An introduction to Phase I dual-agent dose escalation trials

Similar documents
Bayesian Model Averaging Continual Reassessment Method BMA-CRM. Guosheng Yin and Ying Yuan. August 26, 2009

Bayesian Adaptive Designs for Early-Phase Oncology Trials

Continual Reassessment Method

Web-based Supplementary Materials

Bayesian Phase I/II clinical trials in Oncology

CLINICAL TRIALS: Part 2 of 2

A Simple Approach For Incorporating Multiple Toxicity Thresholds In Phase I Trials

Bayesian adaptive designs, with particular reference to dose finding methods in oncology

The CRM for ordinal and multivariate outcomes. Elizabeth Garrett-Mayer, PhD Emily Van Meter

Package bcrm. September 24, 2015

ADDPLAN DF: Designing More Accurate Dose Escalation Studies in Exploratory Drug Development. An ICON White Paper

Bayesian Model Averaging CRM in Phase I Clinical Trials

Adaptive Design for Intra Patient Dose Escalation in Phase I Trials in Oncology

Assessing the Effect of Practical Considerations when using the CRM in Dose Finding Studies

The Promise and Challenge of Adaptive Design in Oncology Trials

Specification of the Bayesian CRM: Model and Sample Size. Ken Cheung Department of Biostatistics, Columbia University

Extending the TITE CRM to Multiple Outcomes with Application to a Phase 1 Clinical Trial in Canine Hemangiosarcoma

Journal of Statistical Software

Clinical Trial Design. Sponsored by Center for Cancer Research National Cancer Institute

The Trials and Tribulations of the CRM: the DFO Experience

Session 6 Clinical Trial Assessment Phase I Clinical Trial

THE RAPID ENROLLMENT DESIGN FOR PHASE I CLINICIAL TRIALS

Robust EM Continual Reassessment Method in Oncology Dose Finding

Likelihood Approaches for Trial Designs in Early Phase Oncology

The Clinical Trials Process an educated patient s guide

By Ying Yuan 1 and Guosheng Yin 2 University of Texas and University of Hong Kong

Establish the Maximum Tolerated Dose in Phase-I Trials using 3+3 Method

TITE-CRM Phase I Clinical Trials: Implementation Using SAS

Package CRM. R topics documented: February 19, 2015

TGN 1412 Welche Änderungen haben sich für die Erstanwendung am Menschen aus Sicht des BfArM ergeben?

Endpoint Selection in Phase II Oncology trials

white paper Adaptive Trial Designs to Advance Oncology Research

Definition of dose-limiting toxicity in phase I cancer clinical trials of molecularly targeted agents

Interactive Software Isotonic Design using Normalized Equivalent Toxicity Score (ID-NETS TM ) for Cancer Phase I Clinical Trials

Clinical Trial Designs for Incorporating Multiple Biomarkers in Combination Studies with Targeted Agents

Phase 1 Trial Design: Is the Best?

Dose-Finding with Two Agents in Phase I Oncology Trials

Longitudinal Modeling of Lung Function in Respiratory Drug Development

Overview of Phase 1 Oncology Trials of Biologic Therapeutics

Not All Clinical Trials Are Created Equal Understanding the Different Phases

Practical considerations in designing a phase I Time to Event Continual Reassessment Method (TiTE CRM) trial in a grant funded CTU

Accelerated Titration Designs for Phase I Clinical Trials in Oncology

Guidance for Industry

COMMITTEE FOR PROPRIETARY MEDICINAL PRODUCTS (CPMP) NOTE FOR GUIDANCE ON THE PRE-CLINICAL EVALUATION OF ANTICANCER MEDICINAL PRODUCTS

Newsletter. WntResearch AB, Medeon Science Park, Per Albin Hanssons väg 41, Malmö, Sweden. Primary Objective:

U.S. Contract Research Outsourcing Market: Trends, Challenges and Competition in the New Decade. N8B7-52 December 2010

Novel Designs for Oncology Clinical Trials. Marc Hoffman, MD Chief Medical Officer

Using the package crmpack: introductory examples

MOLOGEN AG. Q1 Results 2015 Conference Call Dr. Matthias Schroff Chief Executive Officer. Berlin, 12 May 2015

PEER REVIEW HISTORY ARTICLE DETAILS TITLE (PROVISIONAL) An exploratory study to assess the association between healthrelated

Personalized Predictive Medicine and Genomic Clinical Trials

William B. Smith, MD President. Patrick R. Ayd, RN, MBA Chief Operating Officer. SNBL-CPC Baltimore, Maryland

WHITE PAPER: ADAPTIVE CLINICAL TRIALS

MATHEMATICAL MODELS OF TUMOR GROWTH INHIBITION IN XENOGRAFT MICE AFTER ADMINISTRATION OF ANTICANCER AGENTS GIVEN IN COMBINATION

NONCLINICAL EVALUATION FOR ANTICANCER PHARMACEUTICALS

Guidance for Industry

Molecular markers and clinical trial design parallels between oncology and rare diseases?

Statistics and Pharmacokinetics in Clinical Pharmacology Studies

CTC Technology Readiness Levels

Sheffield Kidney Institute. Planning a Clinical Trial

Therapeutic Goods Administration Orphan Drugs Program: Discussion paper

PubH 7470: STATISTICS FOR TRANSLATIONAL & CLINICAL RESEARCH

First In Human Pediatric Trials and Safety Assessment for Rare and Orphan Diseases

Adocia reports positive results from phase IIa clinical study of ultra-fast acting BioChaperone Lispro

Clinical Trial Results Database Page 1

Logistic Regression. Jia Li. Department of Statistics The Pennsylvania State University. Logistic Regression

U.S. Food and Drug Administration

Careers in Biostatistics and Clinical SAS Programming An Overview for the Uninitiated Justina M. Flavin, Independent Consultant, San Diego, CA

NATIONAL HEALTH COUNCIL RESOLUTION Nº 251, DATED 7 AUGUST 1997

Glossary of Methodologic Terms

Ethics and Scientific Oversight for Phase 1 Clinical Trials in Hong Kong. Sydney TANG Chairman, HKU/HA HKW IRB November 21, 2015

Adaptive designs for time-to-event trials

Design and Analysis of Phase III Clinical Trials

Regulatory Pathways for Licensure and Use of Ebola Virus Vaccines During the Current Outbreak FDA Perspective

Uptake of adaptive and Bayesian methods in the design of early phase trials within CTUs

Cancer Biostatistics Workshop Science of Doing Science - Biostatistics

ICH M3 (R2) Guideline on Nonclinical Safety Studies for the Conduct of Human Clinical Trials and Marketing Authorization for Pharmaceuticals

Pharmacotherapy/Psychotherapy Research: Psychotherapy Research T O P C O N F E R E N C E S H A N G H A I, C H I N A A L L A N T A S M A N, M. D.

CLINICAL TRIALS SHOULD YOU PARTICIPATE? by Gwen L. Nichols, MD

Anticancer Drug Clinical Trial Guideline. (version 2.0)

Practical aspects of early phase oncology trials the oncologists view

Risks and Benefits Associated With Novel Phase 1 Oncology Trial Designs. BACKGROUND. Although aggressive dose escalation strategies were designed to

Dose Escalation Methods in Phase I Cancer Clinical Trials

DRUG APPROVAL PROCESS FOR THE TREATMENT OF ALZHEIMER S DISEASE

The Science of Chemical Safety Essential Toxicology - 4. Hazard and Risk. John Duffus & Howard Worth. IUPAC Educators Resource Material IUPAC

Perspectives on Patient Recruitment

Drug Development Process

Clinical trials in haemophilia

ECRIN IA WP4 Breakout session: structuring a European rare disesase clinical research network ECRIN Annual meeting may 2013 Warsaw

Ovarian Cancer and Modern Immunotherapy: Regulatory Strategies for Drug Development

Statistical Considerations for Phase II Trials

People have thought about, and defined, probability in different ways. important to note the consequences of the definition:

Challenges in the Regulation of Pediatric Clinical Trials

Practical Considerations for Adaptive Trial Design and Implementation

Decision trees DECISION NODES. Kluwer Academic Publishers / X_220. Stuart Eriksen 1, Candice Huynh 1, and L.

