Pathophysiological mechanisms of bradycardia in patients with anorexia nervosa

Abstract Background The purpose of this investigation was to examine heart rate variability (HRV), interbeat interval (IBI), and their interrelationship in healthy controls, bradycardic hyperpolarization‐activated cyclic nucleotide‐gated channel 4 (HCN4) mutation carriers, and patients with anorexia nervosa (AN). We tested the hypothesis that neural mechanisms cause bradycardia in patients with AN. Therefore, we assumed that saturation of the HRV/IBI relationship as a consequence of sustained parasympathetic control of the sinus node is exclusively detectable in patients with AN. Methods Patients with AN between the ages of 12 and 16 years admitted to our hospital due to malnutrition were grouped and included in the present investigation (N = 20). A matched‐pair group with healthy children and adolescents was created. Groups were matched for age and sex. A 24‐hour Holter electrocardiography (ECG) was performed in controls and patients. More specifically, all patients underwent two 24‐hour Holter ECG examinations (admission; refeeding treatment). Additionally, the IBI was recorded during the night in HCN4 mutation carriers (N = 4). HRV parameters were analyzed in 5‐minute sequences during the night and plotted against mean corresponding IBI length. HRV, IBI, and their interrelationship were examined using Spearman's rank correlation analyses, Mann‐Whitney U tests, and Wilcoxon signed‐rank tests. Results The relationship between IBI and HRV showed signs of saturation in patients with AN. Furthermore, signs of HRV saturation were present in two HCN4 mutation carriers. In contrast, signs of HRV saturation were not present in controls. Conclusions The existence of HRV saturation does not support the existence of parasympathetically mediated bradycardia. Nonneural mechanisms, such as HCN4 downregulation, may be responsible for bradycardia and HRV saturation in patients with AN.


| INTRODUCTION
Anorexia nervosa (AN) has the highest mortality rate among psychiatric disorders. 1,2 Other than suicide, cardiovascular and respiratory complications are the leading causes of mortality. 3 Typically, these deaths are unexpected and attributed to cardiac arrhythmia. 4 The causes of cardiac arrhythmia and their overall impact on cardiovascular function remain under investigation. Bradycardia, which is commonly observed in patients with AN, likely plays a crucial role in the pathophysiology of AN. 4 To provide insights into the neurophysiology of AN and etiology of bradycardia, heart rate variability (HRV) has been examined in patients with AN in many investigations. 5,6 HRV measures the variations in R-R interval, which reflect a complex interplay of feedback loops, thermogenesis, intrinsic mechanisms of pacemaker cells, and parasympathetic and sympathetic tone. Abnormal HRV has been associated with numerous cardiovascular and non-cardiovascular pathologies. [7][8][9][10][11] In a large number of research articles, high HRV parameters are associated with a high vagal tone. 6,[12][13][14][15] Conversely, low HRV is often associated with sympathetic dominance. 7,9,11,14,15 In patients with AN, mostly high HRV parameters were observed during the acute phase of the disease. It was therefore believed that bradycardia in patients with AN is mediated via a high cardiac vagal tone and high parasympathetic influence. [4][5][6] Interestingly, in the last few years, it has been suggested that HRV is partly a nonlinear surrogate for heart rate (HR). [16][17][18][19] Studies conducted by D'Souza et al imply that alterations in intrinsic HR via the regulation of ionic channels are primarily responsible for HRV and bradycardia in athletes. 18,19 Consequently, the theory of increased vagal tone or decreased sympathetic tone as a cause of starvation-or exercise-induced bradycardia has been questioned by some researchers. [16][17][18][19][20][21] Likewise, HRV metrics as an index of autonomic regulation have been challenged. 16 While nonneural mechanisms of bradycardia, such as HCN4 downregulation or HCN4 c.1737+1G>T mutation, presumably express themselves as a boost in HRV, drug-induced parasympathetic dominance leads to a saturation effect in HRV metrics. [23][24][25] This association shows that HRV may constantly increase with a decline in HR in HCN4 mutation carriers or subjects with HCN4 downregulation. At the same time, HRV uncouples from changes in HR at a certain point and then decreases with longer IBIs in subjects with a high vagal tone.
The underlying mechanism of HRV saturation is probably highintensity vagal nerve discharge during expiration as part of respiratory sinus arrhythmia. If sufficient acetylcholine (ACh) is released during expiration, ACh concentrations in the sinus node area remain high during inspiration, and the parasympathetic effect persists during the whole respiratory cycle, resulting in low HRV and HR ( Figure 1). 23,24 Thus, our hypothesis suggests that HRV saturation occurs in patients with AN, whereas such an effect remains unverifiable in controls and patients with the HCN4 c.1737+1G>T mutation, given that bradycardia in patients with AN is caused by increased cardiovagal modulation and not by nonneural mechanisms, such as HCN4 downregulation. In this investigation, we discuss the pathophysiological mechanisms of bradycardia and reveal the changes in HRV and HR during in-hospital refeeding treatment. Sepp recruited these patients from a former study. 27 Patients did not take any drugs during IBI recording, and they were not athletes.

