A comprehensive regulatory and industry review of modeling and simulation practices in oncology clinical drug development.

Exposure-response (E-R) analyses are an integral component in the development of oncology products. Characterizing the relationship between drug exposure metrics and response allows the sponsor to use modeling and simulation to address both internal and external drug development questions (e.g., optimal dose, frequency of administration, dose adjustments for special populations). This white paper is the output of an industry-government collaboration among scientists with broad experience in E-R modeling as part of regulatory submissions. The goal of this white paper is to provide guidance on what the preferred methods for E-R analysis in oncology clinical drug development are and what metrics of exposure should be considered.

Model enhanced reinforcement learning to enable precision dosing: A theoretical case study with dosing of propofol.

Extending the potential of precision dosing requires evaluating methodologies offering more flexibility and higher degree of personalization. Reinforcement learning (RL) holds promise in its ability to integrate multidimensional data in an adaptive process built toward efficient decision making centered on sustainable value creation. For general anesthesia in intensive care units, RL is applied and automatically adjusts dosing through monitoring of patient's consciousness. We further explore the problem of optimal control of anesthesia with propofol by combining RL with state-of-the-art tools used to inform dosing in drug development. In particular, we used pharmacokinetic-pharmacodynamic (PK-PD) modeling as a simulation engine to generate experience from dosing scenarios, which cannot be tested experimentally. Through simulations, we show that, when learning from retrospective trial data, more than 100 patients are needed to reach an accuracy within the range of what is achieved with a standard dosing solution. However, embedding a model of drug effect within the RL algorithm improves accuracy by reducing errors to target by 90% through learning to take dosing actions maximizing long-term benefit. Data residual variability impacts accuracy while the algorithm efficiently coped with up to 50% interindividual variability in the PK and 25% in the PD model's parameters. We illustrate how extending the state definition of the RL agent with meaningful variables is key to achieve high accuracy of optimal dosing policy. These results suggest that RL constitutes an attractive approach for precision dosing when rich data are available or when complemented with synthetic data from model-based tools used in model-informed drug development.

Exposure driven dose escalation design with overdose control: Concept and first real life experience in an oncology phase I trial.

BACKGROUND: Escalation With Overdose Control (EWOC) designs are increasingly used to ensure dose-toxicity curve of investigational oncology drugs is efficiently characterized during dose escalation steps. We propose a novel EWOC-based method that integrates the longitudinal pharmacokinetic (PK) data of individual patients in a Bayesian forecasting exposure-safety framework. METHODS: The method, called exposure-driven EWOC (ED-EWOC), relies on a population PK model coupled with a Bayesian logistic regression model to make dose recommendation for the next cohort of patients. RESULTS: We applied ED-EWOC to a real oncology clinical trial in parallel to a traditional EWOC approach. We found that for comparable priors, ED-EWOC dose recommendations were equivalent to the one suggested by EWOC when PK is dose proportional with low inter-individual variability. CONCLUSION: This case example demonstrates that ED-EWOC is logistically feasible during a trial conduct when PK bioanalysis can be expedited in the dose escalation phase. Overall, we anticipate that exposure-guided Bayesian designs could benefit patients and drug developers to identify the optimal dose steps of novel compounds entering the clinic with suspected liability in PK or that exhibit large inter-individual variability.

Step-by-step comparison of ordinary differential equation and agent-based approaches to pharmacokinetic-pharmacodynamic models.

Mathematical models in oncology aid in the design of drugs and understanding of their mechanisms of action by simulation of drug biodistribution, drug effects, and interaction between tumor and healthy cells. The traditional approach in pharmacometrics is to develop and validate ordinary differential equation models to quantify trends at the population level. In this approach, time-course of biological measurements is modeled continuously, assuming a homogenous population. Another approach, agent-based models, focuses on the behavior and fate of biological entities at the individual level, which subsequently could be summarized to reflect the population level. Heterogeneous cell populations and discrete events are simulated, and spatial distribution can be incorporated. In this tutorial, an agent-based model is presented and compared to an ordinary differential equation model for a tumor efficacy model inhibiting the pERK pathway. We highlight strengths, weaknesses, and opportunities of each approach.

