Support to early clinical decisions in drug development and personalised medicine with checkpoint inhibitors using dynamic biomarker-overall survival models.

Longitudinal models of biomarkers such as tumour size dynamics capture treatment efficacy and predict treatment outcome (overall survival) of a variety of anticancer therapies, including chemotherapies, targeted therapies, immunotherapies and their combinations. These pharmacological endpoints like tumour dynamic (tumour growth inhibition) metrics have been proposed as alternative endpoints to complement the classical RECIST endpoints (objective response rate, progression-free survival) to support early decisions both at the study level in drug development as well as at the patients level in personalised therapy with checkpoint inhibitors. This perspective paper presents recent developments and future directions to enable wider and robust use of model-based decision frameworks based on pharmacological endpoints.

Nonlinear multilevel joint model for individual lesion kinetics and survival to characterize intra-individual heterogeneity in patients with advanced cancer.

In advanced cancer patients, tumor burden is calculated using the sum of the longest diameters (SLD) of the target lesions, a measure that lumps all lesions together and ignores intra-patient heterogeneity. Here, we used a rich dataset of 342 metastatic bladder cancer patients treated with a novel immunotherapy agent to develop a Bayesian multilevel joint model that can quantify heterogeneity in lesion dynamics and measure their impact on survival. Using a nonlinear model of tumor growth inhibition, we estimated that dynamics differed greatly among lesions, and inter-lesion variability accounted for 21% and 28% of the total variance in tumor shrinkage and treatment effect duration, respectively. Next, we investigated the impact of individual lesion dynamics on survival. Lesions located in the liver and in the bladder had twice as much impact on the instantaneous risk of death compared to those located in the lung or the lymph nodes. Finally, we evaluated the utility of individual lesion follow-up for dynamic predictions. Consistent with results at the population level, the individual lesion model outperformed a model relying only on SLD, especially at early landmark times and in patients with liver or bladder target lesions. Our results show that an individual lesion model can characterize the heterogeneity in tumor dynamics and its impact on survival in advanced cancer patients.

A Pharmacometric Analysis of Patient-Reported Outcomes in Breast Cancer Patients Through Item Response Theory.

PURPOSE: An item response theory (IRT) pharmacometric framework is presented to characterize Functional Assessment of Cancer Therapy-Breast (FACT-B) data in locally-advanced or metastatic breast cancer patients treated with ado-trastuzumab emtansine (T-DM1) or capecitabine-plus-lapatinib. METHODS: In the IRT model, four latent well-being variables, based on FACT-B general subscales, were used to describe the physical, social/family, emotional and functional well-being. Each breast cancer subscale item was reassigned to one of the other subscales. Longitudinal changes in FACT-B responses and covariate effects were investigated. RESULTS: The IRT model could describe both item-level and subscale-level FACT-B data. Non-Asian patients showed better baseline social/family and functional well-being than Asian patients. Moreover, patients with Eastern Cooperative Oncology Group performance status of 0 had better baseline physical and functional well-being. Well-being was described as initially increasing or decreasing before reaching a steady-state, which varied substantially between patients and subscales. T-DM1 exposure was not related to any of the latent variables. Physical well-being worsening was identified in capecitabine-plus-lapatinib-treated patients, whereas T-DM1-treated patients typically stayed stable. CONCLUSION: The developed framework provides a thorough description of FACT-B longitudinal data. It acknowledges the multi-dimensional nature of the questionnaire and allows covariate and exposure effects to be evaluated on responses.

Platform model describing pharmacokinetic properties of vc-MMAE antibody-drug conjugates.

