The Pharmacokinetics UK 2011 Meeting Programme and Abstract

Welcome to the Pharmacokinetics UK 2011 Meeting

Wednesday 9th November – Friday 11th November

Radisson Blu Hotel Frankland Lane Durham DH1 5TA

Programme and Abstract Book

Wednesday 9th November

12:00 Arrival & lunch

Welcome and session 1: Pharmacokinetic issues in the field of Oncology 14:00 Welcome: Steve Toon 14:05 Introduction to the First Session: Peter Milligan and Alan Boddy 14:10 Alan Boddy, Newcastle University Pharmacokinetics and metabolism in early phase clinical trials 14:45 Andrew Stone, AstraZeneca Statistical modelling of clinical data before investing in pivotal trials 15.20 Emma Hansson, Uppsala University PKPD Modeling of VEGF, sVEGFR-2, sVEGFR-3 and sKIT as Biomarkers of Tumor Response, Adverse Events and Survival Following Sunitinib Treatment in GIST 15:55 Coffee break 16.10 Clinton Stewart, St. Jude Children‟s Research Hospital Developmental pharmacology of anticancer drugs in infants and young children with brain tumors: it’s not easy being little 16:45 Gareth Veal, Newcastle University Pharmacokinetic studies in children with cancer – a UK perspective 17:20 Session close 18:30 Poster session & free bar (Sponsored by Pharsight) 20:00 Dinner

Thursday 10th November

Session 2: Clinical Pharmacokinetics and the Paediatric Population 09:00 Introduction to the Second Session: Alison Thomson and Geoff Tucker 09:05 Hussain Mulla, University Hospitals of Leicester Facing the challenges of better medicines for children: A new use for Dried Blood Spots 09:35 Joseph Standing, University College London Inferences from Small First-in-Child Studies 10:05 Leon Aarons, University of Manchester Population Pharmacokinetic Analysis of Ropivacaine and its Metabolite PPX from Pooled Data in Neonates, Infants and Children 10:35 Coffee break 10.55 Lutz Harnisch, Pfizer Model based assessment of Revatio in pediatric pulmonary arterial hypertension (PAH): a viable route to dose recommendation through an integration of PK, biomarker, and clinical endpoints 11.25 Catherijne Knibbe, LACDR, Leiden University Dosage individualisation in children based on predictive population PK-PD models 11:55 Open discussion 12:10 Session close 12:10 Lunch

Session 3: Systems Biology-a new world? 13:30 Introduction to the Third Session: Terry Shepard and Steve Toon 13:35 Piet van der Graaf, Pfizer (How) Can Systems Pharmacology help to improve Phase 2 success? 14:20 David Gavaghan, University of Oxford Developing and applying safety-critical software for systems-level modelling 15.05 Coffee break 15.25 Darrell Abernethy, US Food and Drug Administration Pharmacological Mechanism Based Drug Safety Prediction 16.10 Open discussion 16.25 Session close The Peter Coates Lecture, with an introduction by Steve Toon 16.45 Lawrence Lesko, US Food and Drug Administration Pharmacometric Solutions: A Question-Based Review 20:00 PKUK Banquet

Friday 11th November

Session 4: Open Session 09:00 Introduction to the Open Session: Leon Aarons and Amin Rostami-Hodjegan 09:05 Marc Lavielle, INRIA Saclay A first prototype of the new Clinical Trial Simulator developed within the DDMoRe project 09.25 Zinnia Parra Guillén, University of Navarra Target Mediated Drug Disposition Model to Describe the Expression and Kinetics of IL12 and IFN in Gene Therapy 09.45 Thomas Grandjean, University of Warwick Experimental and Mathematical Analysis of Pitavastatin Hepatic Uptake Across Species 10.05 Kayode Ogungbenro, University of Manchester Semi-mechanistic Modelling of Double Peaks in Pharmacokinetics: LDOPA case study 10:25 Coffee break 11:00 Marco Siccardi, University of Liverpool Pharmacokinetic and Pharmacodynamic Analysis of Efavirenz Dose Reduction Using a Physiologically-Based Dynamic Model 11:20 Ivelina Gueorguieva, Eli Lilly Defining a therapeutic window for the novel TGF-β inhibitor LY2157299 based on a pharmacokinetic/pharmacodynamic (PK/PD) model 11.40 Ashley Strougo, Astellas Pharma “First dose in children”: evaluation of clearance scaling approaches in CYP3A4-metabolized drugs 12:00 Final conclusions, closing remarks and lunch

Speaker Abstracts

Session 1: Pharmacokinetic issues in the field of Oncology

1. Pharmacokinetics and metabolism in early phase clinical trials

Alan Boddy

NICR, Newcastle University

Phase I studies in cancer are performed in patients rather than normal volunteers. Given the narrow therapeutic index for most chemotherapeutic agents, pharmacokinetic investigations are usually performed in “real-time” in such Phase I studies, and may be used to guide dose- escalation, relative to preclinical PK data. Depending on the relative pharmacology of parent drug and metabolites, the early identification and quantitation of metabolites may also form an important part of pharmacological investigations in Phase I studies.

2. Statistical modelling of clinical data before investing in pivotal trials

Andrew Stone

Oncology Therapeutic Area Statistical Expert, AstraZeneca, Alderley Park, Macclesfield, Cheshire, SK10 4TF, UK

The decision to enter into a pivotal trial represents a major milestone in drug development and involves a substantial increase in investment; this decision is also based on many levels of uncertainty. In order to try and quantify the probability that the pivotal trial will be successful, outcomes are modelled carefully. This presentation will describe approaches used in oncology, many of which are applicable to other therapeutic areas. The concept of assurance or unconditional power will be described which has its basis grounded in a bayesian approach together with how an extension when an intermediate endpoint is used to predict a longer term outcome. Finally, the expected shrinkage of treatment effect between phase II and III will be described; an expected phenomenon even if the phase II trial is perfectly conducted.

3. PKPD Modeling of VEGF, sVEGFR-2, sVEGFR-3 and sKIT as Biomarkers of Tumor Response, Adverse Events and Survival Following Sunitinib Treatment in GIST

Emma K. Hansson1, Guangli Ma1,2, Michael Amantea2, Jonathan French2, Peter A. Milligan2, Mats O. Karlsson1, Lena E. Friberg1

1Department of Pharmaceutical Biosciences, Uppsala University, Uppsala, Sweden, 2Pfizer Global Research and Development

Background: A need for new ways of identifying an optimal dose and evaluating response has evolved with the introduction of the targeted therapies in the area of oncology. Identification of predictive biomarkers would improve measures of treatment activity and enable dose optimization. Biomarkers could also serve as an indicator of safety issues and to minimize the occurrence of adverse events by improving clinical interventions.

Aim: To characterize biomarker, tumor size, adverse events and survival relationships following sunitinib treatment in imatinib resistant gastro-intestinal stromal tumors (GIST) with focus on the potential biomarkers VEGF, sVEGFR-2, sVEGFR-3, and sKIT.

Methods: Data on the four biomarkers, tumor size, adverse events (fatigue, hand-foot syndrome (HFS)) and survival were available from 303 patients across four clinical trials following up to 85 weeks of treatment with sunitinib and/or placebo. PK/PD models were developed to characterize the change in biomarkers, tumor size and for the probability of HFS and fatigue following treatment. The overall survival data was described by a parametric time-to event model. Model predicted time-courses of the variables were evaluated as predictors of tumor size, adverse events and overall survival.

Results: The biomarker modulations following sunitinib treatment were described by indirect response models and the longitudinal tumor size data by a previously developed tumor growth inhibition model [1]. The predicted time-courses of the relative changes from baseline for sKIT (most significant) and sVEGFR-3, as well as AUCss, were included as significant descriptors of the change in tumor size. Proportional odds models with a first order Markov element taking into account the previous grade in the probability of a transition between different severity grades were used to describe the probability and severity of HFS and fatigue. The time course of VEGFR-3 described the probability and severity of experiencing HFS and fatigue better than exposure. A parametric time-to-event model with a Weibull distribution described the underlying hazard. The predicted relative change in sVEGFR-3 over time, and tumor size at start of treatment best described survival time and when included in the model, none of the other variables, including tumor response, were of significance.

Conclusion: A modeling framework was proposed linking longitudinal biomarker data with overall survival using PKPD models. sVEGFR-3 was found to be the most promising variable for predicting overall survival and adverse events following sunitinib treatment in GIST.

References: [1] Claret L. et al. JCO. 2010:27, 4103-4108

4. Developmental pharmacology of anticancer drugs in infants and young children with brain tumors: it‟s not easy being little

Clinton F. Stewart, John C. Panetta, Amar Gajjar

St. Jude Children’s Research Hospital, Memphis, TN

Background: Although anti-cancer drugs are widely used in infants and young children with malignant brain tumors, the dosing of these toxic drugs is often scaled based on body size (e.g., BSA or body weight) or arbitrary age cut-offs. This dosing approach does not take into account the many developmental changes that occur in infants and young children. These changes, which include hepatic and renal maturation, may affect the disposition of anticancer drugs, leading to more severe toxicity, particularly myelosuppression, which puts these children at risk of life-threatening infections. Yet the pharmacokinetics and toxicity of many anticancer drugs have not been adequately studied in infants and young children less than 3 years old.

Aim: Thus, we propose to perform clinical pharmacokinetic (PK), pharmacogenetic, and pharmacodynamic (PD) studies of methotrexate (MTX), and cyclophosphamide (CTX) as part of a clinical protocol that treats infants and young children with malignant brain tumors.

Methods: CTX was administered during the first course of treatment (1.5 g/m2 infused over 1 h), and serial samples were collected prior to infusion, at the end of infusion, and 3, 6, and 24 h after the end of infusion. CTX and CEPM were quantitated in plasma with a validated LC-MS/MS method and 4OH-CTX was stabilized by addition of phenylhydrazine to whole blood and quantitated with a validated LC-MS/MS method. A nonlinear mixed-effects model was fit to the data using NONMEM software and covariates were added to the model based on reduction in objective function value and improvement in diagnostic plots. For MTX, patients received four 28-day courses of induction therapy, and HD-MTX (day 1) was given as a 5 gm/m2 infusion over 24 hrs, with PK sampling prior to the infusion and at 6, 23, 42 and 66 h after the start of infusion. Leucovorin was administered starting at 42 hrs and dosed every 6 h until the serum MTX concentration was < 0.1 μM. Patients also received vincristine (day 8 and 15), cisplatin (day 8), and cyclophosphamide (day 9). PK parameters for a two-compartment model were estimated using a Bayesian approach with ADAPT II software and covariates were evaluated using linear mixed effects models.

