Nocturnal sleep duration trajectories in early childhood and school performance at age 10 years

Sleep plays a fundamental role in brain development and resultant functions. The aim was to verify whether nocturnal sleep duration during early childhood has long‐term associations with academic achievement at age 10 years. The present study is part of the Quebec Longitudinal Study of Child Development, a representative cohort of infants born in 1997–1998 in the province of Quebec, Canada. Children with known neurological conditions were excluded from this cohort. Four trajectories of parent‐reported nocturnal sleep duration at ages 2.5, 3, 4, 5 and 6 years were determined using a SAS procedure named PROC TRAJ. Sleep duration at age 10 years was also reported. Teachers provided data on academic performance when the children were age 10 years. These data were available for 910 children (430 boys, 480 girls; 96.6% Caucasians). Univariate and multivariable logistic regressions were performed using SPSS. Children who slept less than 8 hr per night at 2.5 years but normalized later on (Traj1) had three–five times the odds of having grades below the class average in reading, writing, mathematics and science compared with children who slept sufficiently (Traj3–4: 10–11 hr per night). Children who slept about 9 hr per night throughout childhood (Traj2) had two–three times the odds of being below the class average in mathematics and science. Sleep duration at age 10 years was not correlated with the academic performance. These results point to the presence of a very important early period during which sufficient sleep is needed to fine‐tune the functions necessary for academic achievement later on.


Summary
Sleep plays a fundamental role in brain development and resultant functions. The aim was to verify whether nocturnal sleep duration during early childhood has long-term associations with academic achievement at age 10 years. The present study is part of the Quebec Longitudinal Study of Child Development, a representative cohort of infants born in 1997-1998 in the province of Quebec, Canada. Children with known neurological conditions were excluded from this cohort. Four trajectories of parent-reported nocturnal sleep duration at ages 2.5, 3, 4, 5 and 6 years were determined using a SAS procedure named PROC TRAJ. Sleep duration at age 10 years was also reported.
Teachers provided data on academic performance when the children were age 10 years.
These data were available for 910 children (430 boys, 480 girls; 96.6% Caucasians). Univariate and multivariable logistic regressions were performed using SPSS. Children who slept less than 8 hr per night at 2.5 years but normalized later on (Traj1) had three-five times the odds of having grades below the class average in reading, writing, mathematics and science compared with children who slept sufficiently (Traj3-4: 10-11 hr per night).
Children who slept about 9 hr per night throughout childhood (Traj2) had two-three times the odds of being below the class average in mathematics and science. Sleep duration at age 10 years was not correlated with the academic performance. These results point to the presence of a very important early period during which sufficient sleep is needed to fine-tune the functions necessary for academic achievement later on. Sleep is an essential pillar of neurodevelopment. During the first years of life, sleep occupies more than half of each day and is the primary source of brain activity. Sleep thus plays a crucial role in brain development and maturation (Kurth et al., 2013) that lies beneath the astonishingly rapid development observable in early childhood. Sleep is an important player in emotional regulation, memory and several other cognitive domains in children. Cognitive functioning has been consistently related to academic achievement in preschoolers and school-age children (Best et al., 2011;Espy et al., 2004;Willoughby et al., 2012). When investigated directly, the relation between sleep and school functioning was often studied concurrently. This makes it hard to disentangle two possible underlying mechanisms: sleepiness due to current/recent sleep loss versus impairment in neuronal development caused by persistent insufficient sleep in early childhood. For example, a meta-analysis (Astill et al., 2012) of 52 studies investigated the relation between sleep duration and cognition or school performance and found an association. However, few studies focused on preschoolers and even fewer were longitudinal in nature. Among these, our study (Touchette et al., 2007) had shown that trajectories of short nocturnal sleep durations in early childhood were associated with poorer performance on cognitive tasks at ages 5 and 6 years when compared with trajectories of sufficient sleep. We had also shown later that the trajectory of short sleepers had higher odds of presenting poor receptive vocabulary at age 10 years (Seegers et al., 2016). An actigraphy study demonstrated that children with later sleep timing at 30 months had poorer academic performance at 54 months (Hoyniak et al., 2020). Another group has rather demonstrated that the decrease of total sleep duration between ages 2 and 4 years was related to academic achievement when the children were age 7 years (Bernier et al., 2021), likely due to the decrease of daytime sleep duration. Finally, one longitudinal study found that sleep problems trajectories in early childhood were associated with poorer cognitive functioning and poorer school achievement later (Williamson et al., 2020). That latter study did not look at sleep duration per se, and this is of major importance considering that duration was identified as the important sleep parameter for different outcomes in children (Astill et al., 2012;Chaput et al., 2017). Moreover, populational studies reported a growing percentage of children with insufficient sleep duration (Iglowstein et al., 2003;Matricciani et al., 2012).
The objective of the present study was to determine whether the trajectories of sleep duration in early childhood have a longer-term association with academic achievement at age 10 years, as assessed by the teacher, or whether current sleep duration is really the important factor in school success.
Based on previous findings, it is hypothesized that a short sleep duration during early childhood will be associated with poorer academic performance at age 10 years, irrespective of sleep duration at that age.

