A retinal condition called neovascular age-related macular degeneration, often shortened to NVAMD, is one of the leading causes of severe vision loss in older adults. It happens when abnormal blood vessels grow under the retina and leak fluid, damaging the central part of vision. Because the condition is progressive and difficult to treat once established, researchers are keen to understand whether common medications change the risk of developing it.
A research team published a large retrospective study in the journal Ophthalmology asking a specific question: does treatment with a GLP-1 receptor agonist peptide change the likelihood of a person with type 2 diabetes developing NVAMD? The study drew on data from nearly 228,000 people across 12 health databases in the Observational Health Data Sciences and Informatics, or OHDSI, network, covering a period from late 2017 through the end of 2024.
The short answer from the data is that no meaningful difference in NVAMD risk was detected between the GLP-1 peptide group and any of the comparison groups. But the study design, the scale of the data, and the context around it are worth understanding in more detail.
Background on GLP-1 receptor agonist peptides
GLP-1 receptor agonists are a class of peptide-based compounds that mimic glucagon-like peptide-1, a hormone produced in the gut after eating. Researchers have studied these peptides extensively in the context of blood sugar regulation and body weight, and their use among people with type 2 diabetes has grown substantially over the past decade.
Because GLP-1 receptors are found in tissues throughout the body, not just the pancreas, scientists have been exploring whether these peptides might also influence other organ systems, including the eye. Retinal tissue contains GLP-1 receptors, which raised the question of whether GLP-1 peptide exposure could either protect against or, alternatively, accelerate certain eye conditions.
Some earlier observational data had suggested a possible signal linking one GLP-1 peptide to a higher rate of a different eye condition called nonarteritic anterior ischemic optic neuropathy. That finding, reported before this study, created broader interest in thoroughly examining the relationship between GLP-1 peptides and various forms of retinal and optic disease.
Study design and comparison groups
The researchers used two distinct analytical approaches to reduce the chance of reaching a false conclusion. The first was an active-comparator cohort design, which compared new users of the GLP-1 peptide against new users of other diabetes medications. The comparators included two other GLP-1 receptor agonist peptides, dulaglutide and exenatide, as well as three non-GLP-1 agents: empagliflozin, sitagliptin, and glipizide. Using active comparators rather than untreated patients helps control for the fact that people who take any medication for diabetes may differ in meaningful ways from people who take none.
The second approach was a self-controlled case-series analysis, abbreviated as SCCS. In this design, each person who developed NVAMD served as their own control. The analysis compared the rate of NVAMD diagnosis during periods when that person was exposed to the peptide against periods when they were not. This approach is especially good at removing confounding from stable background characteristics, such as age or pre-existing health conditions, because those factors are the same within a single person across time.
The team also used two definitions of NVAMD. One relied on diagnostic condition codes alone. The other required both condition codes and evidence of a related procedure, such as an injection into the eye, which is the standard treatment for NVAMD. Using two definitions helped the researchers check whether their findings were consistent regardless of how strictly NVAMD was identified in the records.
What the numbers showed
Across both analytical methods and both definitions of NVAMD, the study detected no statistically significant difference in risk between people who used the GLP-1 peptide and those who used any of the comparison medications.
In the cohort analysis, hazard ratios comparing the GLP-1 peptide to dulaglutide, empagliflozin, sitagliptin, and glipizide all had confidence intervals that crossed 1.0, meaning the data were consistent with no effect. The p-values ranged from 0.09 to 0.94, none approaching conventional thresholds for statistical significance.
In the self-controlled case-series analysis, the incidence rate ratio for the stricter NVAMD definition was 1.02 with a 95 percent confidence interval from 0.76 to 1.36 and a p-value of 0.92. For the looser definition, the incidence rate ratio was 0.92 with a confidence interval from 0.67 to 1.26 and a p-value of 0.60. Both estimates sit almost exactly at 1.0, the value that represents no difference at all.
A random-effects meta-analysis pooled results across all 12 databases in the network to produce overall estimates, reinforcing the null finding. The literature suggests this kind of network-wide analysis is more robust than any single database study because it draws on a broader and more diverse population.
Limitations worth knowing
Observational studies, even large and carefully designed ones, have known limitations. The data come from health records that were created for clinical care, not for research, so there is always some risk that diagnoses are miscoded or missed entirely. The researchers tried to address this by using two definitions of NVAMD, but some misclassification likely remains.
The study population was limited to adults with type 2 diabetes, which means the findings do not automatically extend to people who might use GLP-1 peptides for other purposes or who do not have diabetes. Age-related macular degeneration is also a relatively rare event in any single year, which means that even a study of nearly 228,000 people may not have enough statistical power to detect small differences in risk.
The retrospective nature of the study means the researchers could not control for every possible confounding variable. People who are prescribed different medications differ from each other in ways that health records may not fully capture, such as dietary habits, sun exposure history, or family history of eye disease.
Relevance to peptide research more broadly
The finding adds a piece to a larger puzzle about how GLP-1 receptor agonist peptides interact with tissues outside the metabolic system. Early data points at a variety of organ-level effects, and the eye has become one active area of inquiry because retinal tissue expresses GLP-1 receptors and because any effect on vision would be clinically significant.
The self-controlled case-series design is particularly informative here because it bypasses many of the biases that plague typical observational comparisons. When a person's own non-exposure periods serve as the control, stable confounders drop out of the equation entirely. The fact that this design also produced a null result strengthens confidence that the absence of a detected association is real rather than an artifact of imperfect comparisons between different groups of people.
Researchers studying other peptides that also act on metabolic pathways or that influence vascular biology may find the methodology of this OHDSI network approach useful as a template. The use of 12 databases simultaneously, combined with meta-analysis, represents a relatively rigorous standard for observational pharmacoepidemiology.
Summary of the evidence
A retrospective study across 12 health databases and nearly 228,000 patients found no evidence that a GLP-1 receptor agonist peptide increases or decreases the risk of neovascular age-related macular degeneration among adults with type 2 diabetes. The null finding held across two study designs, two definitions of the eye condition, and multiple comparison medications. The authors concluded that there was no detectable difference in NVAMD risk associated with this peptide exposure in this population.
For anyone following peptide research, this study is a reminder that documented effects of a compound in one tissue system do not automatically predict effects in another. Careful, large-scale observational work like this is one way researchers work to test and refine those assumptions over time.
