Effects of caffeine therapy for apnea of prematurity on sleep and neurodevelopment of preterm infants at 6 months of corrected age

Abstract

Study Objectives:

To determine the differences in sleep patterns between preterm infants who received caffeine and those who did not and to evaluate the effects of caffeine therapy on early neurodevelopment. Secondarily, actigraphy and polysomnography were compared to evaluate the sleep of preterm infants.

Methods:

Twenty-eight preterm infants ages 28–34 weeks admitted to a single-center Level III neonatal intensive care unit between May 2020 and May 2021 were included. Sleep was assessed by actigraphy for 72 hours with Respironics Mini-Mitter® Actiwatch-2 and Brief Infant Sleep Questionnaire at 6 months corrected age. On the first day of actigraphy, infants underwent polysomnography between 10:00 am and 3:00 pm. Neurodevelopment was evaluated by the Bayley Scales of Infant and Toddler Development-III, the Ages & Stages Questionnaire, and the Hammersmith Infant Neurological Examination.

Results:

There were no significant differences in sleep parameters measured by actigraphy, the Brief Infant Sleep Questionnaire, and polysomnography between infants in the caffeine group (n = 12) and no-caffeine group (n = 16). Sensitivity (91.07%) and agreement rate (77.21%) for the actigraphy against polysomnography were highest at the automatic threshold. No significant differences were observed in the neurodevelopment of infants in the caffeine group compared to the no-caffeine group.

Conclusions:

Sleep parameters and neurodevelopmental outcomes were not different in infants at 6 months of corrected age with regard to caffeine therapy. Actigraphy at the automatic threshold can be used in infants for sleep pattern assessment.

Clinical Trial Registration: Registry: ClinicalTrials.gov; Name: Influence of Caffeine Therapy in Preterm Infants; URL: https://www.clinicaltrials.gov/ct2/show/NCT04376749; Identifier: NCT04376749.

Citation:

Atalah YEY, Barış HE, Akdere SK, et al. Effects of caffeine therapy for apnea of prematurity on sleep and neurodevelopment of preterm infants at 6 months of corrected age. J Clin Sleep Med. 2023;19(12):2075–2085.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Neonatal caffeine therapy is widely used in neonatal care units for prophylaxis and treatment of apnea of prematurity. The effects of neonatal caffeine therapy on the sleep of preterm infants have to be elucidated.

Study Impact: Sleep parameters derived from the Brief Infant Sleep Questionnaire, actigraphy, and polysomnography at 6 months of corrected age were similar in caffeine-treated and noncaffeine-treated preterm infants. Caffeine used in preterm infants does not appear to negatively impact the neurodevelopment of infants assessed at 6 months of corrected age.

INTRODUCTION

The preterm birth rate in Turkey is 11%, which is similar to the global preterm birth rate of 10.6% as reported in 2014.^1,2 Preterm birth and related complications are the leading cause of under-five child mortality.³ These complications include pulmonary conditions such as bronchopulmonary dysplasia and prolonged duration of invasive ventilatory support. Noninvasive ventilation is the preferred method for ventilatory support in most preterm infants, but apnea of prematurity (AOP) may challenge the success of noninvasive ventilation.

The development of sleep-wake cycles is an important process of the maturing human brain during infancy, which is affected by internal physiological factors and external socioenvironmental changes.⁴ Studies revealed differences in sleep outcomes of preterm infants compared to term infants such as reduced sleep duration, increased motor activity during sleep, and lower arousal threshold, some of which extend beyond infancy at older ages.⁵ Premature infants’ brains are particularly vulnerable to environmental insults that may disrupt sleep due to ongoing maturational process outside the protective intrauterine environment.⁴ These insults may include light and noise but also the use of methylxanthines.