Association of circulating protein biomarkers with clinical outcomes of durvalumab in head and neck squamous cell carcinoma.

The potential for durvalumab, a programmed cell death ligand-1 (PD-L1)-blocking monoclonal antibody, to treat head and neck squamous cell carcinoma (HNSCC) is being evaluated in multiple clinical trials. We assessed circulating proteins at baseline to identify potential biomarkers and to understand pathways related to clinical outcomes for durvalumab. Prior to treatment, 66 serum proteins were measured using multiplex immunoassays for 158 durvalumab-treated HNSCC patients in the phase II HAWK and CONDOR trials as a discovery dataset and 209 durvalumab-treated HNSCC patients in the phase III EAGLE trial as a validation dataset. Multivariate Cox modeling of HAWK and CONDOR datasets established that higher baseline concentrations of interleukin-6 (IL-6), C-reactive protein, S100 calcium-binding protein A12, and angiopoietin-2 (ANGPT2) were associated with shorter overall survival (OS), while higher concentrations of osteocalcin correlated with longer OS after durvalumab treatment (p < .05). All five proteins remained significantly correlated with OS after adjusting for baseline clinical factors, with consistent results across clinical efficacy endpoints based on univariate correlation analyses. The validation dataset from the EAGLE trial confirmed the independent association of IL-6 and osteocalcin with OS, and preserved directional trends for the other biomarkers identified in the discovery dataset. Our results demonstrate the important role of immunosuppressive proteins in the resistance of HNSCC to durvalumab treatment. Osteocalcin showed a positive correlation with clinical outcomes, which remains to be further investigated.

Pharmacokinetic-Pharmacodynamic Modelling of Systemic IL13 Blockade by Monoclonal Antibody Therapy: A Free Assay Disguised as Total.

A sequential pharmacokinetic (PK) and pharmacodynamic (PD) model was built with Nonlinear Mixed Effects Modelling based on data from a first-in-human trial of a novel biologic, MEDI7836. MEDI7836 is a human immunoglobulin G1 lambda (IgG1λ-YTE) monoclonal antibody, with an Fc modification to reduce metabolic clearance. MEDI7836 specifically binds to, and functionally neutralizes interleukin-13. Thirty-two healthy male adults were enrolled into a dose-escalation clinical trial. Four active doses were tested (30, 105, 300, and 600 mg) with 6 volunteers enrolled per cohort. Eight volunteers received placebo as control. Following single subcutaneous administration (SC), individual time courses of serum MEDI7836 concentrations, and the resulting serum IL13 modulation in vivo, were quantified. A binding pharmacokinetic-pharmacodynamic (PK-PD) indirect response model was built to characterize the exposure-driven modulation of the target over time by MEDI7836. While the validated bioanalytical assay specification quantified the level of free target (i.e., a free IL13 assay), emerging clinical data suggested dose-dependent increase in systemic IL13 concentration over time, indicative of a total IL13 assay. The target time course was modelled as a linear combination of free target and a percentage of the drug-target complex to fit the clinical data. This novel PK-PD modelling approach integrates independent knowledge about the assay characteristics to successfully elucidate apparently complex observations.

Early predictions of response and survival from a tumor dynamics model in patients with recurrent, metastatic head and neck squamous cell carcinoma treated with immunotherapy.

