The Association between Prenatal Per- and Polyfluoroalkyl Substances Exposure and Neurobehavioral Problems in Offspring: A Meta-Analysis

Exposure to per- and polyfluoroalkyl substances (PFAS) during pregnancy has been suggested to be associated with neurobehavioral problems in offspring. However, current epidemiological studies on the association between prenatal PFAS exposure and neurobehavioral problems among offspring, especially attention deficit/hyperactivity disorder (ADHD) and autism spectrum disorder (ASD), are inconsistent. Therefore, we aimed to study the relationship between PFAS exposure during pregnancy and ADHD and ASD in offspring based on meta-analyses. Online databases, including PubMed, EMBASE, and Web of Science, were searched comprehensively for eligible studies conducted before July 2021. Eleven studies (up to 8493 participants) were included in this analysis. The pooled results demonstrated that exposure to perfluorooctanoate (PFOA) was positively associated with ADHD in the highest quartile group. Negative associations were observed between perfluorooctane sulfonate (PFOS) and ADHD/ASD, including between perfluorononanoate (PFNA) and ASD. There were no associations found between total PFAS concentration groups and neurobehavioral problems. The trial sequential analyses showed unstable results. Our findings indicated that PFOA and PFOS exposure during pregnancy might be associated with ADHD in offspring and that prenatal PFOS and PFNA exposure might be associated with ASD in offspring. According to the limited evidence obtained for most associations, additional studies are required to validate these findings.


Introduction
Perfluoroalkylated substances (PFASs) are forms of persistent organic pollutants, which have the characteristics of bioaccumulation, long-distance migration, and high toxicity, and they have adverse effects on the environment and human health [1][2][3]. Since the 1950s, PFASs have been widely used in various commercial products, including textile, paper, tableware coatings, food packaging, carpets, anti-fouling agents, and other forms of production and life [4][5][6]. PFAS and their terminal degradation products can be stable in soil and water for over 7 years [7,8]. Currently, PFASs have been detected in various environmental media of human lives, such as the atmosphere, water, and soil [9][10][11], including animal and plant-derived foods [12][13][14].
The National Biomonitoring Program from CDC states that at least one PFAS can be detected in blood samples among the U.S. population [15], and numerous evidence indicates that PFAS can also be detected in pregnant women's bio samples [16][17][18]. Moreover, the developing offspring can be affected through the placenta and breastfeeding [19].

Study Selection
The authors (HJY and YYF) separately screened all studies by titles and abstracts and ascertained the eligible studies by reading the full text. The studies that were eventually included met the following criteria: (1) observational study design, including case-control or cohort studies; (2) exposure to at least one PFAS (e.g., perfluorooctanoic acid (PFOA), perfluorooctane sulphonate (PFOS), perfluorononanoic acid (PFNA), perfluorohexane sulfonic acid (PFHxS), and perfluoroalkane sulfonamido substances); (3) measure PFAS exposure during pregnancy using maternal blood samples (e.g., serum, plasma, or whole blood); (4) the results are either prenatal neurological or child neurological as related problems; and (5) risk estimates are provided, including odds ratios (OR), relative risk (RR), or hazard ratio (HR) and 95% confidence interval (CIs). Laboratory research, nonhuman animal research, letters, and reviews were excluded. When the same data appear in multiple studies, only studies with the largest data sets are included in the analysis. The screening process of document screening is shown in the flowchart.

Definition of Outcomes
Due to the limited amount of related research, only two neurobehavioral diseases, ADHD and ASD, were taken into consideration in our study. The diagnosis of ADHD and ASD in offspring may be determined by medical institution registration, expert opinion, or questionnaire survey.

Data Extraction and Quality Assessment
The following elements were extracted from each eligible study: name of the first author; year of publication; country; neurobehavioral outcome; type of PFAS; the number of pregnant women; age of the pregnant women; biosample type; pregnancy week of sample collection; study time; maximum adjustment of ORs, RRs or HRs and 95% CIs; adjustment of covariates. The Newcastle-Ottawa Quality Assessment Scale (NOS) [53] was used as the guideline for evaluating the included articles. The quality assessment of the study includes selection, comparability, exposure (case-control studies), or results (cohort studies). The highest score of NOS is 9 points, and studies with a score of ≥ 7 can be considered as high-quality studies. The data extraction and quality assessment for each study were conducted independently by YHJ and FYY, and the disagreement was resolved via discussions between XQW, HJY, and YYF.

