Early-Life Exposure to Perfluoroalkyl Substances (PFAS) and Child Language and Communication Development: A Systematic Review

Language development starts during the fetal period when the brain is sensitive to endocrine disruptions from environmental contaminants. This systematic review aims to systematically summarize the existing literature on early-life exposure to PFAS and children’s language and communication development, which is an indicator of neurocognitive development. A structured literature search was conducted using three databases, PubMed, Scopus, and CINAHL, last updated in April 2023. The population was defined as children and young adults. PFAS exposure was assessed pre- or postnatally. The outcome was defined as a language and communication ability assessed with validated instruments, parental self-reports, or clinical language disorder diagnoses. In total, 15 studies were identified for subsequent analyses. Thirteen were performed in background-exposed populations and two in highly exposed populations. There were some indications of potential adverse effects; however, these were not consistent across child sex, age of assessment, or PFAS exposure levels. No systematic effect of early-life PFAS exposure on language and communication development was found. These inconclusive findings may partly be explained by the use of general test instruments with limited validity as to children’s language and communication development. Further studies over a wider exposure range using specific language test instruments are needed.


Introduction
Children's language, cognitive, and motor development follows universal milestones and develops in parallel [1].Language development starts during the fetal period [2][3][4] and continues through preschool age [5] and well into adolescence [6].
Developmental language disorder (DLD) is one of the most common developmental disorders and affects 7-8% of all children [7,8].The aetiology of DLD is multifactorial, involving both genetic and environmental risk factors [9,10].Children with DLD have significant difficulties in one or more areas of spoken or written language, comprehension, and communication, including social interactions.If delayed or disordered language and communication persist at 5 years of age, the condition is considered permanent [9,11], and it is likely to be a risk factor for adverse social and educational performance [12].Language and communication development is an important part of children's general development [7,[13][14][15].Late language development or DLD is subsequently associated with other neurodevelopmental disorders and co-occurs with attention-deficit/hyperactivity disorder (ADHD), autism, and intellectual developmental disabilities [14,16,17].
Many hormones affect the development of the fetal brain [18], and prenatal language development is sensitive to the influence of sex hormones [19][20][21].Early-life exposure to endocrine-disrupting chemicals may, thus, harm the developing brain [22][23][24][25][26]. Adverse effects can occur if exogenous substances pass across the placenta and reach fetal circulation or if they interfere with normal placenta functioning [24].Exposure to endocrine-disrupting environmental contaminants such as polychlorinated biphenyls, lead, methylmercury, polybrominated diphenyl ethers, and organophosphates have been associated with the impaired development of both receptive (i.e., understanding of words and sentences) and expressive language (i.e., decreased mean length of utterances) [27].
A group of endocrine-disrupting contaminants that are widely spread in the environment and for which human exposure is ubiquitous are perfluoroalkyl substances (PFAS) [28,29].PFASs are synthetically produced chemicals that are extremely stable and accumulate in humans [30][31][32][33][34]. PFAS transfers effectively from mother to child during pregnancy and lactation [32][33][34][35][36] and may interfere with thyroid functions that are essential for brain development [36].
Animal studies have indicated developmental toxicity as a sensitive target after prenatal PFAS exposure [37].While a few epidemiological studies have reported adverse associations between PFAS serum levels and general neurodevelopment [36,38,39], the overall evidence is inconsistent [37].These previous studies have focused on the potential impact of PFAS on children's general development, but there has been a lack of studies on the impact of children's language and communication, which is an important indicator of neurocognitive development.Therefore, we performed a systematic review of the existing literature on pre-and postnatal exposure to PFAS with a focus on children's language and communication development.

Materials and Methods
We performed a systematic review according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) model [40] and the Swedish Agency for Health Technology Assessment and Assessment of Social Services (2017) [41].
We preregistered the study protocol at the Open Science frameworks on 29 October 2020, https://osf.io/y45ez/.

Eligibility Criteria and Research Definition
We applied the Population-Exposure-Comparison-Outcome model (PECO) [42] to define our research question.The population was defined as children and young adults aged 0-21 years.Exposure was assessed as measured PFAS concentrations or as residential history in a highly exposed population.To this, we added outcomes for language and communication ability based on the use of valid test instruments, parental self-reports, or clinical diagnoses.We included original peer-reviewed scientific studies published in the English language.

