Limited inter-occasion variability in relation to inter-individual variability in chemotherapy-induced myelosuppression

A continued learning approach for model-informed precision dosing: updating models in clinical practice

Model-informed precision dosing (MIPD) is a quantitative dosing framework that combines prior knowledge on the drug-disease-patient system with patient data from therapeutic drug/biomarker monitoring (TDM) to support individualized dosing in ongoing treatment. Structural models and prior parameter distributions used in MIPD approaches typically build on prior clinical trials that involve only a limited number of patients selected according to some exclusion/inclusion criteria. Compared to the prior clinical trial population, the patient population in clinical practice can be expected to include also altered behavior and/or increased interindividual variability, the extent of which, however, is typically unknown. Here, we address the question of how to adapt and refine models on the level of the model parameters to better reflect this real-world diversity. We propose an approach for continued learning across patients during MIPD using a sequential hierarchical Bayesian framework. The approach builds on two stages to separate the update of the individual patient parameters from updating the population parameters. Consequently, it enables continued learning across hospitals or study centers, since only summary patient data (on the level of model parameters) need to be shared, but no individual TDM data. We illustrate this continued learning approach with neutrophil-guided dosing of paclitaxel. The present study constitutes an important step towards building confidence in MIPD and eventually establishing MIPD increasingly in everyday therapeutic use.

Pharmacometrics-Based Considerations for the Design of a Pharmacogenomic Clinical Trial Assessing Irinotecan Safety

PURPOSE

Pharmacometric models provide useful tools to aid the rational design of clinical trials. This study evaluates study design-, drug-, and patient-related features as well as analysis methods for their influence on the power to demonstrate a benefit of pharmacogenomics (PGx)-based dosing regarding myelotoxicity.

METHODS

Two pharmacokinetic and one myelosuppression model were assembled to predict concentrations of irinotecan and its metabolite SN-38 given different UGT1A1 genotypes (poor metabolizers: CLSN-38: -36%) and neutropenia following conventional versus PGx-based dosing (350 versus 245 mg/m2 (-30%)). Study power was assessed given diverse scenarios (n = 50-400 patients/arm, parallel/crossover, varying magnitude of CLSN-38, exposure-response relationship, inter-individual variability) and using model-based data analysis versus conventional statistical testing.

RESULTS

The magnitude of CLSN-38 reduction in poor metabolizers and the myelosuppressive potency of SN-38 markedly influenced the power to show a difference in grade 4 neutropenia (<0.5·109 cells/L) after PGx-based versus standard dosing. To achieve >80% power with traditional statistical analysis (χ2/McNemar's test, α = 0.05), 220/100 patients per treatment arm/sequence (parallel/crossover study) were required. The model-based analysis resulted in considerably smaller total sample sizes (n = 100/15 given parallel/crossover design) to obtain the same statistical power.

CONCLUSIONS

The presented findings may help to avoid unfeasible trials and to rationalize the design of pharmacogenetic studies.

Reinforcement learning and Bayesian data assimilation for model-informed precision dosing in oncology

Model-informed precision dosing (MIPD) using therapeutic drug/biomarker monitoring offers the opportunity to significantly improve the efficacy and safety of drug therapies. Current strategies comprise model-informed dosing tables or are based on maximum a posteriori estimates. These approaches, however, lack a quantification of uncertainty and/or consider only part of the available patient-specific information. We propose three novel approaches for MIPD using Bayesian data assimilation (DA) and/or reinforcement learning (RL) to control neutropenia, the major dose-limiting side effect in anticancer chemotherapy. These approaches have the potential to substantially reduce the incidence of life-threatening grade 4 and subtherapeutic grade 0 neutropenia compared with existing approaches. We further show that RL allows to gain further insights by identifying patient factors that drive dose decisions. Due to its flexibility, the proposed combined DA-RL approach can easily be extended to integrate multiple end points or patient-reported outcomes, thereby promising important benefits for future personalized therapies.

