On this episode of Retina Mentor Moments, John W. Kitchens, MD, is joined by Mathew MacCumber, MD, PhD, to discuss studies on retinal diseases and how big data and artificial intelligence (AI) can be implemented in clinical research to improve retina care.
Dr. MacCumber is a vitreoretinal specialist at Illinois Retina Associates in Chicago, Illinois, as well as a professor and Director of Research for the Department of Ophthalmology at Rush University Medical Center.
Early research in endothelin-based therapies
Dr. MacCumber recounted that his PhD was focused on a peptide called endothelin, which is a potent vasoconstrictor that plays an important role in normal physiology by maintaining vascular tone in various organ systems.1
Endothelin is a key mediator of vascular pathology, stimulating ischemia, inflammation, retinal ganglion cell apoptosis, neovascularization, vascular leak, and fibrosis.2 In his PhD, Dr. MacCumber identified high levels of endothelin expression in the retina, uveal tract, and choroid, and he posited that this is the reason that the pars plana does not bleed significantly when cut into.3-4
Now, 35 years later, clinical research has progressed to identifying ocular diseases that are directly impacted by endothelin, as it has been detected in the aqueous humor of patients with different types of open-angle glaucoma and a range of vascular diseases, such as diabetic retinopathy and branch or central retinal vein occlusion.2,5
Further, findings from a phase 2a clinical trial on PER-001 (Perfuse Therapeutics) were recently presented at Clinical Trials at the Summit 2025. PER-001 is a novel, long-acting endothelin antagonist administered as a bio-erodible intravitreal implant for the treatment of retinal ischemia in patients with moderate to severe non-proliferative diabetic retinopathy.2
The results of the phase 2a trial for PER-001 demonstrated that:2
- The implant was safe and well-tolerated
- There were no serious drug-related ocular or systemic adverse events
- Best-corrected visual acuity (BCVA) remained stable
- There were no changes in intraocular pressure (IOP)
- Patients had an improvement in their low-luminance contrast sensitivity compared to the control group
To learn more about Dr. MacCumber’s research and his experiences with performing clinical research at a large private practice, watch the full interview!
Protocol AM: When to treat or monitor an epiretinal membrane
Dr. MacCumber is a co-chair for Protocol AM, an ongoing DRCR Retina Network study investigating immediate versus deferred surgery for symptomatic epiretinal membranes (ERMs).6
He explained that ERM patients who present with significant macular changes on optical coherence tomography (OCT), but relatively good vision, are typically advised to defer surgery until vision declines to 20/40 or worse. Nevertheless, it remains unclear whether delaying surgery can result in worse visual acuity outcomes compared to performing surgery earlier.6
Protocol AM’s primary outcome is visual acuity at 3 years, with secondary outcomes focusing on other measures of visual function, such as metamorphopsia and reading speed.6 One of his goals for this study is to use AI-based tools to aid in the analysis and identification of anatomic factors on OCT associated with outcomes after surgery to predict which eyes with ERM will progress.6
This data will subsequently be used to provide guidance on when surgeons should operate early to prevent loss of anatomy and vision and when ERM patients can simply be monitored.
Using big data and AI in ophthalmology research
Aside from Protocol AM, Dr. MacCumber explained that he is interested in using AI tools to determine how well geographic atrophy (GA) progression is controlled in the real world with intravitreal complement inhibitors to preserve vision in these patients and whether individuals with previous wet AMD and concurrent GA benefit from using these drugs.
He added that due to the controlled environments of clinical trials, the results from these studies are often not indicative of real-world results, which is where using real-world databases—such as TriNetX, Vestrum Health, RCA Research, and the IRIS Registry—is valuable in understanding the impact of interventions outside of clinical trials.
Dr. MacCumber explained that he’s had the opportunity to use the Vestrum Health database for several big data studies, with two recent analyses outlined below.
How long should surgeons follow up with PVD patients?
One such study evaluated the incidence of complications after posterior vitreous detachment (PVD) through a 6-month follow-up period and identified patient-specific factors associated with a greater incidence of complications.7
In this multicenter, retrospective observational study, eyes with acute PVDs between 2015 and 2019 were identified through the Vestrum Health database. Complications, including vitreous hemorrhage, retinal break, and retinal detachment, were evaluated after acute PVD at presentation and throughout the 6-month follow-up period.
