This study highlights the encouraging visual and anatomical outcomes of patients with DMO refractory to previous intravitreal therapy when treated with aflibercept under “real life” conditions. Patients had statistically significant letter gains at 6, 12 and 18 months in relation to baseline VA, with an overall gain of 7.9 letters sustained at 18 months. Patients had statistically significant less intra-retinal and sub-retinal fluid at six, 12, 18 and 24 months compared to baseline.
In primary treatment of DMO, Nepomuceno et al. showed similar CMT improvements with bevacizumab versus ranibizumab and at some visits (8 and 32 weeks) visual superiority of ranibizumab . Protocol T in primary treatment of DMO at 2 years showed equivalence between bevacizumab versus ranibizumab and aflibercept versus ranibizumab, but superiority of aflibercept over bevacizumab .
Lim et al.  reviewed 21 diabetic eyes receiving bevacizumab/ranibizumab treatment in Singapore and Massachusetts, USA, and found after aflibercept switch, a mean decrease of 107.0 μm in CMT at 5 months, similar to our decrease of 108.6 μm at 6 months. LogMAR at 5 months improved by 0.05, and our study patients demonstrated an improvement of 0.10 LogMAR at 6 months . Rahimy et al.  assessed 50 DMO eyes in Pennsylvania, USA, that received a mean of 13.7 prior bevacizumab/ranibizumab injections prior to switch and found after a 4.6-month follow-up period, a change in VA of −0.14 LogMAR and reduction of 184 µm compared to baseline CMT . Shah and Heier  found following a mean of 16 prior bevacizumab/ranibizumab injections borderline significant VA improvements and significant anatomical improvements after 2 months of aflibercept .
Possible reasons for the significant anatomical improvement with aflibercept are due to its different pharmacokinetic and pharmacodynamic properties. Aflibercept has a higher binding affinity to VEGF-A, leading to possibly sustained VEGF-A inhibition . Furthermore, it can bind to VEGF-B and PIGF which are two additional growth factors released during angiogenesis .
Another reason for the additional clinical response with aflibercept is tachyphylaxis, which patients may eventually succumb to, with repetitive treatment injections of bevacizumab/ranibizumab [10, 15, 16]. This has been established in many studies looking specifically at age-related macular degeneration (AMD). The mechanism is multifactorial, involving macrophage up-regulation of VEGF, varying surface receptor expression, changes in pharmacokinetics and development of neutralising antibodies as part of a systemic immune response . Forooghian et al.  retrospectively looked at 59 patients treated with bevacizumab (1.25 mg) for exudative AMD over a 14-month period and defined tachyphylaxis as a loss of therapeutic response to intravitreal bevacizumab approximately 28 days after administration in an eye which had previously responded to treatment. They identified 5 patients (6 eyes) who developed tachyphylaxis within a median time of 100 weeks (range 31–128 weeks) and the median number of injections to establish tachyphylaxis was 8 injections (range 5–10). Therapeutic response did not improve with a higher dosing regimen (2.50 mg) . Eghoj and Sørensen  looked at 1076 exudative AMD eyes (976 patients) and the potential tachyphylaxis effect with intravitreal ranibizumab. They found a tachyphylaxis rate of 2% (20 patients) in their cohort .
Overall, we found mild improvements in VA which did not match the dramatic anatomical improvements. Possible explanations for this could be the long-term damage to the minute retinal interconnections at the macula as a result of the persistent fluid.
We divided our patients in terms of baseline VA, dividing the cohort into those with better versus worse VA. Those with poor vision had statistically significant improvements in VA at each time point compared to baseline. The difference in VA between the two groups narrowed with time, although the change in LogMAR letters at each time point was not statistically significant. Protocol T described at 1 year that those with worse initial VA did better with aflibercept than those with better baseline VA . In our study, subjects who began aflibercept treatment with worse VA had a trend toward better improvements in vision compared to those who began treatment with good VA. This was not statistically significant possibly due to our small study population. Importantly, in our cohort, there was no difference in the number of injections between the two groups, indicating that subjects with worse VA still had improvements with the same frequency of treatment as those with good VA.
The mean number of bevacizumab/ranibizumab injections in the 6 months prior to switch was 3 (SD 0), representing injections administered weekly for eight weeks. Our centre began to use aflibercept in October, 2014 for select patients before the indication was expanded following protocol T (year 1) in 2015. Prior to aflibercept, the only available anti-VEGF treatment was bevacizumab and ranibizumab. Initially, patients were commenced on six-week regimens and clinical response was monitored. If patients did not respond or plateaued, the intensity of treatment was reduced and subsequently patients were classified as refractory or partial responders prior to the era of aflibercept. This explains why patients may not have been receiving six-week treatment injections in the 6 months prior to aflibercept in our centre. The introduction of aflibercept now offers these refractory and partial responders another option for further gains in VA.
Subjects received a mean of 3.7 (SD 0.8) aflibercept injections in the first 6 months of treatment. This is less than the recommended five injections every 4 weeks within the first 6 months. Rahimy et al.  treated patients with an injection frequency of 4.1 injections within 4.6 months equating to an aflibercept injection every 31 days. Despite more frequent injections in their study, our cohort still had comparable visual and anatomical improvements . A postulated advantage of aflibercept is the longer half-life and therefore potential reduction in injection frequency . After 6 months of aflibercept, our cohort showed a significant trend towards reduction in injection frequency.
This study adds to the limited evidence for aflibercept switch in persistent DMO patients as there are few published studies to our knowledge. The advantages of our study include the fact that the data is from a real-life setting which best reflects daily clinical practice. Our longer follow-up period averaging 22.5 months, when compared to other published studies which only have between 4.6 and 5 months follow-up, shows that improvements in VA and CMT are sustained following aflibercept switch.
Limitations of our study include the retrospective study design, but this best reflects the “real-life” setting. Our study has a small sample size, but this highlights novel experience with aflibercept. In AMD switch studies, it is suggested that such dramatic anatomical improvement with aflibercept may be due to the "loading dose effect" and overcoming tachyphylaxis associated with previous anti-VEGF therapies [18,19,20]. However, our study shows sustained visual and anatomical improvements long after the "loading dose effect" [19, 20]. A control group for comparison in a future randomised control trial would be essential for further evaluation of this effect.
The improvements in VA and CMT in our cohort could be due to the fact that patients were receiving eight weekly injections rather than a 4–6 weekly intensive regimen of bevacizumab/ranibizumab prior to switch. However, again, this represents the “real-life” clinical setting and increasing demands of a medical retina service.