Urinary Volatile Organic Compound Metabolites Are Associated With MASLD/MASH in Humans and Induce Steatosis in Liver Organoids

ABSTRACT

Background

Volatile organic compound (VOC) exposure is an environmental health concern and could, through the liver exposome, be associated with metabolic dysfunction associated steatotic liver disease (MASLD) progression.

Methods

We analysed NHANES 2017–2020, a U.S. population‐based cohort with controlled attenuation parameter (CAP), liver stiffness measurement (LSM) and urinary VOC metabolites. Participants with viral hepatitis, excess alcohol use or missing urine creatinine were excluded. MASLD was defined as CAP ≥ 275 dB/m with metabolic dysfunction, at‐risk MASH as FAST ≥ 0.35 and increased LSM as ≥ 8 kPa. Weighted quantile sum (WQS) regression assessed associations between VOC metabolites and outcomes, adjusting for age, sex, smoking and alcohol. Phenylglyoxylic acid (PGA) and mandelic acid (MA) were further examined using logistic regression for the MA/(MA and PGA) ratio and human liver organoids.

Results

The cohort comprised 2004 participants (41.4% MASLD, 5.4% at‐risk MASH, 9.7% LSM ≥ 8 kPa). Higher VOC metabolite levels were associated with increased risk of MASLD (aOR 1.47 per quartile, 95% CI 1.06–2.04) and at‐risk MASH (aOR 2.69 per quartile, 95% CI 1.23–5.87), primarily driven by N‐Acetyl‐S‐(2‐carboxyethyl)‐L‐cysteine (CEMA) and N‐Acetyl‐S‐(3‐hydroxy‐1‐methylpropyl)‐L‐cysteine (HMPMA), with inverse associations driven by PGA. No significant associations were found for increased LSM, yet a higher MA/(MA + PGA) ratio was associated with increased risk for at‐risk MASH and LSM ≥ 8 kPa. In human liver organoids, PGA exposure increased lipid droplet number and size.

Conclusion

Urinary VOC metabolites show distinct associations with MASLD and at‐risk MASH in the general population. CEMA and HMPMA were associated with increased risk, consistent with prior links to metabolic dysfunction. PGA induced steatosis in liver organoids, suggesting poor metabolising of styrene and ethylbenzene and intracellular PGA accumulation.

Key Points

• Urinary volatile organic compound (VOC) metabolites are significantly associated with metabolic liver disease.

• Elevated urinary levels of CEMA and HMPMA act as the primary drivers for increased MASH and MASLD risk.

• Urinary PGA is inversely associated with liver disease risk, yet directly induces steatosis in human liver organoids.

• Impaired xenobiotic metabolism may make individuals more vulnerable to environmental exposome‐induced liver damage.

Lay Summary

Volatile organic compounds are chemicals that enter the human body through inhalation, ingestion and dermal contact. Urine analysis revealed that exposure to these compounds was associated with MASLD and at‐risk MASH in the general population. Furthermore, when tested in a laboratory setting, these compounds were shown to cause fat buildup in artificial human liver models.

Abbreviations

2HPMA

N‐Acetyl‐S‐(2‐hydroxypropyl)‐L‐cysteine

2MHA

2‐Methylhippuric acid

34‐MHA

3‐ and 4‐Methylhippuric acid

3HPMA

N‐Acetyl‐S‐(3‐hydroxypropyl)‐L‐cysteine

AAMA

N‐Acetyl‐S‐(2‐carbamoylethyl)‐L‐cysteine

AMCC

N‐Acetyl‐S‐(N‐methylcarbamoyl)‐L‐cysteine

ATCA

2‐Aminothiazoline‐4‐carboxylic acid

CAP

controlled attenuation parameter

CEMA

N‐Acetyl‐S‐(2‐carboxyethyl)‐L‐cysteine

CI

confidence interval

CYMA

N‐Acetyl‐S‐(2‐cyanoethyl)‐L‐cysteine

CYP

cytochrome P450

DHBMA

N‐Acetyl‐S‐(3,4‐dihydroxybutyl)‐L‐cysteine

FAST

Fibroscan AST

HDL

high‐density lipoprotein

HMPMA

N‐Acetyl‐S‐(3‐hydroxy‐1‐methylpropyl)‐L‐cysteine

ICO

intrahepatic cholangiocyte organoid

LASSO

least absolute shrinkage and selection operator

LSM

liver stiffness measurement

MA

mandelic acid

MASH

metabolic dysfunction associated steatohepatitis

MASLD

metabolic dysfunction associated steatotic liver disease

MEC

Medical Ethics Committee

MHBMA3

N‐Acetyl‐S‐(4‐hydroxy‐2‐butenyl)‐L‐cysteine

NCHS

National Center for Health Statistics

NHANES

National Health and Nutrition Examination Survey

OR

odds ratio

PFAS

Per and polyfluoroalkyl substances

PGA

Phenylglyoxylic acid

SBMA

N‐Acetyl‐S‐(benzyl)‐L‐cysteine

VOC

volatile organic compound

WQS

weighted quantile sum

1. Introduction

Exposure to volatile organic compounds (VOC) through inhalation, ingestion and dermal contact is considered an environmental health concern [1]. VOCs are organic chemicals that have high vapour pressure at room temperature allowing them to evaporate easily. They originate from a wide range of natural sources (e.g., forest fires), indoor sources (e.g., cooking and heating and personal care products like hair spray, perfumes and deodorants) and other human‐made sources (e.g., industrial processes, vehicle emissions, household products, building materials and cigarette smoke) [1, 2, 3, 4, 5]. VOC exposure is considered harmful and has been associated with metabolic dysfunction, low‐grade systemic inflammation, cardiovascular disease and several types of cancers [6, 7, 8, 9, 10, 11].

VOC exposure has the potential to influence the liver exposome through multiple pathways and can be considered part of the liver exposome that includes diet, infections, medications, alcohol consumption, air pollution and exposure to occupational chemicals [12]. The liver exposome has some important consequences, for example high levels of air pollution increase the risk of MASLD; perfluorinated alkyl substances (PFAS) impact liver metabolism resulting in unfavourable lipid profiles; and microplastics were linked to fibrogenesis [13, 14, 15, 16]. Although VOC exposure has also been linked to increased alkaline phosphatase and metabolic dysfunction, associations with MASLD and fibrosis were not investigated [17, 18].

Exposure to VOCs can be measured in urine as VOCs are primarily metabolised by the liver through its xenobiotic metabolism into different compounds, which are subsequently excreted in urine [1]. In this study we therefore aimed to assess the impact of VOC metabolite concentrations in urine on impaired liver health in the form of MASLD, at‐risk MASH and increased liver stiffness by using the 2017–2020 National Health and Nutrition Examination Survey (NHANES) cohort. The functional relevance of identified associations was subsequently verified in human liver organoids.

2. Methods

2.1. Study Population

NHANES 2017–2020 data was used for this study. NHANES is a retrospective ongoing cohort study focusing on health and nutritional states of the United States population. Data collection included interviews, physical examination and clinical measurements. In addition, blood and urine samples were collected. Especially relevant for this study were the liver health assessment by FibroScan and VOC metabolite analysis in the urine samples. Further details on the aims, procedures and design are available elsewhere [19, 20]. For this specific study, participants with excessive alcohol consumption (> 20 g per day in female/> 30 g per day in male), viral hepatitis or missing data on urine creatine concentrations and/or outcomes were excluded. Data is publicly available from the NHANES database (https://www.cdc.gov/nchs/nhanes/index.htm).

2.2. Liver Health

Transient elastography was performed using the FibroScan model 502 V2 Touch (FibroScan, Echosens, Paris) in individuals who were requested to fast for at least 3 h. There was access to both an M‐probe and XL‐probe, which were used according to the device's instructions. Steatosis was defined as CAP ≥ 275 dB/m [21]. Steatosis together with metabolic dysfunction was defined as MASLD [22]. LSM was considered valid if at least 10 measurements were obtained with an interquartile range < 30% in the absence of congestive heart failure. Increased LSM was defined as LSM ≥ 8 kPa [23, 24]. At‐risk metabolic dysfunction associated steatohepatitis (MASH) was defined according to the FibroScan‐AST score (FAST) with a FAST threshold of ≥ 0.35 [25].

2.3. Urine VOC Metabolites

Urine samples were collected, processed and stored at −30°C prior to shipment to the CDC's National Center for Environmental Health for analysis. Sample handling and quality control procedures followed the NHANES Laboratory Procedures Manuals (2017–2020) and meet Clinical Laboratory Improvement Amendments (1988) standards. The VOC metabolites that were analysed were part of the biomonitoring program of NHANES and were selected based on the known links with specific toxins. VOC metabolites were quantified in urine using ultra‐performance liquid chromatography coupled with electrospray ionisation tandem mass spectrometry. Only metabolites detected in ≥ 80% of samples were included; concentrations below the limit of detection were imputed as lower limit of detection/√2. The complete set of investigated metabolites, including the parent compounds, is available in Table 1. Full methodological details are available at https://wwwn.cdc.gov/nchs/data/nhanes/public/2017/labmethods/UVOC‐J‐MET‐508.pdf.

