Beatriz Guglieri-López (1), Alejandro Pérez-Pitarch (1), Mónica Climente-Martà (2), Matilde Merino-Sanjuán (1).
(1) Department of Pharmacy and Pharmaceutical technology, Faculty of Pharmacy, University of Valencia. Valencia. Spain. (2) Pharmacy Department. Doctor Peset University Hospital. Valencia. Spain. (3) Pharmacy Department . University Clinical Hospital of Valencia. Spain.
Objective: To model the pharmacodynamic M-protein response induced by lenalidomide given in patients with multiple myeloma (MM).
Methods: Data were available from 39 MM patients with measurable M-protein levels who started treatment with lenalidomide between March 2009 and October 2014 in two Spanish Hospitals. A simplified tumour growth inhibition (TGI) model based on a previous published model [1] was used to estimate response metrics based on time profiles of M-protein (taken as a marker of tumour size) after lenalidomide administration. A parameter representing non-monoclonal component (NMC) was added to the model in order to restrict the tumour growth and the effect of lenalidomide to myeloma cells.
dMprot/dt = KL · (Mprot(t)-NMC) – KD(t) · (Mprot(t)-NMC);
KD(t)=KD0 · e–λ·t
KL: tumour growth rate; KD: drug effect; λ: drug resistance.
Model parameters were estimated using non-linear mixed-effects modelling implemented in NONMEM V7.3.0 [2]. Relationship between these TGI metrics and covariates was assessed using Stepwise Covariate Model building tool of PsN v4.2.0. The performance of the model was evaluated using a visual predictive check (VPC) based on 1000 simulated replicates of the development dataset (30 patients) from week 8 onwards. The bootstrap resampling technique was also used for internal validation. External validation was conducted by assessing the ability of the population model to predict M-protein levels from week 8 onwards in a separate group of 9 patients using the normalized prediction discrepancy distribution error (NPDE) add-on package for R [3].
Results: The model was composed of sub-models for tumour growth dynamics, drug effect and drug resistance. The decrease of M-protein levels to normal range at the end of cycle 2 (week 8) was linearly correlated to drug effect and drug resistance. The M-protein level at week 8 was correlated to NMC. The model indicated good prediction and a lack of bias in goodness-of-fit plots and VPC. The external validation showed that NPDE were not different from a normal distribution (global adjusted p-value=0.987).
Conclusion: Internal and external validation techniques demonstrate that this model can be used to predict M-protein profile from week 8 onwards in multiple myeloma patients treated by lenalidomide. The described model could be part of a PK/PD model framework to simulate expected survival taking into account drug exposure.
References:
[1] Chanu P, Claret L, Marchand M, Losic N, Puchalski TA, Bruno R. Population pharmacokinetic/pharmacodynamic models to support dose selection of daratumumab in multiple myeloma patients. PAGE. Abstracts of the Annual Meeting of the Population Approach Group in Europe. ISSN 1871-6032. PAGE 23 (2014) Abstr 3281 [www.page-meeting.org/?abstract=3281]. [2] Beal SL, Sheiner LB, Boeckmann AJ & Bauer RJ (Eds.) NONMEM Users Guides. 1989-2011. Icon Development Solutions, Ellicott City, Maryland, USA.
[3] Comets E, Brendel K, Mentre F. Computing normalised prediction distribution errors to evaluate nonlinear mixed-effect models: the npde add-on package for R. Comput Methods Programs Biomed. 2008;90(2):154-166.
Reference: PAGE 24 () Abstr 3317 [www.page-meeting.org/?abstract=3317]
Poster: Drug/Disease modeling - Oncology