PURPOSE: Numerous dietary factors elevate serum levels of insulin and insulin-like growth factor I (IGF-I), both potent prostate cancer mitogens. We tested whether varying dietary carbohydrate and fat, without energy restriction relative to comparison diets, would slow tumor growth and reduce serum insulin, IGF-I, and other molecular mediators of prostate cancer in a xenograft model. EXPERIMENTAL DESIGN: Individually caged male severe combined immunodeficient mice (n = 130) were randomly assigned to one of three diets (described as percent total calories): very high-fat/no-carbohydrate ketogenic diet (NCKD: 83% fat, 0% carbohydrate, 17% protein), low-fat/high-carbohydrate diet (LFD: 12% fat, 71% carbohydrate, 17% protein), or high-fat/moderate-carbohydrate diet (MCD: 40% fat, 43% carbohydrate, 17% protein). Mice were fed to maintain similar average body weights among groups. Following a preliminary feeding period, mice were injected with 1 x 10(6) LNCaP cells (day 0) and sacrificed when tumors were >or=1,000 mm(3). RESULTS: Two days before tumor injection, median NCKD body weight was 2.4 g (10%) and 2.1 g (8%) greater than the LFD and MCD groups, respectively (P < 0.0001). Diet was significantly associated with overall survival (log-rank P = 0.004). Relative to MCD, survival was significantly prolonged for the LFD (hazard ratio, 0.49; 95% confidence interval, 0.29-0.79; P = 0.004) and NCKD groups (hazard ratio, 0.59; 95% confidence interval, 0.37-0.93; P = 0.02). Median serum insulin, IGF-I, IGF-I/IGF binding protein-1 ratio, and IGF-I/IGF binding protein-3 ratio were significantly reduced in NCKD relative to MCD mice. Phospho-AKT/total AKT ratio and pathways associated with antiapoptosis, inflammation, insulin resistance, and obesity were also significantly reduced in NCKD relative to MCD tumors. CONCLUSIONS: These results support further preclinical exploration of carbohydrate restriction in prostate cancer and possibly warrant pilot or feasibility testing in humans.