Introduction

Autism spectrum disorder (ASD) is a neurodevelopmental condition mainly characterized by social communication deficits as well as restrictive and repetitive sensorimotor behaviors1. ASD is the unified diagnostic name in the diagnostic and statistical manual of mental disorders (DSM-5)2, the diagnostic manual for mental disorders. The prevalence of ASD is 1.09–436 per 10,000 persons (median, 100/10,000 persons) according to a systematic review conducted in 34 countries, which included children- and adult-targeted studies3. ASD is often diagnosed as early as 18–24 months of age and coexists with other developmental disorders, such as attention-deficit/hyperactivity disorder and Fragile X, as well as with temperamental characteristics such as anxiety, and irritability4,5,6. The risk factors for ASD include environmental and genetic factors. Known factors include male sex, older siblings’ ASD status, advanced parental age at birth7,8, maternal antidepressant medication use9, maternal obesity or diabetes10, preterm birth, and low birth weight11.

Early identification and intervention of signs of autism may reduce future autism symptoms12. Adaptive behaviors and expressive and receptive language skills are considered positive consequences of early interventions13. Previous studies have identified the early anatomical, functional, and behavioral signs of ASD. The signs of autism identified before 1 year old involve motor and visuoperceptual functions, such as lack of eye contact, pointing, facial expression, and reaction to loud noises14,15,16,17. Other studies have identified different structures, such as high fractional anisotropy18,19 and increased cerebrospinal fluid20, involved in brain function before 1 year of age among those who later developed ASD. The study of high risk-siblings later diagnosed with autism showed a decline in play and communication and impaired vocal imitation21. The sleep quality and temperament of infants have been suggested as risk factors for ASD. For example, night-awakening at 12 months was associated with autistic traits at 24 months22. In the study following children from 1.5 to 9 years old, children with increasing sleep problem trajectory and children with stable and moderate sleep problems had higher levels of autistic traits23. Diverse temperament features have also been associated with ASD24,25,26. Infants who were later diagnosed with ASD more frequently exhibited crying patterns different from other developmental delay patterns27 or produced pain-related cries with a higher pitch and wider frequency range at 6 months of age28. Additionally, the relationship between sleep and ASD at 1 month of age was suggested in a previous study29, and its relationship may differ according to infants’ sex as sleep duration and temperaments have been suggested to differ between sexes30,31.

To the best of our knowledge, no studies have thoroughly investigated sleep quality and temperament in early infants as predictors of a later ASD diagnosis. Such associations must be ascertained for identifying the mechanism and management of autism. Hence, this study investigates whether the sleep quality and temperament of 1-month-old infants are associated with ASD diagnosis by 3 years old using data from a large-cohort study. We also assess the sex-stratified associations of sleep quality, temperament, and autism. Our findings suggest that sleep quality and temperament are associated with ASD diagnosis.

Methods

Study design and participants

This longitudinal study investigated the association of infants’ sleep quality and temperament at the age of 1 month with an ASD diagnosis by the age of 3 years. We used data (“jecs-ta-201901930” released in October 2019) from a nationwide prospective birth cohort study, the Japan Environment and Children’s Study (JECS), which is registered in the University Hospital Medical Information Network Clinical Trials Registry (number UMIN000030786). The study protocol has been reported elsewhere32,33. Approximately 100,000 pregnant women participated in the JECS, and the infants will be followed up until they turn 13 years old. The recruitment was conducted between January 2011 and March 2014 at 15 Regional Centers. We excluded those with multiple participations, multiple births, miscarriages or stillbirths, congenital anomalies, missing information on maternal age at delivery, birth at <37 or ≥42 gestational weeks, unanswered questions on infant sleep quality or temperament at 1-month-old, and absence in the survey conducted when the children turned 3 years old (Fig. 1). Infants were considered 1 month of age when they were 30 or 31 days old.

Fig. 1: Flow diagram of the selection process of the study participants.
figure 1

Out of 103060 pregnant women registered in the Japan Environment and Children’s Study between January 2011 and March 2014 at Regional Centers, 69751 mothers and infants were selected after excluding those with multiple participations, multiple births, miscarriages or stillbirths, congenital anomalies, missing information on maternal age at delivery, birth at <37 or ≥42 gestational weeks, unanswered questions on infant sleep quality or temperament at 1-month-old, and absence in the survey conducted when the children turned 3 years old.

