HRV for different lengths of beat sequences. We used the following seven lengths of consecutive beats: 50, 100, 500, 1000, 5000, 10000, and 50000. These sequence lengths represent recordings ranging from ~1 minute to ~12 hours, depending on the overall heart rate. We, thus, used the following analytical strategy: for each subject included in the study, we used seven lengths of beat sequences. For each beat sequence, we used lag values from one to ten. For each value of lag we constructed a Poincaré plot and estimated SDLD, SD1, SD2, and SD1/SD2 ratio. For construction of Poincaré plots as well as for estimation of the Poincaré plot indices, we used normal beats only as annotated in the PhysioNet database resource.
Analysis of lagresponse
Considering the facts that mlagged Poincaré plots can describe the autocovariance function, the autocovariance function monotonically decreases with increasing lag for values of lags less than 10, and that the current beat influences only about six to eight successive beats in the context of a short range influence, we expected a pattern of lagresponse which is stronger at the lower values of lag and which attenuates with increasing lag. In normal subjects, therefore, we expected a curvilinear relationship that explains variations in Poincaré plot indices on the basis of lag. To test for curvilinearity we used a quadratic relationship model and plotted the estimates of SDLD, SD1, SD2, and SD1/SD2 ratio against lag. We then fitted a secondorder polynomial curve using the leastsquares method. We assessed the modelfit using R^{2} values. We then compared the average coefficients of the quadratic term in the secondorder polynomial equations in subjects with and without CHF.
Statistical analyses
We compared the linear and nonlinear mlagged Poincaré plot indices in CHF and normal groups by using MannWhitney rank sum test. To study the potential contribution of varying beat sequence length to estimates of the Poincaré plot indices, we used the nonparametric method of Spearman's correlation coefficient (ρ). The curvilinearity of the lagresponse was assessed by using a second order polynomial regression and the modelfit was assessed using the R^{2} value in order to quantify the amount of variation explained by the quadratic lagresponse of the Poincaré plot indices. We wrote dedicated routines in Visual Basic^{®} for estimation of mlagged Poincaré plot indices. Then, we used Stata 7.0^{®} for statistical analysis. For all statistical analysis, we assumed statistical significance as p < 0.05.
Results
Length of ECG recording and Poincaré plot indices
Summary of Poincaré plot indices in study subjects.*
Beat sequence length  SD1  SD2  SD1/SD2 ratio  

CHF  Normal  p  CHF  Normal  p  CHF  Normal  p  
50  0.0474 (0.636)  0.0222 (0.0319)  0.2270  0.0600 (0.0677)  0.0531 (0.0342)  0.2144  0.7503 (0.3518)  0.4013 (0.2305)  2 × 10^{6} 
100  0.0496 (0.0605)  0.0228 (0.0315)  0.1150  0.0640 (0.0619)  0.0595 (0.0302)  0.2215  0.6949 (0.3212)  0.3660 (0.2447)  3 × 10^{6} 
500  0.0541 (0.0626)  0.0249 (0.0247)  0.0696  0.0779 (0.0627)  0.0790 (0.0351)  0.2517  0.6640 (0.3278)  0.3181 (0.2086)  9 × 10^{6} 
1000  0.0505 (0.0559)  0.0248 (0.0224)  0.1004  0.0784 (0.0576)  0.0907 (0.0367)  0.0449  0.6243 (0.3285)  0.2727 (0.1726)  2.5 × 10^{6} 
5000  0.0467 (0.0366)  0.0231 (0.0169)  0.0043  0.0817 (0.0427)  0.1082 (0.0354)  0.0047  0.5534 (0.2532)  0.2146 (0.1240)  4 × 10^{9} 
10000  0.0483 (0.0341)  0.0221 (0.0146)  0.0008  0.0835 (0.0401)  0.1125 (0.0333)  0.0042  0.5457 (0.2328)  0.1967 (0.0972)  2 × 10^{10} 
50000  0.0477 (0.0310)  0.0253 (0.0196)  0.0003  0.0982 (0.0558)  0.1522 (0.0519)  6 × 10^{6}  0.5094 (0.2394)  0.1692 (0.0923)  4 × 10^{10} 
Spearman's rho  0.1766  0.2154  0.6577  0.3157  0.6643  7 × 10^{5}  0.2395  0.4900  0.0042 
Lagged Poincaré plots in CHF and normal sinus rhythm
To illustrate the putative influence of lag on estimates of Poincaré plot indices, we modified the conventional Poincaré plot by plotting (n+m)^{th} beat on the ordinate against the n^{th} beat on the abscissa. Figures 2C–2H depict this technique. It is evident from these Poincaré plots that the scatter of points increases with increasing lag. This indicates that increasing lag corresponds to increasingly unrelated beats. Our hypothesis attempted to characterize this behavior in normal subjects and CHF patients.
Curvilinearity of lagresponse
Coefficients and statistical significance of the quadratic term in equations regressing heart rate variability indices on lag.
HRV index  CHF patients  Normal subjects  

