Association of physical activity and sleep habits during pregnancy with autistic spectrum disorder in 3-year-old infants.

Background: We hypothesized that maternal lifestyle factors, such as physical activity and sleep habits, may be associated with autism spectrum disorder (ASD) in infants. This study aimed to investigate the association between maternal physical activity and sleep before and during pregnancy with infant ASD diagnosed by the age of 3 years.
Methods: We used the data from the Japan Environment and Children's Study between 2011 and 2014. The study included 103,060 pregnant women, among which, 69,969 women were analyzed. Participants were asked about their physical activity and sleep before and during pregnancy using questionnaires during pregnancy. Maternal physical activity was estimated using the international physical activity questionnaire. Based on the levels of physical activity before or during pregnancy, the participants were divided into five groups. Maternal sleep was analyzed based on sleep duration and bedtime. The outcome was diagnosis of ASD in 3-year-old infants.
Results: In mothers with higher physical activity levels during pregnancy, the risk ratios (RR) for ASD in their 3-year-old infants were lower (RR = 0.61, 95% confidence interval (CI) = 0.42-0.90). In contrast, too short (<6 h) and too long (>10 h) sleep durations during pregnancy were associated with higher risk ratios for ASD than 7-8 h sleep duration (too short: RR = 1.87, 95% CI = 1.21-2.90; too long: RR = 1.56, 95% CI = 1.00-2.48). These associations were not observed before pregnancy.
Conclusion: Maternal physical activity and sleep duration during pregnancy may be associated with ASD in infants.

Introduction

The prevalence of developmental disorders such as autism spectrum disorder (ASD) is increasing in developed countries, including Japan[1-3]. ASD is associated with various prenatal, perinatal, and postnatal risk factors, including genetic and environmental factors[4-6].

Maternal lifestyle appears to be associated with offsprings’ development. For example, higher maternal physical activity (PA) during pregnancy has been reported to be positively associated with infants’ development. Two review articles in 2018 concluded that PA during pregnancy was associated with improved language development[7] and total neurodevelopment of the offspring[8]. Another example of such a maternal lifestyle factor is maternal sleep. Short duration of maternal sleep or late bedtime during pregnancy is associated with preterm birth and gestational diabetes mellitus (GDM)[9,10], which are also reported risk factors for ASD[11].

Maternal inflammation and metabolic disorders are possible causal pathways that link maternal lifestyle with infant ASD. Maternal inflammation has been shown to be associated with maternal PA and sleep[12-15], and uterine inflammation is considered to be a cause of developmental disorders[16,17]. Similarly, maternal metabolic disorders, such as obesity and dyslipidemia, affect maternal PA and sleep[15,18], and may be associated with infant developmental disorders through hypertensive disorders in pregnancy (HDP) and GDM[19-23].

For the above reasons, we hypothesized that maternal lifestyle factors, such as PA and sleep habits, may be associated with ASD in their children. However, to the best of our knowledge, these associations have not been reported in interventional studies, such as randomized controlled trials, or in observational studies. In addition, the influence of preconception maternal lifestyle on offspring development has not been reported. We previously reported that higher levels of maternal PA both before and during pregnancy are associated with a decrease in sleeping and developmental problems in 1-year-old infants[24]. Moreover, maternal sleep habits, such as short sleep duration and late bedtime, both before and during pregnancy, are associated with an increase in sleeping problems in 1-year-old infants[25]. Children with ASD frequently have sleeping problems, such as late bedtime and intense night crying, and developmental abnormalities from early infancy[26-28]. We hypothesized that maternal PA and sleep habits before and during pregnancy are associated not only with early infancy sleep and developmental problems but also with the subsequent diagnosis of ASD.

In this observational study, we aimed to investigate the association between maternal PA and sleep habits before and during pregnancy and ASD diagnosis in their 3-year-old toddlers. The result of this study showed that higher maternal PA levels during pregnancy may be associated with a lower risk of ASD in 3-year-old infants, whereas too short or too long sleep durations during pregnancy may be associated with a higher risk of ASD in 3-year-old infants. Our findings suggested that maternal lifestyle during pregnancy may be associated with risk of ASD in their children.

