In the absence of an established clinically important difference in stride length, we consider

25 cm a clinically Protease Inhibitor Library important difference. Again, our 95% CI excludes the possibility that treadmill training worsens stride length to that extent. The walking speed achieved by our experimental group is similar to that achieved by repetitive locomotor training using a mechanical gait trainer (Pohl et al 2007). At six months, Pohl and colleagues (2007) reported a mean walking speed of 0.53 m/s which is almost identical to the 0.57 m/s speed achieved by our treadmill group. Furthermore, our finding that treadmill walking did not have a negative effect on quality is consistent with recent work by Kuys and colleagues (2008a) who found that walking on a treadmill did not result in a deterioration of overground walking AZD6244 molecular weight pattern compared with walking overground in newly ambulating stroke patients undergoing rehabilitation. They (Kuys et al 2008b) also found that increasing the intensity of walking on a treadmill did not adversely affect the walking pattern or quality. Taken together, these findings suggest that one barrier to implementation

of this intervention, ie, the fear that treadmill walking would have a deleterious effect on quality, is unfounded. Another finding suggests that treadmill walking with body weight support results in a greater capacity for walking compared with assisted overground walking. At almost 60 m, the increased capacity is clinically significant. However, during the CI is wide suggesting some uncertainty about the size of the effect. The magnitude of the improvement is similar to that reported by Pohl and colleagues (2007) who found a 44 m difference in favor of the repetitive locomotor group. This increased capacity is accompanied by a 10% higher rating of walking by the experimental group compared to the control group at 6 months. Although this is a positive rating, it may be the result of the participants not being blind to group allocation. However, importantly, participants

undergoing treadmill walking with body weight support do not perceive themselves to be worse off than if they had been assisted to walk overground. There was, however, no difference in community participation between the groups. Our participants had very low levels of community participation as measured by the Adelaide Activities Profile. This is perhaps not surprising given that, on entry to the study, all participants were unable to walk and therefore represent the most disabled people admitted to rehabilitation. Even those who achieved independent walking, regardless of group, walked slowly with a mean speed of less than 0.6 m/s. This is less than half normal elderly speed and only one-third normal young speed. Furthermore it is 0.2 m/s slower than the mean walking speed of people after stroke who met the criteria of community ambulators in the classification devised by Perry and colleagues (1995).