MDRD-derived eGFR is advocated for assessment of kidney function in patients with diabetes . Although validated in CKD , it has recognised limitations outside of this setting. Underestimation of higher GFR levels with overestimation of CKD has been observed . Underestimation of iGFR has been attributed to inaccuracies of measurement, and greater variability in creatinine concentration due to non-renal factors in patients without CKD [4–6].
In this study, IDMS creatinine values were significantly lower than traditional creatinine values. Both formulas introduced significant biases and tended to underestimate GFR. This being most pronounced applying the 186 MDRD formula to patients with iGFR ≥90 ml min−1 1.73 m−2.
This finding contrasts with those of Froissart , who demonstrated a bias of −6.2 ml min−1 1.73 m−2 in a subgroup of 482 European patients with CKD and GFR >90 ml min−1 1.73 m−2. The difference may partly be explained by the fact that our patients were newly diagnosed with type 2 diabetes, many being obese, both factors being known to be associated with hyperfiltration.
We recognise that underestimation may be partly attributable to creatinine values not being calibrated to those of the MDRD laboratory. However, the MDRD equation was also derived using GFR measured by the urinary clearance of 125I-labelled iothalamate rather than 51Cr-EDTA clearance; it does not specify that identical methods of creatinine measurement are required for its use and our findings represent current clinical practice.
Use of IDMS-traceable creatinine values with the 175 MDRD equation partially improved performance of the equation in terms of both bias and accuracy, in the group described. IDMS creatinine values were derived using recognised formulas, which, pending widespread implementation of IDMS-traceable creatinine measurements, again reflects current clinical practice. Evidence that this transformation improves performance is encouraging.
The lower bias observed with the 175 MDRD equation was statistically significant in all subgroups of the study. Furthermore, statistically significant improvements in accuracy of estimation within 10% and 30% of isotopic GFR were observed when the group was studied as a whole, and in patients from group 2. Improved accuracy was also seen in group 1, although this failed to reach statistical significance, possibly due to the small number studied. Precision of equations was unchanged using the 175 equation compared with the 186 MDRD, although improved precision was seen with decline in iGFR.
Overall, use of IDMS-traceable creatinine in the revised 175 MDRD equation did slightly improve estimation of iGFR in patients with diabetes. We conclude that universal standardisation of creatinine measurement will improve the ability of the MDRD equation to detect early changes in GFR.
However, despite improvements in accuracy of creatinine measurement, plasma creatinine concentrations will fluctuate according to non-renal factors such as diet, exercise and BP. This will influence eGFR results, especially when creatinine values are within the normal reference range .
The variability of eGFR is illustrated in our study of mainly white normoalbuminuric patients, where significant underestimation of iGFR remained despite use of IDMS creatinine values and the revised MDRD equation.
While efforts are being made to generate equations which more accurately reflect GFR in diabetes, standardised creatinine values are a positive step towards improved eGFR.
Use of eGFR for assessment of kidney function in patients with diabetes is now established. Our data show that use of IDMS-standardised creatinine improves the overall performance of the MDRD equation; however, significant underestimation of iGFR continues to occur in this patient group.