Associations of per- and polyfluoroalkyl substances and nonalcoholic fatty liver disease in the United States adult population, 2003–2018

Abstract

Background:

Nonalcoholic fatty liver disease (NAFLD) is the most common liver disorder worldwide and a leading cause of liver-related mortality. Prior studies have linked per- and polyfluoroalkyl substances (PFAS) exposure to liver dysfunction and alterations in metabolic pathways, but the extent of a PFAS-NAFLD relationship is unclear. Thus, the aim of the current study was to examine whether there were associations between PFAS exposures and NAFLD in the US adult population over a 16-year period.

Methods:

Data from 10,234 persons who participated in the National Health and Nutrition Examination Survey between 2003 and 2018 were analyzed. Odds ratios and 95% confidence intervals were calculated using multivariable logistic regression for the associations between PFAS and NAFLD, defined by the Hepatic Steatosis Index (NAFLD-HSI), the Fatty Liver Index (NAFLD-FLI), and by Transient Elastography with Controlled Attenuation Parameter (NAFLD-TE-CAP).

Results:

Overall, there was a significant inverse association between total PFAS and NAFLD-HSI (P-trend = 0.04). Significant inverse associations were also found between perfluorohexane sulfonic acid (PFHxS) and NAFLD-HSI (P-trend = 0.04), and NAFLD-FLI (P-trend = 0.03). Analysis by time period, 2003–2010 versus 2011–2018, found that while inverse associations were more apparent during the latter period when total PFAS (P-trend = 0.02), PFHxS (P-trend = 0.04), and perfluorooctanoic acid (PFOA) (P-trend = 0.03) were inversely associated with NAFLD-HSI and PFOA was inversely associated with NAFLD-FLI (P-trend = 0.05), there were no significant interaction effects. No significant associations between the PFAS and NAFLD-TE-CAP were found.

Conclusions:

The current study found no evidence of a positive association between the most common PFAS and NAFLD in the US population.

What this study adds

Nonalcoholic fatty liver disease (NAFLD) is one of the most common liver diseases in many countries, and the prevalence is increasing. While some studies have suggested that environmental chemicals, such as per- and polyfluoroalkyl substances (PFAS), could be associated with liver damage, a link to NAFLD has not been established. Our study examined whether there was an association between PFAS and NAFLD over a 16-year period in a representative sample of the US population. While an association between PFAS and liver enzyme levels was evident, the study found no evidence that PFAS was related to NAFLD.

Introduction

In recent years, nonalcoholic fatty liver disease (NAFLD) has emerged as a leading chronic liver disease worldwide.¹ Considered to be the hepatic manifestation of metabolic syndrome, NAFLD ranges in severity from simple steatosis to nonalcoholic steatohepatitis and is associated with a number of deleterious outcomes, including hepatic cirrhosis and hepatocellular carcinoma (HCC).^2,3 With an estimated global prevalence of 24%¹ and a US prevalence of 33%,⁴ projections suggest that the prevalence of NAFLD will continue to increase. While NAFLD is known to be closely related to other metabolic conditions such as obesity, insulin resistance, dyslipidemia, and diabetes,^5,6 other factors related to increased risk of NAFLD are not as well understood. It has been suggested, however, that environmental contaminants, could be related to NAFLD risk.^7,8

Per- and polyfluoroalkyl substances (PFAS) are a group of man-made organic compounds that, since the 1950s, have been widely used in multiple products, including lubricants, surfactants, food packaging, paints, furniture, carpet, and fire-extinguishing foams.⁹ The primary sources of PFAS exposure in the general population, however, are through diet, and contaminated water, air, and dust.^7,10 Though the production of some PFAS has been terminated over the years, there have been heightened concerns of health consequences in humans given its persistence in the environment, resistance to breakdown, and slow elimination from the human body.^10,11 Four PFAS, perfluorooctane sulfonic acid (PFOS), perfluorooctanoic acid (PFOA), perfluorononanoic acid (PFNA), and perfluorohexane sulfonic acid (PFHxS) have the highest concentrations in human sera and have been examined for associations with various outcomes, including cancer, cardiovascular disease, thyroid dysfunction, ovarian disorders, and mortality.^9,12–14 In addition, persistent exposure to PFAS has been reported to alter key liver functions, increase the risk of liver toxicity, promote liver damage,^7,15 and increase the risk of HCC.^16,17 Studies have also explored PFAS exposure and key NAFLD-related hepatic metabolic pathways,⁷ as well as biomarkers of liver function and liver disease.^15,18–20 Accumulated evidence from these studies has suggested a possible link between PFAS and NAFLD.

