Background
Magnesium (Mg) is essential in metabolic and inflammatory pathways that are central to chronic liver disease and hepatocarcinogenesis. However, the specific pathways remain incompletely understood, particularly across the spectrum of liver disease severity.
Objectives
To define the association of blood Mg and Mg-associated metabolomic signatures with the progression from no liver disease, metabolic dysfunction-associated steatotic liver disease (MASLD), liver cirrhosis, to liver cancer.
Methods
We conducted a matched case-control study in the Mass General Brigham Biobank with electronic healthcare data and blood samples. For 110 liver cancer cases, we matched 1:1 no liver disease, MASLD, and liver cirrhosis controls by age at blood collection, sex, year, and latency time from blood collection to disease. We measured pre- and peri-diagnosis (mean 2.14 years) plasma Mg using Inductively Coupled Plasma Mass Spectrometry. Nightingale metabolomics data were used to derive Mg-associated metabolic signatures via elastic net regression. Associations of Mg and Mg-associated metabolic signatures with liver cancer were assessed using conditional logistic regression models with multivariable adjustment. Mediation analyses were conducted to quantify the proportion of the association mediated by Mg-associated metabolomic signatures.
Results
Higher circulating Mg was associated with lower odds of liver cancer compared with all control groups combined (multivariable OR [95% CI] per SD log-transformed Mg = 0.77 [0.61, 0.96]) and compared with each liver disease control group (OR [95% CI] vs. MASLD = 0.76 [0.57, 1.01]); OR [95% CI] vs. liver cirrhosis = 0.76 [0.58, 1.00]). Mg was strongly associated with lower glucose and higher lipoprotein lipid concentrations (FDR < 0.05; N metabolite = 8). Mg-associated metabolomic signature discriminated liver cancer cases from controls (R² = 0.13 in 30% testing set) and showed graded differences across disease stages (P-diff with liver cancer < 0.05 for all comparisons). This signature was strongly associated with lower odds of liver cancer (multivariable OR [95%CI] = 0.10 [0.04, 0.24]). Mediation analyses demonstrated significant indirect effects of Mg on liver cancer risk through the metabolomic signature (multivariable average causal mediation effect [95%CI]= -0.024 [-0.040, -0.011]; proportion mediated = 60%, P = 0.049).
Conclusions
Blood Mg is inversely associated with liver cancer risk and is linked to a metabolomic signature that partially mediates this relationship. These findings suggest that Mg may play a role in liver carcinogenesis via glycolysis and lipoprotein lipids, highlighting circulating Mg and its related metabolic profile as potential biomarkers for liver cancer prevention across the continuum of liver disease.

Association of the metabolites constituting the metabolic signature with magnesium (Mg) and liver cancer risk.