Which equations should and which should not be employed in calculating eGFR in children?

PurposeWe assessed the reliability of calculating eGFR in children as compared to the iohexol disappearance test (GFR-I), which was performed 417 times in 353 children aged 2 and more.Material/methodseGFR was estimated with equations based on serum creatinine: Schwartz (1: eGFR-Scr), Cockroft–Gault (2: eGFR-CG) and MDRD (3: eGFR-MDRD), and on creatinine clearance (4: eGFR-U), or relying on serum cystatin C: Hoeck (5: eGFR-H), Bokenkamp (6: eGFR-B) and Filler (7: eGFR-F), and on the three Schwartz markers (8: eGFR-S3M). Mean relative error (RE), correlation (R), Bland–Altman analysis and accuracy of GFR-I were studied in all patients and in subgroups: at GFR < 60 ml/min/1.73 m2; in children aged ≤12 and >12.ResultsThe results by eGFR-Scr, eGFR-S3M demonstrated no statistical difference to GFR-I at GFR < 60 ml/min/1.73 m2, but underestimated eGFR at higher filtration values by 11.6 ± 15.1% and 19.1 ± 16.4, respectively (p < 0.0000). The eGFR-B, eGFR-F and eGFR-MDRD equations illustrated important overestimation of reference GFR results (RE: 84 ± 44.2%; 29.5 ± 27.9%, 35.6 ± 62%; p < 0.0000 for all). The MDRD and C–G formulas showed statistically better consistency in children aged >12. A good agreement was achieved by the eGFR-H equation (5.1 ± 21.9%; p < 0.0000; R = 0.78).Conclusions(1) Schwartz equations show a good conformity at GFR < 60 ml/min/1.73 m2, but underestimate the results at higher GFR values. (2) The Bokenkamp equation with original coefficient should not be employed in children. (3) The use of the Hoeck formula in all children and C–G and MDRD formula in children aged >12 is possible. (4) The error of eGFR calculations increases at higher GFR values.