Trials were not excluded on the basis of quality, although quality was taken into account when interpreting the results. Each item on the scale was scored as either ‘yes’ or ‘no’ and the number of items scored as ‘yes’ (excluding the first item, which PCI-32765 solubility dmso relates to external validity) was summed to give a total score out of 10. Trials scoring six or more were considered to be of high quality and trials scoring five or less were considered to be of low quality. For rating the quality of the evidence, the grading of recommendations assessment,

development, and evaluation (GRADE) approach was used. According to this system, the quality of evidence is assessed by rating the outcomes of the trials included in the review. The quality is then categorised as ‘high,’ ‘moderate,’ ‘low,’ or ‘very low’.12 Evidence based on randomised

trials begins as high-quality evidence and is downgraded for the following reasons: limitations in conduct and analysis (ie, risk of bias) of the studies; imprecision of the summary of the estimate of effect; inconsistency of the results across the available studies; indirectness or poor applicability of the evidence with respect to the populations, interventions, and settings where the proposed intervention may be used; 12 and evidence of publication bias. Downgrading for risk of bias could occur for: lack of allocation concealment; Selleck SB203580 non-blinding of participants, personnel, and outcome assessors; incomplete

outcome data; selective outcome reporting; or other sources of bias. 13 Non-blinding of participants and therapists was considered to be a major limitation and also resulted in downgrading. In studies almost with self-reported outcomes, lack of assessor blinding was considered to be a minor limitation and was not downgraded. For judging precision, the clinical decision threshold boundary for absolute difference was set at 1%. If this boundary was met, imprecision was not downgraded. If the absolute size excluded this boundary and if the sample size was small, imprecision was downgraded. 14 To inform this decision, the optimum information size was calculated to be 26 in each group, assuming α of 0.05 and β of 0.02. The difference in means between groups was taken as 1.4 cm, based on previous studies. If assessment of consistency of results indicated heterogeneity between studies, random-effects models were used for meta-analysis where appropriate.