Association of blood manganese, selenium with steatosis, fibrosis in the National Health and Nutrition Examination Survey, 2017–18

Abstract

Background & Aims:

Chronic liver disease is a growing health burden worldwide. Chronic metal exposures may be associated with non-alcoholic fatty liver disease (NAFLD). We aimed to evaluate the association of blood cadmium (Cd), mercury (Hg), lead (Pb), manganese (Mn), and selenium (Se) with two hallmark features of NAFLD: liver steatosis and fibrosis in the general U.S. population.

Methods:

We analyzed transient liver elastography data from participants of the National Health and Nutrition Examination Survey (NHANES) 2017–18, using ordinal logistic regression analyses to evaluate the cross-sectional association between blood metal concentrations and clinical stages of steatosis and fibrosis. We applied survey weights, strata, and primary sampling units and analyses were conducted using the R survey package.

Results:

4,154 participants were included. Median (IQR) for blood Mn and blood Se were 9.28 (7.48–11.39) and 191.08 (176.55–207.16) μg/L, respectively. Per interquartile range increase of natural log transformed blood Mn, the adjusted odds ratio (OR) (95% CI) was 1.59 (1.13–2.23) for a higher grade of steatosis and 1.16 (0.67–2.00) for liver fibrosis. The corresponding OR for steatosis was 2.00 (1.24–3.24) and 2.14 (1.04–4.42) in Black and Mexican American participants, respectively. The corresponding OR for liver fibrosis was 2.96 (1.42–6.17) for females. Per interquartile range increase of natural log transformed blood Se, the adjusted OR was 2.25 (1.30–3.89) for steatosis but 0.31 (0.13–0.72) for liver fibrosis. The inverse association of blood Se with liver fibrosis was also observed in males and White participants. Blood Cd, Hg, and Pb were not associated with liver steatosis and fibrosis in fully-adjusted models overall.

Conclusions:

In NHANES 2017–18, higher blood Mn was positively associated with liver steatosis, and higher Se was positively associated with liver steatosis but negatively associated with liver fibrosis. Longitudinal studies are needed to examine the association of Mn and Se with fibrosis progression.

Graphical Abstract

Introduction

Chronic liver disease is a major health concern globally and in the United States (US). Chronic liver disease/cirrhosis was the 11^th leading cause of death in the US in 2017, and approximately 1.8% of (or 4.5 million) US adults were diagnosed with chronic liver disease in 2018.(1, 2) Non-alcoholic fatty liver disease (NAFLD) is increasingly the most common cause of chronic liver disease, and is characterized by the accumulation of excess fat in the liver.(3, 4) The histologic spectrum of NAFLD ranges from simple steatosis to non-alcoholic steatohepatitis (NASH), fibrosis and eventually end-stage liver disease such as cirrhosis and liver cancer.(5) NAFLD is considered a benign condition, while NASH is characterized by liver steatosis, inflammation, and a different degree of fibrosis and is considered to increase the risk of liver fibrosis and liver cancer.(6, 7) In the absence of a specific treatment for NAFLD, it is important to identify risk factors involved in disease progression allowing for prevention strategies at the individual and population level.

Increasing epidemiologic and mechanistic evidence suggests the role of chronic metal exposures in the pathogenesis of NAFLD.(8, 9) Sources of metal exposures to the general US population occur primarily from diet, drinking water, and air pollution.(10) Cadmium (Cd), mercury (Hg), manganese (Mn), lead (Pb), and selenium (Se) occur naturally in soils across the US, with higher levels of ambient air and water exposures in areas near mining, smelting or industrial sites such as chemical manufacturing and steel production.(11–15) Although the molecular mechanism of Mn in the pathogenesis of fibrosis is not fully established, as shown in vitro, ex vivo, and in mice, it might relate to oxidative stress, inflammation, mitochondrial damage, and the role of bile acids in Mn homeostasis, all key pathways influencing progression of NAFLD.(9, 16, 17) In zebrafish, Cd exposure may initiate abnormal adipocyte differentiation and function, and metabolic dysfunction in the liver, which may lead to NAFLD.(18) In in vitro models of human monocytes, Hg exposure may induce proinflammatory cytokine responses; innate immune cell recruitment to the liver contributes to inflammatory processes that may drive NAFLD progression.(19) In rats, chronic Pb exposure induced hepatic steatosis and decreased evenness and phylogenetic diversity of gut microbiota; in particular, decreased relative abundance of Coprococcus and Oscillospira may contribute to decreased short-chain fatty acids and abnormal metabolism of secondary bile acids that cause Pb-induced fatty liver disease.(20) Potential mechanisms of Se exposure in NAFLD pathogenesis are not as well understood. Most previously published epidemiologic studies used elevated liver enzymes including alanine aminotransferase (ALT) in the absence of excessive alcohol use and viral hepatitis, to define NAFLD/hepatic steatosis and fibrosis.(6, 21) In the 2017–2018 survey cycle, the National Health and Nutrition Examination Survey (NHANES) used, for the first time, liver transient elastography to measure controlled attenuation parameter (CAP), a measure of liver steatosis, and liver stiffness, a measure of liver fibrosis.(22)

In this study, we examined the cross-sectional association of manganese (Mn), selenium (Se), cadmium (Cd), lead (Pb), and mercury (Hg) in blood with liver steatosis (measured by median CAP, in decibels per meter (dB/m)) and liver fibrosis (quantified as liver stiffness, in kilopascals (kPa)), in NHANES 2017–18 participants. We also aimed to evaluate potential effect measure modification by sex, race/ethnicity, alcohol consumption, diabetes, and body mass index (BMI).

Methods

Study Population

The study includes participants from the NHANES 2017–18 survey cycle.(23) NHANES is a series of cross-sectional surveys representative of the general US population.(22) NHANES has a complex, multi-stage probability cluster sampling design and uses a combination of interviews, physical examinations and laboratory measurements to determine the prevalence of major diseases and their risk factors.(22) The National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention conducts NHANES and the NCHS institutional review board approves study protocols.(22) All participants provide written informed consent. As this study uses de-identified and publicly available data from the NCHS, it is exempt from institutional review board approval.

