Concordance and Discrepancies Among 5 Creatinine-Based Equations for Assessing Estimated Glomerular Filtration Rate in Older Adults

IMPORTANCE: There is uncertainty as to which estimated glomerular filtration rate (eGFR) equation should be used among older adults. OBJECTIVE: To compare the 5 most commonly used creatinine-based eGFR equations in older adults, quantifying the concordance among the equations, comparing their discriminative capacity in regards to 15-year mortality, and identifying sources of potential discrepancies. DESIGN, SETTING, AND PARTICIPANTS: This cohort study used data from the Swedish National Study on Aging and Care in Kungsholmen (SNAC-K), a longitudinal study of adults aged 60 years or older in Sweden. Participants were recruited between 2001 and 2004 and followed up for mortality until December 2016. Participants missing creatinine values were excluded. Data were originally analyzed March through July 2022, and were rerun in January 2023. EXPOSURES: Five creatinine-based equations were considered: Modification of Diet in Renal Disease (MDRD), 2009 Chronic Kidney Disease Epidemiological Collaboration (CKD-EPI), Revised Lund-Malmö (RLM), Berlin Initiative Study (BIS), and European Kidney Function Consortium (EKFC). MAIN OUTCOMES AND MEASURES: Concordance between equations was quantified using Cohen κ. Discriminative capacity for mortality was quantified using area under the receiver operating characteristic curve (AUC) and the Harrel C statistic. Calf circumference, body mass index (BMI), and age were explored as correlates of discrepancies. RESULTS: The study sample consisted of 3094 older adults (1972 [63.7%] female; median [IQR] age, 72 [66-81] years). Cohen κ between dyads of equations ranged from 0.42 to 0.91, with poorest concordance between MDRD and BIS, and best between RLM and EKFC. MDRD and CKD-EPI provided higher estimates of GFR compared with the other equations. The best mix of AUC and Harrel C statistic was observed for BIS (0.80 and 0.73, respectively); however, the prognostic accuracy for death decreased among those aged over 78 years and those with low calf circumference. Differences between equations were inconsistent across levels of calf circumference, BMI, and age. CONCLUSIONS AND RELEVANCE: In this cohort study, we found that eGFR equations were not interchangeable when assessing kidney function. BIS outperformed other equations in predicting mortality; however, its discriminative capacity was reduced in subgroup analyses. Clinicians should consider these discrepancies when monitoring kidney function in old age.