Understanding Hematological Toxicities Using Mathematical Modeling

Balancing antitumor efficacy with toxicity is a significant challenge, and drug-induced myelosuppression is a common dose-limiting toxicity of cancer treatments. Mathematical modeling has proven to be a powerful ally in this field, scaling results from animal models to humans, and designing optimized treatment regimens. Here we outline existing mathematical approaches for studying bone marrow toxicity, identify gaps in current understanding, and make future recommendations to advance this vital field of safety research further.

Model-based prediction of myelosuppression and recovery based on frequent neutrophil monitoring

Purpose

To investigate whether a more frequent monitoring of the absolute neutrophil counts (ANC) during myelosuppressive chemotherapy, together with model-based predictions, can improve therapy management, compared to the limited clinical monitoring typically applied today.

Methods

Daily ANC in chemotherapy-treated cancer patients were simulated from a previously published population model describing docetaxel-induced myelosuppression. The simulated values were used to generate predictions of the individual ANC time-courses, given the myelosuppression model. The accuracy of the predicted ANC was evaluated under a range of conditions with reduced amount of ANC measurements.

Results

The predictions were most accurate when more data were available for generating the predictions and when making short forecasts. The inaccuracy of ANC predictions was highest around nadir, although a high sensitivity (≥90%) was demonstrated to forecast Grade 4 neutropenia before it occurred. The time for a patient to recover to baseline could be well forecasted 6 days (±1 day) before the typical value occurred on day 17.

Conclusions

Daily monitoring of the ANC, together with model-based predictions, could improve anticancer drug treatment by identifying patients at risk for severe neutropenia and predicting when the next cycle could be initiated.

Electronic supplementary material

The online version of this article (doi:10.1007/s00280-017-3366-x) contains supplementary material, which is available to authorized users.

Designing More Efficient Preclinical Experiments: A Simulation Study in Chemotherapy-Induced Myelosupression

A new more efficient preclinical study design (referred to as a compact design) is proposed that removes the need for satellite animals for the collection of toxicokinetic (TK) data by sampling from the main study animals, taking no more than one sample in 24 h to build up a full profile over the course of the study. The compact design’s performance was tested with a simulation study, using an example of chemotherapy-induced myelosupression in rats. Data sets were simulated from a model based on available data, following both the compact design and a traditional design using satellite animals, with 100 studies being simulated for each. The effect of the compact design on parameter and variance estimates for the TK and neutrophil models were investigated, as well as the potential effect of interoccasion variability (IOV). The compact design performed equally as well as the traditional design, and had little impact on parameter or variation estimates, indicating that it would be a suitable alternative to traditional satellite designs while reducing the number of animals required. When IOV was present but not accounted for during the TK analysis some parameter estimates were biased and interindividual variation and residual errors inflated; this was reduced by allowing for IOV in the analysis. Using the compact design removes the need for a satellite group, reducing the number of animals required, without affecting the ability to model the data. If large IOV is suspected, caution should be exercised to avoid parameter estimation bias, and inflation of variability and residual error.

Optimizing oncology therapeutics through quantitative translational and clinical pharmacology: challenges and opportunities

Despite advances in biomedical research that have deepened our understanding of cancer hallmarks, resulting in the discovery and development of targeted therapies, the success rates of oncology drug development remain low. Opportunities remain for objective dose selection informed by exposure-response understanding to optimize the benefit-risk balance of novel therapies for cancer patients. This review article discusses the principles and applications of modeling and simulation approaches across the lifecycle of development of oncology therapeutics. Illustrative examples are used to convey the value gained from integration of quantitative clinical pharmacology strategies from the preclinical-translational phase through confirmatory clinical evaluation of efficacy and safety.

