Serum albumin, but not glycated albumin was a potent factor affecting the performance of GFR equation based on serum creatinine.

Background: Recently, Tsuda et al. reported that high HbA1C or high glycated albumin (GA) level is a major factor in overestimation of GFR by Japanese GFR equation based on serum creatinine (Eq-cr). They developed a modified equation of Eq-cr (M-Eq-cr) using GA or HbA1c. Therefore, effect of GA levels on the estimated GFR (eGFR) by Eq-cr was evaluated in Japanese subjects. We validated the accuracy of the modified equation using GA by Tsuda et al. (M-Eq-cr) and new equations that we developed in the present study.
Methods: Seven hundred and fifteen Japanese subjects were included. GFR was measured by inulin renal clearance (Cin). The subjects were divided into two groups by upper limit of the GA reference range (GA-1: GA < 16.3 % and GA-2: GA > 16.4 %). Factors affecting the ratio of eGFR to Cin (eGFR/Cin) were evaluated using multivariate analysis. New equations based on creatinine and albumin (Eq-cr-alb) and based on creatinine, albumin and GA were developed from development dataset (382 subjects). Performances of the equations were validated in validation dataset (333 subjects).
Results: Correlation coefficients between eGFR by Eq-cr and Cin were 0.839 and 0.914 in GA-1 and GA-2, respectively. Slopes (95 % confidential interval) of the regression lines with zero intercepts were 1.013 (0.991 to 1.036) and 0.997 (0.951 to 1.043), respectively. Both slopes were not significantly different from 1.0. Biases were -2.3 ± 19.0 and 0.2 ± 11.7 ml/min/1.73 m(2), respectively. Accuracy (p30; percentage of subjects within 30 % of Cin) (95 % CI) were 78 % (75, 81) and 71 % (62, 78), respectively. There was no significant difference in bias and accuracy between the two groups, indicating a reasonable accuracy of Eq-cr in GA-1 and GA-2. Multiple regression analysis showed that lower serum albumin and higher GA were associated with higher eGFR/Cin. Albumin was a more potent factor affecting eGFR/Cin than GA. M-Eq-cr significantly underestimated GFR and had significantly larger bias compared with Eq-cr in subjects with GA > 20 %, suggesting that the modification of Eq-cr using GA by Tsuda et al. was too much compensation in our subjects. Precisions of Eq-cr-alb were significantly better compared with Eq-cr.
Conclusion: Eq-cr has a reasonable accuracy in GA-1 and GA-2. Lower serum albumin and higher GA were significantly associated with higher eGFR/Cin. The former was a more potent factor affecting eGFR/Cin. Eq-cr-alb showed better performance compared with Eq-cr. M-Eq-cr using GA showed too much compensation and did not improve the accuracy of the equation in our subjects.