Antibody-drug conjugates (ADCs) developed using the valine-citrulline-MMAE (vc-MMAE) platform, consist of a monoclonal antibody (mAb) covalently bound with a potent anti-mitotic toxin (MMAE) through a protease-labile vc linker. Recently, clinical data for a variety of vc-MMAE ADCs has become available. The goal of this analysis was to develop a platform model that simultaneously described antibody-conjugated MMAE (acMMAE) pharmacokinetic (PK) data from eight vc-MMAE ADCs, against different targets and tumor indications; and to assess differences and similarities of model parameters and model predictions, between different compounds. Clinical PK data of eight vc-MMAE ADCs from eight Phase I studies were pooled. A population PK platform model for the eight ADCs was developed, where the inter-compound variability (ICV) was described explicitly, using the third random effect level (ICV), and implemented using LEVEL option of NONMEM 7.3. The PK was described by a two-compartment model with time dependent clearance. Clearance and volume of distribution increased with body weight; volume was higher for males, and clearance mildly decreased with the nominal dose. Michaelis-Menten elimination had only minor effect on PK and was not included in the model. Time-dependence of clearance had no effect beyond the first dosing cycle. Clearance and central volume were similar among ADCs, with ICV of 15 and 5%, respectively. Thus, PK of acMMAE was largely comparable across different vc-MMAE ADCs. The model may be applied to predict PK-profiles of vc-MMAE ADCs under development, estimate individual exposure for the subsequent PK-pharmacodynamics (PD) analysis, and project optimal dose regimens and PK sampling times.

Quantitative characterization of the effects of fulvestrant alone or in combination with taselisib (PI3Kinase inhibitor) on longitudinal tumor growth in patients with estrogen receptor-positive, HER2-negative, PIK3CA-mutant, advanced or metastatic breast cancer.

PURPOSE: Among cases of breast cancer, estrogen receptor-positive (ER +), PIK3CA-mutant, HER2- advanced breast cancer stands as a particularly complex clinical indication where approximately 40% of ER + /HER2- breast carcinomas present mutations in the PIK3CA gene. A significant hurdle in treating ER + breast cancer lies in surmounting the challenges of endocrine resistance. In the clinical setting, a multifaceted approach is essential for this indication, one that not only explores the effectiveness of individual treatments but also delves into the potential gains in therapeutic outcome from combination therapies. METHODS: In the current study, longitudinal tumor growth inhibition (TGI) models were developed to characterize tumor response over time in postmenopausal women with ER + /HER2- advanced or metastatic breast cancer undergoing treatment with fulvestrant alone or in combination with the PI3K inhibitor, taselisib. Impact of clinically relevant covariates on TGI metrics was assessed to identify patient subsets most likely to benefit from treatment with fulvestrant monotherapy or combination with taselisib. RESULTS: Tumor growth rate constant (Kg) was found to increase with increasing baseline tumor size and in the absence of baseline endocrine sensitivity. Further, Kg decreased in the absence of baseline liver metastases both in fulvestrant monotherapy and combination therapy with taselisib. Overall, additive/potentially synergistic anti-tumor effects were observed in patients treated with the taselisib-fulvestrant combination. CONCLUSION: These results have important implications for understanding the therapeutic impact of combination treatment approaches and individualized responses to these treatments. Finally, this work, emphasizes the importance of model informed drug development for targeted cancer therapy. CLINICAL TRIAL REGISTRATION: NCT02340221 Registered January 16, 2015, NCT01296555 Registered February 14, 2011.

Retrospective Assessment of Translational Pharmacokinetic-Pharmacodynamic Modeling Performance: A Case Study with Apitolisib, a Dual PI3K/mTOR Inhibitor.

BACKGROUND AND OBJECTIVES: Despite significant progress in biomedical research, the rate of success in oncology drug development remains inferior to that of other therapeutic fields. Mechanistic models provide comprehensive understanding of the therapeutic effects of drugs, which is crucial for designing effective clinical trials. This study was performed to acquire a better understanding of PI3K-AKT-TOR pathway modulation and preclinical to clinical translational bridging for a specific compound, apitolisib (PI3K/mTOR inhibitor), by developing integrated mechanistic models. METHODS: Integrated pharmacokinetic (PK)-pharmacodynamic (PD)-efficacy models were developed for xenografts bearing human renal cell adenocarcinoma and for patients with solid tumors (phase 1 studies) to characterize relationships between exposure of apitolisib, modulation of the phosphorylated Akt (pAkt) biomarker triggered by inhibition of the PI3K-AKT-mTOR pathway, and tumor response. RESULTS: Both clinical and preclinical integrated models show a steep sigmoid curve linking pAkt inhibition to tumor growth inhibition and quantified that a minimum of 35-45% pAkt modulation is required for tumor shrinkage in patients, based on platelet-rich plasma surrogate matrix and in xenografts based on tumor tissue matrix. Based on this relationship between targeted pAkt modulation and tumor shrinkage rate, it appeared that a constant pAkt inhibition of 61% and 65%, respectively, would be necessary to achieve tumor stasis in xenografts and patients. CONCLUSIONS: These results help when it comes to evaluating the translatability of the preclinical analysis to the clinical target, and provide information that will enhance the value of future preclinical translational dose-finding and dose-optimization studies to accelerate clinical drug development. TRIAL REGISTRY: ClinicalTrials.gov NCT00854152 and NCT00854126.