TABLE 1.

Overview of volatile organic compound urinary metabolite investigated in this study alongside its parent compound.

Parent compound	VOC metabolite detected in urine	Abbreviation
Acrolein	N‐Acetyl‐S‐(2‐carboxyethyl)‐L‐cysteine	CEMA
	N‐Acetyl‐S‐(3‐hydroxypropyl)‐L‐cysteine	3HPMA
Acrylamide	N‐Acetyl‐S‐(2‐carbamoylethyl)‐L‐cysteine	AAMA
Acrylonitrile	N‐Acetyl‐S‐(2‐cyanoethyl)‐L‐cysteine	CYMA
1,3‐Butadiene	N‐Acetyl‐S‐(3,4‐dihydroxybutyl)‐L‐cysteine	DHBMA
	N‐Acetyl‐S‐(4‐hydroxy‐2‐butenyl)‐L‐cysteine	MHBMA3
Crotonaldehyde	N‐Acetyl‐S‐(3‐hydroxy‐1‐methylpropyl)‐L‐cysteine	HMPMA
Cyanide	2‐Aminothiazoline‐4‐carboxylic acid	ATCA
N, N‐ Dimethylformamide, methyl isocyanate	N‐Acetyl‐S‐(N‐methylcarbamoyl)‐L‐cysteine	AMCC
Ethylbenzene and styrene	Phenylglyoxylic acid	PGA
	Mandelic acid	MA
Propylene oxide	N‐Acetyl‐S‐(2‐hydroxypropyl)‐L‐cysteine	2HPMA
Toluene, benzyl alcohol	N‐Acetyl‐S‐(benzyl)‐L‐cysteine	SBMA
Xylene	2‐Methylhippuric acid	2MHA
	3‐ and 4‐Methylhippuric acid	3,4‐MHA

Note: This table shows the VOC urinary metabolites investigated in this study.

Abbreviation: VOC, volatile organic compound.

2.4. Covariates

Blood samples were analysed for various outcomes including liver enzymes, triglycerides, HDL and glucose. Research assistants systematically collected data. The interview included an alcohol questionnaire and details on medical history, such as history of heart failure. Metabolic dysfunction was defined according to the set of criteria provided for the definition of MASLD and included the presence of at least one of the following criteria: overweight, diabetes, hypertension, high waist circumference, dyslipidemia [22].

2.5. Liver Organoids Model

Intrahepatic cholangiocyte organoids (ICOs) were generated from human liver tissue obtained from organ donors during liver transplantation procedures at the Erasmus Medical Center Rotterdam with informed consent and cultured in defined expansion medium as previously described [26, 27]. Two scenarios were tested. Under homeostatic conditions, ICOs were exposed for 48 h to either vehicle, 50 μM phenylglyoxylic acid (PGA) or 50 μM mandelic acid (MA). The 50 μM dose did not induce significant cytotoxicity in vitro. In a steatotic model, organoids were first treated with free fatty acids (oleic + palmitic acid) for 48 h, followed by PGA or MA for an additional 48 h [28]. Organoids were stained with AdipoRed to visualise lipid droplets and Hoechst 33342 for nuclei. Confocal images were acquired and lipid accumulation quantified using ImageJ. In supplementary methods there is a more detailed description of the experiments.

2.6. Statistical Analysis

VOC metabolite concentrations were first transformed to VOC metabolite concentration per mg urine creatinine to account for differences in urine concentrations. Weighted quantile sum (WQS) regression analysis was performed using the corrected urine metabolites, a common approach when assessing the effect of different chemicals on health outcomes [29]. The WQS regression analysis was conducted using a random 60% of the dataset for model training and 40% for validation. The model was fit with 100 bootstrap samples and 100 repeated holdout validations, simultaneously assessing both positive and negative associations within the mixture. The analyses were adjusted for age, sex, smoking (yes/no) and alcohol consumption (gram/day).

In sensitivity analyses, the number of urinary metabolites of VOCs was reduced by first applying the least absolute shrinkage and selection operator (LASSO) analysis, which helped identify the metabolites that contributed most to the associations with MASLD, at‐risk MASH and increased LSM. Furthermore, based on the findings in this study, additional investigation of the relative excretion of MA compared to PGA by using the ratio of MA to PGA + MA. Important to note is that MA is a precursor of PGA and is involved in the styrene and ethylbenzene metabolism.

Finally, for liver organoid experiments, lipid accumulation was quantified by calculating the relative number and area of lipid droplets per cell, obtained by dividing total droplet count or area by the number of nuclei in each image. Data were normalised to the respective control condition (set to 100%). In addition, the surface area of individual lipid droplets (μm²) was measured and size distributions across treatment groups were visualised using scatter plots.

The analysis was performed according to a prespecified analysis plan and sensitivity analyses were performed to assess the robustness of the results. Analyses were performed in R version 4.0.4 (Foundation for Statistical Computing, Vienna, Austria). p‐values < 0.05 were considered statistically significant.

2.7. Ethics

NHANES procedures and protocols were approved by the NCHS Research Ethics Review Board. Participants of the NHANES 2017–2020 cycle provided informed consent. Ethical approval for the use of human liver tissue was obtained from the Erasmus MC Medical Ethics Committee and informed consent for participation in medical research was provided by all donors through the national donation registry (MEC‐2014‐060). This study was conducted according to the principles as outlined in the Declaration of Helsinki and Istanbul.

3. Results

We included 2176 participants from the NHANES 2017–2020 cycle who had data available on urinary VOC metabolite concentrations as part of the biomonitoring program and had complete data on liver outcomes. Of them, 38 were excluded for unavailable data on urine creatinine concentrations, 113 for excessive alcohol consumption and 21 for the presence of viral hepatitis, leaving 2004 participants for final analysis (Figure S1). Among these included participants, the median age was 50 years [34–63], 49% were male and metabolic dysfunction was highly prevalent (Table 2). The prevalence of MASLD, at‐risk MASH and LSM ≥ 8 kPa in this subset was 41.4%, 5.4% and 9.7%, similar to the overall prevalence in the NHANES 2017–2023 NHANES cohort reported previously [30].

TABLE 2.

Participant characteristics.

n	2004
Demographics
Age	50 [34, 63]
Male	981 (49.0)
Ethnicity
Asian	241 (12.0)
Black	506 (25.2)
Hispanic	469 (23.4)
Other	108 (5.4)
White	680 (33.9)
Lifestyle
Alcohol consumption (g/day)	0.46 [0.0, 3.5]
Smoking	310 (15.5)
Metabolic dysfunction
(abdominal) overweight or obesity	1600 (80.2)
(pre)diabetes	990 (49.4)
Hypertension	972 (51.2)
Liver health
MASLD	830 (41.4)
At‐risk MASH	109 (5.4)
LSM ≥ 8 kPa	194 (9.7)
CAP (dB/m)	262 (61)
LSM (kPa)	4.9 [4.1, 6.1]

Note: Data is presented as mean (SD), median [P25–P75] or n and percentage.

Abbreviations: LSM, Liver stiffness measurement; MASH, Metabolic dysfunction–associated steatohepatitis; MASLD, Metabolic dysfunction–associated steatotic liver disease.

We performed WQS analysis, where the quantiles of 15 different urinary metabolites were weighted individually for their relative contribution to positive and negative associations with MASLD, at‐risk MASH and LSM ≥ 8 kPa. The analysis was adjusted for age, sex, smoking and alcohol consumption (Table 3; Table S1). The WQS analysis identified differential urinary metabolite profiles that were either associated with increased or decreased risk for MASLD or at‐risk MASH. For MASLD, increased risk was observed (aOR 1.47 per quartile, 95% CI 1.06–2.04; accounted for 53% by CEMA, HMPMA, ATCA and 3HPMA) and a significant decreased risk (aOR 0.56 per quartile, 95% CI 0.40–0.79; 58% accounted for by MHBMA3, CYMA and PGA). For at‐risk MASH, we demonstrated a stronger increased risk (aOR 2.69 per quartile, 95% CI 1.23–5.87; 54% accounted for by HMPMA, CEMA, MA and 2MHA) and a reduced risk (aOR 0.34 per quartile, 95% CI 0.15–0.74; 48% accounted for by PGA and MHBMA3). On the contrary, for LSM ≥ 8 kPa, no significant associations were found when focused on WQS increased risk (aOR 1.52 per quartile, 95% CI 0.74–3.12; 56% by 3.4MHA, HMPMA, MA and 3HPMA) and WQS decreased risk (aOR 0.61 per quartile, 95% CI 0.31–1.24; 53% by PGA, 2MHA and 2HPMA). The weights for each urinary metabolite of VOCs in the associations with MASLD, at‐risk MASH and increased LSM are presented in Figure 1.

TABLE 3.

The associations between urinary metabolites assessed with WQS analysis separated for a positive and negative association.

