Ietje Perfors

54 Chapter 3 Three studies published multiple articles based on the same data. 16–23 As a result, five RCTs and one CCT were considered eligible for inclusion, which were described in 10 articles. No additional eligible studies were identified in the reference lists of selected studies. Figure 2, table 1, and 2 show a detailed account of the risk of bias, patient population, interventions, outcomes assessed and observed results for each study. Given the various research questions, interventions and heterogeneity of outcome measures, pooling of data, and GRADE assessment were not feasible. Quality of studies The EPOC risk of bias is presented in figure 2. Luker et al 24 and Nielsen/ Kousgaard et al 16, 17 show a high risk of bias, resulting from high risk of selection and information bias. Drury et al 25 scored a medium risk of bias. And the studies of Johnson et al, 26 Johansson et al 23 and Bergholdt et al 18–21 show a low risk of bias. Regarding the RCT by Nielsen/Kousgaard et al 16, 17 several limitations should be kept in mind. The randomisation produced an imbalance, which influenced comparability of outcomes between study groups without corresponding correction in the analyses. Furthermore, it was not reported whether a baseline measurement was performed and the exact timing of the first measurement (table 2). Also, the percentage of missing data was 33% in the intervention and 26% in the control group. 16 Study populations The six eligible studies were conducted in Europe (five) and Australia (one) among different cancer patient populations over the past two decades. Patients with breast cancer were themost commonly studied group (between 33% and 100% of the study populations). Five RCTs included patients with more than one type of cancer, in different stages. Three studies included palliatively treated patients (<25% of total study population). In two RCT’s cancer stage was not specified.