Types of studies

We included randomised controlled trials (RCTs) only. We included reports of RCTs prepared in any language, regardless of their publication status.

Types of participants

Participants had open angle glaucoma (OAG) of any type, including primary and secondary OAG. As there are no universally‐accepted criteria by which glaucoma may be defined, we permitted studies to use their own definitions of glaucoma, provided these were clearly stated. We also included participants with ocular hypertension, normal tension glaucoma, or possible glaucoma (suspects for glaucoma).

We excluded trials with participants with closed angle glaucoma.

We did not apply any restrictions regarding location, setting, or demographics.

Types of interventions

We compared ab interno trabecular bypass surgery with the Hydrus microstent (Ivantis Inc., Irvine, California) to:

• laser treatment (selective laser trabeculoplasty or argon laser trabeculoplasty);

• other minimally‐invasive glaucoma surgery (MIGS) techniques;

• conventional glaucoma surgery (trabeculectomy);

• medical therapy.

Types of outcome measures

We did not use the reporting of particular outcomes as a criterion for eligibility for the review. We did not exclude studies from the review solely on the grounds of not reporting an outcome of interest.

We reported outcomes in the short‐term (six to 18 months), medium‐term (18 to 36 months), and long‐term (longer than 36 months).

Electronic searches

The Cochrane Eyes and Vision Information Specialist conducted systematic searches in the following electronic databases for RCTs and controlled clinical trials. There were no restrictions to language or year of publication. The date of the search was 7 May 2019.

• Cochrane Central Register of Controlled Trials (CENTRAL; 2019, Issue 5 (which contains the Cochrane Eyes and Vision Trials Register)) in the Cochrane Library (searched 7 May 2019; Appendix 1;

• MEDLINE Ovid (1946 to 7 May 2019; Appendix 2);

• Embase Ovid (1980 to 7 May 2019; Appendix 3);

• International Standard Research Clinical Trial Number (ISRCTN) registry (www.isrctn.com/editAdvancedSearch; searched 7 May 2019; Appendix 4);

• US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov; searched 7 May 2019; Appendix 5);

• World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP; www.who.int/ictrp; searched 7 May 2019; Appendix 6).

Searching other resources

We searched the reference lists of included studies for other possible studies, and contacted any individuals or organisations who conducted relevant RCTs. We also searched the website of the manufacturer (Ivantis Inc., Irvine, California; www.ivantisinc.com) for any information on forthcoming trials. We are awaiting additional data on included studies from the manufacturer.

Selection of studies

Four review authors independently screened titles and abstracts of all articles identified by the search, using web‐based online review management software (Covidence). If abstracts were not available, we screened full‐text articles. Two review authors independently assessed the full‐text reports of all potentially eligible studies. If there was disagreement regarding eligibility, a third review author arbitrated. If any full‐text reports were rejected, we recorded the reasons for this.

Data extraction and management

We extracted data from reports of included studies using a data collection form, which was developed, but not piloted on the first five studies included as planned. Three review authors independently extracted study characteristics from reports of each study, and two review authors (AS, GV) entered the data for the studies into Review Manager 5 (RevMan 5 (Review Manager 2014)). Two review authors (GV, FO) independently extracted data for the analyses, and one review author (GV) checked the data, and then entered it into RevMan 5. If there was disagreement, a third review author arbitrated.

Data collected in Appendix 7 were presented in the 'Characteristics of included studies' table. Where data on included studies (or ongoing studies) were missing or unclear, we contacted the individuals or organisations involved to obtain clarification. We collected and used the most detailed numerical data available to facilitate analyses of included studies. We obtained these data from individuals or organisations in preference to less precise methods, such as extracting numeric data from graphs, as indicated.

Assessment of risk of bias in included studies

We used the latest version of the Cochrane 'Risk of bias' tool, as described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions to assess and judge the risk of bias for included studies (Higgins 2017).

Measures of treatment effect

We calculated the risk ratio for the following outcomes: proportion of participants who were medication‐free (not using eye drops); proportion of participants who achieved an IOP of no more than 21 mmHg, 17 mmHg, and 14 mmHg; proportion of participants who required further glaucoma surgery; and proportion of participants who experienced intra‐ and postoperative complications.

When data were available, we calculated the mean difference for the following continuous outcomes: mean change in IOP; mean change in number of IOP‐lowering drops instilled per day; and mean change in quality of life.

Where possible, we checked for the skewness of continuous data (Altman 1996).

Unit of analysis issues

We noted whether studies included one or two eyes from each participant, and whether randomisation was conducted at the level of the participant or the eye. There is a potential for medical treatments, such as topical beta blockers used for one eye, to influence the outcome in the other eye (Piltz 2000). Surgery to lower IOP in one eye may also affect the IOP of the fellow eye (Radcliffe 2010). Therefore, we excluded studies that had adopted a paired design.

Dealing with missing data

We tried to minimise missing outcome data by contacting individuals and organisations to obtain them. Because the level of missing data in each group and reasons for missing data in each group were similar, we analysed available case data. We are waiting for the manufacturer of the Hydrus microstent to provide us with unpublished data, which we will use in the update of this review.

Assessment of heterogeneity

We assessed the heterogeneity between trials by carefully examining the study reports, assessing forest plots, and examining the I² value. We considered I² values greater than 50% to be indicative of substantial heterogeneity, suggesting that meta‐analysis might not be wise. We also considered the consistency of the effect estimates. If all estimates were in the same direction, we pooled the data, even when heterogeneity was evident; we commented on any heterogeneity in the Discussion section.

Assessment of reporting biases

We planned to develop a funnel plot to assess the risk of publication bias if there were more than 10 trials in our review.

Data synthesis

We undertook a meta‐analysis when data appeared clinically, methodologically, and statistically homogeneous. We checked that participants, interventions, comparators, and outcomes were sufficiently similar to give a clinically meaningful result, and that our I² result did not indicate considerable inconsistency (i.e. I² less than 50%). In future updates of this review, we will pool heterogenous data if all estimates are in the same direction. We used a fixed‐effect model, as there were fewer than three trials included in the meta‐analyses.

Subgroup analysis and investigation of heterogeneity

We do not plan to conduct subgroup analyses in future updates of the review.

Sensitivity analysis

We planned to assess the impact of including studies at high risk of bias for an outcome in one or more key domains. However, there were too few included studies to conduct such analyses.