Assessment of the glomerular filtration rate (GFR) is widely accepted as the best index of renal function in patients. Many chemotherapeutic drugs are excreted to a large extent via the kidneys, thus a reliable and accurate measurement of this parameter is particularly important in oncology practice. The relationship between kidney function and the extent of drug exposure is best exemplified by carboplatin, where the dose of drug administered is determined by renal function in both adult and paediatric patients (Calvert et al, 1989; Newell et al, 1993; Thomas et al, 2000). In addition, a measurement of renal function may be important in monitoring the nephrotoxic effects of drugs such as cisplatin and ifosfamide (Skinner et al, 1994).
An accurate determination of GFR can be obtained by measuring the clearance of chromium 51 EDTA (51Cr-EDTA) or similar isotope-based methods (Chantler et al, 1969; Rodman et al, 1993). This approach would be recommended when an accurate prediction of GFR is required, particularly in patients with reduced renal function. However, EDTA is not licenced for use in countries such as the USA. Alternatively, GFR can be estimated from serum creatinine concentration or calculated creatinine clearance (Perrone et al, 1992). These latter methods offer a less precise estimation of renal function, but can generally be performed with minimum patient inconvenience and at a lower cost than the isotopic methods. In paediatrics, problems may arise as these methods have commonly been validated in adults and it is difficult to obtain an accurate collection of urine over a 24-h period. A noninvasive, simple and reliable mathematical model for predicting GFR in a paediatric patient population would be advantageous, particularly if the withdrawal of multiple blood samples could be avoided.
Several mathematical equations and nomograms have been developed to predict renal function (Jelliffe, 1973; Cockcroft and Gault, 1976; Schwartz et al, 1976). The most commonly used of these formulae to predict creatinine clearance, and hence GFR, in adult patients have been those published by Cockcroft and Gault and that of Jelliffe, using age, sex, serum creatinine and either body weight (Cockcroft and Gault) or surface area (Jelliffe) as a measure of body size. In paediatric patients, perhaps the most widely used is that described by Schwartz, which is based on the ratio of height to serum creatinine concentration and includes an adjustment for patient age (Schwartz et al, 1976).
Whereas in adult patient populations these formulae approximate GFR to an acceptable level (Luke et al, 1990), similar studies in paediatric patient populations have highlighted inaccuracies. In a study involving patients aged between 2 and 18 years, 95% of GFR estimates obtained from the Schwartz formula would be expected to lie within 50% of the GFR determined as the clearance of 51Cr-EDTA (Skinner et al, 1994). In clinical practice, a higher level of accuracy is often required and the use of these models has been associated with inaccurate dosing of carboplatin (Calvert and Egorin, 2002).
The statistical methodology routinely used to model 51Cr-EDTA pharmacokinetics does not take into account the sampling variability in the estimates of 51Cr-EDTA clearance. For this reason, approaches based on nonlinear mixed effects models, often referred to as population models, have recently been published and independently assessed (Martin et al, 1998; Wright et al, 2001; Poole et al, 2002; Léger et al, 2002). These methods involve the use of patient specific 51Cr-EDTA plasma data together with supplementary covariate information. The current study applies nonlinear mixed effects modelling to the pharmacokinetics of 51Cr-EDTA, with a view to predicting GFR in paediatric cancer patients.