Development of population based approaches to describe the complex pharmacokinetics of simvastatin in different individuals. Bridging the gap between population and physiologically based pharmacokinetic modelling.
Nikolaos Tsamandouras (1), Gemma Dickinson (2), Yingying Guo (2), Stephen Hall (2), Amin Rostami-Hodjegan (1,3), Aleksandra Galetin (1) and Leon Aarons (1)
(1) Centre for Applied Pharmacokinetic Research, University of Manchester, UK, (2) Eli Lilly and Company, Indianapolis, IN, USA, (3) Simcyp Ltd, Sheffield, UK
Objectives: Simvastatin (SV) is a prodrug with complex pharmacokinetics (PK) due to the interconversion between the parent drug and its main active metabolite, simvastatin acid (SVA) [1, 2]. Both SV and SVA are subject to drug-drug interactions (DDIs)  and are affected by genetic variation in several enzyme/transporter proteins relevant for their disposition [4-7]. The most serious adverse effect of SV is related to skeletal muscle and it can range from myalgia to myopathy and ultimately to the rare but potentially life-threatening rhabdomyolysis [4, 8]. Simvastatin-induced myotoxicity is a concentration-dependent adverse effect which is at least partly of a pharmacokinetic origin . However, our understanding regarding the complex SV/SVA PK and especially their variability in the population is very limited.
This work aims to develop a joint population SV/SVA pharmacokinetic model that incorporates the effects of multiple genetic polymorphisms and clinical/demographic characteristics. Besides that, it also aims to provide a practical example of a covariate analysis where several genetic polymorphisms are tested and the degree of correlation between them through linkage disequilibrium has to be considered [10, 11].
In parallel, this work aims to develop an alternative much more complex but mechanistic population SV/SVA pharmacokinetic model. Such a model has the advantage that it can provide predictions not only for concentrations in plasma, but also in the clinically relevant tissues of muscle (toxicity) and liver (efficacy), while also it allows extrapolation outside the studied conditions. We have recently discussed the various methodological issues related with parameter estimation in such complex models . Therefore, in this work we aim to also explore further methodological aspects such as: incorporation of population variability in system-related parameters which are subject to certain physiological constraints; parameter estimation with the aid of prior distributions; and application of different estimation algorithms.
Methods: SV/SVA plasma concentrations, demographic/clinical data and genotypes for 18 genetic variants were collected from 74 individuals and analyzed with NONMEM 7.2 in order to develop the population pharmacokinetic-pharmacogenetic model . Several structural empirical parent-metabolite pharmacokinetic models were evaluated. Covariate selection was performed using a stepwise forward inclusion – backward elimination process that also examines the degree of linkage disequilibrium between correlated SNPs upon inclusion in each step. Finally the developed model was also used to identify individuals harbouring genetic and other clinical/demographic risk factor combinations that can be associated with clinically important increases on SV/SVA plasma exposure.
For the development of the complex mechanistic pharmacokinetic model, rich SV/SVA plasma concentration data (34 healthy volunteers) were simultaneously analysed with NONMEM 7.2. Additional rich plasma concentration data (28 healthy volunteers) were used for external validation of the model . The implemented mechanistic model has a complex compartmental structure (presented in an earlier version of this work ) that allows interconversion between SV and SVA in different tissues. Prior information for model parameters and (when available) their variability was extracted from physiology, in vitro experiments and in silico methods in order to construct appropriate prior distributions. Variability in system related parameters was incorporated not only by linking them with several well-defined physiological covariates (e.g. body weight), but also by the use of random inter-individual variability terms. Therefore, a generalisation of the logit-normal distribution was applied to constrain some of the system parameters within their physiological range (e.g. gastric residence time) and a multivariate logistic normal distribution was applied on compositional parameters for which the physiological constraints apply on their joint distribution (e.g. fractions of cardiac output that reaches each tissue). The prior functionality in NONMEM [16, 17] was used to integrate prior information with the information from the clinical data and obtain maximum a posteriori (MAP) parameter estimates under the first order conditional estimation method with interaction (FOCE-I) and importance sampling assisted by mode a posteriori (IMPMAP) methods . The developed model was finally employed to simulate concentration profiles in plasma, liver and muscle and investigate the impact of polymorphisms and a range of DDIs (clarithromycin, diltiazem, erythromycin, itraconazole).
Results: The empirical model that best described the data included a two- and a one-compartment disposition model for SV and SVA respectively. Age, weight, Japanese ethnicity and 7 genetic polymorphisms, rs4149056 (SLCO1B1), rs776746 (CYP3A5), rs12422149 (SLCO2B1), rs2231142 (ABCG2), rs4148162 (ABCG2), rs4253728 (PPARA) and rs35599367 (CYP3A4), were identified to significantly affect model parameters. The present work highlighted specific characteristics associated with altered SV/SVA PK and subsequently myopathy cases which cannot be solely attributed to a single genotype. Extensive results of this pharmacogenetic analysis are reported in .
The complex mechanistic model provided a good fit to the plasma SV/SVA concentration data and their variability. The incorporation of priors allowed precise estimation of model parameters without neglecting the random variability in key system and drug related parameters and the uncertainty  on the results of in vitro experiments or in silico predictions that are used to inform them. The FOCE-I and IMPMAP estimates were comparable, with the latter method (parallelised) showing significantly reduced computation times for convergence, similarly to previous studies with complex mechanistic models [20-22]. Simulations with the developed model using reduced SVA hepatic uptake (mimicking a SLCO1B1 polymorphism) recovered the reported increase in exposure of SVA in plasma ; comparable increased fold-exposure was simulated in the muscle, consistent with the risk of toxicity in this tissue; whereas impact on liver exposure was minimal. In addition the developed model successfully predicted the effect of a range of DDIs on the PK of both SV and SVA.
