Proton magnetic resonance spectroscopy (1H MRS) was used to assess chronic liver disease non-invasively. 1H MRS was optimised to study intact liver tissue in vitro and used to derive hepatic metabolic profiles in murine models of non-alcoholic fatty liver disease. The technique was applied to liver biopsy specimens from 47 patients with chronic hepatitis C. In vitro MRS data were used to inform analysis and aid interpretation of in vivo MR spectra. Findings were validated in a separate cohort of 59 patients in vivo. Diagnostic performance of other emerging non-invasive techniques, ultrasound and serum markers, were compared with histology in 67 patients with chronic hepatitis C. Finally, 38 patients underwent comparative 1H MRS, transient elastography and AST to platelet ratio index (APRI). Multivariate pattern recognition enabled distinction of murine liver tissue by strain and diet. There was correct prediction of histologically-defmed steatohepatitis from steatotic and normal samples. Disease severity was correctly predicted in 46 of 47 human liver samples. Lipid and choline resonances contributed predominantly to statistical models. 1H MRS indices of total lipid, lipid saturation and polyunsaturation, and choline were derived in vitro and validated in vivo. Total lipid assessed by 1H MRS was higher in patients with genotype 3 than genotype 1. The degree of fatty acid polyunsaturation decreased with fibrosis stage and steatosis. In 67 patients with hepatitis C, APRI combined with either transient elastography or the serum panel marker, ELF, effectively predicted fibrosis stage, while three or more tests combined led to redundancy of information. Microbubble ultrasound performed least well. In the smaller cohort, 1H MRS improved prediction of hepatic fibrosis when combined with transient elastography or APRI. In conclusion, assessment of hepatic lipid composition by 1H MRS is of biological importance in chronic liver disease and complements data conveyed by other modalities.