Associations between per- and polyfluoroalkyl substances and metabolic dysfunction-associated steatotic liver disease in adolescents and young adults: modifying roles of age, lifestyle factors, and PNPLA3 genotype

Abstract

Studies in youth on per- and polyfluoroalkyl substances (PFAS) exposure and metabolic dysfunction-associated steatotic liver disease (MASLD) are limited. The study aimed to evaluate associations between PFAS and MASLD risk and the modifying role of age, lifestyle factors, and genetic risk. We analyzed data from two independent cohorts: the Study of Latino Adolescents at Risk of Type 2 Diabetes (SOLAR) following adolescents aged 8–13 years over 6 years (n=162; recruited between 2001–2012); and the Metabolic and Asthma Incidence Research (Meta-AIR) study of young adults aged 17–23 years (n=122; recruited between 2014–2018). Eight PFAS were measured in plasma using liquid chromatography-high-resolution mass spectrometry. MASLD was defined as hepatic fat fraction > 5.5% on magnetic resonance imaging plus ≥ 1 cardiometabolic risk factor (high BMI, fasting glucose, blood pressure, triglycerides, and low high-density lipoprotein). Firth’s penalized logistic regression models estimated associations between PFAS and MASLD. We evaluated the interaction between each PFAS and age, lifestyle factors (cigarette smoking, alcohol drinking, physical activity, sleep duration, and diet in adult cohort), and PNPLA3 genotype, a key liver fat–regulating gene. Among adolescents, each doubling in plasma PFOA was associated with higher odds of MASLD (OR = 2.69; 95% CI: 1.16–6.26; p = 0.02), and the association strengthened with increasing age (PFOA × Age OR = 1.45; 95% CI: 1.03–2.06; p = 0.04). Risk was further elevated among older adolescents and those with PNPLA3 risk allele. Among young adults, overall associations were not significant, but MASLD risk was higher among smokers with elevated PFDA, PFHpS, and PFNA levels (interaction p = 0.01, 0.02, 0.02, 0.03, respectively). Our findings support that PFAS may increase the risk of MASLD in youth, with age, genetics, and smoking modifying susceptibility.

1. Introduction

Metabolically-dysfunction-associated steatotic liver disease (MASLD), formerly known as nonalcoholic fatty liver disease (NAFLD), is the most prevalent chronic liver disease and a growing global health concern,^1–3 and a significant and underrecognized concern in pediatric populations. It is estimated to affect 7%-9% of children in the general population and up to 38%-41% of children with obesity.⁴ Additionally, MASLD is strongly associated with an elevated risk of extrahepatic complications.^5,6 The pathogenesis of MASLD in youth is multifactorial, influenced by genetic predisposition, lifestyle factors, and environmental exposures.^7,8 For example, PNPLA3, commonly expressed in liver and fat tissues and encodes a protein that can help regulate the development of adipocytes and the production and breakdown of fats, is strongly associated with an increased risk.⁹

Adolescence and young adulthood are sensitive windows for liver and metabolic function development, with rapid somatic growth, puberty-related hormonal shifts, transient physiologic insulin resistance, and changes in adiposity and bile acid homeostasis that can increase their vulnerability to environmental toxicants.^10–13,13 Early-life and adolescent PFAS exposures can bioaccumulate due to long half-lives and transplacental and lactational transfer, potentially programming lipid metabolism and inflammatory pathways that influence MASLD trajectories into adulthood.^14–17 Studying young populations, therefore, can identify a sensitive exposure window, gene-environment interaction (for example, with PNPLA3), and prevention opportunities when lifestyle interventions and exposure reduction may yield the most significant long-term health impact.

Per- and polyfluoroalkyl substances (PFAS) are a class of synthetic chemicals widely used in consumer products, such as non-stick cookware, water-resistant textiles, and food packaging, and are persistent contaminants in drinking water and food supplies.¹⁸ Due to their long biological half-lives, PFAS bioaccumulate in human tissues, raising concerns about their potential for long-term health effects.¹⁶ Emerging evidence indicates that PFAS are hepatotoxic and are associated with liver injury, lipid metabolism dysregulation, and liver disease.^19,20 Mechanistically, PFAS have been shown to activate peroxisome proliferator-activated receptor-alpha (PPARα) in hepatocytes, which plays a key role in regulating lipid metabolism.²¹ PFAS can also inhibit farnesoid X receptor (FXR) in hepatocytes, which can lead to alterations in bile acid homeostasis and disruption in cholesterol homeostasis.²² Experimental studies in animal models have demonstrated that exposure to certain PFAS leads to hepatomegaly, altered liver enzyme levels, and histological changes consistent with MASLD.^19,23,24

In human epidemiological studies, findings have been mixed, though several studies in multiple countries have reported associations between serum legacy PFAS levels, including perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), perfluorononanoic acid (PFNA), perfluorohexane sulfonic acid (PFHxS), and perfluorodecanoic acid (PFDA), and markers of liver injury.^{19,25(pp2005–2018),26(pp2003–2014),27–30} Additionally, among adults with MASLD, emerging or replacement PFAS such as perfluoroheptanoic acid (PFHpA) has been linked to liver steatosis,³¹ and perfluorobutane sulfonate (PFBS), perfluoroheptane sulfonate (PFHpS), and other emerging PFAS such as perfluorohexanoic acid (PFHxA) have been associated with dysregulated liver function biomarkers in liver cancer patients.³² However, few studies used the gold standard measure of hepatic fat fraction and defined MASLD under the new diagnostic criteria in 2023.³³ Moreover, most epidemiological research on PFAS and liver disease has focused on adults, with limited data available on children/adolescents. A few studies have linked prenatal and childhood exposures to PFAS to metabolic alterations and elevated liver injury biomarkers, suggesting increased severity of nonalcoholic fatty liver disease and greater susceptibility to liver injury in children.^34–36 However, no previous studies have focused on MASLD risk due to PFAS specifically in children or adolescents, likely because imaging-based liver fat quantification is costly and MASLD is emerging in children and adolescents.

