Dean 1997.

Study characteristics
Methods	Study design: RCT Method of randomisation: blocked randomisation ("subjects drew a card from a box that was originally filled with 10 control and 10 experimental cards")
Participants	Number of participants: n = 20 Inclusion criteria: diagnosed with stroke more than one year previously, discharged from rehabilitation, able to understand instructions, able to give informed consent, no orthopaedic problems that could interfere with ability to perform seated reaching tasks and able to sit unsupported for 20 minutes
Interventions	(1) Motor learning group (n = 10) Standardised training programme designed to improve sitting balance through reaching with the unaffected hand "The training for the experimental group was designed to improve sitting balance and involved emphasis on appropriate loading of the affected leg while practicing reaching tasks using the unaffected hand to grasp objects located beyond arm’s length. The reaching tasks were performed under systematically varied conditions. Distance and direction were varied by changing the location of the object. Seat height, movement speed, object weight, and extent of thigh support on the seat were also varied. The training was advanced by increasing the number of repetitions and complexity of the tasks" The individual components delivered are listed in Table 6. Based on the individual components, this intervention is categorised as comprising functional task training Length of intervention period: two weeks Number of sessions and length of individual sessions: 10 sessions over two weeks, average of 30 minutes Intervention provider: "the training programs were carried out by the first author in the subject’s home" (2) Placebo group (n = 10) Cognitive manipulative tasks, involving reaching the unaffected hand over very small distances "Control group had sham training that incorporated the performance of cognitive‐manipulative tasks while seated at a table. Sham training was performed so that subjects would consider themselves involved in a training program and to eliminate any effect due to placebo. They performed manipulative tasks using the unaffected hand over small distances (less than 50% of arm length). Training was advanced over sessions by increasing the repetitions and cognitive difficulty of the tasks. The subjects in the control group performed an equal number of reaching movements as the subjects assigned to the experimental group; however, the nature of the tasks ensured that only a minimum balance perturbation occurred" The individual components delivered are listed in Table 6. Based on the individual components, this intervention is categorised as comprising an attention control Length of intervention period: two weeks Number of sessions and length of individual sessions: 10 sessions over two weeks, average of 30 minutes Intervention provider: "the training programs were carried out by the first author in the subject’s home" This study is classified as intervention (functional task training) versus attention control (cognitive) (Table 8)
Outcomes	Measures of postural control and balance: ground reaction force during reaching; EMG during reaching; maximum distance reached; ground reaction force during rising to stand Measures of voluntary movement: timed 10‐Metre Walk Other measures: time to complete cognitive task Time points when outcomes were assessed: "pretest and posttest group design was used"
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Subjects were randomly assigned to either the experimental or control group. Randomization was blocked to ensure equal numbers in the groups. The procedure involved random sampling without replacement; subjects drew a card from a box that was originally filled with 10 control and 10 experimental cards"
Allocation concealment (selection bias)	High risk	"Subjects drew a card from a box that was originally filled with 10 control and 10 experimental cards" Judged at high risk, as concealment would have been broken once 10 were in one group
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	The assessor was blinded for some assessments. The outcome measures for which there was no blinded assessor were recorded by computer However, the assessor could have encouraged some participants more than others
Incomplete outcome data (attrition bias) All outcomes	Low risk	19/20 completed intervention and final assessment Dropouts accounted for: "One subject from the control group dropped out of the study because of an acute neurological episode that required hospitalization"
Free of systematic differences in baseline characteristics of groups compared?	Low risk	"no significant differences between the groups in terms of age, time since stroke, or walking velocity"
Did authors adjust for baseline differences in their analyses?	Low risk	No information provided
Other bias	High risk	One therapist (principal investigator) carried out all treatments. The use of only one therapist provides a potential source of contamination between groups or the introduction of performance bias