We evaluated 5,492 adults aged 18 years and older with complete liver elastography measurements (median CAP and liver stiffness). Participants who were ineligible for liver elastography (participants less than 12 years of age, are pregnant or could not provide urine to test for pregnancy, had an implanted electronic medical device, or had bandages or lesions where the measurements are taken, n=3,306) or did not have a complete test result (n=456) were excluded.(23, 24) We further excluded participants with a hepatitis C confirmed antibody (n=87) or positive hepatitis B surface antigen (n=26). Hepatitis B surface antigen was evaluated in serum samples using the VITROS HBsAg test.(25) Hepatitis C confirmed antibody was analyzed in serum samples using the manual 16-hour sample incubation test procedure.(26) We additionally excluded participants missing any of the following: body mass index (BMI, n=44), smoking status (participants > 18 years old, n=742; serum cotinine, n=222), alcohol consumption status (n=213), and blood metal data (n=4), leaving 4,154 participants (unweighted sample) for analysis (Figure 1).

Figure 1.

Exclusion criteria of NHANES 2017–18 participants for analysis.

Exposures: Blood metal measurements

Whole blood specimens were collected during the examination, frozen and stored at −30°C and shipped on to the National Center for Environmental Health, Centers for Disease Control and Prevention (CDC) in Atlanta, GA for analysis.(27) Concentrations of Cd, Pb, Hg, Mn, and Se were measured in whole blood using inductively coupled plasma dynamic reaction cell mass spectrometry following simple dilution, as previously described in detail and following NHANES quality assurance and quality control protocols.(27) The lower limit of detection (LLOD) for Cd, Hg, Mn, and Se (μg/L) was 0.10, 0.99, 0.28, and 24.48, respectively; the LLOD was 0.07 μg/dL for Pb.(27) Measurements below the LOD were imputed as the LLOD divided by the square root of 2.(27) The coefficient of variation ranged from 2.5–9.5 for Cd, 1.6–2.8 for Pb, 3.3–4.9 for Mn, 2.7–11.4 for Hg, and 3.1–3.9 for Se.(28)

Outcomes: Liver steatosis and fibrosis

Liver elastography measurements were collected using FibroScan^® at the NHANES Mobile Examination Center.(24) Fibroscan employs ultrasound and vibration controlled transient elastography to measure liver stiffness (median of 10 liver stiffness measurements, in kilopascals, kPa).(24) Fibroscan also measures the ultrasound attenuation related to hepatic steatosis as the controlled attenuation parameter (median CAP measured in decibels per meter, dB/m).(24) A status exam code was provided for each participant. We restricted our analyses to participants with complete exams (fasting time of at least 3 hours, at least 10 complete stiffness (E) measurements, and a liver stiffness interquartile range / median stiffness of < 30%).(27) These are continuous measures that are often categorized based on clinically meaningful, standard cutoffs for liver steatosis: <260 (grades S0-S1, amount of liver with fatty change (steatosis) ≥ 33%), 260 - <290 (grade S2, amount of liver with fatty change >33 – 66%), ≥290 (grade S3, amount of liver with fatty change >66%); and, for liver fibrosis: <7 (reference, no or mild scarring of the liver), 7 - <10.5 (significant fibrosis, moderate scarring of the liver), ≥ 10.5 (cirrhosis, severe to advanced liver scarring).(29, 30)

Covariates

Demographic information on age, sex, race/ethnicity, education, alcohol use, and smoking status was collected using interviewer-administered questionnaires. Race/ethnicity was categorized as Non-Hispanic White, Non-Hispanic Black, Non-Hispanic Asian, Mexican American, and Other (including Other Hispanic and Other, including multiracial). Education was dichotomized as less than high school versus greater than or equal to high school. Alcohol use was categorized as never or ever had an alcoholic drink; and in sensitivity analyses we also adjusted for alcohol use using the following categories based on the Dietary Guidelines for Americans 2020–2025: never drinkers, low-moderate drinkers (2 or less drinks on average per day for males and 1 or less drinks on average for females, on days that the participant consumed alcohol during the past year), or heavy drinkers (greater than 2 drinks on average per day for males and greater than 1 drink on average for females, on days that the participant consumed alcohol during the past year).(31) Smoking status was categorized as never (smoked less than 100 cigarettes over lifetime), former (smoked ≥ 100 cigarettes over lifetime and does not currently smoke), or current smoker (smoked ≥ 100 cigarettes over lifetime and currently smokes). Anthropometric measurements were used to calculate continuous BMI as weight (kg) / height² (m²). Cotinine (μg/L), a metabolite of nicotine, was measured in serum. Urine arsenobetaine (μg/L), a non-toxic seafood-derived organic arsenic species and a specific marker of recent seafood intake, was measured in urine in a random 1/3 subset of participants (N=1,370). Diabetes status was defined (yes/no) according to the American Diabetes Association criteria, as fasting plasma glucose ≥126 mg/dL, use of insulin medication, hemoglobin A1C ≥6.5%, or prior self-reported clinical diagnosis.(32) Total caloric intake was quantified as the daily aggregate of food energy (kilocalories) from all foods and beverages in the 24-hour dietary recall, collected during the interview-administered questionnaire, and calculated using the US Department of Agriculture Food and Nutrient Database for Dietary Studies 2017–18.(23)

Statistical Analysis

Data management and statistical analyses were conducted in R version 4.1.0 using the survey package.(33) Survey weights, strata, and primary sampling units were applied to all analyses to account for the NHANES complex and multi-stage survey design.(23)