Also, no previous studies have evaluated the potential interaction between environmental exposure and other risk factors for MASLD. Lifestyle behaviors such as smoking, alcohol consumption, diet, physical activity, and sleep duration are key determinants of hepatic lipid metabolism, oxidative stress, and inflammation, which are also pathways influenced by PFAS through PPARα and FXR signaling.^7,8,21,22,37 These overlapping biological mechanisms suggest that lifestyle factors may modify the hepatotoxic effects of PFAS. In addition, genetic susceptibility plays a crucial role: the PNPLA3 I148M variant impairs triglyceride hydrolysis in hepatocytes and promotes lipid accumulation, potentially amplifying PFAS-induced disruptions in lipid metabolism.^9,21,37,38 Evaluating these modifiers together can help identify subgroups at greatest risk for PFAS-associated MASLD.

In this study, we aim to evaluate the association between PFAS exposure and the risk of MASLD in two distinct cohorts of adolescents and young adults with a parental history of type 2 diabetes or a history of overweight at enrollment. Studying these high-risk groups provides a unique opportunity to examine PFAS–MASLD associations in metabolically vulnerable groups, where early environmental influences on liver health may be more identifiable. We further evaluate the interactions of PFAS exposure with lifestyle factors in young adults (smoking, alcohol drinking, sleep, physical activity, and diet) and genetic risk in both adolescents and young adults (PNPLA3 genotype). Our findings may contribute to a better understanding of environmental risk factors for MASLD and inform regulatory policies aimed at mitigating PFAS-related health risks.

2. Method

2.1. Study Population

The Study of Latino Adolescents at Risk (SOLAR) participants (n=328) were recruited between 2001 and 2012. Participants were recruited between 2001 to 2003, and between 2010 to 2012. Adolescents were eligible if they met the following criteria: 1) were between 8 and 13 years old; 2) had a body mass index (BMI) ≥85th percentile (age- and sex-adjusted); 3) were free of type 1 or type 2 diabetes; 4) were of Hispanic/Latino ethnicity; 5) had a direct familial history of type 2 diabetes; and 6) were not taking any medications known to influence glucose metabolism. Details of the study design are available elsewhere.^39–42 We included 74 with MASLD status determined at baseline (referred to as SOLAR Phase I cohort) and 88 with MASLD determined only at follow-up (one follow-up visit; mean [sd] follow-up years: 6.01 [3.39]) (referred to as SOLAR Phase II cohort).

The Metabolic and Asthma Incidence Research (Meta-AIR) study participants (n=172) were recruited for a single clinical visit between 2014 and 2018.⁴³ At the time of the visit, participants 1) were between 18 and 23 years old; 2) were free from type 1 or type 2 diabetes; and 3) were not taking any medications known to influence glucose metabolism. Meta-AIR participants were selected from the Southern California Children’s Health Study (CHS) cohort, and all Meta-AIR participants had a history of overweight or obesity (age- and sex-adjusted) in the 9th or 10th grade.⁴⁴ We included 122 with both PFAS and MASLD status.

A flowchart is in Figure S1. The present study intentionally included cohorts of high-risk adolescents and young adults. Blood was drawn at baseline for both cohorts.

The study was approved by the USC institutional review board (IRB Number: HS-19-00338). Written informed assent/consent was obtained.

2.2. PFAS in plasma

In both Meta-AIR and SOLAR studies, PFAS concentrations were quantified using liquid chromatography coupled with high-resolution mass spectrometry (LC-HRMS). Details on the quantification of PFAS in plasma are available elsewhere.⁴⁵ Eight PFAS were quantified in plasma, including perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroheptane sulfonic acid (PFHpS), perfluoroheptanoic acid (PFHpA), perfluoroheptanoic sulfonic acid (PFPeS), and perfluorohexane sulfonic acid (PFHxS). PFHpA was quantified in the Meta-AIR study, but we did not detect PFHpA in the Meta-AIR study participants. We then log2 transformed raw PFAS concentrations to allow a better model fit. Limits of detection (LODs) were 0.43 μg/L for PFOS, 0.05 μg/L for PFHpS, and 0.01 μg/L for PFOA, PFNA, PFDA, PFPeS, PFHxS, and PFHpA. We imputed the missing values as the LOD divided by the square root of 2. Detection frequency was included in Table S1. PFPeS had very low concentrations and detection rates in both cohorts and was not included in the statistical analysis (mean: 0.07 ng/mL).

2.3. MASLD

Hepatic fat fraction (HFF) was assessed via 3-Tesla whole-abdominal magnetic resonance imaging (MRI) scans. The 3-T scans were based on a 3D Iterative Decomposition using Echo-Asymmetry in the Least squares sense (IDEAL) pulse sequences, which has been validated.^46,47

MASLD was determined by 1) HFF > 5.5% and 2) at least one cardiometabolic risk factors.^33,48

1. At least one of the following (adult criteria):

   1. BMI>=25 kg/m²
   2. Fasting serum glucose >= 5.6 mmol/L
   3. Blood pressure >= 130/85 mmHg
   4. Plasma triglycerides >= 1.7 mmol/L
   5. Plasma high-density lipoprotein (HDL) <= 1.0 mmol/L (Male) or 1.3 mmol/L (Female)

We applied pediatric criteria for the SOLAR cohort (adolescent cohort).^33,48

1. BMI>=85^th percentile for age/sex

2. Fasting serum glucose >= 5.6 mmol/L

3. Blood pressure >= 95^th percentile for age < 13yrs

4. Plasma triglycerides >= 1.7 mmol/L for age >= 10yrs and >= 1.15 mmol/L for age < 10yrs

5. Plasma high-density lipoprotein (HDL) <= 1.0 mmol/L

2.4. Covariates

In both cohorts, we included the following potential confounders based on prior literature:⁴⁹ age (recentered at the mean; in years), sex (male and female), body mass index (BMI) (in kilograms per square meter), parental education (below high school, high school, above high school), follow-up time (among people with liver outcomes measured at follow-up; in years). In the Meta-AIR study population, we additionally included race and ethnicity (White, Hispanic, or Other).