Coefficient  P  Coefficient  P  
Beat sequence length = 50  
SDLD  0.00002  0.769  0.00033  3.5 × 10^{5} 
SD1  0.00002  0.741  0.00020  3 × 10^{5} 
SD2  0.00004  0.417  0.00010  0.002 
SD1/SD2  0.00062  0.553  0.00510  0.0003 
Beat sequence length = 50000  
SDLD  1.5 × 10^{6}  0.968  0.00028  1 × 10^{7} 
SD1  2.27 × 10^{6}  0.932  0.00020  1.4 × 10^{7} 
SD2  0.00001  0.502  0.00003  0.078 
SD1/SD2  0.00028  0.531  0.00158  4 × 10^{8} 
We again examined if the increased HRV seen in the CHF patients (described in the previous section) confounded our interpretations of the lagged Poincaré plot analyses. We observed that (Supplementary Table 1 in additional file 1) 10 of the 29 CHF patients and 8 of the 54 control subjects had pNN50 values exceeding 10%. In addition 7 of the control subjects had mean heart rates over 100. To exclude the possibility that these abnormal values for pNN50 and mean heart rates could be responsible for the higher HRV in CHF as against normal subjects, we excluded these 25 subjects and performed a subset analysis on 19 CHF patients and 39 control subjects. The results from these subset analyses are detailed in supplementary tables 2 and 3 (see additional file 1). These results obtained are completely concordant with those obtained without excluding any subject from analysis (Tables 1 and 2). Thus, high pNN50 and mean heart rates in some subjects did not seem to influence our main inference about the loss of a curvilinear lag response in CHF.
Discussion
Our results proffer compelling evidence that normally there is a curvilinear relation between lag and Poincaré plot indices of heart rate variability and that this curvilinearity is impaired in CHF. The main implication of our work is that differences in heart rate variability of CHF patients and normal subjects can be detected when these are not apparent from traditional measures of HRV. In addition, a potential interpretation of these results may be that the short term influence of the lagged beats is importantly determined by the respiratory sinus arrhythmia. Although traditionally viewed as an autonomicmediated phenomenon, respiratory sinus arrhythmia may also involve nonautonomic mechanisms. The presence of complex Poincaré plots in the CHF group suggests that a combination of autonomic and nonautonomic control mechanisms may be responsible for the sinus arrhythmia in some of these patients. Further investigation is needed to better define the interpretation of these analyses in CHF patients.
Traditionally, HRV analysis has been used to assess the autonomic status of the heart and Poincaré plots have been widely used for HRV analysis. But the potential utility of lagged Poincaré plots has not been fully appreciated. Considering the fact that a heart beat is affected by its preceding beats, it follows logically that lag would affect the indices of HRV. We expect this effect to be stronger at the lower values of lag and to become weaker with increasing lag. Therefore, we expect a curvilinear relation between lag and indices of HRV normally. This curvilinearity is lost in CHF. This analysis, then, uncovers a potential means to further risk stratify patients with CHF.
Useful length of beat sequence
Our study also demonstrated that a sequence of 5000 beats and more is useful for analyzing both the shortterm and longterm HRV. Although our study was not designed to identify the useful minimum length of a beat sequence, we did observe the association of the estimates of Poincaré plot indices with varying lengths of beat sequences. We also found that the SD1/SD2 ratio was significantly different in CHF patients as compared to normal subjects for any length of the beat sequences. This observation is consistent with the recent notion that SDSD, SD1, and SD2 only capture the linear aspect of HRV whereas the SD1/SD2 ratio may better relate to the nonlinear component of HRV [2, 16]. Therefore, in the timedomain analysis of ECG, it may be the most informative to use the SD1/SD2 ratio. Our method further capitalizes on this property of the SD1/SD2 ratio by using mlagged SD1/SD2 ratios. Such a modification, by virtue of an implicit property of the autocovariance function, can improve the use of SD1/SD2 ratio as a result of a closer approximation of frequencydomain analysis.
Differences in HRV in CHF and in normal sinus rhythm