Several prior studies of PFAS and NAFLD have been reported,^21–23 but the findings have not been consistent. As a result, the current investigation sought to examine the association between PFAS exposure and NAFLD over a 16-year period in the general adult US population.

Materials and methods

The National Health and Nutritional Examination Survey (NHANES) is a nationally representative, cross-sectional survey that is continuously conducted in 2-year cycles to evaluate the health and nutritional status of the noninstitutionalized US population.^24–26 Survey participants consented to provide biospecimens, take part in a medical examination, and have a household interview. A detailed description of the NHANES protocol is available elsewhere.²⁷

In each NHANES cycle, data on age, sex, race/ethnicity, education level, alcohol consumption, and cigarette smoking are self-reported using detailed questionnaires. Height, weight, and waist circumference in centimeters are measured during the medical examination. Body mass index (BMI) is calculated as weight in kilograms divided by height in meters squared (kg/m²). For the current analysis, obesity was defined as a BMI ≥30 kg/m² and diabetes was defined as either a self-report of a doctor’s diagnosis or having a hemoglobin A1c level of ≥6.5 %.

For the current study, we combined survey cycles of NHANES for the years 2003–2018. The 2013–2014 survey cycle was excluded because surplus serum from only a subset of eligible participants was analyzed. Figure 1 shows the total number of eligible participants, and the participants included in our analytical sample. Of the 70,137 total NHANES participants in the combined cycles, 35,169 participants met the eligibility criteria. Exclusion criteria included: less than age 20 years (n = 31,116), positive for hepatitis B or hepatitis C virus (n = 810), reported heavy alcohol consumption, defined as more than one drink a day for women or more than two drinks a day for men (n = 2166), or pregnant at the time of examination (n = 876). Participants who were not in the random subsample selected for PFAS measurements were also excluded (n = 24,935). The final analytic sample included a total of 10,234 participants.

Figure 1.

Participant flowchart.

In each 2-year cycle, NHANES measures a minimum of 12 PFAS in a one-third randomly selected subset of participants. The current study focused on the four most common PFAS: PFOS, PFOA, PFNA, and PFHxS.²⁸ Beginning in the 2013–2014 cycle, linear and branched isomers of PFOS and PFOA were measured rather than total PFOS and total PFOA. The calculated sum of these isomers is equivalent to the total levels reported in previous NHANES cycles. PFAS were measured in serum by solid-phase extraction-high performance liquid chromatography-turbo ion spray-tandem mass spectrometry (SPE-HPLC-TIS-MS/MS), a method which enabled rapid determination with a limit of detection (LOD) of 0.1 ng/mL and a linear range of 0.01 to 20–50 ng/mL. Values less than the LOD were reported as the LOD divided by the square root of two. A detailed description of the laboratory procedures has been published previously.²⁹

NAFLD was defined using two calculated indices, the fatty liver index (FLI), and the hepatic steatosis index (HSI), and by the transient elastography controlled attenuation parameter score (TE-CAP) (FibroScan, Echosens, Paris, FR). FLI and HSI are noninvasive indices that have been validated for detecting NAFLD.^5,30,31 HSI is calculated as: 8 × (alanine aminotransferase/aspartate aminotransferase [ALT/AST] ratio) + BMI (+2, if female; +2, if diabetes). An HSI >36 is considered evidence of NAFLD.² FLI was calculated as: (exp [0.953 × ln (triglycerides) + 0.139 × BMI + 0.718 × ln (GGT) + 0.053 × waist circumference-15.745])/(1 + exp [0.953 × ln (triglycerides) + 0.139 × BMI + 0.718 × ln (GGT) + 0.053 × waist circumference-15.745]) × 100. An FLI ≥60 is considered evidence of NAFLD.³¹ For ease of discussion, NAFLD by HSI or by FLI will be referred to as NAFLD-HSI and NAFLD-FLI, respectively.

Determination of NAFLD by transient elastography was performed in the NHANES 2017–2018 cycle using a FibroScan model 502 V2 touch equipped with a medium or extra-large probe. A detailed description of procedures and guidelines is described in the NHANES official manual.³² For the current investigation, two TE-CAP cut-points were used to define NAFLD, TE-CAP ≥263 dB/m and TE-CAP ≥285 dB/m. As the TE-CAP analysis of NAFLD was restricted to the 2017–2018 cycle, only 1395 participants could be included for the examination of TE-CAP and PFAS.

Selected liver function biomarkers measured in blood by NHANES and used in the calculations of HSI or FLI include AST (IU/L), ALT (IU/L), and gamma-glutamyl transferase (GGT) (U/L). The following sex-specific cut-points for these biomarkers were considered to be elevated using the NHANES 2017–2018 criteria. For women, high AST was ≥31 U/L, high ALT was ≥31 U/L, and high GGT was ≥33 U/L. For men, high AST was ≥37 U/L, high ALT was ≥40 U/L, and high GGT was ≥51 U/L.

