2.2. Assessment of the Characteristics of Studies

Reviewers (Y.L.T. and Y.C.C.) independently performed quality validity and critical appraisal of the included trials through the use of the Cochrane Risk of Bias Tool [31], PEDro scale score [32], and Modified Jadad Scale [33]. According to the Cochrane Risk of Bias Tool, trials were classified as being high risk if they met less than five of the eight criteria. The PEDro scale score consisted of 11 items, each valued at 1 point if qualified. The higher the total score, the better the methodological quality. Similarly, the Modified Jadad Scale had eight items/points, and studies with 4 to 8 points being considered good to excellent quality.

For non-randomized clinical trials, the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) assessment tool [34] was applied. The number of qualified items was in proportion to the methodological quality of the corresponding study.

Participants in the included studies involved patients with BCRL.

The protocol of ESWT was recorded as intervention: type of ESWT (focused or radial); treatment location, frequency, and intensity; the total number of treatment sessions. We also documented the use of CDT, with or without ESWT, in both the experimental and control groups.

The primary outcomes of our review were the decrease in volume of lymphedema and circumference of the affected limb. The secondary outcomes were the skin thickness at fibrotic areas, increase of shoulder joint range of motion (ROM), and the results from the quick Disability of the Arm, Shoulder, and Hand (qDASH) questionnaire [35]. Studies needed to consist of at least one of these results to be included.

Name of the author, year of publication, the country where the study was conducted, number of patients, characteristics of patients, therapeutic protocol, and outcomes mentioned above were all extracted if possible. We contacted the authors by email for any missing data and other uncertain issues, which would subsequently be marked as N/A or unclear if we received no response. A study would not be included in the meta-analysis if the necessary data were inaccessible. Based on the Cochrane Handbook [31], continuous data were expressed as the mean ± standard deviation and summarized as a Mean Difference (MD) or Standardized Mean Difference (SMD). We used Cochran’s chi-squared test (Q test) and the I-squared test to assess statistical heterogeneity. All results were reported with 95% confidence intervals (CIs), with p < 0.05 considered statistically significant. A fixed-effects model was used when no obvious statistical heterogeneity existed (I² value < 50%); otherwise, a random-effects model was applied (I² value > 50% and p < 0.01). Between-group comparison was conducted if enough data from the RCTs (≧2 studies) could be acquired. In case of high statistical heterogeneity (I² value > 75%), sensitivity analysis or funnel plot would be performed if the number of trials sufficed. We performed the systemic review and meta-analysis with Review Manager software (version 5.4; Cochrane Collaboration, London, UK) for graphical representation of the pooled data.

We adopted the Grading of Recommendations Assessment, Development and Evaluation (GRADE) [36,37] framework to assess the intergroup certainty of evidence.