Five randomised controlled trials (RCTs) (n=135 participants) and one quasi-randomised trial (n=30 participants) were included in the review.
The studies were of moderate quality (mean PEDro score 5.5 from a maximum possible 8). Allocation was concealed in two studies (33%). Assessors were blinded in three studies (50%). More than 15% of participants dropped out in three studies (50%). Intention-to-treat analysis was carried out in two studies (33%).
The intervention had a non-significant beneficial effect on strength immediately after the end of the training period (SMD 0.20, 95% CI -0.17 to 0.56; five studies, n=119 participants). There was no evidence of heterogeneity (I2<30%). Six to 12 weeks after cessation of training, the intervention had no effect on strength (SMD 0.05, 95% CI -0.47 to 0.58; two studies, n=56 participants).
Data from the one study that reported on spasticity could not be extracted.
The interventions had no effect on walking speed (four studies) and a small beneficial effect on activity (two studies). There was an increase of 2% on the Gross Motor Function Measure (95% CI 0 to 4%). This effect was retained but no longer significant six to 12 weeks after the cessation of training (2% increase, 95% CI -4 to 7).
Subgroup analyses showed that neither type of intervention nor the part of the body made a difference to the size of the effect.