Development of a Predictive Pharmacokinetic and Pharmacodynamic Model to Personalize Melphalan Dosing in Autologous Transplant for Patients with Multiple Myeloma

Abstract

Introduction: Melphalan is an interstrand cross-link (ICL)-inducing agent and, in the setting of autologous stem cell transplantation for multiple myeloma, is one of the most effective treatments, providing 30 months of disease stability on average, but with a dramatic progression free survival (PFS) range of 6 months to 12 years. While 200 mg/m2 is the standard dose, there is extensive interpatient variability, and individual melphalan sensitivity with the ability to repair double-stand breaks in primary myeloma cells mirrored by a similar efficiency in peripheral blood mononuclear cells (PBMCs) (Gkotzamanidou M et al, Leukemia and BJC, 2013 and 2014). Our hypothesis is that our PK model will predict > 85% of interpatient variability, AUC achieved as well as measurements of DNA damage ex vivo will correlate with mucositis and duration of neutropenia, and we will be able to create an integrated model to personalize melphalan dosing to maximize myeloma cell killing while minimizing toxicities. Methods: We enrolled 146 patients on a prospective trial using a block randomization scheme based on fat-free mass, calculated creatinine clearance, and hemoglobin as known factors affecting melphalan disposition. Plasma was collected from all patients to assess melphalan pharmacokinetics, and PBMCs were collected to assess ex vivosensitivity to melphalan therapy, including p53 gene expression and WST-1-based cytotoxicity. DNA from PBMCs was collected to assess the presence of SLC7A5 polymorphisms associated with melphalan-induced enteritis (Giglia JL et al, BBMT, 2014). PK/PD modeling and identification of covariates contributing to observed variability in melphalan disposition and outcomes is being achieved using a nonlinear mixed effects approach Results: Melphalan has been quantified in samples from 119 patients, and this data was used in population modeling. Estimated PK parameters (CV%) for the model were ӨCL=0.445 L/min (33.6%), ӨV1=19.4 L (36.6%), ӨQ=0.392 L/min (34.5%) and ӨV2=17.4 L (36.9%). Creatinine clearance (normalized to 70 kg), hematocrit, fat free mass on CL, sex on V1, and body surface area on Q were chosen for the final covariate model. IC50value (CV%=31.2%), and the baseline of p53 mRNA level (47.3%) and viability (20.1%) in donor’s PBMCs were variable. Response of p53 mRNA expression and viability in donor’s PBMCs were dose-dependent. PD modeling, and covariate and SCLA5 polymorphism analyses are underway. Conclusions: This study represents the largest and most comprehensive for identifying variables associated with melphalan pharmacokinetics and pharmacologic activity. An integrated PK/PD model that incorporates PK/PD and other predictive factors form the basis of a prospective randomized clinical trial to validate that the model reduces toxicities while prolonging PFS. Disclosures Hofmeister: Celgene: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Millenium: Honoraria, Research Funding; ARNO therapeutics: Research Funding; Onyx: Membership on an entity's Board of Directors or advisory committees.