Information Sources and Search Methods
A stepwise literature search was performed with the help of two trained librarians at Gothenburg University Library to determine search concepts in the following three areas: (1) language and communication, including stuttering, voice, and swallowing, (2) neurodevelopmental ability problems where an assessment of language and communication abilities may be included and (3) PFAS and terms related to PFAS substances.The first search in PubMed, Scopus, and CINAHL was performed by the librarians on 16 June 2020.The specification for all databases and search concepts is available in a Supplementary File (see Supplementary File S1).An updated search to capture recently published papers was performed in April 2023, and five new records were added.During a pre-search, we found no study with language or communication as the primary study outcome using specific language test instruments.Therefore, the broader search concept was used to include proxy information about language and communication domains through instruments assessing children's general development.Moreover, reviews and systematic reviews were checked for references that might be of interest.

Study Selection
The relevance assessment of record titles and abstracts that included PFAS as the exposure was first conducted by two speech and language pathologists independently, with a pre-specified intention to "include rather than exclude".For inclusion, we required an assessment of children's language and communication and an assessment of PFAS.In the next step, an epidemiologist and a researcher in environmental medicine scrutinized the titles and abstracts that were included to assess whether the exposure assessments and study designs seemed to be relevant.

Data Extraction
We extracted the following metadata from the articles: the author, year, title, continent, country or state, possible connection to other screened articles, aim, population, study design, the name of study or study cohort, ages, counts, gender distribution, and the developmental domains assessed (e.g., cognitive, motor, language, and communication).In addition, information on which test instrument was used for the assessment of language and communication and if this assessment was comprehensive or only a part of broader cognitive testing was extracted.We noted reported PFAS substances, sampling periods (e.g., gestational week and child age), and matrix (whole blood, serum, plasma, or breast milk) or modeled exposure.We focused on the legacy PFAS perfluorohexane sulfonic acid (PFHxS), perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), and perfluorononanoic acid (PFNA).
An additional search in the Supplementary Material was also performed to obtain the information needed for decisions on inclusion and quality assessment.The validation of data followed the extraction of data, and during this process, decisions for inclusion and exclusion were made.Specific information on outcomes related to language ability was extracted from the main and supplementary text, tables, and figures.The summary measures of the results, i.e., medians, means, risk ratio, odds ratio, hazard ratio, and standard deviation, were extracted from the main and Supplementary Material.

Risk of Bias Assessment
The assessment was completed using the Critical Appraisal Skills Programme and checklist for cohorts [43].When necessary, additional information was added from supplements.All questions in the checklist were answered, but we primarily focused on a quality assessment concerning the following questions: -"Were the participants recruited acceptably?" -"Was the exposure accurately measured to minimize bias?" -"Was the outcome accurately measured to minimize bias?" -"Have the authors identified all the important confounding factors?" -"Have they taken into account the confounding factors in the design and/or analysis?"

Study Selection
The final study sample consisted of 15 studies (Table 1) published between 2013 and 2023.The initial literature search identified 164 records, and after the removal of 54 duplicates via EndNote 20, [44], the remaining 110 titles and abstracts were imported to the Rayyan review app [45].After screening titles and abstracts, there were nine articles imported to EndNote in full text.We performed full-text screening for relevance.During this process, one article was excluded because the outcome was irrelevant to the research question.After searching the reference lists of the included articles, we identified and included two additional studies.An updated literature search in April 2023 added five articles that had been published after the initial search.This process is shown in the PRISMA flow diagram (Figure 1).

Study Selection
The final study sample consisted of 15 studies (Table 1) published between 2013 and 2023.The initial literature search identified 164 records, and after the removal of 54 duplicates via EndNote 20, [44], the remaining 110 titles and abstracts were imported to the Rayyan review app [45].After screening titles and abstracts, there were nine articles imported to EndNote in full text.We performed full-text screening for relevance.During this process, one article was excluded because the outcome was irrelevant to the research question.After searching the reference lists of the included articles, we identified and included two additional studies.An updated literature search in April 2023 added five articles that had been published after the initial search.This process is shown in the PRISMA flow diagram (Figure 1).

Study Characteristics
Twelve studies were population-based birth cohorts (Table 1), all with maternal pregnancy or cord blood samples (Table 2).In addition, two studies collected childhood serum samples.
Three studies were initiated in response to a point source of exposure.One was a birth cohort that was part of a program collecting data on contaminants and health outcomes after the collapse of the World Trade Center [46].Two studies were part of large research programs assessing health effects after exposure to contaminated drinking water.The first was performed within the C8 Study Panel framework with exposure dominated by PFOA [47].The second was part of the Ronneby PFAS Research Program, with exposure dominated by PFOS and PFHxS [48,49].Both studies included prenatal exposure assessments.
Thirteen studies focused on children's general cognitive development and included verbal subscales (Table 3), whereas two studies had a specific focus on language and communication development (i.e., Jeddy et al., 2017 andStübner et al., 2023) [49,50].
The publications were from three continents, including six from Asia (China, Japan, Taiwan), five from Europe (Denmark, England, Norway, Spain, Sweden), and four from the U.S. The number of participants ranged from 120 to 11,895.Two publications reported data from the same cohort at different ages of follow-up [51,52].
One study included only girls [50].In the remaining studies, there was an even distribution between the sexes.