Methylxanthines are one of the effective treatments commonly used in the neonatal intensive care unit (NICU) to prevent and treat AOP, thus reducing the need for invasive ventilation. Caffeine is the drug of choice among methylxanthines. In addition to facilitating extubation in the NICU, caffeine reduces the incidence of bronchopulmonary dysplasia and provides improved survival with better neurodevelopmental outcomes.^6–8

Studies about caffeine mainly focus on its respiratory and neurodevelopmental benefits on preterm infants, while effects on sleep changes in preterm infants were rarely investigated. Caffeine acts on adenosinergic receptors that are located in the brain, lungs, and cardiovascular system due to its adenosine-like structure.⁸ Caffeine’s ability to enhance alertness and wakefulness through its adenosine receptor blockade of inhibitory neurotransmitters is already well known, but its effects on the sleep of preterm infants are not well established. Early effects of caffeine on the sleep of preterm infants may be masked by irregular circadian rhythms related to neurodevelopmental immaturity. As such, sleep efficiency and total sleep time in rapid eye movement and non-rapid eye movement sleep of caffeine-treated late preterm infants were not different at the second day of treatment compared to their baseline measurement.⁹ Sleep organization of premature infants born at 33–34 weeks of gestational age (GA) was not affected by maintenance-dose caffeine treatment.¹⁰ Similarly, caffeine had no effects on sleep of preterm infants born at < 28 weeks of GA.¹¹ In contrast, in the group of infants born ≥ 28 weeks, active sleep decreased and wakefulness increased, as caffeine concentrations increased from the first to the fifth day of postnatal age.¹¹ In the long term, neonatal caffeine therapy had no effect on total sleep time of ex-preterm infants at 5–12 years of age.¹² Existing studies differ in study design and participant characteristics, such as GA, techniques used in sleep assessments, caffeine dosage, or timing of sleep assessment in relation to the caffeine administration, resulting in inconsistent results. Besides, effects were measured only after short-term caffeine treatment during short observation periods and mostly during NICU stays.^{9–11,13–16} Except for one study, long-term possible effects were not investigated.¹²

It is assumed that potential alterations in sleep structure related to prematurity may contribute to neurodevelopmental problems in preterm infants. Disturbed sleep may affect the neurodevelopment of preterm infants during this vulnerable period of rapid growth and development.⁵ Caffeine is known as one of the few neuroprotective strategies for preterm infants in the long term. However, studies investigating the effects of caffeine therapy on the sleep of preterm infants did not report the neurodevelopmental differences that may result from potential sleep alterations due to caffeine administration.

This study aimed to investigate the effects of caffeine therapy on sleep-wake periods and neurodevelopment of preterm infants at 6 months corrected age (CA). However, sleep assessment in preterm infants is challenging. Available methods to study sleep, such as polysomnography (PSG) and direct behavioral observation, require expertise or technical equipment.^17–20 Recently, actigraphy has been proposed as a useful tool for objective sleep assessment in preterm infants with less effort, allowing prolonged periods of measurement in the home environment.^19,21 As a secondary aim, two different methods for sleep assessments (actigraphy and PSG) were compared to evaluate the sleep of preterm infants.

METHODS

Participants and settings

This observational comparative study was conducted in a single-center Level III NICU at the Marmara University Hospital located in Istanbul, Turkey, between May 2020 and May 2021. The study was approved by the Marmara University Ethics Committee (09.2019.950) and performed in line with the principles of the Declaration of Helsinki. Written informed consent was obtained from the parents of each infant before recruitment. The trial was registered with ClinicalTrials.gov under the number NCT04376749.

During the study period, 2,371 babies were born, and the annual preterm birth rate was 9.8% (n = 232) at the University Hospital. Eligibility criteria included preterm infants between 28 and 34 weeks of gestation admitted to the University Hospital’s NICU between May 2020 and May 2021. After discharge, parents of all consecutively admitted eligible preterm infants were invited to participate in the study for evaluation of their infants at 6 months CA (Figure 1). Written informed consent was obtained from the parents of each infant before recruitment.

Figure 1. Flow diagram of the study.

Extremely preterm infants, those born before 28 weeks GA, have an increased risk for neonatal morbidity.²² Differentiation of active sleep and quiet sleep states has been demonstrated at 28 weeks of GA, and sleep organization shows no changes beyond 28 weeks of GA. Given the association between GA and sleep organization and the decreased risk for poor neurodevelopmental outcomes, the eligible infants were limited to infants born above 28 weeks.²³ The study population was divided into two groups: infants who received caffeine (caffeine group [CG]) and infants who did not receive caffeine (no-caffeine group [NCG]). The usual protocol of the NICU is to administer caffeine prophylaxis routinely to all infants born at less than 30 weeks of GA. A 20 mg/kg loading dose is followed by a 5–10 mg/kg/d maintenance dose.^6,24 Infants born at > 30 weeks of GA were treated with caffeine only if they showed signs of AOP. Caffeine is rarely administered in infants born at ≥ 35 GA.^6,24 Therefore, preterm infants born between 30 to 34 weeks of GA, who had no AOP and who did not receive caffeine, were selected as the control group (NCG), which was the most similar group in terms of GA to the CG. The caffeine cumulative dose (mg/kg) was calculated for each infant as the sum value of all daily doses the infant received calculated by dividing the daily caffeine dose by the infant’s weight on that day.