2.5. Effect Measure Modifiers

Genomic DNA was isolated from participants’ blood samples. In Meta-AIR, the PNPLA3-I148M genotype was validated via PCR/restriction fragment length polymorphism methods. Genotyping for PNPLA3 in SOLAR was performed using the Applied Biosystems, Inc., Foster City, CA, USA, (ABI) TaqMan system.

In Meta-AIR only, we included cigarette smoking (ever vs never), alcohol drinking (ever vs never), physical activity (continuous; the scale of 1 (least active) to 11 (most active)), nightly sleep duration (in hours), and Healthy Eating Index (HEI). HEI 2015 (continuous) was derived from 24-hour recalls, using the Nutrition Data System for Research (NDSR) (University of Minnesota).

We created a composite lifestyle score to capture overall lifestyle quality across five domains: Healthy Eating Index, physical activity, sleep duration, smoking, and alcohol use. Each component was coded so that healthier behavior received one point and less healthy behavior received zero points, with cut points chosen a priori (higher HEI and physical activity, sleep 7–11 hours, never smoking, and no alcohol). Points were summed to yield a 0–5 score. Participants with fewer than three non-missing components were set to missing. The score was standardized (mean 0, SD 1) for modeling.

2.6. Statistical Analysis

We conducted multiple logistic regression analyses to evaluate the associations between individual PFAS and MASLD, adjusting for the potential confounders, including age, sex, BMI, parental education, follow-up time (only in SOLAR Phase II cohort), and race/ethnicity (only in Meta-AIR cohort). We applied Firth’s bias-reduced penalized maximum-likelihood logistic regression to mitigate small-sample bias. Given the relatively small sample size of each SOLAR Phase I and Phase II, we included pooled analyses for SOLAR cohort participants for all subsequent analyses.

For all models, we included an a priori interaction between age (centered at the cohort mean) and each PFAS because susceptibility to environmental hepatotoxicants plausibly varies across adolescence and early adulthood. During puberty and late adolescence, rapid changes in body composition, insulin sensitivity, bile acid homeostasis, and hepatic nuclear-receptor signaling (e.g., PPARα/FXR pathways implicated in PFAS-related lipid metabolism) may amplify or modify PFAS effects on steatosis. Age also correlates with cumulative exposure duration and elimination kinetics and thus, the PFAS effect on MASLD could reasonably differ by age. A positive PFAS×Age OR (>1) indicates a larger PFAS effect at older ages within a cohort, while an OR (<1) indicates attenuation with increasing age.

We used a novel exposure burden score, derived via item response theory (IRT),⁵⁰ to quantify PFAS mixture effects on MASLD. The PFAS mixture refers to a combined exposure metric that accounts for the joint effect of PFAS to better capture the potential cumulative impact of co-occurring chemicals. The scores follow a standard normal distribution (mean = 0, SD = 1), such that a value of +1 indicates a one-standard-deviation higher PFAS exposure burden. The burden score was created relative to NHANES PFAS levels. In addition to the overall PFAS burden score, we additionally generated perfluoroalkyl carboxylic acids (PFCA; which included PFOA, PFNA, and PFDA) burden score, and perfluoroalkyl sulfonic acids (PFSA; which included PFOS, PFHxS, and PFHpS) burden score.

Given that there are significant sex differences in metabolism or exposure to PFAS and development of MASLD, we further conducted sex stratified analysis.

As a sensitivity analysis, in the Meta-AIR cohort, we additionally controlled for the effect measure modifiers. We repeated analyses with both individual and burden scores.

Lastly, we evaluated the interaction between each of the lifestyle factors as well as the composite score, and we included a three-way interaction among PFAS exposure, age, and lifestyle factors, two-way interaction between PFAS and lifestyle factor in addition to our main model in Meta-AIR. In both SOLAR and Meta-AIR cohorts, we evaluated the interaction between PFAS and PNPLA3 genotype (GG vs. CC/CG), and similarly, we included three-way and two-way interaction terms among PFAS, age, and PNPLA3 genotype. A significant interaction term means that the effect of PFAS is larger among older participants with the genetic risk. We repeated the analysis with individual PFAS and the PFAS burden score.

We applied Benjamini–Hochberg false discovery rate (FDR) correction across PFAS within each cohort. Both nominal and FDR-adjusted p-values are presented. Given the small sample size and independent hypothesis for each PFAS, we still considered a crude p-value < 0.05 as our significance threshold for both main effect and interaction terms.

3. Results

Table 1 shows the characteristics of participants across the three cohorts: SOLAR Phase I (N = 74), SOLAR Phase II (N = 88), and Meta-AIR (N = 122) by MASLD status. The prevalence of MASLD was highest in the SOLAR Phase I cohort (49%), followed by the SOLAR Phase II (27%) and Meta-AIR (18%) cohorts. Across cohorts, participants with MASLD had higher BMI and lower parental education compared to those without MASLD. PFOA, PFHpA, and PFHxS were slightly higher among those with MASLD compared to those without. MASLD prevalence increased with age and pubertal stage, with SOLAR I representing pre- to mid-pubertal children, SOLAR II older adolescents, and Meta-AIR young adults. All participants in SOLAR I and II were Hispanic, whereas Meta-AIR participants were a majority Hispanic and White. Lifestyle factors for Meta-AIR participants were described in Table S2. In Meta-AIR, participants had similar proportions of cigarette smoking, alcohol drinking, and physical activity between those with and without MASLD, but slightly longer nightly sleep duration and a modestly lower Healthy Eating Index among participants with MASLD compared to those without MASLD.

