PURPOSE: To evaluate the standards of harm reporting for glaucoma surgical trials and to develop a classification system for reporting surgical complication severity. DESIGN: Systematic review and Delphi consensus method. METHODS: Systematic review of glaucoma surgical trials published from January 2010 until July 2017 with a quality assessment against the CONSORT checklist for harm. A Delphi method was employed to generate consensus grading (interquartile range ≤ 2) among international glaucoma experts (n = 43) on severity of glaucoma surgical complications, and specifically for trabeculectomy and aqueous shunt complications, from 1 (no clinical significance) to 10 (most severe complication). RESULTS: Forty-seven studies were eligible. The items of the CONSORT checklist for harm that were most frequently missing were use of a validated instrument to report severity (0%), withdrawals due to harm, and subgroup analyses, both reported in 3 publications (6.4%). Most glaucoma experts participating in the Delphi process (80%) completed the second round, and consensus was achieved for all but 1 complication. The least severe complications (graded 2) were "transient loss of vision," "early low intraocular pressure," "choroidal detachment anterior to equator," "small layered hyphema < 1 mm," and "increased lens opacity not clinically significant." The most severe complications (graded 10) were "endophthalmitis" and "permanent severe loss of vision (hand movements or worse)." CONCLUSIONS: Glaucoma surgical randomized controlled trials report frequency of complications, but their severity is rarely reported. The quality of harm reporting is poor. We propose the use of a newly developed system of classification for assessing the severity of surgical complications based on consensus.