Disagreements were resolved by discussion. The inclusion criteria for the review are presented in Box 1. Design • Randomised trials Participants • Adults after total hip replacement Interventions • Post-discharge physiotherapist-directed rehabilitation exercises (outpatient or home-based) Outcomes measured • Muscle strength Comparisons • Post-discharge physiotherapist-directed rehabilitation S3I-201 exercises (outpatient or home-based) versus no intervention Quality: Trials meeting the inclusion criteria were assessed for methodological quality using the PEDro scale ( Maher et al 2003) by two reviewers (CC and JS). Each assessor worked independently. Following assessment, any disagreements were resolved by discussion.

The ten internal validity items of the PEDro scale were reported as a total score ( de Morton 2009). The external validity item, which requires both the source of participants and the eligibility criteria to be reported, was also determined for each trial. The PEDro scale scores were used to characterise the trials but were not used to exclude trials from the review or the meta-analyses. Participants and interventions: Interventions involving early rehabilitation during the hospital inpatient phase, post-acute inpatient rehabilitation, and rehabilitation in residential care (or comparison to any of these) were not considered

by this AZD9291 review. Outcomes: The outcomes considered by the review were muscle strength, gait, function and quality of life. From each trial, data were extracted for these outcome measures, where available, at the beginning of the intervention and at the longest follow-up assessment point. Data were extracted from each trial regarding sample size, population characteristics, details of the interventions, and the effects of interventions. Where outcome measures were reported in two or more trials and were reported ADP ribosylation factor by population descriptors (mean and standard

deviation), meta-analyses were performed using standard softwarea. Where only one trial reported a particular measure, meta-analysis was not used but the data were reported in the text as a between-group difference with a 95% CI. To determine the effect of intervention, experimental and control groups were compared. Where a trial employed two variations of physiotherapy intervention, the outcomes of the two intervention groups within that trial were pooled before performing this meta-analysis. Also, to determine which mode of post-discharge physiotherapy provides better patient outcomes following total hip replacement, we meta-analysed the studies in which outpatient and home-based exercise programs were compared. Forest plots were created to display effect estimates with 95% CIs for individual trials and pooled results. In each case we tested for statistical heterogeneity. This was examined graphically on the forest plot and statistically through the calculation of the I2 statistic.