Imputing Biomarker Status from RWE Datasets-A Comparative Study.

Missing data is a universal problem in analysing Real-World Evidence (RWE) datasets. In RWE datasets, there is a need to understand which features best correlate with clinical outcomes. In this context, the missing status of several biomarkers may appear as gaps in the dataset that hide meaningful values for analysis. Imputation methods are general strategies that replace missing values with plausible values. Using the Flatiron NSCLC dataset, including more than 35,000 subjects, we compare the imputation performance of six such methods on missing data: predictive mean matching, expectation-maximisation, factorial analysis, random forest, generative adversarial networks and multivariate imputations with tabular networks. We also conduct extensive synthetic data experiments with structural causal models. Statistical learning from incomplete datasets should select an appropriate imputation algorithm accounting for the nature of missingness, the impact of missing data, and the distribution shift induced by the imputation algorithm. For our synthetic data experiments, tabular networks had the best overall performance. Methods using neural networks are promising for complex datasets with non-linearities. However, conventional methods such as predictive mean matching work well for the Flatiron NSCLC biomarker dataset.

Parent and Metabolite Concentration-QT Modeling to Evaluate QT-Interval Prolongation at Savolitinib Therapeutic Doses.

Savolitinib is an oral, potent, and highly selective MET-tyrosine kinase inhibitor under investigation in various tumor types. A thorough QT study evaluated effects on QT interval after a 600-mg single savolitinib dose in healthy subjects. We report exposure-response (E-R) modeling from this study to characterize the effects of savolitinib and its metabolites, M2 and M3, on QTc changes. In a novel application, in vitro potencies against hERG current provided mechanistic support to model the metabolites' effects. The hERG IC50 estimates (95% CI) were 25.8 (22.2-29.9) and 22.6 (14.7-34.6) µM for parent and M2, respectively. The E-R was described by both linear and Emax models, with exposure captured by an active moiety that consisted of savolitinib and M2 concentrations, weighted by the hERG IC50 ratio (1.14). The maximal increase in ΔΔQTcF and EC50 estimates (95% CI) was 18.5 (9.2-27.7) ms and 5709 (2889-8529) nM, respectively. Ignoring M2 contribution resulted in under prediction of QTcF prolongation in the hypothetical case of inhibited M2 clearance; at 300 mg Cmax, the mean (90% CI) of ∆∆QTcF was 9.0 (5.7-12.6) and 5.9 (2.9-8.9) ms using the hERG-informed and parent-only linear models, respectively. Simulations in normal setting confirmed modest QTcF prolongation with 600 mg, but not 300 mg. Using the linear model, the mean (90% CI) maximum ΔΔQTcF were 12.3 (8.6-16.2) and 5.5 (2.6-8.5) ms for 600 and 300 mg, respectively. Further clinical studies will monitor cardiac safety to assess the clinical significance of QT-interval prolongation with savolitinib.

A Randomized, Double-Blind, Placebo- and Positive-Controlled, Three-Way Crossover Study in Healthy Participants to Investigate the Effect of Savolitinib on the QTc Interval.

Savolitinib (AZD6094, HMPL-504, volitinib) is an oral, bioavailable, selective MET-tyrosine kinase inhibitor. This randomized, double-blind, 3-way, crossover phase 1 study of savolitinib versus moxifloxacin (positive control) and placebo-evaluated effects on the QT interval after a single savolitinib dose. Healthy non-Japanese men were randomized to 1 of 6 treatment sequences, receiving single doses of savolitinib 600 mg, moxifloxacin 400 mg, and placebo. The primary end point was time-matched, placebo-adjusted change from baseline in the QT interval corrected for the time between corresponding points on 2 consecutive R waves on electrocardiogram (RR) by the Fridericia formula (ΔΔQTcF). Secondary end points included 12-lead electrocardiogram (ECG) variables, pharmacokinetics, and safety. All 3 treatment periods were completed by 44 of 45 participants (98%). Baseline demographics were balanced across treatment groups. After a single savolitinib 600-mg dose, the highest least-squares mean ΔΔQTcF of 12 milliseconds was observed 5 hours postdose. Upper limits of the 2-sided 90% confidence interval for ΔΔQTcF exceeded 10 milliseconds (the prespecified International Council for Harmonisation limit) 3-6 hours postsavolitinib but otherwise remained less than the threshold. Savolitinib showed no additional effect on PR, QRS, QT, or RR intervals. A positive ΔΔQTcF signal from the moxifloxacin group confirmed study validity. Savolitinib was well tolerated, with a low incidence of adverse events. In this thorough QT/QTc study, QTcF prolongation was observed with a single savolitinib 600-mg dose. ECG monitoring will be implemented in ongoing and future studies of savolitinib to assess the clinical relevance of the observed QT changes from this study.

