Test accuracy of glomerular filtration rate estimation with creatinine and cystatin C in adults with moderate chronic kidney disease: prospective cohort study

Objective To study the performance of two contemporary sets of estimating equations for glomerular filtration rate, published by CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) and EKFC (European Kidney Function Consortium) that include one (creatinine or cystatin C only) and combined (creatinine and cystatin C) biomarkers, to assess their accuracy in a population with moderate chronic kidney disease.

Design Prospective cohort study.

Setting Primary, secondary, and tertiary care in six centres in England. Participants were recruited from April 2014 to January 2017.

Participants 1167 adults, aged ≥18 years, with moderate chronic kidney disease (estimated glomerular filtration rate 30-59 mL/min/1.73 m² sustained over at least three months before recruitment).

Main outcome measures  Accuracy of estimating equations CKD-EPI_creatinine, CKD-EPI_cystatin, CKD-EPI_{creatinine-cystatin}, EKFC_creatinine, EKFC_cystatin, and EKFC_{creatinine-cystatin} compared with measured glomerular filtration rate (iohexol clearance). Remodelled 2021 versions of the CKD-EPI equations were also studied. Accuracy was expressed as P30 (percentage of estimates within 30% of measured glomerular filtration rate).

Results Median age was 67.5 years, 58.3% of patients were men, 86.9% were white participants, and 27.8% had diabetes. Median measured glomerular filtration rate was 47.0 mL/min/1.73 m2; 57.0% of participants had albuminuria. Test calibration critically affected measurement of cystatin C. After recalibration of cystatin C, P30 values were 90.2% (CKD-EPI_creatinine), 89.5% (CKD-EPI_cystatin), 94.9% (CKD-EPI_{creatinine-cystatin}), 88.0% (CKD-EPI(2021)_creatinine), 94.9% (CKD-EPI(2021)_{creatinine-cystatin}), 89.4% (EKFC_creatinine), 91.0% (EKFC_cystatin), and 94.9% (EKFC_{creatinine-cystatin}). Creatinine based equations showed varying bias depending on the glomerular filtration rate level; inclusion of cystatin C in the equations improved this effect. Differences in accuracy in age, sex, and glomerular filtration rate level subgroups varied by equation. Equations combining creatinine and cystatin performed equally across age, sex, diabetes status, albuminuria status, and body mass index categories.

Conclusions The CKD-EPI_creatinine equation had acceptable accuracy in a white population in England with moderate chronic kidney disease. Combined dual biomarker equations showed higher accuracy than the CKD-EPI_creatinine equation and their equivalent creatinine only equations. Further research is needed to determine the most accurate equation to use in people of black and South Asian origin living in England.