We hypothesized that the glucose metabolism of prostate cancer is modulated by androgen. We performed in vivo biodistribution and imaging studies of [F-18] fluorodeoxyglucose (FDG) accumulation in androgen-sensitive (CWR-22) and androgen-independent (PC-3) human prostate cancer xenografts implanted in castrated and noncastrated male athymic mice. The growth pattern of the CWR-22 tumor was best approximated by an exponential function (tumor size in mm3 = 14.913 e(0.1086 x days), R2 = .96, n = 5). The growth pattern of the PC-3 tumor was best approximated by a quadratic function (tumor size in mm3 = 0.3511 x days2 + 49.418 x day - 753.33, R2 = .96, n = 3). The FDG accumulation in the CWR-22 tumor implanted in the castrated mice was significantly lower, by an average of 55%, in comparison to that implanted in the noncastrated host (1.27 vs. 2.83, respectively, p < .05). The 3-week maximal standardized uptake value (SUVmax) was 0.99 +/- 0.43 (mean +/- SD) for CWR-22 and 1.21 +/- 0.32 for PC-3, respectively. The 5-week SUVmax was 1.22 +/- 0.08 for CWR-22 and 1.35 +/- 0.17 for PC-3, respectively. The background muscle SUVmax was 0.53 +/- 0.11. Glucose metabolism was higher in the PC-3 tumor than in the CWR-22 tumor at both the 3-week (by 18%) and the 5-week (by 9.6%) micro-PET imaging sessions. Our results support the notions that FDG PET may be useful in the imaging evaluation of response to androgen ablation therapy and in the early prediction of hormone refractoriness in men with metastatic prostate cancer.