Heart rate variability in newborn foals and its association with illness: a pilot study

Abstract This study aimed to investigate differences in heart rate variability (HRV) between healthy and sick neonatal foals with a variety of diagnosis and to verify whether some HRV parameters could be associated with sepsis and/or survival. Twenty-one foals were included in the study: nine were healthy and 12 were sick. Retrospectively, sick foals were divided into sub-groups (non-septic vs. septic and non-survivor vs. survivor) for statistical analysis. Heart rate was recorded daily for 20 min and a 5-min period was used for heart rate variability analysis. Data on HRV variables were analysed using a linear model. Least-square means with their standard errors were reported, and treatment effects were declared significant at p < .05. To isolate the group that differed from the others, Tukey’s test was used as a multiple comparison procedure. Pearson’s correlations between variables were calculated, to estimate their relationship. Standard Deviation of the RR intervals and Standard Deviation 2 by the Poincaré plot was significantly lower in sick foals compared with healthy subjects, but there were no differences between survivors and non-survivors. Healthy foals presented a significantly lower Sample Entropy and heart rate compared with sick foals. HRV analysis is a non-invasive, rapid, and economical measurement tool that can be added to other parameters to improve the accuracy of predicting in-hospital mortality in foals. Further studies should be conducted with a larger sample of foals to confirm these findings and to evaluate the clinical usefulness of HRV analysis during neonatal diseases. Highlights This study investigated differences in heart rate variability (HRV) between sick and healthy neonatal foals Standard Deviation of the RR intervals was significantly lower in sick compared with healthy foals, but there were no differences between survivors and non-survivors. HRV analysis is a non-invasive, rapid, and economical tool that can be added to other parameters to improve the accuracy of predicting mortality in foals.


Introduction
Sepsis in neonatal foals remains a leading cause of morbidity and mortality despite the improvement in survival that has been reported over the last 25 years (Hytychov a and Bezd ekov a 2015; Taylor 2015; Wong et al. 2018). Identifying the right biomarkers is key to an earlier and accurate diagnosis of sepsis in the horse (Pusterla et al. 2006;Bonelli et al. 2015aBonelli et al. , 2015bTaylor 2015;Bonelli et al. 2017). The survival of septicaemic foals varies within a range of 45-81% (Corley and Furr 2003;Slack et al. 2005;Castagnetti and Veronesi 2008). Since intensive care for newborn foals is very expensive, the early prognosis for survival on admission or during the first few hours of hospitalisation should be considered to identify neonates with a higher chance of survival (Castagnetti and Veronesi 2008;Wong and Wilkins 2015).
Heart rate variability (HRV) is a non-invasive measurement that can be used to investigate the balance of sympathetic and vagal activity (Task Force of the European Society of Cardiology, North American Society of Pacing and Electrophysiology 1996). HRV is measured by determining the continuously changing time interval between successive heartbeats (R-R intervals). It is based on the antagonistic oscillatory influences of the sympathetic and parasympathetic nervous systems on the nodus sinuatrialis (Stucke et al. 2015). HRV is increasingly used in animal research to analyse changes in the sympathovagal balance related to disease, psychological and environmental stressors, or individual characteristics, such as temperament and coping strategies (Von Borell et al. 2007;Stucke et al. 2015).
In human medicine, HRV has also been used to predict prognosis in septic patients (Bonjorno Junior et al. 2019;Prabhakar et al. 2019), and several studies have reported that a decrease in HRV indices is associated with higher mortality in critically ill patients (de Castilho et al. 2018). HRV analysis has been used especially in neonatal medicine to predict mortality in infants (Beuch ee et al. 2009;Lake et al. 2014).