Quantitative Assessment of Drug Efficacy and Emergence of Resistance in Patients with Metastatic Renal Cell Carcinoma Using a Longitudinal Exposure-Tumor Growth Inhibition Model: Apitolisib (Dual PI3K/mTORC1/2 Inhibitor) Versus Everolimus (mTORC1 Inhibitor).

Cancer remains a significant global health challenge, and despite remarkable advancements in therapeutic strategies, poor tolerability of drugs (causing dose reduction/interruptions) and/or the emergence of drug resistance are major obstacles to successful treatment outcomes. Metastatic renal cell carcinoma (mRCC) accounts for 2% of global cancer diagnoses and deaths. Despite the initial success of targeted therapies in mRCC, challenges remain to overcome drug resistance that limits the long-term efficacy of these treatments. Our analysis aim was to develop a semi-mechanistic longitudinal exposure-tumor growth inhibition model for patients with mRCC to characterize and compare everolimus (mTORC1) and apitolisib's (dual PI3K/mTORC1/2) ability to inhibit tumor growth, and quantitate each drug's efficacy decay caused by emergence of tumor resistance over time. Model-estimated on-treatment tumor growth rate constant was 1.7-fold higher for apitolisib compared to everolimus. Estimated half-life for loss of treatment effect over time for everolimus was 16.1 weeks compared to 7.72 weeks for apitolisib, suggesting a faster rate of tumor re-growth for apitolisib patients likely due to the emergence of resistance. Goodness-of-fit plots including visual predictive check indicated a good model fit and the model was able to capture individual tumor size-time profiles. Based on our knowledge, this is the first clinical report to quantitatively assess everolimus (mTORC1) and apitolisib (PI3K/mTORC1/2) efficacy decay in patients with mRCC. These results highlight the difference in overall efficacy of 2 drugs due to the quantified efficacy decay caused by emergence of resistance, and emphasize the importance of model-informed drug development for targeted cancer therapy.

gPKPDviz: A flexible R shiny tool for pharmacokinetic/pharmacodynamic simulations using mrgsolve.

GPKPDviz is a Shiny application (app) dedicated to real-time simulation, visualization, and assessment of the pharmacokinetic/pharmacodynamic (PK/PD) models. Within the app, gPKPDviz is capable of generating virtual populations and complex dosing and sampling scenarios, which, together with the streamlined workflow, is designed to efficiently assess the impact of covariates and dosing regimens on PK/PD end points. The actual population data from clinical trials can be loaded into the app for simulation if desired. The app-generated dosing regimens include single or multiple dosing, and more complex regimens, such as loading doses or intermittent dosing. When necessary, the dosing regimens can be defined externally and loaded to the app for simulation. Using mrgsolve as the simulation engine, gPKPDviz is typically used for population simulation, however, with a slight modification of the mrgsolve model, gPKPDviz is capable of performing individual simulations with individual post hoc parameters, individual dosing logs, and individual sampling timepoints through an external dataset. A built-in text editor has a debugging feature for the mrgsolve model, providing the same error messages as model compilation in R. GPKPDviz has had stringent validation by comparing simulation results between the app and using mrgsolve in R. GPKPDviz is a member of the suite of Modeling and Simulation Shiny apps developed at Genentech to facilitate the typical modeling work in Clinical Pharmacology. For broader access to the Pharmacometric community, gPKPDviz has been published as an open-source application in GitHub under the terms of GNU General Public License.

Artificial Intelligence: From Buzzword to Useful Tool in Clinical Pharmacology.

