Autonomic Information Flow Rhythms - IEEE Xplore

CARDIAC RHYTHMS

Autonomic Information Flow Rhythms From Heart Beat Interval to Circadian Variation ©GETTY IMAGES

BY DIRK HOYER, BIRGIT FRANK, _ RAFAL BARANOWSKI, JAN J. ZEBROWSKI, PHYLLIS K. STEIN, AND HENDRIK SCHMIDT

E

very control system must guarantee the stability of closed-loop behavior. This theory is well established in the design of linear control systems. The theory fails in the case of complex and nonlinear control systems. Instead, Lyapunov stability is usually used to study the behavior of these systems. The underlying hypothesis is that the system behaves within an attractor. In the case of the organism, this attractor incorporates the multiple-organ object function of staying alive. Recently, it was shown that specific prediction time intervals of the attractor are associated with specific mechanisms of the cardiovascular control system. The cardiovascular controller incorporates mechanisms at different time scales, such as the heart period (around 0.5–1 s), vagal (around 1–2 s), sympathetic (around 10 s), renin-angiotensin-aldosterone system (scale of minutes), renal volume pressure regulation (scale of hours), and circadian rhythm (which is present in all autonomic and humoral activities). Information flow measures, treated as functions of the time horizon of prediction (time scale of communication), can improve the assessment of particular aspects of complex cardiovascular mechanisms that operate over different scales [1]. We have confirmed in clinical studies that autonomic information flow (AIF) of different time scales, from seconds to minutes, identifies pathophysiologically disturbed autonomic function in several patient groups. Furthermore, outcome in these patients was related to particular AIF measures. This result holds for patients with multiple organ dysfunction syndrome (MODS) [2], heart failure [3], myocardial infarction (MI) [4], idiopathic dilated cardiomyopathy (IDC) [5], and post abdominal aortic surgery (AAS) [6]. Furthermore, heart rate, blood pressure, peripheral resistance, and the release/activity of vasodilating and pressure hormones display circadian variations. The incidence of pathological events, such as sudden cardiac death, stroke, ventricular arrhythmias, arterial embolism, symptoms of coronary heart disease, and MI, also has certain circadian peaks [8]. The objective of the present work was 1) an evaluation of the association of AIF measures with dysfunction and prognosis in selected diseases and 2) a measurement of their circadian rhythm.

Digital Object Identifier 10.1109/MEMB.2007.907091

IEEE ENGINEERING IN MEDICINE AND BIOLOGY MAGAZINE

Materials and Methods

The present study is based on the results of [2], [3], and [7]. Details are outlined later. Patients

The following is the brief detail of the patients in the study. ä Fifty healthy subjects (controls; no history of any cardiovascular symptoms, and normal diurnal activity and sleep [3]) ä Thirty-six patients with MODS (26 survivors, ten nonsurvivors, 28-day mortality [2]), in intensive care unit (ICU) ä Ninety-four patients after AAS (32 length of stay (LOS) >7 days, 62 LOS ( 7 days) [6]), in ICU ä Twenty-seven patients with heart failure (14 low risk— without history of aborted cardiac arrest (CA), 13 high risk—with history of CA), normal diurnal activity and sleep [3]). In all groups, 24-h Holter recordings, as well as 6-h data sets selected during day (between 6:00 a.m. and 3:00 p.m.) and night (between 10:00 p.m. and 6:00 a.m.), were analyzed using AIF methodology that was based on the heart rate variability (HRV) Task Force guidelines [2], [3], [9]. Autonomic Information Flow

Information flow characterizes communication within a time series in terms of information transfer over a prediction time horizon s. It measures the average information of the time shifted signal X(t þ s), which is already contained in the signal X(t). This AIF function is assessed based on the Shannon entropy X pi log2 pi ð1Þ Hx ¼ i

applied to the original time series X(t), leading to HX(t); the part X(t þ t) shifted by time lag t, leading to HX(tþt); and a bivariate presentation, leading to H[X(t), X(tþt)], as AIF½XðtÞ;XðtþsÞ ¼ HXðtÞ þ HXðtþsÞ H½XðtÞ;XðtþsÞ :

ð2Þ

The information flow over particular time horizons s allows the identification of particular physiological oscillators [1].

