dfpk: An R-package for Bayesian dose-finding designs using pharmacokinetics (PK) for phase I clinical trials.

BACKGROUND AND OBJECTIVE: Dose-finding, aiming at finding the maximum tolerated dose, and pharmacokinetics studies are the first in human studies in the development process of a new pharmacological treatment. In the literature, to date only few attempts have been made to combine pharmacokinetics and dose-finding and to our knowledge no software implementation is generally available. In previous papers, we proposed several Bayesian adaptive pharmacokinetics-based dose-finding designs in small populations. The objective of this work is to implement these dose-finding methods in an R package, called dfpk. METHODS: All methods were developed in a sequential Bayesian setting and Bayesian parameter estimation is carried out using the rstan package. All available pharmacokinetics and toxicity data are used to suggest the dose of the next cohort with a constraint regarding the probability of toxicity. Stopping rules are also considered for each method. The ggplot2 package is used to create summary plots of toxicities or concentration curves. RESULTS: For all implemented methods, dfpk provides a function (nextDose) to estimate the probability of efficacy and to suggest the dose to give to the next cohort, and a function to run trial simulations to design a trial (nsim). The sim.data function generates at each dose the toxicity value related to a pharmacokinetic measure of exposure, the AUC, with an underlying pharmacokinetic one compartmental model with linear absorption. It is included as an example since similar data-frames can be generated directly by the user and passed to nsim. CONCLUSION: The developed user-friendly R package dfpk, available on the CRAN repository, supports the design of innovative dose-finding studies using PK information.

Development and performance of npde for the evaluation of time-to-event models.

PURPOSE: Normalised prediction distribution errors (npde) are used to graphically and statistically evaluate mixed-effect models for continuous responses. In this study, our aim was to extend npde to time-to-event (TTE) models and evaluate their performance. METHODS: Let V denote a dataset with censored TTE observations. The null hypothesis (H0) is that observations in V can be described by model M. We extended npde to TTE models using imputations to take into account censoring. We then evaluated their performance in terms of type I error and power to detect model misspecifications for TTE data by means of a simulation study with different sample sizes. RESULTS: Type I error was found to be close to the expected 5% significance level for all sample sizes tested. The npde were able to detect misspecifications in the baseline hazard as well as in the link between the longitudinal variable and the survival function. The ability to detect model misspecifications increased as the difference in the shape of the survival function became more apparent. As expected, the power also increased as the sample size increased. Imputing the censored events tended to decrease the percentage of rejections. CONCLUSIONS: We have shown that npde can be readily extended to TTE data and that they perform well with an adequate type I error.

PharmML in Action: an Interoperable Language for Modeling and Simulation.

PharmML is an XML-based exchange format created with a focus on nonlinear mixed-effect (NLME) models used in pharmacometrics, but providing a very general framework that also allows describing mathematical and statistical models such as single-subject or nonlinear and multivariate regression models. This tutorial provides an overview of the structure of this language, brief suggestions on how to work with it, and use cases demonstrating its power and flexibility.

Model Evaluation of Continuous Data Pharmacometric Models: Metrics and Graphics.

This article represents the first in a series of tutorials on model evaluation in nonlinear mixed effect models (NLMEMs), from the International Society of Pharmacometrics (ISoP) Model Evaluation Group. Numerous tools are available for evaluation of NLMEM, with a particular emphasis on visual assessment. This first basic tutorial focuses on presenting graphical evaluation tools of NLMEM for continuous data. It illustrates graphs for correct or misspecified models, discusses their pros and cons, and recalls the definition of metrics used.

Combined Analysis of Phase I and Phase II Data to Enhance the Power of Pharmacogenetic Tests.

We show through a simulation study how the joint analysis of data from phase I and phase II studies enhances the power of pharmacogenetic tests in pharmacokinetic (PK) studies. PK profiles were simulated under different designs along with 176 genetic markers. The null scenarios assumed no genetic effect, while under the alternative scenarios, drug clearance was associated with six genetic markers randomly sampled in each simulated dataset. We compared penalized regression Lasso and stepwise procedures to detect the associations between empirical Bayes estimates of clearance, estimated by nonlinear mixed effects models, and genetic variants. Combining data from phase I and phase II studies, even if sparse, increases the power to identify the associations between genetics and PK due to the larger sample size. Design optimization brings a further improvement, and we highlight a direct relationship between η-shrinkage and loss of genetic signal.

