Toward Project Optimus for Oncology Precision Medicine: <scp>Multi‐Dimensional</scp> Dose Optimization Enabled by Quantitative Clinical Pharmacology

Abstract

Project Optimus is a major US Food and Drug Administration (FDA) initiative aimed at dose optimization in oncology drug development, moving away from the maximum tolerated dose (MTD) paradigm and prospectively characterizing dose–response for efficacy and safety for patient-focused maximization of benefit vs. risk.1 The critical role of clinical pharmacology principles and prospective dose optimization ahead of designing pivotal registration-enabling trials have been reinforced in recent seminal publications,2 White Papers,3 and cross-sector public workshops4 that have galvanized the oncology drug development community and offered a call to action for reforming current approaches. Clinical Pharmacology and Therapeutics (CPT) has been a home for research articles and reviews illustrating contemporary integrative approaches to inform dose selection of oncology therapeutics, including molecularly targeted small molecules,5-7 immunotherapies,8-11 antibody-drug conjugates,12-14 and cell therapies.15-17 Several examples have catalyzed active scientific discussion contributing to growing appreciation of the biological complexity, population variability, and analytical methodology that demand careful consideration for robust dose optimization in oncology drug development.18-24 In the current issue of CPT, Combes et al.25 illustrate the pivotal role of quantitative clinical pharmacology in optimizing dose selection for a molecularly targeted precision medicine in oncology through their study on asciminib, an allosteric inhibitor of BCR-ABL1 in chronic myeloid leukemia – chronic phase (CML-CP). Asciminib is active against wild-type BCR-ABL1 and several mutant forms of the kinase, including the T315I mutation, albeit with lower potency for T315I mutant BCR-ABL1 as established in cell proliferation assays in vitro, and in preclinical in vivo xenograft models using patient-derived CML cell lines.26 An MTD was not reached in the phase I dose escalation study as assessed by Bayesian logistic regression modeling of dose-related safety data.26, 27 Dose selection for pivotal efficacy studies of asciminib in CML-CP considered the molecular biology of the disease and the mechanism of action of the drug without adopting a one-dose-fits-all approach. Based on quantitative pharmacological analyses of the totality of available data, doses of 80 mg/day for non-T315I and 200 mg b.i.d. (400 mg/day) for T315I mutant BCR-ABL1 harboring CML-CP were recommended. Combes et al.25 describe the development of mechanism-based population pharmacokinetic-pharmacodynamic (PK/PD) models of longitudinal BCR-ABL1 transcript levels in ~ 300 patients with CML-CP across the phase I and phase III clinical program, with model-based simulations reinforcing confidence in these recommended doses currently approved for use in the non-T315I and T315I BCR-ABL1 molecularly defined disease populations. A three-compartment tumor kinetic model with dynamic interplay among proliferating, quiescent, and resistant cells was formulated, with drug effect estimated on the proliferating compartment. Covariate analyses indicated that the T315I mutation was associated with a greater number of resistant cells (where the drug had no effect), a lower number of proliferating cells (sensitive to asciminib), and decreased magnitude of the antiproliferative drug effect. In longitudinal simulations of BCR-ABL1 trajectory from the final model, these covariate effects of the T315I mutation translated to higher dosage requirements for clinically meaningful efficacy, as assessed by major molecular response (MMR) rates. Furthermore, simulations indicated < 5% population coverage for exposures above effective concentration 90% (EC90) in patients with T315I at a dose of 80 mg/day (the recommended dose for non-T315I patients); whereas, > 95% coverage above EC90 was predicted at the 200 mg b.i.d. dose across all evaluated exposure metrics. Taken together with safety data27 and a relatively flat exposure-safety relationship,26 the analyses presented by Combes et al.25 were instrumental in providing scientific justification for differential dosing of asciminib in CML-CP based on the underlying molecular pathology of the disease, with a fivefold higher dosage recommended in patients harboring the T315I mutation in BCR-ABL1. Viewed from a broader perspective, the example of dose optimization for asciminib represents application of a Totality of Evidence28 approach, where evidence is substantiated through the confidence gained from consistency across multiple approaches and data sources integrated in a mechanism-informed manner through modeling and simulation. Prior knowledge from in vitro and in vivo preclinical studies effectively guided selection of appropriate dose ranges for escalation in CML-CP with non-T315I (10 mg b.i.d. to 200 mg q.d.) and T315I mutant (20–200 mg b.i.d.) BCR-ABL1. Dose selection for pivotal evaluation of efficacy was based on integrated exposure-response understanding of asciminib's effects on longitudinal BCR-ABL1 transcript dynamics using the totality of data across the 10 mg b.i.d. to 200 mg b.i.d. dose range incorporating T315 mutation status as a covariate. True to the purpose of a phase I study of a targeted agent, the first-in-human study focused on establishing appropriate phase II dose(s) as opposed to defining an MTD. In fact, an MTD was not defined across study arms evaluating doses up to 280 mg b.i.d., supporting a favorable therapeutic window. What were some of the key enablers that allowed effective learning regarding dose in the asciminib first-in-human study? Unlike many investigational agents with novel nonvalidated targets that may be evaluated in heterogeneous populations of patients with various advanced malignancies in their first-in-human trials, the dose escalation trial of asciminib was conducted in patients with chronic myelogenous leukemia (CML), as appropriate for a molecule with a clearly defined molecular target and disease context at the point of drug design. CML is defined by BCR-ABL1 as a molecular hallmark, with longitudinal molecular monitoring of BCR-ABL1 transcript levels representing a clinically used approach for evaluating response to treatment and relapse. As such, a quantitative efficacy biomarker end point for longitudinal modeling was available, enabling mechanism-based PK/PD modeling. Another disease context where PK/PD modeling of dose-ranging