Results: Pharmacokinetic studies for CTX were performed with 28 patients. The median (range) age was 17.3 months (0.87 – 35.9 months) and median (range) BSA was 0.49 m2 (0.24 – 0.71 m2). A model with one compartment for CTX and each of its metabolites adequately described the data. BSA normalization of the dose reduced the inter-individual variability (IIV) in the volume of distribution of CTX (VCTX) from 32% to 13% and reduced the IIV in the elimination of 4OH-CTX through non-CEPM pathways (k4OH-CTX) from 60% to 38%. The covariate analysis showed that the BSA-normalized clearance of CTX to 4OH-CTX (CLCTX) increased slightly with age, whereas k4OH-CTX increased rapidly with age. The net effect was that exposure to 4OH-CTX decreased with age in a nonlinear fashion. For MTX, 53 patients receiving 176 courses were included in the analysis. The median (range) age was 17.6 months (0.23 – 35.7 months) and BSA was 0.48 m2 (0.22-0.73 m2). The population estimate for MTX clearance (CL) was 91 ml/min/m2. CL increased with age, explaining 39% of interindividual variability (IIV) (p<0.001). Patients at 3 months, 1 yr, 2 yrs, and 3 yrs had an estimated MTX CL of 72, 83, 100 and 120 ml/min/m2, respectively. Calculated glomerular filtration rate (GFR) explained an additional 10% of IIV (p=0.005). Typical MTX CL was higher in the first course (100 ml/min/m2) compared to subsequent courses (88 ml/min/m2), explaining an additional 9.7% of interoccasion variability (p<0.001). HD-MTX was well tolerated; grade 3 mucositis occurred in 4 patients (8%) and no patients experienced grade 4 mucositis. Overall, grade 3 or 4 toxicities occurred in 19 patients (36%), but this may be attributable to the combination of chemotherapy administered during induction therapy.

Conclusion: The preliminary results of this study demonstrate the importance of studying the pharmacokinetics of anticancer drugs in infants and young children, and will provide the basis for the rational design of dosing strategies for infants and young children with brain tumors.

5. Pharmacokinetic studies in children with cancer – a UK perspective

Gareth J Veal

Northern Institute for Cancer Research, Medical School, Newcastle University, Newcastle upon Tyne NE2 4HH

Despite the fact that anticancer drugs have been used successfully to treat childhood malignancies for several decades, relatively limited information is available from pharmacokinetic or pharmacodynamic studies carried out with many of these chemotherapeutics in paediatric patient populations.

Over the past 10 years our group has conducted a portfolio of studies designed to determine key factors which influence the efficacy of currently used and newly-developed drugs or drug formulations, focusing on how this information can potentially be used to optimise the treatment of children with cancer. While therapeutic drug monitoring approaches have been established to guide treatment for drugs such as carboplatin, this type of adaptive dosing is not commonplace in a paediatric oncology setting for the majority of drugs. For many chemotherapeutics we are developing an increased understanding of factors that affect drug disposition which may prove to be clinically beneficial in the future.

Fundamental differences in drug disposition are not only observed between children and adults, but can also be present within paediatric populations between infants and adolescents. For many drugs such differences have the potential to impact on tumour response and the incidence of adverse effects frequently associated with treatment. Challenges are also posed in terms of defining the optimal dosing of subgroups of patients where unpredictable pharmacokinetic profiles may be anticipated, including those suffering from obesity or malnutrition.

It is important to improve our knowledge in this area by coordinating well-planned clinical pharmacology studies in paediatric patient populations. Advances in limited sampling strategies, population pharmacokinetic modeling, and the development of assays that use minimal sample volume in order to maximize patient recruitment will expedite such studies. It is hoped that this will facilitate the development of more appropriate approaches to chemotherapy dosing in children with cancer.

Session 2: Clinical Pharmacokinetics and the Paediatric Population

6. Facing the challenges of better medicines for children: A new use for Dried Blood Spots

Hussain Mulla

Department of Pharmacy, Glenfield Hospital, University Hospitals of Leicester, England.

‘Better medicines for children‟ is the goal that instigated the new paediatric EU legislation, the „Best Pharmaceuticals for Children Act and the Paediatric Research Equity Act‟ in the US and the foundation of the Medicines for Children Research Network (MCRN) in the UK. The aim of EU and US legislation is to improve the evidence base for drugs prescribed to children, whereas networks such as the MCRN aim to facilitate appropriate clinical studies of these drugs.

These welcome initiatives provide a framework for generating information on medicines used in children. Some issues, however, are beyond the scope of this framework. A key but unresolved issue is how best to perform essential pharmacokinetic (PK) studies in children. It is widely accepted that there is a paucity of PK data for children‟s medicines and this information gap is cited as one of the major reasons why the evidence base for medicines use in infants, children and young people is very poor.

Obtaining PK data in young children is difficult. Technical and ethical hurdles relating to repeated blood sampling and the volume of blood remain major obstacles to parents agreeing to take part in studies, particularly those involving preterm infants. Furthermore, many PK studies in children are likely to be performed in „multiples of healthcare settings‟ rather than a single, bespoke clinical trials unit. There remains an urgent need for a method of collecting blood samples that can be used to quantify drug levels robustly in the conduct of a PK-PD study undertaken in a population of any age and from a variety of clinical settings.

Quantification of analytes from blood spotted onto cards has a long standing pedigree in screening for inherited diseases affecting newborn infants. More recently, there has been resurgence in adapting and applying dried blood spots (DBS) methods using micro-volumes of blood (10-20 l) to quantify parent drug, metabolites and biomarkers. The advantages of DBS over current “wet-sample” analysis techniques in the arena of pre-clinical drug development are well recognised. One significant benefit of studies utilising DBS is that fewer animals are needed in comparison to a study based on wet-samples. Reductions in the complexity of data sets and costs are welcome by-products of using fewer animals in drug studies. DBS methodology also has significant potential to impact favourably and significantly in PK-PD studies for children. This presentation reviews and highlights some of the issues that particularly relate to performing DBS-based PK studies in children.

References: Patel P, Mulla H, Tanna S, Pandya H. Facilitating PK studies in Children: A new use for Dried Blood Spots. Arch Dis Child 2010. 95: 484-487 Pandya H, Spooner N, Mulla H. Dried blood spots, pharmacokinetic studies and better medicines for children. Bioanalysis 2011: 3(7): 779-786

7. Inferences from Small First-in-Child Studies

Joseph F Standing

MRC Methodology Fellow Department of Infectious Diseases and Microbiology University College London, Institute of Child Health

In rare diseases or where a drug has previously not been used in children with a certain condition, small PK studies are frequently conducted. The purpose of such studies is often to plan dosing for further pivotal trials, or to provide doing information in situations where further data will be difficult to obtain. Using the data collected in such studies, along with appropriate scaling for size and maturation, two recent examples are presented to show how non-linear mixed effects modelling can be applied for such purposes.

In the first example, recombinant human insulin-like growth factor IGF-1 was administered to 8 children with Crohn‟s disease on two occasions. Period 1 was a single dose study, and Period 2 consisted of multiple dosing; rich PK sampling was performed on both occasions. IGF-1 is a naturally occurring substrate involved in growth hormone (GH) response and children with Crohn‟s disease often have poor growth which is possibly related to low circulating IGF-1. This study aimed to characterise the levels achieved and develop dose guidelines for a future large-scale study. A turnover model was used to describe circulating IGF-1 following a subcutaneous dose. Disease state was found to significantly influence endogenous synthesis rate, and simulations were performed to ascertain the dose required to normalise circulating levels in a long-term study of growth.

In the second example, oseltamivir was administered for treatment or prophylaxis of influenza on a neonatal intensive care unit, and rich pharmacokinetic samples drawn during one of the routine doses in 9 neonates. Samples were assayed for oseltamivir, which is a prodrug, and the active metabolite oseltamivir carboxylate. There are currently no sufficient published neonatal PK data for defining oseltamivir dosing, and due to the fact the drug is only given during influenza outbreaks, planning and executing studies from routine data is difficult. Conversion of oseltamivir to its principle metabolite is catalysed by the enzyme human carboxylesterase 1 (HCE1), the expression of which is reduced in foetal, neonatal and infant samples. A maturation function was fitted to published HCE1 expression data, and this along with renal maturation and size scaling was used in the development of a parent/metabolite model. The model was then used to produce dose simulations based on adult dosing and in vitro concentration/response values.

Acknowledgements The two data examples were analysed in collaboration with Dr Arati Rao & Prof Ian Sanderson at St Bart‟s Medical School, and Prof Maria Tsolia and colleagues at the P. & A. Kyriakou Children‟s Hospital, Athens, Greece.

8. Population Pharmacokinetic Analysis of Ropivacaine and its Metabolite PPX from Pooled Data in Neonates, Infants and Children

L. Aarons(1), B. Sadler(2), M. Pitsiu(3), J. Sjövall(4), J. Henriksson(5)

(1)School of Pharmacy and Pharmaceutical Sciences, University of Manchester, UK; (2)Icon Development Solutions, Ellicott City, MD, USA; (3)Icon Development Solutions, Manchester UK; (4)Clinical Pharmacology & DMPK, AstraZeneca R&D Södertalje, Sweden; (5)Biostatistics, Clinical R&D, AstraZeneca, Södertälje, Sweden

Objectives: The aim of the present study was to characterize ropivacaine and PPX pharmacokinetics and factors affecting them in paediatric anaesthesia.

Methods: Population pharmacokinetics of ropivacaine and its active metabolite PPX were estimated following single and continuous ropivacaine blocks in 192 patients aged 0-12 years from six pooled published studies. Unbound and total ropivacaine and PPX plasma concentration and PPX urinary excretion data were used for non-linear mixed effects modelling by NONMEM. Covariates included age, body weight, gender, ethnic origin, ASA, site and method of administration and total dose.

Results: One-compartment first-order pharmacokinetic models incorporating linear binding of ropivacaine and PPX to α1-acid glycoprotein were used. After accounting for the effect of body weight, clearance of unbound ropivacaine and PPX reached 41% and 89% of their mature values, respectively, at the age of 6 months. Ropivacaine half-life decreased with age from 13 h in the newborn to 3 h beyond 1 year. PPX half-life differed from 19 h in the newborn to 8 - 11 h between 1 and 12 months to 17 h after 1 year. Simulations indicate that for a single caudal block the recommended dose could be increased by a factor of 2.9 (0 to 1 month group) and 6.3 (1 to 12 year group) before the unbound plasma concentrations would cross the threshold for systemic toxicity. Corresponding factors for continuous epidural infusion are 1.8 and 4.9.

Conclusion: Ropivacaine and PPX unbound clearance depends on body weight and age. The results support approved dose recommendations of ropivacaine for the paediatric population.

