Specialty Pharmacy Continuum | August 18, 2023
An artificial intelligence tool combined with a noninvasive retinal eye examination can accurately predict the risk for chronic kidney disease (CKD) before symptoms appear, according to a retrospective study published in npj Digital Medicine (2023;6[1]:114).
An estimated 14% of adults in the United States have CKD, according to the CDC, but as many as nine in 10 people with CKD, and three in 10 of those with severe CKD, do not know they have it. Early-stage kidney disease typically has no symptoms, and many people are not aware of their condition until the disease has reached advanced stages, including kidney failure.
The current method of screening for CKD involves measuring the estimated glomerular filtration rate (eGFR, calculated from serum creatinine) or examining urine for proteinuria, but these biomarkers have significant limitations in the detection of early kidney disease, especially in patients who have preserved kidney function and do not demonstrate abnormalities in blood or urine tests (World J Methodol 2017;7[3]:73-92).
To address this issue, Mediwhale Inc. developed Reti-CKD, a predictive risk score for CKD. Reti-CKD uses retinal photographs to obtain information from the retinal blood vessels, which have shared characteristics with the vessels of the kidney. “Reti-CKD is the first comprehensive AI diagnostic solution for CKD using retinal imaging,” said Mediwhale’s chief medical officer, Tyler Hyungtaek Rim, MD. The process takes about 30 seconds and involves a retinal scan with color fundus imaging, he said, describing it as “very easy to use.”
To train the AI, Reti-CKD’s developers used data from health checkups involving more than 79,000 Koreans, as well as data from another 35,000 individuals from the UK Biobank and Korean Diabetic Cohort for validation of the model. In the UK Biobank, 720 of 30,477 (2.4%) participants had CKD events during a follow-up period of 10.8 years; in the Korean Diabetic Cohort, 206 of 5,014 (4.1%) had CKD events during the 6.1-year follow-up period. The investigators found that the Reti-CKD’s risk score was superior to eGFR-based methods, the current standard for CKD screening, in predicting CKD incidence. (Reti-CKD’s concordance—the probability that its prediction of the outcome is better than chance—was significantly better than that found with eGFR.) Reti-CKD also stratified patients into four classifications of CKD risk, ranging from low to high, and accurately predicted the risk regardless of the presence of underlying diseases such as hypertension or diabetes.
In subgroup analyses, the algorithm’s impact on predicting CKD was relatively greater among patients without diabetes than diabetic patients. “In diabetes patients, other risk factors than retinal abnormalities may serve as powerful surrogates for kidney function status, such as blood glucose level or diabetes duration,” said Dr. Rim, who published the findings alongside authors from Mediwhale and Yonsei University College of Medicine, in Seoul, South Korea. “Nonetheless, although the predictive impact of Reti-CKD may slightly differ regarding the presence of diabetes, it should be noted that, in both patient populations, the effectiveness of Reti-CKD as a predictive marker was significant.”
Reti-CKD has been submitted for approval to the Korean Ministry of Food and Drug Safety, and the company expects it to be approved there in 2024. Mediwhale also plans to apply for U.S. approval sometime next year. “Early kidney disease is a silent killer, and prescreening tools are very important,” Dr. Rim said.