Pharmacometrics Markup Language (PharmML): Opening New Perspectives for Model Exchange in Drug Development.

The lack of a common exchange format for mathematical models in pharmacometrics has been a long-standing problem. Such a format has the potential to increase productivity and analysis quality, simplify the handling of complex workflows, ensure reproducibility of research, and facilitate the reuse of existing model resources. Pharmacometrics Markup Language (PharmML), currently under development by the Drug Disease Model Resources (DDMoRe) consortium, is intended to become an exchange standard in pharmacometrics by providing means to encode models, trial designs, and modeling steps.

New colistin population pharmacokinetic data in critically ill patients suggesting an alternative loading dose rationale.

Colistin is an old antibiotic that has recently gained a considerable renewal of interest as the last-line defense therapy against multidrug-resistant Gram-negative bacteria. It is administered as colistin methanesulfonate (CMS), an inactive prodrug, and it was shown that due to slow CMS conversion, colistin plasma concentrations increase very slowly after treatment initiation, which constitutes the rationale for a loading dose in critically ill patients. However, faster CMS conversion was observed in healthy volunteers but using a different CMS brand, which may also have a major impact on colistin pharmacokinetics. Seventy-three critically ill patients not undergoing dialysis received multiple doses of CMS. The CMS concentrations were measured by liquid chromatography-tandem mass spectrometry (LC-MS/MS), and a pharmacokinetic analysis was conducted using a population approach. We confirmed that CMS renal clearance and colistin concentrations at steady state are mostly governed by creatinine clearance, but we predict a typical maximum concentration of drug in serum (Cmax) of colistin close to 2 mg/liter, occurring 3 h after an initial dose of 2 million international units (MIU) of CMS. Accordingly, the estimated colistin half-life (t1/2) was relatively short (3.1 h), with rapid attainment of steady state. Our results are only partially consistent with other recently published results. We confirm that the CMS maintenance dose should be adjusted according to renal function in critically ill patients. However, much higher than expected colistin concentrations were observed after the initial CMS dose, with rapid steady-state achievement. These discrepancies challenge the pharmacokinetic rationale for a loading dose, which may still be appropriate for rapid bacterial eradication and an improved clinical cure rate.

White paper: landscape on technical and conceptual requirements and competence framework in drug/disease modeling and simulation.

Pharmaceutical sciences experts and regulators acknowledge that pharmaceutical development as well as drug usage requires more than scientific advancements to cope with current attrition rates/therapeutic failures. Drug disease modeling and simulation (DDM&S) creates a paradigm to enable an integrated and higher-level understanding of drugs, (diseased)systems, and their interactions (systems pharmacology) through mathematical/statistical models (pharmacometrics)(1)-hence facilitating decision making during drug development and therapeutic usage of medicines. To identify gaps and challenges in DDM&S, an inventory of skills and competencies currently available in academia, industry, and clinical practice was obtained through survey. The survey outcomes revealed benefits, weaknesses, and hurdles for the implementation of DDM&S. In addition, the survey indicated that no consensus exists about the knowledge, skills, and attributes required to perform DDM&S activities effectively. Hence, a landscape of technical and conceptual requirements for DDM&S was identified and serves as a basis for developing a framework of competencies to guide future education and training in DDM&S.CPT: Pharmacometrics & Systems Pharmacology (2013) 2, e40; doi:10.1038/psp.2013.16; advance online publication 1 May 2013.

Pharmacokinetic and pharmacodynamic variability of fluindione in octogenarians.

In the PREPA observational study, we investigated the factors influencing pharmacokinetic and pharmacodynamic variability in the responses to fluindione, an oral anticoagulant drug, in a general population of octogenarian inpatients.Measurements of fluindione concentrations and international normalized ratio (INR ) were obtained for 131 inpatients in whom fluindione treatment was initiated. Treatment was adjusted according to routine clinical practice. The data were analyzed using nonlinear mixed-effects modeling, and the parameters were estimated using MONOLI X 3.2. The pharmacokinetics (PK) of fluindione was monocompartmental, whereas the evolution of INR was modeled in accordance with a turnover model (inhibition of vitamin K recycling). Interindividual variability (II V) was very large. Clearance decreased with age and with prior administration of cordarone. Patients who had undergone surgery before the study had lower IC50 values, leading to an increased sensitivity to fluindione. Pharmacokinetic exposure is substantially increased in elderly patients, warranting a lower dose of fluindione.

