DRCRnet U 2018.

Methods	Study design: parallel group RCT/within‐person study (40 clinical centres) Unit of randomisation:people and eyes if both eyes eligible Number randomised: 116 people, 129 eyes Unit of analysis: individual eyes Sample size: For a sample size of 70 study eyes per group (increased to 75 to account for approximately 5% loss to follow‐up), a 2‐sided 95% CI for the difference in the 2 means of visual acuity change from randomisation to 24 week visit will extend 2.3 visual acuity letter score in either direction from the observed difference in means, assuming that the common standard deviation is a letter score of 7, not adjusting for correlation between eyes in participants with 2 study eyes.
Participants	Country: USA Mean age: 65 years Sex: 67 women; 62 men Underlying condition: persistent DMO Inclusion criteria: age ≥ 18 years; diagnosis of diabetes mellitus (type 1 or type 2); meets all of the following ocular criteria in at least the one eye: at least 3 injections of anti‐VEGF drug (ranibizumab, bevacizumab, or aflibercept) within the prior 20 weeks, visual acuity letter score in study eye ≤ 78 and ≥ 24 LogMAR letters (approximate Snellen equivalent 20/32 to 20/320), on clinical exam, definite retinal thickening due to DMO involving the centre of the macula, OCT CMT thickness greater than 300 μm before adjusting for sex and machine‐specific factors Exclusion criteria: chronic kidney disease, renal transplant; unstable medical status including blood pressure, cardiovascular disease, and glycaemic control; participation in an investigational trial within 30 days of enrolment; systolic blood pressure > 180 mmHg; acute cardiovascular event within 1 month prior to enrolment; systemic steroid, anti‐VEGF or pro‐VEGF treatment within prior 4 months; pregnant; macular oedema is considered to be due to a cause other than DMO; an ocular condition is present (other than DMO) that, in the opinion of the investigator, might affect macular oedema or alter visual acuity during the course of the study; substantial posterior capsule opacity; history of intravitreal anti‐VEGF drug within 21 days prior to enrolment; history of intravitreal or peribulbar corticosteroids within 3 months prior to enrolment; history of macular laser photocoagulation within 4 months prior to enrolment; laser therapy within 4 previous months; any history of vitrectomy.
Interventions	Intervention: intravitreal ranibizumab (0.3 mg) and dexamethasone implant (0.7g) Comparator:intravitreal ranibizumab (0.3 mg) and sham injection Dexamethasone: sustained dexamethasone drug delivery system. Retreatment: ranibizumab was administered if the visual acuity letter score was less than 84 (approximate Snellen equivalent of 20/25 or worse) or the OCT CST was at or above the pre‐defined cut‐points. At the 4‐week and 8‐week visits, only ranibizumab injections were permitted. At 12 weeks and continuing through 20 weeks, retreatment was according to allocation group. A maximum of 2 injections of either dexamethasone or sham treatment were given in each eye.
Outcomes	Primary outcome measure: mean change in visual acuity letter score Secondary outcome measures: percentage of eyes with at least 10 and at least 15 ETDRS letter gain (increase) or loss (decrease) in visual acuity. visual acuity area under the curve (AUC) mean change in OCT central macular thickness percentage of eyes with worsening or improvement of diabetic retinopathy on clinical exam Length of follow‐up: 6 months (24 weeks)
Notes	Trial registration: NCT01945866 Source of funding: EY14231 (other grant/funding number: National Eye Institute), EY23207 (other grant/funding number: National Eye Institute), EY18817 (other grant/funding number: National Eye Institute) Declaration of interest: Several of the authors report receiving consultancy fees from the manufacturer of the implant, for example, quote "Dr Maturi reported receiving consultancy fees, research grants, payment for manuscript preparation, and travel, accommodations, and meeting expenses from Allergan, Inc" Date study conducted: February 2014 to December 2016
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote "Randomization was performed on the study website (http://www.drcr.net) using a permuted‐block design."
Allocation concealment (selection bias)	Low risk	Quote "Randomization was performed on the study website (http://www.drcr.net) using a permuted‐block design."
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Quote "Study participants and the medical monitor, who reviewed all adverse events, were masked to treatment group assignments. Refractionists, visual acuity testers, and OCT technicians were masked at the 24‐week primary outcome visit. Investigators and study coordinators were not masked."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote "Study participants and the medical monitor, who reviewed all adverse events, were masked to treatment group assignments. Refractionists, visual acuity testers, and OCT technicians were masked at the 24‐week primary outcome visit. Investigators and study coordinators were not masked."
Incomplete outcome data (attrition bias) All outcomes	Low risk	High follow‐up (127/129, 98%)
Selective reporting (reporting bias)	Low risk	Almost all outcomes on clinical trials registry were reported. The exception "Percent of eyes with worsening or improvement of diabetic retinopathy on clinical exam [ Time Frame: 24 weeks after randomization ]" was not an outcome of this review.
Other bias	Unclear risk	Both eyes being eligible to allocation in a single individual poses the theoretical potential to influence outcomes in the fellow eye via systemic absorption.