The analysis conducted in this study makes use of the autocovariance function which allows for a stronger and more robust internal comparison for individual subjects as opposed to the more inaccurate method of group comparisons. We observed a loss of curvilinearity of lagresponse that prevailed in CHF patients with therapy in spite of an improvement in HRV attributable to the drug therapy. It is known that digoxin only improves the clinical profile of CHF patients but does little to improve survival [21, 23]. Thus, even though the patients on therapy seem to have greater HRV than normal, they might still be at risk for early death. This can have great prognostic implications. We propose that patients in whom lagresponse curvilinearity is restored may have a better prognosis than those in whom it is not restored. Further prospective studies are needed to test this hypothesis. Nevertheless, these initial findings suggest that this method offers a unique advantage over the conventional method of 1lagged Poincaré plots.
Study limitations
Our study suffers from several limitations. First, we conducted a secondary data analysis. In addition to the inherent limitations of such analyses, we were restricted to the analysis of the ECG signals only. It would have been more informative to study other pathophysiological and sociodemographic correlates of CHF. Second, the small sample sizes of both the CHF and normal sinus rhythm datasets do not allow wider generalization of our results. Therefore, studies of larger magnitude that can provide more conclusive insights into the utility of these measures of HRV are needed. Third, rather surprisingly, we observed more heart rate variability in CHF patients as compared to the normal subjects, which contrasts the generally accepted notion that HRV is decreased in CHF as compared to normal sinus rhythm. This unexpected result can be partially explained by the facts that i) A large proportion of the CHF patients exhibited patterns of complex Poincaré plots; ii) It is possible that the differential distribution of age, gender, socioeconomic status and other unknown parameters in the comparison group may have confounded the observations on increased heart rate variability in CHF subjects; and iii) all the CHF patients in this study were receiving longterm digoxin therapy [19, 20] which can potentially improve HRV in the CHF patients [23, 24].
Furthermore, though SD1 and SD1/SD2 ratios were higher in CHF patients as compared to normal subjects, SD2 (an index of longterm variability) was higher in the normal subjects for beat sequences 500 and above. This is consistent with two sets of observations: first, CHF patients are more likely to have a high degree of sinus arrhythmia that is not of respiratory origin, that is, they are more likely to have sudden jumps in the NN interval; and second, a study by Huikuri et al reported that depression of long term variability is a predictor of mortality [25]. Thus, CHF patients in this study had lower SD2 and hence a lower longterm heart rate variability than normal subjects, which might be responsible for the poorer prognosis in the former. We also studied three other measures of HRV across the study groups: percentage of beats with change in successive NN intervals exceeding 50 ms (pNN50) and 20 ms (pNN20) as well as the standard deviation of the NN intervals (SDNN). The individual data is shown in Supplementary Table 1 (see additional file 1). Using a MannWhitney test we observed that while in general the values for all the three parameters tended to be higher in the CHF group, these were not statistically significantly different (Supplementary Table 4, see additional file 1). These observations further highlight the value of the proposed method to gain insights into HRV.
Conclusion
In summary, despite the presence of these limitations, our study provides strong evidence for the loss of curvilinearity in the lagresponse of Poincaré plot indices in CHF patients even after receiving longterm therapy and even in cases in which some traditional measures of HRV are similar to those seen in normal individuals. We propose that these approaches of analysis can be used to improve the timedomain analysis of ECG signals and can enhance the diagnostic and prognostic indicators of CHF.
Abbreviations
 CHF:

