LETTERS

Protocol T: What We Don’t Know

The Diabetic Retinopathy Clinical Research Network (DRCRnet) Protocol T study, which addresses the treatment of center involved diabetic macular edema concludes aflibercept (Eylea, Regeneron, Tarrytown, NY) results in more lines of improved vision than bevacizumab (Avastin, Genentech, South San Francisco, CA) or ranibizumab (Lucentis, Genentech) if the presenting visual acuity is worse than 20/50. The study reported mean improvement with aflibercept was 18.9 letters, bevacizumab 11.8 letters, and ranibizumab 14.2. The authors state that the process of randomization created two similar groups (about 220 eyes in each group), with 89 sites participating.

The 11-page paper was published in the New England Journal of Medicine, while a supplemental appendix (52 pages) was published online. This made it challenging for most retinal specialists to review the paper. Hopefully, future papers will be more accessible to the retina community. The study’s conclusion has significant management and financial consequences, and we look forward to the two-year data.

After reviewing the paper and appendix, there are several issues that need clarification. The paper’s most important conclusion addresses visual acuity. Could other factors influence visual acuity, specifically in the subgroup of eyes presenting with 20/50 or worse visual acuity?

The study was not composed of all virgin eyes with recent DME. About one-third in each group had prior focal or grid laser photocoagulation, and/or VEGF-I injections, administered no more recently than one year prior to starting the study. Thus one-third of eyes had DME for more than one year, which was not treated for at least 12 months. The mean duration of DME, prior to enrollment for each treatment arm, and specifically for the <20/50 group, is not stated.

The statisticians, referring to the conclusions of the <20/50 group, argue that the size of the cohort and the process of randomization created similar groups; however, there are no published data to verify this assumption. We really don’t know which eyes in which treatment arm had DME for one, two, three, or more years before enrollment, which could bias the results.

The RISE/RIDE study conclusively showed that the duration of untreated DME (with VEGF-I injections), even if treated with macular laser, had a significant effect on the response to VEGF-I agents (ranibizumab). Eyes not treated with VEGF-I for two years did not regain vision to the level of eyes, which were promptly treated.

The paper does not give the reader any information regarding the duration of DME in any group, again asking the reader to take a leap of faith that the process of randomization created similar groups. The group of eyes treated prior to enrollment and then not treated for a year adds a great number of variables, which could affect the eye’s potential for vision recovery. Would the conclusions of the study be the same if all eyes previously treated, by any means, were purged from the trial?

Laser treatment is lumped into focal and/or grid. Focal laser for circinate threatening the macula has a good visual prognosis and results in minimal burning of the central macula, while macular grid for diffuse DME may significantly and permanently damage central photoreceptors. The number and intensity of grid laser burns are important to know, especially for the <20/50 subgroup; many moderate grid burns can have a dramatic effect on visual acuity and contrast sensitivity. Enlargement of the laser scar with time (RPE creep) can have a negative effect on final visual acuity. It would be helpful to at least know the mean number of laser spots delivered to the macula in this group rather than lumping prior laser into “focal and/or grid.” Also, there is no standardization of “grid” treatment, and with 89 sites participating and likely hundreds of retina specialists treating, grid laser treatment variability is likely. During the trial, the entire cohort sustained up to two additional focal/grid laser treatments (36% of the aflibercept eyes, 56% of the bevacizumab, and 46% of ranibizumab eyes).

We are not informed how many of these eyes, and specifically how many eyes in the <20/50 subgroup, received prior grid laser before entering the trial. Some eyes could have received three or more grid laser treatments and might be compared to eyes in another treatment arm, which had three focal laser treatments or no prior laser treatment. The unintended selection process of eyes with prior grid laser could bias the conclusion that one drug gives more ETDRS letters than another.

Finally, we are not given details about how the fellow eye was managed. In about 56% of cases, the fellow eye in each treatment arm received the same drug as the study eye, but the frequency of injections is not mentioned. Importantly, we are not informed what percentage of fellow eyes or frequency of injections were given in the <20/50 subgroup. If more frequent injections were given to the fellow eye in one arm than another, a crossover effect could influence the outcome.

Before the retinal community accepts the conclusion that aflibercept results in better visual acuity in the subgroup of eyes, which started with 20/50 or worse acuity, we must be shown data that the visual potential for each subgroup treatment arm was similar. For this subset, we would like to know if the process of randomization created equivalent groups with respect to the duration of DME prior to treatment, ETDRS retinopathy severity, A1c levels, the number of pre- and intrastudy grid lasers, prestudy injections, and the frequency and number of fellow eye injections.

I’m not convinced that one drug works better than another and hope more details of the <20/50 group will be made available in the next DRCRnet two-year follow-up publication, and I suggest the report be published in a journal most ophthalmologists routinely read.

Paul E. Tornambe, MD, FACS San Diego, CA

REFERENCES