Exposure
The exposure characteristics of the studies are provided in Table 2. Fourteen studies relied on measured PFAS levels, albeit in different matrices, i.e., whole blood, serum, or plasma.The time of sampling varied between the first pregnancy trimester up to the age of 11 years.Eight studies relied on maternal exposure measurements and two on measurements in the cord blood, whereas four studies used a combination of maternal and cord/child measurements.All studies investigated exposure to PFOA, including fourteen for PFOS, eleven for PFHxS, and ten for PFNA.Three of the studies at background PFAS exposures had higher serum levels than the others [50,56,57].Two studies investigated highly exposed populations living in areas with contaminated drinking water.One was dominated by PFOA exposure, assessed by measured child PFAS levels, and modeled maternal exposure levels during pregnancy [47].The second study used maternal residential history and the municipal distribution of contaminated drinking water over time to construct a proxy variable of prenatal exposure and assessed its validity against measured serum levels in a smaller dataset [49].Here, exposure to PFOS and PFHxS dominated to a lesser extent than PFOA.

Outcome Assessment
The child's age at the outcome assessment ranged from 4 months to 12 years.Seven studies performed repeated outcome assessments over periods spanning from 4 months up to 8 years (Table 1).
Thirteen studies used objective instruments assessing children's general cognitive abilities with subscales for language and communication (Table 3).One study used a valid parental questionnaire targeting children's vocabulary, which was developed specifically to capture language and communication development [50], while another study included preschool teacher reporting [59].Harris et al. [56] used a specific language test instrument assessing vocabulary comprehension as a language sub-domain in their 3-year assessment.Finally, one study defined the clinical outcome of a developmental language disorder according to diagnostic codes (International Classification of Disorders, tenth revision, (ICD-10) [61], set by speech and language pathologists, using data from an administrative healthcare register with the population coverage of children that had been referred to the regional speech and pathology clinic by child nurses performing routine language screening at child health care centers [49].[38] (5 year assessment)

Association between PFAS Exposure and Language and Communication Outcomes
The findings of the included studies are summarized in Table 4.Among the 13 studies with background PFAS exposure levels, the majority found no statistically significant associations.Two studies reported better language ability at higher PFAS levels [53,57].Adverse associations were reported by Luo, Chen, Yu, Huo, Wang, Nian, Tian, Xu, Zhang, and Zhang [51], and two studies reported both positive and negative effect estimates [50,58].
The cohort established after the World Trade Center disaster [46] exhibited PFAS levels comparable to the background exposure cohorts, albeit with a more complex situation of chemical and psychophysiological exposure for pregnant mothers.Here, no effect of PFAS exposure was seen for the verbal scales while simultaneously indicating better general neurodevelopment-mental outcomes with increasing PFAS levels.
Two studies investigated children with substantially higher PFAS exposure.A small study (n = 320) reported no associations between PFOA exposure and language-related test outcomes [47].A large, registry-based study (n = 11,895) found an increased risk for DLD after high exposure dominated by PFOS and PFHxS, but only in girls [49].
Five of seven studies with repeated outcome measurements found different results between examinations [38,46,50,53,58].Eight out of thirteen studies that reported results for several PFAS compounds showed consistent results between the compounds.
Effect modification by sex was found in four of the eleven studies where it was investigated.Three studies had indications of an adverse risk for girls [49,55,58], while Spratlen et al. (2020) [46] found a favorable effect in girls at the age of 2 years.
In summary, the majority of the studies did not demonstrate an association between PFAS and language and communication development, while seven reported favorable or adverse associations.We identified no evident risk of selection bias from conditioning participation on both exposure and outcome in any of the studies.The study by Chen et al. [54], excluded children with physician-diagnosed neurodevelopmental disorders at age 2 years.The study by Skogheim et al. [59], had an enriched sample (20%) of children with ADHD symptoms at age 3 years according to parental interviews, and excluded a few children with high questionnaire scores on autistic symptoms.However, as only a very small proportion of children with neurodevelopmental disorders such as autism or ADHD receive a clinical diagnosis at such an early age [79], we assume the impact to be of minor importance for the generalizability of their results.

