Background: Since the 1950s neuroleptic medication has been extensively used to treat people with chronic mental illnesses such as schizophrenia. These drugs, however, have been also associated with a wide range of adverse effects, including movement disorders such as tardive dyskinesia (TD). Various strategies have been examined to reduce a person's cumulative exposure to neuroleptics. These studies include dose reduction, intermittent dosing strategies such as drug holidays, and neuroleptic cessation. Objectives: To determine whether a reduction or cessation of neuroleptic drugs is associated with a reduction in TD, for people with schizophrenia (or other chronic mental illnesses) who have existing TD. Our secondary objective was to determine whether the use of specific neuroleptics for similar groups of people could be a treatment for TD that was already established. Search strategy: We updated previous searches of the Cochrane Schizophrenia Groups Register (1997), Biological Abstracts (1982-1997), EMBASE (1980-1997), LILACS (1982-1996), MEDLINE (1966-1997), PsycLIT (1974-1997), and SCISEARCH (1997) by searching the Cochrane Schizophrenia Groups Register (July 2003). We searched references of all identified studies for further trial citations. We also contacted the principal authors of trials for further unpublished trials. Selection criteria: We included reports if they assessed people with schizophrenia or other chronic mental illnesses who had established neuroleptic-induced TD, and had been randomly allocated to (a) neuroleptic maintenance versus neuroleptic cessation (placebo or no intervention), (b) neuroleptic maintenance versus neuroleptic reduction (including intermittent strategies), and (c) specific neuroleptics for the treatment of TD versus, placebo or intervention. A post hoc decision was made to broaden comparison (c) to include specific neuroleptics versus other neuroleptics for the treatment of TD. Data collection and analysis: We (KSW, JR) independently inspected citations and, where possible, abstracts, ordered papers, and re-inspected and quality assessed these and extracted data. We analysed dichotomous data using random effects relative risk (RR) and estimated the 95% confidence interval (CI). Where possible we calculated the number needed to treat (NNT) or number needed to harm statistic (NNH). We excluded continuous data if more than 50% of people were lost to follow up, but, where possible, we calculated the weighted mean difference (WMD). It was assumed that those leaving the study early showed no improvement. Main results: We included five trials and excluded 102. One small two week study (n=18), reported on the 'masking' effects of molindone and haloperidol on TD, which favoured haloperidol (RR 3.44 CI 1.1 to 5.8). Two (total n=17) studies found no reduction in TD associated with neuroleptic reduction (RR 0.38 CI 0.1 to 1.0). One study (n=20) found no significant differences in oral dyskinesia (RR 2.45 CI 0.3 to 19.7) when neuroleptics were compared as a specific treatment for TD. Dyskinesia was found to be not significantly different (n=32, RR 0.62 CI 0.3 to 1.26) between quetiapine and haloperidol when these neuroleptics were used as specific treatments for TD, although the need for additional neuroleptics was significantly lower in the quetiapine group (n=47, RR 0.49 CI 0.2 to 1.0) than in those given haloperidol. Authors' conclusions: Limited data from small studies using neuroleptic reduction or specific neuroleptic drugs as treatments for TD did not provide any convincing evidence of the value of these approaches. There is a need for larger trials of a longer duration in order to fully investigate this area.