Applicability evaluation of the GFR equations in Chinese patients with diabetes mellitus complicated with CKD

Objective To explore applicability of Chronic Kidney Disease Epidemiology Collaboration(CKD-EPI)'s glomerular filtration rate (GFR) equations, including modified modification of diet in renal diseases (MDRD) equation, the Feng equation developed for Chinese in recent years, and modified CKD-EPI diabetes equation for Chinese patients with diabetes, in Chinese patients with diabetes mellitus complicated with CKD; furthermore, to provide a more accurate basis for accurate estimation of GFR in patients with diabetes complicated with CKD. Methods A total of 160 patients with type 2 diabetes with CKD were selected. The patient's data on gender, age, height, weight, blood creatinine (Scr) and cysteine protease inhibitor C (Cys C) were collected. GFR determined by ^99mTc-DTPA renal dynamic imaging was used as a reference standard (rGFR). Modified equations for Chinese patients, including MDRD equation, CKD-EPI_Scr equation, CKD-EPI_{Cys C} equation, CKD-EPI_{Scr-Cys C} equation, CKD-EPI diabetes equation, Feng_{Cys C} equation and Feng_{Scr-Cys C} equation, were used to estimate GFR (eGFR). Pearson correlation analysis and Kappa test were used to compare the correlation and staged consistency between eGFR and rGFR for each equation. Bland-Altman analysis was used to compare the consistency limits of eGFR and rGFR for each equation. Bias, precision, 10% accuracy, 30% accuracy and 50% accuracy were used to compare eGFR values for each equation. GFR<60 mL·min^-1·(1.73 m²)^-1 was defined as the diagnostic criteria for renal insufficiency, and the performances of various equations for diagnosing renal insufficiency were compared by ROC curve. Patients were divided into group A (CKD stage 4~5), group B (CKD stage 3), and group C (CKD stage 1~2) according to the rGFR, and the applicability of various equations in each group was compared. Results In the comparison of eGFR between the equations, the CKD-EPI_diabetes equation had the smallest bias, with differences of statistical significance compared to Feng_{Cys C} equation and Feng Scr-Cys C equation (P<0.05); it had the highest precision, with differences of statistical significance compared to MDRD equation and CKD-EPI_{Cys C} (P<0.05). In the comparison of accuracy between the 7 equations,, the 10% accuracy and the 30% accuracy of the CKD-EPI_diabetes equation were the highest, with differences of statistical significance compared to the simplified MDRD equation, CKD-EPI_Scr equation and CKD-EPI_{Cys C} equation (P<0.05). Bland-Altman analysis showed that CKD-EPI_diabetes equation had the best consistency limit. In group A, the difference between CKD-EPI_{Cys C} equation eGFR and rGFR was not statistically significant, with the least bias, and highest 10% accuracy, 30% accuracy and 50% accuracy. Group B were similar to the overall groups, CKD-EPI_diabetes equation had the least bias, the highest precision, 10% accuracy, 30% accuracy and 50% accuracy. Conclusions Among the 7 equations, the CKD-EPI_diabetes equation has the best fit between eGFR and rGFR. In the grouped comparison, the CKD-EPI_{Cys C} equation has the best fit in patients with CKD stage 4~5, CKD-EPI_diabetes equation has the best fit in patients with CKD stage 3. The two equations can be used for estimation of GFR in different groups of patients with diabetes complicated with CKD.