Population pharmacokinetics modelling for clinical dose adjustment of carboplatin in dogs

Abstract

Background: Carboplatin is a human chemotherapeutic agent which is frequently used in dogs for the management of solid tumors. In human patient, its dosage is adjusted carefully, based on the creatinine clearance computation. In dogs however, the pharmacokinetics of carboplatin is poorly known and the dose 300 mg/m2 is based mostly on empirical data. Here, we aimed at characterizing the pharmacokinetics of carboplatin and determined the influence of several covariates, including creatinine plasma concentration and neutering status, in dogs, and used this model to predict myelotoxicity.

Results: Sixteen client owned dogs were included after carboplatin administration (300 mg/m²). For each animals, three to four plasma samples were collected and free plasma concentration of carboplatin was determined by HPLC/MS and analysed using Monolix® software with Non-linear mixed effect modelling. A mono-compartmental model best described the plasma concentration of carboplatin with log plasma creatinine concentration and sterilization status as covariates. After adjustment with the covariates, median population clearance was 3.62 [3.15 - 4.12] L/h/kg and volume of distribution was 3.93 [3.84 - 4.14] L/kg. The application of this model in 14 additional dogs demonstrates that individual drug exposure (model-predicted Area Under the Curve) predicted thrombocyte blood reduction (Pearson coefficient r² = 0.73, p = 0.002) better than dose after 14 days following administration of carboplatin.

Conclusion: Based on our results, plasma creatinine concentration and the sterilization status are relevant explanatory covariates for the pharmacokinetics variability of carboplatin in client owned dogs. Dose adjustment based on these parameters could represent a promising strategy for minimizing thrombocyte toxicity.

Keywords: Carboplatin; Dogs; Modelling; Renal function; Thrombocytopenia.

Fig. 1

A Individual predictions vs. observations expressed in a log10-log10 scale. Blue dots represents the individual observations. Black line represents the identity line and dotted black line the 90% prediction interval. B Distribution of the standardized random effect η_V and η_Cl for Volume of distribution and Clearance respectively

Fig. 2

Correlations between creatinine blood concentration or sterilization status of the animal and clearance. Both relationships were statistically significant according to Pearson correlation test (creatinine blood concentration) or ANOVA (sterilization status). r, Pearson correlation coefficient; Cl, Clearance (L/h/kg); *, p < 0.005

Fig. 3

A Visual Predictive Check (VPC) of the carboplatin concentration (log10 scale) vs. time. The individual values are shown as blue dots. The observed and predicted 10th and 90th percentiles of the interval prediction or empirical percentiles are shown by the blue area or blue line, respectively. Outliers are shown by red circle. B Individual predictions of carboplatin plasma concentration in dogs from the final selected model. Plots of individual observed (blue dot) and individual predicted (black line) concentration time course. Each black line represent the prediction of one occasion. Censored data (below Limit of Quantification) are shown as pink squares

Fig. 4

Thrombocyte and neutrophil blood count between day 0 and 14 days after carboplatin administration. *p < 0.05 between day 0 and day 14

Fig. 5

Relationship between the different pharmacokinetics variable (AUC_0-∞, dose in mg/kg or dose in mg/m²) and thrombocyte or neutrophil blood count reduction between the first day and 14 days. Each black line represent relationship predicted by the E_max model and dotted line represents the 90% prediction band when a significant relationship was evidenced