Rusconi 2002.

Methods	RCT Setting: Italy
Participants	24 randomised (see Notes) but outcome data collected on 20 (The following data describe the 20 participants) Experimental: n = 12, control: n = 8 (experimental is Type 1 and Control is Type 2: see Interventions) Mean age: experimental: 69.8 years, control: 65.1 years Sex (male/female): experimental 5/7, control 3/5 Mean weeks post‐stroke: experimental: 6.92, control 8.38 Inclusion: unilateral right hemisphere stroke assessed by CT scan, right‐handed, symptoms of unilateral neglect, admitted to hospital for rehabilitation 5 weeks post‐stroke Exclusion: dementia
Interventions	The study compared more than 2 interventions. First there is a comparison of 2 types of cognitive training: Type 1 versus Type 2. Each 'type' is then subdivided into whether or not TENS is added (see Notes) Type 1 versus Type 2: both consist of 5 x 1‐hour sessions per week for 2 consecutive months (40 sessions) using 4 procedures requiring the participant to actively scan the visual field (reading sentences and stories, line drawing on a dot matrix, assembling 3D cubes, matching cards containing the name and the visual image of an object. Type 1 and 2 differed in that only Type 1 involved verbal and visuospatial cueing and verbal feedback. Although Type 2 used the same 4 procedures it did not involve cueing or feedback, i.e. the aspects of the training designed to improve awareness and encourage compensation For this review Type 1 was classed as a top‐down approach and Type 2 as an attention control
Outcomes	Assessments were classified as 'functional and neurological' (i.e. BI, standard clinical neurological examination) or 'neuropsychological' (i.e. line cancellation, letter cancellation, line bisection, sentence reading, O'clock test, judgement of drawings, anosognosia, RCPM, facial recognition, position deficit) These were taken at 4 time points: on admission for neurorehabilitation at least 5 weeks post‐stroke (T0), 1 month later (T1) after which eligibility was determined and participants were randomised, after 1 month of intervention (T2) and after 2 months of intervention (T3) For comparability with other studies this review used only the T3 letter cancellation and BI. As intervention continued for 2 months T3 is coded in this review as immediate effects
Notes	Author provided clarification and raw data by personal communication 24 people were randomised: 12 to Type 1 and 12 to Type 2. The authors excluded 4 from the final evaluation because of a "clinical worsening that prevented the conclusion of the treatment". These 4 were all allocated to Type 2 Cancellation scores were for the number correctly cancelled. Separate scores were given for left and right space but this review used the total score. Line bisection data were for mean deviation in mm left (negative) or right (positive) from the midpoint. Line cancellation data could not be used as the experimental group's SD was 0
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	High risk	Paper states that "randomly assigned". No details provided Subsequent information states allocations stored in sequentially numbered, sealed, opaque, envelopes. Concealment of allocation is unlikely
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Blinded outcome assessor
Incomplete outcome data (attrition bias) All outcomes	High risk	None reported in paper but authors reported 24 people were randomised: 12 to Type 1 and 12 to Type 2. The authors excluded 4 from the final evaluation because of a "clinical worsening that prevented the conclusion of the treatment". These 4 were all allocated to Type 2. Not intention‐to‐treat analysis
Free of systematic differences in baseline characteristics of groups compared?	Low risk	Groups look comparable but numbers in each group are very small
Did authors adjust for baseline differences in their analyses?	Low risk	Used a repeated measures ANOVA