The advent of artificial intelligence (AI) in clinical pharmacology and drug development is akin to the dawning of a new era. Previously dismissed as merely technological hype, these approaches have emerged as promising tools in different domains, including health care, demonstrating their potential to empower clinical pharmacology decision making, revolutionize the drug development landscape, and advance patient care. Although challenges remain, the remarkable progress already made signals that the leap from hype to reality is well underway, and AI promises to offer clinical pharmacology new tools and possibilities for optimizing patient care is gradually coming to fruition. This review dives into the burgeoning world of AI and machine learning (ML), showcasing different applications of AI in clinical pharmacology and the impact of successful AI/ML implementation on drug development and/or regulatory decisions. This review also highlights recommendations for areas of opportunity in clinical pharmacology, including data analysis (e.g., handling large data sets, screening to identify important covariates, and optimizing patient population) and efficiencies (e.g., automation, translation, literature curation, and training). Realizing the benefits of AI in drug development and understanding its value will lead to the successful integration of AI tools in our clinical pharmacology and pharmacometrics armamentarium.

Comparison of two-stage and joint TGI-OS modeling using data from six atezolizumab clinical studies in patients with metastatic non-small cell lung cancer.

Two-stage and joint modeling approaches are the two main approaches to investigate the link between longitudinal tumor size data and overall survival (OS) and anticipate clinical trial outcome. We here used a large database composed of one phase II and five phase III clinical trials evaluating atezolizumab (an immunotherapy) in monotherapy or in combination with chemotherapies in 3699 patients with non-small cell lung cancer to evaluate the differences between both approaches in terms of parameter estimates, magnitude of covariate effects, and ability to predict OS. Although the two-stage approach may underestimate the magnitude of the impact of tumor growth rate (KG ) on OS compared to joint modeling approach (hazard ratios [HRs] of 0.42-2.52 vs. 0.25-2.85, respectively, for individual KG varying from the 5th and 95th percentiles), this difference did not lead into poorer performance of the two-stage approach to describe the OS distribution in the six clinical studies. Overall, two-stage and joint modeling approaches accurately predicted OS HR with a median (range) difference with the observed OS HR of 0.02 (0.01-0.18) and 0.03 (0.00-0.19), in all cases considered, respectively (e.g., for IMpower150: 0.80 [0.66-0.95] vs. 0.82 [0.70-0.95], respectively, whereas the observed OS HR was 0.80). In our setting, the two-stage approach accurately predicted the benefit of atezolizumab on OS. Further work is needed to verify if similar results are achieved using phase Ib or phase II clinical trials where the number of patients and measurements is limited as well as in other cancer indications.

Deep-NCA: A deep learning methodology for performing noncompartmental analysis of pharmacokinetic data.

Noncompartmental analysis (NCA) is a model-independent approach for assessing pharmacokinetics (PKs). Although the existing NCA algorithms are very well-established and widely utilized, they suffer from low accuracies in the setting of sparse PK samples. In response, we developed Deep-NCA, a deep learning (DL) model to improve the prediction of key noncompartmental PK parameters. Our methodology utilizes synthetic PK data for model training and uses an innovative patient-specific normalization method for data preprocessing. Deep-NCA demonstrated adequate performance across six previously unseen simulated drugs under multiple dosing, showcasing effective generalization. Compared to traditional NCA, Deep-NCA exhibited superior performance for sparse PK data. This study advances the application of DL to PK studies and introduces an effective method for handling sparse PK data. With further validation and refinement, Deep-NCA could significantly enhance the efficiency of drug development by providing more accurate NCA estimates while requiring fewer PK samples.

Tumor growth and overall survival modeling to support decision making in phase Ib/II trials: A comparison of the joint and two-stage approaches.

Model-based tumor growth inhibition (TGI) metrics are increasingly used to predict overall survival (OS) data in Phase III immunotherapy clinical trials. However, there is still a lack of understanding regarding the differences between two-stage or joint modeling methods to leverage Phase I/II trial data and help early decision-making. A recent study showed that TGI metrics such as the tumor growth rate constant KG may have good operating characteristics as early endpoints. This previous study used a two-stage approach that is easy to implement and intuitive but prone to bias as it does not account for the relationship between the longitudinal and time-to-event processes. A relevant alternative is to use a joint modeling approach. In the present article, we evaluated the operating characteristics of TGI metrics using joint modeling, assuming an OS model previously developed using historical data. To that end, we used TGI and OS data from IMpower150-a study investigating atezolizumab in over 750 patients suffering from non-small cell lung cancer-to mimic randomized Phase Ib/II trials varying in terms of number of patients included (40 to 15 patients per arm) and follow-up duration (24 to 6 weeks after the last patient included). In this context, joint modeling did not outperform the two-stage approach and provided similar operating characteristics in all the investigated scenarios. Our results suggest that KG geometric mean ratio could be used to support early decision-making provided that 30 or more patients per arm are included and followed for at least 12 weeks.

