The Architecture of Early Childhood Sleep Over the First Two Years

Introduction The architecture and function of sleep during infancy and early childhood has not been fully described in the scientific literature. The impact of early sleep disruption on cognitive and physical development is also under-studied. The aim of this review was to investigate early childhood sleep development over the first two years and its association with neurodevelopment. Methods This review was conducted according to the 2009 PRISMA guidelines. Four databases (OVID Medline, Pubmed, CINAHL, and Web of Science) were searched according to predefined search terms. Results Ninety-three studies with approximately 90,000 subjects from demographically diverse backgrounds were included in this review. Sleep is the predominant state at birth. There is an increase in NREM and a decrease in REM sleep during the first two years. Changes in sleep architecture occur in tandem with development. There are more studies exploring sleep and early infancy compared to mid and late infancy and early childhood. Discussion Sleep is critical for memory, learning, and socio-emotional development. Future longitudinal studies in infants and young children should focus on sleep architecture at each month of life to establish the emergence of key characteristics, especially from 7–24 months of age, during periods of rapid neurodevelopmental progress.


Introduction
The architecture of sleep changes markedly over the first two years of age. Despite increasing research interest, gaps remain in our knowledge of infantile sleep, its precise function, and the impact of sleep disruption in infancy and early childhood on cognitive and physical development. Few recommendations exist to educate healthcare providers and parents of young children about the importance of sleep and its impact on overall health (Mukherjee et al., 2015). The majority of research on neonatal sleep generally includes older children, adults, and animal models. Sleep studies focusing on normal sleep in early childhood are necessary to ensure greater understanding of the role sleep plays in cognitive development (Jiang, 2019;Tham et al., 2017).
Neonatal sleep is characterized by Active Sleep (AS), Quiet Sleep (QS), and Indeterminate Sleep (IS). Active Sleep is also called Paradoxical Sleep (Samson-Dollfus et al., 1983); it is characterised by rapid eye movements, irregular breathing, body and limb movements, low voltage electroencephalography (EEG), and high variability in 1 3 heart rate (Barbeau & Weiss, 2017;Mirmiran et al., 1993). AS is believed to play a role in the maturation of the central nervous system (CNS) and facilitate growth and development (Denenberg & Thoman, 1981;Mirmiran et al., 1993). Evidence of the role of AS in brain development comes from research using animal models (Bertelle et al., 2005;Mirmiran, 1995;Mirmiran et al., 1983Mirmiran et al., , 1988Mirmiran et al., , 1993. QS is characterised by reduced eye movements, regular breathing, decreased body movements, slow wave activity on the EEG, and low variability in the heart rate (Barbeau & Weiss, 2017;Mirmiran et al., 1993). It is presumed that QS plays a role in energy maintenance, the release of growth hormones (Bertelle et al., 2005) and has a restorative function (Mirmiran, 1995). After two months of age AS becomes Rapid Eye Movement (REM) Sleep, and QS becomes non-Rapid Eye Movement (NREM) Sleep. When elements of both AS and QS are present, it is described as indeterminate sleep (Bertelle et al., 2005;Korotchikova et al., 2009). This stage is considered a transitional stage or measure of immature sleep (Louis et al., 1997).
Sleep structure in early childhood is very different to sleep in adulthood. At birth, an ultradian rhythm dominates, and infants and young children spend a greater period of time asleep in a 24 h period and have different EEG patterns to older children and adults during sleep (Barry, 2020). Rapid changes occur in sleep structure within the first 12 weeks and continue throughout childhood (Dittrichova, 1966). Throughout this period, sleep is restructuring and reorganising in parallel with rapid brain growth and neuroplasticity (Bes et al., 1988). This is best seen on EEG, where patterns of sleep undergo swift changes especially during the first few months, e.g. the disappearance of tracé alternant, the appearance of sleep spindles (Jenni et al., 2004), and the disappearance of focal sharp waves (Ellingson & Peters, 1980). Sleep has been associated with learning and memory and emotional regulation. However the role of sleep during brain development in early childhood is not well known. (Jiang, 2019). Understanding sleep and its role in development is necessary to; (1) Ensure caregivers have the necessary information to support normative development of infant and young children's sleep  (2) Understand the development of normative sleep and the range of normative sleep variables (Jiang, 2019;Paavonen et al., 2020), and (3) Investigate the potential of sleep as a target for early intervention to optimize development (Tham et al., 2017).
The aim of this narrative review is to discuss available literature relating to sleep development and the impact of sleep on neurodevelopment in the first two years and to identify gaps in the scientific literature at this critical stage of neurodevelopment.

