Branched‐chain amino acids and sleep: a population‐derived study of Australian children aged 11–12 years and their parents

Summary Micronutrients, particularly amino acids, are thought to play an important role in sleep regulation and maintenance. While tryptophan is a known predictor of sleep, less is known about branched‐chain amino acids (BCAAs), which compete with tryptophan for transport across the blood–brain barrier. The aim of this study was to determine the association between BCAAs and actigraphy‐derived sleep duration, timing and efficiency, and self‐reported trouble sleeping. This study examined data on children and adults collected as part of the Child Health CheckPoint study. Linear mixed models, adjusted for geographic clustering, were used to determine the association between BCAAs and sleep characteristics. Complete‐case analysis was conducted for 741 children aged 11–12 years old (51% females) and 941parents (87% mothers). While BCAAs were significantly associated with children's sleep duration, timing and self‐reported trouble sleeping, no associations were observed in adults, in fully adjusted models. In children, higher levels of BCAAs are associated with shorter sleep duration, delayed sleep timing, and more frequent reports of trouble sleeping.

and fewer night-time awakenings (Silber & Schmitt, 2010;Sutanto et al., 2022). Tryptophan is thought to cross the blood-brain barrier where it is involved in the synthesis of serotonina precursor of the sleep-promoting hormone melatonin (Binks et al., 2020). Although tryptophan promotes sleep, it competes with branched-chain amino acids (BCAAs) at the crossing of the blood-brain barrier (Binks et al., 2020;Humer et al., 2020). BCAAs are therefore thought to promote wakefulness by reducing tryptophan transport through the blood-brain barrier and hence the synthesis of serotonin (Humer et al., 2020). Since both tryptophan and BCAAs are essential amino acids and cannot be produced de novo, a diet rich in tryptophan may only improve sleep when BCAAs are low. While competition with tryptophan transport is one mechanism in which BCAAs may influence sleep, other mechanisms may also exist. For instance, BCAAs are involved in the synthesis of de novo glutamate and GABA, neurotransmitters also known to influence sleep (Falup-Pecurariu et al., 2021;Holeček, 2018).
Few studies have examined the role of BCAAs in sleep regulation and maintenance. Studies on BCAA supplements typically focus on muscle mass development, fatigue and performance (Fedewa et al., 2019;Fouré & Bendahan, 2017). While sleep is often not the focus of these studies, BCAA supplements are thought to reduce central fatigue, delay sleep onset, and impair sleep quality, especially if taken later in the day (Hormoznejad et al., 2019;Ord oñez et al., 2017;Portier et al., 2008). In line with these studies, Xiao et al. (2017) in their survey of 277 Chinese adults, reported BCAAs were significantly associated with delayed sleep timing. Consistent findings were also reported by Gordon-Dseagu et al. (2019) in a study of 106 adults.
While biological pathways suggest elevated BCAA levels in healthy individuals have a negative impact on sleep, clinical studies (paradoxically) suggest BCAA supplements are a viable treatment option for sleep disturbances in patients with traumatic brain injury (Elliott et al., 2018;Elliott et al., 2022;Lim et al., 2013). In these cases, BCAAs are provided as supplements for de novo synthesis of glutamate and GABA, which are known to promote sleep, but are deficient after a traumatic brain injury.
To date, population-based studies that have examined the association between BCAAs and sleep in adults are limited and under-explored in children. Moreover, available studies have relied on self-reported measures of sleep duration and timing with their attendant biases (Gordon-Dseagu et al., 2019;Xiao et al., 2017). In line with contemporary conceptualisations of sleep and methods of assessing sleep, it is important to consider actigraphy-derived measures of sleep alongside self-report measures of sleep quality (Buysse, 2014;Matricciani et al., 2018). Given the preliminary evidence and physiological basis for a potential relationship between BCAAs and sleep, as well as the availability and use of supplements containing BCAAs, there is a need to extend our understanding of the relationship between BCAAs and sleep. Therefore, the aim of this study was to test whether higher BCAAs levels (isoleucine, leucine, and valine) were associated with sleep (actigraphic and self-report measures) in a large population-derived sample of Australian children and their parents.