Blood metal concentrations were skewed and log transformed (natural base) for analysis. We first compared participant sociodemographic characteristics overall and across quartiles of blood Mn and Se concentrations. We performed adjusted ordinal logistic regression to evaluate the odds ratio (OR) of a more severe stage of hepatic steatosis, i.e. the odds of progression from stage S0-S1 (steatosis ≤33%) to S2 (steatosis >33–66%), or from stage S2 to stage S3 (steatosis >66%), per blood metal increase (modeled as log transformed) as a continuous variable (all blood metals), and corresponding to the difference between the 75th and 25th percentile (interquartile range, IQR) for blood Mn and Se. We also evaluated the OR of a more severe stage of liver fibrosis, i.e. the odds of progression from no or mild stiffness (stiffness < 7 kPa) to significant fibrosis (7- <10.5 kPa), or from significant fibrosis to cirrhosis (≥10.5 kPa). We evaluated the proportional odds assumption in ordinal logistic regression using the ‘brant’ package in R, and confirmed the proportional odds assumption held.(34)

We conducted an extensive review of the literature to identify risk factors of liver disease and metal exposures, which were used to determine the covariate selection. Model 1 adjusted for age, race/ethnicity (Non-Hispanic White as reference), sex (female as reference), and education (greater than or equal to high school as reference). Model 2 further adjusted for alcohol use (never as reference), smoking status (never as reference), serum cotinine (continuous and natural log transformed), BMI (continuous), and diabetes status (no as reference). Urine arsenobetaine was used to adjust for seafood intake for models with blood Hg. Based on prior studies suggesting potential effect modification, we conducted exploratory analyses assessing effect measure modification, stratified by race/ethnicity, sex, alcohol consumption status, and BMI category.(8, 35) Results for overall analyses and analyses stratified by race/ethnicity and sex were displayed graphically in a forest plot.

Sensitivity analyses

We assessed all models excluding heavy drinkers, based on categories of daily alcohol consumption defined based on the Dietary Guidelines for Americans 2020–2025. We also assessed model 2 adjusting for alcohol consumption based on the Dietary Guidelines,(31) as never (N=452 overall), low-moderate (N=1,469 overall), and heavy (N=1,405) drinkers.

We additionally adjusted for natural log transformed total caloric intake (kilocalories) as measured in the 24 hour dietary recall, as total Mn and Se internal dose may be associated with particular dietary patterns related to NAFLD.

We used linear regression analyses to evaluate associations with continuous measures of liver steatosis (median CAP) and fibrosis (liver stiffness) as outcomes.

Potential non-linear associations between blood Mn and Se with continuous measures of liver steatosis (median CAP) and fibrosis (liver stiffness) were explored with natural cubic splines, using the splines package.(36, 37) Knots were manually supplied at the 10^th, 50^th, and 90^th percentiles of blood Mn (6.13, 9.28, and 13.59 μg/L, respectively) and Se (164.05, 191.08, and 224.19 μg/L, respectively). Boundary knots were selected automatically (by default) based on the range of the data: 1.57 (lower) and 52.0 (upper) for Mn (μg/L), and 89.80 (lower) and 453.62 (upper) for Se (μg/L).(36)

Results

Descriptive statistics

The median (IQR) was 9.28 (7.48–11.39) μg/L for blood Mn and 191.08 (176.55–207.16) μg/L for blood Se. The range was 1.57 to 52.00 μg/L for blood Mn and 89.80 to 453.62 μg/L for blood Se. Participants with higher blood Mn were younger, more likely to be female, had higher BMI, higher median CAP (steatosis), higher liver stiffness (fibrosis), were more likely to be never alcohol drinkers, and had lower prevalence of diabetes (Table 1A). Participants with higher blood Se were older, more likely to be male, had higher measures of liver steatosis and lower measures of fibrosis, and had higher prevalence of diabetes (Table 1B). Black participants were more likely to have lower blood Se and Mn, while Mexican American and Asian participants were more likely to have higher blood Mn. Participants with the most severe grade of steatosis or advanced fibrosis tended to be male, and higher BMI (Table 2).

Table 1A.

Descriptive characteristics of NHANES 2017–18 participants overall and by quartile of natural log transformed blood manganese (μg/L).

	Quartile of blood Mn (μg/L)
	Overall	Quartile 1 <7.48	Quartile 2 7.48 - <9.28	Quartile 3 9.28 - <11.39	Quartile 4 ≥11.39
N	4,154	1,033	1,011	972	1,138
Age, Mean (SE)	47.03 (0.74)	48.82 (0.76)	48.64 (1.12)	46.43 (0.99)	44.22 (0.94)
Body Mass Index, kg/m 2 , Mean (SE)	29.48 (0.29)	28.43 (0.25)	29.27 (0.59)	29.88 (0.44)	30.35 (0.36)
Liver steatosis (CAP, dB/m), Mean (SE)	262.88 (1.71)	258.72 (2.34)	261.06 (4.17)	262.12 (3.44)	269.63 (2.68)
Liver fibrosis (stiffness, kPa), Mean (SE)	5.62 (0.09)	5.57 (0.14)	5.55 (0.21)	5.43 (0.22)	5.92 (0.17)
Serum cotinine, μg/L, Mean (SE)	52.33 (5.31)	64.44 (6.81)	49.93 (5.98)	53.82 (5.95)	41.10 (8.51)
	N	% (N) 1
Categories of liver steatosis
S0-S1 (<260, dB/m)	1992	54.8 (555)	49.6 (474)	51.4 (456)	43.2 (507)
S2 (260-<290, dB/m))	720	16.3 (160)	17.5 (183)	16.2 (169)	17.3 (208)
S3 (≥290, dB/m))	1442	28.9 (318)	32.9 (354)	32.4 (347)	39.6 (423)
Categories of liver fibrosis
No fibrosis (<7, kPa)	3543	86.9 (874)	87.2 (871)	87.9 (826)	84.6 (972)
Significant fibrosis (7-<10.5, kPa)	421	9.7 (105)	8.5 (95)	9.0 (107)	9.9 (114)
Cirrhosis (≥10.5, kPa)	190	3.3 (54)	4.2 (45)	3.2 (39)	5.5 (52)
Race/ethnicity
Non-Hispanic White	1464	66.9 (402)	64.8 (393)	65.8 (351)	54.9 (318)
Non-Hispanic Black	910	16.2 (363)	10.1 (227)	8.0 (175)	7.3 (145)
Non-Hispanic Asian	556	1.0 (28)	3.5 (89)	5.1 (134)	11.5 (305)
Mexican American	602	5.4 (94)	8.5 (141)	10.2 (166)	13.6 (201)
Other	622	10.5 (146)	13.1 (161)	10.9 (146)	12.8 (169)
Sex
Male	2052	60.4 (648)	51.0 (515)	44.8 (458)	40.6 (431)
Female	2102	39.6 (385)	49.0 (496)	55.2 (514)	59.4 (707)
Education
≥ High School	3430	89.6 (838)	90.9 (860)	90.0 (788)	89.3 (944)
< High School	724	10.4 (195)	9.1 (151)	10.0 (184)	10.7 (194)
Alcohol
Never Drink	452	4.9 (73)	6.3 (96)	8.0 (111)	12.1 (184)
Ever Drink	3702	95.1 (977)	93.7 (934)	92.0 (874)	87.9 (963)
Smoking
Never	2491	52.6 (531)	58.6 (586)	62.4 (606)	64.7 (768)
Former	968	26.6 (278)	25.9 (245)	24.9 (226)	20.7 (219)
Current	695	20.8 (224)	15.5 (180)	12.7 (140)	14.6 (151)
Diabetes
No	3365	85.5 (805)	85.7 (821)	87.0 (787)	87.3 (952)
Yes	789	14.5 (228)	14.3 (190)	13.0 (185)	12.7 (186)