Abnormal HRV measurements, such as decreased approximate entropy, decreased sample entropy, reduced variability, and transient decelerations have been significantly associated with sepsis or sepsis-like illness in premature neonates (Beuch ee et al. 2009). Early in the course of neonatal sepsis, there are reduced heart rate variability and transient decelerations similar to the foetal distress that appears before the clinical diagnosis of sepsis or systemic inflammatory response syndrome (Griffin et al. 2005;Bohanon et al. 2015). In animal models, HRV assessment may provide insights into the acute effect of sympathetic and cholinergic anti-inflammatory pathways (Zila et al. 2015). According to some clinical studies, HRV analysis might play an important role in the continuous monitoring of severe infection (Fairchild et al. 2009;Jarkovska et al. 2015). HRV has been evaluated in newborn foals (Nagel et al. 2012;Nagel et al. 2015) however, to the best of the authors' knowledge, no studies have been conducted to analyse HRV in sick or septic foals.
The aim of this study was thus to investigate differences in HRV between healthy and sick neonatal foals with a variety of diagnosis. We wanted to verify whether some HRV parameters could be associated with sepsis or could represent early markers of survivals in neonatal foals referred to the hospital.

Materials and methods
A total of 21 foals were included in the study. Ethical approval (n 2825/2014) by the Ethics Committee on Animal Experimentation of the University of Pisa and an owner's written consent were obtained. Nine of the foals were healthy Standardbreds, that had been born on the same stud farm, had undergone similar management conditions, and were used as a control group. The following inclusion criteria were set for the 'control group': (1) normal gestation length (>320 days) (Lester 2011); (2) unassisted delivery; (3) mares treated for gastrointestinal parasites and vaccinated for equine influenza, tetanus, and equine herpes virus-1; (4) Apgar Score !7 within 5 min of birth (Stoneham 2006); (5) good passive transfer of immunity at 24 h of age (Immunoglobulin G ! 800 mg/dL) 1 ; (6) righting reflex immediately after birth and sucking reflex within 20 min, sternal recumbence within 2 min, standing position within 120 min, first suckling within 180 min (Stoneham 2006); and (7) normal at physical examination throughout the study period.
In addition, 12 sick client-owned foals were included. All the foals were referred with different complaints to two different veterinary teaching hospitals providing secondary health care during one year. The foals were managed under similar circumstances in both the VTHs. Using data from clinical history, physical examination, and clinicopathological analysis collected on admission, a sepsis score was calculated for each foal (Wong et al. 2018). Based on the sepsis score and results of blood culture, each foal was assigned to 1 of 2 patient groups: a sick non-septic group (sepsis score <7 and negative results of bacteriological culture of blood; n ¼ 6 foals) or a septic group (sepsis score !7 with or without positive results of bacteriological culture of blood or sepsis score <7 with positive results of bacteriological culture of blood; n ¼ 6 foals). Retrospectively, the foals enrolled in the study were grouped according to the outcome into survivors (n ¼ 7) and non-survivors (n ¼ 5). The foals were considered survivors if they had been discharged from the hospital, while they were considered non-survivors if they died or had been humanely euthanised due to their severe medical prognosis and not for economic reasons. For details regarding all foals included in this study, see Table 1.
Heart rate was recorded in all foals using a heart rate monitor 2 (connected by a chest belt, as previously described (Nagel et al. 2012;Nagel et al. 2015). In healthy foals, heart rate was recorded daily for 20 min starting at 24 h of life for 12 consecutive days. The heart rate monitor belt was placed on the foals while they were in their usual stall with their dam. A 15-min period of familiarisation preceded the start of the heart rate recording so that the foal could regain calm. Using a closed-circuit television and a chronometer, the same (VV) operator recorded from outside whether the foal was awake or asleep, standing or lying down, or if it was running around the stall. This was used to select 5 min recordings for HRV analysis during which the foals were calm, but not sleeping. Data were always collected between 5:00 and 8:00 pm.
In sick foals, the heart rate was recorded daily for 20 min starting within 24 h from admission until discharge or death/euthanasia. As with the healthy subjects, the Polar belt was strapped to the foals in the hospital box in the presence of the dam. The recordings were performed after a 15-min period of acclimation, between 5:00 and 8:00 pm, without interfering with clinical procedures and avoiding medication or feeding times. Usually, the foals were lying down, however, if they were standing, an operator recorded its activity in the stall as with the healthy foals.
The data were entered into a computer and underwent automatic correction of artefacts using the available software 3 . A 5-min period from each 20 min recording was selected, in which the foals were at rest, also taking into account signal quality. Heart rate variability during this 5 min was subsequently analysed with a free software program designed for HRV analysis 4 . The variables of HRV taken into account were: Heart Rate (HR; bpm); Standard Deviation of the RR intervals (SDRR; ms); Square root of the mean of the sum of the squares of differences between consecutive RR intervals (RMSSD; ms); Geometric Standard Deviations by Poincar e plot (SD1 and SD2; ms) and Sample Entropion (SampEn).
Data on HRV variables were assessed for distribution using the Shapiro-Wilk W test. Data relating to HRV variables were assessed by comparing first healthy, surviving, and non-surviving foals. Subsequently, healthy, septic, and non-septic foals were compared. Data on HRV were analysed by an ANOVA, to evaluate if there are significant differences between foals, classifying them as healthy, survivors, and non-survivors. Subsequently, the same statistical analysis was performed comparing healthy, septic, and non-septic foals.
In both cases, data on HRV were analysed using the following factorial mixed model 5 : where: y ijz ¼ HRV variables: HR, SDRR, RMSSD, SD1, SD2, SampEn; G i ¼ fixed effect of the ith group based on the following classifications of horse status, which were considered independently: (1) healthy vs. survivors vs. no survivors; (2) healthy vs. septic vs. no septic. T j ¼ fixed effect of the jth day of the survey (12 days) H z [G i ] ¼ random effect of the zth horse nested in the group. e ijz ¼ random residual Least-square means with their standard errors were reported, and treatment effects were declared significant at p < .05. To isolate the group that differed from the others, HSD Tukey's test was used as a multiple comparison procedure. Finally, Pearson's correlations between variables were calculated, to estimate their relationship.

Results
Foals included in the study are described in Table 1. In the control group, HR was recorded for 11 consecutive days starting from 24 h of life. In the study group, 4 out of 5 non-survivor foals were euthanised, while one died naturally. In the survivors, the HR was recorded until discharge for 11 days in three foals, for 8 days in two foals, and 4 days in the other two subjects. In non-survivors, the heart rate was recorded until death in each foal for 11, 6, 4, and 1 day, respectively.
HRV parameters for healthy and sick foals, divided into survivor/non-survivors and septic/non-septic are reported in Tables 2 and 3. An analysis of the differences over time revealed that non-survivor foals had a higher HR on all the days except on admission, while the survivor subjects had similar values to healthy foals except on days 7 and 8 when their HR was higher. SDRR and SD2 were significantly lower in sick foals compared with healthy subjects, but there were no differences between survivors and non-survivors. However, after 7 days, in survivor foals, these parameters were similar to those of the healthy groups.
Healthy foals presented a significantly lower SampEn (0.82 vs. 1.19 ms) and higher SDRR (47.27 vs. 25.51 ms) and SD2 (65.72 vs. 34.24 ms) compared with septic foals. Non-septic subjects showed intermediate values for all the variables considered. No significant differences were observed for HR, RMSSD, and SD1 in any of the groups.
An analysis of the differences over the time period in healthy foals revealed that the HR was lower from the 7th day onwards compared with sick foals, but no other differences were observed. Furthermore, no differences were detected between septic and non-septic patients.

Discussion
Our study describes an HRV analysis on neonatal healthy and sick foals with a variety of clinical diagnoses. We found a statistically significant effect of illness on SDRR, SD2, and SampEn. However, no differences were found between survivors and non-survivors, or between septic and non-septic groups on admission.
Our results on HRV parameters in healthy foals are similar to those already described by Nagel and colleagues (Nagel et al. 2015). Data were collected at the same time of the day to prevent circadian interference.
The decrease in SDRR and SD2 in sick foals indicates the decreased long-term variability of cardiac activity (Von Borell et al. 2007). Similarly, in human infants, short-term variability has not been found to change in the presence of sepsis, while long-term variability is significantly lower (Bohanon et al. 2015).
