Introduction Hepatocellular carcinoma (HCC) is characterised by increasing incidence, late diagnosis and high mortality rates. Changes in liver metabolism during HCC development may alter circulating metabolite levels, which could be identified years before HCC diagnosis. Also, a detailed understanding of metabolic dysfunction in HCC may lead to the identification of better prevention strategies. In this context, we aimed to characterise pre-diagnostic circulating biomarkers of HCC risk using an LC-MS untargeted metabolomics approach. Methods A nested case-control study was conducted within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort of > 520,000 participants from 10 European countries. Extensive metabolic profiling was conducted using liquid chromatography-mass spectrometry (LC-MS) and compared between serum samples of 129 HCC cases and their 1:1 matched controls. Multivariable univariate conditional logistic regression models adjusted for anthropometric factors, physical activity, alcohol intake, smoking status and education were used to estimate HCC risk associated with each biomarker and calculated as odds ratios (OR) per 1 standard deviation of log-transformed value. Results Out of 9206 detected metabolic features, 220 significantly discriminated HCC cases from controls after Benjamini-Hochberg correction for multiple testing. Among 37 key compounds that were annotated, 14 were identified with high confidence using authentic chemical standards. Among those, retinol, dehydroepiandrosterone sulfate, glycerophosphocholine, γ-carboxyethylhydroxychroman and creatine were inversely associated with HCC risk (ORs ranging from 0.27 to 0.56). Nine metabolites exhibited positive associations (OR = 6.78 to 2.04): 7-methylguanine, glycocholic acid, glycocheno-deoxycholic acid, L,L-cyclo(leucyl-prolyl), sphingosine, p-hydroxyphenyllactic acid, isatin, N1-acetylspermidine and tyrosine. The remaining 23 annotated metabolites were mainly represented by phospoholipids (mostly inverse associations) and two bilirubin metabolites (positive associations). Conclusions Our findings indicate that dietary factors, altered cholesterol and phospholipid metabolism, DNA damage, and gut bacterial metabolism appear to be involved in HCC development. If validated, these profile changes may be explored as markers of early HCC diagnosis.