In total, 9,635 eyes were included in the analysis, with a rate of any complication of 25.0%, and the following rates for specific sequelae:7
- Isolated vitreous hemorrhage: 13.1%
- Retinal breaks without detachment: 16.0%
- Retinal detachment: 4.2%
While the majority of each complication was observed at presentation, 8.0% of isolated vitreous hemorrhages, 19.2% of retinal breaks without detachment, and 25.8% of retinal detachments were first reported later within the 6-month follow-up period.7
Which PVD patients are at higher risk of complications?
Additionally, men experienced a significantly higher rate of any complication compared to women (30.0% vs. 21.7%) as well as retinal breaks and detachments at both presentation and within the 6-month follow-up period.7
Moreover, pseudophakic patients experienced significantly higher rates of delayed retinal detachment than phakic eyes (odds ratio [OR]: 1.85). Notably, 44.2% of eyes with lattice/peripheral retinal degeneration experienced some complication throughout the clinical course.7
The study confirmed what many retina specialists intuitively know, which is that the presence of a retinal break or detachment in the fellow eye is associated with a significantly increased rate of complication at any time point. Among eyes with vitreous hemorrhage at presentation, 42.0% had a concurrent or delayed retinal break, and 10.5% had concurrent or delayed retinal detachments.7
As such, the study demonstrated that a clinically significant proportion of PVD-related complications were detected late, warranting extended follow-up (up to 6 months), especially in higher-risk groups, such as men, pseudophakic eyes, eyes with lattice/peripheral retinal degeneration, and eyes with a history of retinal breaks or detachment in the fellow eye.7
What is the prevalence of PVR in the fellow eye after bilateral RRD repair?
Another study that utilized the Vestrum Health database investigated the incidence and risk factors of proliferative vitreoretinopathy (PVR) in patients who develop rhegmatogenous retinal detachment (RRD) in their fellow eye after having a prior RRD complicated by PVR.8 In this multicenter, retrospective observational study, the research team identified eyes with retinal detachment and PVR between 2015 and 2023 via the Vestrum Health database.
Risk factors for PVR development were evaluated, such as documented PVR in the fellow eye, gender, age, lens status, and presenting and final visual acuity, with a main outcome measure of the odds ratio for PVR development during the 6-month post-operative period.
Of the 57,264 patients included in the analysis, 11% had PVR in ≥1 eye and, of the 50,989 patients who did not develop PVR after the initial RRD, 4,834 developed RRD in the fellow eye. Further, 166 of these patients developed PVR in their second eye for a PVR rate of 3% in the fellow eye.8
Of the 6,275 patients who developed PVR after primary RRD repair, 406 went on to develop RRD in their fellow eye, and 42 of these patients developed PVR in their second eye for a PVR rate of 10%.8
The odds ratio of PVR development generally decreased with age, and pseudophakic patients had a higher odds ratio for PVR development (OR: 1.48). Initial patients with VA 20/40 to 20/80 had an OR of 2.15, while patients with VA worse than 20/200 had an OR of 2.89 for PVR development.8
Consequently, patients with a history of RRD with PVR in one eye had approximately a 3.5 times higher rate of PVR in their second eye after RRD compared with patients without a history of PVR. This finding potentially impacts surgical decisions and the use of prophylactic anti-PVR therapy if the patient’s second eye has RRD.8
To learn more about how to establish a representative patient population for big data studies, watch the interview!
Conclusion
Studies on diabetic retinopathy using an endothelin antagonist and on ERM management may impact the way we treat these conditions in the future. Further, big data platforms and AI systems have the power to answer complex clinical questions, but their impact depends on clinicians asking the right questions and performing careful analyses.
As these tools evolve, they can provide deeper insights that guide meaningful advances in patient care and answer long-standing clinical questions.
Drs. MacCumber and Kitchens agreed that for new retina specialists, the field is full of promise, with emerging treatments for retinal diseases—particularly IRDs—and technologies set to transform ophthalmology for decades to come.
