Prenatal and childhood exposure to perfluoroalkyl substances (PFAS) and measures of attention, impulse control, and visual spatial abilities

Abstract

Background:

Despite evidence from toxicological studies describing the potential neurotoxicity of perfluoroalkyl substances (PFAS), their role in neurodevelopment remains uncertain amid inconsistent findings from epidemiological studies.

Methods:

Using data from 218 mother-child dyads from the Health Outcomes and Measures of the Environment Study, we examined prenatal and childhood (3 and 8 years) serum concentrations of four PFAS and inattention, impulsivity, and visual spatial abilities. At 8 years, we used the Conners’ Continuous Performance Test-II to assess attention and impulse control and the Virtual Morris Water Maze (VMWM) to measure visual spatial abilities.

Results:

In multiple informant models, there was no evidence to indicate that prenatal or childhood PFAS are associated with attention. However, there was an inverse association between prenatal ln-perfluorooctanoate (PFOA) and errors of commission (β=−2.0, 95% Confidence Interval [CI] −3.8, −0.3). Ln-perfluorononanoate (PFNA) at 3 years was associated with longer (poorer) VMWM completion times of 3.6 (CI 1.6, 5.6). However, higher concurrent concentrations of ln-perfluorohexane sulfonate (PFHxS) (β=−2.4 seconds, 95% CI −4.4, −0.3) were associated with shorter (better) times. Higher prenatal PFHxS was positively associated with percentage of traveling distance in the correct quadrant (β=4.2%, 95% CI 0.8, 7.7), indicating better performance.

Conclusion:

Findings were mixed for prenatal and childhood PFAS concentrations and visual spatial abilities. There is not enough evidence to support that PFAS are associated with visual spatial abilities as assessed by the VMWM or CPT-II measures of inattention or impulsivity in children at age 8 years.

1. Introduction

Perfluoroalkyl substances (PFAS) are a group of synthetic chemicals widely used in commercial and consumer applications for their ability to repel oil and water, while also resisting thermal, chemical, and biological degradation. The pervasive environmental contamination and long half-lives in humans of many PFAS have led to their widespread detection in wildlife and human tissues (Lau et al. 2007). PFAS have been detected in blood of pregnant women and children despite restrictions and the voluntary phase-out of several PFAS by major manufacturers in the United States and Europe (Bjerregaard-Olesen et al. 2017; Fromme et al. 2009; Morck et al. 2015; Okada et al. 2013; WHO 2013; Woodruff et al. 2011; Ye et al. 2018). Further, production of PFOS and PFOS equivalents has increased dramatically in China since 2002, reaching a peak of 250 tons in 2006 (Xie et al. 2013).

There is considerable concern regarding PFAS’ roles in human neurodevelopment given the accumulating data from animal studies highlighting cognitive and behavioral effects in prenatally and neonatally exposed mice (Fuentes et al. 2007; Johansson et al. 2008; Johansson et al. 2009; Viberg et al. 2013). Potential biological pathways for PFAS neurotoxicity include disrupting thyroid hormone homeostasis, affecting neuronal differentiation, altering the cholinergic system, and promoting neuronal cell apoptosis and reactive oxidative stress formation (Berntsen et al. 2017; Eggers Pedersen et al. 2015; Johansson et al. 2008; Johansson et al. 2009; Lee and Viberg 2013; Lee et al. 2013; Lee et al. 2016; Liu et al. 2013; Liu et al. 2015; Long et al. 2013; Reistad et al. 2013; Slotkin et al. 2008; Yu et al. 2016). PFAS have been detected in mouse brain (Yu et al. 2016) and can cross human placenta (Monroy et al. 2008). Thus, PFAS have the potential to impact neurobehavior and development, particularly during critical in utero periods of brain maturation. Postnatal PFAS exposure may also affect neurodevelopment as rapid brain growth continues to age 2 years and synaptogenesis, pruning, and myelination are not complete until adolescence (Rice and Barone 2000). Epidemiological studies on PFAS and inattention, impulsivity, and visual spatial abilities have been fairly limited and with inconsistent results (Fei et al. 2008; Gump et al. 2011; Lien et al. 2016; Stein et al. 2013). Therefore, we examined associations of prenatal and childhood serum PFAS concentrations with measures of inattention, impulsivity, and visual spatial abilities in school-aged children using the Health Outcomes and Measures of the Environment (HOME) Study.

