Teleretinal Screening System Shows Promise in Large Validation Study

By Anne Harding

December 23, 2015

NEW YORK (Reuters Health) - An automated teleretinal screening program can identify almost all patients with diabetic retinopathy, results of a study suggest.

The false-negative rate for the Intelligent Retinal Imaging System (IRIS) was just 2%, Dr. Christina Y. Weng of Baylor College of Medicine in Houston and her colleagues found. "It very rarely missed patients that had true disease," Dr. Weng told Reuters Health in a telephone interview. "That's very, very important for a screening mechanism."

More than 60% of people with type 2 diabetes and nearly everyone with type 1 diabetes will develop diabetic retinopathy within 20 years of their diabetes diagnosis, Dr. Weng and her team note in their report, published online December 17 in JAMA Ophthalmology. While the American Academy of Ophthalmology recommends annual diabetic retinopathy surveillance examinations for people with diabetes, they add, only 50% to 65% of patients actually undergo regular screening.

In the new study, Dr. Weng and colleagues tested IRIS at screening sites across the Harris Health System in Harris County, Texas. Patients who had diabetes and had not undergone retinal screening in the past year were flagged in their medical record. When these patients came to see their primary care doctor, the physician instructed them to have their retinas examined with the IRIS system. Patients were able to have the test done on the spot, without having to make another appointment.

A total of 15,015 consecutive patients underwent nonmydriatic fundus photography with the IRIS system from June 2013 to April 2014. Each patient was classified for referral or observation. An optometrist or ophthalmologist also viewed the image and manually assigned a level of retinopathy using Early Treatment Diabetic Retinopathy Study classification criteria. Based on the manual readings, 15.8% of people screened had sight-threatening diabetic eye disease (STDED).

While 18,025 patients had the IRIS screening, just 83.3% had images that were adequate for analysis. Compared to manual screening, the IRIS algorithm had a sensitivity of 66.4% for identifying STDED and a false-negative rate of 2%. Specificity was 72.8%. Positive predictive value was 10.8% and negative predictive value was 97.8%.

To date, Dr. Weng told Reuters Health, nearly 50,000 patients have undergone the screening test. Based on back-of-the-envelope calculations, she added, this means about 2,500 patients were potentially saved from blindness. The program also improved diabetic retinopathy screening compliance, she added, from 62% to 80% in its first year.

Wait times for an ophthalmology appointment within the Harris Health System have also dropped from three and a half months to one month. "We're saving the clinical resources for those who really need it," Dr. Weng said.

In an editorial accompanying the study, Dr. Jennifer Sun and colleagues from Harvard Medical School and Joslin Diabetes Center in Boston note that the IRIS metrics were lower than thresholds set by the British Diabetic Retinopathy Working Group for detection of diabetic retinopathy of 80% sensitivity and 95% specificity.

"The large number of ungradable images decreases the generalizability of these results and potentially compromises the system's sensitivity for detecting disease because ungradable images may be due to intraocular pathology, necessitating the need for a comprehensive eye examination," the editorialists add.

"Perhaps the most important aspect of this report is that the authors methodically evaluated the algorithm in a large study of eyes with a full range of . . . severity, clearly recognizing the need to validate any telemedicine program for" diabetic retinopathy, Dr. Sun and colleagues state.

"While the validation of individual automated programs for assessment of diabetic retinal disease is critical, what will ultimately be needed is the ability to compare multiple algorithms across different imaging platforms and diverse cohorts of patients with diabetes mellitus," they write.

The authors reported no funding or disclosures.


JAMA Ophthalmol 2015.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: