Dogs for the study were selected from those brought to the Department of Dog & Cat Pediatric Hospital in Tokyo, Japan. Healthy dogs were chosen from among those admitted to the hospital for pet boarding or temporary dog care. Additionally, experimental Beagle dogs from our laboratory at Kitayama Labes in Nagano, Japan, were enrolled between August 2018 and January 2023. G*Power (The G*Power Team, G*Power 3.1.9.7 version, Germany) was used to calculate sample sizes. To adapt a paired t-test, we set α = 0.05, 1-β = 0.8, and effect size (d-family) = 0.5. The sample size was calculated to require at least 34 cases, so efforts were made to collect dogs for the study as a target value. Healthy dogs with normal physical examination and echocardiography were selected as the test subjects for this study. Dogs with obvious pain on physical examination were excluded. Puppies weighing less than 1.0 kg and too small to be fitted with a Holter electrocardiograph were also excluded from the study. All dog owners provided their consent for their pets to participate in the study. Experimental dogs were handled according to the guidelines established by the Institutional Animal Care and Use Committee of the TUAT (Approval number: R05-140).

The Holter electrocardiograph used in this study was manufactured by NIHON KOHDEN CORPORATION (RAC-5203, Japan). Prior to electrode placement, the dogs’ thoraxes were shaved vertically from the sternal scape to the xiphoid process and horizontally around the fifth and sixth intercostal spaces and cleaned with alcohol. Disposable ECG electrodes (XUNDA BRAND, China) were positioned using the M-X induction method and the R-L induction method, which is perpendicular to the M-X method. Subsequently, the induction cords were attached to the electrodes, CH1- (red) electrodes placed on the manubrium of the sternum, CH1+ (yellow) on the xiphoid process, CH2- (orange) on the right 5th ~ 6th intercostal space, CH2+ (blue) on the left 5th ~ 6th intercostal space, and a ground electrode (black) in the middle (24) (Figure 1). Both M-X and R-L leads were recorded. To secure the Holter recorder and leads to the dog, an elastic bandage and a vest utilized (Figure 2). Holter electrocardiogram (ECG) measurements were conducted by veterinarians and clinical laboratory technicians in the TUAT laboratory for 48 h period, during which the animals were allowed free movement within the enclosure.

For the 48-h Holter ECG measurements, the period from 12 p.m. on the first night to 12 a.m. on the second night was designated for 24-h HRV, while the period from 12 p.m. on the first night to 8 a.m. on the second day was earmarked for analysis as sleep HRV. HRV analysis was conducted using the Juntendo University algorithm with MATLAB (MathWorks, R2022a, United States). Both traditional linear analysis and a newer nonlinear analysis method were employed under the following conditions.

Statistical analyses were conducted using R software (R Development Core Team, version 4.1.0, New zealand). The significance level set at p < 0.05. A difference test was adapted to examine the difference between 24-h HRV and sleep HRV. Normality was confirmed using the Shapiro–Wilk test. If the distribution followed a normal distribution, a paired t-test was used. Conversely, if the distribution did not follow a normal distribution, the Wilcoxon signed-rank sum test was applied. Parametric data are presented as means and standard deviations, with 95% confidence intervals also calculated. Nonparametric data are presented as median and interquartile range. The effect size was calculated using R-family. HRV circadian rhythm times were quantified.

A total of 30 dogs were included in the present study, comprising 23 females and 7 males, with ages ranging from 2 months to 8 years (mean [standard deviation], 1.4 [1.6]), and weights ranging from 1.7 to 9.3 kg (5.6 [2.9]). The breeds included Beagles (n = 16), Chihuahuas (n = 4), Miniature Schnauzers (N = 2), Mongrels (n = 3), Miniature Pinscher (n = 1), Yorkshire Terrier (n = 1), Maltese (n = 1), Border Collie (n = 1), and Miniature Dachshund (n = 1). No abnormal rhythm findings that would affect HRV were detected on the Holter ECG.