	Cases/n	Positive WQS (OR, 95% CI)	Negative WQS (OR, 95% CI)
MASLD	830/2004	1.47 (1.06–2.04)	0.56 (0.40–0.79)
At‐risk MASH	109/2004	2.69 (1.23–5.87)	0.34 (0.15–0.74)
LSM ≥ 8 kPa	194/2004	1.52 (0.74–3.12)	0.61 (0.30–1.24)

Note: Results were obtained using logistic regression with weighted quantile sum (WQS) regression to evaluate the association between urinary VOC metabolite mixtures and three outcomes: MASLD, at‐risk MASH and increased LSM. Urinary metabolites of VOCs were transformed into weighted quantiles and combined into positive WQS and negative WQS to capture mixture effects in positive and negative directions, respectively. Models were adjusted for age, sex, smoking status and alcohol intake. Effect estimates for the covariates are available in Table S1.

Abbreviations: LSM, Liver stiffness measurement; MASH, Metabolic dysfunction–associated steatohepatitis; MASLD, Metabolic dysfunction–associated steatotic liver disease; VOC, volatile organic compounds; WQS, weighted quantile sum.

FIGURE 1.

Contribution in percentage towards the positive and negative associations for each volatile organic compound urinary metabolite in WQS analysis. This figure shows the percentage contribution for each VOC urinary metabolite in associations with MASLD, at‐risk MASH and LSM ≥ 8 kPa. The overall associations for the entire VOC urinary metabolite mixture are available in Table 3. The dotted line is placed at 6.67% and indicates the level in case each urinary metabolite contributed equally to the overall associations. This figure illustrates that various urinary metabolites are linked to increased risk of MASLD, at‐risk MASH and increased LSM, whereas PGA is the only consistent parameter contributing to a lower risk. 2HPMA, N‐Acetyl‐S‐(2‐hydroxypropyl)‐L‐cysteine; 2MHA, 2‐Methylhippuric acid; 3,4‐MHA, 3‐ and 4‐Methylhippuric acid; 3HPMA, N‐Acetyl‐S‐(3‐hydroxypropyl)‐L‐cysteine; AAMA, N‐Acetyl‐S‐(2‐carbamoylethyl)‐L‐cysteine; AMCC, N‐Acetyl‐S‐(N‐methylcarbamoyl)‐L‐cysteine; ATCA, 2‐Aminothiazoline‐4‐carboxylic acid; CEMA, N‐Acetyl‐S‐(2‐carboxyethyl)‐L‐cysteine; CYMA, N‐Acetyl‐S‐(2‐cyanoethyl)‐L‐cysteine; DHBMA, N‐Acetyl‐S‐(3,4‐dihydroxybutyl)‐L‐cysteine; HMPMA, N‐Acetyl‐S‐(3‐hydroxy‐1‐methylpropyl)‐L‐cysteine; MHBMA3, N‐Acetyl‐S‐(4‐hydroxy‐2‐butenyl)‐L‐cysteine; SBMA, N‐Acetyl‐S‐(benzyl)‐L‐cysteine; MA, Mandelic acid; PGA, Phenylglyoxylic acid; VOC, Volatile organic compound; LSM, Liver stiffness measurement; MASLD, Metabolic dysfunction–associated steatotic liver disease; MASH, Metabolic dysfunction–associated steatohepatitis; WQS, Weighted quantile sum.

In sensitivity analysis we repeated the analysis using only LASSO selected urinary metabolites of VOCs. Here we found similar overall results for the positive and negative WQS (Table S2). In this analysis, also the negative association between urinary metabolites of VOCs with increased LSM was significant (aOR 0.72, 95% CI 0.52–0.99), but the positive association was not. Due to reduced number of investigated metabolites (five for MASLD, 14 for at‐risk MASH and four for increased LSM), the weight attributed to the overall association increased with CEMA, DHBMA and HMPMA being most important for the observed risk increase and PGA for the observed risk decrease. The full weights are available in Figure S2.

Next, since MA is a precursor of PGA in the styrene and ethylbenzene metabolism, we investigated the relative excretion of MA compared to PGA by using the ratio of MA to PGA + MA adjusted for age, sex, smoking and alcohol consumption. We demonstrate increased risk when MA to PGA + MA ratio increased for at‐risk MASH (aOR 1.39 per 10%; 95% CI 1.08–1.54) and LSM ≥ 8 kPa (aOR 1.16 per 10%, 95% CI 1.01–1.34), but not for MASLD (aOR 1.04 per 10%, 95% CI 0.95–1.15).

