Abstract Objectives To identify factors relevant to implementing constraint-induced movement therapy (CIMT) within the clinical setting. Data sources AMED (1995–January 2007), CINHAL (1982–January 2007), Medline (1996–January 2007) and EMBASE (1996–January 2007) were searched to identify relevant studies. Review methods Criteria for inclusion of trials in this study were that trial participants were over 18 years of age, had had a stroke, and CIMT or modified CIMT was compared with either no intervention, modified CIMT or alternative treatment. Modified CIMT had to include both constraint and training components. The study design was either a quasi-randomised controlled trial or a randomised controlled trial. Trials had to be published in English and score 4 or more on the PEDro scale. Results Twelve eligible studies were identified. The quality of the studies varied, although there was evidence that this improved with more recent studies achieving higher PEDro scores. Patient selection criteria and the components related to the delivery of CIMT were identified as relevant factors. Thirteen different patient selection criteria were identified: age; length of time post stroke; specified side of hemiplegia; hand dominance; spasticity; pain; balance and mobility; hand function; range of active and passive movement; cognitive impairment; perception; sensation; and communication. Ten components were identified as being relevant to the actual delivery of CIMT: type of CIMT; type of constraint; constraint wear time; excluded activities; shaping; shaping dosage; group versus individual treatment; environment; potential harms of CIMT; and compliance. A third relevant consideration was the selection of outcome measures. Significant variability was identified in many aspects of CIMT, although there was evidence of greater standardisation in more recent studies. Conclusion The development of CIMT for stroke patients has provided clinicians with a treatment technique for a defined patient group that is now supported by a considerable evidence base. CIMT is a complex intervention and the optimum intensity and length of treatment remains unknown. Transferring CIMT into the clinical environment has been hampered by the lack of standardisation in many aspects of the intervention. However, there is evidence that this is improving. Implementation and evaluation in the clinical environment would strengthen the evidence base.