We developed and evaluated a method for making early predictions of best overall response (BOR) and overall survival at 6 months (OS6) in patients with cancer treated with immunotherapy. This method combines machine learning with modeling of longitudinal tumor size data. We applied our method to data from durvalumab-exposed patients with recurrent/metastatic head and neck cancer. A fivefold cross-validation was used for model selection. Independent trial data, with various degrees of data truncation, were used for model validation. Mean classification error rates (90% confidence intervals [CIs]) from cross-validation were 5.99% (90% CI 2.98%-7.50%) for BOR and 19.8% (90% CI 15.8%-39.3%) for OS6. During model validation, the area under the receiver operating characteristic curves was preserved for BOR (0.97, 0.97, and 0.94) and OS6 (0.85, 0.84, and 0.82) at 24, 18, and 12 weeks, respectively. These results suggest our method predicts trial outcomes accurately from early data and could be used to aid drug development.

Efficacy and Tolerability of Tremelimumab in Locally Advanced or Metastatic Urothelial Carcinoma Patients Who Have Failed First-Line Platinum-Based Chemotherapy.

PURPOSE: Patients with advanced urothelial carcinoma who fail platinum-containing chemotherapy (treatment fails) have a poor prognosis and limited treatment options. Recent approvals of immune-checkpoint inhibitors confirmed the value of immunomodulatory therapy in urothelial carcinoma. Tremelimumab is a selective human immunoglobulin G2 (IgG2) monoclonal antibody against cytotoxic T-lymphocyte-associated antigen 4 with demonstrated durable response rate in metastatic melanoma. This is the first study to report the efficacy and safety of tremelimumab in urothelial carcinoma. PATIENTS AND METHODS: We report the results of the urothelial carcinoma cohort from a phase II, open-label, multicenter study of patients with advanced solid tumors (NCT02527434). Patients with locally advanced/metastatic urothelial carcinoma were treated with tremelimumab monotherapy (750 mg via intravenous infusion every 4 weeks for seven cycles, then every 12 weeks for two additional cycles) for up to 12 months or until disease progression, initiation of other anticancer therapy, unacceptable toxicity, or consent withdrawal. RESULTS: In 32 evaluable patients with metastatic urothelial carcinoma, objective response rate was 18.8% (95% confidence interval, 7.2-36.4), including complete response (CR) in 2 (6.3%), and partial response in 4 patients (12.5%). Median duration of response has not been reached. Stable disease of ≥12 months was reported in 1 patient (3.1%), yielding a disease control rate at 12 months of 21.9%. Overall, tremelimumab was generally well tolerated; safety results were consistent with the known safety profile. CONCLUSIONS: Tremelimumab monotherapy demonstrated clinical activity and durable responses in patients with metastatic urothelial carcinoma. This study is the first in which CR has been observed with tremelimumab as a single agent in urothelial carcinoma.

Exposure-Response Analysis of Overall Survival for Tremelimumab in Unresectable Malignant Mesothelioma: The Confounding Effect of Disease Status.

Tremelimumab, an anti-cytotoxic T-lymphocyte antigen-4 monoclonal antibody that enhances T-cell activation, was evaluated in a randomized, double-blind, placebo-controlled, phase IIb study (NCT01843374) in patients with unresectable malignant mesothelioma. The study demonstrated no clinically meaningful differences in overall survival (OS). The objective of this analysis was to evaluate the relationship of exposure with OS. A population pharmacokinetic (PK) model adequately described the PK data. Three factors (sex, C-reactive protein, and baseline tumor size) were identified as statistically significant PK predictors (P < 0.05 on clearance). A positive association between exposure and OS was observed. However, an association between key baseline factors with OS (regardless of treatment) and imbalances in prognostic factors favoring patients with higher exposure (upper vs. lower PK quartile) was seen. Taken together, these results suggest that the exposure OS relationship observed for tremelimumab in mesothelioma is likely spurious rather than a true association of exposure with efficacy.

A randomized, placebo-controlled, single ascending-dose study to assess the safety, tolerability, pharmacokinetics, and immunogenicity of subcutaneous tralokinumab in Japanese healthy volunteers.