Statistical Analysis
In this research study, Stata software (version 16) was used for statistical analyses. We used quantitative meta-analysis when there were ≥ 3 qualified studies on the specific association between PFAS exposure in pregnancy and children's neuro-related behaviors that contain identical types of data. A simplified summary was conducted when qualified studies were less than three. Pooled risks and 95% CIs with the fixed-effects model or the random-effects model were assessed as the main statistical method according to the heterogeneity. The heterogeneity of the articles involved was tested with the I 2 and p-value, and p-value < 0.05 was regarded as heterogeneous, with an I 2 statistic < 25% indicating low, 25-50% indicating moderate, and >50% indicating high heterogeneity. A random-effects model was used in cases when I 2 ≥ 50% [54].
When there were more than 3 eligible studies, we conducted a dose-response metaanalysis using linear and non-linear models and selected the significant model with the best goodness-of-fit χ 2 score [55]. When extracting dose data, median doses were selected. The extraction method is shown below. First, in the reference group, the dose was set at zero. Second, when the beginning of the smallest dose group is an open interval, the beginning value is replaced by zero (e.g., 0-10 if < 10). Third, if the end of the largest dose group is an open interval, the width of the previous dose group (e.g., 20−10 = 10) is added to the beginning value of the last group, and the resulting value is set as the end (e.g., 20-30 if > 20). For each dose group, the median was set as the exact dose (e.g., 5 if 0-10) [56]. Since the corresponding PFAS concentrations in each article were different, we recorded the total and quartile concentration of PFAS. For the study that only provides tertile concentrations, we used the tertile percentile as dose concentrations. The pooled odds ratios were calculated for ADHD/ASD associated with per 1 ng/mL increase in PFAS concentrations in continuous analyses and for ADHD/ASD associated with PFAS concentrations by comparing the highest category with the lowest category in quantile analyses [57].
To assess the publication bias in this study, Begg's test and Egger's regression asymmetry test were performed. A funnel chart was used to evaluate the results. Simultaneously, the sensitivity analysis was carried out to observe the stability of the comprehensive results.

Trial Sequential Analysis
The risks of random errors were tested by the trial sequential analysis (TSA) with the Stata package (metacumbounds command), aiming at maintaining an overall 5% risk of a Type I error and a power of 80%. The information size calculation for meta-analyses (cumulated sample size of included trials) was combined with an adapted threshold of statistical significance in the cumulative meta-analysis. With the control event's proportion calculated from the actual meta-analysis in the calculation of the required information size, an intervention effect of a 20% relative risk reduction (RRR) was anticipated [58].

Studies Included in the Meta-Analysis
A total of 1755 articles were retrieved by a literature search of PubMed, Embase, and Web of science. After the removal of 219 duplicates, 1536 studies were further screened for title and abstract. A total of 1424 articles were removed due to systematic reviews, cross-sectional studies, and other unrelated topics. We excluded 101 of 112 articles due to the unavailable concentration of PFASs during maternal pregnancy, lack of quantitative estimates, or duplicate publications. Finally, the remaining 11 articles were included in the present meta-analysis ( Figure 1). Two of the eleven articles contained studies related to both ADHD and ASD, seven articles were about ADHD [42][43][44][46][47][48]50], and six articles were about ASD [29,43,45,[49][50][51]. The NOS score in this meta-analysis ranged from 7 to 9, which is considered a high-quality article (Tables S1 and S2). The characteristics of this article and possible confounding factors are described in Table 1.   Figure 1. Flowchart for the screen of included studies in the current meta-analysis [59].