Procedures and measurements

Demographic and clinical data were noted from the patient’s medical records. GA was determined by the last menstrual period and/or by prenatal ultrasonography and was confirmed by Ballard Scoring System.²⁵ Chronological age was defined as the time elapsed since birth. CA was calculated by subtracting the number of weeks born before 40 weeks of gestation from the chronological age.²⁶

Infants’ illness severity was measured by the Neonatal Therapeutic Intervention Scoring System, which is a validated measure of resource utilization associated with mortality risk estimates and scored for each infant during his or her NICU stay.^27,28

Anthropometric measures were represented as z-scores weight, length, and head circumference both at birth and at 6 months CA.²⁹

Assessment of sleep

After discharge, at 6 months of CA, sleep was assessed by the expanded version of the Brief Infant Sleep Questionnaire, actigraphy, and PSG. On the first day of actigraphy, all infants underwent PSG at the Marmara University Sleep Unit between 10:00 am and 3:00 pm during their daytime naps. Infants wore actigraphy for 72 hours at home. Parents filled out a sleep diary concurrently.

Brief Infant Sleep Questionnaire

The Brief Infant Sleep Questionnaire is widely used for the sleep assessment of children between 0 and 36 months, and the Turkish version was found reliable.^30,31 It includes items about the nighttime and daytime sleep duration, night awakening, wake after sleep onset duration, bedtime, sleep onset latency, bedtime routine, bedtime difficulty, methods of falling asleep, and sleep location.

PSG

PSG was performed at the Marmara University Sleep Laboratory following American Academy of Sleep Medicine AASM guidelines. All infants underwent PSG (Embla N7000 PSG and Embla RemLogic) during their daytime naps between 10:00 am and 3:00 pm. For PSG scoring, the recommendations of The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications Scoring Manual: Rules, Terminology and Technical Specifications, version 2.6³² were followed according to the pediatric criteria, and analysis of PSG records was carried out by an experienced sleep technician blinded for the actigraphy data.

Actigraphy

Philips Respironics Mini-Mitter Actiwatch-2 was placed on the infant’s right ankle. The Actiware software (version 6.0.9) was used to analyze sleep-wake data in 30-second epochs. In accordance with previous studies performed in infants, data were analyzed at low, medium, and high activity thresholds (activity threshold value 20, 40, and 80) in addition to automatic activity threshold (mean activity counts *0.888/epoch length) settings.^17,20 Recorded data was scored as sleep when total activity counts were equal to or less than the activity threshold settings. Daily sleep data derived from sleep diaries completed by caregivers at home were used to verify the actigraphic data.

Measures derived from actigraphy included total sleep time, nocturnal sleep time, nap time, sleep efficiency, sleep onset latency, wake after sleep onset, awakenings, bedtime, and get-up time. Sleep efficiency was defined as total sleep time/total sleep time + total wake time × 100. Wake after sleep onset was defined as the number of waking minutes between sleep start and sleep end time.^33,34

Actiwatch devices used in this study were sponsored in a previous project by Marmara University Scientific Research Commission, BAPKO (SAG-C-TUP-131217-0647).

Actigraphy results in this study were validated against PSG as the gold-standard measure.

Neurodevelopmental assessments

The neurodevelopment was assessed at 6 months of CA by the Ages & Stages Questionnaire-2, the Hammersmith Infant Neurological Examination (HINE), and the Bayley Scales of Infant Development, third edition (BSID-III).