| Ethics statement
All procedures were performed in accordance with the ethical standards of the appropriate ethics review board and with the 1964 Helsinki declaration and its later amendments. All study participants and their parents provided written informed consent for data acquisition and analysis. The study design was approved by the university ethics committee (University of Würzburg, study number 143/16-mk).
Data acquisition from patients with HCN4 c.1737+1G>T mutation was performed by Prof. Robert Sepp, who recruited these patients from a former study. Molecular genetic analyses and ECG recordings in these patients were approved by the Institutional Ethics Committee of University of Szeged, Hungary. For the control group, we used a database containing 151 24-hour ECGs from healthy individuals aged 2 to 23 years. 28 The database was composed of German and Polish children and adolescents. For Polish children and adolescents, data analysis and acquisition were approved by Poznan University of Medical Science Bioethical Committee.

| Design
We examined the effects of nutritional refeeding on HRV parameters and HR in 20 adolescents with AN during in-hospital refeeding. To assess changes in HRV parameters, a 24-hour ECG was performed at admission and during nutritional refeeding. The second ECG was F I G U R E 1 Heart rate variability saturation effect. Respiration-related episodic acetylcholine (ACh) release (heart rate variability [HRV] saturation on the left side, no HRV saturation on the right side). During each breathing cycle, parasympathetic-mediated variation in heart rate (HR) naturally occurs. During inspiration, activated baroreceptors diminish vagal tone via the central autonomic network, which is followed by an increase in HR. In contrast, during expiration, baroreceptors are less active and HR decreases. 30 In patients with anorexia nervosa, these fluctuations in the breathing cycle may attenuate at low HRs during certain sleep stages with a high parasympathetic influence. During expiration, the vagus nerve releases high amounts of ACh, which cannot be sufficiently cleaved by low acetylcholinesterase levels. 45,46 As a result, the expected decline in the concentration of ACh during inspiration in the sinus node area may not occur, and high concentrations of ACh persist during inspiration and expiration (left side). This sequence of events leads to sustained parasympathetic control of the sinus node during the whole breathing cycle, which eliminates respiratory cardiac modulation and reduces HRV at very low HRs. 8 From: de Geus; Should heart rate variability be "corrected" for heart rate? Biological, quantitative, and interpretive considerations; CC BY 4.0; www.onlinelibrary.wiley.com/doi/ full/10.1111/psyp.13287 recorded between fifth and 14th days post-admission when a highcalorie intake (>1500 kcal) and patient compliance were achieved. For the control group, we used a database containing 151 24-hour ECGs from healthy individuals aged 2 to 23 years. 28 Each patient was randomly matched to a control subject by age (±1 year) and sex. The monitoring period started between 07:00 and 12:00 and was discontinued on the second day. The Medilog Darwin ECG system (Holter recorder: Medilog FD4 or Medilog AR4, Schiller, Switzerland; sampling rate 250 Hz) was used for all ECG recordings of controls and patients with AN. An experienced cardiologist edited and inspected all ECG recordings for non-sinus rhythm or artifacts. Notepad++ software was used to count beat annotations. ECGs with <95% sinus beats were excluded. For HCN4 mutation carriers, we used Polar H10 chest belts to record nighttime RR intervals in an ambulant setting. Further, 24-hour ECGs were used to rule out severe dysrhythmia in these patients.
Time-domain (td) and frequency domain (fd) parameters were calculated from exported 5-minute IBI segments using Kubios HRV Pre-    33 Fourth, the relationship between IBI and pNN50 in patients and controls was plotted and fitted to quadratic and linear models ( Figure 2). 33

| Relationship between IBI and HRV metrics
We observed a positive correlation between IBI and LnHF power in In patients with AN, at admission, the quadratic function was superior to the linear one in describing the relationship between IBI and pNN50 ( Figure 2).