phase I data to inform dose selection may be feasible for targeted therapies is multiple myeloma, where circulating M-protein serves as a measure of tumor burden and an efficacy biomarker (analogous to BCR-ABL1 transcript dynamics in CML). Indeed, successful PK/PD modeling of longitudinal M-protein data from phase I/Ib dose escalation studies to inform phase III dose and schedule selection has been described for the anti-CD38 monoclonal antibody isatuximab.29 A key question is whether this paradigm can be applied in solid-tumor settings where imaging-based response assessments are primarily utilized for efficacy evaluation. Indeed, if the phase I dose escalation study is conducted in selected or enriched populations based on the drug's mechanism of action, it should be possible to evaluate exposure-response for antitumor activity through longitudinal tumor kinetic modeling (e.g., of sum of longest diameters of the target lesions). As one example, the phase I dose-ranging study of the phosphoinositide 3-kinase antagonist alpelisib was performed in patients with advanced solid tumors, with a mutation or amplification of the PIK3CA gene, and the resulting data could be used for population PK/PD analyses of tumor kinetics with the purpose of informing dose selection.30 Another scenario where this could be envisioned is in a phase I trial of an antibody-drug conjugate (ADC) that is conducted in a cancer type that (i) expresses the tumor-associated antigen targeted by the ADC, and (ii) is known to be sensitive to the pharmacologic mechanism of action of the ADC-associated payload. With advances in biomarker technologies for molecular monitoring of tumor burden using circulating cell-free plasma-derived tumor DNA (ctDNA) measurements, it should be possible to leverage such longitudinal measurements for PK/PD modeling using semimechanistic tumor kinetic modeling frameworks to inform dose and schedule selection. Tumor kinetic modeling of longitudinal ctDNA concentrations using population methods to describe the dynamics of driver and resistance mutational profiles has been described.31 However, application for characterizing exposure-response relationships for antitumor activity to inform dose selection in drug development remains a largely untapped opportunity. Progress will require collaboration across the disciplines of quantitative clinical pharmacology and molecular biomarker sciences for early and efficient learning of dose/exposure-response relationships in phase I trials via population PK/PD modeling of ctDNA dynamics. Such learning, coupled with complementary dose/exposure-safety understanding, would be particularly valuable to aid dose selection decisions for kinase inhibitors designed to inhibit oncogenic variants of receptor tyrosine kinases, where ctDNA profiling to characterize response and resistance mechanisms is often integrated into clinical protocols. Of note, the FDA recently issued a Draft Guidance on use of ctDNA for early-stage solid tumor drug development.32 This Draft Guidance discusses the use of ctDNA as a measure of response and raises the following opportunity: “ctDNA could be used in early phase clinical trials to aid in signal finding of drug activity and to potentially aid sponsors in their drug development plans.” A natural extension of this opportunity is to explore assessment of exposure-response relationships using ctDNA as a measure of response in early clinical development. Such understanding, where feasible, can be expected to valuably contribute to the purpose of the FDA Optimus initiative for timely dose optimization in oncology drug development. Dose selection in oncology drug development is often a problem of multidimensional optimization. Some dimensions along which optimization is desired include (i) dose/schedule, (ii) patient selection (e.g., tumor mutational/molecular profile), (iii) cancer type (histology), (iv) line of treatment, (v) combination partner selection, and (vi) dose/schedule of the selected combination partner. Besides tumor mutational/molecular profile (which was a key dimension in the case of asciminib), other pertinent features of tumor biology relevant to the drug's mechanism of action may warrant consideration under #2 for patient selection strategy. For example, expression of the tumor-associated antigen for an ADC (e.g., low vs. high) or immunophenotype (e.g., cold vs. inflamed) may be factors to consider depending on the mechanism of action of the investigational agent. Of note, as demonstrated for asciminib, the optimal dose can be expected to vary by biological context. Emerging biomarker technologies and integrative frameworks like quantitative systems pharmacology models11, 33-35 should increasingly enhance our ability to tackle this multidimensional complexity of designing and developing optimized cancer pharmacotherapy. (Relatedly, Pharmacometrics & Systems Pharmacology, another ASCPT journal, has issued a Call for Papers for a themed issue on Dose Optimization in Oncology, which will publish in the fourth quarter of 2023. You can read more about the call here: https://www.ascpt.org/Journals/Call-for-Papers.) Confidence in dose optimization in oncology drug development should be supported by three inter-dependent anchors (Figure 1): (i) rigorous mechanism-based quantitative pharmacological contextualization as a foundational base, (ii) clinical efficacy profile indicative of meaningful clinical benefit, and (iii) long-term (e.g., multi-cycle) safety and tolerability supported by a low incidence of dose modifications and a favorable patient experience. Cultural enablers for success include comfort with an iterative “learning” approach to multidimensional optimization, commitment to a Totality of Evidence mindset, and multidisciplinary collaboration among clinical pharmacologists, biomarker scientists, systems pharmacologists, cancer biologists, pharmacometricians, statisticians, clinical trialists, and oncologists, incorporating the patient voice in the patient-focused design and analysis of clinical trials. Just as a one-dose-fits-all approach may no longer be appropriate in all cases, as elegantly illustrated for asciminib by Combes et al.,25 a case-by-case approach to multidimensional dose optimization is warranted in oncology drug development, with the discipline of clinical pharmacology being foundational to move the needle and realize the promise of our journey toward Project Optimus (Figure 2). The authors declared no competing interest for this work. No funding was received for this work.

Topics & Concepts