Circadian rhythm development in preterm infants. The role of postnatal versus postmenstrual age

Abstract

Background, Aims:

Circadian rhythm maturation may be disturbed in premature infants undergoing neonatal intensive care. We used continuous heart rate recordings across the entire neonatal intensive care period to study circadian rhythm development in preterm infants and to evaluate the roles of postmenstrual (PMA) versus postnatal age (PNA).

Materials and Methods

The circadian rhythm was calculated using a cosine fit of heart rate. The circadian rhythm amplitudes were averaged weekly and studied relative to PMA and PNA using the linear mixed effects models, adjusting for clinical variables that could affect the heart rate. The daily circadian rhythms were used to create grand averages for PMA groups: ≤31, 32–35, and >35 weeks, and for PNA groups: ≤30, 31–60, and >60 days.

Results

Sixty-six infants were evaluated as part of an ongoing prospective study with gestational ages between 23 – 36 weeks. The PMA (1.47×10⁻² beats per minute (bpm)/week, P=2.07×10⁻⁸) and PNA (1.87×10⁻² bpm/day; P=1.86 × 10⁻⁶) were significantly associated with the circadian rhythm amplitude independent of covariates. Infants ≤31 weeks’ PMA and ≤30 PNA, the phase of circadian rhythm amplitude grand averages showed a peak at night and a nadir during the day. Hereafter the circadian rhythm phase reversed to that established for mature individuals. The highest circadian rhythm amplitudes present >35 weeks’ PMA and >60 days PNA.

Conclusions

Our results indicate circadian rhythm matures with advancing gestation. The reversed circadian rhythm phase during the early postnatal period could be due to premature exposure to the ex-utero environment and warrant further study.

Introduction

Each year almost half-million infants in the USA [1] and over 13 million infants worldwide [2] undergo part or all of their third-trimester brain development in an artificial extrauterine environment vastly different from their intrauterine counterparts. This likely plays a major role in the prevalence of neurodevelopmental, behavioral and psychologic long-term adversity experienced by survivors [3]. Circadian rhythm is an endogenous biorhythm that is entrained in the fetus by the mother’s circadian rhythm [4]. Animal models and studies of human fetuses and preterm infants have shown that this endogenous rhythm develops around 30 weeks of gestation [5]. Several factors can alter development of the circadian rhythm in preterm infants, including the separation from transplacental support, loss of maternal entrainment and markedly different environmental stimuli [6] sensory cues (light exposure), and care-providing events. Maternal breast milk contains factors that change diurnally and which help promote the entrainment of the circadian rhythm in the newborn; intake of stored breast milk that is asynchronous with the maternal diurnal rhythm may disrupt entrainment in the premature newborn [6, 7].

The central generator of circadian rhythm is located in the suprachiasmatic nucleus in the anterior hypothalamus,[8, 9] and generates diurnal oscillations at systems, cellular and molecular levels. The circadian rhythm drives phenotypic, physiologic and behavioral responses [10], including variations in body temperature [6], heart rate [8], endocrine secretions such as cortisol, melatonin [10], and even renal function [11]. Surrogate biomarkers for studying circadian rhythm include heart rate, body temperature, respiratory rates and amplitude-integrated EEG [5, 6, 8, 10]^,[12].

While a considerable body of data is available for circadian rhythm function in children and adults,[10, 13, 14] studies of circadian rhythm development during the extrauterine third trimester after premature birth have been few and findings inconsistent, at least in part due to methodological differences [6, 15, 16]. Earlier studies found that the circadian rhythm develops within 15–30 postnatal days for both term and preterm infants [17] [6, 18]. In contrast, other studies have described the effect of postconceptional age (PCA), with circadian amplitude (the maximum deviation of the rhythm from the mean value) at 35–37 weeks PCA being significantly higher than at 32 – 34 weeks PCA, suggesting an earlier emergence of circadian rhythm than previously thought [4, 6, 14, 19]. Another study using sleep/wake behavior measures showed that the duration of day/night cycle exposure (i.e., postnatal age) was a more important determinant of circadian rhythm development than postmenstrual age [7]. A study of continuous light exposure during neonatal intensive care unit (NICU) monitoring showed no consistent pattern of circadian rhythm development [20]. The conflicting results presented in these studies were based on small samples [14, 19], short duration of data collection (about two weeks) [14] or subjective behavioral observations [7].