Research on Research: Learning about Phase 1 Trials

Optimizing Early Phase Oncology Clinical Trials

Statfax. 3 Chapter website: Spring Meeting Workshop

Generating Randomization Schedules Using SAS Programming Chunqin Deng and Julia Graz, PPD, Inc., Research Triangle Park, North Carolina

Pharmacology skills for drug discovery. Why is pharmacology important?

Transcription:

An introduction to Phase I dual-agent dose escalation trials Michael Sweeting MRC Biostatistics Unit, Cambridge 13th March 2013

Outline Introduction to phase I trials Aims, objectives and key elements. The CRM method. Dual-agent phase I trials Escalation strategies. Patient vs. Population gain. Toxicity vs. Efficacy trade-off.

Introduction to Phase I trials Aims Phase I studies involve the first experimentation of a new drug / clinical procedure in human subjects. To seek the highest possible dose subject to toxicity constraints:- This is known as the maximum tolerated dose (MTD). Based on a monotonicity assumption that the benefit (efficacy) of treatment increases with dose. Sample size Typically small: 20-50 patients. Due to ethical considerations patients are added sequentially after side-effects from previous patients have been assessed. Subjects Healthy volunteers for relatively non-toxic agents (e.g. molecular entities for CNS disorders). Patients when drugs are highly toxic (e.g. in oncology).

Introduction to Phase I trials Aims Phase I studies involve the first experimentation of a new drug / clinical procedure in human subjects. To seek the highest possible dose subject to toxicity constraints:- This is known as the maximum tolerated dose (MTD). Based on a monotonicity assumption that the benefit (efficacy) of treatment increases with dose. Sample size Typically small: 20-50 patients. Due to ethical considerations patients are added sequentially after side-effects from previous patients have been assessed. Subjects Healthy volunteers for relatively non-toxic agents (e.g. molecular entities for CNS disorders). Patients when drugs are highly toxic (e.g. in oncology).

Introduction to Phase I trials Aims Phase I studies involve the first experimentation of a new drug / clinical procedure in human subjects. To seek the highest possible dose subject to toxicity constraints:- This is known as the maximum tolerated dose (MTD). Based on a monotonicity assumption that the benefit (efficacy) of treatment increases with dose. Sample size Typically small: 20-50 patients. Due to ethical considerations patients are added sequentially after side-effects from previous patients have been assessed. Subjects Healthy volunteers for relatively non-toxic agents (e.g. molecular entities for CNS disorders). Patients when drugs are highly toxic (e.g. in oncology).

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

Key elements 1. A starting dose that will be given to the first patient:- In psychiatric trials: no observable effect level (NOEL) in animals. In oncology trials: 1 10 LD 10 in mice (one tenth of the lethal dose in 10% of mice). 2. Potential doses for experimentation. 3. Dose regime (single dose or multiple doses per individual). 4. Outcomes:- Safety (Toxicity) outcomes: often binary (e.g. occurrence of a dose-limiting toxicity (DLT) is used in cancer trials). PD measures of efficacy: often biomarker based. 5. A target level (TL) or therapeutic index:- The desired target level that corresponds to the MTD (e.g. cancer trials often propose 30% prevalence of DLT at the MTD). 6. A dose-escalation design:- Rule or model based. Cohort size:- Number individuals treated at each dose level. Sample size / stopping rules.