¹Percents are weighted to account for the complex NHANES survey design and weighting scheme. Percents represent proportion of column total for each metal. Ns are unweighted.

Table 1B.

Descriptive characteristics of NHANES 2017–18 participants overall and by quartile of natural log transformed blood selenium (μg/L).

			Quartile of blood Se (μg/L)
	Overall	Quartile 1 <176.55	Quartile 2 176.55 - <191.08	Quartile 3 191.08 - <207.16	Quartile 4 ≥207.16
N	4,154	1,124	1,044	1,029	957
Age, Mean (SE)	47.03 (0.74)	46.38 (1.01)	46.72 (1.03)	47.24 (0.84)	47.77 (0.89)
Body Mass Index, kg/m 2 , Mean (SE)	29.48 (0.29)	29.16 (0.30)	29.35 (0.38)	29.79 (0.40)	29.64 (0.44)
Liver steatosis (CAP, dB/m), Mean (SE)	262.88 (1.71)	254.47 (2.21)	260.69 (2.42)	266.36 (3.02)	269.98 (3.28)
Liver fibrosis (stiffness, kPa), Mean (SE)	5.62 (0.09)	5.86 (0.18)	5.50 (0.22)	5.74 (0.26)	5.37 (0.11)
Serum cotinine, μg/L, Mean (SE)	52.33 (5.31)	66.60 (8.51)	48.59 (6.22)	49.84 (5.52)	44.32 (6.95)
	N		N (%) 1
Categories of liver steatosis
S0-S1 (<260, dB/m)	1992	55.2 (601)	54.7 (524)	45.9 (461)	43.2 (406)
S2 (260-<290, dB/m)	720	14.8 (186)	14.7 (179)	18.1 (188)	19.6 (167)
S3 (≥290, dB/m)	1442	30.0 (337)	30.6 (341)	36.0 (380)	37.2 (384)
Categories of liver stiffness
No fibrosis (<7, kPa)	3543	86.0 (948)	88.0 (905)	84.6 (867)	88.1 (823)
Significant fibrosis (7-<10.5, kPa)	421	9.2 (111)	7.5 (96)	11.4 (116)	8.9 (98)
Cirrhosis (≥10.5, kPa)	190	4.8 (65)	4.5 (43)	4.0 (46)	3.0 (36)
Race/ethnicity
Non-Hispanic White	1464	60.3 (395)	61.4 (346)	63.7 (363)	66.9 (360)
Non-Hispanic Black	910	13.1 (301)	10.9 (228)	10.0 (216)	7.7 (165)
Non-Hispanic Asian	556	5.8 (126)	4.7 (133)	5.2 (144)	5.4 (153)
Mexican American	602	9.2 (145)	10.4 (171)	9.9 (156)	8.1 (130)
Other	622	11.7 (157)	12.5 (166)	11.2 (150)	11.9 (149)
Sex
Male	2052	43.7 (494)	46.1 (501)	50.8 (540)	56.2 (517)
Female	2102	56.3 (630)	53.9 (543)	49.2 (489)	43.8 (440)
Education
≥ High School	3430	87.1 (905)	89.8 (855)	91.6 (854)	91.3 (816)
< High School	724	12.9 (219)	10.2 (189)	8.4 (175)	8.7 (141)
Alcohol
Never Drink	452	8.4 (125)	6.4 (103)	8.5 (109)	7.9 (115)
Ever Drink	3702	91.6 (999)	93.6 (941)	91.5 (920)	92.1 (842)
Smoking
Never	2491	59.3 (640)	61.7 (648)	59.5 (628)	57.8 (575)
Former	968	22.1 (265)	23.3 (224)	24.4 (236)	28.3 (243)
Current	695	18.7 (219)	15.0 (172)	16.0 (165)	13.9 (139)
Diabetes
No	3365	87.0 (931)	88.4 (850)	84.9 (813)	85.1 (771)
Yes	789	13.0 (193)	11.6 (194)	15.1 (216)	14.9 (186)

¹Percents are weighted to account for the complex NHANES survey design and weighting scheme. Percents represent proportion of column total for each metal. Ns are unweighted.

Table 2.

Descriptive characteristics¹ of NHANES 2017–18 participants by liver transient elastography measures² of liver steatosis and fibrosis measured using median controlled attenuation parameter and median stiffness, respectively.