Association between PFAS and ADHD
With the systematic literature search, seven studies focusesd on pregnancy PFOA (n = 7), PFOS (n = 7), PFNA (n = 3), and PFHxS (n = 3) on ADHD were included. In the total concentration group, the summarized evidence suggests that there is no significant association between prenatal PFAS exposure and ADHD in offspring. Pooled estimates for comparisons of the highest vs. lowest quartile showed that PFOA (OR = 1.21; 95% CI:1.04-1.42; p = 0.016) and PFOS (OR = 0.85; 95% Cl: 0.721-0.996; p = 0.045) were significantly associated with ADHD (Tables 2 and 3; Figures 2 and 3). Dose-response analyses indicated that there is no linear association between pregnancy PFOA and PFOS exposure and ADHD risks in offspring (Figure 4).

Association between PFAS and ASD
We found six studies on PFAS exposure in pregnancy and ASD, including six studies on PFOA, six studies on PFOS, five studies on PFNA, and five studies on PFHxS. Although the pooled estimates in the total concentration group indicated that PFASs were

Association between PFAS and ASD
We found six studies on PFAS exposure in pregnancy and ASD, including six studies on PFOA, six studies on PFOS, five studies on PFNA, and five studies on PFHxS. Although the pooled estimates in the total concentration group indicated that PFASs were

Association between PFAS and ASD
We found six studies on PFAS exposure in pregnancy and ASD, including six studies on PFOA, six studies on PFOS, five studies on PFNA, and five studies on PFHxS. Although the pooled estimates in the total concentration group indicated that PFASs were unlikely associated with ASD, the highest concentration categories of PFOS (OR = 0.79; 95% Cl: 0.64-0.98; p = 0.035) and PFNA (OR = 0.80; 95%Cl: 0.65-0.98; p = 0.032) were negative for ASD compared to the lowest category (Tables 2 and 3; Figures 5 and 6). The dose-response analysis showed that there was no dose-response relationship between PFAS exposure and ASD risk (Figure 7).

Trial Sequential Analysis
The highest quartile arrays of PFOA and PFOS associated with ADHD, including the highest quartile array of PFOS and PFNA associated with ASD, were subjected to trial sequential analyses. Figures S1 and S2 show the results of the trial sequential analysis. All the Z curves did not intersect the TSA boundary (p > 0.05), suggesting that the summary result might be unstable.

Publication Bias
No publication bias was found in this meta-analysis ( Figures S3-S6).

Discussion
In this systematic review and meta-analysis, we systematically gathered the available evidence on the effect of prenatal PFAS exposure with respect to the outcome of neurobehavioral problems in future generations. We included 11 articles in the present study. In the total concentration group, the relationships between PFAS (PFOA, PFOS, PFNA, and PFHxS) and ADHD or ASD could not be identified. However, in the highest quartile group, PFOA is positively associated with ADHD, while PFOS is negatively associated with ADHD and ASD. PFNA is negatively correlated with ASD when compared with the lowest quartile group. There is no publication bias in this study. Unlike the study designed by Joan et al. [60], we directly investigated the effects of maternal PFAS exposure on neurobehavioral problems in offspring, which effectively explored the relationship between PFAS exposure and neurobehavioral problems in offspring.
In the highest quartile group, PFOA exposure increased the risk of ADHD in future generations, which is consistent with the results of previous animal studies [61,62]. Pregnant CD-1 mice exposure to PFOA results in significantly higher activity levels in their male offspring [61], and fertilized zebrafish under PFOA exposure showed morphological changes and hyperactivity [62]. Our study revealed that PFNA exposure potentially

Trial Sequential Analysis
The highest quartile arrays of PFOA and PFOS associated with ADHD, including the highest quartile array of PFOS and PFNA associated with ASD, were subjected to trial sequential analyses. Figures S1 and S2 show the results of the trial sequential analysis. All the Z curves did not intersect the TSA boundary (p > 0.05), suggesting that the summary result might be unstable.

Publication Bias
No publication bias was found in this meta-analysis ( Figures S3-S6).

