Journal of Rehabilitation Medicine 51-5 | Page 5

Records identified through database searching (n = 1,548) 321 Additional records identified through other sources (n = 3,231) Records after duplicates removed (n = 3,602) Records screened (n = 3,602) Records excluded (n = 3,434) Full-text articles assessed for eligibility ( n = 168) Studies included in qualitative synthesis ( n = 50) Recovery-promoting drugs after stroke: a systematic review Studies included in quantitative synthesis ( n = 50) Full-text articles excluded, with reasons (n = 118) • • • • • • • • • • • • Not a human study (n=2) Not an RCT (n=38) Participants are not strokes survivors (n=22) No pharmacological intervention (n=1) No placebo control (n=11) Pharmacological intervention not targeting neuromodulation (n=10) Not targeting motor function outcomes (n=11) Not reported in English (n=6) Not able to access full-text (n=1) Previously excluded (n=10) Was a subgroup analysis (n=5) Non-Invasive Brain Stimulation was a second factor in study design (n=1) Fig. 1. PRISMA 2009 study selection flow diagram. dexamphetamine (6 trials) (37–42) and levodopa (6 trials) (15, 16, 19, 20, 29, 43). Bias scores were mixed across studies (Table SII). Only 4 studies had a low risk of bias across all 6 cri- teria, each study focused on a different RPD interven- tion (36, 44–46). High risk of bias was observed most commonly in attrition bias (27 studies, 54%) (15, 21, 27–30, 32, 33, 37–41, 43, 47–59) and reporting bias (36 studies, 72%) (15–19, 21–33, 37–43, 47, 49–52, 56, 59–63) domains. Study endpoints ranged between 60 min (16, 22) and 2 years (57) after the final RPD dose. Trial endpoints were split evenly between day of final RPD dose (15–26, 28, 44, 45, 47–49, 52–54, 56, 59, 60, 63) and beyond dosing completion (27, 29–43, 46, 50, 51, 55, 57, 58, 61, 62, 64). Fifty-six different efficacy outcome measures were used, with primary efficacy outcome measures desig- nated in 23 studies (16, 20, 32, 35–40, 42–46, 48, 50, 51, 53–56, 61, 62). Fourteen studies were described as safety and efficacy studies (31, 34–36, 40, 44, 46, 50, 51, 58, 59, 61, 62, 64), but only 4 had designated primary safety outcome measures (31, 61, 62, 64). In 22 studies safety was measured using outcome(s) specified in methods and reported in results (21, 29, 31, 33–36, 40, 44, 46, 49–52, 56–59, 61–64), while mortality and adverse events were observed and reported in another 22 studies (15, 17, 20, 22–28, 30, 37–39, 41, 43, 45, 47, 48, 53, 55, 60). In 6 studies, no safety considera- tions were reported (16, 18, 19, 32, 42, 54). No authors reported higher mortality or adverse events in drug in- tervention groups compared with placebo (Table SIII). Of the 28 RPDs identified, 18 (from 25 trials) sho- wed recovery-promoting potential (Table I) (15–18, 20, 22–24, 26, 29, 30, 32, 36, 42, 43, 45, 48, 52–54, 56, 59, 60, 62, 63). Seventeen RPDs were single-drug interventions, and the final RPD was a combination of methylphenidate and levodopa (29). For 13 RPDs, favourable results were reported from a single trial only (17, 23, 26, 29, 30, 42, 48, 52–54, 56, 59, 62). Neutral or unfavourable results were reported in ≥ 1 other trial for 4 (of these 13) drug interventions (amphetamine, dexamphetamine, MLC 601 and selegiline) (27, 37–41, 44, 47, 51, 57). Favourable effects on motor function were reported in ≥ 2 trials for Cerebrolysin®, citalo- pram, fluoxetine, levodopa (+carbidopa) and methylp- henidate (15, 16, 18, 20, 22, 24, 30, 32, 36, 43, 45, 60, 63). However, a beneficial effect on the same outcome was not replicated for any given drug intervention. For example, in 6 studies investigating levodopa, 19 dif- ferent efficacy outcome measures were used (15, 16, 19, 20, 29, 43), with only 2 (9-Hold Peg Test; 9HPT (15, 19), Rivermead Motor Assessment scale (RMA) (15, 43)) utilized in > 1 study. Use of multiple outcome measures in any 1 trial, and variation between trials is shown in Table I (and Table SI). J Rehabil Med 51, 2019