Methods

Research ethics

The Japan Environment and Children’s Study (JECS) protocol was reviewed and approved by the Ministry of the Environment’s Institutional Review Board on Epidemiological Studies (no. 100910001) and by 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. This study was conducted in accordance with the Declaration of Helsinki. Written informed consent, which was also obtained for a follow-up study of the children after birth, was obtained from all recruited pregnant women.

Study participants

Data used in this study were obtained from the JECS, an ongoing large-scale cohort study. The JECS was designed to follow children from the prenatal period to the age of 13 years. The participants were recruited between January 2011 and March 2014 from 15 regional centers throughout Japan, and the follow-up was mainly conducted via a self-administered questionnaire. The detailed protocol of the study and baseline profiles of the JECS participants have been reported previously[29,30]. The participants answered a questionnaire about lifestyle and behavior twice during pregnancy. The questionnaires completed at recruitment (early pregnancy) and later during mid- or late pregnancy were referred to as M-T1 and M-T2, respectively. The husbands of half of the participants also answered a self-administered questionnaire (F-T1) between the mothers’ early pregnancy and 1 month after delivery. The F-T1 covered information such as lifestyle and medical history. The mean gestational weeks (standard deviation, SD) at the time of responding to M-T1 and M-T2 were 16.4 (8.0) and 27.9 (6.5) weeks, respectively. Participants also answered a questionnaire about their offspring 3 years after delivery (C-3y).

Exposure 1: Maternal PA before and during pregnancy

The Japanese short version of the international physical activity questionnaire (IPAQ), for which test-retest reliability and criterion validity were reported elsewhere, was used to evaluate maternal PA[31,32]. Participants reported their duration and frequency of PA lasting ≥10 min, divided by intensity into mild, moderate, and high. Based on that declaration, we estimated their mean PA per week before and during pregnancy in the M-T1 (based on recall) and M-T2 questionnaires, respectively. We calculated PA in terms of the metabolic equivalent of a task (MET), measured as the number of minutes per week (METs-min/week)[31]. PA, as defined in the IPAQ, includes all activities of daily life, such as work, housework, and leisure activities.

We divided the participants into five groups based on their prepregnancy PA levels. We also divided the participants into five groups based on their levels of PA during pregnancy. In each of the five groups, the PA = 0 group consisted of participants whose PA was 0. The other participants were divided into four groups using PA quartile points. The groups were labeled Quartiles 1–4 in ascending order of PA. Quartile 1 referred to the group with the lowest PA levels among the four groups, and Quartile 4 referred to the group with the highest PA levels. Further, the participants only reported PA that lasted ≥10 min; this implies that the PA = 0 group might not have reflected the actual PA. In addition, a general recommended amount of PA during pregnancy has not been determined. Therefore, in this study Quartile 1, which had the least amount of PA, was used as the control group instead of the PA = 0 group.

The abovementioned categorization of maternal PA was performed in the same way as in our previous study[24].

Exposure 2: Maternal sleep before and during pregnancy

In the M-T1 questionnaire, participants were asked about their awakening time and bedtime before pregnancy. We calculated the sleep duration of participants and divided the participants into six groups according to sleep time: <6 h, 6–7 h, 7–8 h (reference), 8–9 h, 9–10 h, and >10 h. Participants were also divided into groups according to bedtime: 21:00 to 00:00 (reference), 24:00 to 03:00, and others (sleep before 21:00 or after 03:00). Since the bedtime for more than 95% of the analyzed participants was between 21:00 and 03:00 and the mode of bedtime was between 22:00 and 24:00, we further divided the participants into groups according to bedtime: 21:00, 24:00, and 03:00.

In the M-T2 questionnaire, participants were also asked about their usual awakening time and bedtime in the previous month. The participants were divided into groups as described above for M-T1.

The abovementioned categorization of maternal sleep was performed in the same way as in our previous research[25].

