A comprehensive 20-year retrospective cohort study evaluated the changing landscape of diabetic retinal disease among U.S. patients with diabetes mellitus and revealed that disease prevalence and incidence has doubled since the early 2000s. However, incidence rates of vision-threatening diabetic retinopathy, diabetic macular edema, and proliferative diabetic retinopathy have improved dramatically.
“The Wisconsin Epidemiologic Study of Diabetic Retinopathy was the sentinel study to assess diabetic retinal disease (DRD) rates and is still cited as a primary source for prevalence data 45 years later,” wrote lead study author Brian L. VanderBeek, MD, MPH, of the Department of Ophthalmology at the Perelman School of Medicine at the University of Pennsylvania, and colleagues in their recent Ophthalmology article. The investigators explained that population-based studies of DRD rates in the United States are more than 15 years old and have fragmented analyses that focus only on specific types of DRD in certain age groups. The current data also don't reflect the advances in medication and technology.
To address the need for updated prevalence and incidence data, the investigators assessed more than 6 million patients with diabetes mellitus (DM) between 2000 and 2022 using a large national claims database of commercially insured and Medicare Advantage patients to track prevalence and incidence rates of DRD and its vision-threatening components: diabetic macular edema (DME), proliferative diabetic retinopathy (PDR), and vision-threatening diabetic retinopathy (VTDR).
DRD diagnosis was based on validated ICD codes, and over 1 million cases were identified. The patient pool was broken down further into 417,155 patients with VTDR, 286,215 with DME, and 225,137 with PDR. The investigators used logistic and Poisson regression with adjustments for demographics, geography, insurance type, and income.
They found that DRD prevalence among patients with DM initially decreased from 13.6% in 2001 to 10.9% in 2007, but increased annually to 20.8% by 2021. Population-wide DRD prevalence (including nondiabetic patients) rose from 0.6% in 2001 to 4.1% in 2021. DRD incidence reached its lowest point in 2014 (16.9 per 1,000 patient-years), then peaked in 2022 at 32.1 per 1,000 patient-years.
VTDR prevalence rose from 5.2% in 2007 to 7.5% in 2016, then decreased to 6.9% by 2021. Incidence decreased from a high of 13.6 per 1,000 in 2002 to 6.1 per 1,000 in 2022.
DME prevalence rose from 2.8% in 2001 to 5.4% in 2016, then declined to 4.9% by 2021. Incidence peaked at 8.6 per 1,000 in 2009 and declined to 5 per 1,000 in 2022.
PDR prevalence fluctuated between 3.2% and 4%, ending at 3.5% in 2021. Incidence steadily declined from 8.3 per 1,000 in 2002 to 2.6 per 1,000 in 2022.
The investigators attributed the doubling of DRD prevalence among patients with DM and 6.8-fold increase in the general population to growing DM prevalence, expanded health insurance access since the Affordable Care Act, and increased screening and telemedicine adoption.
Despite more patients being diagnosed with DRD, incidence of sight-threatening complications has decreased, which could suggest improved diabetes management and earlier detection. Widespread use of anti-VEGF therapy and advances in glucose-lowering medications may have contributed to the declining incidence of DME and PDR. “It is possible if the database were extended, additional years of data could capture an impending wave of VTDR,” the study authors wrote. “However, given the increased systemic DM treatment and monitoring options within a better-insured population (expanding access to those improved treatments), it is unclear if the 5- to 9-year estimation is still relevant to today’s patients,” they added.
The investigators acknowledged limitations of their study, including possible underrepresentation of racial/ethnic minorities, uninsured patients, and Veterans Health Administration patients as a result of the claims-based data set. They also noted that many patients don't receive annual eye exams, which could lead to underestimation of true disease burden. Disease duration was also not able to be assessed, so analyses of progression patters might have been limited. Finally, diagnostic criteria and coding practices have evolved over time (eg, the switch from ICD-9 to ICD-10 and increased OCT use). While coding was validated, the investigators noted that they couldn't be sure that their rates were not affected by these changes. Any misclassification bias as a result of changes in diagnostic criteria “suggests that our estimates are likely to be an underestimate of the overall true rate of disease,” the study authors concluded.
Disclosures can be found in the published research.