Statistical analysis

The multivariable logistic analysis was performed to estimate the odds ratios with 95% confidence interval for the associations between each PFAS and NAFLD, the liver biomarkers, and BMI. The analyses account for the complex sample design, including the sample weighting of the NHANES for the subsample in which PFAS was measured. The analyses to test the association between PFAS and TE-CAP were restricted to participants with a complete examination in the 2017–2018 cycle (n = 1385). PFAS was categorized into quartiles that were specific to each 2-year cycle, with the first quartile used as the referent category. In a sub-analysis, we stratified the study into an earlier period (2003–2010) and a later period (2011–2018) to determine whether there were differences between PFAS and NAFLD during those periods. Multiplicative interaction terms were used to evaluate significant interactions between the two time periods.

All models were adjusted for age (20–39, 40–59, 60–79, ≥80 years), sex (male or female), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and others), education (less than high school, high school or general education diploma some college or associate degree, college graduate or higher), smoking status (never, former, or current), and alcohol consumption status (never, former, or current). BMI (kg/m² <18.5, 18.5–24.9, 25.0–29.9, and ≥30) was not included in the HSI and FLI models, and diabetes (yes, no) was not included in the HSI models as these variables were used in the calculation of those indices. To test for trend across quartiles, we included the adjustment variables in multivariable logistic regression models and an ordinal variable with four levels (with values 0, 1, 2, and 3) for the four quartiles of total PFAS and each individual PFAS from the lowest to the highest quartiles. The SAS software (version 9.4, SAS Institute Inc., Cary, NC) was used for all analyses with P ≤ 0.05 considered statistically significant.

Results

The characteristics of the study population by quartiles of PFAS are presented in Table 1. Univariate analyses found that higher levels of PFAS were evident among persons aged 40–79 years, men, non-Hispanic White persons, those with a BMI ≥25.0, persons with at least a college education, persons with diabetes, former smokers, and current drinkers.

Table 1.

Characteristics of study population by categories of total per- and polyfluoroalkyl substances levels in NHANES 2003–2018 (n = 10,234)^a

	Total PFAS
	Q1	Q2	Q3	Q4
	n = 2557	n = 2559	n = 2561	n = 2557
	%	%	%	%	P trendb
Age (years)					<0.0001
20–39	49.7	39.8	32.7	21.9
40–59	36.0	38.8	37.5	36.3
60–79	11.4	18.4	24.6	33.2
≥80	2.9	2.9	5.2	8.6
Sex					<0.0001
Female	77.7	55.5	40.2	36.5
Male	22.3	44.5	59.8	63.5
Race/ethnicity					<0.0001
Non-Hispanic White	56.2	67.7	72.5	72.6
Non-Hispanic Black	11.7	9.4	9.8	12.2
Hispanic	23.2	15.8	11.5	7.1
Others	8.9	7.1	6.2	8.1
Body mass index (kg/m2)					<0.0001
<18.5	3.8	2.6	1.7	2.4
18.5–24.9	27.9	29.2	28.5	26.4
25.0–29.9	28.1	31.2	35.1	35.8
≥30	40.2	37.0	34.7	35.4
Education level					<0.0001
<High school	19.7	15.7	13.6	15.9
High school/GED	22.2	24.7	21.2	24.3
Some college	32.0	30.7	33.4	29.3
≥College	26.1	28.9	31.8	30.5
Diabetes mellitus					<0.0001
Yes	10.7	10.9	11.4	14.9
No	89.3	89.1	88.6	85.1
Smoking					0.01
Never	61.9	55.8	56.5	57.1
Former	17.8	24.5	24.9	28.6
Current	20.3	19.7	18.6	14.3
Alcohol					0.04
Never	13.5	8.3	10.5	9.7
Former	17.7	19.1	15.3	17.9
Current	68.8	72.6	74.2	72.4

^aData presented as row percentages of participants unless otherwise indicated. Total PFAS levels were used for Quartiles (Q); Quartile-1 < 8.74 ng/mL, Quartile-2 8.74–15.5 ng/mL, Quartile-3 15.5–25.8 ng/mL and Quartile-4 > 25.8 ng/mL.

^bP values for trend were obtained using multivariable logistic regression models.

GED indicates general education diploma; PFAS, polyfluoroalkyl substances.

