Journal of Rehabilitation Medicine 51-2 | Page 5

80 Outcomes of this review were reported in correspondence with the proposed research questions. This firstly includes a description of therapeutic strategies allowing non-ambulant stroke patients to repetitively train walking. Secondly, the 2 feasibility items safety, measured by the incidence of adverse events, and adherence to therapy, defined as the number of drop-outs, were investigated. Thirdly, outcomes on effectiveness were investigated and classified according to the domains of the International Classification of Functioning Disability and Health (ICF) model (32). All included studies had to include the ability to walk independently (primary outcome) as an outcome measure. Secondary gait-related outcomes were included, such as motor impairments of the affected leg and different measures on walking performance. main interventions, body weight supported treadmill training (BWSTT) and robot-assisted gait training (RAGT), in other databases (Cochrane Library, Rehab Data, PEDro) the reviewers searched explicitly for those interventions. After de-duplication and a first phase screening on eligibility, 132 unique studies were included for detailed screening on abstract and after- wards on full-text. Finally, 15 studies were included (see Fig. 1). A revision in August 2018 did not reveal additional eligible studies. Quantitative analysis In the final screening phase, 4 studies were excluded due to insufficient quality. Of the remaining 15 studies, 9 presented good (PEDro score 8 (35–37); 7 (38–41)) and 6 moderate quality (PEDro score 6 (42–47); 5 (48, 49)) (see Table II). A detailed scoring is shown in Table SII 1 . Review Manager software (RevMan 5.3) was used for the quan- titative synthesis on the comparative effectiveness. Therefore, the number of participants in both groups together with the means of post-intervention and follow-up scores and its standard de- viations were entered in RevMan 5.3 by one reviewer (JS) and cross-checked by another reviewer (WS or ST). If the scores were provided in medians and interquartile ranges, an algorithm developed by Wan et al. (33) was used to estimate means and standard deviations. Summary effect sizes (SES) were calculated with 95% confidence interval (95% CI) based on the effect sizes of individual studies. The mean differences (MD) were calcu- lated since identical measures were used per comparison. When dichotomized outcome on walking independence was reported, an odds ratio was additionally calculated. The I 2 statistic was used to determine between-study heterogeneity in results. If heterogeneity was high (I 2  > 50%) a random-effects model was used. In each comparison, a sub-analysis on the intervention type was performed. If at least 3 RCTs could be included in a sub- group, the results were reported separately. In addition, if results of 2 or more subgroups were given, the subgroup difference was established using a χ 2 test. Finally, the level of evidence drawn from the quantitative analysis were graded using a classification adapted from the Scottish Intercollegiate Guideline Network (SIGN) guidelines (34), where the methodological quality of included RCTs and consistency of results (based on the I 2 test for heterogeneity) will be taken into account (see Table I). Methodological quality Outcomes In the 15 studies, a total of 915 participants were treated and evaluated: RAGT was provided to 286 participants and BWSTT to 152 participants, while 425 participants were allocated to the control groups. PubMed 330 Web of Science 464 PEDro 42 RehabData 14 Cochrane Lib 31 Potentially relevant citations identified: 881 Potentially relevant citations identified: 132 1. de-duplication & titel/abstract evaluation, citations excluded: 749 2. detailed titel/ abstract evaluation, citations excluded: 78 Reasons: RESULTS Literature search In PubMed, the search strategy (see Table SI 1 ) led to 330 hits on 24 October 2017 and a similar strategy was used in Web of Science. After identifying the 2 Citations retrieved for more detailed evaluation: B C D RCT: randomized controlled trial. www.medicaljournals.se/jrm 5 S (no RCT) 20 53 54 3. Full text evaluation, citations excluded: 39 Reasons: Conclusion based on … At least 3 RCTs with a low risk of bias with consistent results and a clinical meaningful effect. At least 2 RCTs with a low risk of bias but inconsistent results, or at least 2 RCTs with a moderate risk of bias with consistent results. One RCT with a low risk of bias, or several RCTs with a moderate risk of bias with inconsistent results. Lower. P (chronic) I (no gait training) Table I. Rating the level of evidence adapted from the Scottish Intercollegiate Guideline Network (SIGN) guidelines A J. Schröder et al. P (>1mo post) 11 I (no gait training) 12 P (FAC >3) S (no RCT) S (PEDro 4) Relevant citations: 8 4 4 15 Fig. 1. Flow diagram of study identification and selection process. P: participants; I: intervention; S: study design.