| Participants
This study was part of the Quebec Longitudinal Study of Child Development (QLSCD). All children were recruited from the Quebec Master Birth Registry. A randomized stratified survey design was used to study a representative sample of infants born in 1997-1998 in the province of Quebec, Canada. At the QLSCD inception (March 1998), 2223 children aged 5 months were included. Throughout the years, attrition occurred (n = 1402 at age 10 years). For the present study, we had to exclude the children for whom the teacher's assessments were unavailable (n = 411) and children with known major neurological conditions (autism, epilepsy, mental incapacity; n = 6). Finally, of the remaining sample, we excluded children for whom the sleep duration trajectory could not be drawn (n = 75). Therefore, 910 children, aged 10 years were included. All families signed a consent form before each assessment. The protocol was approved by the "Institut de la Statistique du Québec" ethics committee.

| Nocturnal sleep duration trajectories
Nocturnal sleep duration was measured at 2.5, 3.5, 4, 5, 6 and 10 years old by an open question to the mother: "Indicate how long in total your child sleeps during the night (on average). Do not count the hours that your child is awake". Using all available data points of the QLSCD cohort when children were age 6 years, a trajectory methodology (described in more details elsewhere; Touchette et al., 2007;Touchette et al., 2008)   (e) Below the middle of the class, but above the bottom; or (f ) Near the bottom of the class. Each of the five variables was then dichotomized in the following manner: b + c + d versus e + f (a was considered as missing).

| Potential confounders
Because of their associations with academic outcomes reported in the literature, the following 13 potential confounders were taken into account: sex of the child (Anastas & Reinherz, 1984); prematurity (Patil & Metgud, 2014); low birth weight (Chatterji et al., 2014); ethnic origin (Nitardy et al., 2015); immigrant status of the mother (Sullivan et al., 2016); language spoken at home (Bermejo et al., 2021); insufficient family income (Hair et al., 2015); maternal education level (Tamayo Martinez et al., 2022); coercive parental practices (Alyahri & Goodman, 2008); smoking (Kristjansson et al., 2018); or use of drugs or alcohol during pregnancy (Richardson et al., 2002); and maternal depression (Murray et al., 2010). These confounders were assessed through a computerized questionnaire completed by an interviewer at the home of the families when the child was 5 months old, except for parental practices, which was part of an auto-administered questionnaire completed by the mother. The insufficient family income is defined as a household income below the pre-tax low-income cut-off established by Statistics Canada for that year.

| Statistical analyses
Descriptive statistics compared demographic data of the study sample with those of children lost to follow-up or with missing data. The distribution of the children of the present sample within the four sleep duration trajectories was also compared with that of children lost to follow-up or with missing data, and with that of the original sample from which these trajectories were established. To examine whether sleep duration trajectories in early childhood predict later academic performance, univariate and multivariable logistic regressions were performed.
As aforementioned, the adjusted model took into account the following 13 potential confounders. Odds ratios (ORs) are reported with their 95% confidence intervals (CIs). Considering the well-known association between concurrent sleep duration and academic functioning, we first tested whether current sleep duration (age 10 years) was correlated with the five academic outcomes variables. Alpha levels were set at 0.05, and SPSS version 26 (IBM) was used for all statistical analyses. Table 1 presents sociodemographic data of children who were included in the present study, of children lost to follow-up or with missing data, and of those of the initial representative sample. Compared with both children lost to follow-up or excluded and with the initial sample, the studied sample had fewer boys ( p = 0.003), fewer children of immigrant mothers ( p < 0.001), fewer allophones (p < 0.001), and fewer cases of insufficient family income (p < 0.001).

