Prospective Association of Air-Purifier Usage during Pregnancy with Infant Neurodevelopment: A Nationwide Longitudinal Study—Japan Environment and Children’s Study (JECS)

Fetal exposure to particulate matter (PM) is associated with infant developmental delay likely via neuroinflammation and prefrontal cortex lesions; however, whether air-purifier usage, which can reduce indoor PM levels, is related to infant developmental delay remains unknown. We therefore examined the prospective relationship between air-purifier usage during pregnancy and infant developmental delay by analyzing 82,441 mother–infant pairs using a simple yes/no questionnaire. Developmental delays at 6 and 12 months were assessed in five areas using the Ages and Stages Questionnaire, Third Edition. A generalized linear mixed model analysis was used to derive adjusted odds ratios (AORs) and 95% confidence intervals (95% CIs) while controlling for 20 covariates. The analysis revealed that air-purifier usage was associated with developmental delays in fine motor (AOR: 0.91, 95% CI: 0.83–0.99) and problem solving (AOR: 0.83, 95% CI: 0.77–0.90) at 6 months and in communication (AOR: 0.86, 95% CI: 0.79–0.93), fine motor (AOR: 0.87, 95% CI: 0.82–0.92), problem solving (AOR: 0.83, 95% CI: 0.77–0.88), and personal–social (AOR: 0.79, 95% CI: 0.72–0.86) at 12 months. In conclusion, a negative association exists between air-purifier usage during pregnancy and infant neurodevelopmental delay that strengthens with time. Our results outline the potential role of air purifiers in inhibiting infant neurodevelopmental delay.


Introduction
Pollution is the leading environmental cause of disabilities and early death [1]. Among the different types of pollution, air pollution accounts for the most disability-adjusted life years (DALYs) in the first stage of human life [2]. As such, it is a threat to children's lives [3]. The causative agents of air pollution include particulate matter (PM), such as organic or elemental carbon, polycyclic aromatic hydrocarbons (PAHs), and metals, as well as carbon monoxide, ozone, nitrogen dioxide, and sulfur dioxide [4,5]. Indoor air pollution is considered more harmful than outdoor pollution, as people tend to spend more time indoors [6].

