Visual and anatomical outcomes of early versus delayed anti-VEGF switching after bevacizumab loading in diabetic macular oedema

Purpose To compare 24-month visual and anatomical outcomes associated with early versus delayed switching from bevacizumab in treatment-naive eyes with center-involving diabetic macular edema (DME) and suboptimal early response. Methods This retrospective cohort included 87 eyes with inadequate response (<5 ETDRS-letter gain and/or <20% central macular thickness [CMT] reduction) after three monthly bevacizumab injections. In a month-3 landmark analysis, eyes were classified by the month-3 treatment decision as immediate switch ( n = 37) or continued bevacizumab with later switching ( n = 50). An observed-timing comparison excluded eyes switched at months 4–5 ( n = 7) and compared early switch (month 3; n = 37) with delayed switch (month ≥6; n = 43) to aflibercept 2 mg or ranibizumab. Outcomes from the month-3 decision point were evaluated using adjusted mixed-effects models with categorical time. Results At month 3, BCVA (0.42 versus 0.45 logMAR) and CMT (437 versus 408 µm) were similar in the early- and delayed-switch groups. In the landmark analysis, adjusted 24-month differences were −0.07 logMAR (95% CI, −0.17 to 0.03; P = 0.171), equivalent to 3.5 ETDRS letters favoring immediate switching (95% CI, −1.5 to 8.5), and −36 µm (95% CI, −71 to −1; P = 0.044). In the observed-timing comparison, they were −0.08 logMAR (4.0 letters; 95% CI, −1.5 to 9.5; P = 0.154) and −39 µm (95% CI, −77 to −1; P = 0.044). A ≥ 20% CMT reduction at month 24 occurred in 62.2% versus 37.2%. Total injections numbered 11.2 ± 2.4 versus 14.5 ± 2.6. Conclusions Earlier switching was associated with greater anatomical improvement in both analyses. A 24-month visual advantage was not demonstrated, although a clinically meaningful benefit could not be excluded.