Table 2 displays the prevalence of NAFLD by FLI, HSI, TE-CAP ≥263, and TE-CAP ≥285 and shows the univariate associations of PFAS with each determination of NAFLD. The highest overall prevalence of NAFLD was estimated by HSI (56%), followed by TE-CAP ≥263 (51%), FLI (45%), and finally TE-CAP ≥285 (38%). The examination of each determination of NAFLD by total PFAS quartile found a significant positive univariate association only between NAFLD-FLI and PFAS (P-trend < 0.0001).

Table 2.

Prevalence of nonalcoholic fatty liver disease in NHANES 2003-12, 2014–18 by method of determination and quartile of total per- and polyfluoroalkyl substances (n = 10,234)^a

		Total PFAS
		Q1	Q2	Q3	Q4
	Prevalence	n = 2557	n = 2559	n = 2561	n = 2557
	%	%	%	%	%	P-trendb
Hepatic Steatosis Index (≥36)	56	58.1	55	54.2	54.8	0.09
Fatty Liver Index (≥60)	45	42.9	43.6	45	49.8	<0.0001
TE-CAP (≥263 dB/m)c	51	47.1	46.6	47.8	51.5	0.25
TE-CAP (≥285 dB/m)c	38	32.9	38.5	34.2	40.1	0.24

^aTotal PFAS levels by Quartiles (Q); Quartile-1: <8.74 ng/mL, Quartile-2: 8.74–15.5 ng/mL, Quartile-3: 15.5–25.8 ng/mL, and Quartile-4: > 25.8 ng/mL.

^bP values for trend were obtained using multivariable logistic regression models.

^cTransient elastography controlled attenuation parameter (TE-CAP) measured only in the 2017–2018 cycle, n = 1385. Q1 (n = 345), Q2 (n = 353), Q3 (n = 347), and Q4 (n = 340).

PFAS indicates polyfluoroalkyl substances; TE-CAP, Transient Elastography with Controlled Attenuation Parameter.

Figure 2 shows the prevalence of NAFLD by HSI and FLI from 2003 to 2018 and NAFLD by TE-CAP in 2017–2018. The prevalence of NAFLD-HSI increased from 54% in the 2003–2004 cycle to 61% in the 2017–2018 cycle, while the prevalence of NAFLD-FLI increased from 43% to 51%. During all cycles, there was a higher estimated prevalence of NAFLD as determined by HSI than by FLI. During the 2011–2012 cycle, there was a decrease in NAFLD prevalence by both HSI (53%) and FLI (44%) compared to the previous NHANES cycle (HSI 58% and FLI 48%). The prevalence of NAFLD by both indices increased again, however, in the subsequent cycle. The univariate associations between each of the NAFLD determinations and population covariates are shown in Supplemental Table 1; http://links.lww.com/EE/A256.

Figure 2.

Prevalence of nonalcoholic fatty liver disease (NAFLD) in the study population from 2003 to 2018. In the 2003–2004 cycle, FLI (43%), HSI (54%); 2017–2018, FLI (51%), HSI (61%), TE-CAP ≥263 dB/m (51%), TE-CAP ≥285 dB/m (38%). FLI, fatty liver index; HSI, hepatic steatosis index; TE-CAP, transient elastography with controlled attenuation parameter.

Figure 3 shows the overall concentrations of the PFAS over time. There was a declining trend in all PFAS from 2003 to 2018, most notably in the concentration of PFOS. Between the 2003–2004 cycle and the 2017–2018 cycle, the mean concentration of total PFAS declined from 35.4 to 11.3 ng/mL. PFOS declined from 26.1 to 7.2 ng/mL, PFOA declined from 4.5 to 1.7 ng/mL, PFNA declined from 1.3 to 0.6 ng/mL, and PFHxS declined from 2.6 to 1.7 ng/mL.

Figure 3.

Temporal trend of serum PFAS concentrations (ng/mL) in the study population from 2003 to 2018 (excluding 2013–2014 cycle). Mean serum concentrations (ng/mL) for 2003–2004 and 2017–2018 were, total PFAS 35.4 and 11.3; PFOS 26.1 and 7.2; PFOA 4.5 and 1.7; PFNA 1.3 and 0.6; PFHxS 2.6 and 1.7. PFAS, polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonic acid.

The adjusted odds ratios for associations between PFAS and NAFLD by each definition are presented in Table 3. The analyses of NAFLD-HSI found significant inverse associations with total PFAS (P-trend = 0.04) and PFHxS (P-trend = 0.04). The analyses of NAFLD-FLI also found a significant inverse association with PFHxS (P-trend = 0.03). NAFLD by TE-CAP ≥263 dB/m showed an inverse association with PFHxS (P-trend = 0.06), but the trend did not attain statistical significance. As the TE-CAP analyses were only based on the 2017–2018 participants, the analyses of NAFLD-HSI and NAFLD-FLI restricted to the 2017–2018 participants were also examined. As shown in Supplemental Table 2; http://links.lww.com/EE/A256, PFHxS was inversely associated with NAFLD-HSI (P-trend = 0.04), but not with NAFLD-FLI.

