“…All eight included studies were judged to have a moderately high risk of bias (see table 2). Features increasing the risk of bias were: absence of relevant baseline characteristics (Hanks et al, 2012), not fully concealed allocation (Hsieh et al, 2012), patients serving as their own controls (Anson and Ponsford, 2006b;c;Wolters et al, 2010), matching of (non-randomized) groups (Bradbury et al, 2008;Tiersky et al, 2005), high levels of attrition which was not adequately accounted for (Backhaus et al, 2010; Tiersky et al, 2005) and/or outcome measures which were not formally validated (Backhaus et al, 2016;Backhaus et al, 2010).…”