Search

When a candidate drug enters clinical trials, decisions regarding dosing are mainly based on animal data. Occasionally, toxicity problems are faced in the clinic because of unexpected species differences in pharmacokinetics or pharmacodynamics between humans and preclinical species. Fludarabine and topotecan are examples of such drugs. In the first clinical trials of the new agent CHS 828, the maximum tolerated dose was reached earlier than expected from animal data. This paper discusses the issue of species differences in the development of anticancer drugs, and preclinical models for detection and quantification of such differences. Pharmacokinetic and hematological toxicity data of CHS 828 from studies in rats and humans are presented. In vitro sensitivity to CHS 828 and some established cytotoxic agents was measured in lymphocytes from humans and rats and in a panel of human and rodent cell-lines. 10–100 times higher CHS 828 exposure was tolerated by rats than by patients. In both in vitro cell systems, CHS 828 showed higher potency in human cells compared to rodent cells. A species difference was evident also for fludarabine, but not for doxorubicin and cisplatin. CHS 828 pharmacokinetics were similar across species. In conclusion, the lower tolerance of CHS 828 in humans than in rats could be detected in vitro in cultures of peripheral lymphocytes. Preclinical studies of species differences could help the interpretation of in vivo effect studies as well as the choice of starting dose for clinical trials. We suggest peripheral lymphocytes from different species as a potential model system for such studies.

As myelosuppression is the dose-limiting toxicity for most chemotherapeutic drugs, modelers attempt to find relationships between drug and toxicity to optimize treatment. Mechanistic models, i.e. models based on physiology and pharmacology, are preferable over empirical models, as prior information can be utilized and as they generally are more reliable for extrapolations. To account for different dosing-regimens and possible schedule-dependent effects, the whole concentration-time profile should be used as input into the pharmacokinetic-pharmacodynamic model. It is also of importance to model the whole time course of myelosuppression to be able to predict both the degree and duration of toxicity as well as consecutive courses of therapy. A handful of (semi)-mechanistic pharmacokinetic-pharmacodynamic models with the above properties have been developed and are reviewed. Ideally, a model of myelosuppression should separate drug-specific parameters from system related parameters to be applicable across drugs and useful under different clinical settings. Introduction of mechanistic models of myelosuppression in the design and evaluation of clinical trials can guide in the decision of optimal sampling times, contribute to knowledge of optimal doses and treatment regimens at an earlier time point and identify sub-groups of patients at a high risk of myelosuppression.

Purpose: Quantitative relationships between treatment-induced changes in tumor size and circulating tumor cell (CTC) counts, and their links to overall survival (OS), are lacking. We here present a population modeling framework identifying and quantifying such relationships, based on longitudinal data collected in patients with metastatic colorectal cancer (mCRC) to evaluate the value of tumor size and CTC counts as predictors of OS. Experimental design: A pharmacometric approach (i.e., population pharmacodynamic modeling) was used to characterize the changes in tumor size and CTC count and evaluate them as predictors of OS in 451 patients with mCRC treated with chemotherapy and targeted therapy in a prospectively randomized phase 3 study (CAIRO2). Results: A tumor size model of tumor quiescence and drug-resistance, was used to characterize the tumor size time-course, and was, in addition to the total normalized dose (i.e., of all administered drugs) in a given cycle, related to the CTC counts through a negative binomial model (CTC model). A CTC count≥3/7.5 mL (hazard ratio=3.51, 95% confidence interval: 2.85-4.32), as described by the CTC model, was a better predictor of OS than tumor size changes. The modeling framework was applied to explore if dose-modifications (increased and reduced) would result in a CTC count below 3/7.5 mL after 1-2 weeks of treatment. Conclusions: Time-varying CTC counts can be useful for early predicting OS in patients with mCRC, and may therefore have potential for model-based treatment individualization. Although tumor size had a strong connection to CTC, its link to OS was weaker.

Determination of a maximum tolerated dose (MTD) continues to mainly rely on dose escalation studies using the algorithm-based 3+3 design although it has repeatedly been shown to result in a biased and imprecise MTD. Alternative Bayesian methods, i.e. the Continuous Reassessment Method (CRM), the Escalation with Overdose Control (EWOC), the Bayesian Logistic Regression Model (BLRM), and the modified Toxicity Probability Intervals (mTPI) are increasingly gaining interest for Phase I studies. Here we propose to develop an in silico Clinical Trial Simulation (CTS) framework for multiple comparisons of MTD determination and to highlight the potential benefits for model-based methods. This groundwork was exemplified for a combination therapy in which thrombocytopenia was the most frequent Dose Limiting Toxicity (DLT) in two 3+3 dose escalation trials in solid tumors and in ovarian cancer. The recommended Phase II dose (RP2D) was assessed through simulations from a thrombocytopenic toxicity PKPD model developed using the data of these two trials. Dose finding designs (3+3, CRM, EWOC, BLRM and mTPI) were evaluated for accuracy and precision of the predicted RP2D, percentage of DLTs, proportion of under- and over- dosing patients and dose escalation trajectory. Using this framework, the Bayesian methods were shown to be in better agreement with the reference model-based RP2D and provided an increase of 2 dose levels compared to the 3+3 design approach. Furthermore, they provided a better precision of the RP2D and yielded to more ethical trials. This work is in line with the methodology shift advocated by regulators and academics in phase I oncology studies.