Table 1.

Characteristics of the study population by metabolic dysfunction-associated steatotic liver disease (MASLD) status.

Characteristic	Statistics	SOLAR Phase I (n = 74)		SOLAR Phase II (n = 88)		Meta-AIR (n = 122)
		MASLD (n = 36)	No MASLD (n = 38)	MASLD (n = 24)	No MASLD (n = 64)	MASLD (n = 22)	No MASLD (n = 100)
Baseline PFAS (ng/mL)	mean (sd)
PFDA		0.27 (0.12)	0.30 (0.17)	0.29 (0.21)	0.28 (0.09)	0.19 (0.08)	0.23 (0.10)
PFHpA		0.11 (0.10)	0.11 (0.08)	0.52 (0.73)	0.26 (0.23)	-	-
PFHpS		0.18 (0.05)	0.18 (0.05)	0.35 (0.19)	0.37 (0.19)	0.18 (0.06)	0.19 (0.08)
PFHxS		0.98 (0.63)	1.00 (0.80)	1.78 (1.62)	1.54 (1.21)	1.54 (1.46)	1.36 (1.49)
PFNA		0.80 (0.62)	0.80 (0.45)	0.69 (0.31)	0.68 (0.33)	0.49 (0.20)	0.50 (0.14)
PFOA		1.89 (1.34)	1.80 (0.96)	3.90 (3.25)	3.43 (2.21)	1.35 (0.39)	1.44 (0.52)
PFOS		3.37 (1.24)	4.00 (1.41)	12 (9)	14 (10)	3.18 (1.12)	3.84 (1.93)
PFPeS		0.073 (0.008)	0.074 (0.021)	0.080 (0.020)	0.077 (0.018)	0.07 (0.05)	0.07 (0.04)
Sex	n (%)
Female		14 (39)	14 (37)	13 (54)	28 (44)	12 (55)	52 (52)
Male		22 (61)	24 (63)	11 (46)	36 (56)	10 (45)	48 (48)
Age (years)	mean (sd)	12.03 (1.44)	11.94 (1.45)	17.10 (3.60)	17.23 (3.12)	19.18 (1.30)	19.27 (1.16)
Tanner Stage	n (%)
1 (Prepuberty)		0 (0)	0 (0)	0 (0)	1 (1.6)	-	-
2 (Early puberty)		22 (61)	24 (63)	1 (4.2)	1 (1.6)	-	-
3 (Mid-puberty)		9 (25)	11 (29)	2 (8.3)	7 (11)	-	-
4 (Late puberty)		5 (14)	3 (7.9)	4 (17)	13 (20)	-	-
5 (Postpuberty)		0 (0)	0 (0)	17 (71)	42 (66)	-	-
BMI (kg/m2)	mean (sd)	31.7 (6.0)	24.9 (3.8)	36.0 (7.1)	28.8 (4.7)	34.7 (5.0)	28.4 (4.0)
Parental Education	n (%)
< High School		17 (47)	11 (29)	18 (75)	33 (52)	8 (36)	15 (15)
> High School		8 (22)	10 (26)	2 (8.3)	9 (14)	12 (55)	72 (72)
High School Grad		11 (31)	17 (45)	4 (17)	22 (34)	2 (9.1)	13 (13)
Race/ethnicity	n (%)
White		-	-	-	-	4 (18)	40 (40)
Hispanic		36 (100)	38 (100)	24 (100)	64 (100)	17 (77)	53 (53)
Other		-	-	-	-	1 (4.5)	7 (7.0)

^1.Abbreviations: metabolic dysfunction-associated steatotic liver disease (MASLD), body mass index (BMI), per- and polyfluoroalkyl substances (PFAS), perfluorooctanoic acid (PFOA), perfluorooctane sulfonic acid (PFOS), perfluorononanoic acid (PFNA), perfluorodecanoic acid (PFDA), perfluoroheptane sulfonic acid (PFHpS), perfluoroheptanoic acid (PFHpA), perfluoroheptanoic sulfonic acid (PFPeS), and perfluorohexane sulfonic acid (PFHxS)

^2.Metabolic dysfunction-associated steatotic liver disease (MASLD) was defined as hepatic fat fraction > 5% on MRI plus ≥ 1 cardiometabolic risk factor.

^3.We did not detect PFHpA in Meta-AIR. Limits of detection (LODs) were 0.43 μg/L for PFOS, 0.05 μg/L for PFHpS, and 0.01 μg/L for PFOA, PFNA, PFDA, PFPeS, PFHxS, and PFHpA.

^4.Parental education was calculated using the average parental education level in a single household.

3.1. Associations Between PFAS and MASLD in Adolescents

Among adolescents in the SOLAR Phase II cohort, we found several significant associations between PFAS exposure and MASLD. Specifically, each doubling in plasma PFOA was associated with a 3.48-fold increase in odds of MASLD (95% CI: 1.08, 11.26; p = 0.03), and each doubling in plasma PFHpA was associated with a 1.73-fold increase in odds of MASLD (95% CI: 1.01, 2.96; p = 0.03) (Figure 1, Table 2).

Figure 1.

Cohort-Stratified Forest Plots of Main and Age-Interaction Odds Ratios for Log₂-Transformed PFAS Exposures and MASLD Risk in Meta-AIR, SOLAR Phase I, SOLAR Phase II and SOLAR Pooled

Table 2.