Ages & Stages Questionnaire-2

The Ages & Stages Questionnaire is a simple general developmental screening instrument including questions in five domains: personal-social, gross motor, fine motor, problem-solving, and communication. It helps to evaluate developmental progress in children between the ages of 1 month and 66 months.³⁵ Each domain consists of six questions which are answered in three simple responses: yes, sometimes, or not yet (yes: 10 points, sometimes: 5 points, not yet: 0 points).³⁵ As suggested, the age of the infants was adjusted for prematurity.³⁶ Infants who are close to the cut-off are considered in the low-risk zone. Infants who are 2 standard deviations below the mean (below the cut-off) in any domain are considered in the high-risk zone.³⁷

HINE

The HINE is a simple, easily performed, and scorable neurological examination method for 3- to 24-month-old infants.³⁸ It is a reliable test to detect preterm and term children at risk for cerebral palsy. HINE also helps to identify children who may not reach developmental milestones on time and need rehabilitation.³⁸ It includes 26 items that assess cranial nerve functions (maximum 15 points), posture (maximum 18 points), movements (maximum 6 points), tone (maximum 15 points), and reflexes (maximum 15 points). Each item is scored from 0 to 3 and is added to achieve a global score. The global score can range from 0 to a maximum of 78.³⁸ The cut-off points are as follows: ≥ 70 = normal, 60–70 = low risk, 40–60 = moderate risk, ≤ 40 = high-risk.^39,40

BSID-III

The BSID-III is a standardized 30- to 90-minute tool for children between 0 and 42 months to assess neurodevelopment. Clinical indications for BSID-III include high-risk groups such as preterm infants.⁴¹ It consists of five subtests: cognitive, receptive communication, expressive communication, fine motor, and gross motor.⁴¹ A certified developmental specialist performed the test and calculated composite scores for all subgroups. Composite scores were called to a metric with a mean of 100 and a standard deviation of 15. The cut-off points are as follows: ≥ 70 and < 85: mild delay, ≥ 55 and < 70: moderate delay, < 55 severe delay.^42,43

Statistical analysis

Statistical analyses were performed using SPSS Version 28.0 (IBM Inc., Armonk, NY, USA). Bland–Altman plots were performed using MedCalc Version 20.011. Descriptive data were presented as mean ± standard deviation or median (interquartile range) with significance at the P < .05 level.

Continuous variables were compared by Mann–Whitney U test. Spearman’s correlation test was used to determine the relationship between caffeine cumulative dose and sleep parameters or neurodevelopmental scores. Correspondence between PSG and actigraphy was tested with Cohen’s kappa.⁴⁴ Since sleep and wake were not equally distributed, prevalence-adjusted and bias-adjusted kappa (PABAK) was calculated to provide equal weights to sleep and wake. According to Cohen’s kappa and PABAK, 0.21 ≤ κ ≤ 0.4 was defined as a fair agreement and 0.41 ≤ κ ≤ 0.6 as a moderate agreement.⁴⁴

Actigraphy findings at each threshold setting were compared with PSG at four different threshold settings using the Bland–Altman method.⁴⁵ Scatter plots were drawn for sleep and wakefulness data. The relationship between means and differences was determined according to the distribution between ± 1.96 standard deviation agreement limits.⁴⁵ PSG data were accepted as the gold standard.

Data were analyzed with cross-tabulation and used to calculate five metrics for each threshold setting for validation of actigraphy: agreement rate, sensitivity, specificity, predictive value of sleep, and predictive value of wake (PVW).^17,18

The agreement rate was defined as the overall accuracy of actigraph sleep/wake classification. Sensitivity was the ability of the Actiwatch to predict true sleep. Specificity was the ability of the Actiwatch to predict true wake. Predictive value of sleep was defined as the probability that the Actiwatch data were correct by PSG criteria for sleep and PVW as the probability that the Actiwatch data were correct by PSG criteria for wake.^17,18,20

RESULTS

Characteristics of the infants

Twenty-eight preterm infants were enrolled in the study between May 2020 and May 2021. Characteristics of the study participants are presented in Table 1. Infants in the CG significantly differed from the infants in the NCG in terms of GA at birth since routine caffeine prophylaxis was given to infants born at GA < 30 weeks. As such, infants in the CG had significantly higher Neonatal Therapeutic Intervention Scoring System scores indicating more resource utilization as expected due to their younger GA. Infants in both groups showed similar anthropometric measures with no difference in z-scores.

Table 1.

Demographic and clinical characteristics of preterm infants.