Outcome: ASD in 3-year-old infants

Three years after delivery, information about the infant was collected via the parent-reported questionnaire (C-3y). From among a list of diseases, the participants were asked to select those that their children had been diagnosed with by the age of 3 years. The list of diseases included ASD, which was expressed as “autistic spectrum disorder (e.g., autism, pervasive developmental disorder, Asperger’s syndrome).” We defined outcome based on whether the participants selected ASD.

Covariates

Information on maternal age at delivery, prepregnancy body mass index (BMI), parity, gestational age at birth, infertility treatment, type of delivery, current history of HDP and diabetes/GDM, uterine infection, small for gestational age, and infant sex were collected from medical records transcripts. Parental information regarding smoking habits, alcohol consumption, educational background, history of psychiatric disorders (depression, anxiety disorders, and schizophrenia), autistic traits, and feeding status was collected via self-administered questionnaires.

The parents’ autistic traits were evaluated by a short form of the Japanese version of the autism spectrum quotient (AQ-J-10)[33]. In concordance with previous studies, participants with a score of seven or more were categorized as having autistic traits.

Statistics and reproducibility

Between January 2011 and March 2014, 103,060 pregnancies were included in the study (Fig. 1). We limited the subject of this study to term-born singletons, since gestational age at birth and multiple fetuses are considered risk factors for autism[5,11]. The JECS study also includes several congenital anomalies. Since congenital abnormalities are considered to have a variable influence on infant development, cases with congenital anomalies were also excluded. We excluded 33,091 cases for the following reasons: previous participation in the study (n = 5647), multiple fetuses (n = 948), miscarriage or stillbirth (n = 3520), congenital anomaly or disease at 1 month of age (n = 3550), missing information on maternal age at delivery (n = 11), delivery before 37 weeks or after 42 weeks of gestation (n = 4454), and no response to all questions about maternal PA and sleep habits in the M-T1 and M-T2 (n = 751). Among the remaining 84,719 participants, 14,210 participants (17%) did not respond to the questionnaire regarding their 3-year-old offspring. Finally, the remaining 69,969 participants were included in the analysis. Among these, data of 37,145 fathers were available.

Fig. 1

Study population flowchart.

PA physical activity, M-T1 questionnaire administered at recruitment, M-T2 questionnaire administered during mid- or late pregnancy.

We used a log-binominal regression model to explore the association of each exposure with each outcome and to estimate the risk ratio (RR) of each outcome and the 95% confidence intervals (CIs). We initially adjusted for maternal age at delivery and then further adjusted for other maternal and child factors: smoking habits (never smokers, ex-smokers who quit before pregnancy, or smokers during early pregnancy), alcohol consumption (never drinkers, ex-drinkers who quit before pregnancy, or drinkers during early pregnancy), prepregnancy BMI (<18.5, 18.5–24.9, or ≥25.0 kg/m2), parity (0 or ≥1), infertility treatment (no ovulation stimulation/artificial insemination with husband’s semen or assisted reproductive technology), maternal educational background (<10, 10–12, 13–16, or ≥17 years), maternal history of depression (yes or no), anxiety disorders (yes or no), and schizophrenia (yes or no), maternal autistic traits (yes or no), uterine infection (yes or no), type of delivery (vaginal or cesarean section), gestational age at birth (37, 38, 39, 40, or 41 weeks), small for gestational age (yes or no), infant sex (boy or girl), and feeding (breast milk, formula, or both). The covariates to be added to the multivariate model were determined by referring to previous literature on potential risk factors for ASD[5,11,34,35]. Furthermore, we did not complete the missing data. Thus, the multivariate analysis was limited to participants who had all the covariate data. However, the proportion of participants excluded from the multivariate analysis due to missing information about covariates was only 1%.

When significant associations were found, we also performed subgroup analyses that included the paternal factors (age, smoking, educational background, history of psychiatric disorders, and autistic traits) in the log-binominal model.

In this study, we used a fixed dataset “jecs-ta-20190930,” which was released in October 2019. Stata version 16 (StataCorp LP, College Station, TX, USA) was used for all statistical analyses.