9. Model based assessment of Revatio in pediatric pulmonary arterial hypertension (PAH): a viable route to dose recommendation through an integration of PK, biomarker, and clinical endpoints

Lutz Harnisch1, Pascal Chanu3, Xiang Gao2, Mike Smith1, Rene Bruno3

(1) Pfizer, Pharmacometrics, Sandwich, UK; (2) Pfizer, Clinical Pharmacology, New London, CT, USA; (3) Pharsight, a CertaraTM company, St. Louis, MO, USA

Background: The efficacy of currently registered medications in PAH is mainly based on improvements in exercise capacity: six-minute walk distance (6MWD) or peak oxygen consumption (VO2peak) in adults. In children ≥7 y, 6MWD is not well reproducible and VO2peak is thus the preferred test. In younger patients an exercise capacity test is impractical and hemodynamic endpoints may be suitable alternatives to assess efficacy1,2 instead. Sildenafil (REVATIO®), 20 mg TID, received approval for the treatment of adult PAH based on 6MWD data. A recent FDA analysis3 showed a relationship between changes from baseline in 6MWD and pulmonary vascular resistance index (PVRI) in the adult PAH population. In conjunction with the population PK and an assessment of 6MWD/VO2peak across the age, paediatrics and adults hemodynamic (HD) endpoints play now an important role to support drug development decisions and clinical practice especially in the youngest patients (< 7y) where an exercise capacity measure is not available.

Aim: The objective of this model based analysis was to bridge PK and efficacy measures from adults to children and to provide evidence of the translatability of HD endpoints to clinical efficacy measures with the aim to justify dose selection in the pediatric PAH population.

Methods: Two pivotal sildenafil trials in adult patients: SUPER-14, PACES-15, and one trial in pediatric patients: A1481131 provided data on PK, HDs and exercise capacity. Model- based analyses (218 adults and 219 children) and simulations were conducted to characterize relationships between sildenafil exposure6,7, changes in HD responses8 and changes in exercise capacity outcome9,10. A Bayesian approach adopted the FDA model to contrast the sildenafil experience and to quantify threshold effect sizes on the PD endpoints.

Results: The adult 6MWD and PVRI data and the pediatric VO2peak and PVRI data obtained with sildenafil (in children able to exercise) are consistent with the relationship established by the FDA model. From the 6MWD-PVRI relationship, a minimal target for PVRI improvement can be derived: e.g. a 20-40% improvement in PVRI predicts a 10% improvement in 6MWD. Model-based approaches were also used to quantify exercise capacity (6MWD and VO2peak) and PVR as a function of sildenafil exposure, functional class, etiology and age. Model- based simulations suggest that a dosing regimen of 10/20mg (cut at 20kg) TID is likely to achieve: 1) a sufficient exposure across the whole age range, 2) an exercise capacity outcome of 10% improvement in VO2peak for children 7-17 years of age, comparable to those seen in adults on 6MWD at the labelled dose of 20 mg TID, and 3) in 40% of patients a 20% improvement in change from baseline in PVR, the latter particularly valuable in children too young or otherwise not able to exercise.

Conclusion: Leveraging the FDA assessment on the relationship between 6MWD and PVRI, contrasting it with Pfizer‟s sildenafil data, and utilizing model based approaches to the integration of PK, hemodynamics, and exercise capacity outcomes allowed bridging efficacy from adults to children providing qualified dose recommendations for sildenafil in the pediatric PAH population.

References: [1] L Harnisch, P Chanu, K Dykstra, L Claret, R Bruno, X Gao, Hemodynamics as a substitute for exercise capacity endpoints in patients with pulmonary arterial hypertension (PAH): A model based assessment for sildenafil, ERS, Barcelona, 2010, Presentation 200 [2] J Wagg, L Claret, R Bruno, K Dykstra, X Gao, L Harnisch, Exploratory modeling of exercise capacity and hemodynamic endpoints in patients with pulmonary arterial hypertension, ESC, Stockholm 2010, Presentation 86389 [3] FDA Cardiovascular and Renal Drugs Advisory Committee Meeting, July 29, 2010. http://www.fda.gov/AdvisoryCommittees/Calendar/ucm217266.htm [4] Galie et al, Sildenafil Citrate Therapy for Pulmonary Arterial Hypertension N Engl J Med, 353:2148-57, 2005 [5] Simonneau et al, Addition of Sildenafil to Long-Term Intravenous Epoprostenol Therapy in Patients with Pulmonary Arterial Hypertension, Ann Intern Med, 2008;149:521-530 [6] L Harnisch, N Hayashi Population pharmacokinetic (PK) of sildenafil in paediatric and adult pulmonary arterial hypertension (PAH) patients, ERS, Vienna 2009, P3916. [7] S Watt, N Hayashi, L Harnisch, X Gao, Population pharmacokinetics of sildenafil in paediatric and adult patients with pulmonary arterial hypertension, ESC, Stockholm 2010, Poster 83747 [8] P Chanu, X Gao, M Smith, R Bruno, L Harnisch, A dose selection rationale based on hemodynamics for sildenafil in pediatric patients with pulmonary arterial hypertension (PAH), PAGE, Athens, 2011, Abstr 2104. www.page- meeting.org/?abstract=2104 [9] N Hayashi, L Harnisch, Population PK of Sildenafil and PK/PD assessment of Exercise tolerability in children with Pulmonary Arterial Hypertension (PAH), PAGE, St Petersburg, 2009, Abstr 1523. www.page-meeting.org/?abstract=1523 [10] L Harnisch, N Hayashi Exercise tolerability in children with pulmonary arterial hypertension (PAH), a population PK/PD assessment of the effects of sildenafil, ERS, Vienna 2009, P3890

10. Dosage individualisation in children based on predictive population PK-PD models

Catherijne AJ Knibbe

Department of Clinical Pharmacy, St. Antonius Hospital, Nieuwegein/Division of Pharmacology, LACDR, Leiden University, The Netherlands

Children differ from adults in their response to drugs. These differences may be caused by changes in the pharmacokinetics (PK) and/or pharmacodynamics (PD) between children and adults and may also vary between children of different ages. The maturation rates of these developmental changes vary however between the pathways and receptors and often do not correlate solely with the increase in bodyweight of the child. As a result, instead of an empiric dosing regimen that is based on bodyweight in a linear function, paediatric dosing regimens should be based on an understanding of the PK-PD relationship of the drug in children. Clinical trials in children should therefore consider age-related variability in both PK and PD simultaneously in order to be able to develop rational dosing schemes. In practice, this age-related variability in PK and/or PD must be considered in the context of all other sources of intra- and inter-individual variability resulting from genetic-, environmental- and disease related factors and drug interactions.

If population PK-PD models need to be applied to every single drug in paediatrics, large costs and significant time will be needed to develop evidence-based dosing schedules for each drug. An important question is, to what extent population PK-PD models developed for specific drugs constitute a basis for the development of dosing guidelines for drugs other than those that have actually been studied. In this respect the kinetics of age-related changes in renal function, the functionality of drug metabolizing enzymes, drug transporters, as well as the expression function of pharmacological receptors are patient specific or biological system-specific properties. These system-specific properties, derived from one „model‟ drug, could in principle serve as a basis for the prediction of age related changes in the PK and PD of other drugs (so called extrapolation). Using simulations for drugs other than those used to generate biological system specific information may significantly reduce the time and costs needed to develop drug dosing guidelines for individual drugs. During the presentation, results will be shown of four PhD projects in paediatric PK-PD modelling; 1.glucuronidation (UGT), 2. oxidation (CYP3A), 3. renal function, and 4. liver blood flow. Results of meta PK analyses with novel paradigms for individualized dosing in children are discussed together with validation approaches specific for paediatric datasets. Furthermore, results of proof-of-principle studies are presented on the extrapolation of the system-specific parts of the models to other drugs sharing the same pathway.

It is concluded that future research should focus on the development of validated models for specific elimination routes and targets that have predictive and extrapolation potential, making them of use in designing algorithms to derive first-time-in-child doses and to derive individualized dosing guidelines in paediatrics. This methodology will improve the efficacy/safety balance of dosing guidelines which will be of benefit to the individual child.

References: De Cock RF, Piana C, Krekels EH, Danhof M, Allegaert K, Knibbe CA. The role of population PK-PD modeling in paediatric clinical research. Eur J Clin Pharmacol. 2011 May;67 Suppl 1:5-16. Knibbe CA, Danhof M. Individualized dosing regimens in children based on population PKPD modelling: are we ready for it? Int J Pharm. 2011 Aug 30;415(1-2):9-14. Knibbe CAJ, Krekels EHJ, Danhof M. Advances in paediatric pharmacokinetics. Expert Opin Drug Metab Toxicol. 2011 Jan;7(1):1-8.

Session 3: Systems Biology-a new world?

11. (How) Can Systems Pharmacology help to improve Phase 2 success?

Piet H. van der Graaf

Pfizer, Pharmacometrics/Global Clinical Pharmacology, Sandwich CT13 9NJ, UK. [email protected]

Mechanistic PKPD models are now advocated not only by academic and industrial researchers, but also by regulators. A recent development in this area is based on the growing realisation that innovation could be dramatically catalysed by creating synergy at the interface between Systems Biology and PKPD, two disciplines which until now have largely existed in „parallel universes‟ with a limited track record of impactful collaboration. This has led to the emergence of systems pharmacology. Broadly speaking, this is the quantitative analysis of the dynamic interactions between drug(s) and a biological system to understand the behaviour of the system as a whole, as opposed to the behaviour of its individual constituents; thus, it has become the interface between PKPD and systems biology. It applies the concepts of Systems Engineering, Systems Biology, and PKPD to the study of complex biological systems through iteration between computational and/or mathematical modelling and experimentation. Application of systems pharmacology can now impact across all stages of drug research and development, ranging from very early discovery programs to large-scale Phase 3/4 patient studies, and has the potential to become an integral component of a new „enhanced quantitative drug discovery and development‟ (EQD3) R&D paradigm.

12. Developing and applying safety-critical software for systems-level modelling

David Gavaghan

University of Oxford, Wolfson Building, Parks Road, Oxford, OX1 3QD, UK

Systems level modelling approaches are gradually gaining acceptance as a potentially valuable tool in understanding the effects, and possible side-effects, of potential drug compounds. In this talk, I will begin by illustrate the potential of this approach by discussing some examples of where computational methods have helped in the development and/or safety-assessment of new drugs. Implicit in such approaches is the need to use computational methods to solve the underlying mathematical equations used to model the behaviour of the system. In general little attention, at least in academia, has been paid to the manner in which such computational methods are developed and tested.

I will therefore go on to discuss a new approach to the development of systems biology modelling software that we have developed within our group, that addresses this issue. The underlying premise behind our approach is that such software will eventually form part of the research toolkit used to gain a quantitative understanding of the behaviour of biological systems, and will be used routinely alongside experimental approaches in gaining a greater understanding of biological processes in health and disease, and ultimately in the development of novel therapies and drugs. Such software will therefore have to be developed in a “safety-critical” manner, just as existing software for, for example, air traffic control, is developed. The software that we have developed in Oxford over the last seven years - Chaste (Cancer, Heart and Soft Tissue Environment – see http://www.cs.ox.ac.uk/chaste/) – has been developed using a test-first, agile methodology adapted from the software industry and applied to academic software development. It is a general purpose simulation package aimed at multi-scale, computationally demanding problems arising in biology and physiology.

