Aims ‘Clearance’ is an accepted surrogate for glomerular filtration rate (GFR) that describes the relation between the serum/plasma concentration of a substance, for example, creatinine, and its urinary excretion rate. On a population level the GFR can be estimated from the creatinine concentration by regression analysis. Such estimates inevitably increase the uncertainty compared with the measured quantity, for example, creatinine. Reliable estimates of GFR (eGFR) are crucial for proper administration of drugs with small therapeutic ‘windows’ and require a transformation of ‘normalised’ GFR to ‘individual’ GFR using the individual's anthropometric data. By simulation we compare some common algorithms and discuss their feasibility for drug dosing.

Methods We compared four algorithms for normalised eGFR and one for individual eGFR in a Caucasian man defined by age and simulated the outcome of different S-Creatinine concentrations. Normalised and individual eGFR were interconverted using published anthropometric data of a national average man.

Results Differences between results by the algorithms for eGFR were less than the uncertainty within the diagnostically important interval. Three of the four algorithms were non-continuous at extreme S-Creatinine concentrations. Normalised eGFR values, converted to individual eGFR resulted in an underestimation of 5–40% compared with that by Cockcroft-Gault.

Conclusions The use of algorithms will increase the uncertainty of results. The difference between different algorithms is small. A global staging recommendation is in jeopardy due to the large interindividual biological variation. Dosage of critical drugs, based on conversion of normalised eGFR, may lead to an underdosage compared with the dosage estimated by Cockroft-Gault.