2. Materials and methods

2.1. Study population

The HOME Study is an ongoing prospective pregnancy and birth cohort that enrolled 468 women in the second trimester of pregnancy from the greater Cincinnati area in Ohio (USA) from 2003 to 2006. These women were: 1) ≥18 years of age; 2) ~16±3 weeks of gestation; 3) living in a residential home constructed prior to 1978 (a criterion that was related to one of the original objectives of the HOME Study to examine interventions to reduce lead and injury hazards); 4) receiving prenatal care and planning to deliver at one of the collaborating obstetric practices; 5) HIV negative; and 6) not taking medications for seizures, thyroid disorders, chemotherapy, or radiation treatment. Further details regarding study recruitment, enrollment criteria, specimen and data collection, and follow-up visits have been published elsewhere (Braun et al. 2017b). Of the 390 women who delivered liveborn singleton infants, 218 (56%) mother-child dyads were included in the present study based on available PFAS measurements and assessments of inattention, impulsivity, and visual spatial abilities at age 8 years. This study was approved by the institutional review board (IRB) at the Cincinnati Children’s Hospital Medical Center (CCHMC); the Centers for Disease Control and Prevention (CDC) and collaborating institutions relied on CCHMC’s IRB as the IRB of record.

2.2. Perfluoroalkyl and polyfluoroalkyl substances

Serum concentrations (currently accepted acronyms followed, if pertinent, by previously used acronyms in parentheses) of perfluorooctanoate (PFOA), perfluorooctane sulfonate (PFOS), perfluorohexane sulfonate (PFHxS), perfluorononanoate (PFNA), perfluorodecanoate (PFDA, PFDeA), perfluorooctane sulfonamide (FOSA, PFOSA), 2-(N-methyl-perfluorooctane sulfonamide) acetate (MeFOSAA, Me-PFOSA-AcOH), and 2-(N-ethyl-perfluorooctane sulfonamide) acetate (EtFOSAA, Et-PFOSA-AcOH); were quantified using on-line solid-phase extraction coupled to high-performance liquid chromatography-isotope dilution tandem mass spectrometry (Kato et al. 2011). For quantification, calibration standards were spiked with calibration standards to account for potential matrix effects (Kato et al. 2014). For quality control (QC), each analytical batch included reagent blanks and QC materials; QC concentrations were 0.4-1.0 ng/mL (low-QC) and 2-16 ng/mL (high-QC), depending on the analyte. QC materials had a coefficient of variation of ~6%. The limits of detection (LODs) were between 0.1-0.2 ng/mL, depending on the analyte.

We measured PFAS concentrations in maternal serum samples collected at 16±3 weeks of gestation (n=178, 2003-2006), with some subjects having additional samples collected at 26 weeks of gestation (n=18) and within 24 hours of parturition (n=48). For women who had >1 measurement (n=40), an average was used for prenatal PFAS; otherwise, a single measurement at 16 or 26 weeks of gestation was used. We quantified PFAS concentrations in children’s serum samples collected at 3 (2006-2009) and 8 years (2011-2014). In the present study, we focused on PFOA, PFOS, PFHxS, and PFNA, because detection frequencies for PFDA, PFOSA, MeFOSAA, and EtFOSAA were <75%. Concentrations <LOD were replaced with LOD/√2 (<LOD=0.01% for prenatal PFHxS; all other analytes were detected in 100% of the samples) (Hornung and Reed 1990). Of the 218 children included in our analysis, PFAS were available for 204 prenatally, 141 at 3 years, and 190 at 8 years.