0739-5175/07/$25.00©2007IEEE

NOVEMBER/DECEMBER 2007

19

Specific prediction time intervals of the attractor are associated with specific mechanisms of the cardiovascular control system. The information loss is related to complexity measures [1]. We developed AIF measures based on NN interval series and from resampled time series, corresponding to conventional HRV the Task Force guidelines [9]. These are more completely described in [2], [3].

AIF 1 DECR1

AIF Measures

HF,(m )LF,(m )VLF

DEC(m )

DEClong

0

0

τl

τu

τ

Fig. 1. Sketch of AIF functions. This function describes the information loss (decay) over the prediction horizon s. AIF is normalized to AIF ¼ 1 at s ¼ 0. In the case of frequency band, filtered signals sl and su indicate the corresponding boundary periods. Take care that decay measures DEC* of different AIF functions (nonfiltered, filtered, short term, long term) are sketched together. The decay measures over different time scales are indicated by respective arrows. The superior m indicates the mean value over the frequency band-related periods, otherwise the AIF peak value was chosen within the same range (adapted from [4]).

AIF functions were estimated according to the short- and long-term approaches of the linear measures of Task Force HRV [2], [3], [9]. HRV at short time scales has been clearly related to autonomic mechanisms as follows: HF (highfrequency power) mainly reflects vagal modulation of heart rate; LF (low-frequency power) reflects sympathetic and parasympathetic modulation; the physiological basis of VLF (very-low-frequency power) is inconsistently reported, but a clear parasympathetic influence has been confirmed [10]; and long-term communication reflects slower fluctuations, with high amplitudes reflecting mainly circadian rhythm. Communication over one heartbeat period might additionally indicate inherent cardiac behavior. Typical measures of the AIF functions are their respective decays (DEC*) over prominent prediction time horizons (see Figure 1). These measures were assessed as short-term measures based on the average of short data windows as well as long-term measures based on the entire data set (Table 1). Statistical Analysis

Table 1. Calculation of AIF decay measures. Measure

Calculation

DECR1

AIF (in b/s normalized) decay over s ¼ 0.6 s (s standard beat period in all studies presented here) DECR1 is the mean measure of all 30-min segments based on nonfiltered time series resampled at 5 Hz AIF decay over periods related to the standard HF frequency band boundaries (s ¼ 1.25–3.33 s) (in b/s, normalized) DECHF is the mean measure of all 5-min segments, based on HF-filtered time series AIF decay over periods related to the standard LF frequency band boundaries (s ¼ 3.33–12.5 s) (in b/s normalized) DECLF is the mean measure of all 5-min segments, based on LF-filtered time series AIF decay over periods related to the standard VLF frequency band boundaries (s ¼ 12.5–166.7 s) (in b/s normalized) DECVLF is the mean measure of all 30-min segments, based on VLFfiltered time series AIF decay over s ¼ 100 s This measure reflects the complex communication involving long-term mechanisms based on the nonfiltered entire data set of 6 h or 24 h

DEC(m)HF

DEC(m)LF

DEC(m)VLF

DEClong

20


The comparisons between highand low-risk groups were done by means of the receiver operating curve (ROC) characteristics; pairs of true and false positives for the prediction of prognosis group were estimated over the full range of possible cutoff values. The area under the ROC curve (AUC) of 1 is associated with perfect classification, and AUC of 0.5 corresponds to a random assignment of patients by prognosis group. Significant discrimination was considered to exist if the lower bound of the two-sided 95% confidence interval of the respective AUC was >0.5. Results

Results obtained in different clinical studies are reviewed in Table 2. We found that decreased


Seasonal and circadian variations in the incidence of ruptured abdominal aortic aneurysm have been reported.