Outcome category2	S0-S1	Liver steatosis S2	S3	No fibrosis	Liver fibrosis Sig. fibrosis	Cirrhosis
N	1992	720	1442	3543	421	190
Age, Mean (SE)	43.25 (0.78)	50.26 (1.25)	51.04 (0.81)	46.43 (0.71)	49.90 (1.77)	53.40 (1.41)
Body Mass Index, kg/m 2 , Mean (SE)	26.05 (0.34)	30.23 (0.31)	34.20 (0.42)	28.55 (0.26)	33.87 (0.95)	39.30 (1.01)
Blood manganese, μg/L, Mean (SE)	9.50 (0.10)	9.92 (0.28)	10.07 (0.13)	9.73 (0.07)	9.83 (0.33)	10.34 (0.43)
Blood selenium, μg/L, Mean (SE)	191.27 (0.87)	196.01 (1.91)	195.50 (1.26)	193.72 (1.00)	193.20 (1.91)	189.02 (2.25)
Serum cotinine, μg/L, Mean (SE)	58.52 (6.67)	48.41 (7.20)	45.08 (4.48)	53.22 (5.09)	46.64 (9.88)	46.36 (17.11)
		% (N)			% (N)
Race/ethnicity
Non-Hispanic White	63.6 (684)	60.3 (241)	63.7 (539)	63.0 (1241)	62.3 (144)	66.8 (79)
Non-Hispanic Black	12.3 (514)	11.2 (161)	7.2 (235)	10.3 (767)	12.5 (112)	7.5 (31)
Non-Hispanic Asian	5.4 (281)	6.1 (106)	4.7 (169)	5.6 (507)	3.1 (33)	3.2 (16)
Mexican American	6.7 (206)	9.2 (104)	13.5 (292)	9.2 (495)	11.0 (73)	10.6 (34)
Other	12.1 (307)	13.2 (108)	10.8 (207)	11.9 (533)	11.1 (59)	12.0 (30)
Sex
Female	56.7 (1109)	50.2 (372)	42.3 (621)	52.2 (1842)	42.8 (180)	39.2 (80)
Male	43.3 (883)	49.8 (348)	57.7 (821)	47.8 (1701)	57.2 (241)	60.8 (110)
Alcohol
Never	7.7 (230)	6.3 (74)	8.7 (148)	7.8 (403)	7.5 (39)	8.2 (10)
Ever	92.3 (1762)	93.7 (646)	91.3 (1294)	92.2 (3140)	92.5 (382)	91.8 (180)
Education
≥ High School	91.3 (1680)	87.9 (590)	89.0 (1160)	90.1 (2939)	89.3 (339)	87.5 (152)
< High School	8.7 (312)	12.1 (130)	11.0 (282)	9.9 (604)	10.7 (82)	12.5 (38)
Alcohol status
Never Drink	7.7 (230)	6.3 (74)	8.7 (148)	7.8 (403)	7.5 (39)	8.2 (10)
Ever Drink	92.3 (1762)	93.7 (646)	91.3 (1294)	92.2 (3140)	92.5 (382)	91.8 (180)
Smoking status
Never	63.0 (1249)	57.8 (440)	55.3 (802)	60.2 (2159)	55.2 (235)	54.9 (97)
Former	20.5 (380)	26.4 (176)	29.5 (412)	23.8 (788)	28.4 (111)	31.6 (69)
Current	16.4 (363)	15.8 (104)	15.2 (228)	16.0 (596)	16.4 (75)	13.5 (24)
Diabetes status
No	95.2 (1811)	87.6 (583)	72.7 (971)	89.3 (2984)	71.5 (283)	58.1 (98)
Yes	4.8 (181)	12.4 (137)	27.3 (471)	10.7 (559)	28.5 (138)	41.9 (92)

¹Percents are weighted to account for the complex NHANES survey design and weighting scheme. Percents represent proportion of column total for each subgroup.

²Grades are based on standard cutoffs for liver steatosis: <260 dB/m (grades S0-S1, steatosis ≤ 33%), 260 - <290 dB/m (grade S2, steatosis >33 – 66%), ≥290 dB/m (grade S3, steatosis >66%) and fibrosis: <7 (reference), 7 - <10.5 (significant fibrosis), ≥ 10.5 (cirrhosis).

Ordinal logistic regression analyses

We computed odds ratios (95% CI) for a more severe stage of liver steatosis (comparing the odds of a more severe stage of steatosis, i.e. from stage S0-S1 (steatosis ≤33%) to stage S2 (steatosis >33–66%), or stage S2 to S3 (steatosis >66%)) or liver fibrosis (comparing the odds of a more severe stage of fibrosis, i.e. mild to no scarring (reference) to significant fibrosis, or significant fibrosis compared to cirrhosis). ORs for a higher grade of steatosis, per interquartile range increase in log-transformed blood Mn (IQR = 0.42 μg/L), was 2.60 (1.96–3.44) and 1.59 (1.13–2.23) in models 1 and 2, respectively (Table 3A, Figure S1). The odds ratio (95% CI) for a higher stage of fibrosis was 1.83 (1.12–2.99) and 1.16 (0.67–2.00) in models 1 and 2, respectively.

Table 3A.

Odds ratios (95% confidence interval) for the association between the 75^th vs 25^th percentile of log blood manganese (IQR = 0.42 μg/L) and liver steatosis (S2 vs S0-S1 and S3 vs S2)¹ and liver fibrosis (stiffness category 2 vs 1, category 3 vs 2)¹ using ordinal logistic regression models.