STAT4 is a confirmed genetic risk factor for Sjögren's syndrome and could be involved in type 1 interferon pathway signaling.

Signal transducer and activator of transcription 4 (STAT4) is a transcription factor mainly activated by interleukin 12, which promotes the secretion of type 2 interferon (IFN) by T-helper 1 cells. We assessed the association of STAT4 gene polymorphism and primary Sjögren's syndrome (pSS) and its functional relevance. We analyzed STAT4 rs7582694 polymorphism in an exploratory cohort of 186 pSS patients and 152 controls, and in a replication cohort of 192 pSS patients and 483 controls, all Caucasian. mRNA levels of STAT4alpha, STAT4beta, STAT1, and the type 1 IFN-induced genes PKR, MX1 and IFITM1 were assessed in peripheral blood mononuclear cells (PBMCs) from 30 pSS patients. STAT4 rs7582694 C allele was associated with pSS in both cohorts (odds ratio (OR) 1.57, 95% confidence interval (CI) 1.27-1.93, P=2.3 x 10(-5)). The association was increased for homozygous subjects, which suggests a recessive effect of the STAT4 at-risk allele. STAT4alpha, STAT4beta and STAT1 mRNA levels in PBMCs were not significantly associated with rs7582694 genotypes, however the mRNA levels of STAT4alpha and type 1 IFN-induced genes were strongly correlated: PKR (P=4 x 10(-3), r=0.51), MX1 (P=2 x 10(-4), r=0.63) and IFITM1 (P=8 x 10(-3), r=0.47), suggesting that STAT4 might be involved in not only type 2 IFN production but also in type 1 IFN-mediated effects.

Bioavailability of diazepam after intramuscular injection of its water-soluble prodrug alone or with atropine-pralidoxime in healthy volunteers.

BACKGROUND AND PURPOSE: The aim of this study was to assess the relative bioavailability of diazepam after administration of diazepam itself or as a water-soluble prodrug, avizafone, in humans. EXPERIMENTAL APPROACH: The study was conducted in an open, randomized, single-dose, three-way, cross-over design. Each subject received intramuscular injections of avizafone (20 mg), diazepam (11.3 mg) or avizafone (20 mg) combined with atropine (2 mg) and pralidoxime (350 mg) using a bi-compartmental auto-injector (AIBC). Plasma concentrations of diazepam were quantified using a validated LC/MS-MS assay, and were analysed by both a non-compartmental approach and by compartmental modelling. KEY RESULTS: The maximum concentration (C(max)) of diazepam after avizafone injection was higher than that obtained after injection of diazepam itself (231 vs. 148 ng.mL(-1)), while area under the curve (AUC) values were equal. Diazepam concentrations reached their maximal value faster after injection of avizafone. Injection of avizafone with atropine-pralidoxime (AIBC) had no effect on diazepam C(max) and AUC, but the time to C(max) was increased, relative to avizafone injected alone. According to the Akaike criterion, the pharmacokinetics of diazepam after injection as a prodrug was best described as a two-compartment with zero-order absorption model. When atropine and pralidoxime were injected with avizafone, the best pharmacokinetic model was a two-compartment with a first-order absorption model. CONCLUSION AND IMPLICATIONS: Diazepam had a faster entry to the general circulation and achieved higher C(max) after injection of prodrug than after the parent drug. Administration of avizafone in combination with atropine and pralidoxime by AIBC had no significant effect on diazepam AUC and C(max).

A single tumour necrosis factor haplotype influences the response to adalimumab in rheumatoid arthritis.

OBJECTIVE: To determine whether tumour necrosis factor (TNF) gene polymorphisms and/or the shared epitope are genetic predictors of the response to adalimumab (ADA) in rheumatoid arthritis (RA). METHODS: This ancillary study to the Research in Active Rheumatoid Arthritis (ReAct) Phase IIIb study included a large cohort of Caucasian patients with RA from France (n = 388) treated with ADA plus methotrexate (MTX) (n = 182), ADA plus any other DMARD (n = 98) or ADA alone (n = 108). The primary outcome was ACR50 at 12 weeks. Patients underwent genotyping for HLA-DRB1 and three TNF gene polymorphisms (-238A/G,-308A/G and-857C/T). Extended haplotypes involving HLA-DRB1 and TNF loci were reconstructed using the PHASE program. RESULTS: A total of 151 patients (40%) had an ACR50 response at week 12. Neither the number of HLA-DRB1 shared epitope copies nor presence of the three TNF polymorphisms tested separately was significantly associated with ACR50 response at week 12. However, haplotype reconstruction of the TNF locus revealed that the GGC haplotype (-238G/-308G/-857C) in a homozygous form (i.e. present in more than half of the patients) was significantly associated with a lower ACR50 response to ADA at 12 weeks (34% vs. 50% in patients without the haplotype) (p = 0.003; pa = 0.015). This effect was more important in the subgroup of patients concomitantly treated with MTX. CONCLUSION: This large pharmacogenetic study provides preliminary data indicating that a single TNF locus haplotype (-238G/-308G/-857C), present on both chromosomes is associated with a lower response to ADA, mainly in patients treated with ADA and MTX.

