This review was based on a published protocol (Grimsby 2014), with differences as described here.

In addition to the previously published methodology, the review included plans for GRADE assessment and for preparation of a 'Summary of Findings' tables using GRADEpro GDT software. GRADE assessment is reported in the review, as well as within the 'Summary of findings' table alongside the reportable primary outcome for the main comparison estimable from the available literature.

Due to a combination of low quality available evidence and overall limited numbers of available studies, only one full prospective randomized study was included. There were no high quality descriptive studies available, only low quality descriptive studies by GRADE criteria; therefore, these descriptive studies were not included as planned.

The Newcastle‐Ottawa scoring scale was not utilized for assessment of risk of bias, as the review included no non‐randomized studies (Wells 2012).

Several data collection and analysis steps were not performed since we found no relevant data. For measurement of treatment effect, only dichotomous outcome information was available from the single included study, so no analysis of continuous outcomes data was performed. Similarly, no special analysis was undertaken for unit of analysis issues, as this was not relevant to the single included study, neither was an assessment of heterogeneity of data undertaken. Likewise, data synthesis, subgroup and sensitivity analysis steps were omitted since only one study met inclusion criteria.

We dropped the secondary outcome of number of retreatment per year predefined in the protocol as it was too similar to the primary outcome of retreatment to add value. However, there was no evidence for either outcome.