Nine articles reporting seven trials were included in the review (n=680 participants, range 20 to 189). Three trials were multi-centre and four were single-centre. Three trials were rated as low risk of bias on all items; two were rated as low or unclear risk of bias on all items. One trial was judged at high risk of bias for selective outcome reporting. One trial was judged as high risk of bias for blinding. Outcome assessment post-intervention was four to six weeks after the start of therapy, except in one trial where it was 20 weeks. Duration of follow-up ranged from three to 12 months after stroke/treatment start.
Motor impairment - muscle function (six trials): There was a trend towards benefit using higher-dose therapy; four trials showed statistically significant improvements with higher doses and one found a significant benefit with low-dose therapy. However, five trials showed no statistical differences between doses. Data were only suitable for meta-analysis for the outcome of hand grip force/strength. There was a significant beneficial effect of exercise dose on this outcome in favour of the lower dose (WMD -10.1, 95% CI -19.1 to -1.2; two trials). At the first follow-up point (12 to 26 weeks after stroke/treatment start), there was a significant improvement in two of seven comparisons. There were sufficient data to pool data on arm motricity (Motricity Index Arm score) which showed a significant beneficial effect in favour of the higher dose (WMD 10.7, 95% CI 1.7 to 19.8; two trials). Three comparisons reported outcomes after a second period of follow-up (six months after start of therapy/stroke); none of the comparisons showed a significant effect of exercise dose on outcome.
Motor impairment - movement control (one trial): One trial that reported on two different outcome post-treatment and 18 weeks after the start of therapy found no association between exercise dose and outcome.
Functional activity (seven trials): There was a significant improvement in outcome post-treatment in favour of the high dose in five of 14 comparisons. One comparison showed significant benefit in favour of low-dose therapy; the remaining comparisons showed no significant differences between the different doses. Data were suitable for meta-analysis for two outcomes. There was no significant effect of exercise therapy on Action Research Arm Test (three trials), but there was a significant improvement in comfortable walk speed associated with greater exercise intensity (WMD 0.3, 95% CI 0.1 to 0.5; two trials). At the first follow-up point (12 to 26 weeks after treatment start), there was a significant improvement in favour of higher dose in two of 13 comparisons; two of 13 comparisons showed significant benefit for lower dose of therapy. Meta-analysis was possible for Action Research Arm Test (two trials), Rivermead mobility (two trials), and comfort walk speed (two trials), but none of these showed a significant association with exercise duration. After the second period of follow-up (six to 12 months after stroke/therapy), one of 11 comparisons showed a significant improvement in favour of higher-dose therapy; two of 11 comparisons showed benefit from lower-dose therapy. Meta-analysis was only possible for Action Research Arm Test (three trials), but no significant effect of exercise dose was found.