The trials in these forest plots are arranged to illustrate the subgroup analysis, BTK inhibitor which identified no considerable difference between the low-intensity and moderate-intensity subgroups. Although the best estimate of the overall effect on lymphoedema incidence favoured weight training, this was not statistically significant (RR 0.77, 95% CI 0.52 to 1.15), as presented in Figure 4. See Figure 5 on the eAddenda for a more-detailed forest plot. Again, subgroup analysis identified no considerable difference between the low-intensity and moderate-intensity subgroups. Meta-analysis of four comparisons21, 22, 26 and 39 with upper limb strength as the outcome showed

better results in the weight-training group than the controls, which was statistically significant (SMD 0.93, 95% CI 0.73 to 1.12). The low-intensity and moderate-intensity subgroups again had similar results. This meta-analysis is presented in Figure 6. See Figure 7 on the eAddenda for a more-detailed forest plot. In addition, a study by Kilbreath and colleagues45 reported individual muscle group strength contrary to other studies, which reported bench press, so it was not included in the overall effect estimate. Although one result in this study (horizontal

flexion strength) favoured the control check details group, it was not statistically significant and the other shoulder movements tested showed some improvement with weight-training exercise. Meta-analysis of lower limb strength data from the same four trials21, 22, 26 and 39 also showed significantly better results in the weight-training group than the controls (SMD 0.75, 95% CI 0.47 to 1.04). This meta-analysis is presented in Figure 8. See Figure 9 on the eAddenda for a more-detailed forest plot. The low-intensity and moderate-intensity subgroups again had similar results. The overall effect based on three studies21,

22 and 39 that reported body mass index revealed no significant benefit of weight training (SMD –0.10, 95% CI –0.31 to 0.11), as presented in Figure 10. See Thymidine kinase Figure 11 on the eAddenda for a more-detailed forest plot. All three of these trials used a low-intensity intervention, so no subgroup analysis was performed. Six trials provided data related to quality of life. Three trials26, 39 and 40 reported global quality of life scores whereas the rest21, 22 and 46 reported only individual domains of the quality of life scale. The forest plot in Figure 12 therefore presents pooling by these two subgroups, without a single overall result. A more detailed forest plot is available in Figure 13 on the eAddenda. The global quality of life score showed a positive trend towards the weight-training group. The Physical Health domain score demonstrated a significant overall improvement (SMD 0.34, 95% CI 0.09 to 0.58) in the weight-training group compared to the control group.