Malignant gliomas are known to have a high glycolytic rate, and are dependent on glucose for energy metabolism (the “Warburg effect”). In animal models of glioma, reduced glucose levels via calorie restriction have been associated with reductions in tumor growth and improved survival(1). Ketogenic diets (KDs) are high-fat, low-carbohydrate dietary interventions which reduce serum glucose, induce systemic ketosis, shift energy metabolism to ketone bodies, and have been explored in recurrent glioblastoma(2,3). The modified Atkins diet is a novel KD alternative which has proven efficacious in the management of refractory epilepsy and has been explored in glioma (2).The purpose of this pilot study was to evaluate the feasibility of an 8-week Glioma modified Atkins-based Diet (GLAD), and to explore the biologic activity in both tumor and normal brain regions using proton MR spectroscopy (MRS).5 patients with biopsy-confirmed high grade glioma (2F, age 40±6 yrs) who were post-surgery and more than 7 months post-chemoradiation were evaluated by MRI/MRS at baseline and after 8 weeks of GLAD. GLAD consisted of 2 ‘fasting’ days (caloric restriction to 20% of recommended daily allowance,) interleaved between 5 days following the modified-Atkins diet (carbohydrate intake restricted to 20 gm) each week. MRS was performed at 3T in both the lesion and in a mirror voxel in the contralateral hemisphere (Philips 3T Achieva, 32-channel head coil, sLASER localization, TR/TE 2.2s/34ms, 2nd order shimming, VAPOR water suppression, 2x2x2cm, 128 averages, 5 min scan time). Voxel placement was in a region of the lesion with T2 hyperintensity on MRI, adjacent to the resection cavity, considered by a radiologist to be most likely consistent with residual tumor. Data were analyzed using the ‘LCModel’ program using a basis set consisting of acetone (Ace), beta-hydroxybutyrate (bHb), acetoacetate (AcAc), 2-hydroxyglutarate (2-HG) and citrate (Cit) in addition to a standard set of metabolites, lipid and macromolecule resonances found in normal brain. Concentrations were expressed relative to the brain water signal (i.u), as well as ratios to total creatine (tCr).All patients tolerated GLAD and all achieved measurable ketosis. Spectra and images at 8 weeks in one case are shown in figure 1. Figure 2 shows selected brain metabolite and ketone concentrations at baseline and at 8 weeks. As expected, multiple compounds were different between lesion and normal brain (e.g. at baseline, Glx p<0.02, tNAA p<0.001, tCr, p<0.02, Lac p < 0.04 (2nd visit only)). The only significant difference in metabolite concentrations between baseline and 8 week follow-up was a decrease in tNAA in the contralateral hemisphere(visit 1 [tNAA]=7.29±0.41, visit 2 [tNAA]= 6.89±0.41, p<0.02 (paired t-test)). Of the ketone bodies (Ace, AcAc, bHb), none were detectable by MRS at baseline, however acetone was reliably determined by LCModel at 8 weeks in both lesion ([Ace]=0.41±0.34 i.u., mean CRLB=47%) and normal brain ([Ace]=0.49±0.23 i.u., mean CRLB=28%). In the lesion, MRS was suggestive of stable disease in the cohort as a whole, since no significant differences in tumor markers such as tCho (or other compounds, e.g. tNAA or mI) were observed.Preliminary results of this pilot study demonstrate the ability of brain MRS to monitor the metabolic consequences of dietary interventions in patients with high grade glioma. Consistent with a prior report of MRS in KD (in patients with epilepsy (4)), increases in detectable acetone signal were observed both in lesional and normal brain. The lack of overall changes in tumor markers (such as Cho) is consistent with the clinical stability and lack of progression seen on brain MRI at 8 weeks in 4 of the 5 cases. The origin of the decline in tNAA in the contralateral hemisphere is unclear; possible explanations might include reduced NAA production, since NAA is synthesized in neuronal mitochondria closely related to the TCA cycle, whose flux may be reduced under ketotic conditions, or perhaps a delayed effect of the prior chemoradiation on neuronal health. Future studies of GLAD therapeutic efficacy will require both a larger sample size, and the inclusion of a placebo group for comparison.