2.3. Conners’ Continuous Performance Test–II (CPT-II)

The CPT-II was administered to children at 8 years to assess attentional functioning and impulsivity. The CPT-II is a 14 minute computerized test that presents a series of 360 letters in rapid succession, each appearing singly on the screen for ~250 milliseconds (Conners 1992). Children are informed to press the spacebar whenever any non-“X” letter would flash on the screen and to refrain from pressing it when the letter “X” would appear. We examined 5 CPT-II outcome measures of performance. Errors of omissions are the number of times a child did not respond to a target (non-“X” letters), while errors of commission are the number of times a child incorrectly pressed the spacebar when the letter “X” was displayed. Higher errors of omission reflect inattention, whereas higher errors of commission may indicate impulsivity. Hit reaction time may indicate inattention (longer times) or impulsivity (shorter times) depending on the directionality of the value since it is the average reaction time to respond correctly. Measures of omission, commission, and hit reaction time were analyzed as age-standardized T-scores (mean=50, standard deviation=10). Tau (τ) is the mean of the exponential component of the hit reaction time distribution that is fitted to an ex-Gaussian distribution, with larger values reflecting problems with sustained attention.

2.4. Virtual Morris Water Maze (VMWM)

To assess spatial learning and memory retrieval performance, we used the Virtual Morris Water Maze (Astur et al. 1998; Braun et al. 2017a; Vuong et al. 2017), a computerized version of the validated analogue used in rodent research of visual spatial abilities (Morris 1981). The virtual environment is presented from a first-person perspective, with a circular pool within a square room. Visual cues are positioned on each wall for orientation and to facilitate spatial learning and memory. The platform was located in the center of the northeast quadrant beneath the surface of the water, remaining invisible until located. Children were instructed to locate the hidden platform as quickly as possible. There were four blocks of four trials. The hidden platform would remain stationary throughout the trials even though they would be randomly positioned in different quadrants at the start of each block of trials. Each trial would continue until the platform was reached and an audible cue would sound or if the platform was not located the trial would conclude at 60 seconds (s). After they completed 16 hidden platform trials, we assessed spatial memory retention with a 30 s probe trial that had the platform moved to a new location unbeknownst to the children. Unlike the previous trials, no audible cue would sound if the invisible platform was located.

Spatial learning measures included the average measures of time (s) and distance (pool units) traversed to locate the invisible platform within each block of trials, with shorter values indicating better performance. For assessment of spatial reference memory, the percentage of time and distance searching in the correct quadrant during the probe trial was used, with higher values indicating better retention of spatial location.

2.5. Statistical methods

We analyzed data on PFOS, PFOA, PFHxS, and PFNA, because of their high detection frequency (>99%) in maternal and child serum samples (Supplemental Table S1). We used multiple informant models to assess the relationship between prenatal and childhood ln-transformed (to reduce the influence of outliers) PFAS concentrations and CPT-II and VMWM measures, with each PFAS analyzed separately. These models account for correlations between repeated measures of PFAS, because they are non-standard versions of generalized estimating equations (Horton et al. 1999; Litman et al. 2007; Sanchez et al. 2011). We present separate estimates for each window of exposure, because some interaction terms (PFAS×child age) had a p<0.10. Additional examination of effect modification by child sex was completed with the inclusion of interaction terms (PFAS×sex×age; PFAS×age; PFAS×sex; age×sex), with p<0.10 considered statistically significant for PFAS×sex×age. Final models were adjusted for covariates identified as potential confounders based on a review of the literature (categorized in Table 1). Maternal characteristics and biological measurements (measured/collected at enrollment unless stated otherwise), including age, race/ethnicity, smoking status, alcohol consumption, depression (Beck et al. 1996), marijuana use, IQ (Wechsler 1999), serum concentrations of ∑PCBs (continuous, ng/g lipid), and blood lead levels (continuous, μg/dL) were included. We also adjusted for household income, child sex, and quality and quantity of the caregiving environment using Home Observation for Measurement of the Environment (HOME) scores at age 12 months (Caldwell and Bradley 1984). Child’s experience playing three-dimensional games, interest in playing video games, and history of motion sickness were additionally adjusted for in the VMWM models to determine whether overall conclusions differed. We also examined whether ever being breastfed influenced our overall findings for CPT-II and VMWM measures by including it in the models. Lastly, in a sensitivity analysis we examined associations using PFAS measured at 16±3 weeks gestation instead of the averaged prenatal PFAS concentrations.

Table 1.

