Validation of child serum creatinine-based prediction equations for glomerular filtration rate

Pediatr Nephrol. 2007 Feb;22(2):272-81. doi: 10.1007/s00467-006-0322-0. Epub 2006 Nov 21.


Equations for estimating glomerular filtration rate (GFR) are vital in caring for patients with renal disease and the current standard, the Schwartz formula, lacks precision. We evaluated several child serum creatinine-based GFR prediction equations. Subjects aged 2-21 years who underwent iothalamate GFR (IoGFR) testing between 1999 and 2004 were studied retrospectively. GFR was estimated using: (1) Schwartz formula (SchwartzGFR), using a local k value; (2) Schwartz model (SchwartzMod) using regression-derived coefficients; (3) Leger GFR (LegerGFR) using original coefficients; and (4) Leger model (LegerMod) using regression-derived coefficients. Bias, precision, and diagnostic characteristics were evaluated. There were 195 subjects [61% male; mean (SD) age 12.4 (4.5) years; mean (SD) IoGFR 78.9 (33.4) ml/min per 1.73 m(2)]. Only the LegerGFR overestimated IoGFR (5.5 ml/min per 1.73 m(2)). Precision for all formulae was poor (95% limits of agreement approximately -40 to 40 ml/min per 1.73 m(2)), but >or=72% of estimates were within 30% of IoGFR. Sensitivities for detecting IoGFR <30 and 90 ml/min per 1.73 m(2) were highest using the SchwartzGFR (80%) and SchwartzMod (90%), respectively. The LegerGFR was most specific. Using local coefficients, the Schwartz and Leger models were imprecise estimates of GFR, but the Schwartz model was most unbiased and sensitive. Future research should derive more precise equations for GFR in children.

Publication types

  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Adolescent
  • Adult
  • Bias
  • Child
  • Child, Preschool
  • Creatinine / blood*
  • Female
  • Glomerular Filtration Rate / physiology*
  • Humans
  • Male
  • Models, Theoretical*
  • Predictive Value of Tests
  • Regression Analysis
  • Retrospective Studies
  • Sensitivity and Specificity


  • Creatinine