	No Caffeine (n = 16)	Caffeine (n = 12)	P
Gestational age at birth (weeks)	33.3 [31.3 to 34.1]	29.5 [28 to 30.8]	<.001
Corrected age at the time of study (weeks)	27 [26 to 30]	26.5 [26 to 29]	.81
Female sex, n (%)	5 (50)	5 (41.7)	.66
Cesarean section, n (%)	12 (75)	9 (75)	1.0
5-min APGAR score	8.5 [7.5 to 9.5]	7 [6 to 7]	.005
NTISS score during hospitalization	4.9 [4.5 to 5.9]	6.3 [6.1 to 6.7]	<.001
z-scores of the infants’ physical measurement at birth
Weight	0.5 [0.3 to 0.9]	0.2 [−0.3 to 0.9]	.44
Height	0.6 [0.3 to 1.4]	−0.04 [−0.7 to 0.4]	.04
Head circumference	0.9 [0.3 to 1.1]	0.3 [−0.9 to 0.9]	.20
z-scores of the infants’ physical measurement at 6 months of corrected age
Weight	−0.03 [−1.2 to 1.3]	−0.3 [−0.9 to −0.02]	.31
Height	0.3 [−0.7 to 1.7]	−0.1 [−0.4 to 0.9]	.52
Head circumference	0.1 [−0.5 to 0.5]	−1 [−2.1 to 0.3]	.10

Continuous variables were represented as median [interquartile range]. NTISS = Neonatal Therapeutic Intervention Scoring System.

Sleep analysis

Caregivers completed Brief Infant Sleep Questionnaire at 6 months of CA. Accordingly, sleep parameters were not different between the CG and NCG (Table 2). The total sleep duration of preterm infants was approximately 12.5–13.5 hours at 6 months of CA.

Table 2.

Sleep variables derived from BISQ in the caffeine group and no-caffeine group.

	No Caffeine (n = 16)	Caffeine (n = 12)	P
Nighttime sleep, min	540 [540–630]	600 [547.5–660]	.45
Daytime sleep, min	165 [90–225]	210 [157.5–292.5]	.15
Total sleep, min	735 [667.5–832.5]	810 [750–892.5]	.13
Night wakening, number	2 [2–3.75]	1 [1–3]	.20
WASO, min	35 [18.75– 45]	20 [10–30]	.20
Longest stretch of time the child is asleep during night, min	240 [180–345]	360 [255–465]	.74
Daytime naps, number	3 [2.25–3]	3 [2–3]	.77

Data were represented as median [interquartile range]. BISQ = Brief Infant Sleep Questionnaire, IQR = interquartile range, WASO = wake after sleep onset.

Three-day actigraphy recording was performed in 15 infants in the NCG and in 11 infants in the CG. Recordings from two infants were compromised due to technical failure and were not included in the analyses. Actigraphy recordings showed that preterm infants evaluated at 6 months of CA were sleeping between 10.6 and 12.7 hours daily. No significant differences were detected between the CG and NCG in the sleep parameters derived from actigraphy (Table 3).

Table 3.

Sleep parameters derived from actigraphy at different activity thresholds.

Sleep Variables	Low Threshold		Medium Threshold		High Threshold		Automatic Threshold
	Caffeine (−)	Caffeine (+)	Caffeine (−)	Caffeine (+)	Caffeine (−)	Caffeine (+)	Caffeine (−)	Caffeine (+)
Total sleep, min	639 [591–671]	651 [601–705]	675 [631–716]	689 [625–739]	703 [667–740]	709 [649–765]	745 [723–778]	763 [705–780]
Nighttime sleep, min	460 [407–502]	419 [402–491]	512 [435–528]	443 [421–525]	536 [453–549]	466 [436–548]	569 [482–598]	496 [477–570]
Daytime sleep, min	169 [143–232]	190 [175–232]	176 [147–241]	200 [186–254]	183 [150–248]	216 [194–273]	195 [154–257]	241 [199–296]
Sleep onset latency, min	7 [3–10]	6 [2–8]	7 [3–10]	6 [2–8]	7 [3–10]	6 [2–8]	7 [3–10]	6 [2–8]
Sleep efficiency, %	77 [76 –82 ]	81 [72–84]	83 [81–87]	85 [79–88]	87 [85–90]	89 [86–92]	95 [93–96]	95 [94–96]
WASO, min	107 [89–128]	85 [71–128]	78 [63–89]	55 [49–94]	54 [42–63]	38 [28–64]	11 [7–20]	8 [6–14]
Awakenings, number	86 [67–99]	66 [63–88]	75 [69–89]	71 [57–95]	68 [56–74]	66 [44–75]	21 [17–29]	16 [14–25]

No significant differences were detected between the CG and NCG in the sleep parameters with Mann–Whitney U test. Continuous data are expressed as median [interquartile range]. WASO = wake after sleep onset.

