RESULTS

Circadian HR and HRV patterns showed evening sympathetic predominance and blunted nocturnal para - sympathetic modulation

Across the full 24-hour period, all 15 horses demonstrated consistent patterns of cardiac autonomic regulation. HR showed a clear diurnal pattern, with the highest value during daytime hours (38.88 ± 3.13 beats/min at 15:00–18:00) and the lowest value during nighttime (35.94 ± 2.57 beats/min at 24:00–03:00). Conversely, NN exhibited an inverse trend, as shown in Figures 2A and B.

Time-domain HRV parameters demonstrated relatively stable patterns across the 24-hour period (Figures 2C–F). SDNN values were lowest during 24:00–03:00 (68.87 ± 9.76) and highest during 18:00–21:00 (74.94 ± 8.59), whereas RMSSD values remained relatively consistent, with a peak at 06:00–09:00 (81.04 ± 20.00) and the lowest value at 03:00–06:00 (75.52 ± 15.55). SDANN decreased in the late morning, followed by a gradual increase at night. pNN100 showed a pattern similar to RMSSD. The overall trends of these parameters across the 24-hour cycle are illustrated in Figure 2.

Figure 2: Circadian rhythm of HR and time-domain HRV parameters, including NN, SDNN, RMSSD, SDANN, and pNN100. Data represent mean values across eight time segments throughout the 24-hour period.

Frequency-domain HRV parameters showed modest variation across the 24-hour period. VLF remained relatively stable, with only minor fluctuations throughout the day. LF power and total power peaked during the early evening (18:00–21:00), whereas lower values were observed during late nighttime. HF power exhibited relatively small changes across time intervals without a clear nocturnal elevation. The LF/HF ratio was highest during the evening (2.94 ± 1.31 at 18:00–21:00) and lowest during midday, indicating a transient shift toward sympathetic predominance during this period. The trends of each frequency-domain parameter are shown in Figure 3.

Figure 3: Circadian rhythm of frequency-domain HRV parameters, including VLF power, LF power, HF power, total power, and LF/HF ratio. Data represent mean values across eight time segments throughout the 24-hour period.

Overall, frequency-domain patterns were consistent with time-domain findings. RMSSD, pNN100, and HF showed similar temporal trends, whereas mixed autonomic markers (SDNN and LF) and long-term variability indices (SDANN and VLF) demonstrated comparable fluctuations. However, only modest circadian variation was observed, with a transient evening increase in sympathetic predominance and no clear nocturnal parasympathetic predominance. Detailed mean ± SD values for all parameters across time intervals are provided in Supplementary Table S1.

HR decreased at night, whereas autonomic balance showed minimal day–night variation

Overall, time-domain indices remained largely stable between daytime (06:00–18:00) and nighttime (18:00–06:00), with significant differences observed only for HR and NN. Daytime HR (37.57 ± 2.80 beats/min) was higher than nighttime HR (36.52 ± 2.38 beats/min, p = 0.004), consistent with a small but significant time effect in the mixed-model analysis of variance (p < 0.001, generalized eta-squared = 0.035). Similarly, NN was shorter during daytime (1605.77 ± 125.06 ms) than during nighttime (1649.58 ± 110.75 ms, p = 0.005), consistent with the overall time effect (p = 0.001, generalized eta-squared = 0.030). In contrast, other HRV parameters (SDNN, RMSSD, SDANN, and pNN100) showed no significant day–night differences (all p > 0.05), indicating attenuated circadian modulation. These comparisons are illustrated in Figure 4.

In the frequency-domain, no parameters differed significantly between daytime and nighttime (all p > 0.05). VLF and LF power were slightly lower during daytime than during nighttime. HF power was higher during the daytime than at night, but the difference was not significant. Total spectral power remained comparable between the two periods. The LF/HF ratio increased non-significantly from 2.22 ± 0.86 during daytime to 2.62 ± 1.08 during nighttime. The frequency-domain comparison between daytime and nighttime is illustrated in Figure 5.

Figure 4: Comparison of HR and time-domain HRV parameters between daytime and nighttime. Parameters include NN, SDNN, RMSSD, SDANN, and pNN100. Error bars indicate the standard error of the mean, and significance levels between daytime and nighttime are indicated by **p < 0.01 versus daytime.

Figure 5: Comparison of frequency-domain HRV parameters between daytime and nighttime. Parameters include VLF power, LF power, HF power, total power, and LF/HF ratio. Error bars indicate the standard error of the mean.

Compared with temperate-climate studies that reported pronounced nocturnal parasympathetic elevation, horses in this tropical setting showed only modest HR reduction with blunted changes in RMSSD, pNN100, and HF power, indicating attenuated circadian modulation under these conditions. Detailed mean ± SD values for all parameters during daytime and nighttime are provided in Supplementary Table S2.

Aging horses showed impaired HRV compared with younger horses

To examine age-related differences, HR and HRV parameters were averaged over the 24-hour recording period and compared among groups. Mixed-model analysis of variance revealed significantly large group effects for multiple HRV parameters, including SDNN, RMSSD, pNN100, HF, total power, and LF/HF ratio (all p < 0.05), indicating age-related differences in autonomic regulation. No significant group × time interactions were observed for any parameter.