| RESULTS
In addition, the present sample has a smaller proportion of children in Traj2 and a greater proportion of children in Traj1 compared with the sample from which these trajectories were first established (Table 2).
We pooled together Traj3 and Traj4 for the present analyses as they both respect the sleep duration recommendations for preschool children (Chaput et al., 2018;Hirshkowitz et al., 2015;Paruthi et al., 2016). Moreover, there were no significant differences between Traj3 and Traj 4 on the outcome variables studied. Table 3 presents the sociodemographic data for the present sample as a function of the three nocturnal sleep duration trajectories.
The main results are presented in Figure 2 and Table 4. Compared with children in Traj3-4 (sleeping 10-11 hr per night), a greater percentage of children in Traj1 (very short sleep duration but increasing later on) were rated by their teacher as performing below the class average in reading, writing, mathematics, science and overall (Figure 2; all p < 0.001).
In fact, children in Traj1 had 3.2-5.5 times the odds of performing below the class average in these subjects compared with children who had sufficient sleep during early childhood (Table 4, unadjusted model). Children in Traj2 (9 hr per night) had 2.3-3.2 times the odds of being below the class average in mathematics and science compared with children in Traj3-4. Results remained the same, with slightly lower OR for Traj1 than in unadjusted analyses, even after controlling for the following potential confounders: sex of the child, prematurity, low birth weight, ethnic origin, insufficient family income, maternal immigrant status, language spoken at home, maternal education level, coercive parental practices, smoking or use of drugs or alcohol during pregnancy, and maternal depression (Table 4 adjusted   model).

| DISCUSSION
The present study aimed to determine whether the trajectories of nocturnal sleep duration in early childhood have long-term associations with academic success at age 10 years or whether current sleep duration is the important factor for such success.
Our study showed that short sleep duration, especially if occurring very early in life, is strongly associated with increased risk of having grades below the class average near the end of elementary school. Moreover, sleep duration at age 10 years showed no correlation with the academic performance. In addition, despite similar mean sleep durations at age 10 years for Traj1 and Traj2, children in Traj1 who T A B L E 1 Demographic data of the study sample and comparison with the sample of children lost to follow-up or with missing data  functioning. The first found associations between early childhood sleep profiles and teacher-reported poor social-emotional school adjustment but did not report on academic performance (Williams et al., 2016). The other epidemiological study looked at how sleep trajectories in early childhood affect later cognitive functioning and academic achievement (Williamson et al., 2020). It reported that children in the persistent sleep problems trajectory had more teacher-rated language/literacy and mathematical thinking impairments than did children in the other trajectories. However, in these two studies the trajectories were based on the perception of a sleep problem rather than on sleep duration.
One longitudinal study (Bernier et al., 2021) did investigate early childhood sleep duration trajectories in relation to performance and mathematics in grade 1 (age 6 years). They found that children whose sleep duration decreased more rapidly between ages 2 and 4 years performed better in school.
The key difference between their study and ours is that they measured total sleep duration, which includes daytime naps. It is well known that daytime sleep duration normally and progressively decreases from age 2 years to age 4 or 5 years (Weissbluth, 1995). Also, there is high variability in the age and speed at which daytime sleep subsides and this, as Bernier and colleagues pointed out, seems to reflect brain maturation. In contrast, trajectories of nocturnal sleep duration are more stable over time, and our aim was to look at correlates of short or insufficient sleep duration over childhood.
Finally, we did not find any correlations between sleep duration at age 10 years and school performance at the same age. This is a bit surprising in light of the previously mentioned reviews (Astill et al., 2012;Chaput et al., 2016), showing that even cross-sectional studies showed a relationship in most cases (but not all). However, it must be mentioned that when this relationship is demonstrated, the correlation coefficients can be quite small. In addition, many studies that did not find a notable association between sleep duration and either current or even later school performance included older children (Li et al., 2013;Mayes et al., 2008) or adolescents (Dewald et al., 2010;Eliasson et al., 2002;Loessl et al., 2008). In fact, a metaanalysis of 17 studies (Dewald et al., 2010) reported that effect sizes were larger for studies that included younger participants. Perhaps other measures of sleep such as daytime somnolence or the difference between weekend and school nights sleep duration would have been more telling in older children.