Anticancer regimens are often a delicate compromise between dose intensity and acceptable toxicity, for example neutropenia. The aim of the present study was to develop a theoretical framework using optimal design theory to select the optimal dosing and sampling based on several criteria derived from the predicted neutrophil counts. A semi-physiological PK/PD model for docetaxel's hematological toxicity was used to determine the population typical nadir value of absolute neutrophil count and the time of occurrence of the nadir. An optimization on both time and size of dosing was performed in PopED v.2.11.The optimizations maximized the expected nadir value given a set of clinical criteria using a penalty function. Sampling schedules were also optimized to allow for model identification of the nadir value using D- , C-, MAP-optimal criteria and by using a Sample Reuse Simulation approach. Optimized dosing schedules were found to expose fewer patients to grade 4 neutropenia and total dose could be further increased with recommended dosing intervals. Predicted population nadir was more precisely estimated with a D-optimal design while sampling a true nadir value was more frequently done with a design derived from a Sample Reuse Simulation method. Optimal design methodology can be applied for toxicity monitoring within clinical constraints in oncology studies.

Background and purpose: Infection with Gram-negative bacteria and the immune system’s subsequent recognition of the potent membrane-bound activator endotoxin (ETX), can lead to persistent immune activation. The purpose of the current work was to develop a model-based description of the time-course of ETX concentrations and its effect on the release of the pro-inflammatory, cytokines tumor necrosis factor α (TNF-α) and interleukin 6 (IL-6). Methods: Non-linear mixed effects models were developed based on data from experimental studies in a porcine model of sepsis. Intravenous infusions of E. coli ETX were administered to the piglets at different doses and durations of infusion. ETX, TNF-α and IL-6 plasma levels were measured throughout the infusion time. Results: The concentration-time profile of ETX was well described with a one-compartment model with non-linear elimination. Observation of contamination in early ETX measurements was handled by initializing the central compartment to an estimated parameter. The concentration of ETX over time was used as the driver of the inflammatory response. An indirect response model with ETX stimulated production (Emax model), delayed through a transit chain (three compartments) was used to describe the observed cytokine concentration-time profiles. To describe tolerance to ETX exposure, an exponential time-dependent increase was added to the parameter describing the potency of ETX to stimulate cytokine release (EC50). Conclusions: A mathematical model was developed to depict the time-course of ETX in plasma and its induction of the two immune response markers TNF-α and IL-6. This model-based approach is unique in its description of the three time-courses, and may later be expanded to better understand immune cell release in bacterial infections and sepsis-type pathophysiological changes

Objectives: Colistin is used as a salvage therapy for multidrug-resistant Gram-negative bacterial infections and administered as a prodrug, colistimethate sodium (CMS). Characterizing distribution of colistin at the site of infection is important to optimize bacterial killing. The aims of this analysis were (i) to apply a whole-body physiologically based pharmacokinetic (WPBPK) model structure to describe the pharmacokinetics (PK) of CMS and colistin in critically ill patients and (ii) to predict colistin concentration-time courses and bacterial killing in target tissues combining the WBPBPK model with a semi-mechanistic pharmacokinetic-pharmacodynamic (PKPD) model. Methods: 27 critically ill patients treated with colistin were included in the analysis. A WBPBPK model previously developed in rat was applied to describe CMS and colistin PK data. The model was used to predict tissue concentrations in lungs, skin, blood and kidneys to drive a semi-mechanistic PKPD model on a wild-type (ATCC 27853) or a meropenem-resistant (AUR552) clinical strain P. aeruginosa to predict bacterial killing following the original dosing regimen and by replacing the original initial dose with a loading dose of 9MU. Results: The plasma data were reasonably well described by the WBPBPK model for both CMS and colistin with a slight overprediction at the 1st occasion. High exposure was predicted in kidneys comparable to what had been predicted in previous studies, in rat and healthy subjects. Bacterial load was quickly cleared for both the ATCC 27853 and ARU552 strains in all tissues and at a higher extend in kidney tissue, for all dosing scenarios. Conclusion: The WPBPK model was able to adequately describe the PK of CMS and colistin in critically ill patients. The combination of the predicted PK profiles in tissues of interest with a PKPD model was able to predict the bactericidal effect of colistin at target sites.

