Single voxel proton magnetic resonance spectroscopy (1H-MRS) was used in 30 infants and children with acute central nervous system injuries to determine the value of changes in specific metabolite ratios in predicting outcome. The mean age of all patients was 38 +/- 52 months and the mean time of study after insult was 7 +/- 5 days. 1H-MRS was determined in the occipital gray and parietal white matter (8 cm3 volume, STEAM sequence with TE = 20 ms, TR = 3,000 ms). Data were expressed as ratios of different metabolite peak areas including N-acetylaspartate (NA), choline-containing compounds (Ch), creatine and phosphocreatine (Cr), and lactate (Lac). Statistically significant differences were observed when patients with good/moderate (G/M) outcomes (n = 17; mean age: 46 months) were compared to patients with bad outcomes (n = 10; mean age: 26 months). NA/Cr and NA/Ch were significantly lower in the bad outcome group (NA/Cr = 1.15 +/- 0.38; NA/Ch = 1.18 +/- 0.52) compared to the G/M group (NA/Cr = 1.41 +/- 0.28, P < .05; NA/Ch = 1.98 +/- 0.81, P < .01). Lactate was present in 80% of bad outcome patients and in none of the G/M group (P < .0001). Using a linear discriminant analysis and combining 4 clinical variables (Glasgow Coma Scale score, initial pH and glucose, number of days unconscious at time of 1H-MRS) allows classification of 94% of patients into their correct outcome group. Use of spectroscopy variables (NA/Cr, NA/Ch, Ch/Cr, presence of lactate) alone correctly classified 81% of patients. The combination of clinical and 1H-MRS variables correctly classified 100% of patients. Our findings suggest that 1H-MRS adds information which, in combination with clinical examination, may be useful in outcome assessment in children with serious acute central nervous system injury.