Challenging the Norm: A Multidisciplinary Perspective on Intravenous to Subcutaneous Bridging Strategies for Biologics.

The transition from intravenous (i.v.) to subcutaneous (s.c.) administration of biologics is a critical strategy in drug development aimed at improving patient convenience, compliance, and therapeutic outcomes. Focusing on the increasing role of model-informed drug development (MIDD) in the acceleration of this transition, an in-depth overview of the essential clinical pharmacology, and regulatory considerations for successful i.v. to s.c. bridging for biologics after the i.v. formulation has been approved are presented. Considerations encompass multiple aspects beginning with adequate pharmacokinetic (PK) and pharmacodynamic (i.e., exposure-response) evaluations which play a vital role in establishing comparability between the i.v. and s.c. routes of administrations. Selected key recommendations and points to consider include: (i) PK characterization of the s.c. formulation, supported by the increasing preclinical understanding of the s.c. absorption, and robust PK study design and analyses in humans; (ii) a thorough characterization of the exposure-response profiles including important metrics of exposure for both efficacy and safety; (iii) comparability studies designed to meet regulatory considerations and support approval of the s.c. formulation, including noninferiority studies with PK and/or efficacy and safety as primary end points; and (iv) comprehensive safety package addressing assessments of immunogenicity and patients' safety profile with the new route of administration. Recommendations for successful bridging strategies are evolving and MIDD approaches have been used successfully to accelerate the transition to s.c. dosing, ultimately leading to improved patient experiences, adherence, and clinical outcomes.

Applications of Advanced Natural Language Processing for Clinical Pharmacology.

Natural language processing (NLP) is a branch of artificial intelligence, which combines computational linguistics, machine learning, and deep learning models to process human language. Although there is a surge in NLP usage across various industries in recent years, NLP has not been widely evaluated and utilized to support drug development. To demonstrate how advanced NLP can expedite the extraction and analyses of information to help address clinical pharmacology questions, inform clinical trial designs, and support drug development, three use cases are described in this article: (1) dose optimization strategy in oncology, (2) common covariates on pharmacokinetic (PK) parameters in oncology, and (3) physiologically-based PK (PBPK) analyses for regulatory review and product label. The NLP workflow includes (1) preparation of source files, (2) NLP model building, and (3) automation of data extraction. The Clinical Pharmacology and Biopharmaceutics Summary Basis of Approval (SBA) documents, US package inserts (USPI), and approval letters from the US Food and Drug Administration (FDA) were used as our source data. As demonstrated in the three example use cases, advanced NLP can expedite the extraction and analyses of large amounts of information from regulatory review documents to help address important clinical pharmacology questions. Although this has not been adopted widely, integrating advanced NLP into the clinical pharmacology workflow can increase efficiency in extracting impactful information to advance drug development.

Impact of food and the proton pump inhibitor rabeprazole on the pharmacokinetics of GDC-0941 in healthy volunteers: bench to bedside investigation of pH-dependent solubility.

