Association Between Screen Time Exposure in Children at 1 Year of Age and Autism Spectrum Disorder at 3 Years of Age

Study Design and Participants

We conducted a large birth cohort study in Japan. The Japan Environment and Children’s Study Group operated in collaboration with 15 regional centers across Japan. Its protocol was reviewed and approved by the Ministry of the Environment’s Institutional Review Board on Epidemiological Studies and the ethics committees of all the participating institutions. Approximately 100 000 pregnant women were recruited to participate in the study, and all of the participants provided written informed consent.¹⁸ The research was conducted in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects established by the Ministry of Education, Culture, Sports, Science and Technology and the Ministry of Health, Labour and Welfare. The recruitment period was from January 2011 to March 2014, and the data were analyzed during December 2020. The data used were derived from the strictly controlled jecs-ta-20190930-qsn data set, which was released in October 2019. The target population was selected as follows: first, the study included 100 304 live births of the 104 062 fetal records. Consequently, 382 stillbirths and 1254 miscarriages were excluded. We also excluded 2122 individuals with missing data once we calculated live births, stillbirths, and miscarriages. Next, we excluded 6449 children with cerebral palsy, a congenital condition, at 1 year of age, which may have influenced screen time. We also excluded 9825 children with missing data once we calculated congenital diseases or cerebral palsy. Finally, 84 030 mother-child dyads were included in the analysis (Figure).

Variables

The main exposure variable was screen time at 1 year of age, which was assessed using a questionnaire. When their child turned 1 year of age, the mothers were asked about the number of hours spent per day they let the child watch TV or DVDs. The responses were collected as variables and categorized as, “none (no screen time),” “less than 1 hour,” “1 hour or more but less than 2 hours,” “2 hours or more but less than 4 hours,” and “4 hours or more.” When the participating child turned 3 years of age, we asked mothers the same question. The outcome variable, ASD at 3 years of age, was assessed using a questionnaire. Specifically, mothers were asked the following question when their child turned 3 years of age: “Have they ever been diagnosed with autism spectrum disorder (eg, autism, pervasive developmental disorder, Asperger’s syndrome) by a doctor from the age of 2 years until now?” An option of 2 responses was provided, namely, with and without ASD, which were labeled as “yes (ASD)” or “no (no ASD),” respectively.

Previous studies have indicated that maternal nurturing attitudes and abuse may be associated with screen time.^19,20,21 Therefore, the following factors served as adjustment variables: scores on the Kessler Psychological Distress Scale (K6) and Bonding Scale when the child was 1 year of age; depression, anxiety disorders, integration disorders, and other mental and neurological illnesses; mother’s age at delivery; and household income. In addition, when examining the association between ASD at 3 years of age and screen time at 1 year of age, we considered that predisposition to ASD may affect the outcome and may be attributable to reverse causality. We used participant scores on each of the 5 Ages and Stages Questionnaire [R] (ASQ-3) items, which served as adjustment variables to screen for ASD (communication, gross motor skills, fine motor skills, problem-solving, and personal-social scores) at 1 year of age. The ASQ-3 is valid and reliable in screening for developmental delays in children from 1 month of age to 5 1/2 of age, with age-appropriate questions. Having any of these 5 items is below the cutoff value, the line that prompts a visit to a specialist. Although this screening may not fully reflect the characteristics of ASD, its diagnostic accuracy is more than 80% and accurately detects ASD in most cases.^22,23 Because ASD at 1 year of age is currently difficult to diagnose, the ASQ-3 was used in this study, and all of its 5 items were used as adjustment variables.

Each adjustment variable is described below. The Japanese version of the K6 scale was used to assess depressive tendencies,²⁴ and the cutoff value was set at 5 points or more (Japanese version).²⁵ A score above the cutoff value of K6 (≥5 points) indicated poor mental health. The Japanese version of the Bonding Scale was used to assess mother-child attachment. The scale consists of 10 questions and can yield a maximum score of 30 points; higher scores indicate more negative feelings toward one’s baby.²⁶ However, no clear cutoff value was assigned for this scale; therefore, it was treated as a continuous variable. As younger mothers are at risk of abuse, a cutoff value of 19 years or younger at the time of delivery was used. Because poverty is also a risk factor for abuse, the poverty line reported by the Ministry of Health, Labour and Welfare (annual household income ≤¥ 1.27 million [$11 060.41 USD) was used as the standard,²⁷ and the lowest income item (<¥ 2 million [$17 417.96 USD]) was used as the cutoff value for responses related to annual household income. The calculation of the poverty line reported by the Ministry of Health, Labour and Welfare is based on the standards of the Organization for Economic Cooperation and Development standards. For the 5 ASQ-3 items, we used the cutoff values specified for the Japanese version.²⁸Additionally, because there are sex differences in ASD,^1,29 sex was used as a stratification variable to examine sex differences in the results.