Study Design and Participants
The detailed design and baseline characteristics of the JECS have been published elsewhere [17][18][19]. Briefly, the JECS is a nationwide, government-funded birth cohort study of various environmental factors and children's health and development. In this study, 103,062 pregnancies, typically during the first trimester, were registered via face-to-face recruitment at public agencies across 19 regions, including both rural and urban locations throughout Japan (from the subpolar northern island, Hokkaido, to the southern island, Okinawa), between January 2011 and March 2014. The sample size was determined in advance to have adequate statistical power to examine conditions with ≤1% prevalence. The eligibility criteria for participants (expectant mothers) were as follows: (1) residence in the study areas at the time of recruitment and expected to reside continuously in Japan for the foreseeable future; (2) expected delivery date between 1 August 2011, and mid-2014; and (3) ability to comprehend the Japanese language and complete the self-administered questionnaire. Those residing outside the study areas, even if they visited the cooperating health-care providers within the study areas, were excluded from the study.
The JECS protocol was approved by the Ministry of the Environment Institutional Review Board for Epidemiological Studies (100910001) and by the Ethics Committees of all participating institutions. This study was also approved by the Ethics Committee, University of Toyama (R2018023) on 12 June 2018. The JECS was conducted in accordance with the Declaration of Helsinki and other nationally valid regulations, and written informed consent was obtained from all participants.
Follow-ups were conducted primarily via mailed letters and scheduled in-hospital check-ups until one month postpartum and via mailed letters at 6 and 12 months. Data were acquired using self-administered questionnaires or transcriptions of medical records recorded by physicians, midwives/nurses, and/or research coordinators. The dataset used in this study, jecs-an-20180131, was released in March 2018 and updated in December 2018. The dataset includes measurements for the first and second/third trimesters as well as for 0, 1, 6, and 12 months postpartum.
Among the 103,062 participants in this dataset, 5647 were excluded due to multiple registrations, 949 due to multiple births, and 3676 following miscarriages or stillbirths. Among the remaining 92,790 unique pregnancies with singleton live births, 1693 were further excluded due to no response to the questionnaire requesting information on demographics, psycho-social status, residence type, and dwelling environment characteristics, including the usage of air purifiers, during pregnancy. A further 8656 pregnancies were further excluded due to a lack of responses to the posted Ages & Stages Questionnaire, Third Edition (ASQ-3) [20], at both 6 and 12 months postpartum. Thus, data from 82,441 mother-child pairs were analyzed ( Figure 1). J. Clin. Med. 2020, 9,1924 3 of 14 The JECS protocol was approved by the Ministry of the Environment Institutional Review Board for Epidemiological Studies (100910001) and by the Ethics Committees of all participating institutions. This study was also approved by the Ethics Committee, University of Toyama (R2018023) on 12 June 2018. The JECS was conducted in accordance with the Declaration of Helsinki and other nationally valid regulations, and written informed consent was obtained from all participants.
Follow-ups were conducted primarily via mailed letters and scheduled in-hospital check-ups until one month postpartum and via mailed letters at 6 and 12 months. Data were acquired using self-administered questionnaires or transcriptions of medical records recorded by physicians, midwives/nurses, and/or research coordinators. The dataset used in this study, jecs-an-20180131, was released in March 2018 and updated in December 2018. The dataset includes measurements for the first and second/third trimesters as well as for 0, 1, 6, and 12 months postpartum.
Among the 103,062 participants in this dataset, 5647 were excluded due to multiple registrations, 949 due to multiple births, and 3676 following miscarriages or stillbirths. Among the remaining 92,790 unique pregnancies with singleton live births, 1693 were further excluded due to no response to the questionnaire requesting information on demographics, psycho-social status, residence type, and dwelling environment characteristics, including the usage of air purifiers, during pregnancy. A further 8656 pregnancies were further excluded due to a lack of responses to the posted Ages & Stages Questionnaire, Third Edition (ASQ-3) [20], at both 6 and 12 months postpartum. Thus, data from 82,441 mother-child pairs were analyzed ( Figure 1).

Exposure
The use of air purifiers during the previous year was assessed using simple yes/no questions. The data were collected during the second/third trimester.

Exposure
The use of air purifiers during the previous year was assessed using simple yes/no questions. The data were collected during the second/third trimester.

Outcomes
Children's neurodevelopment at 6 and 12 months after birth was assessed using the ASQ-3, an age-specific, structured, parent-completed child monitoring system [20]. The ASQ-3 is a set of well-validated questionnaires that has been recommended by the United Nations Children's Fund to verify whether children have normal neurological development [21]. The ASQ-3 assesses the following five areas of development: (1) communication: language skills, such as babbling, vocalizing, listening, and understanding; (2) gross motor: arm, body, and leg movements to move and play; (3) fine motor: hand and finger movements; (4) problem solving: problem-solving skills, learning, and playing with toys; and (5) personal-social: self-help skills, solitary social play, and play with toys and others. Each area consists of six developmental items. Each parental response-yes, sometimes, and not yet-yields 10, 5, and 0 points, respectively, and the total score for each area ranges from 0 to 60 points. Screen-positive cases for each area are defined as those with scores on or below the respective threshold values, which are set at −2 standard deviations from the mean (i.e., z-score ≤ −2) to yield a sensitivity of 85-92%, a specificity of 78-92%, and a positive predictive value of 32-64% [20]. Taking early delivery into account, if a parent's completion date at 6 and 12 months was not within ±1 month from the estimated delivery date, the data were treated as missing values in accordance with the scoring guidelines.