| METHODS
This study examines data collected as part of the Child Health Check-Point study, a one-off cross-sectional study nested between Waves 6 and 7 of the Longitudinal Study of Australian Children (LSAC). The CheckPoint study was conducted between February 2015 and March 2016 and involved a comprehensive physical health and biomarker assessments of children aged 11-12 years, and one of their parents.
Further details of the CheckPoint study has been provided elsewhere (Edwards, 2014;Sanson & Johnstone, 2004).

| Ethics and consent
The CheckPoint study protocol was approved by The Royal Children's Hospital Melbourne Human Research Ethics Committee (33225D) and Australian Institute of Family Studies Ethics Committee (14-26). The attending parent/caregiver provided written informed consent for themselves and their child to participate in the study.

| Sleep
Sleep was assessed in terms of actigraphy-derived sleep characteristics and self-reported trouble sleeping.

| Actigraphy-derived sleep
GENEActiv monitors (Activinsights), fitted to the participant's nondominant wrist, were used to assess sleep. Participants were asked to wear the monitor continuously for 8 consecutive days. Raw acceleration data, collapsed into 60 s epochs, were processed using Cobra software  to derive three objective characteristics of sleep examined in this study: • Sleep period (the difference between sleep onset and offset), • Sleep midpoint (the midpoint between sleep onset and offset), • Sleep efficiency (the percent of minutes scored as sleep between onset and offset).
Participants were included for analysis if they had at least 4 nights of sleep data recorded, had an average sleep time > 200 min, and at least 1 weekend night (Friday or Saturday night) of sleep data. These criteria were predetermined by the Child Health CheckPoint team to reflect habitual sleep . Further details of sleep data processing in this study have been reported elsewhere Matricciani et al., 2019). All actigraphy-derived sleep variables were computed for each individual day and then averaged using a 5:2 weighting for weeknight (Sunday-Thursday) and weekend (Friday-Saturday).

| Trouble sleeping
Trouble sleeping was assessed via self-report. Participants were asked to report how often they had trouble sleeping over the past month, using a 5-point Likert scale (never, almost never, sometimes, often, always).

| Covariates
Covariates selected for this study included socio-economic position, maturity (age of adults and pubertal stage of children), sex, and body mass index (BMI). These variables were selected as covariates as they have been associated with both sleep and metabolomics profiles, particularly BCAAs (Dollman et al., 2007;Felden et al., 2015;Jarrin et al., 2014;Ohayon et al., 2004;Olds et al., 2010). For children, pubertal stage rather than age was selected as a covariate as the children examined in this study were within a narrow age range (11-12 years) and since complex changes in amino acid metabolism have been observed during puberty (Cominetti et al., 2020). Puberty was assessed using the Puberty Development Scale, a validated self-report questionnaire that consists of five Likert scale questions (Chan et al., 2010;Petersen et al., 1988). A higher Puberty Development Scale score represents more advanced pubertal development. For parents, age was calculated from date of birth and expressed as years. Socio-economic position (SEP) was determined using a standardised scale derived from the LSAC dataset, which reflects household income, education, and occupation (Baker et al., 2017;Blakemore et al., 2006). Higher SEP scores reflect higher socio-economic position. Adult alcohol consumption was determined from Wave 6 LSAC questionnaire data, using a continuous self-report measure of average daily alcohol consumption, calculated from the midpoint of categories related to frequency and quantity of alcohol consumption. BMI (kg/m 2 ) z-score and waist-to-hip ratio (used in sensitivity analysis) was determined using anthropometry measures. For children, BMI z-score were calculated using the Centers for Disease Control CDC reference dataset .

