The Great Galveston Hurricane of 1900 was the deadliest natural disaster in U.S. history, killing up to 12,000 people. The morning the storm made Texas landfall, the sky was calm with scattered clouds, with no suggestion of the impending storm surge and destruction from which the island would never fully recover.

Since then, weather forecasting has improved tremendously with refinement of complex algorithms that incorporate millions of data points. These systems allow remarkably accurate hurricane tracking three to five days in advance, enabling local preparations and evacuations, saving tens of thousands of lives annually.

Severe nonproliferative diabetic retinopathy without diabetic macular edema provides a similar scenario. This population of about 3 million Americans are typically asymptomatic with excellent visual acuity. While we recognize the risk of these eyes developing proliferative DR or DME, the most common cause of blindness among American working-age adults, severe NPDR remains a major clinical hurdle to initiating intravitreal anti-VEGF injections. The skies are a bit cloudy, but there’s no hurricane in sight, yet.

PANORAMA,1 the first prospective trial involving high-risk NPDR eyes without central-involved DME since the Early Treatment Diabetic Retinopathy Study of the 1980s,2 provides valuable data to inform decision making in the anti-VEGF era. In PANORAMA, about 40 percent of untreated eyes developed PDR or CI-DME at one year, a rate reduced by approximately 75 percent with fixed aflibercept dosing.

But, can we predict which eyes will progress? In the ETDRS, 12, 26 and 52 percent of eyes with DR Severity Scale (DRSS) levels of 43, 47 and 53, respectively, developed PDR within one year.2 While we categorize level 53 eyes as highest risk, even one of eight individuals with level 43 are anticipated to develop PDR within one year. Description of the specific NPDR DRSS score may be irrelevant if we can predict disease progression at an individual level.

As emphasized in previous editorials, I am committed to the promise of artificial intelligence and deep learning. In these pages, two articles describe attempts to harness the power of retinal images and algorithms to predict DR progression.

In the 21st century, it’s ridiculous to consider initiating a coastal hurricane evacuation only once the storm makes landfall. It is equally absurd to consider evacuating the entire Eastern seaboard every time a hurricane forms in the Atlantic. Accurate predictability is key. Similarly, it’s logical to treat DR before the advanced stages of the disease are manifest and VA loss has occurred. Improved prognostic granularity is coming.  ​RS


1. Wykoff CC. Intravitreal aflibercept for moderately severe to severe nonproliferative diabetic retinopathy (NPDR): 1-year of the phase 3 PANORAMA study. On behalf of the PANORAMA investigators. Paper presented at Angiogenesis, Exudation, and Degeneration 2019; Miami, FL: February 9, 2019.

2. Fundus photographic risk factors for progression of diabetic retinopathy. ETDRS report number 12. Early Treatment Diabetic Retinopathy Study Research Group. Ophthalmology. 1991;98(5 Suppl):823-833.