A pharmacokinetic-pharmacodynamic model for the MET tyrosine kinase inhibitor, savolitinib, to explore target inhibition requirements for anti-tumour activity.

BACKGROUND AND PURPOSE: Savolitinib (AZD6094, HMPL-504, volitinib) is an oral, potent, and highly MET receptor TK inhibitor. This series of studies aimed to develop a pharmacokinetic-pharmacodynamic (PK/PD) model to link inhibition of MET phosphorylation (pMET) by savolitinib with anti-tumour activity. EXPERIMENTAL APPROACH: Cell line-derived xenograft (CDX) experiments using human lung cancer (EBC-1) and gastric cancer (MKN-45) cells were conducted in athymic nude mice using a variety of doses and schedules of savolitinib. Tumour pMET changes and growth inhibition were calculated after 28 days. Population PK/PD techniques were used to construct a PK/PD model for savolitinib. KEY RESULTS: Savolitinib showed dose- and dose frequency-dependent anti-tumour activity in the CDX models, with more frequent, lower dosing schedules (e.g., twice daily) being more effective than intermittent, higher dosing schedules (e.g., 4 days on/3 days off or 2 days on/5 days off). There was a clear exposure-response relationship, with maximal suppression of pMET of >90%. Data from additional CDX and patient-derived xenograft (PDX) models overlapped, allowing calculation of a single EC50 of 0.38 ng·ml-1 . Tumour growth modelling demonstrated that prolonged, high levels of pMET inhibition (>90%) were required for tumour stasis and regression in the models. CONCLUSION AND IMPLICATIONS: High and persistent levels of MET inhibition by savolitinib were needed for optimal monotherapy anti-tumour activity in preclinical models. The modelling framework developed here can be used to translate tumour growth inhibition from the mouse to human and thus guide choice of clinical dose and schedule.

Time-to-Event Modeling of Left- or Right-Censored Toxicity Data in Nonclinical Drug Toxicology.

A time-to-event (TTE) model has been developed to characterize a histopathology toxicity that can only be detected at the time of animal sacrifice. The model of choice was a hazard model with a Weibull distribution and dose was a significant covariate. The diagnostic plots showed a satisfactory fit of the data, despite the high degree of left and right censoring. Comparison to a probabilistic logit model shows similar performance in describing the data with a slight underestimation of survival by the Logit model. However, the TTE model was found to be more predictive in extrapolating toxicity risk beyond the observation range of a truncated dataset. The diagnostic and comparison outcomes would suggest using the TTE approach as a first choice for characterizing short and long-term risk from nonclinical toxicity studies. However, further investigations are needed to explore the domain of application of this kind of approach in drug safety assessment.

Randomized clinical study of safety, pharmacokinetics, and pharmacodynamics of RIPK1 inhibitor GSK2982772 in healthy volunteers.