The major application project involving Chaste over the last 4 years was the EU-funded PreDiCT project. This involved a collaboration between academia, Fujitsu, and 5 major pharma companies to use computational approaches to gain a better understanding of mechanisms giving rise to cardiotoxic side effects in existing and potential drug compounds. I will illustrate the potential power of the systems modelling approach by describing one aspect of the project which has resulted in an on-going collaboration with GSK and Astra- Zeneca to develop a novel approach to the assessment of drug cardiotoxicity, as described in Mirams et al (Cardiovasc Res (2011) 91(1): 53-61).

13. Pharmacological Mechanism Based Drug Safety Prediction

Darrell R. Abernethy

Office of Clinical Pharmacology, Center for Drug Evaluation Research, United States Food and Drug Administration, Silver Spring, Maryland 20993

The withdrawal of several drugs from the US market a short time after their approval due to toxicities that had not been predicted during drug development has highlighted the need for further development of risk prediction for drugs in development. At the present time nonclinical toxicological evaluation followed by a rather routine set of safety evaluations during clinical drug development constitutes the understanding of risk at the time a drug is marketed. After the drug is marketed, pharmacoepidemiological approaches to detect “safety signals” after toxic events have occurred are used to characterize risk from exposure to a new drug. Certainly these approaches have been useful to detect unexpected risk, as in the examples of rofecoxib, terfenadine, mibefradil, cisapride, and others. Unfortunately this approach requires injury and sometimes death for patients in the course of detecting and evaluating the risk signal.

To complement these approaches, we are developing methods to predict risk based on pharmacological mechanisms of a drug candidate. These mechanisms are both for on target (intended therapeutic effect) and off target sites of action. This requires a synthesis of understanding of the chemistry of the drug, nonclinical toxicology that is available, and exploration of all biological pathways that such a chemical and therapeutic target may be associated with. This is a “bottom up” approach, evaluating genetic, molecular, cellular, organ, and system potential for toxicity, to complement current “top down” approaches that identify a clinical toxicity and then explore its mechanism. These approaches are meant to be complementary, however add a predictive ability to inform evaluations for risk during late clinical development and postmarket epidemiological surveillance.

The recent advent of systems approaches to chemistry, toxicology, and biology have opened up the possibility of developing these predictive approaches. A broad collaborative effort among disparate disciplines is necessary to achieve the full potential of early risk prediction.

This is a pilot program in the Office of Clinical Pharmacology and preliminary examples of this approach will be presented.

Session 4: Open Session

14. A first prototype of the new Clinical Trial Simulator developed within the DDMoRe project

Marc Lavielle (1,2), Joachim Grevel (3)

(1) INRIA Saclay, (2) Depart. of Mathematics, University Paris-Sud, (3) BAST Inc Ltd., Nottingham, UK

Objectives: One of the objectives of the DDMoRe (Drug Disease Model Resources) consortium (2nd IMI Call) is to develop new tools for model-based drug development (MBDD). A clinical trial simulator (CTS) enables effective implementation of the learn-and- confirm paradigm in MBDD. Through simulations the anticipated success rate of a future trial can be estimated. For various reasons industry has not embraced currently available commercial software. A new CTS tool will fill that gap.

Methodology: The capabilities of the current prototype comprise of:  Basic design of parallel groups used in Phase 2 of clinical drug development,  Simulations of  Patients sampled from known distributions or populations  Exposure to the investigational drug  Drug effects related to exposure (continuous effects, time-to-event effects, categorical effects, count data as effects),  Methods of evaluating the simulated trails with respect to  Optimal sampling times  Individual probabilities or risks  Trial success,  Automatic reporting,  Extensive graphical display. . Conclusions: A new CTS will be developed during the course of the DDMoRe consortium.. Already the first prototype demonstrates that the technology and the methodologies implemented are powerful and that its integration into an efficient workflow is possible. Subsequent versions of the CTS will incorporate dropout and compliance models.

15. Target Mediated Drug Disposition Model to Describe the Expression and Kinetics of IL12 and IFN in Gene Therapy

Zinnia Parra Guillén1, Rubén Hernández-Alcoceba2, Gloria González-Aseguinolaza2, Pedro Berraondo2, Iñaki F. Trocóniz1

1Department of Pharmacy and Pharmaceutical Technology, School of Pharmacy, University of Navarra (Spain); 2Division of Gene Therapy and Hepatology, Center for Applied Medical Research (CIMA), University of Navarra, Spain

Background: Despite the great potential revealed by interleukin-12 (IL12) in the treatment of chronic hepatic diseases, clinical trials have shown severe toxicity at the therapeutic dose ranges needed. Gene therapy approach allows to achieve high local protein expression maintaining low serum concentration levels, and therefore, decreasing undesired side effects. Following this concept, a gutless adenoviral vectors containing a Mifepristone (RU, inductor of the gene expression) inducible system for liver-specific expression of IL12 was developed by Wang L et al (Gastroenterology, 2004). In their work, it was found that IL12 in vivo expression was hampered by a negative feedback mediated by the interferon  (IFN), produced in response to this cytokine.

Aim: To develop a model able to describe the time course of IL12 and IFN expression under different dosing scenarios and including wild-type and knock-out mice for the receptor of the IFN.

Methods: Knock-out and wild type mice were infected with two doses (low and high) of a gutless adenoviral vectors containing a Mifepristone-inducible system for the expression of IL12. Daily induction with constant or increasing doses of RU was performed to produce liver protein expression, and serum levels of IL12 and IFN were measured. NONMEM VII, Berkeley-Madonna, and R softwares were used to develop and evaluate the model.

Results: The model developed using data from wild-type mice had the following features: (i) IL12 expression was modelled through a zero order rate process which was linearly and non-linearly (EMAX model) dependent on the DNA and RU dose levels administered, respectively; (ii) a target drug mediated disposition (TMDD) model was selected to describe the kinetics of IL12 in serum, (iii) an indirect response model where the zero order rate of synthesis was controlled by the receptor bound IL12 was used to describe the time course of IFN in serum, and (iv) IFN exerted a negative feed-back on IL12 expression. However, simulations from that model, where decreasing elimination rates of IFN were used to emulate the knock-out mice conditions, were not able to describe the profiles observed in those mice. Incorporation of a TMDD for IFN, and modifying the negative feed-back effect on IL12 expression by considering the presence of a modulator which synthesis is controlled by the receptor bound INF allowed a good description of the complete set of experimental data.

Conclusions: A computational model able to describe jointly the IL12 and IFN profiles observed both, in wild type and knock-out mice, through diverse experimental conditions has been developed, increasing the understanding of the underlying biological mechanisms and providing a tool for the optimisation of dosing regimens in future animal experimentation.

16. Experimental and Mathematical Analysis of Pitavastatin Hepatic Uptake Across Species

T. R. B. Grandjean1, A. Lench2, J. W. T. Yates3, M. J. Chappell1, C. J. O’Donnell3

1 School of Engineering, University of Warwick, Coventry, CV4 7AL, UK 2 Departments of Pharmacy and Pharmacology, University of Bath, Bath, UK 3 DMPK Department, Astrazeneca, Alderley Park, Macclesfield, UK

Background: Pitavastatin is an enzyme used to treat hypercholesterolaemia, which is actively uptaken into hepatocytes, medited mainly by the Organic Anion Transporting Polypeptide (OATP) 1B1[1,2] . OATP is of particular interest as it is capable of transporting a large array of structurally divergent drugs. The current widely accepted view is that the rate of diffusion of Pitavastatin into the cell is the same at both 4°C and 37°C but that the transporter action only occurs at 37°C (Shimada et al [1]). Data are normally collected at both 4°C and 37°C, and the 4°C data are used to estimate diffusion whilst the 37°C are used to estimate the transporter affinity. It is suspected that the current Shimada et al assumption is not valid since fitting the 4°C and 37°C simultaneously with the same diffusion rates gave very poor fits. Fits are therefore carried out for 4°C and 37°C data individually to test the hypothesis.

Aim: To investigate the nonlinear kinetics of in vitro hepatic uptake the OATP substrate, Pitavastatin, and quantify the mechanisms present both structurally and numerically across three species (rat, dog and human).

Methods: Experiments were conducted using freshly isolated hepatocytes, utilising the „oil spin‟ methodology described by Hassen et al [3]. Briefly, freshly isolated rat, dog and human hepatocytes were incubated with Pitvastatin (5 – 300µM, 1 – 650µM and 1 – 100µM respectively). At 10, 30, 50 and 70 seconds 100µL aliquots were spun through a silicone oil layer to separate the hepatocytes from the media. [Pitivastatin]hepatocyte was detemined using Liquid Chromatography double Mass Spectrometry (LC-MS/MS).

Modelling: Two candidate non-linear pharmacokinetic compartmental models have been derived to characterise the uptake process. Structural identifiability analyses were performed to establish that all unknown parameters could be identified from the experimental observations available. Furthermore, structural indistinguishability analysis was performed to establish that both candidate models input/output relationships were structurally different. A kinetic modelling software package, FACSIMILE (MPCA Software, UK), was used to obtain numerical solutions for the system equations and for parameter fitting. Sensitivity analysis and model robustness analyses were also performed.

Conclusions: Uptakes to rat, dog and hum hepatocytes were saturable and progressed according to Michaelis-Menten kinetics. Model fits gave good agreement with the in vitro data and suggest the current widely accepted assumption that the rate of diffusion of Pitavastatin in the hepatocytes is the same at both 4°C and 37°C is not valid. Results also suggest that both diffusion rate and transporter affinity are similar across species. Once fully validated the model has the potential to perform robust, predictive simulations to ascertain optimal levels of uptake and the effects of the use of appropriate inhibitors. This work was supported by AZ & TG was supported by AZ and the MRC Capacity Building initiative.

17. Semi-mechanistic Modelling of Double Peaks in Pharmacokinetics: LDOPA case study

Kayode Ogungbenro, Henry Pertinez and Leon Aarons

Centre for Applied Pharmacokinetic Research, School of Pharmacy and Pharmaceutical Sciences, The University of Manchester, Oxford Road, Manchester, M13 9PL, United Kingdom

Following oral administration, concentration-time profiles of some drugs show double peaks which can be explained by a number of different mechanisms: delayed gastric emptying; variable absorption; enterohepatic recirculation; drug secretion etc. The profiles of such drugs may not be adequately described by classical compartmental models. Levodopa (LDOPA) the principal drug in the treatment of Parkinson‟s disease has demonstrated double peak profiles in almost all patients in clinical studies which has been attributed to the effect of gastric emptying. Gastric emptying has also been shown to be important in the systemic availability of LDOPA due to the metabolism of the drug by decarboxylase enzymes in the gastric mucosa. Data from simultaneous scintigraphy and a paracetamol absorption test with and without LDOPA were obtained from the literature. Gastric emptying profiles obtained from scintigraphy in the absence of LDOPA showed rapid decline. In the presence of LDOPA the gastric emptying profile showed a period of interruption with no emptying and this is associated with double peaks in both paracetamol and LDOPA concentration-time profiles. A semi-mechanistic emptying model, with separate compartments for the stomach, small intestine, central and peripheral compartments, was developed to describe LDOPA pharmacokinetics and its double peak. A feedback mechanism was introduced via an effect compartment that links the plasma concentration of LDOPA to the rate of gastric emptying. This allows LDOPA pharmacokinetics to control the rate of gastric emptying as observed in the profiles which give rise to multiple peaks. The model was applied to the data with and without scintigraphy data and in both cases the model produced a very good fit to the data. Even in the absence of scintigraphy data the model predicted stomach profile adequately.

