BACKGROUND Stroke patients have impaired physical fitness and this may exacerbate their disability. It is not known whether improving physical fitness after stroke reduces disability. OBJECTIVE The primary aims of the review were to establish whether physical fitness training reduces death, dependence and disability after stroke. The secondary aims of the review included an investigation of the effects of fitness training on secondary outcome measures (including, physical fitness, mobility, physical function, health and quality of life, mood and the incidence of adverse events). METHODS We searched the Cochrane Stroke Group Trials Register (June 2003). In addition, the following electronic databases were searched: Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, 2002 Issue 4), MEDLINE (1966 to December 2002), EMBASE (1980 to December 2002), CINAHL (1982 to December 2002), SPORTDiscus (1949 to December 2002), Science Citation Index Expanded (1981 to December 2002), Web of Science Proceedings (1982 to December 2002), Physiotherapy Evidence Database (December 2002), REHABDATA (1956 to December 2002) and Index to UK Theses (1970 to December 2002). We hand searched relevant journals and conference proceedings and screened reference lists. To identify unpublished and ongoing trials we searched trials directories and contacted experts in the field. METHODS Randomised controlled trials were included when an intervention represented a clear attempt to improve either muscle strength and/or cardiorespiratory fitness, and whose control groups comprised either usual care or a non-exercise intervention. METHODS Data from eligible studies were independently extracted by two reviewers. The primary outcome measures were death, disability and dependence. The lack of common outcome measures prevented some of the intended analysis. RESULTS A total of twelve trials were included in the review. No trials reported death and dependence data. Two small trials reporting disability showed no evidence of benefit. The remaining available secondary outcome data suggest that cardiorespiratory training improves walking ability (mobility). Observed benefits appear to be associated with specific or 'task-related' training. CONCLUSIONS There are few data available to guide clinical practice at present with regard to fitness training interventions after stroke. More general research is needed to explore the efficacy and feasibility of training, particularly soon after stroke. In addition more specific studies are required to explore the effect of content and type of training. Further research will require careful planning to address a number of issues peculiar to this type of intervention.