Association between basedline exposure to per- and polyfluoroalkyl substances (PFAS) and metabolic dysfunction-associated steatotic liver disease (MASLD) among Meta-AIR and SOLAR participants.

PFAS1,2,3,4	SOLAR Phase I (n = 74)			SOLAR Phase II (n = 88)			SOLAR Pooled (n = 162)			Meta-AIR (n = 122)
	OR [95% CI]	p value	q value	OR [95% CI]	p value	q value	OR [95% CI]	p value	q value	OR [95% CI]	p value	q value
PFDA	1.31 [0.65, 2.66]	0.47	0.97	1.67 [0.53, 5.26]	0.45	0.63	1.18 [0.69, 2.03]	0.56	0.94	0.77 [0.36, 1.62]	0.51	0.85
PFDA*Age	1.76 [0.79, 3.93]	0.21	0.55	1.39 [0.7, 2.74]	0.42	0.67	1.55 [0.89, 2.7]	0.14	0.29	1.06 [0.66, 1.72]	0.82	0.82
PFHpA	1.14 [0.76, 1.71]	0.55	0.97	1.73 [1.01, 2.96]	0.04	0.21	1.37 [1, 1.86]	0.05	0.27	-	-	-
PFHpA*Age	1.11 [0.88, 1.41]	0.38	0.55	1.05 [0.82, 1.36]	0.72	0.72	1.06 [0.92, 1.23]	0.44	0.48	-	-	-
PFHpS	0.73 [0.14, 3.8]	0.72	0.97	0.49 [0.08, 2.96]	0.47	0.63	0.62 [0.18, 2.1]	0.45	0.94	0.9 [0.32, 2.49]	0.85	0.85
PFHpS*Age	2.09 [0.57, 7.58]	0.28	0.55	1.2 [0.75, 1.93]	0.47	0.67	1.3 [0.89, 1.88]	0.19	0.29	0.77 [0.35, 1.67]	0.54	0.69
PFHxS	0.92 [0.45, 1.9]	0.84	0.97	1.38 [0.59, 3.27]	0.50	0.63	1.02 [0.58, 1.8]	0.94	0.94	1.3 [0.79, 2.13]	0.35	0.85
PFHxS*Age	0.98 [0.62, 1.55]	0.93	0.93	1.11 [0.77, 1.61]	0.60	0.67	1.09 [0.82, 1.45]	0.56	0.56	0.94 [0.67, 1.32]	0.74	0.82
PFNA	0.92 [0.35, 2.38]	0.87	0.97	2.1 [0.61, 7.26]	0.31	0.63	1.07 [0.52, 2.21]	0.86	0.94	1.2 [0.33, 4.41]	0.80	0.85
PFNA*Age	2.36 [1.02, 5.47]	0.04	0.35	1.35 [0.57, 3.19]	0.55	0.67	1.78 [1.01, 3.14]	0.04	0.21	0.42 [0.18, 0.98]	0.05	0.43
PFOA	1.46 [0.44, 4.84]	0.60	0.97	3.48 [1.08, 11.26]	0.04	0.21	2.69 [1.16, 6.26]	0.02	0.25	0.68 [0.24, 1.92]	0.50	0.85
PFOA*Age	2.29 [0.9, 5.85]	0.09	0.43	1.41 [0.89, 2.26]	0.17	0.67	1.45 [1.03, 2.06]	0.04	0.21	0.56 [0.26, 1.18]	0.14	0.43
PFOS	0.74 [0.22, 2.55]	0.65	0.97	0.52 [0.13, 2.12]	0.40	0.63	0.59 [0.23, 1.53]	0.29	0.94	0.79 [0.32, 1.97]	0.65	0.85
PFOS*Age	1.66 [0.67, 4.1]	0.29	0.55	1.15 [0.83, 1.58]	0.45	0.67	1.17 [0.92, 1.49]	0.23	0.29	0.73 [0.38, 1.42]	0.40	0.67
Total PFAS Burden	0.9 [0.47, 1.71]	0.76	0.97	1.37 [0.38, 4.92]	0.66	0.76	0.94 [0.43, 2.02]	0.87	0.94	0.87 [0.49, 1.55]	0.66	0.85
Total PFAS Burden*Age	1.31 [0.83, 2.07]	0.27	0.57	1.22 [0.85, 1.76]	0.32	0.67	1.28 [0.97, 1.68]	0.09	0.26	0.82 [0.55, 1.22]	0.36	0.67
PFCA Burden	1.06 [0.56, 2]	0.87	0.97	0.88 [0.49, 1.6]	0.70	0.76	1.03 [0.65, 1.63]	0.90	0.94	0.82 [0.46, 1.44]	0.50	0.85
PFCA Burden*Age	1.39 [0.88, 2.2]	0.17	0.55	1.2 [0.82, 1.77]	0.38	0.67	1.31 [0.95, 1.79]	0.10	0.26	0.72 [0.47, 1.09]	0.14	0.43
PFSA Burden	0.76 [0.4, 1.46]	0.43	1.00	0.91 [0.27, 3.04]	0.89	0.63	0.8 [0.38, 1.67]	0.56	0.94	0.9 [0.5, 1.61]	0.75	0.85
PFSA Burden*Age	1.07 [0.67, 1.71]	0.81	0.78	1.19 [0.84, 1.67]	0.37	0.67	1.2 [0.91, 1.59]	0.20	0.29	0.84 [0.55, 1.28]	0.45	0.67

^1.All logistic models were adjusted for age, sex, BMI, parental education, race/ethnicity (Meta-AIR only), Tanner stage (SOLAR only), and follow-up time (SOLAR Phase II only). Age of the participants was recentered at the mean of each study.