Tralokinumab is a human monoclonal antibody in clinical development for asthma and atopic dermatitis that specifically neutralizes interleukin-13. This phase I, single-blind, randomized, placebo-controlled, single ascending-dose study assessed the safety, tolerability, pharmacokinetics (PK), and immunogenicity of subcutaneous tralokinumab (150, 300, or 600 mg) in thirty healthy Japanese adults. The most frequent treatment-emergent adverse event (TEAE) in all treatment groups was injection-site pain. The frequency and severity of TEAEs was similar across tralokinumab doses. Cmax, AUC(0-t), and AUC(0-inf) increased in a dose-proportional manner, and mean t1/2 ranged from 20 to 25 days. No anti-drug antibodies were detected. A post-hoc pooled population PK modeling analysis, incorporating PK data from this study, demonstrated that Japanese individuals had greater systemic exposure to tralokinumab than non-Japanese individuals. This difference was not clinically relevant and was primarily due to differences in body weight, with lower body weight associated with greater PK exposure. Japanese ethnicity was not a significant predictor of tralokinumab PK. This study indicates that single-dose subcutaneous administration of tralokinumab 150-600 mg was well tolerated in Japanese healthy volunteers, and supports the 300 mg dose selection for Japanese patients with asthma in ongoing clinical trials.

Dose-Exposure-Response Relationship of the Investigational Anti-Interleukin-13 Monoclonal Antibody Tralokinumab in Patients With Severe, Uncontrolled Asthma.

Interleukin (IL)-13 is involved in the pathogenesis of some types of asthma. Tralokinumab is a human immunoglobulin G4 monoclonal antibody that specifically binds to IL-13. Two placebo-controlled phase II studies (phase IIa, NCT00873860 and phase IIb, NCT01402986) have been conducted in which tralokinumab was administered subcutaneously. This investigation aimed to characterize tralokinumab's dose-exposure-response (forced expiratory volume in 1 s (FEV1 )) relationship in patients with asthma and to predict the most appropriate dose for phase III. An integrated population pharmacokinetic-pharmacodynamic (PK/PD) modeling analysis was required for phase III dose selection, due to differing phase II patient populations, designs, and regimens. Analysis of combined datasets enabled the identification of tralokinumab's dose-exposure-FEV1 response relationship in patients with asthma. Near-maximal FEV1 increase was predicted at a dose of 300 mg SC once every 2 weeks (Q2W). This dose was chosen for tralokinumab in the phase III clinical development program for treatment of severe, uncontrolled asthma.

Population Modeling of Tumor Kinetics and Overall Survival to Identify Prognostic and Predictive Biomarkers of Efficacy for Durvalumab in Patients With Urothelial Carcinoma.

Durvalumab is an anti-PD-L1 monoclonal antibody approved for patients with locally advanced or metastatic urothelial carcinoma (UC) that has progressed after platinum-containing chemotherapy. A population tumor kinetic model, coupled with dropout and survival models, was developed to describe longitudinal tumor size data and predict overall survival in UC patients treated with durvalumab (NCT01693562) and to identify prognostic and predictive biomarkers of clinical outcomes. Model-based covariate analysis identified liver metastasis as the most influential factor for tumor growth and immune-cell PD-L1 expression and baseline tumor burden as predictive factors for tumor killing. Tumor or immune-cell PD-L1 expression, liver metastasis, baseline hemoglobin, and albumin levels were identified as significant covariates for overall survival. These model simulations provided further insights into the impact of PD-L1 cutoff values on treatment outcomes. The modeling framework can be a useful tool to guide patient selection and enrichment strategies for immunotherapies across various cancer indications.

Population Pharmacokinetics of Durvalumab in Cancer Patients and Association With Longitudinal Biomarkers of Disease Status.