GDC-0941 is an orally administered potent, selective pan-inhibitor of phosphatidylinositol 3-kinases (PI3Ks) with good preclinical antitumor activity in xenograft models and favorable pharmacokinetics and tolerability in phase 1 trials, and it is currently being investigated in phase II clinical trials as an anti-cancer agent. In vitro solubility and dissolution studies suggested that GDC-0941, a weak base, displays significant pH-dependent solubility. Moreover, preclinical studies conducted in famotidine-induced hypochlorhydric dog suggested that the pharmacokinetics of GDC-0941 may be sensitive to pharmacologically induced hypochlorhydria. To investigate the clinical significance of food and pH-dependent solubility on GDC-0941 pharmacokinetics a four-period, two-sequence, open-label, randomized, crossover study was conducted in healthy volunteers. During the fasting state, GDC-0941 was rapidly absorbed with a median Tmax of 2 h. The presence of a high-fat meal delayed the absorption of GDC-0941, with a median Tmax of 4 h and a modest increase in AUC relative to the fasted state, with an estimated geometric mean ratio (GMR, 90% CI) of fed/fasted of 1.28 (1.08, 1.51) for AUC0-∞ and 0.87 (0.70, 1.06) for Cmax. The effect of rabeprazole (model PPI) coadministration on the pharmacokinetics of GDC-0941 was evaluated in the fasted and fed state. When comparing the effect of rabeprazole + GDC-0941 (fasted) to baseline GDC-0941 absorption in a fasted state, GDC-0941 median Tmax was unchanged, however, both Cmax and AUC0-∞ decreased significantly after pretreatment with rabeprazole, with an estimated GMR (90% CI) of 0.31 (0.21, 0.46) and 0.46 (0.35, 0.61), respectively for both parameters. When rabeprazole was administered in the presence of the high-fat meal, the impact of food did not fully reverse the pH effect; the overall effect of rabeprazole on AUC0-∞ was somewhat attenuated by the high-fat meal (estimate GMR of 0.57, with 90% CI, 0.50, 0.65) but unchanged for the Cmax (estimate of 0.43, with 90% CI, 0.37, 0.50). The results of the current investigations emphasize the complex nature of physicochemical interactions and the importance of gastric acid for the dissolution and solubilization processes of GDC-0941. Given these findings, dosing of GDC-0941 in clinical trials was not constrained relative to fasted/fed states, but the concomitant use of ARAs was restricted. Mitigation strategies to limit the influence of pH on exposure of molecularly targeted agents such as GDC-0941 with pH-dependent solubility are discussed.

A physiologically based pharmacokinetic (PBPK) approach to evaluate pharmacokinetics in patients with cancer.

Potential differences in pharmacokinetics (PK) between healthy subjects and patients with cancer were investigated using a physiologically based pharmacokinetic approach integrating demographic and physiological data from patients with cancer. Demographic data such as age, sex and body weight, and clinical laboratory measurements such as albumin, alpha-1 acid glycoprotein (AAG) and hematocrit were collected in ~2500 patients with cancer. A custom oncology population profile was built using the observed relationships among demographic variables and laboratory measurements in Simcyp® software, a population based ADME simulator. Patients with cancer were older compared with the age distribution in a built-in healthy volunteer profile in Simcyp. Hematocrit and albumin levels were lower and AAG levels were higher in patients with cancer. The custom population profile was used to investigate the disease effect on the pharmacokinetics of two probe substrates, saquinavir and midazolam. Higher saquinavir exposure was predicted in patients relative to healthy subjects, which was explained by the altered drug binding due to elevated AAG levels in patients with cancer. Consistent with historical clinical data, similar midazolam exposure was predicted in patients and healthy subjects, supporting the hypothesis that the CYP3A activity is not altered in patients with cancer. These results suggest that the custom oncology population profile is a promising tool for the prediction of PK in patients with cancer. Further evaluation and extension of this population profile with more compounds and more data will be needed.

The convergence of pharmacometrics and quantitative systems pharmacology in pharmaceutical research and development.

Quantitative systems pharmacology (QSP) models are an important facet of pharmaceutical and clinical research as they combine mechanistic models of physiology in health and disease with pharmacokinetics/pharmacodynamics to predict systems-level effects. The quantitative clinical pharmacology toolbox has traditionally included both mechanistic modeling and population approaches, collectively called pharmacometrics, but the current landscape requires the optimization and use of multiple models together. Here, we explore several case studies in drug development that exemplify three approaches for using QSP and pharmacometrics models together - parallel synchronization, cross-informative use, and sequential integration. While these approaches are increasingly applied in drug development, achieving a true convergence of QSP and pharmacometrics that fully exploits their synergy will require new tools and methods that enable greater technical integration, in addition to nurturing scientists with diverse modeling expertise that enable cross-discipline strategy. Extensions of existing methods used in each approach as well as additional resources including machine learning models, data-at-scale, end-to-end computation platforms, and real-time analytics will enable this convergence.

Risk Factors of Hyperglycemia After Treatment With the AKT Inhibitor Ipatasertib in the Prostate Cancer Setting: A Machine Learning-Based Investigation.

