Early Nonproliferative Diabetic Retinopathy Regressed More Often Than It Progressed
In a large teleophthalmology cohort drawn from the Indian Health Service, diabetic retinopathy did not move in only one direction. Among patients with very mild or mild nonproliferative diabetic retinopathy at baseline, regression was common enough to matter clinically, especially in the very mild group, where reversal outpaced progression.
The study followed patients with type 2 diabetes who had retinal imaging in 2015 and a subsequent exam within 6 to 36 months. Among 5,735 eligible patients, 4,367 had very mild nonproliferative diabetic retinopathy at baseline and 1,368 had mild disease. Follow-up data were available for 2,348 patients in the very mild group and 705 in the mild group. Average age was 56 years, diabetes duration averaged 13 years, and mean HbA1c was 8.8% in the overall follow-up cohort.
The clinical message is direct: early retinopathy can improve, not only worsen.
Regression Was More Common in Very Mild Disease
Among patients who started with very mild nonproliferative disease, 47% experienced regression, with a 95% confidence interval of 45% to 50%. By comparison, 27% progressed. The remaining patients were split between stability and higher-grade progression. In the mild nonproliferative group, 36% regressed, with a 95% confidence interval of 33% to 40%, while 34% progressed.
Those proportions tell an important story about disease biology at the earliest stages. Very mild retinopathy is often driven by microvascular injury that may still be partially reversible if glycemic exposure improves and capillary stress eases. Once lesions accumulate and retinal ischemia deepens, the balance shifts toward persistent damage. The data here suggest that the window for reversal is still open in a meaningful share of patients seen through routine primary care-linked eye screening.
The follow-up interval averaged 22.8 months, which gives the observations practical relevance for chronic disease management rather than a short-lived fluctuation in grading. Even in a program serving a population with substantial diabetes burden, the direction of change was not uniformly adverse.
HbA1c Carried the Strongest Signal for Reversal
Lower HbA1c was associated with greater odds of regression in both the very mild and mild cohorts. That association fits the pathophysiology of diabetic retinal injury, where chronic hyperglycemia damages pericytes, destabilizes capillaries, and promotes microaneurysm formation. If glycemic exposure falls, endothelial stress may ease enough for some lesions to recede or for retinal grading to improve at the next exam.
Insulin use moved in the opposite direction and was associated with lower odds of regression. The study does not imply harm from insulin itself; rather, insulin use often marks more advanced or more difficult-to-control diabetes, a clinical phenotype less likely to show spontaneous retinal improvement over time. Shorter diabetes duration was also linked with greater odds of regression, but only in the very mild cohort. Earlier disease may simply leave more retinal reserve intact, making improvement more plausible.
Age and sex were included in the multivariable models, along with diabetes therapy and duration, but the most consistent signal across strata was glycemic control. The analysis also used a stricter significance threshold of p < .01, which reduces the chance of overcalling weaker associations in a retrospective dataset.
Teleophthalmology Captured Disease Change in Routine Care
This cohort came from the Indian Health Service teleophthalmology program, a validated program in which patients were screened during primary care visits and retinal images were graded by certified and licensed optometrists under ophthalmologic oversight. More than 80% of examinations used nonmydriatic ultra-widefield imaging, with the remainder obtained by nonmydriatic fundus photography. The imaging approach matters because early retinal lesions can be subtle, and broader field coverage improves the chance of identifying microaneurysms, hemorrhages, and other small changes that define nonproliferative disease.
For patients with very mild disease at baseline, the study separated progression into mild nonproliferative retinopathy from progression to any higher grade. That distinction helps clinicians see where the cohort was headed. Some eyes remained near the starting line; others moved into a more established stage of microvascular injury. In mild disease at baseline, the outcome set was even more granular, capturing regression to no retinopathy, regression to very mild disease, stability, or progression.
The retrospective design cannot prove causality, but the pattern is biologically coherent. Retinal microvascular injury is tightly linked to metabolic control, and the likelihood of improvement appears highest before the retinal circulation crosses into more advanced damage.
Bedside Meaning for Diabetes Eye Care
For clinicians, the main consequence is practical. A patient with very mild nonproliferative retinopathy should not be treated as though progression is inevitable. Early-stage disease leaves room for improvement, especially when HbA1c comes down and diabetes management tightens.
The findings also support a more explicit conversation about behavior change. Increased physical activity, weight loss, and other measures that reduce HbA1c may carry retinal benefits alongside broader metabolic gains. In a setting where many patients already had long-standing diabetes and elevated HbA1c, regression still occurred at meaningful rates. That makes glycemic optimization a retinal strategy, not just a laboratory target.
Prior work from this same cohort showed that a substantial fraction of patients with no retinopathy later developed mild disease or worse, and a smaller subset progressed to proliferative retinopathy. Against that background, the current results add an underappreciated counterpoint: early disease can move backward as well as forward.
The unanswered clinical question is how best to identify the subgroup most likely to regress in real-world practice, and whether more aggressive metabolic intervention can reliably shift that probability over time. A retrospective teleophthalmology cohort can map the terrain, but it cannot determine how much of the improvement came from treatment changes, lifestyle change, or natural fluctuation in early retinal findings.