Finally, we experimentally evaluated the impact of PGA and MA on lipid accumulation in human liver‐derived organoids (Figure 2). We first preliminarily tested the effects of both compounds on host cell viability using a series of concentrations (Figure S3) and selected 50 μM for follow‐up experiments in organoids. PGA treatment led to a significant increase in intracellular lipid accumulation in organoids cultured in both expansion medium and FFA‐enriched medium (Figure 2A,B). Specifically, the average number of lipid droplets per cell was significantly elevated in the control group following PGA exposure (p < 0.0001, Figure 2C), whereas in the FFA‐induced group, a similar trend was observed but without reaching statistical significance (Figure 2D). A consistent increase in average lipid droplet area per cell was also detected in the homeostatic condition after PGA treatment (p < 0.0001, Figure 2E), while no significant difference was found in the FFA‐exposure group (Figure 2F). Notably, lipid droplet size was significantly increased in both conditions after PGA exposure (Figure 2G). However, MA treatment did not affect lipid droplet number or area in the homeostatic condition but led to a slight increase in droplet size under the FFA‐exposure condition (Figure 2G).

FIGURE 2.

The effect of phenylglyoxylic acid (PGA) and mandelic acid (MA) on fatty acid‐induced lipid droplet accumulation in human‐derived liver organoids. (A) Representative confocal images of lipid droplets (red, AdipoRed) and nuclei (blue, Hoechst) in liver organoids cultured in expansion medium representing homeostatic condition (vehicle control; Ctrl), treated with 50 uM PGA (Ctrl+ PGA) and 50 uM MA (Ctrl+ MA) for 48 h. Scale bar, 25 and 50 μm. (B) Representative confocal images of lipid droplets (red, AdipoRed) and nuclei (blue, Hoechst) in liver organoids cultured in free fatty acid‐enriched medium modelling steatosis (vehicle control; FA), treated with 50uM PGA (FA + PGA) or 50uM MA (FA + MA) for 48 h. Scale bar, 25 μm and 50 μm. (C) Average lipid droplet number per cell in the Ctrl group (quantified from 13 independent images of four biological replicates), Ctrl + PGA group (13 images from four biological replicates) and Ctrl + MA group (eight images from four biological replicates), normalised to the Ctrl group (set as 100%). (D) Average lipid droplet number per cell in the FA group (nine images from four biological replicates), FA + PGA group (11 images from four biological replicates) and FA + MA group (11 images from four biological replicates), normalised to the FA group (set as 100%). (E) Average lipid droplet area per cell in the Ctrl group (13 images from four biological replicates), Ctrl + PGA group (13 images from four biological replicates) and Ctrl + MA group (eight images from four biological replicates), normalised to the Ctrl group (set as 100%). (F) Average lipid droplet area per cell in the FA group (nine images from four biological replicates), FA + PGA group (11 images from four biological replicates) and FA + MA group (11 images from four biological replicates), normalised to the FA group (set as 100%). (G) The size of captured lipid droplets in Ctrl group (n = 113 lipid droplets from 13 images of four biological replicates), Ctrl + PGA group (n = 4283 lipid droplets from 13 images, four biological replicates), Ctrl + MA group (n = 51 lipid droplets from eight images, four biological replicates), FA group (n = 12 356 lipid droplets from nine images, four biological replicates), FA + PGA group (n = 18 911 lipid droplets from 11 images, four biological replicates), FA + MA group (n = 10 909 lipid droplets from 11 images, four biological replicates). Data are presented as mean ± SEM. Statistical analysis was performed using Mann–Whitney test. *p < 0.05, **p < 0.01, ***p < 0.001. ****p < 0.0001. Ctrl, control group; FA free fatty acid enriched medium group; MA, mandelic acid; PGA, phenylglyoxylic acid.

4. Discussion

We demonstrated that VOC metabolite concentrations in urine were associated with the MASLD spectrum. In particular, higher concentrations of CEMA and HMPMA were associated with a higher risk of MASLD and at‐risk MASH but not with increased LSM. Higher urinary concentrations of primarily PGA were associated with a lower risk of MASLD and at‐risk MASH and higher MA relative to PGA was also associated with LSM ≥ 8 kPa.