Table 3.

Adjusted odds ratios for associations between per- and polyfluoroalkyl substances levels and the presence of nonalcoholic fatty liver disease by each definition in NHANES 2003–2018^a

	Hepatic Steatosis Index	Fatty Liver Index	TE-CAP ≥263 dB/m	TE-CAP ≥285 dB/m
	n = 10,234	n = 10,234	n = 1385	n = 1385
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Total PFAS
Q1	1	1	1	1
Q2	0.87 (0.71, 1.06)	0.83 (0.69, 0.98)	0.71 (0.45, 1.11)	0.93 (0.55, 1.56)
Q3	0.83 (0.69, 0.98)	0.78 (0.65, 0.94)	0.63 (0.35, 1.13)	0.62 (0.31, 1.23)
Q4	0.84 (0.70, 0.99)	0.86 (0.72, 1.03)	0.68 (0.37, 1.24)	0.77 (0.37, 1.59)
P-trend	0.04	0.12	0.23	0.30
PFOS
Q1	1	1	1	1
Q2	0.84 (0.69, 1.01)	0.84 (0.70, 1.01)	0.74 (0.43, 1.27)	0.87 (0.50, 1.48)
Q3	0.93 (0.79, 1.09)	0.88 (0.74, 1.04)	0.76 (0.45, 1.29)	0.82 (0.46, 1.45)
Q4	0.82 (0.69, 0.97)	0.86 (0.71, 1.04)	0.61 (0.35, 1.06)	0.62 (0.32, 1.19)
P-trend	0.07	0.19	0.13	0.17
PFOA
Q1	1	1	1	1
Q2	1.06 (0.91, 1.24)	0.99 (0.84, 1.17)	0.76 (0.43, 1.35)	0.83 (0.46, 1.49)
Q3	0.91 (0.78, 1.07)	0.95 (0.79, 1.15)	0.77 (0.42, 1.38)	0.83 (0.44, 1.56)
Q4	0.97 (0.79, 1.17)	0.97 (0.79, 1.18)	0.74 (0.47, 1.16)	0.78 (0.41, 1.50)
P-trend	0.36	0.64	0.29	0.49
PFNA
Q1	1	1	1	1
Q2	0.95 (0.79, 1.14)	0.90 (0.76, 1.07)	0.58 (0.33, 1.02)	0.66 (0.34, 1.28)
Q3	0.99 (0.84, 1.19)	0.92 (0.77, 1.10)	0.52 (0.35, 0.76)	0.65 (0.37, 1.11)
Q4	1.03 (0.87, 1.22)	0.96 (0.81, 1.13)	0.73 (0.44, 1.22)	0.83 (0.43, 1.59)
P-trend	0.60	0.69	0.22	0.58
PFHxS
Q1	1	1	1	1
Q2	0.97 (0.81, 1.17)	0.98 (0.83, 1.16)	0.92 (0.49, 1.69)	1.19 (0.58, 2.43)
Q3	0.92 (0.77, 1.09)	0.92 (0.75, 1.13)	0.65 (0.35, 1.20)	0.68 (0.36, 1.31)
Q4	0.86 (0.73, 1.01)	0.84 (0.70, 1.01)	0.66 (0.39, 1.12)	0.81 (0.41, 1.62)
P-trend	0.04	0.03	0.06	0.22

^aAdjusted covariates in model: age, sex, race–ethnicity, education levels, smoking status, alcohol drinking status.

CI indicates confidence interval; OR, odds ratio; PFAS, polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonic acid; TE-CAP, Transient Elastography with Controlled Attenuation Parameter.

Table 4 displays the adjusted odds ratios for associations between PFAS and NAFLD stratified by time period (2003–2010 and 2011–2018). During the 2003–2010 period, there were no significant associations between PFAS and either NAFLD-HSI or NAFLD-FLI. In contrast, during the 2011–2018 period, total PFAS was inversely associated with NAFLD-HSI (P-trend = 0.02), PFHxS was inversely associated with NAFLD-HSI (P-trend = 0.04), and PFOA was inversely associated with both NAFLD-HSI (P-trend = 0.03) and NAFLD-FLI (P-trend = 0.05). Despite the apparent differences by time period, the interaction analyses found little support for significant differences between the early and later time periods. The only significant interaction effects were for NAFLD-FLI associations in which the period-specific trends themselves were not significant (PFOS, P-interaction = 0.05; PFNA, P-interaction = 0.03; PFHxS, P-interaction = 0.04).