Statistical Analysis
Separate generalized linear mixed models, with 19 regions set as a random effect and logit as a link function, were used to calculate odds ratios (ORs) and their 95% confidence intervals (CIs). Nonuse of air purifiers was used as the reference. Outcome variables were screen-positive cases in the five developmental areas: (a) communication; (b) general motor; (c) fine motor; (d) problem solving; and (e) social-personal (in the ASQ-3); as well as (f) the total number of screen-positive cases over the five areas at 6 and 12 months. The exposure variable was the yes/no answer to the question regarding the use of air purifiers.
The forced entry method was used to include covariates in the multivariate analysis. In the crude model, only the region was set as a random effect. In model 1, the regression models were adjusted for maternal age, parity, smoking status, passive smoking status, folic acid intake, marital status, highest educational level, annual household income, type of residence, number of rooms in the house/apartment, materials covering the floor of the living room, and age of house/apartment building, with the region set as a random effect. In model 2, the regression models were adjusted for the covariates used in model 1: BMI, alcohol intake, number of hours spent outdoors, physical activity, employment status, high-rise living, house renovation/interior completion after becoming pregnant, and number of years living in the current place of residence. SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and R 4.0.0 were used for all statistical analyses.

Missing Data
The effective response rates at 6 and 12 months postpartum were 88.01% (n = 81,669) and 85.60% (n = 79,427), respectively, while 0.83% (n = 772) of the mothers did not respond at 6 months but responded at 12 months. Of the 82,441 mother-infant pairs included in the study, the missing-data rate was ≤1% for all covariates except for parity (2.51%; n = 2073), number of years living in the current place of residence (3.13%; n = 2580), physical activity (3.33%; n = 2749), average number of hours spent outdoors (4.04%; n = 3329), and annual household income (6.92%; n = 5704). For exposure measures, the item regarding the use of air purifiers had a missing-data rate of <0.37% (n = 298). Each score of the five areas of the ASQ-3 had a <5.99% (max n = 4931) missing-data rate at 6 months and <10.57% (max n = 8712) at 12 months. A total of 18,985 (23.03%) and 21,908 (26.57%) mother-infant pairs had at least one missing value at 6 and 12 months, respectively.
We conducted imputation using chained equations [26] to obtain 24 imputed datasets. All data were imputed simultaneously, regardless of the measured time points. When conducting multiple imputations, auxiliary variables that were likely associated with the analyzed variables were also included to avoid violating the missing-at-random assumption. The estimates from each dataset after multiple imputation were combined using Rubin's rule [27].

Sensitivity Analysis
The resulting OR patterns from the complete case datasets (n = 60,660-63,790) were compared with those from the multiply imputed datasets (n = 82,441 each) to assess the differences between the strategies for addressing missing values.
Instead of using the cases, the raw scores for each area (ranging from 0 to 60; the higher, the better) were analyzed using the same setting except for using a Gaussian distribution.

Backgrounds
The characteristics of mothers and dwelling environments together with data regarding the use of air purifiers during pregnancy are presented in Table 1. Higher usage of air purifiers was generally associated with high education levels, high income, less smoking, married status, parity, and living in a newer house.