Information Bias
All PFAS analyses were performed in established laboratories.The studies that used a modeled exposure assessment provided cross-validation between modeled and measured data [47,49].Moreover, in these studies, the exposure levels spanned over a very wide range, which reduced the risk of exposure misclassification.By contrast, analytical uncertainty has a larger relative impact at low exposure levels.
There was a lack of information about blinding regarding exposure in three studies with objective outcome assessments [46,54,59] and in the study using parental questionnaires [50].The remaining eleven studies with objective outcome assessments had assessors blinded to participants' PFAS exposure levels.
Most studies relied on instruments assessing general cognitive abilities with verbal subscales.Although these tests can provide essential information about language difficulties, it is important to remember that such tests are not primarily designed to assess language ability.

Confounding
In general, studies were adjusted for the confounders that we a priori considered to be relevant (i.e., maternal age, socioeconomic status, parity, and smoking during pregnancy), and presented a theoretical ground as to why they were included.All studies presented both unadjusted and adjusted results in the main text or the Supplementary Materials.

Discussion
This systematic review identified fifteen studies that investigated the association between early-life exposure to PFAS (PFOA, PFOS, PFHxS, and PFNA) and language and communication development.Overall, there were no consistent findings for associations between early-life exposure to PFAS and language and communication development, neither adverse nor favorable.Sex dimorphic effects were reported in some studies, albeit in different directions.
All but two studies were performed in birth cohorts, with substantial variation concerning the timepoint for outcome assessments.The majority of studies used well-known, validated developmental or cognitive test instruments, capturing crude measures of language and speech domains as subtests within general test batteries.In marked contrast, one study relied on standardized expert diagnoses of DLD [49], thus detecting clinically important outcomes.
One explanation for the inconsistent findings might be related to the lack of specific language test instruments for outcome assessments.Although developmental tests or cognitive tests aim to assess the intellectual capacity of children and adolescents and can provide valuable information about possible language difficulties, these tests were not primarily designed to assess language difficulties.For example, children with expressive language disorder who struggle with grammar and speech sound pronunciation problems may still have full marks on these cognitive tests.It is also possible for children with reading and writing difficulties to pass the test because the tasks are administered and answered verbally.Thus, there is a risk that estimates may have become attenuated from outcome misclassification when general instruments are used.
Specific language test instruments are more sensitive to picking up nuances in language and communication, and they are, thus, expected to increase the validity of the outcome assessment.In addition, the use of caregiver questionnaires on language development can be a cost-and time-efficient method to capture important aspects of functional everyday communication [80,81].This approach was used in two of the included studies [50,59].
An alternative strategy for outcome assessment is to use clinical developmental language disorder diagnoses to define the outcome when healthcare data are available.In Stübner et al. (2023) [49], the outcomes were defined as (a) routine language screening with validated instrument child healthcare nurses followed by referral to a speech and language pathology clinic and (b) a subsequent diagnosis set by a speech and language pathologist.An advantage of this approach is that it captures clinically relevant outcomes, but it requires routine healthcare data for the study population to be available.
Only two studies were based on cohorts with high exposure levels of PFAS, and they reported diverging findings.
The remaining 13 studies investigated background-exposed cohorts with small exposure contrasts.Thus, there is insufficient evidence to evaluate exposure-response relationships, particularly at intermediate exposure levels.
Self-selection in longitudinal cohorts is an inherent risk, and parents with better socioeconomic characteristics are more likely to enroll as well as have higher compliance over time [82].In populations with background exposure, a higher PFAS exposure has been associated with higher parental socioeconomic status (SES) [83].Given that language ability is positively associated with parental SES [84], the association between language ability and PFAS exposure risk is biased toward the null in background-exposed birth cohorts.In contrast, the study by Stübner et al. (2023) [49] used routinely collected administrative healthcare data to define the outcome status for the entire population.With such a study design, selection bias can be avoided.

Conclusions
We found no consistent association between early-life exposure to PFAS and language and communication development.Most previous research was performed in populations with background levels of exposure; the timepoint of exposure and outcome assessment varied substantially, and most studies used general cognitive instruments for outcome assessment.However, the observation of an increased risk for the clinical diagnosis of developmental language disorders in highly exposed girls is of concern.Thus, research at intermediate exposure levels is needed to clarify the exposure-response relationship between early-life PFAS exposure and speech and language development.In future studies, instruments developed to assess language and communication abilities, preferably in collaboration with speech-and language pathologists with clinical expertise, should be used.

Table 1 .
General characteristics of studies included in the systematic review.

Table 3 .
Characteristics of outcome assessment tests or questionnaires including the domains tested and subtests including language and communication.

Table 4 .
Summary of study results.