A laboratory-based algorithm to predict future kidney function decline in older adults with reduced estimated...

Area of Science:

• Nephrology
• Geriatric Medicine
• Clinical Chemistry

Background:

• Reduced estimated glomerular filtration rate (eGFR) in adults aged 60-75 is common, necessitating differentiation between normal aging and significant kidney disease.
• Developing predictive models using routine laboratory data is crucial for effective patient triage.

Purpose of the Study:

• To develop and validate a predictive model for estimating glomerular filtration rate (eGFR) decline in older adults.
• To utilize routine laboratory data for improved risk stratification of kidney function decline.

Main Methods:

• General linear models were developed to predict eGFR regression slopes using urine protein measurements (urine albumin/creatinine ratio [UACR] or urine protein/creatinine ratio [UPCR]) and other routine lab data.
• Models were trained on a large US laboratory dataset (n=43,523) of individuals aged 60-75 with baseline eGFRs of 30-59 mL/min/1.73m².
• Model performance was validated on a separate dataset (n=11,979) using receiver operator characteristic (ROC) statistics and areas under the curves (AUC).

Main Results:

• Key predictors of eGFR slope included UACR or UPCR, initial eGFR, serum albumin, chloride, glucose, and aspartate aminotransferase (AST) or serum calcium.
• The predictive models demonstrated strong performance, with AUCs ranging from 0.716 to 0.900 for UACR models and 0.751 to 0.868 for UPCR models.
• Validation results closely mirrored training data performance, confirming model robustness.

Conclusions:

• A laboratory-based predictive model can effectively distinguish older adults with eGFR 30-59 mL/min/1.73m² who are at risk for further kidney function decline.
• This model offers a valuable tool for clinicians to better manage and triage elderly patients with reduced kidney function.