Congestive Heart Failure
 ECG:

Electrocardiogram
 HRV:

Heart Rate Variability
 RR:

RR interval (these are considered to be measured from normal beats, NN)
 SD1:

Standard Deviation 1 (long axis of ellipse fitted to Poincaré plot)
 SD2:

Standard Deviation 2 (short axis of ellipse fitted to Poincaré plot)
 SDLD:

Standard Deviation of Lagged Deltas
 SDRR:

Standard Deviation of RR intervals
 SDSD:

Standard Deviation of Successive Differences
 pNN50:

percentage of beats with change in successive NN intervals exceeding 50 ms
 pNN20:

percentage of beats with change in successive NN intervals exceeding 20 ms
 SDNN:

standard deviation of the NN intervals
Declarations
Acknowledgements
The authors wish to thank Drs. Hemant Kulkarni and Manju Mamtani, Lata Medical Research Foundation, Nagpur, India for their constant help and discussion during conceptualization and conduct of the study. We also thank Dr. Phyllis Stein, Dr. Michal Javorka and Dr. Jacques Regnard for the insightful reviews.
Authors’ Affiliations
References
 Bergfeldt L, Haga Y: Power spectral and Poincare plot characteristics in sinus node dysfunction. J Appl Physiol. 2003, 94 (6): 22172224.View ArticlePubMedGoogle Scholar
 Lerma C, Infante O, PerezGrovas H, Jose MV: Poincare plot indexes of heart rate variability capture dynamic adaptations after haemodialysis in chronic renal failure patients. Clin Physiol Funct Imaging. 2003, 23 (2): 7280. 10.1046/j.1475097X.2003.00466.x.View ArticlePubMedGoogle Scholar
 Tulppo MP, Makikallio TH, Seppanen T, Airaksinen JK, Huikuri HV: Heart rate dynamics during accentuated sympathovagal interaction. Am J Physiol. 1998, 274 (3 Pt 2): H8106.PubMedGoogle Scholar
 Woo MA, Stevenson WG, Moser DK: Comparison of four methods of assessing heart rate variability in patients with heart failure. Am J Crit Care. 1996, 5 (1): 3441.PubMedGoogle Scholar
 Woo MA, Stevenson WG, Moser DK, Middlekauff HR: Complex heart rate variability and serum norepinephrine levels in patients with advanced heart failure. J Am Coll Cardiol. 1994, 23 (3): 565569.View ArticlePubMedGoogle Scholar
 Woo MA, Stevenson WG, Moser DK, Trelease RB, Harper RM: Patterns of beattobeat heart rate variability in advanced heart failure. Am Heart J. 1992, 123 (3): 704710. 10.1016/00028703(92)905103.View ArticlePubMedGoogle Scholar
 Stys A, Stys T: Current clinical applications of heart rate variability. Clin Cardiol. 1998, 21 (10): 719724.View ArticlePubMedGoogle Scholar
 Casolo G, Balli E, Taddei T, Amuhasi J, Gori C: Decreased spontaneous heart rate variability in congestive heart failure. Am J Cardiol. 1989, 64 (18): 11621167. 10.1016/00029149(89)908710.View ArticlePubMedGoogle Scholar
 Casolo G, Balli E, Fazi A, Gori C, Freni A, Gensini G: Twentyfourhour spectral analysis of heart rate variability in congestive heart failure secondary to coronary artery disease. Am J Cardiol. 1991, 67 (13): 11541158. 10.1016/00029149(91)90887Q.View ArticlePubMedGoogle Scholar
 Casolo GC, Stroder P, Sulla A, Chelucci A, Freni A, Zerauschek M: Heart rate variability and functional severity of congestive heart failure secondary to coronary artery disease. Eur Heart J. 1995, 16 (3): 360367.PubMedGoogle Scholar
 Guzzetti S, Cogliati C, Mezzetti S, Turiel M, Lombardi F, Malliani A: Heart rate and systolic arterial blood pressure variabilities in the progression of chronic heart failure. Clin Sci (Lond). 1996, 91 Suppl: 3739.View ArticleGoogle Scholar