To date there has been no longitudinal study of circadian rhythm development in preterm infants throughout their NICU stay with continuous monitoring of physiological signals. Our goal was to fill this void, by characterizing circadian rhythm development in preterm infants born between 23 and 36 weeks’ gestation, using continuous heart rate measurements. We sought to investigate the effect of postnatal and postmenstrual ages on circadian rhythm.

Materials and Methods

Infants recruited in this ongoing prospective observational study were admitted to our level IV NICU for prematurity between May 2020 and July 2023. The inclusion criteria included birth ≤ 36 weeks of gestation and < 14 days of life at admission. Exclusion criteria included significant parenchymal brain injury on routine head ultrasound in the first week of life (e.g., periventricular leukomalacia, intraventricular hemorrhage (IVH) grades III-IV, stroke). Infants with confirmed or suspected genetic conditions were also excluded. In our NICU, routine care includes covering of isolette for extremely preterm infants while infants who have graduated (typically 1600 – 1800 grams for infants with adequate growth, which usually coincides around 32 weeks) based on weight to open cribs are exposed to natural daylight and dimmed lights overnight. We abstracted relevant clinical variables from the medical records. The study was approved by the Children’s National Hospital Institutional Review Board and informed consent was obtained in all cases.

Physiological Variables

Physiological variables including heart rate, respiratory rate, indwelling arterial blood pressure, and oxygen saturation were retrieved continuously from the bedside monitor (Intellivue MP70, Philips Healthcare, Andover, MA, USA) in health level 7 (HL7) format. The sampling rate was 0.2 Hz. The data were retrieved from the time of admission until NICU discharge.

Brain Magnetic Resonance Image

Brain magnetic resonance images (MRI)s were taken on a 3T scanner (Discovery MR 750, GE Healthcare, Milwaukee, WI, USA) every four postnatal weeks as part of this study. The term equivalent MRIs were reviewed by a neuroradiologist (JM) and scored for brain injury using the Kidokoro score [21], which yields a score between 0 – 12+. Based on the Kidokoro Score, the infants were stratified into Normal 0–3, Mild 4–7, Moderate 8–11, and Severe 12+.

Circadian Rhythm Characterization

The circadian rhythm characterization followed a cosine fit to the heart rate as previously described [8, 22]. For every postnatal day, the heart rate was selected from 6 am to the following day 6 am. Spurious beats (heart rate<50 beats per minute (bpm) or heart rate >250 bpm) [23] were removed using an automated program. The heart rate was partitioned into 20-minute non-overlapping epochs, and the average was calculated for heart rate in each epoch, which yielded 72 samples for 24 hours of data. The average heart rate $H{R}_t$ was modeled using a cosine fit to identify the presence of a 24-hour rhythm as follows: $H{R}_t=A\mathrm{c}\mathrm{o}\mathrm{s}\left(\frac{2\pi }{24}t+\varphi \right)+M$, with $A$ being the amplitude of the rhythm, $M$ is the mean value of the heart rate and $\varphi$ being the phasor which measures the location of the amplitude. The goodness of fit was assessed by calculating Pearson’s correlation coefficient between the best-fit rhythm and $H{R}_t$. The probability $\left(P_P\right)$ that Pearson’s correlation coefficient is zero is also calculated. The fit is considered reliable if $P_P<0.05$; otherwise, the data from this day was disregarded. Figure 1a shows an example of a good cosine fit, and 1b is a poor fit. If more than two hours of physiological data were unavailable (e.g., due to the transport of an infant for a procedure), we discarded data from those days. We then determined the percentage of days during the study period when discernable circadian rhythms were present. Once a discernable circadian rhythm was confirmed, we calculated the circadian rhythm amplitude for each day.

Figure 1.

Circadian rhythm fit. a) A good cosine fit and b) a poor cosine fit.

We then averaged the daily circadian rhythm amplitudes to derive a weekly circadian rhythm amplitude. The relationships between daily amplitudes and postmenstrual/postnatal age were studied to assess circadian rhythm maturation over time. The circadian rhythm that reached statistical significance $\left(P_P<0.05\right)$ were used to construct the group average. We considered three PNA groups (≤30, 31 – 60, and >60 days) and three PMA (≤31, 32 – 35, and > 35 weeks) groups. The rationale for using these categories was to ensure a uniform distribution of days with significant circadian rhythm in each category. All analyses were performed offline in MATLAB 2023a (Mathworks Inc, Natick, MA, USA).

