Use of Imaging-based Pharmacodynamics For Dose Selection And Decision making

Transcription

1 Use of Imaging-based Pharmacodynamics For Dose Selection And Decision making Dominic G Spinella, Ph.D. Head of Translational Medicine Pfizer Oncology

2 Proof of Mechanism (POM) and Proof of Concept (POC) The overarching question of Exploratory Development: If experimental drug X hits and modulates its intended target, does clinical benefit ensue? The first part of this question (does drug X hit and modulate its intended target) is what we call POM; the second part (association with clinical benefit) we call POC. We try to a achieve POM in Phase 1 (usually an expanded cohort setting of selected patients at MTD) and POM is required for project progression into Phase 2 POC studies. Both POM and POC depend on a biomarker endpoints preferably markers that are linked to drug mechanism and/or clinical outcome. Given unsustainable Phase 2 attrition rates, the old paradigm of Phase 1 dose escalation solely to obtain PK and safety/tolerability data, followed by Phase 2 studies to seek objective response endpoints is no longer viable.

3 We try to avoid biopsy-based biomarkers wherever possible Many cancer targets are expressed preferentially if not exclusively in tumors (that s why we picked them) and direct assessment of POM is theoretically best done in pre and post-treatment biopsy BUT: Many patients do not present with readily biopsiable tumor; those that do predominantly present with only a few tumor types (e.g. melanoma) in which the targeted mechanism may not even be in play. Serial biopsies even FNAs are difficult to obtain in the clinic (patients often resist them), and can dramatically slow down enrollment. Biopsy-based biomarkers are usually obtained at a single dose at or near the MTD. This leads to a lack of exposure/response data and makes it difficult to define an optimal biological dose. Levels of biomarker variability within tumors in real patients are typically much higher than in cell lines used preclinically. Typical Phase 1 (and even Phase 1b) sample sizes are far underpowered to show significant treatment effects on these biomarkers. Since each anti-cancer mechanism is different, one must develop and validate new biomarkers for every program with little cross-project learning or utility.

4 Functional Imaging can provide POM and is the preferred clinical biomarker Virtually all anti-cancer agents, regardless of their proximate mechanism of action, ultimately exert their activity by interfering with tumor metabolism, or proliferation, inducing tumor apoptosis, or reducing tumor microvascular density. These anti-tumor effects are all image-able (both preclinically and clinically) using non-invasive imaging techniques. Drug effects on functional imaging biomarkers are typically observable very quickly usually far in advance of any morphological or clinical effects on tumor lesions. Preclinical translational biomarker work should focus on establishing: - Which imaging biomarker is most appropriate to the target? - When is the optimal time point to collect the imaging data? - What is the quantitative relationship between the imaging endpoint and any true target/mechanism biomarker (e.g. phosphorylation status of a kinase substrate). - What is the relationship of the imaging endpoint to drug exposure?

5 Can functional imaging provide true Proof of Mechanism? In some cases, yes. For example, a cdk 4/6 inhibitor would be expected to induce cell cycle arrest in a proliferating tumor at the G1 phase, and prevent the entry into S-phase. A hallmark of entry into S-phase is the up-regulation of thymidine kinase in the cell, and the resultant phosphorylation of thymidine, which becomes trapped inside the cell. Hence, imaging of a tumor with the thymidine analog and PET tracer 18 FLT would be expected to show a significant decrease in signal after exposure to a cdk inhibitor relative to baseline. FLT uptake then, is a classic mechanism biomarker for this target and can be used to infer that such a molecule has or has not hit and modulated its intended target. In many cases, no. For example, a cytotoxic agent might be expected to reduce FDG uptake in a tumor simply because, as the cells die off, the tumor as a whole is less metabolically active. In this case, FDG uptake is not a true mechanism biomarker, but is really an Outcome biomarker. The risk in assessing POM using an Outcome biomarker is that one cannot know whether a failure to observe the desired change in the biomarker endpoint is a result of failure to hit and modulate the intended target ( bad drug ) or whether actual modulation of the target was achieved but did not have the desired effect on the tumor ( bad target ). Nevertheless, we are willing to abandon both the compound and the mechanism based on a negative study result if certain conditions are met as outlined in the next slide.