There was also no significant difference in sleep parameters according to the cumulative caffeine dose at different activity threshold settings.

Sleep variables obtained during daytime naps derived from PSG did not differ between the CG and NCG (Table 4).

Table 4.

Comparison of the polysomnography findings according to the caffeine therapy.

Sleep Variables	Caffeine (−)	Caffeine (+)	P a
	(n = 16)	(n = 12)
Total recording time, min	108.8 [69.7–157.8]	89.1 [74–148.5]	.85
WASO, min	17.8 [7.2–34.4]	16.9 [11.1–37]	.85
Total sleep time, min	81.5 [51–116]	68.3 [49.8–95.8]	.50
Sleep latency, min	4.5 [2–8.2]	7.5 [2.8–11.5]	.35
Sleep efficiency, %	78.7 [67.4–84.3]	68.2 [57.8–80.4]	.21

Data are represented as median [interquartile range]. ^aMann–Whitney U test. WASO = wake after sleep onset.

Epoch-by-epoch analysis

A total of n = 6,098, 30-second epochs were studied to calculate agreement of actigraphy against PSG. The PSG data of 2,206 minutes (72%) sleep and 843 minutes (28%) wakefulness were matched with the actigraphy data in binary form (sleep: 1, wakefulness: 0) at four different thresholds. Sensitivity (91.07%), PVW (63.65%), and agreement rate (77.21%) for the actigraphy against PSG were highest at the automatic threshold setting (Table 5). The highest specificity (80.37%) and predictive value of sleep (90.52%) were observed at the low threshold setting.

Table 5.

Sensitivity and specificity analysis of actigraphy against PSG and epoch by epoch agreement for actigraphy with PSG.

	Actigraphy Activity Threshold
	Low	Medium	High	Automatic
Sensitivity, %	71.65	76.34	79.96	91.07
Specificity, %	80.37	74.50	69.40	40.93
PVS, %	90.52	88.68	87.24	80.14
PVW, %	52.00	54.61	56.96	63.65
Agreement, %	74.06	75.83	77.04	77.21
Kappaa	0.445	0.457	0.462	0.360
PABAKb	0.481	0.517	0.540	0.544

^aCohen’s kappa. ^bPrevalance and bias adjusted kappa. PSG = polysomnography, PVS = predictive value for sleep, PVW = predictive value for wakefulness.

At all thresholds, the agreement was largely equivalent with moderate kappas (0.44–0.46) except for the automatic threshold with a low kappa coefficient (Table 5). Moderate PABAK coefficients (0.48–0.54) were calculated at all thresholds for actigraphy and PSG.

Bland–Altman plots for concordance between PSG and actigraphy of sleep and wake duration at all activity threshold settings are shown in Figure 2. At the automatic and high threshold setting, 95% of the observations were within the limits of agreement, indicating that mean sleep and wake data of PSG and actigraphy were consistent.

Figure 2. Plots of between methods (actigraphy against polysomnography) difference against means for sleep and wake duration at all low, medium, high, and automatic activity threshold levels.

Neurologic assessments

Infant neurodevelopmental assessments were performed by the Ages & Stages Questionnaire, BSID-III, and HINE at 6 months of CA, and results were shown in Table 6. There were no significant differences in the neurodevelopment of CG and NCG according to scores from BSID-III and the Ages & Stages Questionnaire. The HINE scores of infants in the CG were lower compared to the infants in the NCG; however, when corrected for GA, significance disappeared, since infants in the CG had lower median GA and were expected to have lower HINE scores. According to HINE scores, none of the infants were at high risk for developmental delay.

Table 6.

Comparison of neurodevelopment of preterm infants with or without caffeine therapy.

