Perez-Castello, I(1); Mangas-Sanjuan, V(2,3); Gonzalez-Alvarez, I(2); Marco-Garbayo, JL(1); Troconiz, I(3); Bermejo, M (2)
(1) Hospital of Gandia; (2) Pharmacokinetics and Pharmaceutical Technology Area, Miguel Hernandez University; (3) Pharmacometrics and Systems Pharmacology. University of Navarra
Introduction: Lithium is an antidepressant used as primary treatment for the prevention of episode recurrences in bipolar disorder, acute treatment of mania and to a lesser extent depression. The main objective of this work is to predict the individual compliance per treatment cycle of the population using a Markov model [1].
Methods: 96 psychiatric patients were enrolled in this study. Lithium carbonate was administrated to all patients at different dose levels (200, 300, 400, 600 and 800 mg) and administration intervals (8, 12 and 24 h). Patients received several treatment cycles and one plasma concentration measurement for each patient was obtained always before starting next cycle (pre-dose) at steady state. Lithium concentrations were categorized below (0), within (1) and above (2) the therapeutic interval. Experimental data were fitted using non-linear mixed-effects modelling implemented in NONMEM 7.2. Different approaches were implemented in order to capture the concentration profiles observed: (1) IOV on bioavailability dose fraction (F1), (2) Markov model on F1 based on the previous categorical lithium state. Model selection was based on the lowest and significant OFV, final parameter estimates and RSE. Model evaluation of the number of (i) transitions per cycle, (ii) transitions per individual or (iii) total number of transitions were performed.
Results: Plasma observations were described using a two-compartment model. Creatinine clearance (CrCl) was selected as significant covariate on typical clearance parameter with a power relationship. The empirical model including IOV on F1 allows for a adequate description of the data. The number of predicted transitions was greater than experimentally observed. A Markov model including the drug compliance on F1 was successfully applied. Markov model predicted more precisely the number of transitions, number of transitions per cycle and number of transitions per individual compared to IOV model.
Conclusions: The final model was able to characterize the number of individuals/observations out of the therapeutic interval with more precision compared to the other approaches proposed.
Reference:
[1] Troconiz IF, et al. Modelling overdispersion and Markovian features in count data. JPKPD. 2009.
Reference: PAGE 25 (2016) Abstr 5991 [www.page-meeting.org/?abstract=5991]
Poster: Drug/Disease modeling - CNS