Prenatal and concurrent PFAS serum concentrations (ng/mL), HOME Study^a

		Prenatal			8 years			Omission	Commission		Latency (s)	Distance
		PFOS	PFOA		PFOS	PFOA		Mean±SD	Mean±SD		Mean±SD	Mean±SD
	n	GM±GSD	GM±GSD	n	GM±GSD	GM±GSD	n			n
All participants	204	12.5±1.8	5.2±1.8	190	3.9±1.7	2.4±1.5	213	61.2±19.6	53.0±7.9	186	46.1±16.3	17.1±8.0
Age, years
<25	56	12.2±2.1	5.5±1.9	54	3.3±1.6*	2.0±1.5*	56	64.8±23.0	51.4±6.9	47	47.2±17.6	17.2±7.5
25-34	118	12.3±1.7	4.9±1.7	109	4.0±1.7*	2.5±1.5*	125	60.4±18.5	53.4±8.5	109	45.9±16.4	17.5±7.9
≥35	29	13.9±1.7	6.0±1.8	26	4.8±1.7*	3.0±1.6*	31	58.4±17.2	54.4±6.7	29	46.1±13.8	16.0±9.2
Race/ethnicity
Non-Hispanic White	121	14.1±1.7*	5.7±1.7*	111	4.5±1.7*	2.8±1.5*	128	56.8±14.1*	54.5±7.6*	115	44.0±15.0*	16.5±8.1
Non-Hispanic Black and Others	82	10.5±1.9*	4.6±1.8*	78	3.1±1.5*	2.0±1.5*	84	68.0±24.4*	50.8±7.9*	70	50.0±17.7*	18.2±7.8
Family Income
<$40,000	87	11.0±2.0*	4.8±1.9*	83	3.6±1.7*	2.1±1.5*	89	65.3±23.9*	52.3±7.3	75	48.5±17.7	18.3±6.6
$40,000-$79,999	65	12.1±1.5*	5.3±1.6*	57	3.8±1.6*	2.5±1.5*	69	58.1±14.9*	54.1±7.9	61	45.1±15.7	15.6±8.5
≥$80,000	51	16.4±1.7*	6.1±1.8*	49	4.7±1.7*	3.0±1.5*	54	58.6±15.9*	52.8±8.7	49	44.1±14.4	17.3±9.0
Maternal smoking status
Non-smoker	171	13.0±1.8	5.4±1.8	156	4.1±1.7*	2.5±1.6*	178	60.3±18.8	53.2±8.1	159	45.9±15.7	17.0±8.6
Environmental tobacco smoke	19	11.5±1.8	4.5±1.8	19	3.3±1.5*	2.0±1.4*	18	71.1±29.7	50.7±7.1	16	45.9±17.3	16.7±6.2
Active smoker	13	8.8±1.6	5.0±1.6	14	3.1±1.4*	2.1±1.3*	16	61.0±10.4	53.7±5.5	10	52.7±23.3	19.6±6.5
Maternal alcohol consumption
Never	116	12.4±1.9	5.0±1.8	105	3.6±1.7	2.3±1.5	119	62.0±18.5	52.0±8.1	106	44.7±17.0	16.7±7.9
<1 per month	59	12.6±1.8	5.3±1.7	59	4.2±1.7	2.5±1.6	63	60.9±22.4	53.5±7.7	54	48.1±15.3	18.5±8.4
>1 per month	28	12.8±1.6	6.1±1.7	25	4.4±1.8	2.7±1.6	30	59.0±18.4	55.7±6.7	25	48.9±14.8	16.3±7.6
Maternal depression
Minimal/mild	182	13.3±1.7*	5.5±1.7*	168	4.0±1.7	2.5±1.5*	191	60.5±18.6	53.3±7.8	167	47.0±15.9	17.4±8.0
Moderate/severe	19	7.3±2.2*	3.7±1.8*	19	3.1±1.4	2.0±1.4*	19	68.1±28.0	49.7±8.5	17	40.0±18.9	14.4±7.7
Home Observation for Measurement of the Environment score
≥40	115	13.8±1.6*	5.5±1.7	106	4.3±1.7*	2.8±1.5*	122	57.3±13.6*	53.8±8.1	108	45.5±14.8	16.4±8.2
35-39	41	12.0±1.8*	5.3±1.6	39	3.3±1.5*	1.9±1.4*	40	62.0±20.1*	52.8±7.0	37	42.6±14.4	17.0±6.6
<35	32	10.3±2.2*	4.8±2.0	30	3.6±1.9*	2.2±1.6*	34	69.7±27.2*	51.3±8.0	28	51.3±19.3	19.0±8.5
Marital status
Married/living with partner	148	12.8±1.8	5.2±1.8	134	4.2±1.7*	2.6±1.5*	156	59.0±17.1*	53.8±7.9*	138	44.6±15.1*	16.8±8.3
Not married, living alone	55	11.9±1.8	5.4±1.7	55	3.2±1.5*	2.0±1.5*	56	67.5±24.4*	50.8±7.4*	47	51.0±18.6*	18.1±7.2
Maternal drug use
No	191	13.0±1.8*	5.3±1.8	176	4.0±1.7*	2.5±1.5	199	60.9±20.0	53.0±8.0	174	45.7±15.8	17.1±8.1
Yes	12	7.1±2.0*	4.3±1.8	13	2.8±1.4*	2.0±1.3	13	65.8±11.7	53.7±5.8	11	54.6±21.2	17.9±5.8
Child sex
Male	92	11.8±1.8	4.9±1.8	87	3.8±1.7	2.3±1.6	95	63.5±22.3	52.7±8.1	86	43.3±16.7*	16.2±7.2
Female	112	13.0±1.9	5.5±1.8	103	4.0±1.7	2.5±1.5	118	59.4±16.9	53.3±7.7	100	48.6±15.6*	17.9±8.6
	n	Pearson r	Pearson r	n	Pearson r	Pearson r	n	Pearson r	Pearson r	n	Pearson r	Pearson r
Maternal IQ
(Mean±SD 105.6±15.1)	191	0.12	0.01	179	0.14	0.31*	199	−0.15*	0.09	175	−0.12	−0.12
Maternal PCBs
(GM±GSD 44.2±1.8 ng/g lipid)	185	0.17*	0.08	165	0.16*	0.23*	187	−0.06	0.10	171	−0.12	−0.12
Maternal blood lead
(GM±GSD 0.6±1.4 μg/dL)	203	<0.001	0.01	189	−0.11	−0.08	212	0.19*	0.02	185	0.07	0.07