	No Caffeine	Caffeine	P
ASQ total score	270 [260–290]	278 [253–295]	.87a
ASQ groups			.97b
Normal	10 (62.5%)	7 (58.3%)
Low risk	5 (31.3%)	4 (33.3%)
High risk	1 (6.3%)	1 (8.3%)
ASQ communication	60 [53–60]	60 [58–60]	.35a
ASQ gross motor	58 [43–60]	53 [43–60]	.59a
ASQ fine motor	60 [60]	60 [53–60]	.22a
ASQ problem-solving	60 [50–60]	60 [48–60]	.84a
ASQ personal social	50 [50–60]	58 [45–60]	.66a
BSID-III cognitive scale score	100 [95–100]	100 [100–105]	.29a
BSID-III language scale score	103 [103–105]	105 [100–106]	.77a
BSID-III motor scale score	100 [97–197]	100 [99–112]	.57a
BSID-III total score
Normal	16 (100%)	12 (100%)
HINE score	75.5 [70–76]	70.5 [65–72]	.01a
HINE groups			.18b
Normal	13 (81.3%)	7 (58.3%)
Low risk	3 (18.8%)	5 (41.7%)

Continuous data are expressed as median [interquartile range]. ^aMann–Whitney U test. ^bChi-square test. ASQ = Ages & Stages Questionnaire, BSID-III = Hammersmith Infant Neurological Examination, third edition, HINE = Hammersmith Infant Neurological Examination.

DISCUSSION

Our study compared the sleep patterns of the caffeine-treated and noncaffeine-treated premature infants and found no significant differences in sleep parameters derived from the Brief Infant Sleep Questionnaire, actigraphy, and PSG at 6 months of CA. Neurodevelopmental outcomes of infants in the CG and NCG were also similar.

Despite its widespread use in preterm infants, only a few studies have investigated the sleep impacts of caffeine or other methylxanthines in preterm infants. Most recently, in the PSG-based study of Seppä-Moilanen et al, there was no effect of caffeine on the sleep of 21 preterm infants born at 28–33 weeks of GA on the second day of treatment.⁹ Similarly, Curzi-Dascalova et al performed PSG in 10 preterm infants at 33–34 postmenstrual age during the phase of maintenance caffeine treatment and observed similar sleep organization compared to the five preterm infants not on caffeine.¹⁰ In the earlier, smaller-sized polysomnographic study of Gabriel et al, sleep characteristics were not affected during treatment or after withdrawal.¹⁶ Other studies used behavioral observation for sleep assessment of caffeine-treated infants. Koch et al demonstrated GA-dependent changes in sleep distribution of preterm infants with increased caffeine concentrations over the first 5 days.¹³ With increasing wakefulness, active sleep decreased, whereas quiet sleep remained unchanged in infants with GA ≥ 28, but no caffeine effects were evident in sleep-wake behavior. Thoman et al observed four preterm infants at least 1 month after theophylline and found decreased active sleep compared to the NCG and term infants.¹⁵ Hassanein et al evaluated cerebral cortical activity by amplitude-integrated electroencephalography 2 hours after the caffeine administration and detected decreased quiet and active sleep with increased alertness compared to baseline.¹¹ Increased sleep-related movements measured by actigraphy were noted by Hayes et al with increased methylxanthine duration. However, wakefulness and sleep-related movements decreased at night hours, which was attributed to accumulated sleep dept secondary to the stimulatory effects of ongoing methylxanthine exposure.¹⁴ Contradictory results from studies may be related to the study design and timing of sleep assessment. Except for the study of Thoman et al, all studies at infancy focused on the acute effects during loading dose or maintenance therapy.¹⁵ Our study is unique in the evaluation of the midterm effects of caffeine treatment on sleep at 6 months of CA when substantial brain maturation is still ongoing. In addition, we evaluated sleep by two different methods (PSG and actigraphy) and found no differences in sleep parameters of preterm infants in the CG and NCG.