 Guzzetti S, Magatelli R, Borroni E, Mezzetti S: Heart rate variability in chronic heart failure. Auton Neurosci. 2001, 90 (12): 102105. 10.1016/S15660702(01)002740.View ArticlePubMedGoogle Scholar
 Guzzetti S, Mezzetti S, Magatelli R, Porta A, De Angelis G, Rovelli G, Malliani A: Linear and nonlinear 24 h heart rate variability in chronic heart failure. Auton Neurosci. 2000, 86 (12): 114119. 10.1016/S15660702(00)002393.View ArticlePubMedGoogle Scholar
 Stefenelli T, BerglerKlein J, Globits S, Pacher R, Glogar D: Heart rate behaviour at different stages of congestive heart failure. Eur Heart J. 1992, 13 (7): 902907.PubMedGoogle Scholar
 Wijbenga JA, Balk AH, Meij SH, Simoons ML, Malik M: Heart rate variability index in congestive heart failure: relation to clinical variables and prognosis. Eur Heart J. 1998, 19 (11): 17191724. 10.1053/euhj.1998.1148.View ArticlePubMedGoogle Scholar
 Brennan M, Palaniswami M, Kamen P: Do existing measures of Poincare plot geometry reflect nonlinear features of heart rate variability?. IEEE Trans Biomed Eng. 2001, 48 (11): 13421347. 10.1109/10.959330.View ArticlePubMedGoogle Scholar
 Goldberger AL, Amaral LA, Glass L, Hausdorff JM, Ivanov PC, Mark RG, Mietus JE, Moody GB, Peng CK, Stanley HE: PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals. Circulation. 2000, 101 (23): E21520.View ArticlePubMedGoogle Scholar
 The PhysioNet Database. [http://www.physionet.org]
 Krum H, Bigger JTJ, Goldsmith RL, Packer M: Effect of longterm digoxin therapy on autonomic function in patients with chronic heart failure. J Am Coll Cardiol. 1995, 25 (2): 289294. 10.1016/07351097(94)00417O.View ArticlePubMedGoogle Scholar
 Goldsmith RL, Bigger JT, Bloomfield DM, Krum H, Steinman RC, SacknerBernstein J, Packer M: Longterm carvedilol therapy increases parasympathetic nervous system activity in chronic congestive heart failure. Am J Cardiol. 1997, 80 (8): 11011104. 10.1016/S00029149(97)006164.View ArticlePubMedGoogle Scholar
 Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Circulation. 1996, 93 (5): 10431065.
 Jiang W, Hathaway WR, McNulty S, Larsen RL, Hansley KL, Zhang Y, O'Connor CM: Ability of heart rate variability to predict prognosis in patients with advanced congestive heart failure. Am J Cardiol. 1997, 80 (6): 808811. 10.1016/S00029149(97)005262.View ArticlePubMedGoogle Scholar
 Flapan AD, Goodfield NE, Wright RA, Francis CM, Neilson JM: Effects of digoxin on time domain measures of heart rate variability in patients with stable chronic cardiac failure: withdrawal and comparison group studies. Int J Cardiol. 1997, 59 (1): 2936. 10.1016/S01675273(96)028938.View ArticlePubMedGoogle Scholar
 Binkley PF, Haas GJ, Starling RC, Nunziata E, Hatton PA, Leier CV, Cody RJ: Sustained augmentation of parasympathetic tone with angiotensinconverting enzyme inhibition in patients with congestive heart failure. J Am Coll Cardiol. 1993, 21 (3): 655661.View ArticlePubMedGoogle Scholar
 Huikuri HV, Makikallio TH, Airaksinen KE, Seppanen T, Puukka P, Raiha IJ, Sourander LB: Powerlaw relationship of heart rate variability as a predictor of mortality in the elderly. Circulation. 1998, 97 (20): 20312036.View ArticlePubMedGoogle Scholar
 The prepublication history for this paper can be accessed here:http://www.biomedcentral.com/14712261/6/27/prepub
Prepublication history
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.