Abbreviations: GM, geometric mean; GSD, geometric standard deviation; SD, standard deviation.

^aCollection years were 2003-2006 and 2011-2014 and for maternal serum at 16±3 weeks gestation and for child serum at age 8 years, respectively.

^*p < 0.05

3. Results

3.1. Participant characteristics

Median PFOS and PFOA concentrations were higher than PFHxS and PFNA concentrations during both gestation and childhood (Supplemental Table S1). The GMs of PFOA and PFOS were lowest at age 8 years, with at least half the concentration of maternal serum. GMs of PFHxS and PFNA at age 8 years were fairly similar to prenatal levels. PFAS were significantly correlated with each other within the same measurement period and between windows of exposure (Supplemental Table S2). Concentrations of prenatal PFAS were significantly higher among children of mothers who were non-Hispanic white, minimally/mildly depressed, did not use marijuana during pregnancy, and were from households with higher incomes and caregiving environment scores (Table 1). These differences were similarly observed with PFAS at 8 years. PFOS (prenatal, 8 years) and PFOA (8 years) concentrations were positively correlated with maternal serum ∑PCBs (r_pearson=0.16-0.23). Prenatal PFOA at age 8 years was positively correlated with maternal IQ. HOME Study children had higher errors of omission (61.2±19.6) and commission (53.0±7.9) than the than the population mean (50±10). Higher errors of omission were noted among children who had mothers who identified as non-Hispanic black or others and who were not married or living alone. Children with higher errors of omission were also more likely to be from households with lower incomes and HOME scores. Commission errors were significantly higher among children of mothers who were non-Hispanic white and who were married or living with a partner. These children were also more likely to have lower completion times on the VMWM. Significantly lower completion times on the VMWM were also observed among male children compared to females. Children excluded from the present study due to incomplete information were significantly more likely to have mothers who were married, had a lower IQ, and had lower levels of blood lead during pregnancy than mothers of children included in the analyses (Supplemental Table S3). However, participants were similar with respect to prenatal and childhood serum PFAS, socioeconomic status, home environment, and maternal lifestyle/behavioral factors.