Several studies used actigraphy in the sleep assessment of preterm infants,^{17,19,20,46,47} even though validity studies revealed variable results depending on the method used for comparison, GA at birth, or age and setting at assessment. However, compared with other methods of sleep assessment in infants, such as PSG and direct observation, actigraphy is less time-consuming and laborious, allowing prolonged periods of measurement in the home environment.^48,49 In this study, we contribute to the growing evidence about the use of actigraphy in preterm infants by comparing sleep variables derived from actigraphy with PSG. The automatic threshold showed the highest sensitivity (91%) and agreement rate (77%) can be used in preterm infants to predict sleep-wake patterns at 6 months of CA. Of note, use of actigraphy may be limited due to poor predictive ability to detect wake after sleep, as we report low values for PVW. Supporting our findings, So et al previously validated actigraphy against PSG and reported similar sensitivity and agreement rate but low PVW in a group of term and preterm infants at 5–6 months of age, which was attributed to the relatively low number of wake epochs.¹⁷ Due to reliability issues reported in the literature, other reliability statistics are recommended to report.⁴⁸ Accordingly, PABAK analysis in this study showed moderate agreement at all thresholds, and Bland–Altman statistics revealed concordance between sleep and wake in high and automatic activity thresholds. In addition, we used actigraphy recordings with concurrent sleep diaries as suggested in pediatric sleep studies.⁵⁰

Studies about the use of caffeine therapy for AOP demonstrated proven benefits for premature infants with regard to decreasing bronchopulmonary dysplasia and severe retinopathy of prematurity and improved survival. Besides, it is one of the few neuroprotective strategies with shown benefits in reducing cerebral palsy at 18–21 months of CA and improved motor function at 5 and 11 years of age, possibly due to reduced exposure to hypoxemia episodes.^6–8 These studies focused on the long-term neuroprotective benefits of caffeine treatment for AOP. On the other hand, sleep is crucial in the neurocognitive development of infants, and one can hypothesize that the sleep-suppressing effects of caffeine may affect the neurodevelopment of infants. In this study, we showed that there was no demonstrable effect of caffeine on early neurodevelopment related to the sleep pattern changes.

Our study has several limitations. It is a single-centered, small-sized study, due to intensive sampling criteria, and findings could be more precise in a larger sample size. The sample size is not unique in terms of GA, which might affect sleep characteristics. However, various methods of sleep assessments (PSG, actigraphy, and the Brief Infant Sleep Questionnaire) and neurological assessments were employed, each supported by evidence in the literature for utilization in infants. Second, the PSG recordings included only daytime naps. Of note, a total of 6,098 epochs of PSG and actigraphy recordings were compared. Another limitation is that infants were not randomly assigned to the caffeine and no-caffeine group due to ethical considerations, and this resulted in older GA in the NCG, which was also described in other studies. Therefore, HINE scores were corrected for GA when examining the differences in neurological outcomes. It should also be noted that the findings do not represent the acute effects of caffeine treatment on the sleep of preterm infants, since it was beyond the scope of this study. Lastly, although we found similar sensitivity and specificity values of actigraphy reported in the literature, comparison of PABAK values was not possible due to the limited availability of other reliability statistics in previous reports. Therefore, one should consider the moderate agreement between PSG and actigraphy according to kappa statistics and strengthen the actigraphic sleep data with other tools, such as sleep diaries, depending on the research or clinical purpose. Future studies on the use of actigraphy in preterm infants should report kappa or Bland–Altman statistics to provide a comprehensive assessment of the agreement.

In conclusion, neonatal therapeutic caffeine administration was not associated with changes in sleep structure at 6 months of CA. In addition, caffeine used in preterm infants does not appear to negatively impact the neurodevelopment of infants assessed at 6 months of CA. Future large-scale studies are needed to support the findings.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. This study was conducted in a single-center Level III neonatal intensive care unit at the Marmara University Hospital located in Istanbul, Turkey. Actiwatch devices used in this study were sponsored in a previous project of Perran Boran by Marmara University Scientific Research Commission, BAPKO (SAG-C-TUP-131217-0647). The authors declare that no additional funds, grants, or other support were received during the preparation of this manuscript. The authors have no relevant financial or nonfinancial interests to disclose.

ABBREVIATIONS

AOP

apnea of prematurity

BSID-III

Bayley Scales of Infant Development, third edition

CA

corrected age

CG

caffeine group

GA

gestational age

HINE

Hammersmith Infant Neurological Examination

NCG

noncaffeine group

NICU

neonatal intensive care unit

PABAK

prevalence-adjusted and bias-adjusted kappa

PSG

polysomnography

PVW

predictive value of wake