HF-related AIF (increased HF-related complexity) in MODS, measures, which are focused on source behavior characincreased VLF-related AIF (decreased VLF-related comteristics. Phase synchronization statistics allow the plexity) after abdominal aortic surgery, and decreased AIF assessment of couplings between rhythms with different over one heartbeat period (increased beat interval related frequencies, an important characteristic of coupled physicomplexity) in the heart failure patients were associated with ological oscillators. Information flow characterizes the higher risk. information transfer of a representative state variable as Selected underlying AIF decay measures are shown in well as between different variables. Traditional HRV Table 3. In the higher-risk MODS group, DECHF is increased analysis [9] is based on a linear model and cannot approxboth during day and night; however, DECLF is increased durimate these aspects of complex autonomic control and ing nighttime. In the heart failure patients with history of communication. cardiac arrest, DECR1 is significantly increased during dayWe investigated the hypotheses that 1) complex pathologitime only. cal dysfunction can be assessed by disease-specific AIF time The circadian patterns of corresponding parameters are horizons and 2) these AIF measures reflect the circadian patpresented in Figure 2. For these calculations, arrhythmic beat tern of complex autonomic control. intervals were not excluded. In controls, the heartbeat intervalIn MODS, sepsis mediators and toxins disturb the cardiac related AIF is reduced during nighttime (increased cellular signal transduction, leading to attenuated cellular DECR1). In the context of reduced HRV amplitude during nightresponses reflected in the increased randomness (comtime, this result indicates reduced modulating influences, leadplexity) of the heartbeat period reflected as the cardiac ing to higher complexity due to background noise. The clearly cycle time scale (increased DECR1). In this case, the HF increased complexity that is associated with decreased reduced DEC during nighttime may reflect the more regular nighttime breathing pattern, reflected in decreased complexity. Table 2. Direction of higher risk of significant AIF decay measures (for values, see Table 3). If DECLF is viewed as resulting Results of the daytime and the nighttime data sets are presented by pairs of arrows, results from the interaction of both vagal of the 24-h data sets are presented by single arrows. and sympathetic activity, the more regular vagal activity also results High Risk Versus Low Risk DECR1 DECmHF DEC(m)LF DECVLF DEClong in a reduction in complexity but to MODS [2] -›› -› --a lesser extent. The remaining AAS: LOS > 7 Versus LOS 7 Days [6] ﬂ ﬂ complexity of LF during the day History of Cardiac Arrest [3] ›----time can be interpreted by a greater extent of systematic modulations of the cardiovascular system. In all patient groups, the Table 3. Selected AIF-based risk predictors. Significant discrimination (lower AUC circadian rhythm is dramatically confidence value > 0.5) in bold. reduced or lost, but the remaining patterns clearly differ between the High Risk, Low Risk, AUC patients groups. Mean SD Mean SD (95% CI) Discussion

The stable behavior of the organism requires the adjustment between multiple physiological control mechanisms acting at similar and different time horizons. The identification of particular mechanisms and their interrelationships is limited because of their nonlinearity and complexity. Established approaches of nonlinear dynamics and statistics are Lyapunov stability and entropy

MODS DECR1 CA DECR1 MODS DECmHF CA DECmHF MODS DECLF CA DECmLF


Day Night Day Night Day Night Day Night Day Night Day Night

1.69 1.69 1.37 1.50 0.93 0.93 0.93 0.90 0.83 0.84 0.91 0.90

0.34 0.56 0.23 0.26 0.03 0.03 0.04 0.03 0.02 0.02 0.01 0.01

1.55 1.67 1.21 1.34 0.91 0.91 0.94 0.91 0.81 0.81 0.90 0.89

0.47 0.51 0.35 0.21 0.04 0.05 0.01 0.03 0.04 0.04 0.01 0.01

0.58 (0.38, 0.78) 0.51 (0.29, 0.73) 0.75 (0.56, 0.94) 0.69 (0.49, 0.90) 0.71 (0.52, 0.90) 0.75 (0.55, 0.95) 0.54 (0.31, 0.77) 0.56 (0.33, 0.78) 0.59 (0.39, 0.79) 0.78 (0.62, 0.94) 0.65 (0.44, 0.86) 0.63 (0.42, 0.84)


21

The multitude of biological rhythms in the organism and their complex interactions manage the systemic function of homeostasis.