3.2. Perfluoroalkyl substances and CPT-II

There was no evidence to show that prenatal and childhood PFAS were associated with errors of omissions (Figure 1; Supplemental Table S4). While associations between childhood PFOA concentrations and errors of commission were negligible, a ln-unit increase in prenatal PFOA concentrations was associated with lower errors of commission scores (β=−2.0, 95% CI −3.8, −0.3), indicating better performance. No relationship was noted between PFAS at any time and hit reaction time or τ.

Figure 1.

Estimated score differences and 95% CIs in CPT-II measures of inattention and impulsivity by a ln-increase in serum concentrations of PFAS in each exposure assessment window, HOME Study. Adjusted by maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, HOME Score, marital status, maternal marijuana use, maternal IQ, maternal serum PCBs, maternal blood lead levels, and child sex.

3.3. Perfluoroalkyl substances and VMWM

Higher concentrations of PFOS (β=−2.1 s, 95% CI −4.9, 0.6 per ln-unit increase) and PFHxS (β=−2.4 s, 95% CI −4.4, −0.3) at 8 years were associated with shorter VMWM completion times, indicating better performance (Figure 2; Supplemental Table S5). In contrast, a ln-unit increase in PFNA concentrations at ages 3 and 8 years was associated with longer completion times of 3.6 s (95% CI 1.6, 5.6) and 2.0 s (95% CI −0.6, 4.5), respectively. For visual spatial memory retention, prenatal concentrations of PFHxS were associated with higher percentage of the total distance traveled (β=4.2%, 95% CI 0.8, 7.7 per ln-unit increase) searching in the quadrant that the invisible platform was originally located, denoting better reference memory. A positive, albeit non-statistically significant, association was present between PFHxS at 3 years and percentage of distance traveled (β=2.3%, 95% CI −0.9, 5.6 per ln-unit increase).

Figure 2.

Estimated mean differences and 95% CIs in VMWM measures by a ln-increase in serum concentrations of PFAS in each exposure assessment window, HOME Study. Adjusted by maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, HOME Score, marital status, maternal marijuana use, maternal IQ, maternal serum PCBs, maternal blood lead levels, and child sex. Higher values on VMWM block trials indicate poorer performance. Lower values on VMWM probe trials indicate poorer performance.

3.4. Child sex differences

We observed one consistent trend for both PFOS and PFHxS at age 8 years, which was that males had better performance, but females tended to have either worse performance or null findings (Figure 3). Specifically, male children had lower errors of omission with increased concurrent concentrations of PFOS (β=−7.3, 95% CI −13.0, −1.7) and PFHxS (β=−4.5, 95% CI −10.0, 1.0), while females had higher omission errors (PFOS: β=4.3, 95% CI −1.2, 9.9; PFHxS: β=3.2, 95% CI −1.1, 7.4). Higher concentrations of concurrent PFOS and PFHxS were also associated with shorter hit reaction times in males, but null associations in females. A ln-unit increase in PFHxS at age 8 years was also associated with lower times on the VMWM (β=−4.2 s, 95% CI −6.9, −1.6) in males, but were not associated with visual spatial learning in females (β=−0.9, 95% CI −4.0, 2.1).

Figure 3.

Estimated associations between prenatal and childhood serum concentrations of PFAS (ng/mL) and CPT-II and VWMW measures at 8 years by child sex, HOME Study. Adjusted by maternal age, race/ethnicity, household income, maternal smoking status, maternal alcohol consumption, maternal depression, HOME Score, marital status, maternal marijuana use, maternal IQ, maternal serum PCBs, maternal blood lead levels, and child sex.