Search Strategy
The search strategy for this narrative review was conducted according to the 2009 Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines (Liberati et al., 2009). Inclusion and exclusion criteria were established, databases were identified, search terms and data to be collected were agreed upon by authors SL and LF. In May 2021, SL searched four databases (OVID Medline, Pubmed, CIMAHL, and Web of Science). The full search strategy is shown in Table 1. No limitations were applied to the publication years. The papers were extracted using EndNote™ Version X9. Duplicates were removed. Extracted papers were included or excluded based on title and abstract. Full articles were screened for inclusion based on the below criteria. The reference lists of all relevant articles were checked manually for papers potentially missed by the search.

Inclusion Criteria
Studies were included if: • Participants were healthy infants and young children up to the age of 2 years. • More than five subjects were included. • Objective (polysomnography, actigraphy) or subjective (self-reported) measures of sleep were recorded, such as sleep duration, night-waking, sleep latency, longest sleep period, daytime naps, sleep architecture, and sleep efficiency. • Sleep variables and neurodevelopmental or cognitivedevelopmental outcomes were reported.

Exclusion Criteria
Studies were excluded if: • Participants were born preterm (< 37 weeks' gestational age) or admitted to a Neonatal Intensive Care Unit (NICU). • Cohort had a mean age > 24 months. • Behavioural interventions were performed. • Studies were validation of methodology or computational models, and did not contribute information on sleep. • Cohorts were duplicated, unless assessed at different time points.
1 3 • Sleep data or development data were not reported in results. • Type of publication was a review article, book chapter, dissertation, or conference abstract. • No English translation was available.

Data Collection
Author, year, sample size, assessment age, sleep variables reported in the studies, and developmental assessments were collected. Reported hours of sleep were collected from each article for 24 h and/or 12 h for; total sleep time (TST), AS, QS, IS, REM, and NREM. The weighted mean was calculated for each variable that had data available using the formula: weighted mean = sum of (number x weighting factor)/ sum of all the weights.

Results
Results from the search strategy are shown in Fig. 1. Ninetythree studies assessed sleep in approximately 90,000 subjects. Table 2 lists the characteristics of the studies, including sample size (male:female), cohort age, sleep variables, sleep assessment tools, and development assessment. Tools included EEG, actigraphy, questionnaires, and sleep diaries. Most studies used one tool to assess sleep, eighteen used two, and four used three tools. Forty-one studies assessed sleep using one assessment in the first year, 31 studies used multiple assessments over the first year, 6 studies used one assessment in the second year, and two studies used multiple studies over the second year. Four studies used one assessment with the cohort ranging over the first two years. Table 3 shows the weighted means of daytime sleep, nighttime sleep, and TST over 24 h. Figure 2 provides a visual summary of the development of sleep in the first year, showing the appearance and disappearance of tracé alternant, the appearance of slow wave sleep and spindles, and the number of hours of sleep in a 24 h period.

Night-time Sleep
Data for night-time sleep were collected from 17 studies (See Table 3). Night-time sleep increased between 1-7 months (6.7 h to 11.7 h) and decreased at eight months (9.5 h). Sleep increased again at nine months (10 h) and decreased at 24 months (7.4 h). High amounts of night-awakenings were observed between 17-20 weeks (Giganti et al., 2001). Night-awakening decreased between 6-18 months (Blair et al., 2012). Infants who were crawling between 5-8 months of age had more disrupted sleep compared to their precrawling peers (Scher & Cohen, 2005;Scher, 2005a). At six months (n = 388) of age, 37.6% of infants slept for less than six consecutive hours at night, while 43% of infants slept for eight consecutive hours. At 12 months (n = 369), 72.4% of infants slept six consecutive hours at night and 56.6% slept eight consecutive hours (Pennestri et al., 2018). Occasional sleeping throughout the night did not prevent  1 3 future night-awakenings (Pennestri et al., 2020). Swaddling resulted in decreased night-awakenings (Franco et al., 2005). Breastfed infants had significantly more night-awakenings compared to formula-fed infants (mean (SD) 1.63 (1.24) vs 0.94 (0.87), p = 0.003 ) but this resolved within the first year (Eaton-Evans & Dugdale, 1988;Mindell et al., 2012;Tikotzky et al., 2010b), and infants who were nursed back to sleep in the first year also had more night-awakenings (Ramamurthy et al., 2012). Cosleeping was reported as a risk factor for shorter sleep duration and more frequent night-awakenings from 6-18 months (Hysing et al., 2014). One hour of exposure to electronic screen-based media at four months resulted in 13 min less nocturnal sleep (Ribner et al., 2019).

