Oncology Dose Finding A Case Study: Intra-patient Dose Escalation Jonas Wiedemann, Meghna Kamath Samant & Dominik Heinzmann, pRED Biostatistics, Valerie Cosson & Sylvie Retout, pRED TRS F. Hoffmann-La Roche picture placeholder

Outline Oncology Dose Finding Why is This of Interest? Imaging Study - Intra-patient Dose Escalation – Pros & Cons Why is This of Interest? Imaging Study Statistical Methodology Lessons Learned & Further Development

Statistical Methodology Lessons Learned & Further Development Oncology Dose Finding - Intra-patient Dose Escalation – Pros & Cons Why is This of Interest? Imaging Study Statistical Methodology Lessons Learned & Further Development

Oncology Dose Finding Overview Several different approaches are more or less commonly seen: Conventional rule based “3+3” Continual Reassessment Methodology (CRM) More advanced methods combining toxicity and efficacy Intra-patient Dose Escalation Commonly acknowledged that more advanced and innovative methods are needed using accumulated information – such as Bayesian methodologies

FDA point of view A need for innovative designs Increasing spending of biomedical research does not reflect an increase of the success rate of pharmaceutical development. Many drug products were recalled due to safety issues after regulatory approval. Critical path initiative In its 2004 Critical Path Report, the FDA presented its diagnosis of the scientific challenges underlying the medical product pipeline problems. Advancing innovative trial designs: Use of prior experience or accumulated information in trial design Insufficient exploration of the dose-response curve is often a key shortcoming of clinical drug development In 2004 the Critical Path Initiative was developed to address the growing concern that many of the new basic science discoveries made in recent years may not quickly yield more effective, more affordable, and safe medical products for patients. The Critical Path Initiative proposed the development of a toolkit which would contain powerful new scientific and technical methods, including new clinical evaluation techniques to improve the efficiency and effectiveness of the clinical trial process, including trial design, endpoints, and analyses. In the 2006 Critical Path Opportunities List, the Adaptive Trial Design was specifically proposed.

Accelerated Titration Designs A direct comparison to “3+3” In 2008 Penel et. al. compared the performance of ATD and “3+3” in 270 (1997–2008) published phase I trials ATD had been used in only 10% of the these studies ATD had permitted to explore significantly more dose levels (seven vs. five) ATD reduced the rate of patients treated at doses below phase-2 recommended dose (46% vs. 56%,) Nevertheless, ATD did not allow a reduction in the number of enrolled patients, shorten the accrual time nor increase the efficacy However, still support ATD as an effective clinical trial design over a standard “3+3”

Intra-patient Dose Escalation Pros & cons Intra-patient dose escalation designs are generally used in ethical grounds, i.e. to address the fact that in cancer research it may be unethical to only provide sub therapeutic doses to cohorts of patients Fewer patients needed, i.e. lower costs, faster study conduct Meaningful if no toxicity is expected If analyzed properly, they can provide information about inter-patient variability in dose–response effects The succession of dose levels is not necessarily determined completely by choices made before the onset of the trial

Intra-patient Dose Escalation Pros & cons However, though appealing these designs are not commonly applied due to some theoretical and practical objections Successive observations in a single patient are correlated. Hence, difficult to know if toxicity is due to current dose or cumulative exposure (same potential issue for PD markers) May not be feasible due to the fact that most patients in phase 1 studies would only stay on drug for 2 to 3 cycles of therapy due to rapidly progressive disease Could potentially create some selection bias (prognostics, characteristics, etc.)

Oncology Dose Finding - Intra-patient Dose Escalation – Pros & Cons Why is This of Interest? Imaging Study Statistical Methodology Lessons Learned & Further Development

Why is This of Interest? Project overview Anti-body, angiogenesis inhibitor (inhibits growth of new blood vessels, especially by inhibiting vascular permeability) Tested in first-in-man multiple dose ascending study with a dose of up to 3 mg/kg, no observed toxicity, and a ½ life of ~ 9 days Dose schedule simulated and a q2w approach chosen DCE-MRI* as angiogenic PD marker – values (Ktrans, Kep, AUC90, Ve) directly related to: Blood volume Blood flow Extracellular Extra-vascular Space - ESS Rate of extravasation In addition, low within-patient variability * Dynamic Contrast Enhanced-Magnetic Resonance Imaging

DCE-MRI methodology – Excellent reproducibility ml/ml/min 2 paired pre-treatment scans (Ktrans: wSD ~ 0.10-0.11)

Why is This of Interest? Decision to go for intra-patient dose escalation Angiogenesis inhibition confirmed and DCE-MRI as angiogenic PD marker – low within-patient variability No observed toxicity and tentative dose found in first-in-man study – However, still uncertainty about actual therapeutic dose −> alternative approach needed Modeling and simulation methods explored and tools in place, i.e. Bayesian, WinBugs, EDC, etc. −> practical feasible By introducing large dose-escalating steps / relatively short half life −> faith in observed Toxicity/PD dose- response Phase I intra-patient dose escalation imaging study to establish PD dose-relationship measured as DCE-MRI