18. Pharmacokinetic and Pharmacodynamic Analysis of Efavirenz Dose Reduction Using a Physiologically-Based Dynamic Model

Marco Siccardi1, Lisa Almond2, Alessandro Schipani1, Chantal Csajka3, Catia Marzolini4, Duncan Edwards2, Andrew Owen1, David Back1.

1- Pharmacology Research Laboratory, Institute of Translational Medicine, University of Liverpool, UK 2- Simcyp Limited, Sheffield, UK 3- Division of Clinical Pharmacology and Toxicology, Département de Médecine, Hôpital de Beaumont, Centre Hospitalier Universitaire Vaudois, Lausanne, Switzerland 4- Division of Infectious Diseases and Hospital Epidemiology, University Hospital Basel, Basel, Switzerland

Background: Efavirenz (EFV) pharmacokinetics (PK) is characterized by large inter-patient variability and a correlation between plasma exposure and efficacy has been described. CYP2B6 is the main enzyme responsible for EFV metabolism and 516G>T (rs3745274) is considered to be the main genetic variant. The aim of this study was to develop a physiologically-based model to simulate EFV PK and PD in virtual human subjects.

Materials and Methods: In vitro data describing the chemical properties, ADME of EFV and the effect of CYP2B6 516 genotype on CYP2B6 protein expression in liver tissue were obtained from published literature. These data were used to simulate EFV (600mg once daily) PK using Simcyp Simulator. Simulated PK parameters, such as Ctrough, Cmax, AUC, and the impact of 516G>T genotype were compared with clinical observations. PK/PD was characterised using data from Csajka et al. and Marzolini et al. and a significant correlation between C8-16hr and viral suppression was identified. This was incorporated into the IVIVE and the effect of dose reduction to 400mg once daily on PK and viral load simulated.

Results: In a virtual Caucasian cohort of 500 patients the simulated Ctrough was equal to 2275 ± 2045 ng/ml and Cmax 3438 ± 2246 ng/ml. Clearance (CL/F) was 15.6 l/hr for 516 GG, 11.4 l/hr for 516GT and 6.4 l/hr for 516 TT. Overall, the mean (95% CI) probability of achieving viral suppression was 80% (44-99) and 75.5% (33-99), 81.6% (41-99), 88.8% (61- 99) in patients with 516 GG, 516 GT, 516 TT genotypes, respectively. Following dose reduction to 400 mg, overall mean probability of viral suppression was decreased to 73.4% (34-99), and to 69.2% (26-97), 76.7% (34-99), 84.9% (53-99) for 516 GG, 516 GT, 516 TT genotypes, respectively.

Conclusions. The model developed predicted the PK and PD of EFV in individuals with different CYP2B6 genotypes. These simulations indicate that genotype-guided dose reduction could be used in patients without compromising viral suppression.

19. Defining a therapeutic window for the novel TGF-β inhibitor LY2157299 based on a pharmacokinetic/pharmacodynamic (PK/PD) model

Ivelina Gueorguieva1, Ann Cleverly1, Anja Stauber2, N. Sada Pillay2, Jordi Rodon3, Colin Miles1, Jonathan Yingling2 and Michael Lahn

1 Lilly Research Laboratories, Sunninghill Road, Windlesham, Surrey, UK 2 Lilly Research Laboratories, Indianapolis, Indiana, USA 3 Medical Oncology, Vall d’Hebron, Barcelona, Spain

Background: Transforming Growth Factor-Beta (TGF- β) signalling has been recognized as an important regulation of tumour growth in advanced forms of cancer. A number of development programs aimed at identifying TGF- β receptor kinase inhibitors, have been stopped due to severe toxicities observed in animals. We investigated several activin receptor-like kinase (ALK) 5 inhibitors from which LY2157299 was selected for clinical evaluation. While we also describe similar cardiovascular toxicities as have been reported for ALK5 inhibitors, we used a PK/PD model to prospectively establish a therapeutic window that would be safe for human clinical investigation.

Aim: To successfully identify a therapeutic window for LY2157299 in the treatment of cancer patients based on preclinical PK/PD model.

Methods: The preclinical model for LY2157299, combined plasma concentrations, biomarkers in tumour and tumour size (Bueno et al., 2008). Using this model with allometric scaling to human, we simulated plasma concentration and biomarker dynamics to prospectively predict pharmacological doses in patients. In addition to the efficacy data, the toxicity data was considered in identifying a safe dose range for patients. . LY2157299 (short for LY2157299 monohydrate) was administered as a tablet to glioblastoma patients on an empty stomach. During LY2157299 monotherapy, there was a cohort-by-cohort safety and PK analysis prior to each dose escalation. LY2157299 monotherapy was given intermittently in patients with glioma at a dose range of 80, 120 and 150mg twice daily (morning and evening) and to all comers at lower doses (20 and 40mg BID). Population PK analysis was performed on 37 patients with total 717 observed concentration measurements. Demographic characteristics were tested as potential covariates. Simulations with patient population PK/PD model were performed to describe tumour biomarker levels in patients. In preparation for future trials with LY2157299 anticipated plasma exposure levels were simulated and optimal sampling windows were proposed.

Results: We present the application of PK/PD model during FHD study of LY2157299 and how it allowed defining and confirming therapeutic window of LY2157299. The lower bound of the prospective therapeutic window was at 150mg/day dose where simulated median exposure reached 30% biomarker inhibition level (pSMAD) and the upper bound around 250 mg/day dose, where 80th percentile simulated patient exposure was equivalent to the no effect level for cardiac toxicity in rats. At these high levels, an anticipated reduction of 40% in pSMAD was expected. Driven by the necessity to fully understand exposure variation between patients we expanded cohorts for PK, once we were within desired dose/exposure range. The oral mean population apparent clearance of LY2157299 was 38 L/h with between-subject variability of 46%. The initial volume of distribution 100 L, steady-state volume 193 L and absorption rate constant was 2.2 h-1. The between-occasion variability on apparent CL was estimated at 18%. In some patients PD effects (pSMAD) indicated target inhibition starting at 80mg BID dose level.

Conclusion: This approach allowed to overcome some of the concerns indicated by the preclinical toxicology studies and to successfully identify the therapeutic window for LY2157299 in the treatment of cancer patients.

References: Bueno L, de Alwis DP, Pitou C, Yingling J, Lahn M, Glatt S, Trocóniz IF. Semi-mechanistic modelling of the tumour growth inhibitory effects of LY2157299, a new type I receptor TGF-beta kinase antagonist, in mice. Eur J Cancer. 2008 Jan;44(1):142- 50. Jordi Rodon, Jose Baselga, Emiliano Calvo, Joan Seoane, Irene Brana, Elisabet Sicart, Ivelina Gueorguieva, Ann Cleverly, Michael Lahn, N. Sada Pillay, Matthias Holdoff, Jaishri Blakeley and Michael Carducci. First Human Dose (FHD) Study of the oral transforming growth factor-beta (TGF-β) receptor I kinase inhibitor LY2157299 in patients with treatment-refractory malignant glioma. ASCO MEETING ABSTRACTS, June 9, 2011:3011.

20. “First dose in children”: evaluation of clearance scaling approaches in CYP3A4-metabolized drugs

Ashley Strougo1-2, Claire Monnereau1-2, Ashraf Yassen1, Meindert Danhof2, Jan Freijer3

1Global Clinical Pharmacology & Exploratory Development, Astellas Pharma Europe, Leiderdorp, The Netherlands;2Division of Pharmacology, Leiden/Amsterdam; 3Center for Drug Research, Leiden University, Leiden, The Netherlands

Introduction: Dose selection for “first in children” trials often relies on scaling of the clearance from adults to children. Previously reported approaches are mechanism-based and allometric scaling in combination with maturation of clearance for early life [1,2]. In this investigation we aimed to evaluate the performance of these two approaches for drugs metabolized mainly by CYP3A4 enzyme.

Methods: From the literature, the clearance in adults and children of various age were randomly selected for 20 drugs mainly metabolized by CYP3A4 enzymes. The clearance in adults were scaled to children using allometric scaling in combination with maturation function and using a mechanism-based approach, which relies on the concepts of the well stirred model and takes into account maturation of different physiological parameters/process, such as liver blood flow, protein binding, liver size and maturation of the CYP3A4 enzyme activity in the gut and in the liver. Additionally, the maturation function of CYP3A4 enzyme activity in the liver reported by Edginton et al and Johnson et al were evaluated in combination with each scaling approach [1,2]. The four types of predictions in children were then compared with observed clearances.

Results: For all the 20 drugs there are in total 187 observations in children and 38 in adults. All four types of predictions were adequate in older children and slightly to strongly biased in infants younger than 3 months. In children younger than 3 months, the use of the CYP3A4 maturation function reported by Johnson et al [2] strongly compromised the accuracy of the predictions independent of the approach used (more 75% of the predictions in children outside of the 2-fold range). The most optimal approach in children younger than 3 months was found to be the mechanism-based approach using the maturation function for the CYP3A4 enzyme in the liver as reported by Edginton et al.

Conclusion: These results support the use of a more mechanistic approach for scaling the clearance of CYP3A4 drugs in order to select the doses of “first in children” trials. Additional caution in dose selection of neonates and young infants is warrant.

References: 1 Edginton et al, Clin Pharmacokinet 2006; 45 (7): 683-704 2 Johnson et al, Clin Pharmacokinet 2006; 45 (9): 931-956

Poster Abstracts

A Joint Selection Model for Efficacy and Dropout using WinBUGS

1,2 1 2 Eunice Yuen , Ivelina Gueorguieva , Leon Aarons

1 2 Global PK/PD/TS, Eli Lilly & Co., Windlesham Surrey, School of Pharmacy and Pharmaceutical Sciences, University of Manchester

Background: Selection models are models which incorporate both the disease progression and dropout, using a joint distribution. The major advantage of selection models over other approaches such as imputation, is the ability to simulate from the model to predict dropout rates and drug efficacy or disease progression for future clinical trials. Bayesian methods were chosen to develop the joint model since it has been described as a more natural way to handle random-coefficient models, in particular shared-parameter models.

Aim: To develop a selection model for a drug with categorical trial endpoints to aid future clinical trial design and to determine the influence of dropout on the drug disease model and vice versa.