| Branched-chain amino acids
All three branched-chain amino acids (BCAAs) (leucine, isoleucine, and valine) were examined in this study. BCAA concentrations were derived using semi-fasted venous blood samples, taken from consenting children and adults participating in the CheckPoint study. Appropriately trained researchers or phlebotomists collected venous blood samples within the assessment centres. Blood collection included four Becton Dickinson (BD) Vacutainer tubes collected in the order of: 2.7 mL EDTA, 9 mL EDTA, 9 mL serum, 7.5 mL lithium heparin, which were then processed at an on-site processing laboratory (Ellul et al., 2019).
Samples were spun at room temperature for 10 min at 550Â g relative centrifugal force before 0.5 mL aliquots were taken of plasma, serum, buffy coat lymphocytes, whole blood and/or a blood clot (Ellul et al., 2019). Samples were then immediately frozen at À80 C for batch analysis at the Melbourne Children's Bioresource Center (Murdoch Children's Research Institute). Metabolomic profiling was done using the Nightingale NMR metabolomics platform (Helsinki, Finland) using the 2016-version quantification algorithm (Ellul et al., 2019).
A high-throughput experimental setup was used to simultaneously measure a range of metabolites, including (but not limited to) routine lipids, lipoproteins, fatty acids, glycolysis-related metabolites, and amino acids, measured from 0.35 mL serum (Ellul et al., 2019). This process generated 228 serum metabolites including nine amino acids (Ellul et al., 2019). Further detail of blood collection, storage, and metabolomics processing has been reported elsewhere (Ellul et al., 2019).

| Statistical analysis
Data management and analyses were undertaken in R (version 4.1.0). The association between each of the three BCAAs (predictor) and the different sleep characteristics (outcome) were assessed using linear mixed models, with the postal code entered as a random intercept to account for the geographic clustering. Mixed models are a suitable technique as they allow for non-independent correlation structures and account for between-and within-group variance (Harrison et al., 2018). Geographical clustering was an important consideration as data were collected in specific towns and suburbs across Australia where differences might exist in health-related outcomes. To account for potential differences related to geographic location, mixed models were used. Postal code was selected as it is a large spatial unit that correspond to towns and suburbs. The performance package in R was used to confirm linear mixed model assumptions were met (linearity, homogeneity of variance, collinearity, normality of residuals, normality of random effects, posterior predictive check, influential observations) (Lüdecke et al., 2021). Complete case analysis was undertaken in R, version 4.1.0, using the lme4 package (Bates et al., 2015). Two models were undertaken for each of the four sleep characteristics. The first model was adjusted for age (of parents)/puberty stage (of children), sex, SEP, average alcohol consumption (of parents), and fasting time. The second model adjusted sociodemographic characteristics in addition to BMI z-score. Analyses were undertaken for children and adults separately. Sensitivity analysis was performed with waistto-hip ratio instead of BMI z-score to verify the robustness of results.

| Participant characteristics
Of the 1874 children and their parents who took part in the Check-Point Study, complete data were available for 741 children and 941 adults. Participants in this study were relatively healthy and few had diabetes (2% of adults and 0.1% of children). Demographic details of the included participants are presented in Table 1. As shown, the mean age of adults was 38.8 years, with the majority being female (87%). The mean age of children was 11.9 (SD 0.4) years, with an approximately equal number of boys and girls. Participants had a higher SEP z-score than the population-based sample of LSAC B cohort participants (mean = 0.32, SD = 0.90 vs. mean = 0.00, SD = 1.00) (Blakemore et al., 2009). Self-report trouble sleeping was poorly correlated with actigraphy-derived sleep parameters in children (r = À0.05 to 0.07) and adults (r = À0.07 to 0.07).
While significant associations were observed, effect sizes were small.
For instance, one standard deviation increase in leucine was associated with a À 0.07 standard deviation decrease in sleep duration and 0.09 standard deviation increase in sleep timing, reflecting a decrease of 3.2 min/day in sleep duration and a 4.1 min/day delay in sleep timing. Associations observed for sleep efficiency were in the direction opposite to expected (higher BCAAs were associated with higher sleep efficiency), however, effect sizes were very small (0.36%). Sensitivity analysis was performed with waist-to-hip ratio instead of BMI z-score to verify the robustness of results. Results relating to the relationship between BCAAs and sleep were not affected and are presented in Appendix A.