Oncology Dose Finding - Intra-patient Dose Escalation – Pros & Cons Why is This of Interest? Imaging Study Statistical methodology Lessons Learned & Further Development

Imaging Study Overall target Establish exposure – PD relationship for single agent Identify the minimal PD effective dose Confirm MoA Confirm feasibility of DCE-MRI 100 250 750 2500 3000 Dose (mg) by applying intra-patient dose escalation with 3 initial dose steps by applying a Bayesian approach

Study Overview Initial Test Cohort 6-10 subjects Study Overview Highest dose DCE-MRI signal DCE-MRI signal First intra-patient Dose Escalation Cohort 6-10 subjects Terminate study non-interpretable DCE-MRI signal Adapted Intra-patient Dose Escalation Cohorts 6-10 subjects pr cohort Parallel Fixed Dose Cohorts 6-10 subjects pr cohort Allows timing of PD/BM adjustment dose scheme adjustment Adapted Confirmatory Parallel Fixed Dose Cohorts 6-10 subjects pr cohort Allows Further adjustment of timing and no of assessments Tumor Biopsy Evaluation Cohort 10 subjects on lowest efficacious dose Up to 50 subjects will be evaluated in total

Oncology Dose Finding - Intra-patient Dose Escalation – Pros & Cons Why This Interest? Imaging Study Statistical methodology Lessons Learned & Further Development

Primary PK/PD Modeling Bayesian approach – Primary model A direct* inhibitory Imax model Two unknown parameters to be estimated, i.e. Imax and IC50 (both assumed to be Gaussian distributed with mean and precision) With E the DCE-MRI parameter, i.e. Ktrans, Kep, Ve, Vp and iAUC, E0 the DCE-MRI parameter at baseline, Cp the drug concentration at the time of DCE-MRI assessment, Imax the maximum decrease of the DCE-MRI parameter (0<Imax<1), IC50 the drug concentration at which 50% of max inhibition is reached. * If possible, an exploratory indirect model to investigate time delay in DCE-MRI

Primary PK/PD Modeling Bayesian approach – General principles - unknown parameters are interpreted in terms of probability Prior distribution on IC50 (and Imax) + Observed PD data ~ A posterior mean value and precision Bayesian estimation

Bayesian Method Advantages Combines a priori knowledge, including uncertainty, with new data Allows an increase of that knowledge, even with a low number of subjects Basis for formal approach to incremental model building, parameter estimation and other statistical inference as knowledge and data are accumulated Implemented in Winbugs 1.4.3 Issues Construction of prior distributions is a somewhat subjective process Apparently very sensitive to the choice of the priors Bayesian inference is based on Monte Carlo Markov Chain Iterative process which eventually converges to the posterior distribution Requires high number of samples (5000 – 10000) => time consuming

Oncology Dose Finding - Intra-patient Dose Escalation – Pros & Cons Why This Interest? Imaging Study Statistical methodology Lessons Learned & Further Development

Lessons Learned - so far Regulatory feedback (EU) Study approved in 3 EU countries without major issues: Validation of analytical methods required for future studies Concern about high dose for Initial Test Cohort Feedback from clinicians/operational Internal Open minded lead clinician – could have been an issue!!! Some opposition from operational External Investigators very open and helpful in setting up study Status: Study still ongoing – 4 patients enrolled in Initial Test Cohort Status: Good feedback on DCE-MRI data quality However, some issues with too large tumors since DCE-MRI here is less sensitive

Further Development Current dilemmas? Phase Ib/IIa combination study planed in recurrent Glioblastoma (GBM) Target: to estimate the treatment benefit of combined treatment (with launched anti-angiogenic agent) Endpoint: Progression-free-survival DCE-MRI as PD and clinical marker? Future dose when moving into a combination treatment Should be based on a toxicity/efficacy trade off? Possibility to adjust the dose of the launched agent? Phase 3 gating? Further disease areas? – difficulties in generalizing

References Simon, R. Accelerated Titration Designs for Phase I Clinical Trials in Oncology, JNCI, 1997 Orloff, J. The future of drug development: advancing clinical trial design, NATURE, 2009 Whitehead, J. Easy-to-implement Bayesian methods for dose-escalation studies in healthy volunteers, Biostatistics, 2001 Thall, P. F., Dose-Finding Based on Efficacy-Toxicity Trade-Offs, Biometrics, 2004 Chang, M. A Hybrid Bayesian Adaptive Design for Dose Response Trials, Journal of Biopharmaceutical Statistics, 2005 Penel, N., “Classical 3+3 design” versus “accelerated titration designs”: analysis of 270 phase 1 trials investigating anti- cancer agents, Invest New Drugs, 2009

Thanks! Contact info: Jonas.wiedemann@roche.com

We Innovate Healthcare