HRV amplitude reflects loss of systematic control and the remaining stochastic noise. As for the MODS prognosis, increased DECHF might reflect decreased vagal influence to the heart pacemaker in MODS. The pathogenetic importance of vagus-related information flow was discussed in [2]. An intact vagus may be a major precondition for the body’s rapid anti-inflammatory action in sepsis and MODS [12], [13]. Thus, a loss of vagal activity may be pathophysiologically relevant for an increase in inflammatory mediators and a consecutive spill over to the circulation.

DECR1

DECmHF

Controls

2

0.9

1,5

0.85

1

0.8

0

6

12

18

0.9 0.85 0.8 0

24

2,5

MODS

0.95

0.75

0,5

6

12

18

24

0.9

1,5

0.85

1

0.8 6

12

18

24

0.95

2

0.9

1,5

0.85

1

0.8

0,5

0.75 0

6

12

18

2,5

0.95

2

0.9

1,5

0.85

1

0.8

6

12

18

24

0

6

12

18

24

24

0

6

12

18

24

0

6

12

18

24

0

6

12

18

24

0.95 0.9 0.85 0.8 6

12

18

24

0.95 0.9 0.85 0.8

0.75

0,5

18

0.8

0

24

12

0.85

0

2,5

6

0.9

0.75 0

0 0.95

0.95

2

0,5

AAS

DECmLF

0.95

2,5

CA

Intensive care patients also have a derangement of several hormonal axes. The hypothalamic-pituitary-thyroid axis is significantly disturbed, as characterized by an attenuation of the nocturnal surge of thyroid-stimulating-hormone (TSH), which seems to be prognostically significant [14]. Additionally, animal and human studies suggest that the function of the hypothalamic-pituitary-adrenal axis is also compromised in critically ill patients. In healthy controls, plasma leptin and cortisol seem to have reciprocal circadian rhythms with high nocturnal leptin levels and low nocturnal cortisol

0

6

12

18

24

Time (h) Fig. 2. Circadian course of AIF decay measures: Information loss over one heartbeat period DECR1, the HF-related time horizon DECmHF, and the LF-related time horizon DECmLF; in healthy controls, and patients with MODS, after AAS, and with heart failure (CA). Median quartiles. In all patient graphs, characteristics in the control group are included as benchmarks.

22



Every control system must guarantee the stability of closed-loop behavior.

concentrations. In critically ill patients with sepsis, this relationship is abolished and this has prognostic implications [15]. Sedation, given to critically ill patients so that they can tolerate the mechanical ventilation, may be a major source of alterations of the biological clock. In healthy persons, melatonin levels are normally high during nighttime and low during daytime and are suppressed by bright light. This circadian rhythm of melatonin release might be suppressed in critically ill patients, and a dyssynchronization of the melatonin secretion rhythm occurs [16]. Abnormalities of sleep are common in critically ill patients with total lack of activity and under constant light in the ICU. Thus, nearly half of total sleep time occurs during the daytime, and circadian rhythm is markedly diminished or lost. Mechanical ventilation itself may cause sleep disruption. In healthy subjects, sleep deprivation can have a negative impact on immune function [17]. Seasonal and circadian variations in the incidence of ruptured abdominal aortic aneurysm have been reported [18]. There may also be a significant association between periods of low barometric pressure and high incidence of ruptured aneurysm [19]. After AAS there is considerable inflammation associated with healing and the patients have undergone cardiopulmonary bypass. Prolonged recovery after AAS was associated with simplified long-term control (reduced DEClong) indicating altered integrative complex autonomic control (for details see [7]). The loss of circadian rhythm might partly be caused by similar reasons like in the MODS group. In the heart failure patients the heartbeat interval-related AIF (DECR1) was the only prognostically significant index, which is in line with the cardiac origin of the sudden cardiac death. Increased DECR1 might reflect cardiac destabilization because of disturbed (less divergent) attractor-like behavior. The incidence of sudden cardiac death shows a circadian rhythm as well as systematic variations during the week (highest incidence on Mondays) and year (highest incidence from December to February [20]). During the daytime, the highest incidence of out-of-hospital cardiac arrests is reported between 6:00 and 12:00 a.m. in Belgium [21] and Germany [20], but between 8:00 and 12:00 p.m. in Greece [22]. These findings suggest that the onset of sudden cardiac death may be associated with different endogenous rhythms and environmental influences. Different autonomic rhythms are further associated with psychiatric diseases. In acute schizophrenia, circadian autonomic variation is altered, i.e., the vagal information flow is reduced during nighttime [23]. In depression, the increased broadband AIF indicates the complex autonomic control [24]. The latter results reflect the diversity and complexity of influences to cardiovascular rhythm.


