Background: Human retinal pigment epithelial (RPE) cells produce levodopa and their transplantation into the striatum might improve continuity of administration compared with that achieved with oral levodopa. We aimed to assess the safety, tolerability, and efficacy of transplantation of microcarrier-bound human RPE cells versus a sham surgery control in patients with advanced Parkinson's disease. Methods: In this randomised, double-blind study eligible patients were aged 36-70 years, had been symptomatic for at least 5 years, were in Hoehn and Yahr stage 3-4 and had unified Parkinson's disease rating scale (UPDRS) motor scores of 38-70 when off medication (off state), and had symptoms that responded to oral levodopa but were insufficiently controlled by optimised pharmacotherapy. Randomisation was done in a 1:1 ratio. Only the neurosurgical team was aware of treatment assignments. During stereotactic transplantation around 325 000 cells per side were injected into the postcommissural putamen; sham surgery patients received partial burr holes. The primary efficacy endpoint was change in UPDRS off-state motor score at 12 months. This study is registered with ClinicalTrials.gov, number NCT00206687. Findings: Of 71 enrolled patients, 35 underwent cell transplantation and 36 sham surgery. Change in mean motor scores did not differ significantly between groups (-10·5 [SD 10·26] for transplantation vs -10·1 [SD 12·26] for sham surgery, p=0·9). The overall rate of adverse events was similar in the two study groups, although the number attributable to surgery or RPE cells (mostly neurological or psychiatric) was higher in transplant recipients. Two and seven patients died in the sham surgery and transplantation group, respectively; one death in the latter group was possibly related to surgery or RPE cells. Interpretation: Transplantation of human RPE cells provided no antiparkinsonian benefits compared with sham surgery. Funding: Bayer HealthCare AG.