Hepatocellular carcinoma (HCC) is the commonest primary hepatic malignancy worldwide. Current serum diagnostic biomarkers, such as alpha-fetoprotein, are expensive and insensitive in early tumor diagnosis. Urinary biomarkers differentiating HCC from chronic liver disease would be practical and widely applicable. Using an 11.7T nuclear magnetic resonance system, urine was analyzed from three well-matched subject groups, collected at Jos University Teaching Hospital (JUTH), Nigeria. Multivariate factor analyses were performed using principal components analysis (PCA) and partial least-squares discriminant analysis (PLS-DA). All patients were of Nigerian descent: 18 hepatitis B surface antigen (HBsAg)-positive patients with HCC, 10 HBsAg positive patients with cirrhosis, and 15 HBsAg negative healthy Nigerian controls. HCC patients were distinguished from healthy controls, and from the cirrhosis cohort, with sensitivity/specificity of 100%/93% and 89.5%/88.9%, respectively. Metabolites that most strongly contributed to the multivariate models were creatinine, carnitine, creatine and acetone. Urinary (1)H MRS with multivariate statistical analysis was able to differentiate patients with HCC from normal subjects and patients with cirrhosis. Creatinine, carnitine, creatine and acetone were identified as the most influential metabolites. These findings have identified candidate urinary HCC biomarkers which have potential to be developed as simple urinary screening tests for the clinic.