Methods: A PK/PD model describing weekly scores on a categorical scale from 0-10 following 3 dose levels of a drug had been previously developed using NONMEM. In the current analyses, the same PK/PD model was developed in WinBUGs with the form:

and

2 where ij = jth score for individual i,  = residual error variance, Base = score at baseline, PE = magnitude of placebo effect, C = first order rate constant describing the onset of placebo effect, Emax = maximum effect, EC50 = concentration required to achieve 50% of maximal effect, K= first order rate constant describing the onset of drug effect. Uninformative priors were used since this would mirror the nonlinear mixed effects estimation methods. The dropout model was modelled both simultaneously and separately from the efficacy model and had the following structure:

where Pr = probability of dropout, 1=intercept, 2 and 3 = parameters associated with time and 4 = parameter associated with the change from baseline model-predicted score at time of dropout. Three sets of priors for 1 to 4 were investigated (informative, vague or diffuse and non-informative priors) to test the sensitivity of the model to its supplied priors. Convergence was inspected by visual and diagnostic checks.

Results: Convergence was achieved with all different sets of priors. Regardless of the chosen priors, all the model parameter estimates from the joint (simultaneous) estimation models were similar to those where the efficacy and dropout were modelled separately. Some parameters of the dropout model estimated using uninformative priors had larger standard deviations compared to the estimates using vague or informative priors.

Conclusion: A selection model incorporating both efficacy and dropout was developed for a drug with categorical trial endpoints. This will enable more accurate simulations for future clinical trials with similar drugs and/or indications. The use of separate models for efficacy and dropout were supported in these analyses, similar to other reports in literature. The main advantage of separate modelling approaches lie in the simplification of model codes and datasets, as well as analysis run times.

Population Pharmacokinetic Analysis of Ropivacaine and its Metabolite PPX from Pooled Data in Neonates, Infants and Children

L. Aarons(1), B. Sadler(2), M. Pitsiu(3), J. Sjövall(4), J. Henriksson(5)

Objectives: The aim of the present study was to characterize ropivacaine and PPX pharmacokinetics and factors affecting them in paediatric anaesthesia.

Conclusion: Ropivacaine and PPX unbound clearance depends on body weight and age. The results support approved dose recommendations of ropivacaine for the paediatric population.

Predicting the Effect of CYP2D6 Polymorphism on Pharmacodynamic Response to Metoprolol

K. Abduljalil1, L. Gaohua1, R. Rose1, T.N. Johnson1, M. Chetty1, M. Jamei1, D. Edwards1, A. Rostami-Hodjegan1,2

1-SIMCYP Limited, Sheffield, UK 2-School of Pharmacy, Manchester University, UK

Background: The polymorphism of CYP2D6 enzyme is believed to be an important determinant of variation in the clinical response to standard doses of metoprolol in ultrarapid metabolisers (UMs), extensive metabolisers (EMs) and poor metabolisers (PMs). Plasma concentrations and effects on heart rate have been shown to correlate signiﬁcantly with CYP2D6 metabolic phenotype in clinical studies. The prevalence of some phenotypes is not adequately high to discern the differences in PKPD of drugs by conduct of small clinical studies. It would be of value to use the in vitro information on metabolism together with PKPD information in prevalent phenotypes of CYP2D6 to conduct virtual clinical studies with a view to assess the potential pharmacological differences in various less frequent phenotypes prior to conduct of any clinical studies.

Objectives: To simulate the reduction of heart rate due to a standard 100 mg dose of metoprolol in virtual healthy Caucasian populations stratified for their CYP2D6 phenotypes using Simcyp simulator.

Methods: Simulations of metoprolol PK and the decrease in heart rate effects in UMs, EMs and PMs were performed using Simcyp V11. The PKPD relationship was taken from Kirchheiner et al 20041, and was assumed to be the same regardless of CYP2D6 genotype. Simulations were compared with clinical observations from 2 studies1,2.

Results: The simulated contribution of the CYP2D6 phenotype to metoprolol PK/PD within Simcyp is based on the propagation of the differences in CYP2D6 abundance to the PD response via changes in the plasma concentration profile. In general both PK and PD profiles were predicted successfully. Simulated/observed ratios for AUC, Tmax, Cmax, and CLpo values are 1.26-, 1.12, 0.84, and 0.85 for PMs, 0.70, 0.88, 0.63, and 1.22 for EMs and 1.11, 1.11, 1.03, and 0.90 for UMs, respectively. The simulated CL (Dose/AUC) of UMs group is 16- and 2-fold higher than that of PMs and EMs group, respectively. Simulated mean PD profiles showed that the area under the effect curve in PMs were 6-fold higher than that in UMs, and 2-fold higher than that in EMs. The simulated/observed ratios for the maximum reduction in heart rate and absolute area under effect curve are 0.94 and 1.51 for PMs, 0.95 and 0.96 for EMs, and 0.94 and 0.73 for UMs groups, respectively.

Conclusions: The Simcyp Simulator and its PKPD module is a seamless tool to assess the propagation of key PK factors, such as metabolic activity or drug-drug interactions, through to a PD effect. Simulation results showed consistency with clinical observations in terms of significant differences of metoprolol PKPD profiles between PMs and UMs with a marginal change between EMs and UMs. UMs may not achieve optimal target concentrations of metoprolol, which can lead to a lower benefit from the standard 100mg dose of the drug compared with PMs. Although POPPK studies have been valuable to inform investigators of such differences, these studies should be powered adequately to recognise the differences. Clinical trial simulations similar to the one shown in this study can be used to investigate the design of POPPK studies and their power.

References: Kirchheiner J et al., Clin Pharmacol Ther. 2004;76(4):302-12. Sharma A et al., J Pharmacol Exp Ther. 2005 ;313(3):1172-81. Dried Blood Spots and Sparse Sampling: A perfect combination for minimally invasive PK/PD studies in children

Parul Patel (1), Hitesh Pandya (2), Neil Spooner (3), Oscar Della Pasqua (3), Sonya Gade (3), Venkatesh Kairamkonda (4), Graham Lawson (1), Sangeeta Tanna (1) and Hussain Mulla (5)

(1) Leicester School of Pharmacy, De Montfort University, UK; (2) University of Leicester, UK; (3) GlaxoSmithKline, R&D, UK; (4) Neonatal Intensive Care Unit, University Hospitals of Leicester NHS Trust, UK; (5) Centre for Therapeutic Evaluation of Drugs, Glenfield Hospital, UK

Background: Dried blood spots (DBS) have recently received considerable interest for application to PK studies. The technique requires a micro (~50µl) blood-volume sample, and is therefore particularly advantageous to studies involving children and neonates. In-vitro validation of DBS based quantification techniques indicate a comparable performance to methods based on large blood-volumes. However, validation in the clinic is necessary to ascertain the robustness of the DBS sampling methodology as a means of generating population PK data in children.

Objective: To perform a clinical validation of the DBS technique in preterm neonates receiving caffeine therapy for apnoea of prematurity.

Method: In a prospective study, between 1-10 (15µl) DBS samples (total 338) were collected opportunistically from 67 preterm neonates at random times intervals post caffeine dose. Neonates received oral and iv caffeine doses according to the local protocol. Caffeine exhibits low plasma protein binding, does not bind to red blood cells (RBC) and has a blood- to-plasma ratio of 1. Therefore conversion of blood values for comparison was not necessary. The DBS caffeine concentration data was used to develop a Pop-PK model, and compared with a previously reported model based on conventional plasma caffeine concentrations.

Results: A 1 compartmental model with zero and first order absorption described the DBS data well. Parameters derived from DBS data were estimated with precision (RSE <10%) and were comparable to CL and V estimates from plasma (6.83 vs. 6.96ml/h/kg and 614 vs. 851ml/kg, respectively). Weight and postnatal age were the most influential covariates in the CL model which is in accordance to previous findings (Charles, 2008). Similar to Charles et al. the BOV in CL (31.2%) was higher than the BSV in CL (24.7%) which has important implications for caffeine TDM. Model evaluation using bootstrap and PC-VCP confirmed the robustness of the model.

Conclusion: The DBS based population model enabled precise estimation of caffeine PK parameters in preterm neonates. Furthermore, estimates were comparable to plasma literature values derived from a demographically similar neonatal population. DBS is potentially a more practical and ethical sampling technique in PK /PD studies involving young children provided there is sufficient understanding of the behaviour of the drug with respect to RBC association and protein binding.

References: Charles, B. G., Townsend, S. R., Steer, P. A., Flenady, V. J., Gray, P. H. & Shearman, A. 2008. Caffeine Citrate Treatment for Extremely Premature Infants With Apnea: Population Pharmacokinetics, Absolute Bioavailability, and Implications for Therapeutic Drug Monitoring. Therapeutic Drug Monitoring, 30, 709-716.

The influence of severe hepatic dysfunction on the metabolic capacity of the liver in children: Composition and characterization of a liver tissue bank

L. De Bock1, K. Boussery1, M. Van Winckel2, P. De Paepe3, X. Stephenne4, E. Sokal4, J. Van Bocxlaer1

1 Laboratory of Medical Biochemistry and Clinical Analysis, Faculty of Pharmaceutical Sciences, Ghent University, Harelbekestraat 72, 9000 Gent, Belgium, 2 Ghent University Hospital, Paediatric Gastroenterology Dpt., De Pintelaan 185, 9000 Gent, Belgium, 3 Heymans Institute of Pharmacology, Ghent University, De Pintelaan 185, 9000 Gent, Belgium, 4 Catholic University of Louvain and St Luc Clinics, Paediatric Department (HPED), PEDI unit, Laboratory of Paediatric Hepatology and Cell Therapy, Hippocrate Avenue 10, 1200 Brussels, Belgium

Introduction: The influence of liver disease on the hepatic biotransformation of drugs has mainly been studied in adult populations, whereas for paediatric populations similar data are missing. Information on the impact of hepatic failure on the activity and abundance of the main CYP450 isoforms is however essential for predicting the need for dosage adjustments when a (new) drug is used in a child with liver disease. Therefore, this study aims to fill this knowledge gap by composing and characterizing a liver tissue bank comprising pathological paediatric liver samples.

Methods: All patients undergoing liver transplantation in a Belgian university hospital at age 12 or lower were included in the study, except when infectious diseases were diagnosed or suspected. Liver tissue samples were snap frozen after explantation, and were processed into subcellular fractions (microsomes and cytosolic fraction) for further analysis. Patient files were consulted for the relevant pre-operative history. The in vitro activities of CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP2E1, and CYP3A4, will be determined through the incubation of the microsomes with the specific probe substrates phenacetin, tolbutamide, S- mephenytoin, dextromethorphan, chlorzoxazone, and midazolam, respectively. An indirect ELISA will be used for the determination of the abundance of the isoforms. The most important SNPs of the highly polymorphic isoforms will be analyzed using TaqMan® Drug Metabolism Genotyping Assays.

Results: Up to now, samples from thirty-one children with diverse pathologies (23 with biliary atresia, the others with PFIC II , cystic fibrosis, 1-antitrypsine deficiency, neonatal hemochromatosis, or acute liver failure) were collected and processed. Elaborated patient details will be discussed in the presentation. An incubation protocol and a UPLC-MS/MS method were optimized and validated for the determination of the activities of the aforementioned CYP isoforms. The activities of these isoforms in the 31 microsomal samples were determined, as well as in commercially available adult pools. Some patients showed very low activities compared to a pool of adult microsomes, whereas others showed extremely high activities (up to 300 times the adult activity).

