Retinal and optic nerve measures after initiation of empagliflozin plus metformin versus metformin alone in early type 2 diabetes: a prospective observational comparative study

To explore whether empagliflozin added to metformin is associated with short-term changes in optical coherence tomography (OCT) and visual evoked potential (VEP) parameters in adults with newly diagnosed early type 2 diabetes mellitus (T2DM).

In this prospective observational comparative study, adults with newly diagnosed T2DM received metformin alone (M) or empagliflozin plus metformin (EM) according to routine clinical care. Outcomes were macular ganglion cell complex (GCC) and peripapillary retinal nerve fibre layer (RNFL) thickness measured by OCT, and pattern-reversal VEP parameters. Confounding by indication was addressed using propensity-score overlap weighting. Six-month change was analysed using baseline-adjusted overlap-weighted regression with HC3 robust standard errors. Secondary endpoints were controlled using the Benjamini–Hochberg false discovery rate. An additional exploratory repeated-measures mixed-effects analysis incorporating baseline, 3-month, and 6-month data was also performed. Only right eyes were included.

Eighty participants were analysed (40 per group). Overlap weighting achieved excellent covariate balance (maximum weighted absolute standardized mean difference, 0.017). At 6 months, mean GCC change favoured EM over M (β = +3.33 μm; 95% CI, 2.90 to 3.76; p < 0.001). Secondary outcomes also favoured EM, including shorter VEP P100 latency (β = −10.06 ms; 95% CI, − 12.07 to − 8.05; q < 0.001) and greater VEP N75–P100 amplitude (β = +4.17 µV; 95% CI, 3.21 to 5.14; q < 0.001), with directionally consistent findings across RNFL and GCC sectoral measures.

In this exploratory prospective observational study, empagliflozin added to metformin was associated with short-term OCT and VEP changes compatible with a potential neuroprotective signal in early T2DM. Findings require confirmation in longer-term randomised or robust quasi-experimental studies.