Conclusions: The population based approaches developed in this work overall provide further insight into the PK of simvastatin and the related population variability. These approaches can be of significant clinical application due to the widespread use of simvastatin and the clinical burden of muscle toxicity. In addition the present work illustrates the feasibility of combining traditional PBPK with population approaches and information from clinical data in order to develop mechanistically sound population models with clinical relevance.
 Vickers S, Duncan CA, Chen IW, Rosegay A & Duggan DE, Metabolic disposition studies on simvastatin, a cholesterol-lowering prodrug. Drug Metab Dispos, 1990. 18(2): p. 138-145.
 Prueksaritanont T, et al., Interconversion pharmacokinetics of simvastatin and its hydroxy acid in dogs: Effects of gemfibrozil. Pharm Res, 2005. 22(7): p. 1101-1109.
 Elsby R, Hilgendorf C & Fenner K, Understanding the critical disposition pathways of statins to assess drug-drug interaction risk during drug development: It's not just about OATP1B1. Clin Pharmacol Ther, 2012. 92(5): p. 584-598.
 Wilke RA, et al., The clinical pharmacogenomics implementation consortium: CPIC guideline for SLCO1B1 and simvastatin-induced myopathy. Clin Pharmacol Ther, 2012. 92(1): p. 112-117.
 Pasanen MK, Neuvonen M, Neuvonen PJ & Niemi M, SLCO1B1 polymorphism markedly affects the pharmacokinetics of simvastatin acid. Pharmacogenet Genomics, 2006. 16(12): p. 873-9.
 Kim K-A, Park P-W, Lee O-J, Kang D-K & Park J-Y, Effect of polymorphic CYP3A5 genotype on the single-dose simvastatin pharmacokinetics in healthy subjects. J Clin Pharmacol, 2007. 47(1): p. 87-93.
 Niemi M, Transporter pharmacogenetics and statin toxicity. Clin Pharmacol Ther, 2009. 87(1): p. 130-133.
 Thompson PD, Clarkson P & Karas RH, Statin-associated myopathy. JAMA, 2003. 289(13): p. 1681-90.
 Niemi M, Pasanen MK & Neuvonen PJ, Organic anion transporting polypeptide 1B1: A genetically polymorphic transporter of major importance for hepatic drug uptake. Pharmacol Rev, 2011. 63(1): p. 157-181.
 Lehr T, Schaefer H-G & Staab A, Integration of high-throughput genotyping data into pharmacometric analyses using nonlinear mixed effects modeling. Pharmacogenet Genomics, 2010. 20(7): p. 442-450.
 Bertrand J & Balding DJ, Multiple single nucleotide polymorphism analysis using penalized regression in nonlinear mixed-effect pharmacokinetic models. Pharmacogenet Genomics, 2013. 23(3): p. 167-174.
 Tsamandouras N, Rostami-Hodjegan A & Aarons L, Combining the “bottom-up” and “top-down” approaches in pharmacokinetic modelling: Fitting PBPK models to observed clinical data. Br J Clin Pharmacol. Accepted, doi: 10.1111/bcp.12234.
 Tsamandouras N, et al., Identification of multiple polymorphisms effect on the pharmacokinetics of simvastatin and simvastatin acid using a population modeling approach. Clin Pharmacol Ther, 2014. Accepted.
 Polli JW, et al., Evaluation of drug interactions of GSK1292263 (a GPR119 agonist) with statins: From in vitro data to clinical study design. Xenobiotica, 2013. 43(6): p. 498-508.
 Tsamandouras N, Galetin A, Dickinson G, Hall S, Rostami-Hodjegan A & Aarons L, A mechanistic population pharmacokinetic model for simvastatin and its active metabolite simvastatin acid. PAGE 22 (2013) Abstr 2739 [www.page-meeting.org/?abstract=2739].
 Gisleskog PO, Karlsson MO & Beal SL, Use of prior information to stabilize a population data analysis. J Pharmacokinet Pharmacodyn, 2002. 29(5): p. 473-505.
 Langdon G, Gueorguieva I, Aarons L & Karlsson M, Linking preclinical and clinical whole-body physiologically based pharmacokinetic models with prior distributions in NONMEM. Eur J Clin Pharmacol, 2007. 63(5): p. 485-498.
 Bauer JR, NONMEM users guide: Introduction to NONMEM 7.2.0. 2011, Ellicot City, Maryland: ICON Development Solutions.
 Gertz M, Tsamandouras N, Säll C, Houston JB & Galetin A, Reduced physiologically-based pharmacokinetic model of repaglinide: Impact of OATP1B1 and CYP2C8 genotype and source of in vitro data on the prediction of drug-drug interaction risk. Pharm Res, 2014. Accepted (DOI: 10.1007/s11095-014-1333-3).
 Gibiansky L, Gibiansky E & Bauer R, Comparison of nonmem 7.2 estimation methods and parallel processing efficiency on a target-mediated drug disposition model. J Pharmacokinet Pharmacodyn, 2012. 39(1): p. 17-35.
 Bauer R, Guzy S & Ng C, A survey of population analysis methods and software for complex pharmacokinetic and pharmacodynamic models with examples. The AAPS Journal, 2007. 9(1): p. E60-E83.
 Bulitta J & Landersdorfer C, Performance and robustness of the monte carlo importance sampling algorithm using parallelized s-adapt for basic and complex mechanistic models. The AAPS journal, 2011. 13(2): p. 212-226.