Journal of Rehabilitation Medicine 51-5 | Page 4

320 N. Firth et al. disability when taking risk of harm into account (7). The number of patients included in the studies was too small to be able to draw firm conclusions regarding the effect of amphetamines on recovery from stroke (7). Conversely, another Cochrane Review (n = 52 trials, 4,059 patients) provided “tantalizing evidence” that selective serotonin reuptake inhibitors (SSRIs) appear to improve dependence, disability, neurological impair- ment, and anxiety and depression after stroke (8). Both reviews recommended larger, well-designed trials be undertaken to clarify efficacy, and to overcome issues with heterogeneity and methodology seen in studies across both drug classes. As both reviews targeted singular drug classes, neither could provide judgement comparing the outco- mes of the drugs with each other. To address this gap, the aim of this systematic review was to investigate the efficacy and safety of drug interventions trialled to enhance motor recovery post-stroke (3, 4). METHODS Protocol and registration This systematic review was registered on PROSPERO (refe- rence number: CRD42016048035). Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) state- ment provided the framework for the article (9). Data sources and searches Six electronic databases (Cochrane CENTRAL, CINAHL, Embase, MEDLINE, SCOPUS and Web of Science) were searched from database inception to 2 May 2017. Reference lists of included studies were entered into the Web of Science to identify relevant studies from forward citations. The search term “recovery-promoting drug” was not widely recognized. Relevant drug studies identified through a scoping search were mined for terminology describing the concept of “promoting neurorecovery”. The resulting search strategy was curated carefully, containing key words and MeSH terms associated with target pathways, anatomy and processes (e.g. “efferent pathways”, “motor cortex” and “neurogenesis”), and molecules involved in neural plasticity and neural repair (e.g. more broadly: “nerve growth factors”, “psychotropic drugs”; specifically “biogenic amines” and “dopamine”) with the in- tention of selecting motor recovery-specific studies from the broader pool of neurorecovery trials (Figs S1–S7). Study selection Study inclusion criteria were: (i) randomized placebo-controlled trial design; (ii) commencement of 1 or more RPD intervention/s > 24 h post-stroke (10); and (iii) a measure of motor outcome of components of the motor system within the domains of body functions and structures and activity, as defined by the Inter- http://www.medicaljournals.se/jrm/content/?doi=10.2340/16501977-2536 1 www.medicaljournals.se/jrm national Classification of Functioning, Disability and Health (ICF) (11). Non-English publications and aphasia trials were excluded, the latter being a language disorder, not attributable to motor function. Titles and abstracts were reviewed and shortlisted by author NF, and by author JB if inclusion was unclear. Eligibility was determined through independent assessments of full-text ver- sions of shortlisted articles by authors NF and KH, while author JB confirmed eligibility when necessary. Data extraction and analysis, and risk of bias assessment Study details (sample size, time post-stroke, age, sex, stroke severity), experimental design descriptors (drug and control intervention details, adjuvant physical therapy, treatment/ follow-up endpoints), outcome measures and corresponding measures of central tendency were extracted by author NF and corroborated by author KH utilizing standardized pro forma (12). Authors of included studies were emailed for missing data. Physical rehabilitation interventions within each trial were recorded as “adjuvant therapies”. The endpoint was defined as final assessment of outcome; whether occurring at final dose of drug intervention or end of follow-up was noted, along with whether primary outcome measures were designated. The extent of safety monitoring and adverse events were recorded, and whether they were pre-specified outcomes or general obser- vations. Safety assessment was based on mortality and severe adverse events (SAEs) associated with drug intervention, e.g. haemorrhage, neurological deterioration. Risk of bias was asses- sed by NF and KH using the Cochrane risk of bias assessment tool (13). Between-group endpoint estimates and change scores were extracted for motor outcomes. When statistically signi- ficant p-values were noted, effect sizes were calculated (NF) from raw data using Cohen’s d method, where possible (14). RESULTS Database searching yielded 1,548 results, with 29 studies eligible for inclusion. These studies contained 3,231 references, used to identify further studies, which led to the inclusion of a further 21 studies (Fig. 1). Therefore, a total of 50 studies (n = 5,643 participants) were included. Duplicate citation (25%) or not RCT (40%) were the most common reasons for exclusion. Included studies were published between 1973 and 2017 (Table SI). Methodologies varied, including cros- sover trials (n = 10) (15–24) and short-term studies with outcomes measured ≤ 24 h post-RPD administration (n = 11) (16–26). Sample sizes ranged from 8 to 1,099 participants (median: 40, IQR: 18.5–83), mean age spanned 53 (25) to 78 years (27). Participants were predominantly male (range 32–100%); only 12 (24%) studies had ≥ 50% females. Twenty-eight different RPDs were investigated (Ta- ble I). Four studies compared 2 drug intervention arms with placebo (28–31). Four studies evaluated MLC 601 (NeuroAid ™ ) and involved the largest proportion of participants (n = 2,099, 37.2%) (32–36). The most frequently studied pharmacological interventions were