Etienne Chatelut, Alan V. Boddy, François Doz
on behalf of Centre Claudius Regaud (Toulouse), Cancer Research Unit (Newcastle), Institut Curie (Paris)
Ultrafiltrable carboplatin AUC has a major impact on its haematologic toxicity and seems to correlate with the therapeutic outcome. We performed two separate studies by using NONMEM to correlate ultrafiltrable carboplatin clearance (CL) to patient characteristics. A two-compartment pharmacokinetic model and proportional error model for the residual and inter-patient variabilities were used.
In adults, since our previous study [J Natl Cancer Inst 87: 573, 1995], we performed pharmacokinetics in 31 new patients (56 cycles). No bias was observed between the clearances predicted according to our formula:
0.134·weight + (218·weight·(1-0.00457·age) ·(1-0.314·sex))/creatinine(uM)
with sex = 0 if male and sex = 1 if female, and the actual values (median percent error + 4%, range -51 to +47%, 10th and 90th percentiles, respectively, -25 and +23%). Its accuracy was also studied in a subpopulation of 25 obese patients. By using the actual weight, CL were significantly overpredicted (by more than 20% for 7/25 pts). By using the mean value between the ideal and the actual weight, a good prediction of CL was obtained: the percent error ranged from -33 to 39% and was comprised between -22 and 22% for 23/25 pts. This formula is used for carboplatin dosing either in standard regimens or in clinical trials.
In 57 children, CL was significantly correlated to three co-variates: weight, serum creatinine, and nephrectomy status [Clin Pharmacol Ther 1996; 59: 436 – 43]. The formula obtained from the data of 117 children (CL in mL/min = 2.9·weight·(1 -0.00368·Scr)·(1 – 0.239·Np) + 9.8 ; with Np = 1 or 0 for unilateral nephrectomy or not) was similar to the initial equation obtained from the 57 first studied children. The possibility of obtaining accurate value of CL from one sample per patient (at 1 hr postinfusion) was investigated prospectively in 23 patients by using Bayesian estimation under NONMEM (with the formula and a data base of the remaining patients). Absolute percent of error between calculated and actual value of CL never exceded 20%.
In conclusion, regression formulas obtained by NONMEM analysis allow to calculate the first carboplatin dose according to the patient characteristics. The following doses can be adjusted according to a small number of observed carboplatin concentrations and tolerance from the previous cycle.
Reference: PAGE 6 () Abstr 592 [www.page-meeting.org/?abstract=592]
Poster: oral presentation