Conclusions: A liver tissue bank containing microsomes, cytosolic fractions and DNA extracts from 31 paediatric liver patients was collated. As expected, hypervariable activities in this population were detected for all 6 isoforms. Due to the diverse characteristics (age, pathology) of the patients, interpretation of the activity data will be an intricate task. In order to expand the data set, additional data will be collected in the near future, such as information on the most important SNPs leading to altered activity, and on the abundance of the isoforms. Furthermore, the potential inducing or inhibiting pre-operative drug use, as well as the relevant clinical biology results should be taken into account.

Internal and external validation for aminoglycoside population pharmacokinetic model in patients with cystic fibrosis

Alghanem S1, Touw D2, Thomson AH1,3

1Strathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, Glasgow, 2Apotheek Haagse Ziekenhuizen and Haga Teaching Hospital, PO Box 43100, 2504 AC The Hague, Netherlands, 3Pharmacy Department, Western Infirmary, NHS Greater Glasgow and Clyde, Glasgow.

BACKGROUND: An aminoglycoside population pharmacokinetic model was developed using a cystic fibrosis data set from Glasgow. A new, independent data set of tobramycin data was available from patients with cystic fibrosis who were monitored in a hospital laboratory in The Netherlands.

AIMS: To evaluate the developed model using internal and external validation methods. The internal validation methods were bootstrapping and prediction corrected visual predictive check (pc VPC). The external validation was conducted using the new data set, which comprised data collected between 2002 and February 2011.

METHODS: Internal validation was conducted for two possible final models. Model 1 included creatinine clearance (CLcr) in clearance (CL) and height in volume of distribution of the central compartment (V1); model 2 included both CLcr and height in CL and height in V1. Internal validation was performed using NONMEM 7.1 (1) and PsN 3.2.12 (2-3) using 1000 bootstrap samples and prediction-corrected VPCs.. External validation was conducted with the Dutch data using the POSTHOC option in NONMEM to determine population and individual parameter estimates and predicted concentrations. Bias and imprecision were calculated using the Sheiner method (4).

RESULTS: Bootstrap results showed that the parameter estimates of the original model and the replicates coincided well for both models. However, the pc VPC looked slightly better for model 1 compared to model 2. In total, 1452 concentration measurements from 165 patients with 415 courses of therapy were available for external validation. Patients in the validation dataset were significantly older, taller and heavier than patients in the model development dataset but there was no difference in the distributions of serum creatinine. The majority (54%) of the concentrations were sampled mid-dose, 38% were peaks and 9% were troughs. Population concentrations were significantly over estimated with model 1 but not with model 2. Individual concentration predictions and population versus individual parameter estimates were unbiased and precise with both models. There was no difference in the model 2 population parameter estimates obtained from the Glasgow data alone and those obtained with the combined UK and Dutch data.

CONCLUSIONS: Bootstrapping and prediction corrected VPC showed good predictive ability for both models, with a slight preference for model 1. However, validation with a new data set in which the patient‟s characteristics were different demonstrated that model 2, which included both height and estimated CrCl, was more suitable for patients with cystic fibrosis.

REFERENCES: 1. Beal S, Sheiner LB, Boeckman A, Bauer R. NONMEM User's Guides (1989-2009). Ellicott City, MD, USA: Icon Development Solutions2009. 2. Lindbom L, Pihlgren P, Jonsson N. PsN-Toolkit--A collection of computer intensive statistical methods for non-linear mixed effect modeling using NONMEM. Computer Methods and Programs in Biomedicine. 2005;79(3):241-57. 3. Lindbom L, Ribbing J, Jonsson E. Perl-speaks-NONMEM (PsN)- a Perl module for NONMEM related programming. Computer Methods and Programs in Biomedicine. 2004;75:85-94. 4. Sheiner LB, Beal S. Some suggestions for measuring predictive performance Journal of Pharmacokinetics and Biopharmaceutics. 1981;9:503-12. A Pharmacokinetic Study of Ranitidine in a Paediatric Population

P. Westwood, P. Collier, S. Yakkundi

School of Pharmacy, Faculty of Medicine, Health and Life Sciences, Queen’s University of Belfast, Northern Ireland, United Kingdom.

Objectives: Ranitidine is a histamine-2-receptor-antagonist widely used in intensive care units as prophylaxis against stress ulcer syndrome, gastro oesophageal reflux, persistent vomiting and gastric aspiration, or to negate the harmful effects of steroids. The purpose of this study was to use sparse data to investigate the pharmacokinetic (PK) profile of both i.v. (infusion and intermittent bolus dosing) and oral ranitidine in a paediatric population and to determine the influence of patient demographics; age, gender, weight, concomitant drugs and disease states.

Methods: The population PK analysis was performed in NONMEM (v.6.1). Several models were tested including one- and two-compartment disposition and double peak absorption models. Influence of the patient demographics was assessed as both continuous and categorical covariates. Development of the final model was guided by the relevant plots, reduction in the errors, and the change in the objective function using a multi-stage forward and backward stepwise elimination modeling approach. Three methods were used to evaluate the final model for the ranitidine dataset; a variation on the Jack-Knifing technique, Bootstrapping and Principal Component Analysis.

Results: Data from 78 children attending The Royal Belfast Hospital for Sick Children between 1998 and 2006 (mean age 4.57±4.48 years and mean weight 16.27±12.24kg), provided 248 opportunistically drawn samples with a median of 2 samples per patient (range 1 to 13). Conditions were separated into five main categories including stomach surgery and management of heart defects. There were 247 concomitant drug therapies identified from the individual patient records. A one compartment model best described the data. The final parameter estimates for the population were 32.1L/hr (CV 60%) for total clearance and 285L (CV 85%) for volume, both allometrically scaled for a 70kg adult and final estimates for the typical absorption rate constant and bioavailability of 1.31hr-1 and 27.5%, respectively. Weight was the most significant covariate in the model and the presence of heart-related conditions was shown to significantly reduce ranitidine clearance by 54%.

Conclusion: This PK study of ranitidine in a paediatric population found that the presence of a heart condition significantly decreased the clearance, and dose adjustments and careful monitoring are recommended for paediatric patients with heart conditions who are receiving ranitidine.

Using Stochastic Control Methods and Pharmacokinetics to Individualise Drug Therapy: A Case Study with the Enzyme Inhibitor Imatinib

Ben Francis1, Andrea Jorgensen1, Andrea Davies2, Richard Clark2 and Steven Lane1.

1University of Liverpool, Department of Biostatistics. L69 3GS. 2University of Liverpool, Department of Molecular and Clinical Cancer Medicine. L69 3GA.

Background: The list of drugs which exist in healthcare to treat various conditions and diseases is vast with new drugs added every year. Many drugs cause adverse events which are dose dependent, consequently there is a need to identify the correct dose for the patient which minimises the chances of these adverse events developing, while at the same time maximising the efficacy of the drug. However, finding the correct dose is complicated by the inter-individual variability in the pharmacokinetics of each individual patient. Post population analysis is required to classify and account for all the sources of variability in drug dose response. The drug imatinib, in this case study used to treat chronic myelogenous leukaemia, is often administered at the 400mg dose regardless of patient dosing needs with 800mg doses utilized if there is an apparent resistance to the drug.

Aim: To enable greater individualisation of drug therapy for patients, providing informed doses which aim to induce therapeutic plasma concentration levels as quickly as possible.

Methods: The drug dose algorithm is developed from stochastic control methods utilising post population pharmacokinetic data to make response estimates which can then be compared against noisy measurements of the response from each specific patient. Using noisy measurements to update the analysis allows the system to become interactive; ultimately seeking to reduce the overall uncertainty of prediction and providing dose estimates which are tailored to the patient‟s requirements.

Results: In the Imatinib case study, dosage regimens derived by the drug dose algorithm advised alterations in ten out of twelve patients. In simulation, the new dosage regimens kept patient trough levels between 0-5.4% away from the therapeutic trough level of 1000ng/ml. However, continuing with the prescribed dose showed that patient trough levels were between 2.1 and 48.9% away from the therapeutic trough level.

Conclusion: There is an important need for individualised dosing regimens that maximise efficacy of the drug, whilst at the same time minimising the risk of adverse drug reactions; results from this case study show that patients who continue on standard dosing protocol in all but two cases will fail, in theory, to achieve a therapeutic trough level whereas stochastic control methods can be utilized to derive a dosing regimen which will encourage a therapeutic trough level. This methodology can be applied to almost any drug with an established compartmental pharmacokinetic model. Further work will be undertaken to provide clinicians with percentage chances of therapeutic effect and adverse event.

Metabolic drug-drug interactions (DDI) in Paediatric vs. Adults

Farzaneh Salem1, Trevor N. Johnson2, Amin Rostami-Hodjegan1,2

1The University of Manchester, 2SIMCYP Limited, Sheffield, UK

Background: Drug-drug interactions (DDI) are an important and avoidable cause of drug toxicities and occur when one drug (perpetrator) changes the effectiveness or toxicity of another drug (victim drug) in a patient, due to pharmacokinetic (PK) or pharmacodynamic (PD) reasons. Some of the DDI which relate to metabolic inhibition or induction of clearance can be predicted from the in vitro information on the drug (metabolic routes and inhibitory potency etc). However, an accurate prediction of a DDI can be challenging as there are many factors that may be contributing to the observed changes. Many studies and case reports are available on the magnitude of DDIs for various drug combinations when administered in adults but such information is not readily available in paediatrics. Consequently, decisions on handling potential DDIs in paediatrics are largely based on information in adults. Due to discordance between the growth and developmental of various metabolic and excretory elimination routes from birth to adolescents, assumption of DDIs being similar in adults and paediatrics might not be rational.

Aim: The aims of this study were to explore drug combinations that cause DDI in paediatrics, find corresponding DDI in adult populations, and to compare the magnitude of reported DDIs in paediatrics with those in adult populations when the paediatric data are stratified for recognition of any trends.

Methods: A systematic literature review was undertaken to identify reports on paediatric DDIs, these were analysed by the type of study conducted. Most of the DDIs reported in paediatrics were:

 Clinical studies conducted in critically ill children  Reports on toxicity/adverse reaction cause by interactive pairs  Retrospective data analysis (i.e. analysis of previously available pharmacokinetic data on interactive pairs from laboratories and clinical studies) and POPPK studies

Where reports corresponding DDI information, in the form of AUC before and after interaction, were available in both paediatric and adult populations an attempt was made to compare the magnitude of the interaction and explore the possible trends (age, elimination routes etc).

Results: A total of 148 reports on DDIs in paediatric patients were found from the literature review over the age range birth to 20 years, of these 71(47.98%) were prospective clinical studies, 60 (40.54%) were case reports and 17 (11.48%) retrospective analysis of data. The magnitude of DDIs between adults and children was assessed in 23 drug pairs and showed that in 10 vases the magnitude of DDI in paediatrics were higher than adults over the investigated age range. For the remaining cases the level of interaction in paediatrics was similar or lower than in adults. No obvious trend could be recognised.