3.5. Sensitivity analyses

Additionally adjusting for breastfeeding did not change our results. For instance, we still noted longer times (β=3.4 s, 95% CI 1.3, 5.5) in the VMWM with an increase in ln-PFNA concentrations at age 3 years, and shorter VMWM times (β=−2.3 s, 95% CI −4.4, −0.3) and higher percentages of the distance traveled (β=4.3%, 95% CI 0.8, 7.7) in the correct quadrant with concurrent and prenatal PFHxS, respectively. No differences in the study results were observed after further adjustment for child’s experience or interest with three-dimensional games and history of motion sickness (results not shown). Using serum concentrations at 16±3 weeks did not change our main findings (results not shown). However, effect measure modification by child sex was no longer present between prenatal PFOS associations and hit response time and VMWM visual spatial learning times.

4. Discussion

Our study does not provide evidence that prenatal or childhood PFAS serum concentrations are associated with inattention measured by Conners’ CPT-II at age 8 years. However, higher prenatal PFOA concentrations were significantly associated with lower errors of commission, indicating better impulse control. These findings are discordant with those reported by Gump et al. (2011), where several PFAS were associated with shorter inter-response times in a cross-sectional study of children at 10 years, suggesting problems with impulse control. In terms of sustained attention, our null findings align with those of the Danish National Birth Cohort (DNBC) that concluded no relationship between prenatal PFOS and PFOA and attention at 18 months (Fei et al. 2008). In contrast, protective associations with cord serum PFOA and PFNA have been reported in a Taiwanese birth cohort of children at 7 years (Lien et al. 2016) and in the C8 Heath Project with childhood PFOA (2-8 years) (Stein et al. 2013). Epidemiological studies examining PFAS and ADHD-like behaviors and diagnoses have also shown inconsistent results, reporting adverse (Hoffman et al. 2010; Liew et al. 2015), null (Ode et al. 2014; Quaak et al. 2016; Strom et al. 2014), and mixed results (Stein and Savitz 2011; Stein et al. 2014).

Our findings regarding the associations between PFAS and VMWM measures were mixed. For visual spatial learning, we observed that children with higher PFNA concentrations at 3 and 8 years took more time to locate the hidden platform, indicating worse spatial learning. In contrast, concurrent PFOS and PFHxS concentrations were associated with shorter completion times. A positive association was additionally observed between prenatal PFHxS concentrations and visual spatial memory retention. Only one epidemiological study has examined PFAS and visual spatial abilities. The C8 Health Study reported higher (better) scores for visual spatial processing among children with higher PFOA concentrations at age 2-8 years, but no associations with maternal concentrations (Stein et al. 2013).

In the present study, only PFNA at age 3 years was identified as having a potential adverse association with visual spatial learning, whereas results for prenatal and concurrent concentrations of PFHxS were associated with higher percentages of traveling distance in the correct quadrant and shorter completion times, respectively. In addition, prenatal PFOA was associated with lower errors of commission at age 8 years. PFAS neurotoxicity varies depending on the carbon backbone, functional moieties, molecular weight, and hydrophobicity. While PFOS produces more oxidative stress and has higher genotoxic and bioaccumulation potential, PFNA causes more membrane instability and damage (Liu et al. 2013). The mean concentration of PFNA in human brain tissue (29.7 ng/g wet weight) has been found to be substantially higher than that of PFOS, PFHxS, and PFOA (4.9, 3.2, and <2.40 ng/g wet weight, respectively) (Perez et al. 2013). The protective association observed with PFOS and PFOA may be due to their role as peroxisome proliferator-activated receptor-gamma (PPARγ) agonists (Vanden Heuvel et al. 2006). PPARγ agonists mitigate inflammation associated with chronic and acute neurological insults, and are hypothesized as having neuroprotective effects (Kapadia et al. 2008).