The objectives of this analysis were to develop a population pharmacokinetics (PK) model of durvalumab, an anti-PD-L1 antibody, and quantify the impact of baseline and time-varying patient/disease characteristics on PK. Pooled data from two studies (1,409 patients providing 7,407 PK samples) were analyzed with nonlinear mixed effects modeling. Durvalumab PK was best described by a two-compartment model with both linear and nonlinear clearances. Three candidate models were evaluated: a time-invariant clearance (CL) model, an empirical time-varying CL model, and a semimechanistic time-varying CL model incorporating longitudinal covariates related to disease status (tumor shrinkage and albumin). The data supported a slight decrease in durvalumab clearance with time and suggested that it may be associated with a decrease in nonspecific protein catabolic rate among cancer patients who benefit from therapy. No covariates were clinically relevant, indicating no need for dose adjustment. Simulations indicated similar overall PK exposures following weight-based and flat-dosing regimens.

Pharmacokinetics of tralokinumab in adolescents with asthma: implications for future dosing.

AIMS: Tralokinumab, an investigational human immunoglobulin G4 monoclonal antibody, potently and specifically neutralizes interleukin-13, a central mediator of asthma. Tralokinumab has shown improvements in clinical endpoints in adults with uncontrolled asthma. The present study explored the pharmacokinetics (PK) and safety of a single tralokinumab dose, and utilized a population PK modelling and simulation approach to evaluate the optimal dosing strategy for adolescents. METHODS: Adolescent subjects with asthma, using daily controller medication, received a single subcutaneous dose of tralokinumab 300 mg. Safety, immunogenicity and PK data were collected during a 57-day follow-up. A population PK model was developed using data from the present study and prior studies in adults. Simulations were performed to evaluate dose adjustment requirements for adolescents. RESULTS: Twenty adolescents (12-17 years) were enrolled; all completed the study. No clinically relevant safety findings or antidrug antibodies were detected. PK parameters were similar to those observed in adults. PK modelling showed that body weight was a minor predictor of tralokinumab PK; after incorporating body weight into the PK model, a 15% (nonparametric 95% confidence interval 5%, 26%) lower clearance was found in adolescents compared with adults [173 (151, 209) vs. 204 (191, 229) ml day(-1)]. Simulations showed no therapeutically relevant differences in exposures between adolescent and adult populations, and similar PK profiles for weight-based (4 mg kg(-1)) and fixed (300 mg) fortnightly subcutaneous doses of tralokinumab. CONCLUSION: Single-dose administration of tralokinumab 300 mg in adolescents was well tolerated, with a PK profile similar to that in adults. Exposure predictions suggest that dose adjustment is not required for adolescents.

Pharmacodynamic modeling of cough responses to capsaicin inhalation calls into question the utility of the C5 end point.

BACKGROUND: Inhaled capsaicin elicits cough reproducibly in human subjects and is widely used in the study of cough and antitussive therapies. However, the traditional end points C2 and C5 (the concentrations of capsaicin inducing at least 2 or 5 coughs, respectively) display extensive overlap between health and disease and therefore might not best reflect clinically relevant mechanisms. OBJECTIVES: We sought to investigate capsaicin dose responses in different disease groups. METHODS: Two novel capsaicin cough challenges were compared in patients with chronic cough (CC; n = 20), asthmatic patients (n = 18), and healthy volunteers (HVs; n = 20). Increasing doubling doses of capsaicin (0.48-1000 µmol/L, 4 inhalations per dose) were administered in challenge 1, whereas the order of the doses was randomized in challenge 2. A nonlinear mixed-effects model compared dose-response parameters by disease group and sex. Parameters were also correlated with objective cough frequency. RESULTS: The model classified subjects based on maximum cough response evoked by any concentration of capsaicin (Emax) and the capsaicin dose inducing half-maximal response (ED50). HVs and asthmatic patients were not statistically different for either parameter and therefore combined for analysis (mean ED50, 38.6 µmol/L [relative SE, 28%]; mean Emax, 4.5 coughs [relative SE, 11%]). Compared with HVs/asthmatic patients, patients with CC had lower ED50 values (14.7 µmol/L [relative SE, 28%], P = .008) and higher Emax values (8.6 coughs [relative SE, 11%], P < .0001). Emax values highly correlated with 24-hour cough frequency (r = 0.71, P < .001) and were 37% higher in female compared with male subjects, regardless of disease group (P < .001). CONCLUSIONS: Nonlinear mixed-effects modeling demonstrates that maximal capsaicin cough responses better discriminate health from disease and predict spontaneous cough frequency and therefore provide important insights into the mechanisms underlying CC.