Full size image

Data collection

The caregivers responded to self-administered questionnaires during the first and second or third trimesters of pregnancy and the first month and third year after birth. In addition, the infants’ medical record transcriptions at birth were collected.

Infant sleep quality assessment

We assessed the infants’ sleep quality at 1 month of age based on the awakening frequency and length of daytime and nighttime sleep34. All sleep quality data were evaluated based on the caregivers’ self-reported answers regarding the infants’ sleep. The caregivers checked the time-frames (from 00:00 to 23:30, with 30-min intervals) on the answer sheet when the infant was asleep on the day before data collection. Then, we considered the frequency of sleep interruption as “awakening” and the duration for which the infant slept as “length of sleep.” We considered ≥5 nocturnal awakenings (between 8:00 p.m. and 7:59 a.m.) as a high awakening frequency, based on the awakening frequency range (1.0–5.0 between 8:00 p.m. and 7:59 a.m.) for 2-week-old neonates35. We compared the lengths of daytime and nighttime sleep based on the sleep duration in the nighttime (8:00 p.m.–7:59 a.m.) and daytime (8:00 a.m.–7:59 p.m.)36. We defined a longer daytime than nighttime sleep period as unusual.

Infant temperament assessment

The behaviors recognized in autistic infants are characterized by unresponsive and irritable/overreactive types37. We focused on the latter type, which appeared to include more sleep problems, whereas the former type involved continuous sleep34. Thus, we assessed the infants’ temperaments based on their mood, crying frequency, and crying intensity34. The caregivers’ self-reported answers were used for assessment. The infants’ mood was identified by the question “frequency of having difficulty while holding the baby due to his/her affect and/or behavior (e.g., crying and bending backward).” The possible answers were “often,” “sometimes,” “seldom,” and “never.” We considered “often” as an indicator of a “bad mood.” The question on crying occurrence was “intensity and frequency of crying (baby),” and the answers were “quite often and long,” “sometimes and short,” or “hardly.” We considered “quite often and long” to mean “frequent crying, for long periods.” The crying intensity was identified by the question, “I have trouble calming down my crying baby.” The answers were “yes” or “no.” We considered “yes” to mean “intense crying.” All these categorizations have been defined in previous studies23,31.

ASD assessment

We identified autistic infants based on the caregivers’ answers on whether their infant was diagnosed with ASD (e.g., autism, pervasive developmental disorder, and Asperger syndrome) by the age of 3 years. This assessment was used because the study’s questionnaire was developed during the transitioning period of DSM -IV38 to DSM-52.

Covariates

The covariates adjusted in the regression models were the infant’s sex, small for gestational age (<10th percentile of birth weight standards by gestational age39), breastfeeding status at 1 month after birth, as well as mother’s age at delivery, smoking habits, alcohol consumption, gestational age at delivery, parity, educational background, household income, and postpartum depressive symptoms at 1 month after delivery (Edinburgh Postnatal Depression Scale score40,41,42 ≥ 9). We selected the covariates adjusted with the previous study related to the association between sleep problems and temperament of 1-month-old infants29.

Statistics and reproducibility

We analyzed all the data using descriptive statistics and then conducted log-binominal regression analyses to estimate the risk ratios (RRs) of ASD concerning sleep quality (≥5 nighttime awakenings and longer daytime than nighttime sleep) and temperament (bad mood, frequent crying for a long period, and intense crying). For regression analyses, we constructed two models. Model 1 was adjusted for infants’ variables, such as infant sex, small for gestational age, and breastfeeding status at 1 month after birth. Model 2 was adjusted for maternal variables, such as maternal age at delivery, smoking habits, alcohol consumption, gestational age at delivery, parity, educational background, household income, and postpartum depressive symptoms. We also conducted an analysis stratifying the participants by sex. All analyses were conducted using STATA version 16.1 (StataCorp LLC, College Station, TX, USA). The dataset used for this study was the jecs-ta-20190930 dataset released in September 2019.