^2.PFAS concentrations were log2 transformed. The OR can be interpreted as the increased risk per doubling of PFAS.

^3.We used item response theory to quantify PFCA, PFSA and total PFAS burden. PFCA stands for Perfluoroalkyl Carboxylic Acids, while PFSA stands for Perfluoroalkyl Sulfonic Acids

^4.We applied Benjamini–Hochberg false discovery rate (FDR) correction across PFAS within each cohort.

In the SOLAR Phase I cohort, although no significant main effects were observed, we detected a significant interaction between PFNA and age, indicating that older adolescents exposed to higher PFNA levels had greater MASLD risk (OR per doubling PFNA × year of age = 2.36; 95% CI: 1.02, 5.47; p = 0.04). We did not observe any significant associations between PFAS mixture burden scores and MASLD risk in either adolescent cohort.

In the pooled SOLAR cohort, the results were similar to the results in each SOLAR cohort. PFOA was positively associated with MASLD, with each doubling linked to higher odds (OR 2.69, 95% CI 1.16–6.26, p=0.02), and the association increased with age (PFOA × Age OR 1.45, 95% CI 1.03–2.06, p=0.04). Age also modified the PFNA association (PFNA × Age OR 1.78, 95% CI 1.01–3.14, p=0.04). PFHpA showed a borderline positive main effect (OR 1.37, 95% CI 1.00–1.86, p=0.05). Other individual PFAS and the PFAS mixture scores were not associated with MASLD.

When examining genetic susceptibility, we observed significant interactions between PFHxS and PFOS with PNPLA3 genotype (GG vs CC/CG) in the SOLAR Phase I cohort. Adolescents with higher PFHxS exposure and the PNPLA3 GG risk genotype had the greatest odds of MASLD (OR 6.52, 95% CI 1.55–27.31; Figure 2, Table S3). We also saw age-dependent effects for PFHpS with PNPLA3 GG (OR 0.72, 95% CI 0.54–0.97, p=0.03) and for PFOS with PNPLA3 GG (OR 1.55, 95% CI 1.04–2.29, p=0.03), with borderline evidence for PFDA with PNPLA3 GG (OR 0.68, 95% CI 0.48–0.98, p=0.05) and a similar trend for PFHpA (OR 0.86, 95% CI 0.72–1.02, p=0.09). In pooled SOLAR analyses, PFHxS with PNPLA3 GG remained significant (OR 3.01, 95% CI 1.27–7.14, p=0.01) and the PFHxS with PNPLA3 GG effect varied with age (OR 1.62, 95% CI 1.06–2.49, p=0.03).

Figure 2.

Marginal predicted probability curves (±95 % CI) for MASLD over the range of PFHxS (log₂ scale), stratified by PNPLA3 genotype (CC/CG in red; GG in blue) in SOLAR Phase I.

3.2. Associations Between PFAS and MASLD in Young Adults

In the Meta-AIR young adult cohort, we did not observe significant associations between individual PFAS or PFAS mixture scores and MASLD in the main analysis (Figure 1, Table 2). We also did not observe any interaction between PFAS and PNPLA3 genotype (Table S3). Further, sensitivity analyses adjusting for lifestyle factors yielded similar null main effects (Table S4). However, we identified significant interactions between PFDA, PFHpS, and PFNA with cigarette smoking, suggesting that PFAS exposure may confer greater MASLD risk among smokers (Table S5). No other lifestyle factors, nor the composite lifestyle factor score, interacted with PFAS for MASLD.

3.3. Sex stratified analysis

Shown in Table S6, among males, PFAS effects strengthened with age in SOLAR. In pooled SOLAR males, PFHpA × Age, PFHpS × Age, PFOA × Age, and PFOS × Age were all positive and statistically significant, with odds ratios per one-year increase of 1.53, 2.47, 3.08, and 1.68, respectively. SOLAR Phase I males showed additional age modification effects, including total PFAS burden × Age and PFCA burden × Age (each positive and significant) and PFNA × Age, which were borderline significant. Main effects in males were largely null across cohorts, and Meta-AIR male estimates were mostly imprecise and null. Among females, the associations were mainly null. PFHpA showed a borderline positive main effect in SOLAR Phase I and a similar magnitude in the pooled SOLAR female sample. The only significant age modification in females was PFCA burden × Age in SOLAR Phase I. Other female associations were small and not statistically significant in SOLAR Phase II and Meta-AIR.

Discussion

Our multi-cohort analysis was conducted in adolescents and young adults with a higher risk of developing MASLD, spanning pre- to mid-pubertal children, older adolescents, and young adults. Our study showed that PFAS–MASLD associations varied by age and PNPLA3 genotype, with stronger associations observed in the SOLAR cohort. In SOLAR II, higher PFOA and PFHpA were linked to higher odds of MASLD per doubling of exposure, while in SOLAR I, we observed limited main effects but clear modification by age and PNPLA3, including stronger effects of PFHxS and PFOS among adolescents with the PNPLA3 GG genotype. These genotype-dependent patterns persisted when we pooled SOLAR participants together. In Meta-AIR, main effects were mostly null, yet interactions with cigarette smoking indicated higher MASLD risk among smokers with elevated PFDA, PFHpS, and PFNA. PFAS mixture scores were generally null across cohorts.

