MO513: Implications of Using Various Equations to Estimate Glomerular Filtration Rate in the GNC (NAKO) Study

Abstract

Abstract BACKGROUND AND AIMS Glomerular filtration rate (GFR) is the most common measure of kidney function. Measuring GFR is often infeasible in routine clinical practice and population studies, where it is estimated instead from blood biomarkers. Three currently recommended equations, eGFR(crea), eGFR(cystC) and eGFR(cr + cy), use either creatinine, cystatin C or both, respectively, to calculate eGFR [1]. In addition, three alternative equations were introduced recently: two of them update eGFR(crea) and eGFR(cr + cy) by excluding the race term used in them [2], and the full age spectrum (FAS) equation (creatinine-based), which was specifically developed to estimate GFR across the full age range [3]. We aimed to assess the impact of these six equations on GFR estimates and the proportion of participants with chronic kidney disease (CKD) in the German National Cohort (GNC, NAKO) Study. METHOD The NAKO study is a prospective cohort study in Germany. Data to estimate GFR (eGFR) using the six equations was available for 112 843 participants aged 19–75 years, mostly of European ancestry. Pairwise Spearman correlation coefficients between GFR estimates were calculated. CKD was defined as an eGFR <60 mL/min/1.73 m² for each calculation. Agreement between CKD status derived from the different GFR equations was assessed using Cohen's κ. Measured GFR was not available. RESULTS Median eGFR/equation ranged from 92.2 mL/min/1.73 m² using the new FAS-equation [3] to 105.7 mL/min/1.73 m² using the updated race-free eGFR(cr + cy)2 (Fig. 1). eGFR values of the three equations using creatinine only, as well as eGFR values from the two equations using both creatinine and cystatin C, correlated well (r > 0.95), while the correlation of GFR estimates based on equations either using creatinine or cystatin C was only moderate (r = 0.6, Fig. 1). The frequency of CKD ranged from 1.0% to 2.3%, a >2-fold difference. The agreement assessed as Cohen's κ was the smallest when comparing the CKD status based on eGFR(cy) [1] with the CKD status based on the updated race-free eGFR (crea) [2] (Cohen's κ = 0.38). While both equations led to a mutual CKD classification in 707 participants (0.6%), 613 (0.5%) were only classified as having CKD when the updated race-free eGFR (crea) [2] was used, and 1580 (1.4%) were only classified as having CKD using eGFR (cy) [1]. Interestingly, the respective serum biomarker that led to the discordant classification of CKD was abnormally high, while the other biomarker was mainly in the normal range, highlighting the potential influence of non-kidney related factors on biomarker levels that influence estimates of CKD prevalence. CONCLUSION The use of different GFR equations leads to variable estimates, affecting the proportion classified as having CKD. Non-GFR determinants with an impact on the levels of either creatinine or cystatin C may lead to incorrect estimates of GFR and affect CKD prevalence estimates, which underscores the utility of a joint view using various equations.