Main Results
Multicollinearity was not detected among all covariates; that is, all generalized variance inflation factors were below 2.73.
The prevalence, cases, crude ORs, and adjusted ORs (models 1 and 2) for developmental delay in each developmental area at 6 and 12 months according to the use or nonuse of air purifiers are presented in Table 2.  The mean and SD values of the total numbers of screen-positive cases over the five developmental areas and their crude ORs and adjusted ORs (models 1 and 2) according to the use or nonuse of air purifiers are presented in Table 3. Table 3. Total numbers of screen-positive cases over the five developmental areas in the ASQ-3 and their ORs according to the use or nonuse of air purifiers. . Crude = crude model only, with 19 regions set as a random effect. Model 1 = partial model adjusted for maternal age, parity, smoking status, passive smoking status, folic acid intake, marital status, highest educational level, annual household income, type of residence, number of rooms in the house/apartment, materials covering the floor of the living room, and age of house/apartment building, with the region set as a random effect. Model 2 = full model adjusted for all the covariates in model 1: body mass index, alcohol intake, number of hours spent outdoors, physical activity, employment status, high-rise living, house renovation/interior completion after becoming pregnant, and number of years living in the current place of residence. Figure 2 shows the summary of the adjusted ORs (model 2) and their 95% Cis, which are presented in Tables 2 and 3. and age of house/apartment building, with the region set as a random effect. Model 2 = full model adjusted for all the covariates in model 1: body mass index, alcohol intake, number of hours spent outdoors, physical activity, employment status, high-rise living, house renovation/interior completion after becoming pregnant, and number of years living in the current place of residence. Figure 2 shows the summary of the adjusted ORs (model 2) and their 95% Cis, which are presented in Tables 2 and 3. Figure 2. Adjusted ORs (model 2) and 95% confidence intervals (bars) of air-purifier usage for screen-positive cases of developmental delay assessed using the ASQ-3 with air-purifier nonuse set as a reference. This graph was drawn using the data shown in Tables 2 and 3. OR, odds ratio; ASQ-3, Ages & Stages Questionnaire, Third Edition; Total number, total number of screen-positive cases over five developmental areas in the ASQ-3. Model 2 = full model adjusted for maternal age, body mass index, parity, smoking status, passive smoking status, alcohol intake, number of hours spent outdoors, physical activity, folic acid intake, marital status, highest educational level, employment status, annual household income, type of residence, high-rise living, number of rooms in the house/apartment, materials covering the floor of the living room, age of house/apartment building, house renovation/interior completion after becoming pregnant, and number of years living in the current place of residence, with 19 regions set as a random effect.

Results of Sensitivity Analysis
The results of the sensitivity analysis using the complete case dataset ( Figure S1) were not meaningfully different from those calculated using the multiply imputed dataset.
The analysis using continuous scores of each developmental area in the ASQ-3 is summarized in Table S1.

Discussion
To our knowledge, this is the first study to examine the prospective relationships between the use of air purifiers and child neurodevelopment using a nationwide birth cohort study dataset (n > 80,000) [17][18][19] while controlling for up to 20 potential confounders. Analysis revealed that adjusted ORs (models 1 and 2) of air-purifier usage for screen-positive cases of developmental delay in the ASQ-3 were significantly lower than the reference (i.e., air-purifier nonuse) in the areas of fine motor and problem solving at 6 months and in communication, fine motor, problem solving, and personal-social at 12 months. In addition, adjusted ORs (models 1 and 2) at 12 months were lower than those at 6 months. Therefore, we found a negative relationship between the use of air purifiers during pregnancy and cases of neurodevelopmental delay that worsened over time until the child reached at least one year of age.
Although the usage of air purifiers was associated with the areas of communication, fine motor, problem solving, and personal-social, it is not clear which brain areas correspond to each developmental area in the ASQ-3. That said, the prefrontal cortex would be deeply involved in these areas because the prefrontal cortex is known to be responsible for higher intellectual functions, such as goal-driven behavior, including action planning and motor control, executive functioning, working memory, and adequate social response [28,29]. These functions are involved in the problem solving, communication, fine motor, and personal-social areas of the ASQ-3. This overlap suggests that air-purifier usage might be promote intact development in the prefrontal cortex. Further studies using magnetic resonance imaging and other developmental tests [7] will be interesting.
Given the underlying mechanism by which the usage of air purifiers prevents impaired neurodevelopment in infants via elimination of airborne PM [4,[7][8][9]15], other methods that reduce PM would exert a similar effect in principle. Examples include wearing a mask, closing windows to shut out polluted air [30], ventilating a room [15,31], and temporarily residing in places with clean air. However, it should be noted that this working hypothesis regarding PM is one possibility among others. In fact, the factors associated with reduced risks for developmental delay through air-purifier usage are unknown. For example, other air pollutants, such as PAHs, can also mediate this relationship because prefrontal damage may be caused by PAH [8,32,33], whereas air purifiers can eliminate both vapor-and PM-phase PAHs [34]. Alternatively, developmental delay might be caused by something with no relation to air pollutants. Even in large-scale observational studies such as this one, it is virtually impossible to control all confounders and to come to a causal conclusion. Therefore, the observed relationships do not necessarily guarantee that the introduction of air purifiers will prevent infants from developing developmental delay. To examine the net effect of air-purifier usage on the prevention of developmental delay, it is necessary to conduct randomized controlled trials.
The crude ORs hardly changed even when 12 covariates were used (model 1) and all 20 covariates were used (model 2). The stability suggests that all covariates served only as very weak confounders. Thus, it would be very difficult to explain the observed association in terms of these variables, although air-purifier usage was clearly related to many variables, such as high education levels, high income, less smoking, being married, multipara, and living in a newer house (Table 1). Interestingly, these stable ORs were more salient at 12 months than at 6 months. This tendency implies that the common backgrounds, which could not be explained by the 20 covariates used, might emerge over time. Unfortunately, the JECS birth cohort data are currently available only until the first year of age, but data at later time points will be available in a few years. When the time is ripe, it will be fruitful to examine whether the differences between the use and nonuse of air purifiers increase or decrease over time.
Several limitations of this study should be addressed. First, we evaluated the usage of air purifiers using simple yes/no questions. Thus, important information regarding the performance of the devices, the types of air filters used, and their usage patterns (when, where, and how often the devices were used) was not available. In addition, we did not have data on the actual decrease in indoor PM levels after use of the air purifiers or on the dose-response relationship. Accordingly, the current estimation of ORs was rough. Second, the relationship between maternal PM2.5 exposure and subsequent developmental delay has not yet been reported for this population, so there is a gap in our hypothesis that air-purifier usage might prevent infant neurodevelopmental delay. We cannot deny the possibility that the observed relationship was a spurious association with the true cause existing outside the current framework. Third, although the ASQ-3 is a validated, parent-completed questionnaire, its aim is screening, not diagnosis. Thus, the developmental delay in this study could be overestimated. Fourth, our participants were recruited within the last 10 years from several areas across Japan, including both urban and rural areas, but the extent to which the results are applicable to specific populations cannot be determined. Finally, because our study used a prospective birth cohort, the observed prospective relationship cannot be regarded as a causal relationship.
Despite these limitations, we found inverse prospective relationships between the use of air purifiers during pregnancy and infant neurodevelopmental delay at 6 and 12 months postpartum that were salient over time in 82,441 mother-infant pairs while controlling for up to 20 potential confounders. Our results outline the potential role of air purifiers in inhibiting infant neurodevelopmental delay. This study provides a rationale for conducting randomized controlled trials to address the causal relationship between the use of air purifiers and infant neurodevelopmental delay and in-depth studies elucidating what factors are related to reduced risk for developmental delay in relation to air-purifier usage.
Supplementary Materials: The following are available online at http://www.mdpi.com/2077-0383/9/6/1924/s1, Figure S1: Adjusted ORs (model 2) and 95% confidence intervals (bars) of air-purifier usage for screen-positive cases of developmental delay assessed using the ASQ-3 with air-purifier nonuse set as a reference, Table S1: Mean and SD values, βs and 95% CIs for the score of each developmental area in the ASQ-3 according to the use or nonuse of air purifiers, SAS code for main analysis. Funding: The JECS is funded by the Ministry of the Environment, Japan. This funding source played no role in the study's design; in the collection, analysis, or interpretation of data; in the writing of the report; or in the decision to submit this paper for publication.