Boehringer Ingelheim and Carelon Research have conducted a validation study of the Klinrisk model, a machine-learning tool developed to predict the risk of chronic kidney disease (CKD) progression at all stages of the disease. Using data from a diverse population of over 4 million U.S. adults, the model was more than 80% accurate in predicting CKD progression over a five-year period. These findings were presented in an oral session at the American Society of Nephrology (ASN) Kidney Week conference in Philadelphia, being held from November 1-5, 2023.
“Effective disease-modifying therapies rarely reach the people with chronic kidney disease who are most likely to need them, due to limited recognition of early-stage disease,” said Mohamed Eid, M.D., M.P.H., M.H.A., vice president, Clinical Development & Medical Affairs, Cardio-Renal-Metabolism & Respiratory Medicine, Boehringer Ingelheim Pharmaceuticals, Inc. “This model may have the potential to help healthcare professionals better identify patients at risk of CKD progression using simple lab results. Physicians need novel tools to evaluate risk of CKD progression, which could assist with earlier diagnosis and treatment.”
The Klinrisk model was developed to predict the risk of CKD progression, defined as either a 40% decline in estimated glomerular filtration rate (eGFR) or kidney failure (dialysis, kidney transplant or eGFR below 10 mL/min/1.73 m2). The model uses machine learning to evaluate risk using age, sex and routinely collected laboratory data, including complete blood cell counts, chemistry and metabolic panels and urinalysis (when available). At least one serum creatinine measurement was required for the validation, and patients were excluded if they did not have at least one year of follow-up time with continuous insurance enrollment after laboratory data was initially collected.
The validation study used data provided by Carelon Research from a diverse population of 4.6 million U.S. adults enrolled in commercial, Medicare and Medicaid insurance plans to assess that the Klinrisk model can predict risk of CKD progression. The model correctly predicted CKD progression in 80-83% of individuals over two years and in 78-83% of individuals over five years, depending on the insurance provider. When urinalysis data was available, the model correctly predicted CKD progression in 81-87% of individuals over two years and in 80-87% of individuals over five years.
“The model is a useful tool to help identify people with high-risk CKD early, before kidney function is lost,” said Navdeep Tangri, M.D., scientific founder of Klinrisk. “We look forward to advancing this model as we aim to bring it to U.S. physicians.”
“As this model requires only demographic information and routine laboratory data, it may have the potential for broad application in a clinical setting to help identify individuals at risk of CKD progression,” said Mark Cziraky, president of Carelon Research. “Earlier identification of CKD risk could help to inform and improve care decisions.”
