Background Animal experimental studies have shown that arsenic exposure contributes to hepatic lipid accumulation, but epidemiological findings are inconsistent. Moreover, the role of arsenic metabolism is still unclear.

Objective To evaluate the potential causal association between arsenic metabolism and non-alcoholic fatty liver disease (NAFLD).

Methods A total of 1020 participants from the Dongfeng-Tongji cohort with urinary arsenic metabolites and genotype data were included in the present study (NAFLD group, n=529; non- NAFLD group, n=491). Epidemiological information was obtained by questionnaire survey, liver ultrasound was obtained by physical examination, arsenic metabolites in urine were measured by high-performance liquid chromatography-inductively coupled plasma mass spectrometry, and DNA from leukocytes was extracted for genome-wide genotype. NAFLD was diagnosed if the following two criteria were met: (1) positive fatty liver according to abdominal ultrasonography; (2) excluding participants reporting history of excessive alcohol consumption (≥30 g·d⁻¹ for men; ≥20 g·d⁻¹ for women) and/or fatty liver with other known causes. Genetic risk score (GRS) and weighted genetic risk score (w-GRS) were constructed using single nucleotide polymorphisms (SNPs) related to arsenic metabolism reported in previous studies to predict the estimated arsenic metabolism. Logistic regression models were used to analyze the association between arsenic metabolism and NAFLD; linear regression models were used to analyze the association between GRS/w-GRS and arsenic metabolism, and Mendelian randomization analysis was performed using GRS method, inverse variance weighting, Egger regression, and weighted median.

Results The mean age of the 1020 participants was (68.14±7.45) years, of which 64% were female, and 529 (51.9%) were NAFLD cases. The median (P₂₅, P₇₅) level of total arsenic in urine was 18.34 (11.93, 27.14) μg·L⁻¹ with a geometric mean and standard deviation of (15.86±1.81) μg·L⁻¹. The proportions of inorganic arsenic (iAs%), monomethylarsenic (MMA%), and dimethylarsenic (DMA%) in the total arsenic were 13.90%±9.90%, 9.49%±4.97%, and 76.60%±11.00%, respectively. After adjustment for potential confounders, the ORs (95%CIs) for NAFLD risk by per standard deviation increase in iAs% and MMA% were 1.21 (1.06, 1.38) and 0.62 (0.51, 0.74) respectively. Each unit increase in GRS constructed from 77 SNPs was associated with a 0.16% increase in MMA% and a 0.19% decrease in DMA%, and each unit increase in w-GRS was associated with a 0.17% increase in MMA% and a 0.14% decrease in DMA%. After further exclusion of SNPs with linkage disequilibrium (r²>0.3) and pleiotropic effect, a total of 25 SNPs were included in the Mendelian randomization analysis. The GRS method showed that the OR (95%CI) for NAFLD risk by per unit increase in MMA% expectation was 0.95 (0.90, 0.99), and the inverse variance weighting method also showed a significant association between MMA% and NAFLD, with OR (95%CI) of 0.91 (0.84, 0.99).

Conclusion There is a negative causal association between MMA% and NAFLD.