Conclusion

The multitude of biological rhythms in the organism and their complex interactions manage the systemic function of homeostasis. The identification of key parameters of this complex and nonlinear process is the challenge of biosignal analysis and interpretation. The magnitude of different physiological controllers acting at different time scales has implications with regard to understanding physiological and pathophysiological function and risk stratification, and developing therapeutic interventions. AIF functions of heart rate data enable the identification of underlying controllers and their complex communication. Time scales of AIF functions corresponding to the heartbeat period, the traditional short- and long-term HRV rhythms, and the circadian rhythm, seem to be valuable tools for risk stratification in cardiac and ICU patients. Acknowledgments

We thank Konstanze Ernst for her valuable contribution in technical assistance. This work was supported by Deutsche Forschungsgemeinschaft (Ho1634/8-1,2; Ho1634/9-1,2). Dirk Hoyer received a Ph.D. in biomedical engineering and information processing in 1985 and an Sc.D. (habilitation) in biosystems analysis in 2001, both from the Technical University, Ilmenau. In 2005, he received an Sc.D. from the Friedrich Schiller University, Jena. He is currently a professor for biosystems analysis at the Friedrich Schiller University, Jena, where he leads a research team on systems analysis in the Biomagnetic Center, Department of Neurology. His research interests are the analysis of complex cardiovascular control and central nervous information processing. His current projects deal with prenatal development and risk stratification, cardiovascular dysfunctions in adults, heart failure, stroke, intensive care monitoring, and psychiatric diseases. Birgit Frank received the diploma in 1985 and the Ph.D. in 1989, both in physics from the Friedrich Schiller University, Jena, Germany. Until 2000, her research interests were focussed on nonlinear modeling of superconducting devices. Currently, she develops methods of nonlinear time series analysis in the context of fetal maturation and pathophysiology. Rafal Baranowski received a Ph.D. degree in extended Holter ECG analysis in healthy subjects, habilitation in


23

repolarization analysis in Holter ECG. He is an assistant professor at Institute of Cardiology, Warsaw, Poland. His research interests are electrocardiology, biomedicine, and telemedicine.

_ Jan J. Zebrowski received an M.S.E.E. from the Warsaw University of Technology in 1974. He was a graduate student (special status) at the California Institute of Technology in 1977–1979, where he did experimental research in magnetic bubbles. He obtained the doctorate in physics at the Institute of Physics of the Warsaw University of Technology in 1981, specializing in the theory of solid state magnetism. Since 1983, his main interest has been nonlinear dynamics (chaos theory), initially in solid state magnetism and since 1991 in physiology and medicine. He obtained his Ph.D. in physics in 1989. He is now a full professor at the faculty of physics at Warsaw University of Technology and head of the cardiovascular physics group. His current main interests are nonlinear methods of time series analysis and modeling in cardiovascular physics. Phyllis K. Stein is a research associate professor of medicine at the Washington University School of Medicine, St. Louis, Missouri, and director of the Washington University Heart Rate Variability Laboratory, specializing in research-quality Holter scanning for both grant- and industry-funded studies, state-of-the-art HRV analysis, and novel ECG-based analysis, including analyses from overnight polysomnography. Her research interests are focused on the use of HRV and other ECG-based measures, including t-wave alternans as markers of risk and benefit. She is also interested in HR and HRV patterns and their relationship to sleep. Current projects include HRV in the cardiovascular health study, in depressed and nondepressed cardiac patients, and in participants with Gulf War illness. She is the author of more than 80 research papers, invited papers, and book chapters. Hendrik Schmidt received his MD in 1998 at the MartinLuther-University Halle-Wittenberg. In 1997, he was a research student at the Heart Failure Unit of the National Heart and Lung Institute/Royal Brompton Hospital in London. In 2005, he obtained his specialty in internal medicine and in 2006 in cardiology. He is also a specialist in sports medicine. He now works in the heart catheter lab at Martin-Luther University Halle-Wittenberg. His research focus is the autonomic dysfunction seen in critically ill patients with cardiogenic shock, acute coronary syndromes, and sepsis. Address for Correspondence: Dirk Hoyer, Biomagnetic Center, Department of Neurology, University Hospital, 24