Time-domain analysis showed minimal differences in HR and NN among groups. In contrast, short-term variability indices demonstrated clear age-related effects. SDNN was significantly lower in Group 3 than in Group 1 (p = 0.025) and Group 2 (p = 0.011), whereas Groups 1 and 2 did not differ. A similar pattern was observed for RMSSD, with Group 3 showing significantly lower values than Group 1 (p = 0.003) and Group 2 (p = 0.004), with no difference between Groups 1 and 2. pNN100 showed the clearest separation, with Group 3 having significantly lower values than both Group 1 and Group 2 (both p = 0.021), whereas Groups 1 and 2 did not differ. In contrast, long-term variability, as assessed by SDANN, showed no significant differences among groups. These comparisons are illustrated in Figure 6.

Figure 6: Comparison of HR and time-domain HRV parameters among the three age groups: Group 1 (4–7 years), Group 2 (8–14 years), and Group 3 (15–20 years). Parameters include NN, SDNN, RMSSD, SDANN, and pNN100. Error bars indicate the standard error of the mean, and significance levels between groups are indicated by *p < 0.05 and **p < 0.01.

When frequency-domain HRV parameters were compared among groups, VLF power remained relatively similar across all age groups, whereas LF power showed only modest differences. Conversely, HF power showed the clearest group separation, with Group 3 showing significantly lower values than Groups 1 and 2 (both p = 0.0159). Total spectral power followed a similar pattern, with Group 3 showing significantly lower values than Group 1 (p = 0.039) and Group 2 (p = 0.018). The LF/HF ratio increased progressively with age, with Group 3 showing significantly higher values than Group 1 (p = 0.005) and Group 2 (p = 0.017). These findings demonstrate reduced parasympathetic activity and a shift toward sympathetic predominance in older horses. The across-group comparisons are illustrated in Figure 7.

These findings indicate that older horses (15–20 years) displayed markedly reduced short-term vagal activity and elevated LF/HF ratios, extending previous age-related findings by demonstrating that impaired HRV occurs even at rest under tropical conditions. Detailed numerical values are provided in Supplementary Table S3.

Environmental parameters were associated with HRV parameters

Feels-like temperature and light intensity were significantly higher during daytime, whereas humidity increased at night. PM2.5 and AQI showed relatively minor day–night differences but remained highly variable. Environmental values during daytime, nighttime, and the 24-hour period are shown in Table 3.

Figure 7: Comparison of frequency-domain HRV parameters among the three age groups: Group 1 (4–7 years), Group 2 (8–14 years), and Group 3 (15–20 years). Parameters include VLF power, LF power, HF power, total power, and LF/HF ratio. Error bars indicate the standard error of the mean, and significance levels between groups are indicated by *p < 0.05 and **p < 0.01.

Environmental parameter | Day time (mean ± SD) | Night time (mean ± SD) | p-value
Temperature (°C) | 28.65 ± 1.45 | 26.43 ± 1.09 | <0.001
Humidity (%) | 54.08 ± 2.86 | 61.85 ± 5.04 | <0.001
Feels-like temperature (°C) | 29.46 ± 2.04 | 27.09 ± 1.73 | 0.008
Light intensity (lux) | 6214.55 ± 751.84 | 5.09 ± 5.47 | <0.001
AQI | 134.33 ± 30.87 | 133.04 ± 25.08 | 0.99
PM2.5 (µg/m³) | 54.12 ± 17.66 | 52.03 ± 14.94 | 0.95

Correlation analysis showed no significant associations between temperature or feels-like temperature and any HRV parameter. However, humidity was significantly negatively correlated with HR (r = −0.230, p < 0.001) and positively correlated with NN (r = 0.230, p < 0.001). Light intensity showed a significant positive correlation with HR (r = 0.166, p < 0.001) and negative correlations with NN (r = −0.166, p < 0.001) and SDANN (r = −0.111, p = 0.029).

Both AQI and PM2.5 demonstrated consistent associations with cardiorespiratory parameters. Specifically, both variables were negatively correlated with HR (AQI: r = −0.318, p < 0.001; PM2.5: r = −0.320, p = 0.002) and positively correlated with NN (AQI: r = 0.318, p < 0.001; PM2.5: r = 0.320, p = 0.003). In addition, the respiratory cycle was reduced with increasing air pollution levels (AQI: r = −0.323, p < 0.001; PM2.5: r = −0.332, p < 0.001). AQI and PM2.5 were also significantly negatively correlated with parasympathetic-related indices, including RMSSD (AQI: r = −0.239, p < 0.001; PM2.5: r = −0.235, p < 0.001), pNN100 (AQI: r = −0.238, p < 0.001; PM2.5: r = −0.231, p < 0.001), and HF power (AQI: r = −0.300, p < 0.001; PM2.5: r = −0.295, p < 0.001).

Conversely, both AQI and PM2.5 showed significant positive correlations with LF/HF ratio (r = 0.318, p < 0.001 for both), indicating a shift toward sympathetic predominance. Total power was also negatively associated with AQI (r = −0.323, p < 0.001) and PM2.5 (r = −0.332, p < 0.001).

Positive correlations between PM2.5/AQI and sympathetic predominance, reflected by LF/HF ratio, provide direct evidence linking ambient air pollution to impaired autonomic function in horses. Correlations between environmental variables and HRV parameters, including respiratory cycle, are summarized in Figure 8 and Supplementary Table S4.

Figure 8: Correlation between environmental parameters and HRV indices. Environmental parameters include temperature, humidity, feels-like temperature, light intensity, AQI, and PM2.5. HRV parameters include HR, NN, SDNN, RMSSD, SDANN, pNN100, VLF power, LF power, HF power, total power, and LF/HF ratio. Respiratory cycle is included. Statistical significance levels: ● = p < 0.05, ●● = p < 0.01, and ●●● = p < 0.001.