Table 4.

Adjusted odds ratios for associations between polyfluoroalkyl substances levels and the presence of nonalcoholic fatty liver disease in NHANES 2003–2008 and 2011–2018^a

	Hepatic Steatosis Index			Fatty Liver Index
	2003–2010	2011–2018		2003–2010	2011–2018
	n = 6027	n = 4207		n = 6027	n = 4207
	OR (95% CI)	OR (95% CI)	P-int	OR (95% CI)	OR (95% CI)	P-int
Total PFAS			0.33			0.09
Q1	1	1		1	1
Q2	0.99 (0.83, 1.18)	0.73 (0.50, 1.07)		0.97 (0.80, 1.17)	0.69 (0.52, 0.92)
Q3	1.01 (0.83, 1.23)	0.63 (0.47, 0.86)		0.94 (0.72, 1.25)	0.63 (0.46, 0.85)
Q4	0.95 (0.78, 1.16)	0.70 (0.51, 0.97)		0.97 (0.81, 1.18)	0.75 (0.53, 1.05)
P-trend	0.69	0.02		0.75	0.10
PFOS			0.12			0.05
Q1	1	1		1	1
Q2	0.91 (0.70, 1.19)	0.77 (0.56, 1.06)		0.92 (0.74, 1.16)	0.77 (0.57, 1.04)
Q3	1.01 (0.86, 1.28)	0.80 (0.60, 1.05)		1.03 (0.81, 1.30)	0.74 (0.58, 0.95)
Q4	0.86 (0.70, 1.05)	0.78 (0.58, 1.07)		0.92 (0.75, 1.12)	0.81 (0.56, 1.17)
P-trend	0.32	0.14		0.65	0.25
PFOA			0.76			0.48
Q1	1	1		1	1
Q2	1.18 (1.00, 1.39)	0.93 (0.71, 1.22)		1.22 (1.02, 1.46)	0.78 (0.58, 1.05)
Q3	1.14 (0.96, 1.37)	0.68 (0.50, 0.92)		1.22 (0.98, 1.52)	0.70 (0.50, 0.97)
Q4	1.16 (0.95, 1.42)	0.75 (0.52, 1.06)		1.26 (1.00, 1.57)	0.70 (0.49, 1.00)
P-trend	0.25	0.03		0.06	0.05
PFNA			0.19			0.03
Q1	1	1		1	1
Q2	1.15 (0.95, 1.40)	0.77 (0.56, 1.06)		1.10 (0.92, 1.33)	0.74 (0.56, 0.99)
Q3	1.05 (0.84, 1.31)	0.95 (0.71, 1.26)		0.98 (0.78, 1.23)	0.90 (0.67, 1.19)
Q4	1.16 (0.96, 1.41)	0.90 (0.65, 1.24)		1.05 (0.87, 1.27)	0.90 (0.66, 1.22)
P-trend	0.26	0.79		0.92	0.76
PFHxS			0.21			0.04
Q1	1	1		1	1
Q2	0.98 (0.78, 1.22)	0.96 (0.70, 1.31)		1.02 (0.84, 1.25)	0.94 (0.70, 1.26)
Q3	1.08 (0.87, 1.35)	0.72 (0.54, 0.97)		1.04 (0.84, 1.29)	0.79 (0.55, 1.14)
Q4	0.90 (0.74, 1.08)	0.80 (0.61, 1.06)		0.86 (0.70, 1.06)	0.82 (0.60, 1.12)
P-trend	0.42	0.04		0.16	0.12

^aAdjusted covariates includes age, sex, race–ethnicity, education levels, smoking status, alcohol drinking status.

^bP-int: P values for interaction.

CI indicates confidence interval; OR, odds ratio; PFAS, polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonic acid.

Because BMI and liver function biomarkers are used in the calculations of FLI and HSI, we also examined their relationships with PFAS. The proportion of NAFLD participants with elevated biomarkers and obesity is shown in Supplemental Table 3; http://links.lww.com/EE/A256. The adjusted odds ratios are shown in Table 5. BMI was significantly inversely associated with total PFAS (P-trend = <0.0001), PFOS (P-trend = 0.01), PFOA (P-trend = 0.03), PFNA (P-trend = 0.05) and PFHxS (P-trend = 0.01). ALT was significantly positively associated with total PFAS (P-trend = 0.03), PFOA (P-trend = 0.02), and PFNA (P-trend = 0.03), while GGT was significantly positively associated with PFOA (P-trend = 0.05). There were no significant associations between AST and any PFAS.

