External validation of a model based on cystatin C to predict carboplatin clearance
A. Schmitt(1), L. Gladieff(1), A. Lansiaux(2), C. Bobin-Dubigeon(3), M.C. Etienne(4), F. Pinguet(5), F. Serre-Debeauvais(6), A. Floquet(7), C. le Guellec(8), M. Boisdron-Celle(9), E. Billaud(10), N. Penel(2), M. Campone(3), R. Largillier(4), M. Fabbro(5), M. Mousseau(6), N. Houede(7), P. Bougnoux(8), O. Capitain(9), J. Medioni(10), E. Chatelut(1).
(1) Institut Claudius Regaud (Toulouse), (2) Centre Oscar Lambret (Lille), (3) Centre René Gauducheau (Nantes), (4) Centre Antoine Lacassagne (Nice), (5) Centre Val d’Aurelle Paul Lamarque (Montpellier), (6) Centre Hospitalo-Universitaire de Grenoble, (7) Institut Bergonié (Bordeaux), (8) Centre Hospitalier Régional Universitaire de Tours, (9) Centre Paul Papin (Angers), (10) Hopital Européen Georges Pompidou (Paris).
Objectives: The individual dosing of drugs that are mainly eliminated unchanged in the urine is made possible by assessing renal function. Carboplatin is one of the drugs for which elimination is most dependent on glomerular filtration rate (GFR). The formulas actually used for individual carboplatin dosing are all based on serum creatinine (SCr) as the unique biological covariate (together with demographical and morphological covariates) [1, 2]. Thomas et al. [3] recently proposed a formula including plasma cystatin C level (CysC), an other endogenous marker of GFR. A clinical trial was conducted in 12 centers, to identify pharmacodynamics covariates of toxicity. A first ancillary study was performed to assess the Thomas formula for prediction of carboplatin clearance (CL).
Methods: The patients were receiving 1 hour-infusion of carboplatin as part of established protocols. Samplings were done 5 min before the end of infusion, 1, and 4 hours after the end of infusion. A population pharmacokinetic analysis was performed using the nonlinear mixed effect modelling NONMEM program and FOCE estimation method. Data from 260 patients were used to evaluate Thomas formula, which takes into account SCr, CysC, body weight (BW), age and sex.
Results: In a first time, individual POSTHOC CL were compared to values predicted by the Thomas equation. The Mean Percentage Error (MPE) was 1% with [-25%; +37%] as 5th-95th percentiles, and the Mean Absolute Percentage Error (MAPE) was 14%. In a second time, a covariate analysis was performed. The best covariate equation was: CL(mL/min){vs. previous value of Thomas formula} = 105,5{110}*(SCr/75)-0,332{-0.512}*(CysC/1,00)-0,473{-0.327} *(BW/65)0,616{0.474}*(age/56)-0,178{-0.387}*0,864{0.854}sex, with SCr in µmol/L, CysC in mg/L, BW in kg, age in years, and sex = 0 for male. Deletion of each covariate was associated with a significant increase of the objective function value (p<0.005). Finally, the model was validated by both a visual predictive check and bootstrapping with simulation on the external validation dataset.
Conclusions: External validation is the highest degree of validation for PK model. The Thomas formula has been validated at a multi-center level. These results confirm definitively the benefit of cystatin C as a marker of renal elimination of drugs. However, it should be used with other morphological and demographical covariates. Serum creatinine and cystatin C are not completely redundant marker of GFR.
References:
[1] Calvert, A.H., et al., Carboplatin dosage: prospective evaluation of a simple formula based on renal function. J Clin Oncol, 1989. 7(11): 1748-56.
[2] Chatelut, E., et al., Prediction of carboplatin clearance from standard morphological and biological patient characteristics. J Natl Cancer Inst, 1995. 87(8): 573-80.
[3] Thomas, F., et al., Cystatin C as a new covariate to predict renal elimination of drugs: application to carboplatin. Clin Pharmacokinet, 2005. 44(12): 1305-16.
