Glomerular filtration rate (GFR) in children can be estimated by the formula GFR=k x BH/Pcr (where BH is body height in centimetres and Pcr is the plasma creatinine concentration in micromoles per litre). For k, several values have been reported: k=38 (Counahan), k=40 (Morris) and k=48.7 (Schwartz). In this study the predictive performance of these formulae was compared with that of newly developed formulae. GFR measurements based on inulin concentration time curves were divided into an index (n=58) and a validation data set (n=48). In the index data set a value for k was derived by application of nonlinear mixed-effect modelling. This approach was also used to develop a formula that better explained the relationship between patient factors and GFR. Bias and precision of all formulae were calculated for the validation data set. In the index data set a value of 41.2 was found for k, which was close to the value k=40 (Morris). Both formulae estimated GFR well (bias <5%; precision 25%). Further modelling of the relationship between patient factors and GFR did not improve the predictive performance. In our hospital GFR was best estimated by the formula with k=40 and k=41.2. It is recommended that the optimal value for k be assessed locally.