Linear HRV data were summarized in Table 1. Frequency analysis nHF was significantly higher in sleep HRV compared to in 24-h HRV (mean [standard deviation], 95% confidence interval, p value, r-family: 0.24 [0.057] vs. 0.23 [0.045], 0.006–0.031, p = 0.005, r = 0.49). Meanwhile, there was no statistically significant difference in LF/HF between 24-h HRV and sleep HRV (0.87 [0.38] vs. 0.84 [0.49], −0.11-0.15, p = 0.72, r = 0.067). In time domain analysis, both SDNN and RMSSD were significantly higher in sleep HRV than in 24-h HRV (SDNN: 179.7 [66.9] vs. 156.6 [53.2], 14.5–31.7, p < 0.001, r = 0.71) (RMSSD: 187.0 [74.0] vs. 165.4 [62.2], 13.2–30.0, p < 0.001, r = 0.70).

Nonlinear HRV data were summarized in Table 1. In the geometric analysis, both SD1 and SD2 were significantly higher sleep HRV than 24-h HRV (SD1: 132.4 [52.4] vs. 117.1 [44.0], 9.3–21.1, p < 0.001, r = 0.70) (SD2: 215.0 [80.5] vs. 185.9 [62.0], 17.6–40.6, p < 0.001, r = 0.69). However, there was no significant difference in SD1/SD2 between 24-h HRV and sleep HRV (median [interquartile range], 0.63 [0.59–0.66] vs. 0.60 [0.54–0.65], p = 0.43, r = 0.041). Additionally, DFA in fractal analysis showed no statistically significant difference between 24-h HRV and sleep HRV (0.71 [0.11] vs. 0.71[0.12], −0.024-0.040, p = 0.62, r = 0.092).

An average example of circadian rhythm elements for HRV indicators is shown in Figure 3, while a representative example is illustrated in Figure 4. The parasympathetic indicators SDNN (mean [standard deviation]: 6.42 [5.07]), RMSSD (7.45 [6.41]), SD1 (7.45 [6.41]), and SD2 (5.8 [3.82]) exhibited positive peaks during sleep HRV. However, the positive peak for nHF (9.92 [8.06]), a parasympathetic index, was observed to fall outside the range defined as sleep HRV.

One of the main objectives of the present study was to compare 24-h HRV with sleep HRV to delineate the differences. The findings of the present study suggest that the parasympathetic indices nHF, SDNN, RMSSD, SD1, and SD2 predominantly reflect parasympathetic activity during sleep. Moreover, as the positive peaks of the circadian rhythm elements of the parasympathetic indices fall within the range defined as HRV during sleep, HRV during sleep can serve as an indicator of the peak parasympathetic activity during the day. These results support the hypothesis that differences exist between 24 h HRV and sleep HRV.

Previous studies in humans have suggested that parasympathetic activity increases during the night with the circadian rhythm components of the HRV parasympathetic index exhibiting a positive peak during nighttime (34). The data presented in the present study suggest that sleep HRV may be useful metric in dogs, as a valuable analytical tool for early detection and prognosis of various diseases, as they exhibit similar changes to those observed in humans. This study distinguishes itself from previous studies by utilizing circadian rhythms to identify the maximum peak of circadian rhythm elements of parasympathetic indices within a 24-h period focusing specifically on the measurement range of sleep HRV. Rasmussen et al. (35) defined sleep HRV as the period starting from 30 min after the dog enters sleep and extending to 6 h. Blake et al. (28) also defined resting HRV as the period from 0:00 to 6:00 and activity HRV as the period from 12:00 to 18:00. Since both measurements are based on activity records, the measurement times are back and forth. Therefore, it would enhance the external validity of the results if the peak time of the HRV circadian rhythm item could be utilized as a reference when evaluating the HRV index.

