Glaucoma, a chronic degenerative disease of the optic nerve, is the leading cause of irreversible blindness worldwide. Although lowering the intraocular pressure (IOP) has been shown to be effective to slow optic nerve degeneration, a significant portion of glaucoma patients continue to develop progressive loss in vision despite adequate control of IOP. Development of neuroprotective therapy to prevent optic nerve degeneration by mechanisms other than IOP- lowering is critical to reduce the burden of glaucoma blindness. With 76 million glaucoma patients in 2020 worldwide, the need to investigate neuroprotection for glaucoma is pressing. While metformin is a widely adopted oral hypoglycemic medication for treatment of type 2 diabetes mellitus (DM), increasing evidence from clinical studies has shown that metformin can decrease the risk of many age-related diseases including neurodegenerative diseases. In a retrospective study of 150,016 patients with DM, those taking metformin at >1500mg/day had a 25% reduced risk of development of open-angle glaucoma than those who took no metformin. Metformin has a high safety profile. The investigators aim to investigate whether metformin can be repurposed to a neuroprotective therapy for glaucoma patients in a randomized controlled trial.

The investigators propose to conduct a 24-month, double-blind, placebo-controlled, parallel group, randomizing 125 primary open angle glaucoma patients who have progressive retinal nerve fiber layer (RNFL) and/or ganglion cell inner plexiform layer (GCIPL) thinning in at least one eye, as determined by optical coherence tomography Trend-based Progression Analysis (TPA) or Guided Progression Analysis (GPA), to receive metformin 1500mg/day or placebo. All patients will be followed up at 2-month intervals for IOP, RNFL thickness, and visual field (VF) measurements. The objectives are to compare (1) the rates of change of average RNFL thickness (primary outcome measure), and (2) the rates of change of VF mean deviation (MD) (secondary outcome measure) between treatment groups. The investigators hypothesize that patients treated with metformin have a slower rate of RNFL thinning, and a slower rate of VF MD decline compared with those treated with placebo at similar levels of IOP over the 24-month follow-up. The proposed study has the potential to mark a paradigm shift in the management of glaucoma patients by demonstrating that neuroprotection is attainable with metformin, which will alleviate the increasing burden of glaucoma blindness in China and other Asian countries where glaucoma patients with normal levels of IOP are prevalent. Furthermore, it will inform and impact the study design in future neuroprotection trials which can expedite the development of neuroprotective therapy for glaucoma.

125 POAG patients who show progressive RNFL/GCIPL thinning by TPA or GPA in at least one eye will be consecutively recruited from the HKU Eye Centre, Grantham Hospital after obtaining written informed consent at the screening visit. All investigations will be performed at HKU Eye Centre. The study will be conducted in accordance with the ethical standards stated in the 2013 Declaration of Helsinki and in compliance with Good Clinical Practice. They will have clinical examination, Goldmann applanation tonometry (GAT), OCT RNFL imaging, and perimetry at the screening, baseline, and follow-up visits. POAG is defined by eyes with open angles by dark room gonioscopy and glaucomatous optic disc damage (narrowed neuroretinal rim, RNFL defects, and optic disc excavation), with or without visual field defects, and without iris trabecular contact for >180 degrees by dark room anterior segment OCT.

Patients will be randomly assigned in a 1:1 ratio to receive, twice daily, oral metformin 750mg (i.e. 1500mg/day) or identical-appearing oral placebo. The investigators and patients will be blinded to the treatment assignment. Randomization with stratification by age, gender, VF MD, spherical equivalent, mean IOP measurement over the past 3 years, and number of glaucoma medications will be performed using an open-source computer program MinimPy. Patients and investigators will be unaware of treatment allocations and have no access to the randomization sequence or codes. The metformin and placebo tablets will be in identical appearance and uniformly packaged. OCT imaging of the RNFL, and VF testing will be performed at baseline, 1 month, 4 months, and every 2 months thereafter up to 24 months, with two repeated OCT and VF measurements at baseline, 1 month, and 24 months. Compliance with oral metformin/placebo use will be assessed by evaluating opened and unopened metformin/placebo packs at each follow-up visit.

Clinical examination including VA measurement, slit-lamp biomicroscopy for the anterior and posterior segments, and IOP measurement will be performed at the screening, baseline and all follow-up visits. Dark room indentation gonioscopy, anterior segment OCT imaging, axial length and central corneal thickness (CCT) measurements, and refraction will be performed at the screening and the last follow-up visits. Corneal hysteresis will be measured at the baseline visit using the Ocular Response Analyzer (Reichert Ophthalmic Instruments, Depew, NY, USA). Dilated fundus examination will be performed at the baseline visit and then yearly. Montreal Cognitive Assessment will be administered at the screening visit. Patients with a score of ≤25 will be excluded.

RNFL/GCIPL imaging will be performed with a swept-source OCT (Triton OCT, Topcon, Japan) using a wide-field scan covering the parapapillary region and the macula (512x256 pixels in 12x9mm2). Scans with a signal strength <40 (manufacturer recommendation), motion artifact, poor centration, RNFL segmentation errors or missing data (e.g. blinking) are discarded with re-scanning performed in the same visit. The average RNFL thickness will be measured for analysis of the primary outcome measure (i.e. the rate of change of average RNFL thickness). To identify patients with progressive RNFL/GCIPL thinning for study inclusion, RNFL/GCIPL thickness data of the individual pixels in serial RNFL/GCIPL thickness maps before the time of recruitment will be exported for TPA. The algorithm of TPA has been described. In brief, TPA performs pixel-by-pixel linear regression analysis between RNFL thickness and time for evaluation of progressive RNFL thinning after registering and aligning serial OCT scans in corresponding retinal locations of an eye. To minimize type I errors consequential to multiple testing in an eye, the RNFL thickness maps will be condensed from 512x256 pixels to 128x64 superpixels with a false discovery rate (FDR) controlled at 5% (an FDR of 5% suggests that 5% of the superpixels detected with a significant negative slope between RNFL thickness and time would be false positives). Progressive RNFL/GCIPL thinning is defined when there are ≥20 contiguous superpixels with a significant negative trend for at least two consecutive visits. Patients with progressive RNFL/GCIPL thinning detected by TPA or GPA (Carl Zeiss Meditec) in at least one eye will be invited for study inclusion.

Perimetry will be performed with the Humphrey Field Analyzer 3 (Carl Zeiss Meditec, Dublin, CA) using the 24-2 SITA standard program. A reliable VF test has fixation losses <20% and false positive rate <15%. Eyes with unreliable test results will receive repeated testing in the same visit. Definition of VF progression VF progression is identified when there are ≥3 points that showed significant change compared with 2 baseline examinations for at least 3 consecutive tests (i.e. "likely progression" in the Guided Progression Analysis) according to the EMGT criteria.

Inclusion Criteria:

• age ≥18 years
• best corrected VA ≥20/40
• IOP ≤24mmHg at the screening and baseline visits
• progressive RNFL and/or GCIPL thinning by TPA or GPA over the past 3 years in at least one eye.

Exclusion Criteria:

• patients with DM, kidney or liver diseases
• pathological myopia
• cognitive impairment (e.g. Alzheimer's disease)
• diseases that may cause visual field loss or optic disc abnormalities other than glaucoma
• inability to perform reliable visual field
• suboptimal quality of OCT images.

Both eyes of a patient will be included in the analysis if both eyesare eligible for inclusion.