The use of biomarkers for risk stratification is currently largely limited to research settings, and there has been intense scrutiny. My group has been interested in the prognostic significance of hyperreflective foci in AMD and DME. Larger hyperreflective foci have historically been understood in the context of lipid exudation and poor prognosis, with the potential for deposition under the fovea that may lead to vision loss. However, there are smaller hyperreflective foci that often have no shadowing behind them. Some people have suggested that these are not lipids but rather are activated microglia or an indicator that other inflammatory processes are in play. 

We know that both DME and AMD are multifactorial diseases. The presence of these smaller hyperreflective foci could potentially be a signal that we can use to guide early switching or starting with more aggressive therapy, whether that may be combination inhibition of VEGF and Ang2 or the addition of steroids. Now, should we make treatment changes based on finding smaller hyperreflective foci? In my opinion, this is still more of a research topic and not yet actionable, as we need prospective data to make treatment decisions. But it is something to keep in mind for the future. 

I think that there is less controversy surrounding the prognostic significance of subretinal hyperreflective material (SHRM) in patients with nAMD. SHRM is a well-established precursor for the development of fibrosis, which damages long-term visual acuity. There is compelling evidence that you want to be very aggressive with treatment in the setting of SHRM and not relax the frequency of therapy. 

The use of AI in predicting treatment outcomes for patients with nAMD and DME is something to look for in the future. Researchers are taking advantage of existing data sets to test algorithms and to train AI models. Fascinating research is being conducted to try to predict how an optical coherence tomography (OCT) image would appear if various anti-VEGF therapies were applied. This is not ready for clinical use, but AI is going to be a very important tool for us going forward.

After 3 injections of anti-VEGF therapy in a patient with DME, you would expect to see a very nice response on OCT, if a patient was going to respond. The patient who may need other therapies, or additional mechanisms of action, after 3 injections generally still has cystic spaces and may have started off with a very thick, swollen retina at baseline. 

As Dr Sadda alluded to, our holy grail is to be able to look at a patient’s baseline imaging and predict how many treatments will be required and how their disease will respond. Unfortunately, in clinical studies, we have not been able to identify very many predictive baseline factors aside from some crude markers. For instance, patients with nAMD have worse outcomes if they have intraretinal fluid centrally rather than subretinal fluid. In DME, patients who start off with thicker retinas at baseline generally have worse outcomes.

We need more precise markers to accurately predict treatment responses. Researchers around the world are evaluating the use of AI to identify the features that humans cannot see but a computer can and that may predict who will respond. This is very exciting work because we are starting to identify promising markers. For example, hyperreflective foci are seen in both nAMD and DME, although the prognostic significance differs in each disease. Since computers are better at picking up details than humans, I think that AI will assist us in the future.

I think that Drs Kaiser and Sadda covered this topic very well. What I would add is that the lack of an initial response to anti-VEGF treatment alerts us that a patient’s disease is difficult to treat. There is a lot of controversy about defining persistent nonresponsive edema. But if you identify this in a patient after a period of anti-VEGF treatment, you know that the disease is going to be more difficult to treat, so you tend to treat these patients much more aggressively. This means not extending treatment intervals and considering the addition of steroids to the VEGF inhibitor or switching patients to faricimab.

Dr Sadda alluded to the multifactorial nature of these diseases, and the anti-inflammatory component of treatment may be important for some patients. We may think of steroids first, but one of the effects of blocking Ang2 is to also block some of the inflammatory component, the influx of bone marrow–derived cells. That is potentially an added benefit of dual VEGF/Ang2 suppression. The outcomes that we have seen thus far with faricimab suggest that there might be longer-duration effects on DME, and it would be interesting to see whether that holds up and whether it might be due to an anti-inflammatory component. That said, I think that the steroids probably offer a greater anti-inflammatory effect, and, while it is possible that the use of faricimab might reduce the need for steroids in some patients, there are some other patients who have a significant inflammatory component, or multiple factors in addition to VEGF; for those patients, steroids may be helpful.