Statistical Analysis

If normally distributed, continuous data was represented as mean (standard deviation); otherwise, it was represented as median (minimum, maximum). The ordinal and nominal data were represented as numbers (percentages). The dependence of circadian amplitude on postmenstrual and postnatal ages was studied using a linear mixed effects model for repeated measures with random intercepts and random slopes. We controlled for covariates with potential chronotropic effects during each week, including medications/procedures, such as caffeine and/or pressors including dopamine, epinephrine, dobutamine, adenosine, as well as dexmedetomidine, ventilator support, and infection by including them as variant covariates. We also controlled for sex and brain injury quantitated using the Kidokoro Score by including them as invariant covariates. Model included gestational age at birth as the invariant covariate in the assessment of circadian amplitude and postnatal age. A P<0.05 was considered statistically significant. The statistical analyses were performed in MATLAB 2023a using the statistical toolbox.

The grand averages of circadian rhythm in the different age categories were represented as mean + standard error.

Results

A total of 67 premature infants were enrolled in this ongoing prospective study. One subject without continuous heart rate data was removed from further analysis. Clinical data are shown in Table 1. The median (minimum, maximum) postnatal age at admission was 0 (0,7) days. The mean (standard deviation) gestational age at birth was 29.87 (3.65) weeks. Thirty-four subjects (51%) were females. Sixteen (24%) were on conventional ventilators and/or high-flow oscillator ventilator support during the study period, 19 (28%) developed infection and were treated with antibiotics, 44 (66%) received caffeine treatment, and 18 (27%) required pressor support. The median length of NICU stay was 61.94 (2.82, 245.94) days. Ultrasound assessment at admission showed unilateral grade I IVH in 3 (5.36%) infants, unilateral grade II IVH in 5 (8.93%), bilateral grade I IVH in 1 (1.79%) infant, and bilateral grade II IVH in 6 (10.71%) infants.

Table 1.

Clinical data (n = 66).

Clinical Variable	n(%) or mean(sth) or median (min, max)
GA at birth mean(std) weeks	29.84 (3.65)
Postnatal age at admit median(min,max) days	0 (0,7)
Ventilation support n(%)	16 (24)
Caffeine treatment n(%)	44 (66)
Pressor support n(%)	18 (27)
Female n(%)	34 (51)
Infection treatment n(%)	19 (28)
Apgar 1 minute mean(std)*	5.39 (2.44)
Apgar 5 minutes mean(std)**	7.31 (1.78)
Length of stay median(min, max) days	61.94 (2.82, 245.94)
Clinically significant PDA n(%)	18 (27.27)
Brain Injury (Based on Kidokoro Score)*** n(%)
Normal	34 (62)
Mild	19 (35)
Moderate	2 (4)
Severe	0 (0)
IVH**** n(%)
Unilateral Grade I	3 (5.36)
Unilateral Grade II	5 (8.93)
Bilateral Grade I	1 (1.79)
Bilateral Grade II	6 (10.71)
No IVH	41 (73.21)

^*Calculated from 62 infants

^**Calculated from 61 infants

^***Calculated from 55 infants

^****Calculated from 56 infants

GA: Gestation age; PDA: Patent Ductus Arteriosus; IVH: Intraventricular Hemorrhage.

In total there were 3189 days of data from all infants. Of which, 2191 days (68.7%) had significant circadian rhythms. The deviations of heart rate from the mean value and the cosine fit for an infant born at 27 weeks’ gestation for the postnatal days of 10 and 118 are shown in Figures 2a and b, respectively. There is an increase in the circadian amplitude from day 3 to day 117. The peak represents the active state and the trough represents the quiet (state) period. In the plots, the peaks occur during the day and troughs during the night.

Figure 2.

Circadian rhythm characterization using cosinor fit for a preterm infant born at 27 weeks gestation age on a) 10 days and b) 117 days.

