Fig. 3

Risk of bias assessment. ¹ random sequence generation: none of the articles described random sequence generation in detail. The study by Bialoszewski et al. [12] is affected by an even higher risk since patients were not randomized primarily but only if they developed oedema during treatment. Chan et al. [21] recruited patients with and without meniscal surgery which might be medically reasonable but is methodologically disputable. Windisch et al. [11] performed no randomization but used a historical control. Gulenc et al. [17] describe randomization “ based on the rank of admission” in their study on kinesiotaping after knee arthroscopy. ² Allocation concealment is not described or doubtful (picking of envelopes)

³ Blinding of participants and personnel is not feasible in this context since the effect of sham taping with an alternative material has not been explored and control treatment like manual lymphatic drainage or intermittent pneumatic compression cannot be concealed either. ⁴ All but one articles fail to mention a blinding of the assessor, only Donec et al. [22], Ristow et al. [16, 19, 20] and Tozzi et al. [21] name the assessor. Balki et al. [14] describe a separation of assessor and researcher. ⁵ Bialoszewski et al. [12] miss to report the exact duration of treatment as well as the exact timing of assessment, Chan et al. do not mention the exact timing of assessments. Donec et al. [22] fail the reporting of basic measurements preoperative and retrospectively retrieve data on use of analgesics from patients’ charts. Tozzi et al. [21] do not report the beginning of treatment.

⁶ In spite of the overall high risk of bias in all the studies a tendency for selective reporting cannot be observed.