	Liver steatosis		Liver fibrosis
	Model 12	Model 23	Model 12	Model 23
Overall	2.60 (1.96–3.44)***	1.59 (1.13–2.23)**	1.83 (1.12–2.99)*	1.16 (0.67–2.00)
By sex
Male	2.45 (1.54–3.89)***	1.60 (0.89–2.86)	1.00 (0.46–2.20)	0.62 (0.24–1.63)
Female	2.77 (1.90–4.05)***	1.62 (1.05–2.51)*	3.85 (1.92–7.71)***	2.96 (1.42–6.17)**
By race/ethnicity
Non-Hispanic White	2.68 (1.75–4.12)***	1.33 (0.81–2.20)	1.60 (0.72–3.55)	0.89 (0.37–2.15)
Non-Hispanic Black	2.64 (1.80–3.87)***	2.00 (1.24–3.24)**	2.22 (1.15–4.27)*	1.84 (1.00–3.39)
Non-Hispanic Asian	2.14 (1.07–4.27)*	1.42 (0.67–3.02)	1.53 (0.46–5.09)	1.25 (0.38–4.12)
Mexican American	2.44 (1.19–5.01)*	2.14 (1.04–4.42)*	1.46 (0.53–4.02)	1.44 (0.44–4.72)
Other	2.31 (1.03–5.19)*	1.32 (0.59–2.96)	3.91 (0.64–23.70)	1.58 (0.43–5.75)
By alcohol status
Never	1.60 (0.43–5.96)	1.76 (0.75–4.11)	2.66 (0.58–12.24)	3.26 (0.84–12.63)
Ever	2.64 (1.97–3.53)***	1.54 (1.07–2.21)*	1.81 (1.03–3.19)*	1.10 (0.58–2.07)
By BMI, kg/m 2
< 25	0.87 (0.25–3.09)	0.86 (0.19–3.79)	0.66 (0.06–7.12)	0.71 (0.09–5.67)
25 - <30	1.60 (0.84–3.05)	1.78 (0.87–3.63)	1.39 (0.68–2.86)	1.38 (0.67–2.87)
≥ 30	1.91 (1.25–2.94)**	1.98 (1.32–2.96)**	1.38 (0.93–2.04)	1.48 (0.94–2.33)

N = 4,154 overall, 2,052 male, 2,102 female, 1,464 non-Hispanic White, 910 non-Hispanic Black, 556 non-Hispanic Asian, 602 Mexican American and 622 Other.²

P values < 0.001: ***, < 0.01: **, < 0.05: *

¹Categories are based on standard cutoffs for liver fibrosis: <7 kPa (reference), 7 - <10.5 kPa (significant fibrosis), ≥ 10.5 kPa (cirrhosis), and liver steatosis: <260 dB/m (grades S0-S1, steatosis ≤ 33%), 260 - <290 dB/m (grade S2, steatosis >33 – 66%), ≥290 dB/m (grade S3, steatosis >66%).

²Model 1: Adjusts for race/ethnicity, sex, age, and education. Stratified models are adjusted for covariates not stratified on.

³Model 2: Adjusts for race/ethnicity, sex, age, education, alcohol use, smoking status, BMI, log serum cotinine and diabetes status. Stratified models are adjusted for covariates not stratified on.

Findings for the association between blood Mn and liver steatosis in fully adjusted models over all participants were largely consistent with model results stratified by sex, race/ethnicity, alcohol consumption status, and in participants with BMI 25 kg/m² and higher, but not in participants with BMI <25 kg/m² (Table 3A). Significant positive associations were observed among female, non-Hispanic Black and Mexican American participants, Ever alcohol drinkers and among individuals with BMI ≥ 30 kg/m² in fully adjusted analyses. The association between blood Mn and liver fibrosis differed by sex and race, with inverse associations observed among males and a significant, positive association observed among females.

The odds ratio (95% CI) for a higher grade of liver steatosis, per interquartile range increase in log-transformed blood Se (IQR = 0.16 μg/L), was 2.35 (1.34–4.11) and 2.25 (1.30–3.89) in models 1 and 2, respectively (Table 3B, Figure S1). The corresponding odds ratio (95% CI) for a higher grade of liver fibrosis was 0.47 (0.20–1.11) and 0.31 (0.13–0.72) for models 1 and 2, respectively.

Table 3B.

Odds ratios (95% confidence interval) for the association between the 75^th vs 25^th percentile of log blood selenium (0.16 μg/L) and liver steatosis (S2 vs S0-S1 and S3 vs S2)¹ and liver fibrosis (stiffness category 2 vs 1, category 3 vs 2)¹ using ordinal logistic regression models.

	Liver steatosis		Liver fibrosis
	Model 12	Model 23	Model 12	Model 23
Overall	2.35 (1.34–4.11)**	2.25 (1.30–3.89)**	0.47 (0.20–1.11)	0.31 (0.13–0.72)**
By sex
Male	4.12 (1.88–9.04)***	3.70 (1.64–8.35)**	0.26 (0.06–1.08)	0.15 (0.05–0.46)***
Female	1.21 (0.47–3.11)	1.27 (0.53–3.02)	1.00 (0.45–2.23)	0.88 (0.25–3.12)
By race/ethnicity
Non-Hispanic White	1.63 (0.80–3.35)	1.67 (0.66–4.23)	0.30 (0.09–0.96)*	0.19 (0.05–0.68)*
Non-Hispanic Black	4.58 (1.15–18.16)*	6.76 (1.93–23.71)**	1.57 (0.36–6.89)	2.35 (0.45–12.26)
Non-Hispanic Asian	5.15 (1.68–15.74)**	4.54 (1.29–15.95)*	1.23 (0.06–26.13)	0.20 (0.01–2.81)
Mexican American	5.64 (1.41–22.53)*	5.29 (0.84–33.33)	1.05 (0.19–5.83)	0.87 (0.16–4.90)
Other	5.04 (0.68–37.27)	2.58 (0.30–22.39)	0.55 (0.05–5.66)	0.15 (0.01–1.47)
By alcohol status
Never	0.34 (0.03–3.27)	2.31 (0.15–35.28)	0.22 (0.03–1.92)	0.83 (0.02–38.63)
Ever	2.71 (1.45–5.08)**	2.19 (1.21–3.98)**	0.49 (0.20–1.22)	0.28 (0.12–0.68)**
By BMI, kg/m 2
< 25	0.49 (0.06–3.79)	0.37 (0.05–2.73)	0.13 (0.01–2.56)	0.13 (0.01–2.04)
25 - <30	5.15 (1.87–14.23)**	4.71 (1.48–14.95)**	0.48 (0.07–3.29)	0.43 (0.06–3.02)
≥ 30	2.16 (0.72–6.50)	2.52 (0.76–8.35)	0.56 (0.16–1.94)	0.55 (0.15–2.09)

P values < 0.001: ***, < 0.01: **, < 0.05: *. N = 4,154 overall, 2,052 male, 2,102 female, 1,464 non-Hispanic White, 910 non-Hispanic Black, 556 non-Hispanic Asian, 602 Mexican American and 622 Other.