The variation of circadian amplitude (quantified using heart rate) is shown in Figure 3a for postmenstrual age and Figure 3b for postnatal age. The unadjusted PMA increase (estimate: 1.23 × 10⁻² beats per minute (bpm)/week, confidence interval: 1.01 × 10⁻² – 1.46 × 10⁻² bpm/week, P = 1.01 × 10⁻²⁴) and postnatal increase (estimate: 0.02 bpm/day, confidence interval: 0.01 – 0.02 bpm/day; P = 1.05 × 10⁻¹³) are statistically significant and remained significant after adjusting for clinical variables defined above (adjusted estimate for PMA: 0.15 bpm/week; confidence interval: 9.7 × 10⁻² – 1.97 × 10⁻² bpm/week; P = 2.07 × 10⁻⁸ and for postnatal age: 0.02 bpm/day; confidence interval: 0.01 – 0.27 bpm/day; P = 1.87 × 10⁻⁶).

Figure 3.

The longitudinal evolutions of circadian amplitude calculated using the heart rate are shown in a) postmenstrual and b) postnatal ages. The insets show the beta estimates and P-values calculated using the un-adjusted linear mixed effects models.

Table 2 summarizes the number of days with significant circadian rhythm in the three PMA and PNA categories. Circadian rhythm at distinct gestational age windows was highest for the group in >35 weeks (890 out of 1387) postmenstrual age compared to the groups in 32–35 weeks (669 out of 970) and 23 – 31 weeks (632 out of 832) of postmenstrual age (see Figure 4a). The rhythm for the 32–35 weeks group had the lowest amplitude of the other two groups. In the postnatal analysis, circadian rhythm amplitude was highest for the group in >60 days of life (619 out of 920) compared to the groups in 30 – 60 days of life (639 out of 947) and ≤ 30 days of life (935 out of 1322) (see Figure 4b). The rhythm amplitude increased progressively with postnatal age.

Table 2.

Percentages of days with significant circadian rhythm (P_P < 0.05).

PMA weeks	% (number of days/total days)	PNA days	% (number of days/total days)
≤ 31	76 (632/832)	≤ 30	70.7 (935/1322)
32 – 35	68.9 (669/970)	31 – 60	67.5 (639/947)
>35	64.3 (892/1387)	>60	67.1 (619/920)

Figure 4.

Grand averages were calculated for a) three different postmenstrual age groups and b) three postnatal age groups. The vertical bars present mean + one standard error.

In postmenstrual and postnatal age analyses, the rhythm amplitude for the youngest groups (PMA≤ 31 weeks and PNA≤ 30 days) peaked during the night with the nadir occurring during the day. Confining the postmenstrual age analysis to infants born ≤ 28 weeks of gestation age showed similar results to the ones obtained for the postnatal age analysis (Figure 4b), suggesting that postnatal age maturation is more important than the postmenstrual age maturation.

Discussion

In this study, we describe the circadian rhythm maturation in preterm infants born between 23- and 36-weeks’ gestational age. We show a significant increase in circadian amplitude as a function of both postnatal and postmenstrual ages. These results are independent of potential confounders, such as infection status, ventilator support, brain injury, and use of medications that alter heart rate. Furthermore, we also show that the circadian rhythm is present during the first four weeks of life but with reversed phase, peak amplitude at night and lowest amplitude during the day, which is independent of PMA and PNA. This longitudinal study further supports the use of heart rate as a surrogate biomarker for the study of circadian maturation.

Although several studies have shown a lack of circadian rhythm at birth, our results show the presence of this rhythm during the first postnatal week. Reasons for these discrepancies may be several, including a relatively small variation in the circadian rhythm surrogate used (e.g., body temperature) [4, 6, 14, 24], or loss of circadian rhythm amplitude when circadian rhythm phase changes are not considered during the averaging of measurements. Our results support the findings of previous studies demonstrating that postnatal age played a more prominent role in circadian rhythm maturation than postmenstrual age [4, 7]. In assessing the role of postnatal age, we also show that the circadian rhythm amplitude is lower during first 30 postnatal days and that circadian rhythm phase during the 24-hour clock is reversed during this period. After 30 postnatal days the circadian rhythm amplitude is significantly higher and the circadian rhythm phase is the same as in more mature subjects [4, 8, 25]. The reversed circadian rhythm phase early in the postnatal period of these prematurely born infants is similar to the changes observed in cortisol concentrations in preterm infants, which have been argued to be indicative of the disruptive nature of the NICU environment [26, 27]. Other possibilities include the abrupt cessation of circadian rhythm entrainment by maternal/placental hormones, and developmentally inappropriate extrauterine environmental exposures, such as increased caregiving events for extremely and very preterm infants, increased sensory stimulation, among others and use of medications that could impact the sleep/wake cycles [28, 29]. Further studies are needed to examine the relationship between exposures in the NICU environment and circadian rhythm development in premature infants during the early postnatal period.

