A quiet blood draw may soon warn you that diabetes is nibbling at your eyesight years before you notice a single blurry road sign.
Story Snapshot
- A new artificial intelligence model, Pro-DRN, uses 71 blood proteins to flag early diabetic retinal nerve damage before symptoms appear.[1][3][7]
- Retinal neurodegeneration often starts years before classic “leaky vessel” diabetic retinopathy shows up on an eye exam.[4][5][6]
- The model reportedly beats the best existing risk tool by about 26%, but it is still research, not a clinic-ready lab test.[1][3][7]
- Without large, real-world trials proving it prevents vision loss, this technology sits on the knife edge between breakthrough and overhyped promise.[1][3][5]
A blood test that listens for nerve damage before your eyes complain
Researchers in China recently reported an artificial intelligence assisted blood test, dubbed Pro-DRN, that tries to do something your annual eye photo cannot: hear the whisper of nerve damage before diabetes shouts.[1][3][7] They pulled a detailed protein “snapshot” from the blood of people with type 2 diabetes and looked for patterns tied to subtle retinal nerve injury, long before the classic bleeding and leaking of advanced diabetic eye disease.[1][3]
Instead of one magic marker, Pro-DRN relies on 71 different plasma proteins—molecules involved in inflammation, cell maintenance, and other stress responses that shift as diabetes grinds away at delicate nerve tissue in the back of the eye.[1][3][7] The team fed these protein levels into a machine learning model and trained it to distinguish people with early diabetic retinal neurodegeneration from those whose retinas still looked healthier on imaging.[1][3]
Why doctors now see nerve loss as diabetes’s first strike on the eye
For decades, diabetic retinopathy was framed as a blood vessel disease first, with nerves and vision failing later as collateral damage. Longitudinal work has flipped that script. A careful follow-up study in people with diabetes but little or no visible retinopathy showed that neuroretinal degeneration clearly preceded the small blood vessel changes and progressed with diabetes duration.[4] That degeneration did not simply track with average blood sugar, age, or sex, which undercuts the idea that glycated hemoglobin alone tells the whole story.[4]
Other research reviews have walked through the same pattern: neuronal dysfunction, glial stress, and nerve cell death can appear before obvious microangiopathy in diabetic retinas.[5] Ophthalmology education sources now openly describe neurodegeneration as a major early driver of diabetic retinopathy, and mention that sophisticated electrical tests of retinal function can be abnormal even when standard photographs still look “normal.”[6] In one model, changes in multifocal electroretinography predicted future retinopathy over three years with very high reported sensitivity and specificity, reinforcing that the eye starts misfiring before it visibly falls apart.[6]
How Pro-DRN tries to turn 71 proteins into a crystal ball
The Pro-DRN investigators leaned on this evolving biology: if the retina is injured early, the damage should leave fingerprints in the blood.[1][3][7] Using high-throughput plasma proteomics, they scanned for proteins whose levels consistently tracked with subtle retinal nerve damage, then used artificial intelligence to combine those signals into a single risk score.[1][3] In their report, that composite score outperformed the best existing predictive model by roughly 26%, an eye-catching bump for any risk tool.[1][3][7]
A simple blood draw folded into the lab work people with diabetes already endure could quietly sort who needs aggressive eye surveillance before any vision changes. The authors themselves framed it this way: early retinal nerve damage in diabetes leaves measurable signals in the blood, and coupling those signals with explainable artificial intelligence might move care from reacting to damage toward true early risk stratification.[3][7]
Why this is not yet the “simple blood test” headlines promise
There is a hard line between a clever research model and a lab slip your primary care doctor can order. Pro-DRN currently lives firmly on the research side of that line. The published reporting describes a correlation-based classifier built on 71 proteins, not a single standardized, regulated clinical assay.[1][3][7] The model is rooted in association, not proven mechanism, which means it recognizes patterns tied to retinal damage but does not necessarily reveal what causes what.[1][3]
A blood test combining 71 protein biomarkers can predict diabetic retinal neurodegeneration years early, enabling preventive screening and treatment.https://t.co/K56V4qZgAo#MedAIDigest #PLoSMed
— MedAI Digest (@MedAIDigest) June 3, 2026
Key practical questions also remain unanswered in the available coverage. Critical details—such as the exact sample size for each cohort, the mix of ethnic backgrounds, the presence of kidney disease or other inflammatory conditions, and how well the model kept its accuracy in fully independent populations—are not yet fully visible to the public.[1][3] No one has shown that changing care based on this score actually prevents blindness, which is what matters to patients, families, and insurers.
Where this could fit alongside existing diabetes eye care tools
The broader diabetes eye-care world is already experimenting heavily with artificial intelligence, but from the other side of the eyeball. Several systems now use machine learning to read retinal photographs and autonomously flag diabetic retinopathy, with the United States Food and Drug Administration clearing some image-based tools for real-world screening. Those tools still wait for visible damage; Pro-DRN aims to move upstream, turning blood chemistry into an early warning radar for nerve injury before the camera sees much at all.[1][3][7]
A balanced reading says this: the biological foundation is solid, the technical approach is plausible, and the early performance signals look promising.[4][5][6][1][3] But calling it a ready-made “blood test that predicts blindness before it starts” runs ahead of the evidence. To earn a place in everyday practice—especially in a health system already skeptical of shiny algorithms—Pro-DRN will need large, prospective validation, standardized assays, transparent performance reporting, and proof that acting on its warnings truly saves sight.[1][3][5]
Sources:
[1] Web – A New Blood Test Could Predict This Diabetes Symptom Before It Starts
[3] Web – Risk of diabetic retinopathy and retinal neurodegeneration in …
[4] Web – Retinal neurodegeneration may precede microvascular changes …
[5] Web – Neurodegeneration in diabetic retinopathy: does it really matter?
[6] Web – Neurodegeneration in Diabetic Retinopathy – EyeWiki
[7] Web – Proteins in blood could help doctors predict retinal degeneration in …