¹Categories are based on standard cutoffs for liver stiffness: <7 kPa (reference), 7 - <10.5 kPa (significant fibrosis), ≥ 10.5 kPa (cirrhosis), and liver steatosis: <260 dB/m (grades S0-S1, steatosis ≤ 33%), 260 - <290 dB/m (grade S2, steatosis >33 – 66%), ≥290 dB/m (grade S3, steatosis >66%).

²Model 1: Adjusts for race/ethnicity, sex, age, and education. Stratified models are adjusted for covariates not stratified on.

³Model 2: Adjusts for race/ethnicity, sex, age, education, alcohol use, smoking status, BMI, log serum cotinine and diabetes status. Stratified models are adjusted for covariates not stratified on.

Findings for the positive association between blood Se and liver steatosis in fully adjusted models over all participants were largely consistent with model results stratified by sex, race/ethnicity, alcohol drinking status, and in participants with BMI 25 kg/m² and higher, but not in participants with BMI <25 kg/m² (Table 3B). Significant positive associations were observed among male, non-Hispanic Black and Asian participants, Ever alcohol drinkers and among individuals with BMI 25 - <30 kg/m² in fully adjusted analyses. Significant inverse associations were observed between blood Se and liver fibrosis among males, non-Hispanic White participants, and Ever drinkers.

The observed odds ratios computed per IQR increase were consistent with the analyses using continuous log transformed concentrations of Mn and Se as predictors (Table S1A–B).

Additional blood metals

The results of ordinal logistic regression models evaluating the association between additional blood metals (log μg/L), Cd, Pb, and total Hg, with liver steatosis and fibrosis are provided in Table S2. Significant positive associations were observed between blood Cd and Hg (per log μg/L increase) with a higher stage of liver steatosis among non-Hispanic White participants in fully adjusted analyses. All other associations for blood Cd, Pb and Hg were null in fully adjusted analyses.

Sensitivity analyses

Excluding heavy alcohol drinkers, adjustment for alternative alcohol status

Findings from ordinal logistic regression analyses excluding heavy drinkers, defined based on the Dietary Guidelines for Americans 2020–2025, are similar to results from the main analyses (Table S3). Blood Mn was significantly positively associated with a higher grade of steatosis overall and among females in fully adjusted analyses. Due to smaller sample sizes, odds ratios could not be computed for Non-Hispanic Black, Asian, and Mexican American participants, so results stratified by race/ethnicity are not shown. Findings from analyses adjusting for alcohol consumption defined based on the Dietary Guidelines for Americans (never, low-moderate, and heavy drinkers) were similar to analyses dichotomizing alcohol consumption as never vs ever drinkers) (Table S4).

Adjusting for total caloric intake

In ordinal logistic regression models additionally adjusting for natural log transformed total caloric intake (kilocalories), the associations between blood Mn and Se with liver steatosis and fibrosis are similar to results from our main analyses (Table S5).

Linear regression analyses

Results for the linear associations between continuous blood Mn and liver steatosis (measured as median CAP) and liver fibrosis (measured as median stiffness) are in general consistent with the findings from the ordinal logistic regression models (Table S6). The fully adjusted mean difference in median CAP and median stiffness levels per increase in blood Mn (modeled as log-transformed) was 6.31 (2.45–10.17) dB/m and 0.09 (−0.56–0.74) kPa, respectively. The adjusted mean difference in median CAP and median stiffness levels per increase in blood Se (modeled as log-transformed) was 16.08 (5.97–26.20) dB/m and −2.29(−3.16 to −1.42) kPa, respectively. Results for the stratified analyses were similar.

Non-linear regression analyses

Natural cubic splines were used in generalized linear regression analyses to explore non-linear dose-response associations between natural log transformed blood Mn and Se (μg/L) with continuous measures of steatosis (median CAP) and fibrosis (liver stiffness) overall (Figure S2). Graphs of predicted median CAP per increase in blood Mn suggest a non-linear association between the 10^th to 90^th percentiles of blood Mn when additionally accounting for alcohol use, smoking and diabetes status (model 2). Predicted median CAP generally increases between the 10^th to 90^th percentile of blood Se; however, at concentrations above the 90^th percentile of blood Se, predicted median CAP decreases.

Discussion

Of the 5 metals we evaluated, only Mn and Se were associated with the liver outcomes studied in the general US population. Overall, we found increasing blood Mn was associated with increased odds of an elevated grade of liver steatosis. Overall, higher blood Se was positively associated with liver steatosis but inversely associated with liver fibrosis. Differences in these associations were also observed by race/ethnicity. An inverse association between higher blood Se and liver fibrosis was observed in males and Non-Hispanic White participants.

Cross-sectional epidemiologic evidence suggests the role of chronic Mn exposure in the pathogenesis of hepatic steatosis and fibrosis in US populations, however available studies are limited.(8) Ahmad et al. (2020) found that bile acids influence intestinal control of Mn homeostasis through the manganese transporter Slc30a10, which is expressed in mRNA and induces Mn efflux, reflecting a major breakthrough in the field.(9) These findings have important implications for liver outcomes as well as diabetes because rates of 12 alpha hydroxylated bile acids synthesis are higher in insulin resistance/ type 2 diabetes.(38) Hypermanganesemia, an inherited disorder characterized by elevated levels of Mn in blood (due to Slc30a10 gene mutations that affect the Mn transporter), has been observed in patients with liver cirrhosis.(39, 40)

Se is an essential element and plays an important role in the defense against oxidative stress.(41) We found opposite associations between Se and steatosis, which was positive, and Se and liver fibrosis, which was negative. These opposite trends were largely consistent in most stratified analyses. Results from non-linear analyses suggest that a negative association between blood Se and median CAP (steatosis) occurred at higher concentrations (above the 90^th percentile) of blood Se. We do not have a good explanation for these findings. The inverse association with fibrosis was also found for the risk of advanced liver fibrosis, determined using the NAFLD Fibrosis Score (NFS), in NHANES 1988–94 participants.(35) The positive association with steatosis was consistent with a study in Chinese adults, which found higher plasma Se was associated with an increased odds of NAFLD, reported by ultrasonography.(42) In NHANES 2011–16 elevated serum Se levels showed a non-linear association with NAFLD, defined by serum alanine aminotransferase activity.(43) Given the complexity of these cross-sectional findings for Se, prospective studies in a population with long-term follow-up and data on the different stages of NAFLD is needed to understand the association of Se with liver disease and potential differential roles of Se in earlier vs. more advanced stages, notwithstanding that some of the observed associations might be related to reverse causality.

