Aim To determine whether early vision gains predict long-term visual outcomes in the BEVORDEX randomised clinical trial of bevacizumab or dexamethasone implants for diabetic macular oedema.

Methods Post hoc analysis of 68 study eyes (77%) that completed 2 years follow-up of the BEVORDEX multicentre randomised clinical trial set in Australia (ClinicalTrials.gov identifier: NCT01298076). Study eyes from both groups were combined and stratified by visual acuity (VA) change in the first 12 weeks in to three groups: (a) suboptimal gain: <5 letters gain (includes VA loss), (b) moderate gain: 5–9 letters gain, (c) pronounced gain: ≥10 letters gain. This was correlated with VA outcome at 104 weeks taking into account treatment allocation and baseline lens status.

Results The change in VA in the first 12 weeks was significantly correlated with VA change at 104 weeks (p<0.001). This was independent of treatment allocation (p=0.353) and lens status at baseline (p=0.593). The change in central macular thickness at 12 weeks did not correlate with VA gain at 104 weeks (p=0.847).

Conclusion Short-term visual gain at 12 weeks was strongly correlated with long-term vision improvement independent of treatment allocation or baseline lens status. Early improvement in central macular thickness was not predictive of long-term visual outcomes.

Trial registration number NCT01298076, Post-results.