From January 2004 to March 2006, 520 neonates were admitted to the NICU at Miechuo Medical Center. All of them were monitored with electrocardiogram (ECG) for heart rate (HR), respiration rate (RR), and with pulse oxymetry on the wrist or the feet for saturation of pulse oxymetry oxygen (SpO₂) and pulse rate (PR) throughout their stay in the NICU. Monitored physiological information was transformed as measurement variables at 10-second intervals by the Wave Achieving System (WAS-J; Philips Electronics Japan, Tokyo, Japan) through the local area network in the NICU. The data were recorded for 24 hours for the following postnatal periods: Period 1: days 0–3; Period 2: days 4–6; Period 3: days 7–13; and Period 4: days 14–21. Subjects with continuously disrupted data for more than 1 minute were excluded from the study. A total of 187 neonates (114 boys and 73 girls) were recorded from period 1 to period 4.

The NICU was maintained under a light-dark cycle. The light was dimmed (less than 30 lux) during the night from 21:00 pm to 07:00 am, while it was maintained at a higher level (300–580 lux) during the daytime. NICU staff also varied according to time of day: the number of attendants at night was one third that of attendants during daytime hours. Parent's visitations were allowed three times a day (11:00 to 12:00 in the morning, 14:00 to 15:00 in the afternoon, and 17:00 to 21:00 in the evening). Bathing and measurement of body weight were conducted daily in the morning. Medical examinations, such as blood sampling, radiography, or ultrasonography, were mostly provided in the morning if necessary.

Written informed consent was obtained from the parents, and the study was approved by the ethical committee of the institute. Demographics and health status information's were obtained from the medical records.

Physiological rhythmicity was analyzed for HR, PR, RR and SpO₂ with spectral analysis (periodogram) with SPSS 11.5 software (SPSS Inc. Chicago, IL), as previously reported 28. Briefly, 24 hours sessions were run in 10-second intervals and were aggregated into 1-minute time blocks. Periodogram analysis was performed with a time series of 1440 minutes (N = 1440 observations). The Fisher test was used to test the statistical significance of the cyclic components (N = 1440, α = 0.05) 2829. Among the significant cycles, the cycle with the largest peak in the periodogram was considered to be the dominant cycle for each time series data and was used for further analysis 28. All dominant cycles were confirmed by Fourier analysis, and further circadian cycles were confirmed by cosinor analysis with a significance of p < 0.05 by least square analysis (Figure 1). The amplitude, the distance between mesor and the highest value of the cosine curve, was calculated for each dominant cycle. In addition, an amplitude index was calculated as follows:

Amplitude index = amplitude ÷ mean of variables × 100.

Data were analyzed with SPSS and Statview. ANOVA was used to evaluate the differences between gestational age groups. The Pearson correlation coefficient was used to analyze the relationships between postconceptional age (PCA) and rhythmicity parameters. Univariate analysis using Mann-Whitney U-test for continuous variables or Fisher's exact test for categorical variables was used to compare clinical variables according to the development of physiological rhythmicity. A multiple logistic regression analysis was performed using a step-wise approach to determine the independent relationship of significant variables found in the univariate analysis.

The demographics of neonates are shown in Table 1. The median gestational age (GA) was 34 weeks (range: 23–42 weeks), and the median birth weight was 1968 g (range: 454–4132 g). Among these neonates, 9.1% were born at < 28 weeks of gestation age and 14.4% had birth weight of less than 1000 g. The median age at hospitalization was 0 day (range: 0–9 day) and the median duration of hospitalization was 32 days (range: 5–182 days). One hundred eleven neonates (59.4%) were intubated and 72 neonates (38.5%) received oxygen.

Results of the analyses of rhythmicity are summarized in Table 2. To ensure the accuracy of rhythmicity analysis, parameters missing more than 7% of total data were excluded from the analysis in each study. Among 461 time series recorded for each parameters, eligible samples were obtained in 304 for HR, 379 for PR, 372 for RR, and 383 for SpO₂ within the 4 periods. Among eligible samples, rhythmicity was observed in more than 90% of neonates in each period for HR, PR, RR and SpO₂ (Table 2). The percentage was not much lower (HR: 89%, PR: 90%, RR: 79%, SpO₂: 76%) after Bonferroni correction for multiple testing (p < 0.0001).

Without correction for multiple testing, circadian cycle (1440 minutes) was observed among 23.8% neonates in HR, 20% in PR, 27.8% in RR and 16% in SpO₂ in Period 1. Because many samples were excluded from HR analysis, and the percentage of eligible samples was consistently lower than for PR, further analysis of cardiac rhythmicity used PR instead of HR.

Rhythmicity was analyzed in four gestational age groups: < 28 wks, 28–32 wks, 33–36 wks, ≥ 37 wks. The distribution of circadian cycles in each gestational age groups and periods is summarized in Table 3. In PR, the percentage of circadian cycles was highest (40%) at <28 wks of GA, decreasing with GA, and lowest (14.3%) at ≥ 37 wks of GA in Period 1. A similar tendency was observed in each period in PR; however, there was no consistent tendency in percentages of circadian cycle in RR and SpO₂.

Amplitudes and amplitude indexes of all detected cycles in PR in each period are shown in Figure 2. An increase of circadian amplitude with gestational age was observed in PR. and significant differences were present among gestational age groups in all periods (Figure 2A). These changes were not observed in RR and SpO₂ (data not shown). Amplitude indexes showed similar tendency to amplitudes in PR (Figure 2B). There were no significant associations between cycles and amplitudes in any parameter in each period (data not shown).

In examining the relationship with postconceptional age (PCA), correlation of coefficient was performed using amplitudes and amplitude indexes in each period for all parameters. Amplitudes and amplitude indexes of PR were positively correlated with PCA in all four periods (Figure 3).

To determine whether clinical conditions may affect the emergence and development of rhythmicity, clinical factors were determined according to cycle length with circadian cycles (1440 minutes) or ultradian cycles (≤ 720 minutes). On univariate analyses in Period 1, circadian cycle (1440 minutes) was significantly associated (p < 0.05) only with oxygen administration at data sampling in PR (Table 4), while there were no significant associations in RR or SpO₂ (data not shown). In Periods 3 and 4 in PR, gestational age was found to be significantly associated with circadian cycle (p < 0.01) as well as with oxygen administration (p < 0.05). Neither gestational age nor oxygen administration qualified as an independent factor for existence of circadian cycle in multivariate logistic regression models. Clinical parameters were not associated with the existence of significant cycles in amplitude or amplitude index.