Blood Mn and Se have been shown to differ by race/ethnicity group in previous studies of NHANES: Non-Hispanic Caucasian participants had higher Se while Non-Hispanic Black and Mexican American participants had lower Se(35). Conversely, Asian and Mexican-American participants have been found to have higher Mn.(44) In our study, though Non-Hispanic Black participants tended to have lower blood Mn, blood Mn was positively and significantly associated with increased odds of steatosis in Non-Hispanic Black and Mexican American participants. In the US, Mexican Americans have a higher burden of NAFLD and Black Americans have a lower burden, with increasing prevalence among Asian Americans.(45, 46) These differences may be attributed in part to diet as well as socioeconomic status: consumption of nutrient rich foods that contain Se (i.e. meat, seafood, and nuts) may be linked to socioeconomic status, as these food items are typically more expensive.(35, 47).

In the US population, sex differences in blood Mn and Se have previously been observed: females had higher levels of blood Mn and lower levels of Se compared to males.(48, 49) Sex differences in the association of Mn and Se with health effects have also been reported.(48, 50, 51) Published reports of higher blood Mn among female vs male NHANES participants may suggest sex-specific differences in metabolism and regulation of Mn.(44) Although women are less likely to have liver cancer compared to men, women are more commonly affected by toxin-mediated liver diseases such as alcohol- and drug-induced liver injury and failure.(52, 53) There are sex-related differences in liver outcomes reported in the literature, however the mechanisms behind the effect modification of the association between Se and liver phenotypes, by sex, remain to be determined.(54) Further studies are needed to understand sex-specific differences in the hepatotoxicity of Mn and Se.

We also observed differences by BMI category; positive associations were observed between blood Mn and liver steatosis among participants with BMI ≥30 kg/m², and between blood Se and liver steatosis among participants with BMI from 25-<30 kg/m². In the US, NAFLD often co-occurs with obesity, insulin resistance and type 2 diabetes, and the prevalence of obesity has been linked to the prevalence of NAFLD.(55) In univariate analyses, blood Se was positively associated (p < 0.05) with liver steatosis among participants with BMI > 25, however this was not observed with blood Mn (Table S1A–B).

In fully adjusted analyses, we found no overall association between Pb, Cd, and Hg with liver disease outcomes measured by ultrasound. These findings are inconsistent with the Korea National Health and Nutrition Examination Survey, where blood Pb, Hg and Cd were positively correlated with hepatic steatosis and fibrosis, defined as hepatic steatosis index and Fibrosis-4.(56) In NHANES 1988–94, urinary Cd was positively associated with steatosis, defined as the presence of moderate to severe steatosis with normal liver enzymes levels.(57, 58) A potential explanation for this discrepancy may be related to different metal concentrations among distinct populations, as well as different outcome definitions. Mean blood Pb, Hg and Cd concentrations were higher in Chung et al. (mean (range) = 1.81 (0.20–20.16), 3.89 (0.29–42.80) and 1.08 (0.10–6.62) μg/L, respectively) compared to our study (mean (range) = 1.08 (0.09–42.48), 1.31 (0.20–63.64) , and 0.43 (0.07–13.03), though the ranges varied.(56) Previous studies of the associations of metal exposure and NAFLD used elevated liver enzyme or blood-based, risk scores of liver fibrosis as the outcome(35). Studies have concluded that vibration controlled transient elastography performs well to detect and stage liver fibrosis and to evaluate liver steatosis.(59–61) Thus, compared to prior studies, our study using transient elastography provides a more refined definition of the outcome.

Potential limitations to this study include the cross-sectional study design of NHANES, which does not allow determination of temporality, and that alcohol consumption is self reported. Additionally, smaller numbers of Non-Hispanic Asian and Mexican American participants limits our power to assess differences between subgroups, and some of the findings might be unstable.(23) A majority of Asian participants in this study were in quartile 4 of blood Mn, highlighting the importance of continued research to evaluate metal exposures to Asian Americans and implications for metabolic and chronic disease risk. Shaheen et al. observed that prevalence of severe hepatic steatosis is higher in Mexican Americans participants but not higher among Non-Mexican (Other) Hispanic participants, compared to other racial/ethnic groups.(62) Though we aggregated Other Hispanic and Other, multi racial participants for stratified analyses, we also found a slightly stronger association with Mexican American participants compared to Other participants.

An additional limitation is that blood Mn may not reflect total Mn exposure, particularly in highly exposed occupational settings, or long-term cumulative Mn exposures, potentially due to its short half-life.(63, 64) Smith et al. observed that blood Mn was associated with the cumulative respiratory exposure index among workers, while plasma and urine Mn was not.(63) Toenails and hair may also improve the validity of estimates of cumulative Mn exposures.(64) As Mn was only measured in whole blood specimens in NHANES 2017–18 participants, we cannot compare to urinary or other biomarkers; however, in combination with emerging mechanistic evidence, our analyses suggest a potential association between Mn exposure, as measured in blood, and liver steatosis.

Conclusions

Our study provides evidence that higher Mn is specifically associated with liver steatosis. In addition, higher Se is positivity associated with liver steatosis and negatively associated with liver fibrosis. Longitudinal studies are needed to examine the associations of Mn and Se exposure with fibrosis progression, as well as how liver fibrosis progression influences Mn and Se distribution in the body. The natural history of NAFLD and its progression to cirrhosis represents a complex interaction between genetic and environmental factors.(65) This highlights the need for future research and interventions to reduce disparities in exposure-disease risk.