Circadian rhythm is considered to include a series of six to eight ultradian (sleep/wake) cycles [30, 31]. Several studies in preterm infants have shown increased sleep/wake cycling with maturation [32–35]. However, there may be numerous disruptions to the natural sleep/wake cycles in the NICU environment, including day/light exposures, monitor alarms, routine care sessions, and painful procedures, which may, in turn, negatively impact circadian rhythm development [4, 36]. Such caregiving events in the NICU have been shown to alter the maturation of autonomic physiology, especially of sympathetic tone [37–40]. In a chronically implanted ovine fetal model after sympathectomy and vagotomy there was a reduction in the heart rate and a shift in the circadian amplitude in sympathectomy group compared to controls [22]. Together these data, including the low amplitude circadian rhythm in our group-level analysis, suggest that impaired maturation of circadian rhythm and the autonomic nervous system may have shared pathways.

Apart from indicating the maturational aspect of the circadian system, our findings may have several other maturation-related implications. The ability to measure peak and trough circadian rhythm will help guide design future studies to understand the neurodevelopment and autonomic tone development around those periods [41–43]. Furthermore, with a slight modification to the approach, identification of ultradian rhythms (sleep/wake states) is possible, which could be used to optimize NICU caregiving schedules according to infants’ state [32]. Finally, the relationship between autonomic and circadian rhythm development is incompletely understood and is an important area of future research [44].

Our work has several strengths, including prolonged (up to 245 days) continuous heart rate recordings and an objective approach to characterize circadian rhythm. Some of the limitations of our work deserve mentioning. Non-exposure to light during the isolette monitoring might have impacted the circadian rhythm and subsequent maturation after transitioning into the open crib. We did not control for the duration of the isolette monitoring, which we will pursue in future studies. Previous studies have shown the importance of infant exposure to day/night in generating circadian rhythm in premature infants [7]. We did not measure the intensity and duration of light exposure experienced by our subjects, but plan to in future prospective studies. Feeding mother’s milk versus formula milk has been shown to affect circadian rhythm development [4, 36]. Furthermore, diurnal fluctuations in human milk melatonin levels have been implicated in the entrainment of diurnal rhythm in the newborn [45]. Thus, for example, breastmilk pumped in the morning but fed at night might disrupt the entrainment process [4]. We aim to study the timing of breastmilk pumping and feeding in future studies. Other common events in premature infants, such as bradycardia and oxygen desaturation spells, are thought to affect brain development [46] we plan to study the effects of such events on circadian rhythm development in future prospective studies. With maturation, ultradian rhythms become stronger and might interfere with the reliable estimation of the circadian cycles, and this may explain the smallest number of significant circadian rhythm observed in the mature category (PMA>35 weeks and PNA> 60 days). Future work will explore using a different window length to average the heart rate for the later age category.

Conclusions

In summary, we have demonstrated that in premature infants, the circadian rhythm is present soon after birth, albeit with low amplitude and reversed phase; after the first month, amplitude increases significantly and circadian rhythm phase is similar to that in more mature subjects. Postnatal age appears to play a more important maturational role than postmenstrual age. The reason for the early circadian rhythm phase reversal is unclear and warrants further study. Development of the circadian rhythm within the context of overall brain development in premature infants is another as yet unexplored question awaiting future study.

Highlights.

Circadian rhythm evolves longitudinally with postmenstrual and postnatal ages Below 30 postnatal days, the phase of the rhythm is reversed, normalizes thereafter Nocturnal caregiving events could impact the rhythm development below 30 days

Data Availability

The data underlying this article cannot be shared publicly due to the IRB restriction. The data will be shared on reasonable request to the corresponding author.