Child sex modified associations between PFAS and attention and visual spatial abilities with varying directionalities for males and females depending on the window of exposure. It should be noted, however, that effect estimates were imprecise due to our limited sample size. We observed that prenatal PFOS and PFHxS concentrations were associated with worse performance on the CPT-II errors of omission for males, but weakly better performance or null findings in females. This may partially explain the overall null findings when both sexes were combined. In contrast, for concurrent concentrations, there was a reverse in the directionality of the association for males and females, with males having a better performance for CPT-II errors of omission, hit reaction time, and VMWM spatial learning time. Previously, the C8 Health Study reported that sex modified the relationship between PFOA at 2-8 years and ADHD Index scores at 6-12 years, with a favorable association in males and an adverse association in females (Stein et al. 2014). One possible explanation may be variations in PFAS excretion rates. Longer half-lives were reported in males for PFHxS (7.4 years) and PFOS (4.6 years) compared to females (PFHxS=4.7 years; PFOS=3.1 years) (Li et al. 2018). However, we could not find any biological mechanism that would explain the observed shifts in the directionality of the associations between males and females by timing of exposure. Further, our study is underpowered to draw conclusions regarding sex differences given our limited sample size.

The present and previous studies have yielded inconsistent results. Differences in PFAS concentrations and timing of measurements may have contributed to discordant results. Participants in the C8 Health Study had higher concentrations of PFOA due to contaminated drinking water, with an average maternal PFOA concentration of 115.9 ng/mL compared to median serum concentrations of ~5-6 ng/mL in the HOME Study and the DNBC. Childhood serum PFOA concentration in the C8 Heath Study (91.9 ng/mL at 6-12 years) far exceeded that observed in the HOME Study (2.7 ng/mL at 8 years). Secondly, the HOME Study is more generalizable to the U.S. population than the C8 Health Study, having more racial/ethnic diversity with ~40% non-Hispanic Black and other racial/ethnic groups compared to ~5% in the C8 Health Study. Third, we used multiple informant models to examine repeated measures of PFAS, which had not been done in previous studies with repeated PFAS concentrations. Lastly, we utilized the CPT-II to assess inattention and impulsivity, whereas many studies used other assessment batteries or assessed attention as a component of ADHD. In addition, studies that assessed visual spatial abilities examined processing and perception, while we focused on learning and reference memory.

Our study has several notable strengths. The study was conducted using a well-established pregnancy and birth cohort that had PFAS serum reference ranges similar to the U.S. general population, except for higher PFOA concentrations. We had repeated measures of serum PFAS and detailed data on socioeconomic status, the home environment, maternal IQ and depression, and other potential neurotoxicants. We also used a statistical model that enabled us to test for susceptible windows of neurotoxicity. Lastly, the VMWM is comparable to the Morris Water Maze test, which is a widely accepted assessment of visual spatial learning and memory in rodent studies.

Our study has several limitations. First, due to our limited sample size we were underpowered to draw major conclusions. Our results were mixed with regard to the children’s performance on the VMWM, thus conclusions from this study should be interpreted cautiously as null findings. Further, sex-specific associations between prenatal and childhood PFAS and CPT-II and VMWM measures were limited by sample size. Thus, our estimates for effect modification by child sex were imprecise. Selection bias may be present given study attrition. Children lost to follow-up were more likely to have married mothers, with a lower IQ, and a lower blood lead level during pregnancy than those included (Supplemental Table S3). Third, residual confounding may be an issue as we do not have information on other factors that may affect CPT and VMWM measures, including neighborhood characteristics, detailed dietary intake, school settings, and video gaming use. However, we have information on the home environment with regard to the overall quality of the child’s cognitive stimulation and emotional support offered by the family, as measured by the HOME scores with an observation of how nurturing the home environment was. Further, we adjusted for household income, which may affect neurodevelopment through various mediators, including parental care, availability of resources for stimulating child development, and caregiver involvement and interaction. Fourth, multiple comparisons remains a potential concern. While using multiple informant models correspondingly reduced the total number of models and produces more precise confidence intervals for each window of susceptibility, it does not reduce type 1 error if we do not assume equal associations over time. Lastly, serum PFAS concentrations are moderately to highly-correlated. Thus, it is difficult to fully differentiate associations of one individual compound from the associations of other PFAS.

5. Conclusions

We observed mixed associations between prenatal and childhood serum PFAS concentrations with children’s attention and visual-spatial abilities, with both better and worse performance on the VMWM. Further research is need to investigate the associations between prenatal and childhood PFAS concentrations and measures of attention, impulsivity, and visual spatial abilities using cohorts with larger sample sizes so that sex specific differences can also be explored.