The depression of HRV in conditions of systemic inflammation has been repeatedly shown in humans (Griffin et al. 2005;Prabhakar et al. 2019), experimental animals (Jarkovska et al. 2015;Zila et al. 2015), and adult horses (McConachie et al. 2016a(McConachie et al. , 2016bVitale et al. 2020). As no changes were observed in RMSSD and SD1, we cannot associate the decrease in HRV in sick animals with a decreased parasympathetic tone but rather with a possible autonomic dysfunction (Vitale et al., 2020). Bonjorno Junior et al. (2019 already speculated that the reduced HRV in sepsis may indicate early autonomic dysfunction due to attenuation of the adrenergic response at the cardiomyocyte level. When considering the HR in foals we must take into account that the physiology of the cardiovascular system is slightly different from those of adults. Foals cannot readily increase their stroke volume, thus, to Table 3. HRV parameters presented as mean ± SEM for healthy, non-septic, and septic foals and results of the statistical comparison.  maintain an adequate cardiac output they have to increase their HR (Carr, 2014). This might influence the HRV which is difficult to quantify at this point due to the low number of cases included. Nevertheless, HR was not found different between septic, non-septic and healthy foals while other HRV parameters were.
Our study demonstrates that also in foals, HRV is different according to the state of health. The lack of significant differences between the different groups of sick foals during the initial 24 h suggests that HRV analysis performed on admission may not be useful when predicting survival or septicaemia. It is also true that the number of sick foals enrolled in this study was low which might have been a bias for the statistical analysis.
A recent study reported that when the HRV responses to porcine moderate sepsis without organ failure and to a porcine progressive lethal septic shock were compared, no differences were found (Jarkovska et al. 2015). This probably indicates that the sensitivity of HRV to systemic inflammation is high but, on the other hand, HRV does not allow the illness to be scaled. A reliable staging of inflammatory processes likely requires an integrative approach with additional parameters, such as clinical, hematological, and biochemical variables (Peek et al. 2004;Bohanon et al. 2015;Jarkovska et al. 2015;Wong et al. 2018).
Although the accuracy of the sepsis score developed by Brewer and Koterba (1988), has been questioned (Corley and Furr 2003;Slack et al. 2005;Castagnetti and Veronesi 2008;Wong and Wilkins 2015;Wong et al. 2018), it remains the best predictor of survival, with the survival rate decreasing as the sepsis score increases (Peek et al. 2004;Wilkins 2018;Wong et al. 2018).
Nevertheless, a variety of 'new vital signs', such as tissue oxygenation and HRV, have been emerging in medicine. These can be associated with blood analysis and other vital parameters to obtain more detailed information on the health status of the patient, and thus produce a more reliable prognosis (Bohanon et al. 2015). In addition, in studies on humans, several data points are usually obtained to predict the risk of death in these patients (Bonjorno Junior et al. 2019). The values of SDRR and SD2 in the sick survivors became similar to those of the healthy foals starting from the seventh day of observation, while in the non-survivor group the values remained higher. These results support the hypothesis that HRV parameters return to normal values in foals that recover from illness.
When comparing foals according to their septic status, the changes observed were similar. This is probably because the majority of septic foals died (4/6) and the majority of non-survivors were septic (4/5).
Several limitations need to be considered in the interpretation of the current results. Firstly, the recruitment of patients was limited as it was related to the equipment and manpower available. Secondly, the ages of sick foals are quite different, and this could have influenced the results. Thirdly, although we selected our HRV parameters based on a literature review, there may have other suitable HRV variables for a comparison between groups.

Conclusions
In conclusion, HRV analysis is a non-invasive, rapid, and economical measurement tool, which requires a common heart rate monitor which is nearly always freely available in a referral equine hospital, as the software for HRV analysis. Adding HRV to other commonly used parameters may improve the accuracy of predicting in-hospital mortality in foals. Further studies should be conducted with a larger sample of foals to confirm these findings and to evaluate the clinical usefulness of HRV analysis during neonatal diseases.