Statistical Analysis

First, we aggregated the key variables by screen time at 1 year of age and at 3 years of age, and examined their attributes. Thereafter, odds ratios (ORs) and 95% CIs were calculated to examine the association between screen time and ASD. In addition, because sex differences have been reported in the prevalence of ASD, we examined the interaction by sex and then conducted a sex-stratified analysis. In the trend test in the adjusted model, the categorical variable (screen time) was statistically examined as an ordinal variable (continuous variable). Jonckheere-Terpstra tests were conducted to examine the association between screen time at 1 year of age and 3 years of age. The statistical significance level was set at .05 and 2-tailed. SPSS, version 27 (IBM), was used for statistical analysis.

Aggregate Proportions of ASD at 3 Years of Age to Screen Time at 1 Year

A total of 84 030 mother-child dyads were analyzed. In 330 (0.4%) of the 84 030 children included in the analysis, ASD had been diagnosed at 3 years of age (Table 1). Of these, 251 were boys (76.0%) and 79 were girls (24.0%) (eTable 4 in Supplement 1). There were 83 237 responses for screen time at 1 year of age and 74 554 responses for screen time at 3 years of age (Table 1). Irrespective of whether the child had ASD at 3 years of age, at the age 1 year, less than 1 hour was the most commonly provided response for daily screen time (Table 1). The proportion of children with ASD increased as screen time increased (Table 1). The attributes of children and mothers by screen time are shown in Table 1 and eTable 3 in Supplement 1.

Association Between Screen Time at 1 Year and ASD at 3 Years of Age

Logistic regression analysis of the association between screen time at 1 year and ASD at 3 years of age is shown in Table 2. Longer screen time at 1 year of age was associated with the statistically significantly higher odds of ASD at 3 years of age. Furthermore, the longer screen time at 1 year of age was associated with the statistically significantly higher odds of ASD at 3 years of age in boys. The distribution of screen time was similar across the sexes (Table 1). However, among girls, no association between screen time and ASD was found.

Association Between Screen Time at 3 Years and ASD at 3 Years of Age

The results of the trend test between screen time at 1 year of age and at 3 years of age showed that screen time at 1 year of age was statistically significantly associated with screen time at 3 years of age (Table 3). Namely, as the screen time increased, the proportion of children with ASD at 3 years of age also increased (Table 2). Moreover, logistic regression analysis reported that screen time at 3 years of age was not associated with ASD at 3 years of age (Table 3).

Strengths and Limitations

The strength of this study is that it provides new insights into the association of screen time with the onset of ASD in early childhood, which had not yet been examined. Furthermore, the results of this study are reliable because the analysis was conducted using a large birth cohort data set representative of the Japanese population.⁵²

A limitation of this study is that ASD (outcome) and screen time (exposure) were assessed based on parental reports, which may have led to reporting bias. Data on the exact amount of time children are gazing at the screen are not available, and it may differ from what parents report. Medical institutions diagnose ASD based on the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) but only when ASD can be diagnosed with certainty because it is still in a developmental stage when children are 3 years of age. Therefore, mild cases may not be diagnosed as ASD at 3 years of age, when the child is at a developmental stage, and may be observed in subsequent years. As a result, the study may be biased toward children with severe ASD. To account for potential risks, we conducted a multivariable analysis with ASD predisposition (ASQ-3) at 1 year of age. Additionally, external (eg, prenatal, living, and childcare environment) and internal factors (eg, genetic factors, diseases, and disabilities) other than screen time, which were examined as exposure factors in this study, were not adequately considered. Therefore, this study conducted a sensitivity analysis of unmeasured confounding as an additional analysis (the E-value),^{53,54,55,56,57} and it was found that the effect of residual confounding remained (eTable 5 in Supplement 1). Further research is needed to examine other factors involved in the association between ASD and screen time, and the combination of factors at a given time of the year associated with increased risk of ASD.