Most prior epidemiological studies linking PFOA to MASLD have focused on adults.^{29(pp2015–2017),51,52,53(pp2017–2018)} For example, several studies using NHANES data found that PFOA was associated with higher odds of fatty liver disease (OR 1.79; 95% CI: 1.07-2.99) in adults.^{26(pp2003-2014),27} Similarly, data from the Korean National Environmental Health Survey showed PFOA was associated with an increased risk of MASLD and liver fibrosis (OR 1.09-1.39).^54,55 PFAS exposure, including PFOA, has also been linked to MASLD severity through altered hepatic lipid metabolism.³⁷

However, evidence among children/adolescents and young adults remains limited and inconsistent. One NHANES-based study of liver FibroScan found no significant association between PFOA and MASLD in adolescents.⁵⁶ In contrast, A recent predictive equation risk score study of Chinese college students reported a significant association between PFOA and increased risk of MASLD.²³ Our study provides important evidence that PFOA is associated with MASLD risk in adolescents, but evidence in young adults was less clear. The possible explanation is that young adult participants were recruited in recent years compared to the adolescent cohort. Several PFAS have been banned or phased out of the market since 2000, resulting in decreasing levels of PFAS in blood.^57,58

Few studies have investigated the role of PFHpA in MASLD. A study in France found that PFHpA concentrations were higher in patients with grade 3 steatosis compared to grade 1.³¹ Likewise, older Chinese adults with detectable PFHpA levels had a 54% increased risk of MASLD.⁴⁹ Children in the multicenter Pediatric Research Network were found to have complex relationships of serum PFAS to liver histology.⁵⁹ A study in the US found that among patients with liver disease, PFHpA was associated with increased alanine aminotransferase (ALT) levels.⁶⁰ To our knowledge, ours is the first study to show an association between PFHpA and MASLD in the pediatric population.

Although the Age×PFAS interaction appeared positive in the adolescent SOLAR cohorts and negative in Meta-AIR young adults, we interpret these patterns as age-dependent susceptibility. First, puberty and late adolescence are periods of dynamic hepatic and metabolic remodeling.^12,13 Processes directly linked to PFAS-responsive pathways (PPARα, FXR) potentially sensitize adolescents to PFAS-related steatosis.^11,14,21,22 Second, SOLAR participants were enrolled in earlier calendar years with higher legacy PFAS levels, whereas Meta-AIR participants were enrolled later with lower background PFAS levels, which may reduce contrast and attenuate age gradients in young adults. Third, lifestyle factors (e.g., smoking) that cluster in young adulthood may modify PFAS effects (we observed significant PFAS×Smoking interactions in Meta-AIR), shifting the apparent age pattern. Finally, the age span differs by cohort (broad in SOLAR vs narrow in Meta-AIR), so an increasing PFAS effect across adolescence could naturally plateau or attenuate by early adulthood. Together, these points support the view that adolescence is a potentially sensitive window for PFAS-associated MASLD risk, with effects that decline as individuals transition into early adulthood.

Animal studies suggest that gestational and lactational exposure to PFOA, PFOS, and PFHxS disrupt liver proteome regulation, affecting pathways related to xenobiotic metabolism, lipid regulation, and liver damage.⁶¹ PFAS may alter hepatocyte function through negative regulation of gap junction assembly, glutathione biosynthesis, and pathways involved in inflammation, steatosis, and bile flow.²³ More specificially, PFOA could disrupt lipid transport and metabolism, even in peroxisome proliferator-activated receptor-alpha (PPARα)-knockout mice, implicating both PPARα-dependent and - independent pathways.^62,63 The mTOR signaling pathway, important in regulating cell growth and metabolism, is also disrupted by PFOA, suggesting shared molecular mechanisms involving cellular metabolism and energy homeostasis across PFAS compounds.^64,65

Very few studies evaluated the interaction between smoking and PFAS on MASLD or other liver outcomes. One study in NHANES found no evidence of interaction between PFAS and smoking.²⁷ Interestingly, resmetirom, the first and only approved medication in the US to treat non-alcoholic steatohepatitis (NASH), is a thyroid hormone receptor beta agonist.⁶⁶ Resmetirom can activate this receptor, which is expressed in the liver, and through this mechanism regulate lipid metabolism.⁶⁷ PFAS have been shown to disrupt thyroid function and induce thyroid cancer in human and animal studies.⁶⁸ Specifically, exposure to PFAS can lead to reduced thyroid hormones,⁶⁹ and hypothyroidism is a risk factor for MASLD.⁷⁰ One previous study showed that smoking modifies the associations between PFAS and thyroid hormones.⁷¹ Therefore, it is likely that PFAS can interact with smoking through their impact on thyroid hormones and more studies are needed to confirm our findings.

Furthermore, we found evidence that PFOS and PFHxS may interact with PNPLA3, leading to an increased risk of MASLD. However, it is important to note that these results should be interpreted with caution due to the small sample size. PNPLA3 rs738409 polymorphism is a well-established genetic risk factor for increased risk of various liver diseases, and in a meta-analysis of 21 studies with 14,266 participants, the PNPLA3 gene variant was associated with 4 times higher risk of developing MASLD.³⁸ We interpret the PNPLA3–PFAS findings as a modification of genetic susceptibility rather than proof that PFAS chemically interacts with the PNPLA3 gene itself. PNPLA3 encodes a protein that can regulate triglycerides in hepatocytes and retinyl esters in hepatic stellate cells.⁷² I148M variant can lead to a loss of function and thus can promote triglyceride accumulation in hepatocytes.⁷² A few studies have proposed the potential interaction between PNPLA3 and PFAS.^54,73 In contrast, the inverse or null associations observed for PFHpS, PFDA, and PFHpA among PNPLA3 GG carriers are likely due to small-sample variability and correlations within the PFAS mixture rather than reflecting true protective effects. Although not significant, total PFAS and PFSA burden showed positive effect estimates for the interaction with PNPLA3.