Discussion
In this systematic review and meta-analysis, we systematically gathered the available evidence on the effect of prenatal PFAS exposure with respect to the outcome of neurobehavioral problems in future generations. We included 11 articles in the present study. In the total concentration group, the relationships between PFAS (PFOA, PFOS, PFNA, and PFHxS) and ADHD or ASD could not be identified. However, in the highest quartile group, PFOA is positively associated with ADHD, while PFOS is negatively associated with ADHD and ASD. PFNA is negatively correlated with ASD when compared with the lowest quartile group. There is no publication bias in this study. Unlike the study designed by Joan et al. [60], we directly investigated the effects of maternal PFAS exposure on neurobehavioral problems in offspring, which effectively explored the relationship between PFAS exposure and neurobehavioral problems in offspring.
In the highest quartile group, PFOA exposure increased the risk of ADHD in future generations, which is consistent with the results of previous animal studies [61,62]. Pregnant CD-1 mice exposure to PFOA results in significantly higher activity levels in their male offspring [61], and fertilized zebrafish under PFOA exposure showed morphological changes and hyperactivity [62]. Our study revealed that PFNA exposure potentially decreases the risk of ASD in the offspring, and PFOS exposure potentially reduces the risk of ADHD and ASD in future generations in the highest quartile group, which is contrary to the results of some animal studies [62][63][64][65][66][67]. Several studies show that exposure to PFOS during pregnancy results in hypermobility in rat offspring [63], and hyperactivity was found in fertilized zebrafish when exposed to PFOA [62,64]. Ghassabian et al. found that PFOS exposure is positively associated with children's behavioral difficulties [65]. The significant positive correlation between PFNA exposure and the total SDQ score was presented in the Hoyer and Luo studies, respectively [66,67]. One plausible reason for the inconsistent results may be that the doses exposed to animals and humans are different. The exposure concentrations used in animal experiments are usually higher than the human routine exposure. Animals exposed to a higher dose may experience other serious side effects such as increased mortality and birth defects, which could make the observed neurotoxicity biased [68,69].
In the total concentration group, there is no significant association between PFAS exposure and ADHD, which is consistent with many of the studies [70][71][72][73][74]. Previous studies of continuous-dose PFAS exposure have also yielded no significant results [70][71][72][73][74]. One of the possible hypotheses is that prenatal PFAS exposure and neurobehavioral problems in offspring might have a non-linear relationship. The study by Christou et al. revealed that zebrafish larval hyperactivity or underactivity was related to PFAS concentrations and the age of the zebrafish [75]. Some studies show that there are gender differences in the relationship between PFAS exposure and neurobehavioral problems [39,72,76]. Moreover, there is an uncertain threshold of PFAS exposure on human health [70,77,78]. Stein et al. found that there was an inverted "J"-shaped association between PFOA exposure and ADHD [38]. Low-dose prenatal PFOS exposure leads to an overactive basal swimming speed, but the offspring continue to be exposed after delivery, and the overactive state gradually disappears [79]. Liew et al. concluded that, assuming uncontrolled confounding and selection bias due to fetal death, the observed association between PFAS exposure and ADHD is either positive or negative [80]. This may explain the inconsistency between our results and existing studies.
There are several advantages in the present study. This is one of the first studies to comprehensively and systematically evaluate the relationship between prenatal exposure to PFAS and neurobehavioral problems in offspring. More than 8000 participants were included in this study, which makes the summary results more reliable. The use of various statistical methods also improved the effectiveness of meta-analysis. However, due to the lack of existing studies, sequential analyses suggested that the summary result might be unstable. In addition, hierarchical analyses cannot be conducted. According to the different study designs and methods, the dose-response analyses cannot be accurately assessed based on the dose-response relationship. Therefore, more in-depth studies are needed.

Conclusions
In summary, PFOA, PFOS, and PFNA exposure during pregnancy were found to be associated with ADHD and ASD in offspring in the highest quartile group. The findings indicated that prenatal PFAS exposure and neurobehavioral problems in offspring are not linear associations, but the research result is unstable. More research should be encouraged to further study its associations and mechanisms of action.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijerph20031668/s1, Table S1. Quality of case-control studies included in meta-analysis according to the Newcastle-Ottawa Scale; Table S2. Quality of cohort studies included in meta-analysis according to the Newcastle-Ottawa Scale; Figure S1.

Data Availability Statement:
No new data were created or analyzed in this study. Data sharing is not applicable to this article.

Conflicts of Interest:
The authors declare no conflict of interest.