Conclusion: Care should be given when applying the knowledge of DDI from adults to paediatrics.

Evaluating the effect of using a non-ionic surfactant vesicles on the delivery of luciferin to various sites in the body in real-time

Shaw C. D., Al Gawhari F., Mullen A.B., Coombs G.H, Williams R.A.M., Wiese M., Thomson A.H., and Carter K. C.

Strathclyde Institute of Pharmacy and Biomedical Sciences, University of Strathclyde, 161 Cathedral Street, Glasgow, G4 0RE, UK.

Background: Measuring the amount of bioluminescence released by luciferase-expressing cells, after exposure to luciferin solution, is used as a non-invasive tool to monitor in vivo disease progression. The production of bioluminescence is dependent on the ability of luciferin to reach luciferase-expressing cells and therefore reflects in vivo delivery of luciferin to luciferase-expressing cells within animals. We have shown in previous studies that non- ionic surfactant vesicles (NIV) can be used to modulate the pharmacokinetics of entrapped drugs and thus enhance their uptake by tissues after intravenous injection. It should be possible to monitor the ability of NIV to alter the in vivo pharmacokinetics of luciferin solution by comparing the amount of bioluminescence emitted from animals with luciferase- expressing cells which show different tissue tropisms when luciferin is given as a solution or as luciferin-NIV and by injection or inhalation.

Aim: To compare the pharmacokinetics of luciferin free in solution or as luciferin-NIV in mice treated by intravenous injection or inhalation.

Methods: BALB/c mice, inoculated with three types of luciferase-expressing cells (B16 F0 luc cancer cells which home to the lungs, L. donovani which home to the liver and L. major which remain at the cutaneous site of injection) were treated 5 minutes later with luciferin solution or luciferin-NIV by intravenous injection (IV) or inhalation. The delivery of luciferin to the lungs, liver and footpad of the mice was monitored every two minutes for up to 35 minutes after treatment using an IVIS® Spectrum (Caliper Life Sciences, Runcorn, UK) Thirty minutes after imaging ended, the amount of bioluminescence released from the lungs, liver and footpad was also determined. Bioluminescence AUCs were determined for each site over the first 30 minutes.

Results: Effectiveness of delivery to the lungs was in the order luciferin-NIV IV route > luciferin-solution IV route > luciferin-NIV inhalation = luciferin solution inhalation. In contrast, delivery to the liver was highest for mice given luciferin solution by the IV route with the other 3 treatments being equally effective (p < 0.01 compared to IV luciferin solution, p < 0.0001 compared to both inhaled formulations). The order for delivery to the footpad was luciferin solution IV route > luciferin-NIV IV route > luciferin-NIV inhalation = luciferin-NIV inhalation. There was no difference in the lung AUC for both formulations. IV treatment with luciferin- NIV gave the highest AUC in the liver (p < 0.01 compared to luciferin solution IV, p < 0.0001 compared to both inhaled formulations). Higher AUCs were obtained formulations for the footpad when formulations were given by the IV route compared to inhaled (p < 0.0001), but there was no difference between the AUC for each formulation given by the same delivery route.

Conclusion: In vivo imaging has been used to monitor delivery of luciferin to different locations in mice.

Modeling the sequence-sensitive gemcitabine/docetaxel combination using the Virtual Tumor

1 1 1 2 2 Eric Fernandez , David Orrell , Frances Brightman , Caroline Mignard , Zina Koob , Damien 2 2 2 1 1 France , Nicolas Hoffman , Francis Bichat , David Fell , Christophe Chassagnole .

1Physiomics plc, Oxford, United Kingdom; 2Oncodesign SA, Dijon, France.

Background: In recent years there has been great interest in determining synergistic drug combinations. A difficulty, however, is that the number of different possible schedules increases combinatorially when more than one drug is considered, so it is very hard to anticipate which schedules will have the highest efficacy.

Aim: To determine whether the scheduling predictions made using Physiomics‟ Virtual Tumour would lead to improved responses in xenograft models.

Methods: We have developed a predictive PK-PD tumour model called the “Virtual Tumour” that allows us to rationally design and optimise schedules for anti-cancer drug in isolation or in combination. Experimental data for each drug in isolation were generated to calibrate the model, which have then been used to simulate the outcome of combination regimens.

Results: Here we present the dramatic improvement of the gemcitabine-docetaxel combination regimen in mice bearing MX-1 breast tumours. From PK time courses, xenograft and biomarker data generated experimentally for single injection of each drug in isolation, we have calibrated a Virtual Tumour capable of simulating the outcome of various regimens using the gemcitabine-docetaxel combination and proposed new optimal administration schemas. We accurately predicted a schedule that would lead to an absence of synergy between the two drugs and conversely a schedule that revealed to be 50% more efficient without increasing dose and toxicity.

Conclusion: The Virtual Tumour platform can therefore be used to design new regimens with proprietary compounds as well as standard of care, small molecules or biotherapeutical agents; help test possible schedules for combinations of different drugs that would be effectively impossible to investigate experimentally; and allow prioritisation of the most effective drug combinations.

Optimal Study Design for Oral Ciprofloxacin in Malnourished Children

Wanchana Ungphakorn1, Alison H. Thomson1,2

1Strathclyde Institute of Pharmacy & Biomedical Sciences, University of Strathclyde, Glasgow, 2Pharmacy Dept, Western Infirmary, Pharmacy and Prescribing Support Unit, Glasgow

Background: Oral ciprofloxacin has been considered as an alternative antimicrobial agent for children with severe malnutrition. Recently, a population pharmacokinetic study using sparse data (2-4 samples per patient) was conducted to determine the drug concentration time profile in severely malnourished children [1]. The study was not designed using optimal sampling methodology and therefore the samples may not contain sufficient information.

Aim: To develop an optimal sampling design and sampling windows for future pharmacokinetic studies of oral ciprofloxacin in malnourished children.

Methods: The optimal sampling design was developed based on the population Fisher Information Matrix and modified Fedorov exchange algorithm implemented in the PopDes program [2]. A grid size of 0.25 was used for the optimisation. The structural pharmacokinetic model, pharmacokinetic parameters and their variability were obtained from the previous population pharmacokinetic study [1]. The maximum number of patients to be recruited in the study was fixed at 52 and the number of samples was constrained between 3 to 4 samples per patient. Sampling time limits were set to between 0 to 12 h. Optimal sampling windows with an efficiency of at least 80% compared to the fixed D-optimal time points were also determined.

Results: The previous design was three groups of similar numbers of subjects (n = 17, 18 and 17) and the sampling times in this design were at 2, 4, 8, 24 hours; 3, 5, 8, 12 hours and 1, 3, 6, 10 hours. The percent coefficient of variation (%CV) of Ka, CL and V were 19.7, 5.5 and 6.4, respectively. The optimal design was composed of 2 groups (A and B). The optimal sampling times were 1, 1.5, 3.5, 12 h for group A (n = 34) and 0.75, 1, 2.75, 12 h for group B (n = 18). The expected %CV of Ka, CL and V were 7.8, 5.5 and 6.0, respectively. The efficiency of the optimal design was 211% compared to the previous design. The sampling windows which had at least 80% efficiency were approximately 10 min around the fixed sampling times (D-optimal). The optimal sampling times for a design of one group of 52 subjects were 1, 1.75, 6.5 and 12 h. The expected %CVs for this design were 15.0, 5.5 and 6.4 with an efficiency of 132% and 63% compared to the previous design and optimal design, respectively.

Conclusion: Optimal sampling times and sampling windows have been derived for future pharmacokinetic studies in malnourished children. The results suggest that the previous study design has sufficient information to obtain precise estimates for CL and V but can be simplified by using one or two group of subjects and adjusting the sampling times. If Ka is a parameter of interest, the optimal design is recommended.

References: [1] Thuo, N., et al., Dosing regimens of oral ciprofloxacin for children with severe malnutrition: a population pharmacokinetic study with Monte Carlo simulation. J Antimicrob Chemother, 2011. 66(10): p. 2336-45. [2] Gueorguieva, I., et al., A program for individual and population optimal design for univariate and multivariate response pharmacokinetic-pharmacodynamic models. Comput Methods Programs Biomed, 2007. 86(1): p. 51-61.

Development and application of an in silico pharmacokinetic „post bariatric surgery model‟ in a morbidly obese population to assess drug absorption and metabolism from gastrointestinal tract

A. Darwich1, D. Pade2, K. Rowland-Yeo2, D. Turner2, M. Jamei2, A. Rostami-Hodjegan1

1University of Manchester, 2Simcyp Limited

Background: Bariatric surgical procedures, such as Sleeve Gastrectomy (SG), Roux-en-Y Gastric Bypass (RYGBP) and Biliopancreatic Diversion with Duodenal Switch (BPDDS) involve partial restriction/reduction of the gastrointestinal tract (GIT). They have proven to be clinically effective and cost beneficiary for treating morbid obesity. The invasive nature of these surgeries impacts oral drug bioavailability however the magnitude and direction of the effect has not been consistent. Clinical data to establish underlying factors is sparse.

Aim: To evaluate the change in oral bioavailability of various drugs in morbid obese population „post bariatric surgery‟ through simulation utilising in silico mechanistic physiologically-based pharmacokinetic model.

Methods: The Advanced Dissolution, Absorption, Metabolism (ADAM) model within the Simcyp® Simulator was used to predict bioavailability. The morbidly obese population template within the „Simcyp Library‟ was altered to mimic the characteristics of RYGBP, SG and BPDDS by modifying systems parameters relating to GIT following surgery. Drugs with therapeutic relevance to obese patients (e.g. simvastatin, atorvastatin, omeprazole, diclofenac, cyclosporine, fluconazole and ciprofloxacin) were simulated at varying therapeutic doses (D0) and different small intestinal transit times (SIT). Plasma drug concentration-time (C-t) profiles were compared pre- and post-surgery.

Results: Different drugs showed variations in their sensitivity to different types of surgeries. For example, C-t profiles for post RYGBP simulations indicated that the CYP3A4 substrate simvastatin showed an increase up to 1.22 fold in AUC due to an increased FG (fraction escaping gut wall metabolism). This effect was more pronounced at longer SIT‟s and at low D0. Conversely, C-t profiles for cyclosporine displayed a reduced AUC post RYGBP . BPDDS simulations displayed a pronounced reduction in AUC compared to RYGBP. Simulations for SG did not show significant changes in bioavailability of most drugs from those of pre-surgery values.

Conclusion: The results reflect the variation in the invasiveness of the surgical procedure regarding the impact on GIT and highlights the difference in parameters determining the bioavailability of various drugs at given doses (such as solubility, dissolution, permeability or first-pass gut metabolism) after surgery. ADAM could be a useful tool to identify the most likely effect of these surgeries on different drug relevant factors.