HMPMA and CEMA in urine were associated with a higher risk of MASLD and at‐risk MASH. These findings extend on previous evidence identifying metabolites of 1,3‐butadiene and acrolein as an important risk factor for metabolic dysfunction and elevated alkaline phosphatase [10, 17, 18]. These are the metabolites of crotonaldehyde and acrolein, respectively. Acrolein is present in various foods prepared at high temperatures (typically fried or grilled) and is formed from carbohydrates, vegetable oils, animal fats and amino acids during the heating of foods as well as by combustion of petroleum fuels and smoking tobacco products [31]. 1,3‐butadiene is found in synthetic rubber, engine exhaust and cigarette smoke and is considered carcinogenic [32]. Higher exposure to these VOCs could be linked to a suboptimal lifestyle with a diet containing many fried foods; however, acrolein levels did not change much when individuals were on a no‐fry diet [33, 34]. Crucial in the metabolism of acrolein is the NF‐kappa‐B pathway, for which higher activity was previously associated with fibrogenesis and hepatocellular carcinoma, which could contribute to current findings and could support a direct effect on the liver [31, 35]. However, which part of the demonstrated associations is explained by or mediated through metabolic dysfunction remains to be explored.

PGA's parent compounds are ethylbenzene and styrene, which have previously been linked to markers of liver injury and steatosis [36]. Ethylbenzene is primarily the precursor of styrene and is present in engine exhausts, paint thinners and air fresheners [37, 38]. Styrene is one of the most important chemicals used worldwide to manufacture plastics, synthetic rubber (as co‐polymer with butadiene) and resins. Furthermore, it is present in food, tobacco and engine exhaust and has been labelled as possibly carcinogenic [37].

Higher concentrations of PGA in urine were associated with a lower observed risk of MASLD and at‐risk MASH. This association between higher urine PGA and a lower risk of MASLD‐related disease aligns with previous findings showing that higher urine PGA concentrations are associated with a lower risk of metabolic syndrome and its components [18]. Considering the origins of styrene and ethylbenzene, this finding could also be explained by inadequate excretion and (intrahepatic) PGA accumulation, instead of an actual beneficial effect. Indeed, higher relative excretion of MA compared to PGA in urine was associated with increased risk of at‐risk MASH and LSM ≥ 8 kPa, supporting this explanation. In general, around 60%–70% of ethylbenzene and styrene is excreted as MA and the remainder is further metabolised into PGA [37, 38]. This suggests that altered metabolism may drive the seemingly hepatoprotective associations of urinary PGA in an epidemiological setting.

Our hypothesis that PGA itself was not hepatoprotective was further confirmed in our human organoid model, where the potential benefits of PGA were no longer observed. In fact, exposure to PGA induced lipid droplet accumulation and lipid droplet size growth in both the steatosis and control human liver organoids, confirming that PGA itself is not the factor explaining the potential hepatoprotective association of PGA in urine. MA, the precursor of PGA in liver organoids, seems to have less steatogenic characteristics; however, in the steatosis organoid model, it did increase the lipid droplet size. This fuels a hypothesis that lower PGA levels in urine are due to poor metabolising of (metabolites of) ethylbenzene and styrene.

Aligning with this, individuals with CYP2E1 variants have impaired ability for PGA excretion [39, 40]. For example, the CYP2E1*5B variant has been linked to exacerbate the harmful effects of ethanol and other toxins including volatile organic compounds due to increased oxidative stress, endoplasmic reticulum stress and inflammation leading to fibrosis [41, 42]. This might be the driver for the demonstrated association between higher concentrations of PGA in urine with lower MASLD‐related disease risk and suggests that individuals with impaired xenobiotic metabolism may be more prone to exposome‐induced liver disease. Additional studies are required to further focus on genetic variants, blood and urine concentrations of styrene and ethylbenzene metabolites and the consequences for liver health.

This study reveals that VOC metabolite concentrations in urine are associated with MASLD related disease. Although the exact mechanisms remain unknown and likely include alterations in hepatic lipid metabolism, these findings are suggestive of the potential disruptive effects of VOC exposure on liver health and could therefore be considered part of the liver exposome [43]. VOC exposure, which has increased over the years, may therefore help to explain (in addition to the surge of metabolic dysfunction) the increase in impaired liver health and can be considered a public health concern [44]. Also, in the presence of other risk factors such as poor lifestyle, VOC exposure may lower the threshold for the point when steatosis or liver damage occurs, in line with the concept that poor lifestyle interacts with other risk factors regarding mortality in MASLD patients [16, 45].

4.1. Limitations

This is one of the first studies to assess VOC metabolite concentrations in urine and their associations with liver health on a large scale in a general population setting. Nonetheless, the following limitations should be noted. First, this study utilised a cross‐sectional design, which precludes any definitive conclusions regarding causality. Second, the potential confounding or mediating effects of metabolic dysfunction traits (e.g., type 2 diabetes or obesity) were not extensively adjusted for in our primary models. This decision was guided by both statistical and biological considerations. Methodologically, the relatively low number of cases for ‘at‐risk MASH’ and increased liver stiffness combined with the splitting of data required for the repeated holdout procedure, limited the statistical power to include additional covariates without risking model overfitting. Biologically, emerging data suggest that certain VOCs or altered metabolism may directly affect lipid metabolism and induce hepatic oxidative stress, implying that metabolic dysfunction could act as a downstream consequence or mediator rather than a classic confounder [41, 42, 43]. Third, MASLD, at‐risk MASH and increased LSM were assessed non‐invasively and may not have classified all individuals accurately, which could have led to underestimation of the reported effect sizes. Fourth, no blood levels for volatile organic compounds were available, neither was genetic data and concepts such as poor metabolising could not be evaluated. However, through the human liver organoid experiments, we could refute the observed hepatoprotective associations with PGA. Fourth, no data was available on gallstones, which has been previously identified to be associated with MASLD, but could not be explored in this study [46]. Fifth, whilst we could exclude individuals with viral hepatitis, no data were available for the presence of other liver diseases such as autoimmune hepatitis or hemochromatosis, whose prevalence is, however, low in a general population setting. Sixth, alcohol consumption was based on a questionnaire and is likely underreported; since it is used as a confounder only the main results are unlikely to have been affected by this, but the effect size of alcohol itself may be inflated [47]. Finally, simultaneous data on liver health and urinary VOC metabolites were only present in a relatively small NHANES subset of 2176 individuals before exclusions and selection bias could have occurred in this study.

5. Conclusion

Urinary metabolites of VOCs and their relative exposure have distinct associations with MASLD, at‐risk MASH and increased LSM in the general population. In particular, CEMA and HMPMA were associated with increased risk of MASLD and at‐risk MASH and have previously been associated with metabolic dysfunction traits. PGA was the primary metabolite inversely associated with MASLD‐related disease. Contrastingly, PGA induced steatosis in human liver organoids and the seemingly beneficial association might be explained by poor metabolism of styrene and ethylbenzene and intracellular PGA accumulation, which was further supported by the observed increased risk for higher MA excretion relative to PGA.

Author Contributions

Collection of data: Laurens A. van Kleef, Xincheng Li, Qiuwei Pan and Willem P. Brouwer. Study design, data analysis, writing of the manuscript: Laurens A. van Kleef, Xincheng Li, Qiuwei Pan and Willem P. Brouwer. Critical review of the manuscript, writing of the manuscript, approval of final version and approval of submission: All authors.

Funding

This work was supported by the Stichting voor Lever‐en Maag‐Darm Onderzoek.

Conflicts of Interest

H.L.A.J. received grants from AbbVie, Arbutus, Gilead Sciences, Janssen and Roche and is a consultant for Arbutus, Arena, Enyo, Gilead Sciences, GlaxoSmithKline (GSK), Janssen, Merck, Roche, Vir Biotechnology Inc. and Viroclinics F.T. has received research grants (funding to the institution) from AstraZeneca, MSD, Gilead, Agomab. Frank Tacke has received honoraria for consulting or lectures from Gilead, Abbvie, Falk, AstraZeneca, Boehringer, Madrigal, MSD, GSK, Ipsen, Pfizer, Novartis, Novo Nordisk, Sanofi. W.P.B. received speakers fees for Eli Lilly, is part of the advisory board of Novo Nordisk and participates in trials of 89BIO, Boehringer Ingelheim, Novo Nordisk, Inventiva Pharma. The remaining authors declare no conflicts of interest.

Supporting information

Table S1: The associations between urinary metabolites assessed with WQS analysis separated for a positive and negative association including covariates.

Table S2: The associations between LASSO selected urinary metabolites assessed with WQS analysis separated for a positive and negative association.

Figure S1: Flowchart of study.

Figure S2: Contribution in percentage towards the positive and negative associations for LASSO‐selected volatile organic compound urinary metabolite in WQS analysis.

Figure S3: The effects of phenylglyoxylic acid (PGA) and mandelic acid (MA) on viability of human liver Huh7 cells. (A–D).

van Kleef L. A., Li X., Li P., et al., “Urinary Volatile Organic Compound Metabolites Are Associated With MASLD/MASH in Humans and Induce Steatosis in Liver Organoids,” Liver International 46, no. 8 (2026): e70783, 10.1111/liv.70783.

Data Availability Statement

Data are publicly available from the NHANES database (https://www.cdc.gov/nchs/nhanes/index.htm).