Disagreements on eligibility were first resolved by discussion and decided by a third reviewer (CL) if disagreement persisted. Design • Repeated measures between raters Participants • Symptomatic and

asymptomatic individuals Measurement procedure • Performed passive (ie, manual) physiological or accessory movements in any of the joints of the shoulder, elbow, or wrist-hand-fingers Outcomes • Estimates of inter-rater reliability Description: We extracted data on participants (number, age, clinical characteristics), raters (number, profession, training), measurements (joints and movement direction, position, movement performed, method, outcomes Vandetanib datasheet reported), and inter-rater reliability (point estimates, estimates of precision). Two reviewers (RJvdP and EvT) extracted data independently and were not blind to journal, authors, or results. When disagreement between reviewers could not be resolved by discussion, a third reviewer (CL) made the final decision. Quality: No validated instrument is available for assessing check details methodological quality of inter-rater reliability studies. Therefore, a list of criteria for quality was compiled derived from the QUADAS tool, the STARD Statement, and criteria used for assessing studies on reliability of measuring

passive spinal movements ( Bossuyt et al 2003a, Bossuyt et al 2003b, Van Trijffel et al 2005, Whiting et al 2003). Criteria were rated ‘yes’, ‘no’, or ‘unknown’ where insufficient information was provided ( Box 2). Criteria 1 not to 4 assess external validity, Criteria 5 to 9 assess internal validity, and Criterion 10 assesses statistical methods. External validity was considered sufficient if Criteria 1 to 4 were rated ‘yes’. With respect to internal validity, Criteria 5, 6, and 7 were assumed to be decisive in determining risk of bias. A study was considered to have a low risk of bias if Criteria 5, 6, and 7 were all rated ‘yes’, a moderate risk if two of these criteria were rated ‘yes’, and a high risk if none or only one of these criteria were rated ‘yes’. After training, two reviewers (RJvdP, EvT) independently assessed methodological quality

of all included studies and were not blind to journal, authors, and results. If discrepancy between reviewers persisted after discussion, a decisive judgement was passed by the third reviewer (CL). 1. Was a representative sample of participants used? Data were analysed by examining ICC and Kappa (95% CI). ICC > 0.75 indicated an acceptable level of reliability (Burdock et al 1963, cited by Kramer and Feinstein 1981). Corresponding Kappa levels were used as assigned by Landis and Koch (1977) where <0.00 = poor, 0.00–0.20 = slight, 0.21–0.40 = fair, 0.41–0.60 = moderate, 0.61–0.80 = substantial, and 0.81–1.00 = almost perfect reliability. In addition, reliability was analysed relating it to methodological quality and risk of bias.