Two bootstrapping routines for obtaining imprecision estimates for nonparametric parameter distributions in nonlinear mixed effects models.

When parameter estimates are used in predictions or decisions, it is important to consider the magnitude of imprecision associated with the estimation. Such imprecision estimates are, however, presently lacking for nonparametric algorithms intended for nonlinear mixed effects models. The objective of this study was to develop resampling-based methods for estimating imprecision in nonparametric distribution (NPD) estimates obtained in NONMEM. A one-compartment PK model was used to simulate datasets for which the random effect of clearance conformed to a (i) normal (ii) bimodal and (iii) heavy-tailed underlying distributional shapes. Re-estimation was conducted assuming normality under FOCE, and NPDs were estimated sequential to this step. Imprecision in the NPD was then estimated by means of two different resampling procedures. The first (full) method relies on bootstrap sampling from the raw data and a re-estimation of both the preceding parametric (FOCE) and the nonparametric step. The second (simplified) method relies on bootstrap sampling of individual nonparametric probability distributions. Nonparametric 95% confidence intervals (95% CIs) were obtained and mean errors (MEs) of the 95% CI width were computed. Standard errors (SEs) of nonparametric population estimates were obtained using the simplified method and evaluated through 100 stochastic simulations followed by estimations (SSEs). Both methods were successfully implemented to provide imprecision estimates for NPDs. The imprecision estimates adequately reflected the reference imprecision in all distributional cases and regardless of the numbers of individuals in the original data. Relative MEs of the 95% CI width of CL marginal density when original data contained 200 individuals were equal to: (i) -22 and -12%, (ii) -22 and -9%, (iii) -13 and -5% for the full and simplified (n = 100), respectively. SEs derived from the simplified method were consistent with the ones obtained from 100 SSEs. In conclusion, two novel bootstrapping methods intended for nonparametric estimation methods are proposed. In addition of providing information about the precision of nonparametric parameter estimates, they can serve as diagnostic tools for the detection of misspecified parameter distributions.

Evaluation of the nonparametric estimation method in NONMEM VI: application to real data.

The aim of the study was to evaluate the nonparametric estimation methods available in NONMEM VI in comparison with the parametric first-order method (FO) and the first-order conditional estimation method (FOCE) when applied to real datasets. Four methods for estimating model parameters and parameter distributions (FO, FOCE, nonparametric preceded by FO (FO-NONP) and nonparametric preceded by FOCE (FOCE-NONP)) were compared for 25 models previously developed using real data and a parametric method. Numerical predictive checks were used to test the appropriateness of each model. Up to 1000 new datasets were simulated from each model and with each method to construct 90% and 50% prediction intervals. The mean absolute error and the mean error of the different outcomes investigated were computed as indicators of imprecision and bias respectively and formal statistical tests were performed. Overall, less imprecision and less bias were observed with nonparametric methods than with parametric methods. Across the 25 models, t-tests revealed that imprecision and bias were significantly lower (P < 0.05) with FOCE-NONP than with FOCE for half of the NPC outcomes investigated. Improvements were even more pronounced with FO-NONP in comparison with FO. In conclusion, when applied to real datasets and evaluated by numerical predictive checks, the nonparametric estimation methods in NONMEM VI performed better than the corresponding parametric methods (FO or FOCE).