This investigation demonstrated that the relationship between IBI
and HRV showed signs of saturation in patients with AN and in two patients with the HCN4 c.1737+1G>T mutation. In contrast, the signs of HRV saturation were barely present in controls. Additionally, a significant increase in HR during nutritional refeeding was observed, whereas HRV parameters either increased or decreased.

| HRV and HR changes during starvation and refeeding
Our findings are highly consistent with studies measuring HRV before nutritional refeeding in patients with acute AN. 34

| Relationship between IBI and HRV in controls and patients with AN
Recent studies suggest that HRV is partly a nonlinear surrogate for HR. [37][38][39] Therefore, it was asserted that HRV parameters fundamentally increase with a decrease in HR and an increase in IBI for mathematical and physiological reasons. 16,[40][41][42] Boyett et al concluded that high HRs with steep pacemaker potentials produce relatively (little variability between successive heartbeats) small changes in IBI. In contrast, low HRs with flat pacemaker potentials produce relatively large changes in IBI. 16 Consequently, HRV parameters were corrected for the chronotropic state. 41,43,44 Here, we demonstrate that HRV is not simply a surrogate for HR or IBI in patients with AN. We observed dissociation between HRV parameters and IBI in patients with AN (Appendix A and Table 2).
We calculated Spearman's correlation coefficients for the IBI-LnHF power relationship for each patient and control subject. The fd parameter of LnHF power has been shown to quantify the magnitude of respiratory sinus arrhythmia. 45 The correlation between IBI and LnHF power was much weaker in patients with AN than in controls.
In eight of the 20 patients with AN, we even found a negative Spearman's correlation coefficient. A Spearman's correlation coefficient less than zero implies that most LnHF power values and corresponding IBIs are likely to be located at the descending limb of a quadratic HRV-IBI function. Consequently, HRV saturation was present in at least eight patients with AN at admission and at least seven patients with AN during refeeding treatment (Table 2).
Spearman's correlation coefficients were similar to those in controls in some patients with AN may be due to the differences in individual IBI-LnHF power relationships. Moreover, HRV saturation may only occur in patients and controls if a specific IBI is exceeded.
Consequently, the period in which distinct HRV saturation is detectable may be restricted to severe starvation phases, which some patients may not have experienced. Evidence of this may be that signs of HRV saturation partly disappeared during refeeding treatment (Table 2). During nutritional refeeding, Spearman's correlation coefficients increased in 11 of the 20 patients, implying a shift toward the ascending limb of the curve (Table 2).

| Relationship between IBI and HRV in patients with HCN4 mutations
In patients with the HCN4 c.1737+1G>T mutation, our HRV and IBI analysis revealed contradictory results. In two patients, a strong correlation between IBI and LnHF power was present, which confirmed the theory of HRV boost in HCN4 mutation carriers. 25 Two patients had a negative Spearman's correlation coefficient, which implied HRV saturation. In addition, the pNN50/IBI graph indi-

| Neural or nonneural mechanisms?
We revealed that HRV saturation might not result from sustained parasympathetic control of the sinus node, as stated by Goldberger et al. 23,46 HRV saturation also occurred in HCN4 mutation carriers; thus, we assume that HRV saturation is a phenomenon that becomes visible in all subjects with strong bradycardia. The level of bradycardia that is necessary for HRV saturation may be subject specific. Our meaning that less I f current is available at diastolic potentials. 48 Consequently, diastolic depolarization time is prolonged, and HR slows down. However, this effect cannot explain HRV saturation. One possible theory is that if cAMP insensitivity and low cAMP concentrations drastically prolong diastolic depolarization via I f , other rhythmic intrinsic pacemaker mechanisms may be less influenced by autonomic tone than the funny channel, become dominant, and intervene to prevent sinus arrest. [49][50]

CONFLICT OF INTEREST
The authors declare there is no conflict of interest. Reiner Buchhorn and Christoph Baumann (corresponding author) had full access to all the data in this study and take complete responsibility for the integrity of the data and the accuracy of the data analysis.

TRANSPARENCY STATEMENT
The lead author Reiner Buchhorn affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.

DATA AVAILABILITY STATEMENT
The datasets analyzed in this study are available from Reiner Buchhorn or Christoph Baumann on reasonable request.