6 Avoiding false negatives Failure to achieve POM based upon a biomarker study (imaging or otherwise) is grounds for program termination. It is therefore critical that the program be given fair chance to succeed by ensuring that the POM study represents the best possible test of the target and clinical hypothesis. This requires that: The POM study is performed at a sufficiently high dose (usually the MTD)*. The study is performed in an expanded cohort appropriately sized and powered to demonstrate a significant treatment effect on the biomarker (see later slide). The expanded cohort POM study is performed in an appropriate patient population in which the target is known to be relevant. There are preclinical data demonstrating the association between Imaging/Outcome marker and an appropriate Target/Mechanism marker.

7 Selecting patients for expanded cohort POM studies Patients for a POM expanded cohort should have tumors in which the drug target is known in advance to be relevant (activating mutation, over-expression, etc.). Tumors should not have known activating mutations in signal transducing molecules located downstream of the drug target (e.g. an EGFR inhibitor may be ineffective in a tumor with mutant K-ras). Sometimes selection can be made solely by resort to particular tumor types CML for bcr-abl, Mantle cell lymphoma for cdk, GIST for kit, etc. Sometimes selection can be made by molecular analysis of archival biopsy. The POM expanded cohort should represent the best test of the molecular and clinical hypothesis: If the test drug has no effect in this population, it is unlikely to work anywhere else. Important components of the PRISM process are the Molecular Profiling and Translational Research plans that will enable the POM patient selection strategy.

8 What is a significant pharmacodynamic effect? It is recognized that the Phase 1 setting is not the place to require rigorous statistical demonstration of pharmacodynamic effects. What is desired is more of a trend towards appropriate modulation of the mean biomarker value in the post- vs. pre-treatment expanded cohort population. This is fit for purpose of supporting the decision to move into larger Phase 2 studies. In general, the expanded cohort study should be sized to have an 80% power to detect a biomarker change between pre- and post-treatment mean values (across the entire cohort) with a p value (α value) of < For typical imaging biomarker studies in which the C.V. is ~15-20% and a clinically meaningful effect size is on the order of 25-30%, this will mean a cohort size of ~ 5-10 patients. It is recognized that patients who represent the best test of the clinical hypothesis may be rare; hence, recruitment of the POM cohort may be slow. To mitigate the risk of unduly slowing a promising candidate, if the first three patients in the study ALL show an effect size of greater than 25% (e.g. reduction from baseline in mean PET SUV of target lesions from 5 to 3.75), POM will be considered achieved.

9 Imaging Pharmacodynamics to Select Dose? The vast majority of oncology agents are dose-selected by toxicity and administered at the maximum tolerated dose (MTD). Is there enough dynamic range and precision in imaging endpoints to generate complete exposure/response models to allow selection of optimal biologic dose? It may be possible to use imaging-based pharmacodynamics to ascertain whether maximum pharmacodynamic effect occurs at some dose lower than MTD potentially allowing Phase 2 testing at more than a single dose as is routine in most other therapeutic areas. The potential utility of this approach is currently under exploration using dose de-escalation strategies in Phase 1. The use of doses below MTD in an oncology setting will depend in part on the nature, severity and manageability of the dose limiting toxicities.

10 Oncology Phase 1: Towards a standard paradigm Dose escalate in standard all-comers phase I design (usually 3 pts per dose) to obtain PK, safety, tolerability and MTD no imaging/biomarkers need be collected at this point. Enroll expanded of cohort (powered appropriately as discussed) of selected patients at MTD for collection of baseline and post-treatment functional images. Proof of mechanism is defined as >25% response in each of the first three patients in the expanded cohort OR significant response across the entire cohort at a p value of <0.2. Absent POM, the program is terminated. Objective tumor response trumps negative biomarker data and constitutes esoe.

11 Does it work? Pre-treatment Baseline FLT PET (Day -6) FLT PET Post-treatment treatment (Day 21) Note dramatic decrease in uptake (intensity) of the PET tracer in i n virtually all lesions in the post-treatment treatment scan relative to the baseline scan.

12 Mean lesion SUV after one treatment cycle: Drug X P < baseline Cycle 1

13 Change in mean lesion SUV over time: Drug Y Mean Lesion SUV (FDG-PET) B C1 d14 C1 d21 C1 d28 C2 d14 C1 d28

14 Dose/Response of SUV: Drug Z Mean SUV * * QD Dose (mg/kg) * P < 0.2 Baseline SUV Cycle 1 SUV