Background: High dose melphalan is standard conditioning for autologous stem cell transplantation in the treatment of multiple myeloma. Appropriate dosing is critical as melphalan may be associated with profound toxicity, including cytopenia and gastrointestinal toxicity, which can lead to significant treatment-related mortality, while insufficient dose intensity can lead to disease relapse. The aim of this study was to develop and evaluate a population pharmacokinetic model for melphalan in myeloma patients and identify factors that affect melphalan pharmacokinetic parameters.
Methods: Total melphalan concentrations were measured in 6-12 plasma samples collected after a single intravenous dose from 69 myeloma patients aged between 36 and 73 years. The standard 200 mg/m2dose was administered to 40 patients, while the remainder received reduced doses due to obesity, impaired renal function or other reasons. Creatinine clearance (CLCR, L/h) was estimated using the Cockcroft and Gault equation  using fat free mass (FFM, ) in the equation instead of total body weight. Population pharmacokinetic modelling was performed with the FOCE method (η-ε interaction) in NONMEM (version VI), using an exponential random effects model to describe interpatient variability and a combined additive and proportional error model to estimate the residual unexplained error. Patient covariates, including CLCR (L/h), weight (kg), weight with an allometric scaling function (WT0.75, kg), FFM (kg), gender, body mass index (kg.m-2) and age (y) were tested for significant (p<0.05) relationships with pharmacokinetic parameters, then incorporated into the model in a forward stepwise manner followed by backwards elimination. Model stability was assessed using the bootstrap option in the Wings for NONMEM software (N = 872).
Results: A two compartment model best fitted the concentration-versus-time data. The following structural models for clearance (CL) and volume of distribution of the central compartment (V1) were obtained:
(1) CL (L/h) = 15.5 x (FFM/53)0.5 + 11.2 x CLCR /4
(2) V1(L) = 12.9 x FFM/53
Interpatient variability (co-efficient of variation) was 25% for CL and 57% for V1.
Mean bootstrap estimates of population pharmacokinetic parameters were comparable with those obtained using the original data set indicating that the covariate model had acceptable accuracy and stability: differences were within ± 2% for all parameters except the additive component of residual variability, which was -5.2%.
Conclusions: A validated population pharmacokinetic model for melphalan in myeloma patients has been developed. CLCR and FFM were important determinants of melphalan CL, while FFM was an important determinant of V1. This data may be used in conjunction with pharmacodynamic data to develop an optimal dosing schedule for melphalan in myeloma patients.
 Cockcroft DW and Gault MH. Nephron 1976; 31.
 Janmahasatian S et al. Clin. Pharmacokinet. 2005; 44: 1051.