The ability to document the extent of hepatic injury and predict the outcome of fulminant hepatic failure would be helpful in identifying those patients who might benefit from liver transplantation. The aim of the present study was to determine whether in vivo phosphorus-31 magnetic resonance spectroscopy (31P MRS) accurately assesses the severity of liver damage and is of prognostic value in a D-galactosamine (D-galN)-induced model of acute liver failure. Adult male Sprague-Dawley rats (n = 36) received an intraperitoneal dose of D-galN (1.0 g/kg), and MRS examinations were performed at peak (48 hours) and in subsequent experiments, just prior to peak (30 hours) hepatic injury. Rats not exposed to D-galN served as controls. The concentration of hepatic phosphorylated metabolites decreased in proportion to the severity of liver injury at 48 hours. Significant correlations were detected between hepatic adenosine triphosphate (ATP) and serum aspartate aminotransferase, bilirubin, and percentage of hepatocyte necrosis identified histologically (r = -.91, -.74, and -.92, respectively; p < .001). Prior to peak hepatic injury (30 hours), 31P MRS was able to predict with 100% accuracy those rats that would survive (ATP > 2.3 mM) and those that would not (ATP < 1.5 mM). When an intermediate cutoff value of 2.0 mM was selected, ATP levels were able to correctly predict survival and death with 80% and 60% accuracy, respectively. These findings indicate that hepatic ATP levels as measured by 31P MRS provide a noninvasive indication of the severity of liver damage and serve as a useful prognostic indicator of outcome in this model of acute liver failure.