Ethical approval

The JECS protocol was reviewed and approved by the Ministry of the Environment’s Institutional Review Board on Epidemiological Studies (No. 100910001) and the Ethics Committees of all participating institutions: the National Institute for Environmental Studies that leads the JECS, the National Center for Child Health and Development, Hokkaido University, Sapporo Medical University, Asahikawa Medical College, Japanese Red Cross Hokkaido College of Nursing, Tohoku University, Fukushima Medical University, Chiba University, Yokohama City University, University of Yamanashi, Shinshu University, University of Toyama, Nagoya City University, Kyoto University, Doshisha University, Osaka University, Osaka Medical Center and Research Institute for Maternal and Child Health, Hyogo College of Medicine, Tottori University, Kochi University, University of Occupational and Environmental Health, Kyushu University, Kumamoto University, University of Miyazaki, and University of Ryukyu. The study was conducted following the Declaration of Helsinki and Japan’s Ethical Guidelines for Epidemiological Research issued by the Japanese Ministry of Education, Culture, Sports, Science, and Technology and the Ministry of Health, Labor, and Welfare. Written informed consent was obtained from all participants.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Results

Out of the 103,060 registered pregnancies, we analyzed data from 69,751 eligible maternal and infant dyads (Fig. 1). In total, 320 (0.5%) infants were diagnosed with ASD, of whom 247 (0.4%) were male and 73 (0.1%) were female. As presented in Tables 1, 64.2% of mothers were ≥30 years old. The highest percentage of mothers had 12–16 years of schooling (64.7%), and 67.3% had household incomes between 2 and 6 million yen. More than half of mothers had already had children. Those who had ever smoked were 49.4% and had ever consumed alcohol were 65.4%. Those who had postnatal depressive symptoms were 13.7%. Regarding infants, as indicated in Table 2, around half were male. Those who were categorized in a small for gestational age were 7.2% and a majority of infants were born at 38–40 weeks gestation. Of all 4429 infants (6.3%) had ≥5 nighttime awakenings, and 13,141 (18.8%) slept longer during the day than at night. As for temperament, 4381 (6.3%) experienced bad moods; 12,109 (17.4%), frequent crying for long periods; and 13,826 (19.8%), intense crying.

Table 1 Baseline characteristics of mothers according to infant sleep quality and temperament categories in the Japan Environment and Children’s Study (2011–2014).
Full size table
Table 2 Baseline characteristics of infants according to infant sleep quality and temperament categories in the Japan Environment and Children’s Study (2011–2014).
Full size table

Risk ratio of ASD according to infant sleep quality

As presented in Table 3, infants who slept longer in the daytime at 1 month old were at a high risk of ASD by 3 years old after adjusting for infant and maternal factors (RR: 1.33, 95% confidence interval [CI]: 1.01–1.75). Those with ≥5 nighttime awakenings did not significantly differ in ASD risk.

Table 3 Association of infant sleep quality and temperament at 1 month of age with a diagnosis of autism spectrum disorder by 3 years of age.
Full size table

Risk ratio of ASD according to infant temperament

As presented in Table 3, after adjusting for infant factors, it was revealed that infants who experienced bad moods (RR: 1.79, 95% CI: 1.27–2.53) and those who experienced frequent long crying (RR: 1.34, 95% CI: 1.03–1.74) had a higher risk of ASD by 3 years old. Those who experienced intense crying had a higher risk of ASD after adjusting for infant and mother factors (RR: 1.31, 95% CI: 1.00–1.72).

Sex-stratified RR of ASD

Table 4 presents the sex-stratified RR of ASD. The risk of association between a bad mood at 1 month of age and later ASD significantly varied between sex (P < 0.001). Female infants with bad moods had a higher risk of later ASD (RR: 3.59, 95% CI: 1.91–6.75) after adjusting for infant and mother factors. However, there was no significant increase in relative risk for ASD diagnosis associated with the mood in male infants (RR: 0.90, 95% CI: 0.55–1.47). Male infants who experienced intense crying were at increased risk of ASD after adjusting for infant factors (RR: 1.45, 95% CI: 1.09–1.92). Female infants who experienced intense crying were at a higher risk of ASD after adjusting for infant factors (RR: 1.75, 95% CI: 1.05–2.91).

Table 4 Sex-stratified association of infant sleep quality and temperament at 1 month of age with a diagnosis of autism spectrum disorder by 3 years of age.
Full size table

Discussion

To the best of our knowledge, this is the first study to demonstrate an association between sleep quality and temperament at 1 month of age and later ASD. Infants who slept for longer during the day at 1 month old showed a higher risk of developing ASD by age three. Bad moods and frequent, persistent, or intense crying at 1 month of age were also associated with a higher risk of ASD by 3 years old, after adjusting for infant-related factors. A sex-stratified analysis found the sex of the infant not affecting sleep. However, frequent bad moods were an ASD risk factor in female, but not male, infants. We also identified a slightly higher tendency to intense crying in females than males. These associations may be early prognostic indicators of later ASD.

