What do we know about anti–vascular endothelial growth factor (VEGF) therapy for neovascular age-related macular degeneration (AMD)? Briefly, I think we know that pegaptanib is safe, but compared with newer anti-VEGF drugs, not as effective, and ranibizumab and aflibercept are safe and effective. To my view, bevacizumab seems to be safe and effective with outcomes similar to those seen with ranibizumab. I think we know that monthly and as-needed treatment approaches for bevacizumab and ranibizumab yield similar outcomes. We also know that treating at 3-month intervals with ranibizumab yields suboptimal outcomes. We know that bevacizumab therapy is less expensive than the other therapies.

What are some things I would like to know about anti-VEGF therapy for neovascular AMD? I would like to know if there are subtle safety differences between the treatment options, for how long (how many years) we need to treat, and how infrequently we can treat and schedule visits yet achieve outcomes comparable with those of the registration trials? I would also like to know whether tachyphylaxis exists, and whether geographic and or photoreceptor atrophy, glaucoma, or other possible long-term adverse effects are real findings? Importantly, although the outcomes using the various drugs seem comparable for populations, I would like to learn if there are there individual patients who do better with one drug as opposed to one of the others? Better still would be if we can identify those patients early on as we initiate treatment.

This issue of the Journal contains 3 articles based on retrospective study designs that look at switching, so these studies offer initial looks at whether some patients do better on one drug or the other? Nonrandomized switching studies are challenging to interpret because of potential behavioral and technical influences such as the possibility of regression to the mean. We tend to switch subjects who are not doing well. Some are truly not doing well, but others are only apparently not doing well, either because of temporary random variation or because they in fact will do well with longer follow-up or treatment. When switched, those who are truly failing may respond to the new therapy, but those who would have done well had they stayed the course are apparent winners with the new treatment (but would also have won with the old one and more time or more of it).

Investigators from Boston report on 102 eyes of 94 patients with either refractory or recurrent neovascular AMD switched from bevacizumab or ranibizumab to aflibercept. At a mean of 18 weeks, vision was stable and anatomic measures, for example, central macular thickness, improved. In Iowa City, 36 eyes from 31 patients resistant to bevacizumab or ranibizumab were switched to aflibercept. Vision over time did not seem to change, but again, an anatomic response was seen based on optical coherence tomography. Finally, 96 eyes of 85 subjects in Atlanta with persistent, recurrent, or worsening exudation were switched from either bevacizumab or ranibizumab to aflibercept. Acuity was stable, and again, there seemed to be an anatomic response.

Retrospective study designs have flaws, yet still may have teaching value with implications for improved patient care. More certainty is seen with, among other things, clear inclusion and exclusion criteria, longer and more complete follow-up, larger sample sizes, and outcomes reported at standard time points. It is key to understand who was studied and what clinical parameters led to the switching. In Boston “subjects were identified…with neovascular AMD [who were switched].” There was a refractory group with persistent intraretinal or subretinal fluid despite monthly injections and a recurrent group who initially responded well, but required frequent maintenance injections to maintain a dry macula. In Iowa City, subjects who were resistant were switched. There had to have been initial anatomic responses followed by recurrence or persistence. Finally in Atlanta, switching to aflibercept was based on “persistent, recurrent, or worsening exudative fluid or hemorrhage on examination [or optical coherence tomography].”

For treatment-naïve subjects, aflibercept achieves similar outcomes compared with ranibizumab and, after 3 monthly initial doses, can be administered every 2 months. Regular, ranibizumab or bevacizumab injection every 2 months has not been studied and could work also. The every-8-week aflibercept dosing offers an advantage over ranibizumab monthly, but since the Complications of Age-Related Macular Degeneration (CATT) and IVAN (A randomised controlled trial of alternative treatments to Inhibit VEGF in Age-related choroidal Neovascularisation). treatment trials reports were published, I have learned that monthly dosing of ranibizumab and bevacizumab are not needed (although monthly visits are, if one follows the CATT and IVAN protocols).

The 3 reports in this issue offer evidence that some patients with persistent activity or apparent lack of response to ranibizumab or bevacizumab benefit anatomically by switching to aflibercept. We already have seen in CATT that anatomic outcomes differ between ranibizumab and bevacizumab, yet to my view, vision outcomes do not. To understand if switching will allow improved vision outcomes will require prospective studies with a priori determined definitions of terms such as lack of response and adequacy of prior treatment. Such studies are underway. It would be ideal if these studies randomly switched half of the subjects and maintained a contemporaneous nonswitched comparison group followed on the same schedule and with similar indications for reinjection.

It would be nice if genetic make-up predicted in which patients CNV or AMD progression, or both, develop. It seems to for populations, which is very useful for clinical trial design, but the incremental predictive value for individual subjects is small. For me, when asked to predict likelihood of progression, I enjoy the simplicity of the Age-Related Eye Disease Study simple system. Studies concerning predictions of a response to ranibizumab or bevacizumab have yielded discrepant findings, but one of the largest and most robust looks at the issue, in CATT, did not identify a genetic marker that predicts response to one drug or the other.

While we wait for more definitive answers, what should we do in the meantime? Although one can debate the meaning of a few letter difference in various outcomes or hints of systemic safety issues that will require very large data sets to confirm or refute, those of us who are comfortable using ranibizumab probably will use it, and those who for their own reasons favor bevacizumab or aflibercept will use that drug. We need a uniform agreed-on definition of nonresponder. After that definition is met, it makes sense to switch drugs.

I congratulate Drs Yonekawa (Boston), Bakall (Iowa City), and Ho (Atlanta) and their colleagues for their very nice first steps looking into the switching to aflibercept question.