The nHF observed in the frequency analysis was consistent with expectations, showing high values during HRV in sleep. However, the circadian rhythm item of the parasympathetic index deviated from expectations by falling outside the range defined in this study as HRV during sleep. Identifying the peak time of parasympathetic activity proves challenging even when utilizing circadian rhythms due to significant individual variation in nHF and the discrepancy between the average peak derived from each case and the peak calculated from the average curve. nHF demonstrates respiratory variability due to interference from the respiratory center and reflexive input from the periphery, which is transmitted to the sinus node to become HF. Therefore, it is plausible that nHF may not be considered a pure indicator of the cardiac vagus nerve activity (20, 36). SDNN in time domain analysis is known to be influenced by the duration of analysis. In short-term recordings, the primary source of SDNN variability is parasympathetically mediated RSA, whereas in 24-h recordings, sympathetic nerves contribute significantly to SDNN (20, 31). Therefore, it is expected that the difference in analysis time between 24-h HRV and 1-h sleep HRV would impact SDNN values. RMSSD is considered a superior indicator of parasympathetic activity compared to SDNN. However, RMSSD is less affected by respiration than SDNN, but more influenced by RSA (37). RSA also depend on a number of control mechanisms due to interference from cardiovascular centers in the medulla oblongata, the degree of lung distension, and reflexive input from right atrial wall distension (38). Therefore, interpretations other than cardiac vagal activity must be considered, but this potential influence is rarely taken into account in linear analysis methods (39). Consequently, it was hypothesized that dogs with physiological respiratory arrhythmias might be affected by this phenomenon. Linear analysis of HRV during sleep in dogs may be susceptible to several biases, including those related to respiration, duration, and respiratory arrhythmia. Therefore, utilizing linear analysis for HRV measurement in dogs may not be an appropriate analytical method.

On the contrary, nonlinear analysis is an analytical method that circumvents the drawbacks associated with linear analysis. While linear analysis entails examining the time series of RR intervals obtained. Life is inherently nonlinear, meaning that the relationship between variables cannot be plotted as a straight line. Nonlinear analysis serves as a method of assessing the unpredictability of a time series and reveals correlations when the complexity arises from the same underlying process (26). In humans, nonlinear analysis has attracted attention due to its potential to predict the onset of heart failure (4). Additionally, HRV measurement during sleep, which is less susceptible to environmental factors compared to daytime measurements, has been proposed (22). Therefore, considering that SD1 and SD2 in this study were nonlinear analyses and showed significant increases during sleep compared to the 24-h period, and that the maximum peak of circadian rhythm items occurred during sleep, we believe that future investigations into heart disease in dogs utilizing sleep HRV and nonlinear analyses could be advantageous for early detection of heart disease mirroring approaches used in humans.

One limitation of this study is the lack of consideration for breed differences, as well as the wide age range of the subjects, spanning from 2 months to 8 years. In humans, studies such as those by Bonnemeier et al. (40) have shown that HRV indices experience the most significant decline between the ages of 20 and 30. Additionally, Almeida-Santos et al. also reported that RMSSD decreases between the ages of 40 and 60, followed by an increase after the age of 70 (41). Therefore, age may have influenced the findings of the present study. As a perspective for future studies to clarify the effects of breed and age on HRV in dogs, it may be necessary to equalize breeds or differentiate between small, medium, and large dogs, or to conduct HRV studies by age stratification. Another limitation is the inclusion of brachycephalic breeds. It has been suggested that brachycephalic breeds may exhibit higher cardiac vagal activity compared to non-brachycephalic breeds (42). Furthermore, the heterogeneity of the rearing environment of the experimental animals is another limitation. If future studies are conducted with domestic dogs living in human households, the circadian rhythm may be influenced by the human life rhythm, potentially impacting HRV measurements. In practice, we are investigating the early detection of doxorubicin-induced myocardial damage. If myocardial damage is detected before irreversibility, it can be treated. Sleep HRV may detect smaller myocardial changes.