For continuous data, standardised mean differences (otherwise known as effect sizes), with 95% CIs were calculated by dividing the post-intervention means by the pooled standard deviation (Hedges g). Where means and standard deviations were not reported, data were estimated according to recommendations outlined by Higgins and Deeks (2009) (see Appendix 2 on the eAddenda for statistical equations).

A meta-analysis was conducted where a minimum of two trials were clinically homogenous. To account for clinical, methodological, or statistical heterogeneity, a pooled random effects model was applied using RevMan 5 a. Statistical heterogeneity was examined by calculating the quantity I2 where a value of 0% indicates no observed heterogeneity, Cilengitide less that 25% is considered to have low levels, and a value of 100% indicates a completely heterogeneous sample ( Higgins et al 2003). The search strategy identified 2375 papers. Following removal of duplicates, screening of titles and abstracts, and the inclusion of one paper identified through citation tracking

and one through hand searching of reference lists, 29 potentially relevant papers remained. After reapplication of inclusion criteria to full-text copies of these 29 papers, 14 papers remained (Figure 1). These 14 papers represented 13 separate BTK inhibitor library trials because two papers reported data from the same trial at different time points. The other 15 studies obtained as full text were excluded. Five were not randomised or quasi-randomised controlled trials (Altissimi et al 1986, Amirfeyz and Sarangi 2008, Clifford, 1980, Liow et al 2002, MacDermid et al 2001), one was not available in English (Grønlund et al 1990), one was published only as an abstract (Bache et al 2000), and Mephenoxalone eight had insufficient information about the exercise therapy intervention (Davis and Buchanan, 1987, de Bruijn, 1987, Dias et al 1987, Gaine et al 1998, Lozano Calderón et al 2008, McAuliffe et al 1987, Millett and Rushton, 1995, Oskarsson et al 1997). Design: A single trial evaluated the effects of exercise and home advice

compared to a no-intervention control group in patients with a distal radius fractures ( Kay et al 2008). In the remaining 12 trials, differing amounts of exercise and advice were incorporated in both control and intervention groups. Three trials compared exercise introduced earlier in rehabilitation with delayed introduction of exercise following a proximal humeral fracture ( Agorastides et al 2007, Hodgson et al 2003, Lefevre-Colau et al 2007), while in four trials patients received supervised exercise in addition to a home exercise program compared to simply a home exercise program ( Christensen et al 2001, Maciel et al 2005, Pasila et al 1974, Revay et al 1992). Five trials compared physiotherapy, which included supervised exercise plus a home exercise program, with a home exercise program ( Bertoft et al 1984, Krischak et al 2009, Lundberg et al 1979, Wakefield and McQueen 2000, Watt et al 2000).