GSK2982772 is a highly selective inhibitor of receptor-interacting protein kinase 1 (RIPK1) being developed to treat chronic inflammatory diseases. This first-in-human study evaluated safety, tolerability, pharmacokinetics (PK), and exploratory pharmacodynamics (PD) of GSK2982772 administered orally to healthy male volunteers. This was a Phase I, randomized, placebo-controlled, double-blind study. In Part A, subjects received single ascending doses of GSK2982772 (0.1-120 mg) or placebo in a crossover design during each of 4 treatment periods. In Part B, subjects received repeat doses of GSK2982772 (20 mg once daily [QD] to up to 120 mg twice daily [BID]) or placebo for 14 days. Part C was an open-label relative bioavailability study comparing 20-mg tablets vs capsules. Safety, tolerability, pharmacokinetics (PK), RIPK1 target engagement (TE), and pharmacodynamics (PD) were assessed. The most common adverse events (AEs) were contact dermatitis and headache. Most AEs were mild in intensity, and there were no deaths or serious AEs. The PK of GSK2982772 was approximately linear over the dose range studied (up to 120 mg BID). There was no evidence of drug accumulation upon repeat dosing. Greater than 90% RIPK1 TE was achieved over a 24-hour period for the 60-mg and 120-mg BID dosing regimens. Single and repeat doses of GSK2982772 were safe and well tolerated. PK profiles showed dose linearity. The high levels of RIPK1 TE support progression into Phase II clinical trials for further clinical development.

Pharmacology-based toxicity assessment: towards quantitative risk prediction in humans.

Despite ongoing efforts to better understand the mechanisms underlying safety and toxicity, ~30% of the attrition in drug discovery and development is still due to safety concerns. Changes in current practice regarding the assessment of safety and toxicity are required to reduce late stage attrition and enable effective development of novel medicines. This review focuses on the implications of empirical evidence generation for the evaluation of safety and toxicity during drug development. A shift in paradigm is needed to (i) ensure that pharmacological concepts are incorporated into the evaluation of safety and toxicity; (ii) facilitate the integration of historical evidence and thereby the translation of findings across species as well as between in vitro and in vivo experiments and (iii) promote the use of experimental protocols tailored to address specific safety and toxicity questions. Based on historical examples, we highlight the challenges for the early characterisation of the safety profile of a new molecule and discuss how model-based methodologies can be applied for the design and analysis of experimental protocols. Issues relative to the scientific rationale are categorised and presented as a hierarchical tree describing the decision-making process. Focus is given to four different areas, namely, optimisation, translation, analytical construct and decision criteria. From a methodological perspective, the relevance of quantitative methods for estimation and extrapolation of risk from toxicology and safety pharmacology experimental protocols, such as points of departure and potency, is discussed in light of advancements in population and Bayesian modelling techniques (e.g. non-linear mixed effects modelling). Their use in the evaluation of pharmacokinetics (PK) and pharmacokinetic-pharmacodynamic relationships (PKPD) has enabled great insight into the dose rationale for medicines in humans, both in terms of efficacy and adverse events. Comparable benefits can be anticipated for the assessment of safety and toxicity profile of novel molecules.

Comparing probabilistic and descriptive analyses of time-dose-toxicity relationship for determining no-observed-adverse-effect level in drug development.

The no-observed-adverse-effect level (NOAEL) of a drug defined from animal studies is important for inferring a maximal safe dose in human. However, several issues are associated with its concept, determination and application. It is confined to the actual doses used in the study; becomes lower with increasing sample size or dose levels; and reflects the risk level seen in the experiment rather than what may be relevant for human. We explored a pharmacometric approach in an attempt to address these issues. We first used simulation to examine the behaviour of the NOAEL values as determined by current common practice; and then fitted the probability of toxicity as a function of treatment duration and dose to data collected from all applicable toxicology studies of a test compound. Our investigation was in the context of an irreversible toxicity that is detected at the end of the study. Simulations illustrated NOAEL's dependency on experimental factors such as dose and sample size, as well as the underlying uncertainty. Modelling the probability as a continuous function of treatment duration and dose simultaneously to data from multiple studies allowed the estimation of the dose, along with its confidence interval, for a maximal risk level that might be deemed as acceptable for human. The model-based data integration also reconciled between-study inconsistency and explicitly provided maximised estimation confidence. Such alternative NOAEL determination method should be explored for its more efficient data use, more quantifiable insight to toxic doses, and the potential for more relevant animal-to-human translation.