Friedrich Schiller University, 07740 Jena, Germany. E-mail: [email protected]. References [1] D. Hoyer, B. Pompe, K. H. Chon, H. Hardraht, C. Wicher, and U. Zwiener, ‘‘Mutual information function assesses autonomic information flow of heart rate dynamics at different time scales,’’ IEEE Trans. Biomed. Eng., vol. 52, no. 4, pp. 584–592, 2005. [2] D. Hoyer, H. Friedrich, U. Zwiener, B. Pompe, R. Baranowski, K. Werdan, U. Mueller-Werdan, and H. Schmidt, ‘‘Prognostic impact of autonomic information flow in multiple organ dysfunction syndrome patients,’’ Int. J. Cardiol, vol. 108, no. 3, pp. 359–369, 2006. _ [3] D. Hoyer, H. Friedrich, B. Frank, B. Pompe, R. Baranowski, J. J. Zebrowski, and H. Schmidt, ‘‘Autonomic information flow improves prognostic impact of Task Force HRV monitoring,’’ Comput. Methods Programs Biomed., vol. 81, no. 3, pp. 246–255, 2006. [4] D. Hoyer, B. Frank, C. Go¨tze, P. K. Stein, J. J. Zebrowski, R. Baranowski, M. Palacios, M. Vallverdu, P. Caminal, A. Bayes De Luna, G. Schmidt, and H. Schmidt, ‘‘Interactions between short-term and long-term cardiovascular control mechanisms—A new aspect in risk stratification,’’ Chaos, vol. 17, no. 1, pp. 015110-1–015110-8, 2007. [5] D. Hoyer, M. Watanabe, R. Schneider, H. Hoyer, and G. Schmidt, ‘‘Risk stratification after acute myocardial infarction by autonomic information flow,’’ unpublished. [6] M. Palacios, H. Friedrich, C. Go¨tze, M. Vallverdu, A. B. de Luna, P. Caminal, and D. Hoyer, ‘‘Changes of autonomic information flow due to idiopathic dilated cardiomyopathy—A time scale specific complexity approach,’’ Physiol. Meas., vol. 28, no. 6, pp. 677–688, 2007. [7] D. Hoyer, H. Friedrich, P. K. Stein, P. P. Domitrovich, H. Hoyer, C. Faldu, and T. G. Buchman, ‘‘Autonomic information flow improves prognostic value of heart rate pattern after abdominal aortic surgery,’’ to be published. [8] B. Lemmer, ‘‘Circadian rhythm regulations of the cardiovascular system in rats and mice,’’ IEEE Eng. Med. Biol. Med., vol. 26, no. 6, pp. xx–xx, 2007. [9] Task Force of the ESC and NASPE, ‘‘Heart rate variability, standards of measurement, physiological interpretation and clinical use,’’ Circulation, vol. 93, no. 5, pp. 1043–1065, 1996. [10] J. A. Taylor, D. L. Carr, C. W. Myers, and D. L. Eckberg, ‘‘Mechanisms underlying very-low-frequency RR-interval oscillations in humans,’’ Circulation, vol. 98, no. 6, pp. 547–555, 1998. [11] J. V. Borovikova, S. Ivanova, M. Zhang, H. Yang, G. I. Botchkina, L. R. Watkins, H. Wang, N. Abumrad, J. W. Eaton, and K. J. Tracey, ‘‘Vagus nerve stimulation attenuates the systemic inflammatory response to endotoxin,’’ Nature, vol. 405, no. 6785, pp. 458–462, 2000. [12] K. J. Tracey, ‘‘The inflammatory reflex,’’ Nature, vol. 420, no. 6917, pp. 853–859, 2002. [13] C. Libert, ‘‘Inflammation: A nervous connection,’’ Nature, vol. 421, no. 6921, pp. 328–329, 2003. [14] M. Leon-Sanz, J. A. Lorente, L. Larrodera, P. Ros, J. Alvarez, A. E. Esteban, and L. Landin, ‘‘Pituitary-thyroid function in patients with septic shock and its relation with outcome,’’ Eur. J. Med. Res., vol. 2, no. 11, pp. 477–482, 1997. [15] S. R. Bornstein, J. Licinio, R. Tauchnitz, L. Engelmann, A. B. Negrao, P. Gold, and G. P. Chrousos, ‘‘Plasma leptin levels are increased in survivors of acute sepsis: Associated loss of diurnal rhythm, in cortisol and leptin secretion,’’ J. Clin. Endocrinol. Metab., vol. 83, no. 1, pp. 280–283, 1998. [16] K. Olofsson, C. Alling, D. Lundberg, and C. Malmros, ‘‘Abolished circadian rhythm of melatonin secretion in sedated and artificially ventilated intensive care patients,’’ Acta Anaesthesiol. Scand., vol. 48, no. 6, pp. 679–684, 2004. [17] S. Parthasarathy and M. J. Tobin, ‘‘Sleep in the intensive care unit,’’ Intensive Care Med., vol. 30, no. 2, pp. 197–206, 2004. [18] R. Manfredini, F. Portaluppi, M. Gallerani, A. Tassi, R. Salmi, P. Zamboni, F. Chierici, S. Occhionorelli, F. Mascoli, E. Rizzioli, A. Liboni, I. Donini, and C. Fersini, ‘‘Seasonal variations in the rupture of abdominal aortic aneurysms,’’ Jpn. Heart J., vol. 38, no. 1, pp. 67–72, 1997. [19] D. W. Harkin, M. O’Donnell, J. Butler, P. H. Blair, J. M. Hood, and A. A. Barros D’Sa, ‘‘Periods of low atmospheric pressure are associated with high abdominal aortic aneurysm rupture rates in Northern Ireland,’’ Ulster Med J., vol. 74, no. 2, pp. 113–121, 2005. [20] R. H. Arntz, S. N. Willich, C. Schreiber, T. Bru¨ggemann, R. Stein, and H. P. Schultheiss, ‘‘Diurnal, weekly and seasonal variation of sudden death,’’ Eur. Heart J., vol. 21, no. 4, pp. 315–320, 2000. [21] Belgian Cardiopulmonary Cerebral Resuscitation Study Group, ‘‘Further prospective evidence of a circadian variation in the frequency of call for sudden cardiac death,’’ Intensive Care Med., vol. 21, no. 1, pp. 45–49, 1995. [22] C. Savopoulos, A. Ziakas, A. Hatzitolios, C. Delivoria, A. Kounanis, S. Mylonas, M. Tsougas, and D. Psaroulis, ‘‘Circadian rhythm in sudden cardiac death: A retrospective study of 2,665 cases,’’ Angiology, vol. 57, no. 2, pp. 197– 204, 2006. [23] S. Boettger, D. Hoyer, K. Falkenhahn, M. Kaatz, V. K. Yeragani, and K. J. Ba¨r, ‘‘Altered diurnal autonomic variation and reduced vagal information flow in acute schizophrenia,’’ Clin. Neurophysiol., vol. 117, no. 12, pp. 2715– 2722, 2006. [24] S. Boettger, D. Hoyer, K. Falkenhahn, M. Kaatz, V. K. Yeragani, and K. J. Ba¨r, ‘‘Non-linear broad band dynamics are less complex in major depression,’’ to be published.
