Lum 2006

Methods	Randomly allocated Patients were stratified by initial Fugl‐Meyer score and side of stroke and randomly assigned to 1 of 4 groups Following interim analysis the randomisation schedule was changed from providing the same number of participants to each group so that subsequent participants could only be allocated to 2 of the groups, therefore participants did not have an equal chance of entering 1 of the 4 groups The change in randomisation during the trial may have introduced bias
Participants	30 participants (only 2 groups of participants ‐ 14 participants ‐ relevant to this review) Inclusion criteria: single CVA, 1 to 5 months post‐stroke, no upper‐limb joint pain or ROM limitations that would limit ability to complete training, no unstable cardiovascular, orthopaedic or neurological conditions, > 21 on MMSE
Interventions	Group 1 (9 participants): robot‐unilateral group, 12 reaching tasks progressing from easiest robotic‐mode to most challenging mode Group 2 (5 participants): robot‐bilateral group, practised same 12 reaching tasks but in bilateral mode rhythmic circular movements also performed Training lasted 1 hour per session for 15 sessions over 4 weeks Training was supervised by an occupational therapist
Outcomes	Primary outcome: performance in activities of daily living: Functional Independence Measure (self‐care and transfer sections only) Secondary outcome: motor impairment ‐ motor impairment scales: Fugl‐Meyer (proximal and distal upper limb sections), Motor Status Score (movement scale and synergy scale) and Modified Ashworth scale (proximal and distal scores) (Fugl‐Meyer (proximal upper limb section) selected for analysis); strength outcomes: Motor power examination (several joints across proximal upper limb)
Notes	This study included assistive technology, however it compared a bilateral and unilateral group both receiving robotic assistance, therefore we decided that this was relevant to include as bilateral training versus unilateral training 4 groups were included in this trial: robot‐unilateral, robot‐bilateral, robot‐combined and control Only robot‐unilateral and robot‐bilateral relevant to this review Participants in the other 2 groups (16 participants) not included in any analysis Average gains data presented in paper and used in analysis Standard deviations calculated from presented standard error of the mean to allow for inclusion in statistical pooling
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	Initially allocation concealment was possible, however an interim analysis was carried out to compare the groups and, as a result the randomisation process was changed so that later participants could only be entered into 1 of the 2 groups, not 1 of the 4
Blinding of outcome assessor?	Low risk	An occupational therapist blinded to group assignment completed the assessments
Intention to treat analysis?	Unclear risk	Not stated; all participants completed the training and post‐treatment evaluations
Baseline similarity	Low risk	No significant differences at baseline