Table 1

Association between maternal physical activity before or during pregnancy and diagnosis of autism spectrum disorder in infants aged 3 years (parent-reported; Japan Environment and Children’s Study).

	No. of participants	No. of outcomes		Maternal age-adjusted model			Multivariable modela
			%	RR	95% CI		RR	95% CI
Physical activity
Before pregnancy
0	11,816	57	0.5	0.99	0.70	1.40	0.97	0.68	1.38
Q1	14,144	70	0.5	Reference			Reference
Q2	14,254	76	0.5	1.08	0.78	1.49	1.12	0.80	1.55
Q3	14,096	58	0.4	0.85	0.60	1.20	0.85	0.60	1.21
Q4	14,305	56	0.4	0.85	0.60	1.21	0.81	0.57	1.16
During pregnancy
0	15,210	80	0.5	0.94	0.68	1.29	0.91	0.66	1.26
Q1	12,796	71	0.6	Reference			Reference
Q2	12,281	48	0.4	0.71	0.49	1.02	0.69	0.48	0.99
Q3	13,382	67	0.5	0.92	0.66	1.28	0.87	0.62	1.21
Q4	12,839	43	0.3	0.63	0.43	0.92	0.61	0.42	0.90

CI confidence interval, RR risk ratio.

aAdjusted for maternal age at delivery, smoking habits, alcohol consumption, prepregnancy body mass index, parity, infertility treatment, maternal educational background, history of depression, anxiety disorders and schizophrenia, autistic traits, uterine infection, type of delivery, gestational age at birth, small for gestational age, infant sex, and feeding status.

Table 2

Association between maternal sleep before or during pregnancy and diagnosis of autism spectrum disorder in infants aged 3 years (parent-reported; Japan Environment and Children’s Study).

			No. of participants	No. of outcomes			Maternal age-adjusted model			Multivariable modela
					%	RR		95% CI		RR	95% CI
Before pregnancy	Sleep duration (hours)	<6	4590	29	0.6	1.38		0.92	2.08	1.30	0.86	1.96
		6–7	13,809	73	0.5	1.15		0.85	1.54	1.09	0.80	1.47
		7–8	23,659	108	0.5	Reference				Reference
		8–9	17,124	63	0.4	0.81		0.60	1.11	0.91	0.66	1.25
		9–10	6966	29	0.4	0.94		0.62	1.42	1.13	0.75	1.72
		>10	3018	17	0.6	1.38		0.83	2.31	1.44	0.84	2.41
	Bedtime	21:00–24:00	46,913	286	0.6	Reference				Reference
		24:00–03:00	20,431	133	0.7	1.27		1.01	1.61	1.08	0.85	1.39
		Other	1822	15	0.8	1.73		0.97	3.10	1.41	0.75	2.50
During pregnancy	Sleep duration (hours)	<6	3267	26	0.8	2.04		1.32	3.16	1.87	1.21	2.90
		6–7	10,490	53	0.5	1.28		0.91	1.81	1.20	0.85	1.70
		7–8	21,737	85	0.4	Reference				Reference
		8–9	19,739	83	0.4	1.08		0.80	1.46	1.18	0.87	1.60
		9–10	9801	47	0.5	1.26		0.88	1.80	1.46	1.03	2.11
		>10	4354	24	0.6	1.57		1.00	2.47	1.56	1.00	2.48
	Bedtime	21:00–24:00	51,213	216	0.4	Reference				Reference
		21:00–24:00	16,676	98	0.6	1.44		1.13	1.83	1.21	0.94	1.55
		Other	1499	4	0.3	0.67		0.25	1.81	0.64	0.24	1.71

CI confidence interval, RR risk ratio.

aAdjusted for maternal age at delivery, smoking habits, alcohol consumption, prepregnancy body mass index, parity, infertility treatment, maternal educational background, history of depression, anxiety disorders and schizophrenia, autistic traits, uterine infection, type of delivery, gestational age at birth, small for gestational age, infant sex, and feeding status.