OBJECTIVE To compare the efficacy and safety of adjunct pergolide therapy versus bromocriptine in patients with Parkinson's disease, already established on levodopa and suffering the long-term complications of therapy. METHODS Electronic searches of MEDLINE, EMBASE and the Cochrane Controlled Trials Register. Handsearching of the neurology literature as part of the Cochrane Movement Disorders Group's strategy. Examination of the reference lists of identified studies and other reviews. Contact with Eli Lilly Company and Sandoz Limited. METHODS Randomised controlled trials of pergolide versus bromocriptine in patients with a clinical diagnosis of idiopathic Parkinson's disease and long-term complications of levodopa therapy. METHODS Data was abstracted independently by each author and differences settled by discussion. RESULTS Three short-term trials fulfilled the inclusion criteria for the review. Pergolide was superior to bromocriptine regarding UPDRS and NYPDS motor and NYPDS ADL scores in two trials. More patients recorded a 'marked' or 'moderate improvement' in clinician's global impression score with pergolide than bromocriptine in two studies. Insufficient evidence on fluctuations and dyskinesia was available to draw any conclusions. No significant differences between the agonists were seen in levodopa dose reduction, drop outs or adverse events. CONCLUSIONS Although pergolide is superior to bromocriptine in reducing motor impairments and disability, no firm conclusions regarding levodopa-induced motor complications can be reached. Levodopa dose reduction, adverse events and withdrawals from treatment are similar for the two agonists. The small advantage of pergolide in efficacy does not take into account its additional cost compared with bromocriptine.