Early Prediction of Disease Progression in Small Cell Lung Cancer: Toward Model-Based Personalized Medicine in Oncology.

Predictive biomarkers can play a key role in individualized disease monitoring. Unfortunately, the use of biomarkers in clinical settings has thus far been limited. We have previously shown that mechanism-based pharmacokinetic/pharmacodynamic modeling enables integration of nonvalidated biomarker data to provide predictive model-based biomarkers for response classification. The biomarker model we developed incorporates an underlying latent variable (disease) representing (unobserved) tumor size dynamics, which is assumed to drive biomarker production and to be influenced by exposure to treatment. Here, we show that by integrating CT scan data, the population model can be expanded to include patient outcome. Moreover, we show that in conjunction with routine medical monitoring data, the population model can support accurate individual predictions of outcome. Our combined model predicts that a change in disease of 29.2% (relative standard error 20%) between two consecutive CT scans (i.e., 6-8 weeks) gives a probability of disease progression of 50%. We apply this framework to an external dataset containing biomarker data from 22 small cell lung cancer patients (four patients progressing during follow-up). Using only data up until the end of treatment (a total of 137 lactate dehydrogenase and 77 neuron-specific enolase observations), the statistical framework prospectively identified 75% of the individuals as having a predictable outcome in follow-up visits. This included two of the four patients who eventually progressed. In all identified individuals, the model-predicted outcomes matched the observed outcomes. This framework allows at risk patients to be identified early and therapeutic intervention/monitoring to be adjusted individually, which may improve overall patient survival.

The impact of composite AUC estimates on the prediction of systemic exposure in toxicology experiments.

Current toxicity protocols relate measures of systemic exposure (i.e. AUC, Cmax) as obtained by non-compartmental analysis to observed toxicity. A complicating factor in this practice is the potential bias in the estimates defining safe drug exposure. Moreover, it prevents the assessment of variability. The objective of the current investigation was therefore (a) to demonstrate the feasibility of applying nonlinear mixed effects modelling for the evaluation of toxicokinetics and (b) to assess the bias and accuracy in summary measures of systemic exposure for each method. Here, simulation scenarios were evaluated, which mimic toxicology protocols in rodents. To ensure differences in pharmacokinetic properties are accounted for, hypothetical drugs with varying disposition properties were considered. Data analysis was performed using non-compartmental methods and nonlinear mixed effects modelling. Exposure levels were expressed as area under the concentration versus time curve (AUC), peak concentrations (Cmax) and time above a predefined threshold (TAT). Results were then compared with the reference values to assess the bias and precision of parameter estimates. Higher accuracy and precision were observed for model-based estimates (i.e. AUC, Cmax and TAT), irrespective of group or treatment duration, as compared with non-compartmental analysis. Despite the focus of guidelines on establishing safety thresholds for the evaluation of new molecules in humans, current methods neglect uncertainty, lack of precision and bias in parameter estimates. The use of nonlinear mixed effects modelling for the analysis of toxicokinetics provides insight into variability and should be considered for predicting safe exposure in humans.

Model-based prediction of the acute and long-term safety profile of naproxen in rats.

