Abnormal covariance pattern of regional metabolism associated with Parkinson disease (PD) is modulated by dopaminergic pharmacotherapy. Using high-resolution ¹⁸F-FDG PET and network analysis, we previously derived and validated a parkinsonism-related metabolic pattern (PRP) in nonhuman primate models of PD. It is currently not known whether this network is modulated by experimental therapeutics. In this study, we examined changes in network activity by striatal implantation of human levodopa-producing retinal pigment epithelial (hRPE) cells in parkinsonian macaques and evaluated the reproducibility of network activity in a small test–retest study. Methods:¹⁸F-FDG PET scans were acquired in 8 healthy macaques and 8 macaques with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP)–induced bilateral nigrostriatal dopaminergic lesions after unilateral putaminal implantation of hRPE cells or sham surgery. PRP activity was measured prospectively in all animals and in a subset of test–retest animals using a network quantification approach. Network activity and regional metabolic values were compared on a hemispheric basis between animal groups and treatment conditions. Results: All individual macaques showed clinical improvement after hRPE cell implantation compared with the sham surgery. PRP activity was elevated in the untreated MPTP hemispheres relative to those of the normal controls (P < 0.00005) but was reduced (P < 0.05) in the hRPE-implanted hemispheres. The modulation observed in network activity was supported by concurrent local and remote changes in regional glucose metabolism. PRP activity remained unchanged in the untreated MPTP hemispheres versus the sham-operated hemispheres. PRP activity was also stable (P ≥ 0.29) and correlated (R² ≥ 0.926; P < 0.00005) in the test–retest hemispheres. These findings were highly reproducible across several PRP topographies generated in multiple cohorts of parkinsonian and healthy macaques. Conclusion: We have demonstrated long-term therapeutic effects of hRPE cell implantation in nonhuman primate models of PD. The implantation of such levodopa-producing cells can concurrently decrease the elevated metabolic network activity in parkinsonian brains on an individual basis. These results parallel the analogous findings reported in patients with PD undergoing levodopa therapy and other symptomatic interventions. With further validation in large samples, ¹⁸F-FDG PET imaging with network analysis may provide a viable biomarker for assessing treatment response in animal models of PD after experimental therapies.