Diabetic retinopathy is the leading cause of vision loss in working-age adults and develops as chronically elevated blood glucose damages the small vessels of the retina. Clinicians conventionally describe it as a progressive disease, moving from very mild and mild nonproliferative stages through more severe forms and, ultimately, proliferative retinopathy in which abnormal new blood vessels grow. What has been far less documented is regression: the phenomenon in which retinal findings such as microaneurysms and hemorrhages diminish over time and the eye appears to step back to a less severe disease state. Because regression is rarely reported, its true prevalence in early disease has remained uncertain, leaving a gap in how eye care management strategies are designed and communicated.

To address that gap, the research team turned to a uniquely valuable data source: the Indian Health Service teleophthalmology program for diabetic retinopathy evaluation. This program brings retinal imaging to patients in American Indian and Alaska Native communities, many of which face significant barriers to specialty eye care. Trained readers grade the images using standardized scales, creating a longitudinal record of retinal status that is unusually consistent across time. The investigators identified patients who had a baseline retinal examination between 2015 and 2017 and a follow-up examination within three years, then tracked how their retinopathy status changed.

The cohort was substantial. A total of 2,348 patients had very mild nonproliferative diabetic retinopathy at baseline, and 705 had mild disease. Across the full sample, the mean age was 56 years with a standard deviation of 12.8 years, 55 percent were female, the average duration of diabetes was 13.0 years, and the mean hemoglobin A1c, a measure of average blood glucose control over roughly three months, was 8.8 percent. These characteristics describe a population in mid-adulthood with long-standing diabetes and glycemic control well above recommended targets, a profile typical of many patients at risk for retinopathy complications.

The central result was striking. Among patients starting with very mild nonproliferative diabetic retinopathy, 47 percent, with a confidence interval of 45 to 50 percent, showed regression on follow-up imaging, while only 27 percent progressed. Among those starting with mild disease, 36 percent regressed, with a confidence interval of 33 to 40 percent, compared with 34 percent who progressed. In other words, for the very mildest form of retinopathy, improvement was nearly twice as likely as worsening over the follow-up window. Even in the mild stage, regression matched or exceeded progression. The findings directly challenge the assumption that early diabetic retinopathy inevitably advances.

Beyond simply counting transitions, the researchers used multivariable analyses to untangle which factors were associated with the odds of regression, accounting simultaneously for other patient characteristics. Three variables emerged as significant. Lower hemoglobin A1c levels were associated with greater odds of regression in patients with either very mild or mild baseline disease, reinforcing the central role of glycemic control. Insulin use, by contrast, was associated with lower odds of regression, likely reflecting that patients requiring insulin tend to have more severe or longer-standing diabetes. Shorter diabetes duration was associated with greater odds of regression, but only among patients who began with very mild disease, suggesting that the window for spontaneous or treatment-supported reversal narrows as diabetes persists.

These associations carry mechanistic weight. Sustained hyperglycemia damages retinal capillaries through well-characterized pathways, including the accumulation of advanced glycation end products, increased vascular permeability, and chronic low-grade inflammation. When average glucose levels fall, the metabolic drivers of vascular injury weaken, and the retina’s supporting cells and vessels can partially recover. Clinical trials have long shown that intensive glycemic control slows retinopathy progression, but the new study extends that principle in a different direction: better control is not merely protective against worsening, it is statistically linked to measurable improvement in early disease. Behavioral modifications such as increased physical activity and weight loss, which lower hemoglobin A1c and thereby reduce the risk of progression, are important to emphasize in diabetic eye care, the authors conclude.

The population context matters as much as the biology. American Indian and Alaska Native communities experience among the highest rates of type 2 diabetes in the United States, and geographic isolation often delays access to ophthalmologists. The Indian Health Service teleophthalmology program was built to close that gap by bringing fundus photography and standardized grading directly into primary care settings. Because the same imaging and grading protocols are applied across sites and years, the program produces longitudinal data of a consistency that is difficult to achieve in fragmented private health systems. That infrastructure is what made it possible to observe thousands of transitions in retinal status over a defined three-year window, and it demonstrates how teleophthalmology can serve simultaneously as a care delivery mechanism and a surveillance engine for population eye health.

For clinicians, the practical implications are immediate. Patients told that they have very mild or mild diabetic retinopathy often assume that vision-threatening disease is inevitable, a message that can foster fatalism and disengagement from self-care. The new evidence supports a different framing: early retinopathy is a warning sign, but it is also a stage at which improvement is common, particularly when glucose control improves, diabetes is caught early, and the disease has not yet required insulin therapy. That message could strengthen motivation for the very behaviors, diet, exercise, adherence to glucose-lowering treatment, and regular monitoring, that the data link to better retinal outcomes. It also argues for continued regular retinal screening even after regression is documented, since the underlying metabolic disease persists and the risk of future progression remains.

The study also leaves open questions that future research must address. As a retrospective cohort analysis, it can identify associations but cannot prove that lowering hemoglobin A1c causes regression; randomized or prospective designs would be needed to establish causation. The three-year follow-up window captures short-term dynamics but not the longer arc of disease. And the findings, drawn from a specific population served by a single national program, may not generalize identically to all populations with diabetes. Nevertheless, the scale and consistency of the results make a compelling case that the earliest stages of diabetic retinopathy deserve to be described to patients not as a one-way slide toward vision loss but as a reversible crossroads, one at which the choices a patient makes about glucose control, activity, and weight can measurably tilt the odds toward a healthier retina.

Subject of Research: Rates of nonproliferative diabetic retinopathy (NPDR) regression among patients with very mild or mild NPDR in an Indian Health Service teleophthalmology program, 2015 – 2020

Article Title: Rates of nonproliferative diabetic retinopathy (NPDR) regression among patients with very mild or mild NPDR in an Indian Health Service teleophthalmology program, 2015 – 2020

Article References: Bursell, S.-E., Fonda, S. J., Borik, K., Stainback, J., & Shahon, D. (2026). Rates of nonproliferative diabetic retinopathy (NPDR) regression among patients with very mild or mild NPDR in an Indian Health Service teleophthalmology program, 2015 – 2020. PLOS One, 21(10), e0361052. https://doi.org/10.1371/journal.pone.0361052

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0361052

Keywords: diabetic retinopathy, regression, Indian Health Service, teleophthalmology, nonproliferative diabetic retinopathy, HbA1c, glycemic control, retinal screening, diabetes complications, eye disease, Rates, nonproliferative