Total Sleep Time
Total sleep time over a 24 h period was available from 10 studies. The average TST was 13.13 h at one month, 15.7 h at two months, 12.2 h at one year, and 13 h at 24 months (Bamford et al., 1990;Figueiredo et al., 2016;Jacklin et al., 1980;Wooding et al., 1990). The number of sleep episodes decreased from 6.1 at two weeks to 5.2 at six months (Figueiredo et al., 2016). The duration of time awake at two weeks increased from 8.7 h to 10 h at six months, while the longest period of sleep increased from 3.2 h at two weeks to 5.6 h at six months (Figueiredo et al., 2016). Sleep efficiency (ratio of TST to time spent in bed) and periodicity increased by 12 months, with a shift towards night sleep, over daytime sleep (Louis et al., 1997;Wooding et al., 1990).
Summer born infants were reported to have shorter TST (Karki et al., 2019;Kärki et al., 2020). Parental presence at bedtime, frequency of night-awakenings, and less TST at one month of age predicted poor sleep at six and twelve months (Henderson et al., 2020). Greater parental involvement in both daytime and night-time care at three months predicted more consolidated maternal and infant sleep at six months (Tikotzky et al., 2015).

Sleep States
The duration of QS was higher on the day of birth compared to one day later in 19 infants born vaginally and 17 born by emergency Caesarean-section (CS) after a period of labour. This may be a temporary response to the stress of labour (Carroll et al., 1999). From birth, AS occupies a greater percentage of TST (Ellingson & Peters, 1980;Hoppenbrouwers et al., 1988;Korotchikova et al., 2016), with QS occupying less than half and IS occupying 5-13% of TST (Ellingson & Peters, 1980). Sleep in neonates begins in AS rather than QS/NREM sleep (Ashton, 1971;Ellingson & Peters, 1980;Hoppenbrouwers et al., 1988). As sleep matures over the first year, AS-onset gives way to QS-onset, and the percentage of TST spent in AS (between 50-80% to less than 50%) and IS decreases. The percentage of TST spent in QS increases to approximately 35-50% in the first year (Anders, 1978;Dittrichová et al., 1972;Ellingson & Peters, 1980;Fagioli & Salzarulo, 1997;Fattinger et al., 2014;Hoppenbrouwers et al., 1988;Jenni et al., 2004). After two months, AS and QS become REM sleep and NREM sleep, and the stages of NREM sleep begin to appear (Sankupellay et al., 2011). By six months, all sleep should begin with NREM onset (Ellingson & Peters, 1980). The increase in NREM sleep duration is associated with the appearance of slow-wave sleep (SWS), or NREM stage 3, showing the maturational restructuring of sleep (Ktonas et al., 1995;Peirano et al., 1993). A potential increase in the proportion of SWS was seen in exclusively breast-fed 3-4 month old infants (Yoshida et al., 2015).