Table 5.

Adjusted odds ratios for associations between per-and polyfluoroalkyl substances levels and the presence of abnormal liver biomarkers and body mass index in NHANES 2003–2018^a

	Obesity	ALT	AST	GGT
	n = 10077	n = 9030	n = 9022	n = 9097
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Total PFAS
Q1	1	1	1	1
Q2	0.85 (0.72, 0.99)	1.21 (0.91, 1.59)	1.15 (0.85, 1.57)	1.12 (0.87, 1.45)
Q3	0.76 (0.64, 0.90)	1.42 (1.04, 1.94)	1.29 (0.91, 1.82)	1.25 (0.95, 1.65)
Q4	0.75 (0.63, 0.90)	1.44 (1.04, 2.00)	1.16 (0.66, 1.58)	1.14 (0.83, 1.56)
P-trend	<0.0001	0.03	0.27	0.34
PFOS
Q1	1	1	1	1
Q2	0.81 (0.68, 0.96)	1.17 (0.92, 1.49)	1.21 (0.89, 1.64)	1.09 (0.86, 1.38)
Q3	0.83 (0.71, 0.99)	1.21 (0.88, 1.67)	1.19 (0.86, 1.66)	1.08 (0.81, 1.44)
Q4	0.76 (0.64, 0.90)	1.46 (1.00, 1.96)	1.12 (0.81, 1.56)	1.00 (0.74, 1.35)
P-trend	0.01	0.07	0.51	0.99
PFOA
Q1	1	1	1	1
Q2	0.98 (0.84, 1.14)	1.25 (0.91, 1.72)	1.38 (1.00, 1.90)	1.34 (1.02, 1.76)
Q3	0.86 (0.72, 1.02)	1.44 (1.09, 1.90)	1.33 (0.97, 1.84)	1.26 (0.96, 1.62)
Q4	0.85 (0.72, 1.01)	1.43 (1.01, 2.02)	1.39 (1.01, 1.92)	1.39 (1.02, 1.92)
P-trend	0.03	0.02	0.08	0.05
PFNA
Q1	1	1	1	1
Q2	0.87 (0.74, 1.02)	1.18 (0.90, 1.53)	1.19 (0.84, 1.70)	0.89 (0.68, 1.16)
Q3	0.85 (0.72, 1.02)	1.15 (0.85, 1.55)	1.01 (0.74, 1.38)	1.00 (0.77, 1.31)
Q4	0.85 (0.72, 0.99)	1.45 (1.07, 1.96)	1.25 (0.92, 1.70)	1.27 (0.95, 1.70)
P-trend	0.05	0.03	0.30	0.08
PFHxS
Q1	1	1	1	1
Q2	0.96 (0.80, 1.15)	0.97 (0.71, 1.33)	1.12 (0.77, 1.62)	1.03 (0.79, 1.33)
Q3	0.81 (0.68, 0.96)	1.12 (0.85, 1.46)	1.03 (0.73, 1.44)	1.05 (0.81, 1.66)
Q4	0.82 (0.69, 0.98)	1.27 (0.93, 1.74)	1.31 (0.94, 1.63)	0.94 (0.69, 1.25)
P-trend	0.01	0.07	0.17	0.67

^aAdjusted covariates in model age, sex, race–ethnicity, education levels, BMI (excluded when obesity is outcome), smoking status, alcohol drinking status.

ALT indicates alanine aminotransferase; AST, aspartate aminotransferase; CI, confidence interval; GGT, gamma-glutamyl transferase; OR, odds ratio; PFAS, polyfluoroalkyl substances; PFHxS, perfluorohexane sulfonic acid; PFNA, perfluorononanoic acid; PFOA, perfluorooctanoic acid; PFOS, perfluorooctane sulfonic acid.

Discussion

In this study of 10,234 participants from NHANES 2003–2018, the overall analysis found that total PFAS was inversely associated with NAFLD-HSI, and PFHxS was inversely associated with both NAFLD-HSI and NAFLD-FLI. NAFLD determined by TE-CAP was not associated with any PFAS.