| Branched-chain amino acids and sleep of adults
As shown in Table 3, BCAAs were not significantly associated with any of the actigraphy-derived sleep parameters in adults. Isoleucine (β = 0.10; CI = 0.03 to 0.16; p = 0.004) and leucine (β = 0.09; CI = 0.02 to 0.15; p = 0.011) were significantly associated with selfreported trouble sleeping (Model 1), even after adjusting for BMI z-score (isoleucine; β = 0.08; CI = 0.01 to 0.15; p = 0.022; leucine; β = 0.07; CI = 0.00 to 0.14; p = 0.040) (Model 2). Sensitivity analysis was performed with waist-to-hip ratio instead of BMI z-score to verify the robustness of results. Results relating to the relationship between BCAAs and sleep were not affected and are presented in Appendix B.

| DISCUSSION
This study examined the association between BCAAs and different sleep characteristics in a large, population-derived sample of  Although BCAAs were not associated with adult sleep, significant associations were observed for children's sleep duration, timing, and self-reported trouble sleeping, even after adjusting for sex, puberty stage, SEP, and BMI z-score. Biological processes related to growth and development (puberty) may explain why different associations are observed for children (Cominetti et al., 2020;Terasawa, 2005). In particular, puberty has been shown to involve changes in amino acid metabolism, with complex patterns specifically observed for BCAAs (Cominetti et al., 2020). Further, while the mechanisms of puberty onset are complex, laboratory-based studies suggest the GABAergic neuronal system plays an important role, with changes observed in glutamate and gamma-aminobutyric acid (GABA) concentrations, substrates that BCAAs are known to be involved in the synthesis of (Holeček, 2018;Terasawa, 2005). However, this area of research has not been extensively explored. Dietary patterns and parenting styles associated with socioeconomic position may also, in part, explain findings. Whole foods, such as chicken, eggs, salmon, nuts, and brown rice are rich in BCAAs, but are atypically preferred by children. Higher consumption of BCAAs in children may therefore reflect a more health-conscious parenting style, which may also include more after-school activities and stimulating pre-bed activities (such as homework) that may result in later bedtimes and poorer sleep. Reverse causality may also be possible, however, the cross-sectional nature of this study precludes our ability to infer causality.

| Strength and weaknesses
This study examines the association between BCAAs and sleep in a large, population-derived sample of Australian children and their parents. Key strengths include the large samples of both children and adults, as well as a comprehensive range of sleep parameters measured by both actigraphy and self-report. Despite these strengths, there are also a number of limitations that need to be acknowledged.
Firstly, care must be taken when generalising findings as included participants were of slightly higher SEP than the general population with a z-score 0.32 (0.90) vs 0.00 (1.00). Adult participants were all parents, and most (87%) were mothers. Child participants were from a narrow age range (11-12 years). Secondly, since tryptophan was not present in the publicly available CheckPoint dataset used in this study, the relative influence of tryptophan remains unclear. It is also important to note that this study only examined the role of BCAAs. Other large neutral amino acids, such as tyrosine and phenylalanine, may also compete with tryptophan in passing the blood-brain barrier. Thirdly, although analyses were adjusted for fasting time, diet prior to fasting was not measured. Fourthly, this study was a cross-sectional analysis and cannot imply causality. Lastly, although different dimensions of sleep were examined, actigraphy-derived sleep parameters were assessed over a relatively short period of a week and single-item selfreport trouble sleeping has not been validated as a specific dimension of sleep (Matricciani et al., 2022). Maurizio Costabile for reading a draft of this manuscript.

CONFLICT OF INTEREST
The authors have no conflicts to declare.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available  β, standardised coefficient.