| Sensitive period for the role of sleep in development
Our findings point to a very important period for the effect of sleep on development. Using the same trajectories, we have shown previously that the children who slept insufficiently during early childhood, but whose sleep duration later normalized to about 10 hr of sleep per night (Traj1), were nonetheless more at risk for poor performance on cognitive tasks (Touchette et al., 2007). They were also more at risk of being overweight or obese at age 6 years, even after adjusting for (2.7-11.5) < 0. The reference for sleep duration trajectories in the logistic regressions was Traj3-4. c Unadjusted OR (95% CI) without controlling for potentially confounding factors.
d Adjusted OR (95% CI) while controlling for sex of the child, prematurity, low birth weight, ethnic origin, language spoken at home, maternal immigrant status, insufficient family income, maternal education level, coercive parental practices, smoking or use of drugs or alcohol during pregnancy, and maternal depression.
CI, confidence interval; OR, odds ratio. many potentially confounders (Touchette et al., 2008). Likewise, a longitudinal twin study showed that poor sleep consolidation during the two years of life, which often results in sleep curtailment, was a risk factor for language skills at age 5 years (Dionne et al., 2011).
Compared with many species, humans are very immature at birth.
Brain development must then continue after birth, structurally and functionally, and sleep contributes greatly to this development. A relation was shown between the immaturity of a newborn mammal and the amount/percentage of rapid eye movement (REM) sleep it has at birth (Zepelin, 2005). Human infants sleep 14-18 hr a day, and REM sleep occupies half of that time. This percentage slowly decreases to reach adult levels around age 5 years, but stays relatively high during the first 3 years. Similarly, non-REM sleep activity and connectivity patterns undergo huge transformations during the first years of life, and these changes reflect brain maturation (Kurth et al., 2013). Consequently, if sustained sleep loss occurs during that sensitive period, some functions will likely suffer. The prefrontal cortex, the seat of higher executive functions and emotional regulation, has been shown to be particularly sensitive to sleep deprivation (Horne, 1993). Moreover, it is the region that matures the latest (Teffer & Semendeferi, 2012); which makes it more vulnerable to disruptions occurring in the first years of life. Indeed, it was shown that higher volume in orbitofrontal and prefrontal cortex (as well as temporal cortex, precuneus and supramarginal gyrus) was correlated with longer sleep duration in children (Cheng et al., 2021). Moreover, higher volume in these regions was associated with higher cognitive scores (Cheng et al., 2021).
The differential academic outcomes obtained between children in Traj1 and Traj2 might be explained by the presence of genetically determined true short sleepers mixed in the group of children of Traj2.
True short sleepers are individuals who really need less sleep to function normally. Indeed, a longitudinal twin study observed a strong heritability (71% of the variance) for the short persistent trajectory (Traj2; Touchette et al., 2013). These individuals will be sleeping less throughout their life without significant consequences on their daily functioning. By opposition, only common and unique environment accounted for the variance in the short increasing sleep trajectory (Traj1), which showed more negative outcomes (Touchette et al., 2013).

| Preventive and corrective actions
Contrary to popular belief, early childhood sleep problems do not always spontaneously disappear, especially night wakings, which are responsible for the reduction in sleep duration (Touchette et al., 2005). In some cases they do but in many others they tend to persist over several years if left untreated (Byars et al., 2012;Kataria et al., 1987). What then is there to do to insure sufficient sleep for a specific child in that very important period? First, as already mentioned, there are true short sleepers, even at this young age, who will not suffer any functional daytime deficits. It is an uncommon and heri-  (Touchette et al., 2005;Yu et al., 2017). A review by the American Academy of Sleep Medicine (Mindell et al., 2006) on the management of bedtime problems and night wakings in young children focuses indeed on helping parents adopt appropriate bedtime behaviours (Mindell & Williamson, 2018) with their child. In fact, strong empirical support has been reported for the following paediatric sleep recommendations to parents (Allen et al., 2016): ensuring age-appropriate sleep duration, establishing bedtime routines, and teaching children to develop their sleep autonomy. However, more research is needed to quantify the long-term impacts of those recommendations on child development, including cognition.

| Limitations of the study
The first limitation is that sleep duration trajectories were not based on objective measures but rather on parental reports, which usually overestimate sleep duration (Dayyat et al., 2011

DATA AVAILABILITY STATEMENT
Data has been obtained from a third party: The data analyzed in this study was obtained from the Institut de la statistique du Québec and, as stipulated in the clauses 10 and 11 of the Institut de la statistique's Québec Act (Canada), the access to the data is restricted to the parties identified in the partnership agreement signed to ensure the conduct of the study and which describes the author's right. In the QLSCD cohort, the participants only consented to share their data to the study's financial partners and affiliated researchers and their collaborators. Those partners and researchers have only access after signing a data sharing agreement. Requests to access these data can be directed to the Institut de la statistique du Québec's Research Data Access Services -Home (www.quebec.ca). For more information, contact Marc-Antoine Côté-Marcil (SAD@stat.gouv.qc.ca).