Renal function equations before and after living kidney donation: a within-individual comparison of performance at different levels of renal function

Abstract

Background and objectives: The Modification of Diet in Renal Disease (MDRD) study equation and the Cockcroft-Gault (CG) equation perform poorly in the (near-) normal range of GFR. Whether this is due to the level of GFR as such or to differences in individual characteristics between healthy individuals and patient with chronic kidney disease (CKD) is unknown.

Design, setting, participants, & measurements: We evaluated the performance of MDRD, CG per BSA (CG/(BSA)) and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations compared with measured GFR (mGFR; I-iothalamate) at 4 months before and 2 months after donation in 253 consecutive living kidney donors.

Results: mGFR declined from 103 ± 15 to 66 ± 11 ml/min per 1.73 m(2) after donation. All equations underestimated mGFR at both time points. Arithmetic performance analysis showed improved performance after donation of all equations, with significant reduction of bias after donation. Expressed as percentage difference, mGFR-estimated GFR (eGFR) bias was reduced after donation only for CG/(BSA). Finally, in 295 unselected individuals who were screened for donation, mGFR was below the cutoff for donation of 80 ml/min per 1.73 m(2) in 19 individual but in 166, 98, and 74 for MDRD, CDK-EPI, and CG/(BSA), respectively.

Conclusions: A higher level of GFR as such is associated with larger absolute underestimation of true GFR by eGFR. For donor screening purposes, eGFR should be interpreted with great caution; when in doubt, true GFR should be performed to prevent unjustified decline of prospective kidney donors.

Figure 1.

Individual values of the change (Δ) in bias over donation (Unx). Δ was calculated as (predonation bias − postdonation bias) in ml/min per 1.73 m² for arithmetic values (left) and as a percentage difference for relative values (right). Error bars represent median bias. The arithmetic values show a positive median bias for all equations with a large variation between the individual donors. Relative values show a small negative median bias for MDRD and CKD-EPI study equations, again with large variation.

Figure 2.

Regression of eGFR on mGFR/_BSA for all three equations for predonation (dark dots; solid line) and postdonation (gray diamonds; dotted line) values. Pre- and postdonation slopes of the regression lines are, respectively, 0.36 ± 0.05 and 0.55 ± 0.05 for MDRD, 0.43 ± 0.06 and 0.47 ± 0.04 for CKD-EPI, and 0.26 ± 0.04 and 0.39 ± 0.04 for CG/_BSA.

Figure 3.

Pre- and postdonation bias as arithmetic (left) and relative values. Arithmetic bias was calculated as (mGFR − eGFR) and relative bias as [(mGFR − eGFR)/mGFR * 100]. Bias was divided according to tertiles of predonation mGFR/_BSA; therefore, pre- and postdonation tertiles contain the same donors. Tertile median (IQR) values of predonation mGFR/_BSA were 87 (82 to 92), 102 (99 to 106), and 118 (113 to 125), respectively. Corresponding values for postdonation mGFR/_BSA were 56 (53 to 62), 66 (61 to 71), and 73 (68 to 77) ml/min per 1.73 m². Bars represent medians with 95% confidence intervals. *P < 0.05 versus first tertile; #P < 0.05 versus second tertile.

Figure 4.

Predicted probability of mGFR <80 when eGFR was <80 ml/min per 1.73 m² by Bayes theorem. Percentages represent probabilities calculated by Bayes theorem, for MDRD (M), CKD-EPI (C-E), and CG/_BSA (CG) equations in 295 subsequent individuals who were referred for donor screening without previous selection. Break-ups are made for gender, age, and BMI. The number of individual in each subgroup is as follows: ^an = 20, ^bn = 41, ^cn = 25, ^dn = 45, ^en = 39, ^fn = 43, ^gn = 34, and ^hn = 48.