AimsWe evaluated the performance of creatinine-based equations that are currently used to estimate glomerular filtration rate (GFR) in people with type 2 diabetes compared to measured GFR using gold-standard methods. MethodsIn this post-hoc analysis, 32 participants underwent repeated measurement of GFR by inulin clearance (mGFR). GFR was estimated by serum creatinine using the MDRD (eGFRMDRD) and CKD-EPI (eGFRCKD-EPI) equations four times over the course of one month. Performance was evaluated using measurements of bias (mean difference), precision (SD), and inaccuracy (proportion of eGFR that differed by >20 % of mGFR). Treatment and time effects on bias were evaluated using linear mixed effects models. ResultsAt baseline, participants (38 % female) were age 60 ± 8 years, had diabetes duration of 9 ± 7 years, HbA1c 56 ± 9 mmol/mol (7.2 ± 0.8 %), and BMI 31.0 ± 6.2 kg/m2. Mean mGFR was 113 ± 24, mean eGFRMDRD was 93 ± 12, and mean eGFRCKD-EPI was 94 ± 9 mL/min/1.73 m2. When 128 observations (32 participants measured 4 times) were evaluated, both equations substantially underestimated mGFR. For eGFRMDRD, mean bias was −21.5 mL/min/1.73 m2, precision was 22.7 mL/min/1.73 m2, and 46 % of observations differed by >20 %. Results were similar for eGFRCKD-EPI. No time or treatment effects on bias were observed. ConclusionsIn adults with type 2 diabetes and preserved renal function, eGFR equations underestimated mGFR, lacked precision and accuracy, and performance was lower at higher ranges of mGFR. Current eGFR equations by serum creatinine are inaccurate in adults with type 2 diabetes with preserved renal function, highlighting the necessity to develop new methods to measure kidney function at earlier stages of diabetic kidney disease.
Read full abstract