Overview of model-building strategies in population PK/PD analyses: 2002-2004 literature survey.

AIMS: A descriptive survey of published population pharmacokinetic and/or pharmacodynamic (PK/PD) analyses from 2002 to 2004 was conducted and an evaluation made of how model building was performed and reported. METHODS: We selected 324 articles in Pubmed using defined keywords. A data abstraction form (DAF) was then built comprising two parts: general characteristics including article identification, context of the analysis, description of clinical studies from which the data arose, and model building, including description of the processes of modelling. The papers were examined by two readers, who extracted the relevant information and transmitted it directly to a MySQL database, from which descriptive statistical analysis was performed. RESULTS: Most published papers concerned patients with severe pathology and therapeutic classes suffering from narrow therapeutic index and/or high PK/PD variability. Most of the time, modelling was performed for descriptive purposes, with rich rather than sparse data and using NONMEM software. PK and PD models were rarely complex (one or two compartments for PK; E(max) for PD models). Covariate testing was frequently performed and essentially based on the likelihood ratio test. Based on a minimal list of items that should systematically be found in a population PK-PD analysis, it was found that only 39% and 8.5% of the PK and PD analyses, respectively, published from 2002 to 2004 provided sufficient detail to support the model-building methodology. CONCLUSIONS: This survey allowed an efficient description of recent published population analyses, but also revealed deficiencies in reporting information on model building.

An in vivo pharmacokinetic/pharmacodynamic model for antiretroviral combination.

PURPOSE: The objective of this study was to investigate the pharmacokinetic/pharmacodynamic properties of an antiretroviral therapy in HIV-infected patients. METHOD: Eight HIV-infected patients received zidovudine, lamivudine, and indinavir as their first antiretroviral treatment. Pharmacokinetic data were analyzed separately using a one-compartmental model with first-order absorption, and the individual estimates were used to simulate drug concentrations. To determine the relationship between drug concentrations and the antiviral effect, an in vitro E(max) model was tested. Alternatively, a dynamic model was built describing the viral and cellular pathophysiology, including the turnover of viral replication in infected cells and the production of virus under treatment. RESULT: The E(max) model fit poorly the experimental data. The complex model was not identifiable with the data available in this study, however a simplified model allowed us to estimate the pharmacodynamic parameters reflecting the decrease of both infected cells and viral load under antiretroviral treatment. CONCLUSION: Using potent highly active antiretroviral therapy, the treatment was so effective that it was not possible to estimate the parameters of the relationship between drug concentrations and the reduction of viral load. Even if the relationship does exist, the direct response model of antiretroviral agents cannot be demonstrated, however the simplified model provides an understanding of the synergy of such a combination and offers suggestions as how to prevent the emergence of viral resistant strains.

Evaluation of tests based on individual versus population modeling to compare dissolution curves.

This study is aimed at comparing dissolution curves obtained for several batches of sustained release formulations of octreotide, a somatostatin analog used in acromegaly. We evaluated four tests to compare the parameters of two series of curves: a Mann-Whitney-Wilcoxon test comparing individual estimates; a likelihood ratio test after population analysis (NONMEM), and two Wald tests using population parameters estimated by two-stage methods. The four approaches were evaluated by simulations. The Mann-Whitney-Wilcoxon test and the population approach had adequate type-I error, and the latter provided the highest power under the alternatives simulated. The two-stage method including the uncertainty on individual estimates was disappointing.

Prediction of fluindione maintenance dosage hampered by large intraindividual variability.