BACKGROUND AND PURPOSE: Despite the increasing importance of biomarkers as predictors of drug effects, toxicology protocols continue to rely on the experimental evidence of adverse events (AEs) as a basis for establishing the link between indicators of safety and drug exposure. Furthermore, biomarkers may facilitate the translation of findings from animals to humans. Combined with a model-based approach, biomarker data have the potential to predict long-term effects arising from prolonged drug exposure. Here, we used naproxen as a paradigm to explore the feasibility of a biomarker-guided approach for the prediction of long-term AEs in humans. EXPERIMENTAL APPROACH: An experimental toxicology protocol was set up for evaluating the effects of naproxen in rats, in which four active doses were tested (7.5, 15, 40 and 80 mg·kg(-1) ). In addition to AE monitoring and histology, a few blood samples were also collected for the assessment of drug exposure, TXB2 and PGE2 levels. Non-linear mixed effects modelling was used to analyse the data and identify covariate factors on the incidence and severity of AEs. KEY RESULTS: Modelling results showed that besides drug exposure, maximum PGE2 inhibition and treatment duration were also predictors of gastrointestinal ulceration. Although PGE2 levels were clearly linked to the incidence rates, it appeared that ulceration severity is better predicted by measures of drug exposure. CONCLUSIONS AND IMPLICATIONS: These results show that the use of a model-based approach provides the opportunity to integrate pharmacokinetics, pharmacodynamics and toxicity data, enabling optimization of the design, analysis and interpretation of toxicology experiments.

Model-based analysis of thromboxane B2 and prostaglandin E2 as biomarkers in the safety evaluation of naproxen.

The assessment of safety in traditional toxicology protocols relies on evidence arising from observed adverse events (AEs) in animals and on establishing their correlation with different measures of drug exposure (e.g., Cmax and AUC). Such correlations, however, ignore the role of biomarkers, which can provide further insight into the underlying pharmacological mechanisms. Here we use naproxen as a paradigm drug to explore the feasibility of a biomarker-guided approach for the prediction of AEs in humans. A standard toxicology protocol was set up for the evaluation of effects of naproxen in rat, in which four doses were tested (7.5, 15, 40 and 80 mg/kg). In addition to sparse blood sampling for the assessment of exposure, thromboxane B2 and prostaglandin E2 were also collected in satellite groups. Nonlinear mixed effects modelling was used to evaluate the predictive performance of the approach. A one-compartmental model with first order absorption was found to best describe the pharmacokinetics of naproxen. A nonlinear relationship between dose and bioavailability was observed which leads to a less than proportional increase in naproxen concentrations with increasing doses. The pharmacodynamics of TXB2 and PGE2 was described by direct inhibition models with maximum pharmacological effects achieved at doses >7.5 mg/kg. The predicted PKPD relationship in humans was within 10-fold of the values previously published. Moreover, our results indicate that biomarkers can be used to assess interspecies differences in PKPD and extrapolated data from animals to humans. Biomarker sampling should be used systematically in general toxicity studies.

Feasibility of a fixed-dose regimen of pyrazinamide and its impact on systemic drug exposure and liver safety in patients with tuberculosis.

Historically, dosing regimens for the treatment of tuberculosis (TB) have been proposed in an empirical manner. Dose selection has often been the result of efficacy trials in which drugs were administered regardless of the magnitude of the effect of demographic factors on drug disposition. This has created challenges for the prescription of fixed-dose combinations with novel therapeutic agents. The objectives of this investigation were to evaluate the impact of body weight on the overall systemic exposure to pyrazinamide (PZA) and to assess whether the use of one fixed dose, without adjustment according to weight, would ensure target exposure and safety requirements across the overall patient population. Using a population pharmacokinetic model, simulation scenarios were explored based on population demographics from clinical trials in TB patients and on historical hepatotoxicity data. The systemic drug exposure (area under the concentration-time curve [AUC]), peak concentrations (the maximum concentration of drug in serum [C(max)]), the time above the MIC (t > MIC), and the risk of hepatotoxicity were evaluated for the current weight-banded regimen and compared to fixed doses under the assumption that pharmacokinetic differences are the primary drivers of toxicity. Evaluation of the standard weight banding reveals that more than 50% of subjects in the weight range of 45 to 55 kg remain below the proposed target exposure to PZA. In contrast, the use of a fixed 1,500-mg dose resulted in a lower proportion of subjects under the target value, with a 0.2% average overall increase in the risk of hepatotoxicity. Our results strongly support the use of a fixed-dose regimen for PZA in coformulation or combination with novel therapeutic agents.