Our study demonstrated that a longer daytime sleep pattern at 1 month of age was associated with a later ASD diagnosis. A similar result was obtained in a previous retrospective study among 1–6-year-old children34. Also, our study demonstrated that the association of autism with the sleep of infants is evident at an earlier age than that reported in a previous study. This association supports the suggestion by Mike et al. that fetal development of circadian chronobiology may be immature or disrupted in infants with sleep problems34. This is thought to result from low maternal melatonin secretion, which has been shown to increase the risk that the fetus will later develop ASD43. Problems with clock-controlled genes have also been reported to disturb sleep cycle rhythms in autistic children44. Some of these genes (e.g., PER1, NPAS2) were found to be associated with ASD development and sleep disturbances in a study of autistic children and parents45. Sleep problems may also lead to the development of an irritable or overreactive temperament in infants29, which, in turn, increases the risk of later ASD34. Our findings supported this relationship between difficult temperament and later ASD. However, Nguyen et al. found no association between infant sleep quality at 12 months, either at night or during the day, with ASD-type behaviors at 24 months, as identified using the modified checklist for autism in toddlers (M-CHAT)22. This differs from our findings. This discrepancy may be because, as stated in the Discussion section of the Nguyen et al. study, M-CHAT scores may not provide sufficient accuracy for a precise determination of ASD risk. Further research is needed to confirm this association.

We found that female infants who exhibited frequent bad moods had a higher risk of later ASD. There are known to be sex-specific differences in infant temperament, including higher reactivity to perceived environmental threats in female than male infants31,46,47. Such sex-specific dispositions might have affected some characteristics of subsequent ASD. Further studies are needed to elucidate sex differences involving temperament and ASD, as they could be important indicators of ASD in female infants.

In our analysis, a difficult temperament at 1 month of age, as evidenced by bad moods and frequent crying was associated with an increased risk of ASD by the age of three. This has been reported in previous studies;48,49 however, these have not studied infants younger than 6 months50. Thus, our findings add to the literature by evidencing risk factors for ASD in 1-month-old infants51. An interesting finding of our study was that there was no significant relationship between infant temperament and later ASD development when the mothers’ covariates were included in the assessed model. As indicated in the descriptive analyses of the participants, mothers who perceived their infant to have a poor temperament were more likely to be primiparous, had more depressive symptoms, and were less likely to be breastfeeding than those who did not. Such factors may have moderated the association between temperament and ASD. For example, primiparous mothers may react more to the duration and intensity of infant crying than multiparous mothers52. Also, maternal factors related to the risk for later development of autism will differ between mothers. For example, previous studies suggested both maternal smoking53 and maternal drinking during54 pregnancy were linked to the later development of autism in the child. However, according to meta-analyses, current evidence does not support any independent association between alcohol consumption and ASD54.

This study, using a large sample size, addressed the difficulty of ASD diagnosis by demonstrating that infant sleep quality and temperament at only 1 month of age may be associated with ASD development. Such behaviors can be observed even by caregivers and can thus enable early intervention. However, this study had some limitations. First, the assessment of sleep was based on caregivers’ reports. Since these are external observations of sleep, their accuracy is not guaranteed. Reports of infant sleep quality might be overestimated if there is no behavioral evidence of poor sleep; whereas, the rate of night awakenings may be underestimated due to brief awakenings in which the infant does not cry and wake their caregiver55. Therefore, the caregivers’ responses may have influenced our study findings. Second, ASD outcome was identified according to caregivers’ self-report on the infants’ ASD diagnosis. Therefore, infants who might be subsequently diagnosed with ASD were not considered, and the actual number might be underestimated as the rate in this study was lower than in another study in Japan56. Third, because the data were obtained from a large-scale study, minor differences may have exhibited significance. Fourth, this study may not have covered all infant temperaments and could be improved in future research.

In conclusion, this study demonstrated that infant sleep patterns and temperament are associated with the risk that the infant will develop ASD before the age of three. Specifically, infants prone to longer daytime sleeping; more frequent bad moods; and intense, frequent, or prolonged crying showed a higher incidence of ASD at 3 years old. In particular, the association between a bad mood and ASD was characteristic of female infants. These findings may serve as important indicators for identifying infants requiring intervention to reduce the risk of future ASD.