Background and purpose Colistin is a last-line antibiotic administered as the prodrug colistin methanesulfonate (CMS) for the treatment of multidrug resistant Gram-negative bacterial infections. Whole-body physiologically based pharmacokinetic (WBPBPK) models are valuable tools to understand and characterize drug disposition, predict tissue distribution and interpret exposure-response relationship. The aim of this work was to develop a WBPBPK model for colistin and CMS in five animal species and evaluate the utility of the model for predicting colistin and CMS disposition in human. Methods A nonlinear mixed-effects WBPBPK model previously developed in rats was extended to describe CMS and colistin plasma data of animals from 5 different species (40 mice, 6 rats, 3 rabbits, 3 baboons and 2 pigs) that had received single doses of CMS. CMS renal clearance and hydrolysis to colistin were allometrically scaled based on glomerular filtration rate (GFR) and tissue volumes, respectively. For the non-renal colistin clearance, three scaling models were evaluated: volume based allometric scaling, volume and maximum lifespan potential (MLP) based allometric scaling, and estimation of specie-specific parameters. Tissue concentrations were predicted for all species. The WBPBPK model was then used to predict human plasma concentrations, which were compared to observed human plasma PK data extracted from literature. Results The description of the plasma PK of CMS and colistin in mice, rats, rabbits, baboons and pigs was satisfactory. The volume and MLP based allometric scaling of the non-renal clearance of colistin was best at characterizing colistin concentration-time course, even if a misprediction remained in pigs. In human however, allometric scaling without MLP was closest to the observed data, with satisfactory prediction of the CMS plasma profiles and a slight overprediction of colistin plasma PK profiles. Conclusions Interspecies WBPBPK models were developed to describe the

Colistin is a polymyxin antibiotic which is used to treat patients infected with multidrug resistant Gram negative bacteria (MDR-GNB). The objective of this work was to develop a whole-body physiologically based pharmacokinetic (WBPBPK) model in rat for colistin and its prodrug, CMS. The Kp prior values of colisin and CMS used in the WBPBPK model were either measured ex vivo in rat tissue homogenates or calculated using an in silico model. The PK parameters were estimated fitting plasma concentrations from rats receiving an i.v. bolus of CMS. In the WBPBPK model, the tissue distribution was assumed to be well-stirred and perfusion-limited. Three scenarios were investigated: estimating the Kp values using in silico Kp prior values (I), estimating the Kp values using the Kp prior values from ex vivo experiments (II) and fixing the Kp values to the experimental ex vivo Kp values (III). The WBPBPK model well described CMS and colistin plasma concentration-time profiles. Colistin Kp values in kidneys were higher than in the other tissues. The predicted concentrations in tissue were highest for kidneys and brain which might be due to a high affinity for these tissues and/or active transport processes that remain poorly elucidated. The clearance estimates of CMS and colistin were in agreement with previously reported values in the literature. The model developed in this study might be a valuable tool in drug development to understand the disposition of colistin or new polymyxin candidates as well as to guide for optimal dosing regimens.

Objectives: The primary aim of this work was to develop a whole-body physiologically based pharmacokinetic (WBPBPK) model to describe CMS and colistin disposition in human plasma and urine. The secondary aim of this analysis was to use the WBPBPK model to predict CMS and colistin tissue distribution in typical individuals with different pathophysiological changes and receiving different dosing regimens. Methods: Twelve healthy males were included in the analysis. They received a single dose of 80 mg CMS (1 million unit) through a 1-h intravenous infusion. Venous blood was collected between 0 and 18 h post dose. Fractionated urine samples were collected between 0 and 24 h after dose. A WBPBPK model initially developed in rat was further detailed with the addition of a specific urinary tract (UT) model. The Kp values of CMS and colistin were estimated for all tissues using experimental Kp prior values from rat tissue homogenates. Results: The model adequately described CMS and colistin concentrations over time in plasma and in urine. A shared first order elimination rate constant was estimated to depict the hydrolysis of CMS in plasma and tissues. A separate hydrolysis rate constant for CMS was estimated in urine, and was lower than in plasma and tissues. A shared non-renal elimination rate constant of colistin was estimated in plasma and in tissues. CMS and colistin disposition in urine was well characterized by the UT model. The tubular reabsorption of colistin was best described by a saturable model estimating the colistin affinity constant, KM. Non-specific binding of colistin in the UT lumen was accounted for using a linear relationship. Conclusion: The WBPBPK developed in this study characterized plasma and urine PK of CMS and colistin in human well. This model was used as a new framework to predict colistin exposure in the tissues of interest under different physiological conditions. The model can be easily refined when new data are available and can be combine