In a previous study, the authors proposed a method to individualize fluindione dosage regimen, based on a pharmacokinetic/pharmacodynamic model describing the evolution of the International Normalized Ratio (INR). In this method, daily maintenance dosage for a target INR depends on the product of individual Cl and C50. The present work shows the results of a follow-up study in 50 patients for whom target INR was 2.5. INR measurements and dosage regimens were recorded both during hospital stay and during the 1st month of treatment. Patients were defined as equilibrated after 1 month if the last two INRs were in the range 1.5-3.5 under a stable dosage regimen. Actual maintenance dose was compared with the dose predicted using the three first INRs measured in the hospital. Intraindividual variability of Cl*C50 between hospital stay and after 1 month was evaluated. After 1 month, only 27 patients (54%) were equilibrated. Actual maintenance dose varied from 5 to 30 mg daily. There was no bias between predicted and actual maintenance dose (1.4 mg), but a large root mean squared error (8 mg) was found. The intraindividual variability in Cl*C50 between hospital and maintenance regimen was high (93%), which may explain the dispersion in the predicted maintenance dose.

Modeling the kinetics of release of octreotide from long-acting formulations injected intramuscularly in rabbits.

Long-acting repeatable formulations (LAR) based on polymeric microspheres were manufactured to deliver octreotide, an octapeptide analogue used in acromegaly. We developed a model to describe the complex triphasic concentration versus time profile observed in rabbits after intramuscular (i.m.) injection of these LAR formulations. A 5-mg x kg(-1) dose of octreotide in a reference LAR formulation was given im to eight rabbits; two groups of four rabbits each received a different formulation. In each animal, 26 blood samples were taken over 49 days. Concentrations of octreotide were assayed by radioimmunoassay. A model describing the concentration profile was developed. Octreotide release was described using the succession of an exponential model, a semiempirical non-Fickian model, and a delayed Weibull model. Parameters were estimated using nonlinear regression, first for the eight rabbits that received the reference formulation and then for the other eight animals. The model provides an adequate description of the concentration versus time profile for the three formulations. For the reference phase, erosion accounted for 87% of total drug release. The formulation encapsulating an octreotide-acetate complex showed a prolonged diffusion phase.

Nonparametric analysis of the absorption profile of octreotide in rabbits from long-acting release formulation OncoLAR.

Octreotide (octreotide-acetate, Sandostatin(R)) is a somatostatin analogue, used in long-term treatment of acromegaly. The present study describes the absorption profile in rabbits of octreotide after release from the long-acting formulation OncoLAR (denoted as octreotide-LAR). In a first experiment, the disposition kinetics of octreotide was studied for 24 h in six rabbits after intravenous (i. v.) injection of 0.025 mg of a solution of octreotide. In a second experiment, release kinetics was studied in eight rabbits for 49 days after an i.m. injection of 5 mg/kg of octreotide-LAR. Concentrations were determined by radioimmunoassay. After i.v. injection of octreotide, one- and two-compartment models were compared for each rabbit. A typical disposition profile was computed using the mean parameters. After i.m. injection of octreotide-LAR, deconvolution was performed using the point-area method. Individual absorption profiles were characterised using natural splines. The number of breakpoints was selected using the generalised cross-validation criterion. The two compartment model was selected based on the i.v. study. After i.m. administration, octreotide exhibited a triphasic absorption profile, with large interindividual variability. A transient peak followed the initial burst phase. The third phase covered 85% of total drug released. The approach allows a model-independent description of the in vivo absorption profile of octreotide-LAR.

Modeling INR data to predict maintenance fluindione dosage.

This study was designed to construct a pharmacokinetic/pharmacodynamic model describing the evolution of International Normalized Ratio (INR) under oral anticoagulation treatment by fluindione in patients and to develop a method for individualization of fluindione dosage. Three indirect response models describing the concentration-INR relationship were tested using a nonparametric estimation method. INR was modelled as a quantity being produced and eliminated. According to a log-likelihood ratio test, the evolution of INR was best modelled as an inhibition of its elimination by fluindione. The selected model was evaluated in 24 additional patients with INR measurements (after 2, 3, 4, 6, and 10 doses). Using a Bayesian method with data until day 4, INR was correctly predicted for days 6 and 10. The population characteristics of fluindione were estimated, pooling the two groups of patients. A Bayesian method for individualization of dosage regimen was developed, based on a risk function for INR at steady state. Prescription rules for fluindione were derived using this method retrospectively on the 73 patients in this study.