Dose schedule optimization and the pharmacokinetic driver of neutropenia

Toxicity often limits the utility of oncology drugs, and optimization of dose schedule represents one option for mitigation of this toxicity. Here we explore the schedule-dependency of neutropenia, a common dose-limiting toxicity. To this end, we analyze previously published mathematical models of neutropenia to identify a pharmacokinetic (PK) predictor of the neutrophil nadir, and confirm this PK predictor in an in vivo experimental system. Specifically, we find total AUC and Cmax are poor predictors of the neutrophil nadir, while a PK measure based on the moving average of the drug concentration correlates highly with neutropenia. Further, we confirm this PK parameter for its ability to predict neutropenia in vivo following treatment with different doses and schedules. This work represents an attempt at mechanistically deriving a fundamental understanding of the underlying pharmacokinetic drivers of neutropenia, and provides insights that can be leveraged in a translational setting during schedule selection.

Platelet immunology in fungal infections

Up to date, perception of platelets has changed from key players in coagulation to multitaskers within the immune network, connecting its most diverse elements and crucially shaping their interplay with invading pathogens such as fungi. In addition, antimicrobial effector molecules and mechanisms in platelets enable a direct inhibitory effect on fungi, thus completing their immune capacity. To precisely assess the impact of platelets on the course of invasive fungal infections is complicated by some critical parameters. First, there is a fragile balance between protective antimicrobial effects and detrimental reactions that aggravate the fungal pathogenesis. Second, some platelet effects are exerted indirectly by other immune mediators and are thus difficult to quantify. Third, drugs such as antimycotics, antibiotics, or cytostatics, are commonly administered to the patients and might modulate the interplay between platelets and fungi. Our article highlights selected aspects of the complex interactions between platelets and fungi and the relevance of these processes for the pathogenesis of fungal infections.

The shape of the myelosuppression time profile is related to the probability of developing neutropenic fever in patients with docetaxel-induced grade IV neutropenia

PURPOSE

Chemotherapy-induced neutropenia is associated with the risk of developing febrile neutropenia (FN). The aim was to describe the time course of myelosuppression in breast cancer patients treated with docetaxel and to investigate how the shape of the predicted myelosuppression time course and earlier proposed risk factors influence the probability of developing FN.

METHODS

Neutrophil counts from 140 breast cancer patients with observed grade IV neutropenia during the first course of docetaxel treatment were included. Twenty-six of the patients (19%) experienced FN. The myelosuppression time course was described using a semi-mechanistic myelosuppression model in NONMEM. The individual myelosuppression model parameters [baseline neutrophil count, mean transit time (MTT) and drug effect parameter (EC(50))], myelosuppression descriptors (nadir, duration of grade IV neutropenia) and earlier suggested risk factors (age, performance status, haemoglobin and liver function) were explored to be related to FN by logistic regression.

RESULTS

The neutrophil time course following docetaxel treatment was well described by the model. EC(50) and MTT were both significantly related to the probability of developing FN where low parameter values result in a rapid decline, low nadir and an increased risk of FN. None of the evaluated risk factors or myelosuppression descriptors were significant.

CONCLUSION

The probability to develop FN in patients who experience grade IV neutropenia was dependent on the myelosuppression time profile. Patients with a rapid neutrophil decline and high drug sensitivity had a higher probability to develop FN. Model-based parameter estimates were superior predictors over descriptive values such as the nadir or duration of neutropenia.

Population pharmacokinetics of busulfan in children: increased evidence for body surface area and allometric body weight dosing of busulfan in children

PURPOSE

To evaluate the best method for dosing busulfan in children, we retrospectively analyzed two different data sets from three different dosing regimens by means of population pharmacokinetics using NONMEM.

EXPERIMENTAL DESIGN

The development data set consisted of plasma samples from 94 children, in the age range of 0.4 to 18.8 years, receiving either oral or intravenous busulfan. The external model evaluation data set comprised 24 children, in the age range of 0.1 to 18.9 years, who belonged to the once-daily intravenous busulfan dosing regimen. A one-compartment model with first-order absorption using body surface area (BSA) or allometric body weight (BW) as covariate on clearance (CL) and BW as covariate on volume of distribution (V) were used to describe the results sufficiently. In addition to interindividual variability on all pharmacokinetic parameters, interoccasion variability was included for CL and V.