Electroencephalogram Characteristics of Sleep
Within 6-12 h of birth, the EEG of healthy term neonates shows continuous, symmetrical, and synchronous activity at amplitudes of approximately 15-150 µV across all sleep states, with well-developed sleep-wake cycles (SWC) (Korotchikova et al., 2016). Markers of sleep homeostasis (low-frequency delta activity and declining theta activity throughout the night) are present in the first months (Jenni et al., 2004). REM or AS in healthy term neonates is seen on the EEG as continuous irregular, low-voltage activity with amplitudes of 15-35 µV at frequencies between 5-8 Hz. NREM/QS is continuous and synchronous with higher amplitudes of 50-150 µV.
Tracé alternant pattern is the most frequent EEG pattern detected during QS in healthy term neonates and is present from birth. Tracé alternant is characterized by bursts of high amplitude slow-wave activity (SWA) with intermixed faster frequencies, separated by lower amplitude mixed frequency activity (Eiselt et al., 2001;Ellingson & Peters, 1980) and begins to disappear rapidly from the EEG from 2 weeks of age and is not seen after six weeks (Bes et al., 1988;Ellingson & Peters, 1980). Sleep spindles are a hallmark of stage 2 NREM sleep. Sleep spindles should be present in NREM stage 2 on EEG recording by three months (Corsi-Cabrera et al., 2020;Ellingson & Peters, 1980;Horne et al., 2003;Jenni et al., 2004;Louis et al., 1992;Navelet et al., 1982;Sankupellay et al., 2011;Sterman et al., 1977). Spindles and SWS are believed to stimulate the development of thalamocortical networks by supplying endogenous neural signals with repetitive and co-ordinated activity (Jenni et al., 2004). The density of sleep spindles, and NREM proportions with sleep spindles was more frequent when the infant was placed supine rather than prone (Horne et al., 2003). In stage 2 NREM sleep, another characteristic waveform called the K complex can appear by as early as 5 months and although its exact function during sleep is not precisely known it is believed to play a role in sleep promotion and arousal. Vertex sharp waves usually appear in the EEG at the age of 5-6 months. Between 5-16 months, Cyclic Alternating Pattern (CAP) is present during NREM sleep as a physiologic oscillating pattern. CAP is important for the build-up and maintenance of sleep. It has 2 phases, A and B, with A having three subtypes based on EEG patterns. Subtype A1 is based on the prevalance of EEG synchrony, subtype A3 is based on the prevalence of EEG desynchronization, and subtype A2 is a combination of both. Miano et al. showed that a decreased frequency of the CAP A1 subtype may indicate maturation of the arousal system (Miano et al., 2011).

Sleep and Growth
Between 4-17 weeks, sleep was temporally related to growth (Lampl & Johnson, 2011). Short sleep duration was linked to weight gain and obesity in young children (Tikotzky et al., 2010b). Infants who slept less than 12 h per day at three months had higher body mass index (BMI) and shorter body length (Zhou et al., 2015). The search returned only these studies investigating the relationship between sleep and growth hormone secretion suggesting there is a gap in the literature.

Sleep and Mode of Delivery
Two studies reported the influence of mode of delivery on sleep. Neonates delivered vaginally and by emergency CS (following a period of labour) showed more sleep periods in the daytime compared to neonates born by elective CS (Korte et al., 2004). A significant decrease in AS and increase of QS in the EEG recordings of neonates delivered by elective CS was observed compared to neonates born vaginally or by emergency CS (Korotchikova et al., 2016). One study reported that mode of delivery did not affect sleep (Scher et al., 1995).

Sleep Deprivation
Neonates spent more time asleep after a period of sleep deprivation (Anders & Roffwarg, 1973). Short-term sleep deprivation in 6-18 week old infants was associated with development of obstructive sleep apnoea and a significant increase in the threshold for auditory arousal in sleep immediately after sleep deprivation (Franco et al., 2004).