PURPOSE: Hyperglycemia is a major adverse event of phosphatidylinositol 3-kinase/AKT inhibitor class of cancer therapeutics. Machine learning (ML) methodologies can identify and highlight how explanatory variables affect hyperglycemia risk. METHODS: Using data from clinical trials of the AKT inhibitor ipatasertib (IPAT) in the metastatic castrate-resistant prostate cancer setting, we trained an XGBoost ML model to predict the incidence of grade ≥2 hyperglycemia (HGLY ≥ 2). Of the 1,364 patients included in our analysis, 19.4% (n = 265) of patients had HGLY ≥2 events with a median time of first onset of 28 days (range, 0-753 days), and 30.0% (n = 221) of patients on an IPAT regimen had at least one HGLY ≥2 event compared with 7.0% (n = 44) of patients on placebo. RESULTS: An 11-variable XGBoost model predicted HGLY ≥2 events well with an AUROC of 0.83 ± 0.02 (mean ± standard deviation). Using SHapley Additive exPlanations analysis, we found IPAT exposure and baseline HbA1c levels to be the strongest predictors of HGLY ≥2, with additional predictivity of baseline measurements of fasting glucose, magnesium, and high-density lipoproteins. CONCLUSION: The findings support using patients' prediabetic status as a key factor for hyperglycemia monitoring and/or trial exclusion criteria. Additionally, the model and relationships between explanatory variables and HGLY ≥2 described herein can help identify patients at high risk for hyperglycemia and develop rational risk mitigation strategies.

A disease model predicting placebo response and remission status of patients with ulcerative colitis using modified Mayo score.

The Mayo Clinical Score is used in clinical trials to describe the clinical status of patients with ulcerative colitis (UC). It comprises four subscores: rectal bleeding (RB), stool frequency (SF), physician's global assessment, and endoscopy (ENDO). According to recent US Food and Drug Administration guidelines (Ulcerative colitis: developing drugs for treatment, Guidance Document, https://www.fda.gov/regulatory-information/s. 2022), clinical response and remission should be based on modified Mayo Score (mMS) relying on RB, SF, and ENDO. Typically, ENDO is performed at the beginning and end of each phase, whereas RB and SF are more frequently available. Item response theory (IRT) models allow the shared information to be used for prediction of all subscores at each observation time; therefore, it leverages information from RB and SF to predict ENDO. A UC disease IRT model was developed based on four etrolizumab phase III studies to describe the longitudinal mMS subscores, placebo response, and remission at the end of induction and maintenance. For each subscore, a bounded integer model was developed. The placebo response was characterized by a mono-exponential function acting on all mMS subscores similarly. The final model reliably predicted longitudinal mMS data. In addition, remission was well-predicted by the model, with only 5% overprediction at the end of induction and 3% underprediction at the end of maintenance. External evaluation of the final model using placebo arms from five different studies indicated adequate performance for both longitudinal mMS subscores and remission status. These results suggest utility of the current disease model for informed decision making in UC clinical development, such as assisting future clinical trial designs and evaluations.

Multistate Pharmacometric Model to Define the Impact of Second-Line Immunotherapies on the Survival Outcome of the IMpower131 Study.

Overall survival is defined as the time since randomization into the clinical trial to event of death or censor (end of trial or follow-up), and is considered to be the most reliable cancer end point. However, the introduction of second-line treatment after disease progression could influence survival and be considered a confounding factor. The aim of the current study was to set up a multistate model framework, using data from the IMpower131 study, to investigate the influence of second-line immunotherapies on overall survival analysis. The model adequately described the transitions between different states in patients with advanced squamous non-small cell lung cancer treated with or without atezolizumab plus nab-paclitaxel and carboplatin, and characterized the survival data. High PD-L1 expression at baseline was associated with a decreased hazard of progression, while the presence of liver metastasis at baseline was indicative of a high risk of disease progression after initial response. The hazard of death after progression was lower for participants who had longer treatment response, i.e., longer time to progression. The simulations based on the final multistate model showed that the addition of atezolizumab to the nab-paclitaxel and carboplatin regimen had significant improvement in the patients' survival (hazard ratio = 0.75, 95% prediction interval: 0.61-0.90 favoring the atezolizumab + nab-paclitaxel and carboplatin arm). The developed modeling approach can be applied to other cancer types and therapies to provide a better understanding of efficacy of drug and characterizing different states, and investigate the benefit of primary therapy in survival while accounting for the switch to alternative treatment in the case of disease progression.