No previous studies have formally evaluated the interaction of PNPLA3 and PFAS exposure. It is thus worthwhile to explore whether PNPLA3 genotype can modify the PFAS-MASLD association or the role of PFAS in the expression of PNPLA3 to understand the risk of MASLD among people with PNPLA3 I148M variant and high PFAS exposure, which may not be as evident among people without this variant. A biological interaction remains plausible and deserves study. PNPLA3 I148M impairs triglyceride mobilization in hepatocytes and alters retinyl ester handling in stellate cells, and PFAS have been linked to lipid remodeling, endoplasmic reticulum stress, and PPAR-mediated lipid metabolism.^{9,21,22,24,37,65,72} These processes converge on hepatic triglyceride storage and fibrogenic signaling, which could amplify disease risk in carriers. Our data cannot distinguish statistical effect modification from a mechanistic interaction at the molecular level. Replication in larger cohorts and functional experiments using PNPLA3-edited hepatocytes or stellate cells exposed to PFAS would help determine whether PFAS modify PNPLA3 expression or activity. Future work should also evaluate interaction on the additive scale to quantify excess risk attributable to joint exposure and genotype.

Our findings provide limited support for sex as a modifier of the PFAS–MASLD association, but indicate that age-related susceptibility is stronger among males in adolescence. This pattern may reflect sex differences in pubertal/endocrine differences and hepatic lipid metabolism that shape PFAS toxicity pathways. However, the results should be interpreted with caution, given the relatively small sample size, and future studies should investigate potential sex-dependent effects.

Strengths and Limitations

This study has several notable strengths. We leveraged two well-characterized and diverse cohorts of adolescents and young adults, using imaging-based hepatic fat fraction assessment, the current gold standard, for accurate identification of MASLD. The use of high-resolution mass spectrometry enabled precise quantification of multiple PFAS compounds, including some that are rarely evaluated in pediatric studies. Our analysis also incorporated effect modification by both genetic susceptibility (PNPLA3 genotype) and key lifestyle factors, enhancing our ability to identify vulnerable subgroups. Finally, we applied a novel PFAS burden score using item response theory to address chemical mixture effects, a growing concern in environmental health research.

Nonetheless, several limitations merit consideration. First, the modest sample sizes limited our statistical power to detect more subtle associations and may have constrained the interpretation of some interaction effects. Second, while we evaluated multiple PFAS and stratified by several effect modifiers, we did not adjust for multiple comparisons. Our analyses were guided by a priori hypotheses informed by prior literature, and many findings were consistent with previously observed biological and epidemiological patterns, reducing the likelihood that results are due to chance alone. Still, we acknowledge the potential for type I error, and these findings should be interpreted in the context of replication and biological plausibility. Third, PFAS exposure was assessed at a single time point, which may not fully capture cumulative exposure, however, given the long elimination half-lives of these chemicals and their strong stability over time, a single baseline measurement serves as a reliable estimate of long-term exposure.⁷⁴ In addition, our analyses were cross-sectional, which limited our ability to establish a causal association between PFAS and MASLD. Reverse causation cannot be ruled out, and residual confounding by dietary factors, such as consumption of fish or ultra-processed foods that may also contribute to PFAS exposure, remains possible.

Conclusion

In this multi-cohort study, higher plasma levels of specific PFAS, especially PFOA and PFHpA, were linked to increased risk of MASLD in adolescents. Risk was further elevated in those with the PNPLA3 GG genotype and among young adults who smoked, highlighting both genetic and lifestyle factors as modifiers of PFAS-related liver effects.

These findings suggest that adolescence may represent a sensitive period for PFAS-associated liver effects. However, we cannot rule out that differences in SOLAR and Meta-AIR participant characteristics and PFAS exposure levels may contribute to differences in the associations between PFAS and MASLD. Future large, prospective cohort studies involving both adolescents and young adults should confirm our findings. Given the widespread and persistent nature of PFAS exposure, these results may have important public health and regulatory implications for reducing environmental risk factors contributing to MASLD in youth.

Highlights.

• Two high-risk cohorts studied: adolescents (SOLAR) and young adults (Meta-AIR)

• PFOA and PFHpA in adolescents were associated with higher MASLD odds

• PFAS were not associated with MASLD in young adults

• Age, PNPLA3, smoking may modify PFAS-MASLD associations

Abbreviations

ABI

Applied Biosystems, Inc.

BMI

body mass index

CHS

Children’s Health Study

CI

confidence interval

EPA

U.S. Environmental Protection Agency

FXR

farnesoid X receptor

HFF

hepatic fat fraction

HEI

Healthy Eating Index

IDEAL

Iterative Decomposition using Echo-Asymmetry in the Least–squares sense

IQR

interquartile range

IRB

institutional review board

IRT

item response theory

LC-HRMS

liquid chromatography–high-resolution mass spectrometry

LOD(s)

limit(s) of detection

MASLD

metabolic dysfunction-associated steatotic liver disease

Meta-AIR

Metabolic and Asthma Incidence Research

MRI

magnetic resonance imaging

NAFLD

nonalcoholic fatty liver disease

NHANES

National Health and Nutrition Examination Survey

NDSR

Nutrition Data System for Research

NIEHS

National Institute for Environmental Health Sciences

OR

odds ratio

PCR

polymerase chain reaction

PFAS

per- and polyfluoroalkyl substances

PFCA

perfluoroalkyl carboxylic acids

PFDA

perfluorodecanoic acid

PFOA

perfluorooctanoic acid

PFHpA

perfluoroheptanoic acid

PFHpS

perfluoroheptane sulfonic acid

PFHxS

perfluorohexane sulfonic acid

PFNA

perfluorononanoic acid

PFOS

perfluorooctane sulfonic acid

PNPLA3

patatin-like phospholipase domain-containing protein 3

PPARα

peroxisome proliferator-activated receptor-alpha

SD

standard deviation

SOLAR

Study of Latino Adolescents at Risk of Type 2 Diabetes

Data availability

Data used in the manuscript is confidential and the code is available upon request.