RESULTS

CL values in the present study did not reflect the shape of the CL versus weight curve reported in previous investigations. By external model evaluation, we were able to confirm these findings. Furthermore, bioavailability was calculated to be between 93% and 99% for the development data set. On the basis of the final models, we simulated two dosing schemes according to allometric BW and BSA showing that we estimated to include about 30% more patients into the proposed therapeutic area under the curve (AUC) range of 900 to 1,500 μM*min and could, furthermore, achieve a reduction in the AUC variability when dosed according to the labeled European Medicines Agency (EMA) dosing recommendation.

CONCLUSION

We recommend a BSA or an allometric BW dosing regimen for individualizing busulfan therapy in children to reduce variability in busulfan exposure and to improve safety and efficacy of busulfan treatment.

Two-stage model-based design of cancer phase I dose escalation trials: evaluation using the phase I program of barasertib (AZD1152)

Introduction Modeling and simulation of pharmacokinetics and pharmacodynamics has previously been shown to be potentially useful in designing Phase I programs of novel anti-cancer agents that show hematological toxicity. In this analysis, a two-stage model-based trial design was evaluated retrospectively using data from the Phase I program with the aurora kinase inhibitor barasertib. Methods Data from two Phase I trials and four regimens were used (n = 79). Using barasertib-hydroxy QPA plasma concentrations and neutrophil count data from only study 1A, a PKPD model was developed and subsequently used to predict the MTD and a safe starting dose for the other trials. Results The PKPD model based on data from the first study adequately described the time course of neutrophil count fluctuation. The two-stage model-based design provided safe starting doses for subsequent phase I trials for barasertib. Predicted safe starting dose levels were higher than those used in two subsequent trials, but lower than used in the other trial. Discussion The two-stage approach could have been applied safely to define starting doses for alternative dosing strategies with barasertib. The limited improvement in efficiency for the phase I program of barasertib may have been due to the fact that starting doses for the studied phase I trials were already nearly optimal. Conclusion Application of the two-stage model-based trial design in Phase I programs with novel anti-cancer drugs that cause haematological toxicity is feasible, safe, and may lead to a reduction in the number of patient treated at sub-therapeutic dose-levels.

Predictive ability of a semi-mechanistic model for neutropenia in the development of novel anti-cancer agents: two case studies

In cancer chemotherapy neutropenia is a common dose-limiting toxicity. An ability to predict the neutropenic effects of cytotoxic agents based on proposed trial designs and models conditioned on previous studies would be valuable. The aim of this study was to evaluate the ability of a semi-mechanistic pharmacokinetic/pharmacodynamic (PK/PD) model for myelosuppression to predict the neutropenia observed in Phase I clinical studies, based on parameter estimates obtained from prior trials. Pharmacokinetic and neutropenia data from 5 clinical trials for diflomotecan and from 4 clinical trials for indisulam were used. Data were analyzed and simulations were performed using the population approach with NONMEM VI. Parameter sets were estimated under the following scenarios: (a) data from each trial independently, (b) pooled data from all clinical trials and (c) pooled data from trials performed before the tested trial. Model performance in each of the scenarios was evaluated by means of predictive (visual and numerical) checks. The semi-mechanistic PK/PD model for neutropenia showed adequate predictive ability for both anti-cancer agents. For diflomotecan, similar predictions were obtained for the three scenarios. For indisulam predictions were better when based on data from the specific study, however when the model parameters were conditioned on data from trials performed prior to a specific study, similar predictions of the drug related-neutropenia profiles and descriptors were obtained as when all data were used. This work provides further indication that modeling and simulation tools can be applied in the early stages of drug development to optimize future trials.