Discussion
Rapid changes occur in sleep architecture over the first year and this review highlights the importance of sleep for neurodevelopment. Over the first two years, AS decreases while QS increases. The number of sleep episodes, TST, and nightawakenings all decrease with age, while the longest periods of sleep and wakefulness increase. This decrease in TST is not linear, as shown in Fig. 2. This may be due to the attainment of different developmental milestones, such as crawling and pull-to-stand as mentioned in this review (Atun-Einy & Scher, 2016;Scher, 2005a). Parents often report a period of unsettled or disrupted sleep around specific developmental milestones and some experts refer to this as sleep regression (Foley, 2020). There are inconsistencies in how sleep variables are reported in the literature. This makes comparisons between studies complex in many cases. Serious ethical concerns about studies involving sleep deprivation in infants also exist. Ander and Roffwarg used foundling infants (infants left by their parents at hospitals/churches etc. (The Foundling Museum, 2020)) in their study (Anders & Roffwarg, 1973), which raises ethical concerns. Sleep depriving 6-18 week old typically developing infants (Franco et al., 2004) also raises ethical concerns.
Some sleep studies were conducted during daytime naps, and others during nocturnal sleep. This can make it challenging to compare sleep variables between studies as daytime records are often shorter and daytime naps have disproportionately less REM as development progresses (Coons & Guilleminault, 1984). This can lead to differences in reference values (Coons & Guilleminault, 1984;Ellingson & Peters, 1980;Figueiredo et al., 2016;Louis et al., 1992).
A limitation observed in EEG studies was that some sleep recordings did not capture a full SWC, influencing the interpretation and analysis of the different sleep stages (Dittrichova, 1966;Korotchikova et al., 2016). Korotchikova et al. observed that a complete SWC was not captured for over 50% of their neonatal cohort, despite having an hour of EEG recording for each neonate. Based on their data, the authors recommended sleep EEG recordings of 150 min to ensure capture of a complete SWC (Korotchikova et al., 2016). The number of channels used for the EEG recording and the frequency of EEG recordings was a noted limitation in some studies (Sankupellay et al., 2011;Yoshida et al., 2015). Future studies should comprehensively characterise sleep EEG biomarkers over the first two years of life and correlate them with neurodevelopmental outcomes (Ventura et al., 2022).
One study using actigraphy raised concerns about reliability (Scher & Cohen, 2015a) in infant and young children studies, with another study arguing that actigraphy should be accompanied by a sleep diary (Bernier et al., 2010). Scher et al. highlighted they captured three nights of actigraphy instead of the recommended five nights (Scher & Cohen, 2015a).
The search terms used in this paper may have limited the scope of this review and missed some elements of sleep development, for example K-Complexes and vertex waves. In a review on sleep neurophysiology and maturation in infants and young children, Dan & Boyd discussed the development of K complexes and vertex sharp waves during sleep (Dan & Boyd, 2006). K-Complexes are usually present between 5-6 months of age and are well established by 18 months (Grigg-Damberger et al., 2007;Metcalf et al., 1971;Verma & Baisakhiya, 2021). However, the search strategy used in this review did not capture any studies studying these features. This should be taken into consideration when reading this review.
A strength of this narrative review was that the search strategy was developed using the PRISMA guidelines, resulting in the search being completed in a systematic 1 3 way with predefined inclusion and exclusion criteria. The narrative review allowed us to look at all methods used for assessing sleep and allowed for a wider scope for inclusion of literature in this review, as systematic reviews are often restrictive with their research question.

Conclusion
Despite a considerable volume of research on sleep in the first year, research on the impact of sleep on neurodevelopment is lacking. Infants and young children spend most of the time asleep, with QS increasing and AS decreasing during maturation (Figueiredo et al., 2016). Studies highlighted the importance of sleep for learning and memory. While it is difficult to investigate the impact of sleep disruption and sleep deprivation on neurodevelopment, infants admitted for long periods in the NICU may shed some light on the effect of sleep disruption in early life on long-term outcome (Levy et al., 2017;van den Hoogen et al., 2017). However, this cohort may have sleep disruption and poor neurodevelopmental outcomes due to underlying pathology. An example of a population which may be more suited to longitudinal study of sleep disruption and neurodevelopmental outcome are infants and young children with moderate-to-severe eczema. These young children are typically developing but have disrupted sleep due to itch (Camfferman et al., 2010).
A lack of literature documenting sleep in healthy 7-24-month-old children was found during this review. As a result, gaps in knowledge about the maturation of sleep in the first two years of age exist. Longitudinal studies require a significant commitment, and monthly appointments over a 2-3 year period may be inconvenient for participating parents. To fill in these knowledge gaps, future studies should consider assessing sleep in the second year of life, as this continues to be an important period of plasticity and development (Cusick & Georgieff, 2017).
Uniformity in scoring sleep stages is recommended, as well as a consensus on whether night-time or daytime sleep studies are best. Alternatively, it may be necessary to compare daytime sleep studies with other daytime studies and similar for nocturnal studies to ensure consistent and reliable reference values. Given the importance of sleep for neurodevelopment, future longitudinal studies may need to focus more closely on specific time points in the first two years and use EEG to identify normal sleep neuro-biomarkers.