The relationship between PFAS and NAFLD has been examined in several prior studies that used NHANES data. In a study of NHANES 1999–2014 participants that only examined total PFAS, no association was found with NAFLD (OR = 0.99; 95% CI = 0.90, 1.08).²³ A second study examined a subset of the NHANES 2005–2018 participants and reported a significant positive association between NAFLD and total PFAS among women but not among men.²² A third study used only one NHANES cycle, NHANES 2017–2018, to examine 1135 participants and reported a significant positive trend (P = 0.03) only between NAFLD and PFHxS.²¹ Reasons that the results of these prior studies vary among themselves, and with the current study, are likely related to several factors. The studies had different sample sizes, employed different participant exclusions, and used different definitions of NAFLD. In addition, some included the NHANES cycle that was restricted to a subset of the participants, and some combined measurements of urinary and serum PFAS levels in their analyses. As reported by a number of publications, including a recent one from NHANES, PFAS levels in urine are considerably lower than they are in serum, so including both urinary and serum levels in a single analysis is unlikely to provide accurate results.³³

Why the PFAS-NAFLD association would appear to differ by time period is not clear but could be related to the inverse association between PFAS and obesity shown in the current study, as BMI is a variable in the calculation of both FLI and HSI, and the prevalence of obesity in NHANES rose from 32.2% in 2003–2004 to 42.2% in 2017–2018. In contrast, the PFAS-NAFLD analyses, using either TE-CAP ≥263 dB/m or TE-CAP ≥285 dB/m as the determinant of NAFLD, found no significant relationships. Prior studies of the PFAS-BMI relationship have produced inconsistent results, perhaps because many of the studies have been conducted in specialized populations, such as mother-child pairs, pregnant women, or persons with preexisting medical conditions.³⁴ At least two prior studies, however, have used NHANES data to examine PFAS and variables associated with BMI.^35,36 In an analysis of NHANES 2003–2004 data, few associations between PFAS and body weight were found, with the exception of an inverse association between PFOS and body weight among males.³⁵ Similarly, in an analysis of NHANES 2007–2014 data, while waist circumference and PFNA were positively associated, waist circumference and PFOA were inversely associated.³⁶

The study’s findings of positive associations between the PFAS and ALT levels were consistent with prior NHANES studies and with a previously published meta-analysis of 24 epidemiologic studies which found ALT was positively associated with PFOA (P < 0.001), PFOS (P < 0.001) and PFNA (P = 0.02).¹⁶ Similar to the current study, the meta-analysis also found that GGT was positively associated PFOA (P < 0.001).

In agreement with previous reports, the current analysis also found an increasing prevalence of NAFLD in the US adult population between 2003 and 2018 by both HSI (54%–61%) and FLI (43%–51%). Different estimates of NAFLD prevalence by FLI and HSI were not unanticipated as prevalence estimates by HSI have been routinely higher than those by FLI.^37–39 The current study’s results are consistent with previous studies that reported a growing prevalence of NAFLD in the United States and worldwide.^1,6,40–42 On the global scale, a recent meta-analysis found the prevalence of NAFLD was 26% in studies published in 2005 or earlier, but 38% in studies published in 2016 or later.⁴⁰ To date, TE-CAP is one of the most sensitive and noninvasive methods for identifying NAFLD,⁴³ but cut-points differ by study. The current study used two cut-points for identifying NAFLD in 2017–2018. Our study and that of Le et al.³⁹ reported 38% NAFLD prevalence by TE-CAP ≥285 dB/m, while our finding of 51% NAFLD prevalence by TE-CAP ≥263 dB/m was similar to that of Zhang et al (56.7%),⁴³ and Jones et al (47.6%)³⁷ which used TE-CAP ≥248 dB/m and ≥260 dB/m as cut-points. As shown in the current study, TE-CAP ≥263 dB/m and NAFLD-FLI found identical estimates of NAFLD, 51%, in 2017–2018.

Over the past two decades, the US Environmental Protection Agency and several large-scale manufacturers and importers of PFAS have taken action to terminate or limit the production and emission of some PFAS in the United States.⁴⁴ As expected, these actions were evident in our findings which showed a significant downward trend of serum PFAS concentrations in the general US population, most notably for PFOS and PFOA. These trends are also seen in several European countries that have also been regulating and monitoring the use of various PFAS.⁴⁵

The current study has several key strengths, including the use of multiple years of data from a large, nationally representative sample, the use of high-quality data to derive the fatty liver indices, and the generalizability of the results to the overall US population. The study also had several limitations. NAFLD was defined by calculated indices for most of the study period. While the indices are only estimates of NAFLD, it is notable that NAFLD by FLI found an identical estimate of NAFLD as did TE-CAP ≥263 dB/m in the 2017–2018 cycle. In addition, due to the cross-sectional study design of NHANES, no conclusions concerning the temporal association between PFAS and NAFLD are possible.

In conclusion, the current study found an overall inverse association between total PFAS and NAFLD. Thus, the hypothesis that exposure to PFAS could be associated with the development of NAFLD was not supported.

Conflicts of interest statement

The authors declare that they have no conflicts of interest with regard to the content of this report.

Supplementary Material

ee9-8-e284-s001.docx ^{(27.7KB, docx)}