A dose-response curve Response θ 1 2 3MTD 4 Dose level 5 6 θ = Target level

The Continual Reassessment Method (CRM) The CRM is an adaptive design that adjusts the dose for the next cohort based on all information accrued to date. The aim is to choose the dose that is closest to the target level. d(i): dose administered to patient i. π(d;α): probability of a DLT at dose d with parameter α. f (α): prior distribution for α. Posterior distribution for α: f (α y n ) = f (α)l(α;y n ) 0 f (u)l(u;y n ) du. The dose whose posterior probability of DLT closest to the TTL, θ, is chosen, that is:- d(n + 1) = arg min E [π(ξ ;α)] θ ξ {1,...,k}

The Continual Reassessment Method (CRM) The CRM is an adaptive design that adjusts the dose for the next cohort based on all information accrued to date. The aim is to choose the dose that is closest to the target level. d(i): dose administered to patient i. π(d;α): probability of a DLT at dose d with parameter α. f (α): prior distribution for α. Posterior distribution for α: f (α y n ) = f (α)l(α;y n ) 0 f (u)l(u;y n ) du. The dose whose posterior probability of DLT closest to the TTL, θ, is chosen, that is:- d(n + 1) = arg min E [π(ξ ;α)] θ ξ {1,...,k}

The Continual Reassessment Method (CRM) The CRM is an adaptive design that adjusts the dose for the next cohort based on all information accrued to date. The aim is to choose the dose that is closest to the target level. d(i): dose administered to patient i. π(d;α): probability of a DLT at dose d with parameter α. f (α): prior distribution for α. Posterior distribution for α: f (α y n ) = f (α)l(α;y n ) 0 f (u)l(u;y n ) du. The dose whose posterior probability of DLT closest to the TTL, θ, is chosen, that is:- d(n + 1) = arg min E [π(ξ ;α)] θ ξ {1,...,k}

Example escalation scheme 1.00 Cohort 0 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 1 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 2 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 3 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 4 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 5 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 6 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 7 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 8 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 9 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Example escalation scheme 1.00 Cohort 10 0.75 Probability of DLT 0.50 5 0 1 2 3 4 5 Dose level Previous slide Next slide

Extending to dual-agent phase I trials Drug B Synergy Antagonism Probability DLT 0.1 0.3 0.5 0.7 Drug A

Comparison of single and dual-agent phase I trials Aim Single-agent To identify a single maximum tolerated dose (MTD). Dual-agent To identify numerous MTD combinations. Escalation Single-agent Quick and safe escalation to the target level (TL). Dual-agent Quick and safe escalation to the TL, with varied experimentation of doses at the TL. Prior information Single-agent Pre-clinical, Learning from similar compounds. Dual-agent Pre-clinical, Learning from similar compounds and combinations, Results from single-agent Phase I-III studies.

Comparison of single and dual-agent phase I trials Aim Single-agent To identify a single maximum tolerated dose (MTD). Dual-agent To identify numerous MTD combinations. Escalation Single-agent Quick and safe escalation to the target level (TL). Dual-agent Quick and safe escalation to the TL, with varied experimentation of doses at the TL. Prior information Single-agent Pre-clinical, Learning from similar compounds. Dual-agent Pre-clinical, Learning from similar compounds and combinations, Results from single-agent Phase I-III studies.

Comparison of single and dual-agent phase I trials Aim Single-agent To identify a single maximum tolerated dose (MTD). Dual-agent To identify numerous MTD combinations. Escalation Single-agent Quick and safe escalation to the target level (TL). Dual-agent Quick and safe escalation to the TL, with varied experimentation of doses at the TL. Prior information Single-agent Pre-clinical, Learning from similar compounds. Dual-agent Pre-clinical, Learning from similar compounds and combinations, Results from single-agent Phase I-III studies.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Series of 1-agent CRMs Fix one drug at a single dose, escalate the other, in a series of sub-trials. Identified MTD from one sub-trial can be used to limit dose range for next sub-trial. 1 6 5 Drug B 4 3 2 1 1 2 3 4 5 6 Drug A 1 Yuan, Y. and Yin, G. Statistics in Medicine 2008; 27: 5664-78.

Using 1-dimensional CRM with partial ordering If a prior ordering of the dose combinations can be assumed then a 1-dimensional CRM can be applied. 2 Alternatively, the likelihood of each possible ordering can be assessed after each patient is recruited. 3 Can become problematic if many dose levels are used (many possible orderings). Drug B 2 1 1 2 3 Drug A 2 Kramar, A. et al. Statistics in Medicine 1999; 18: 1849-64. 3 Wages, N. A. et al. Clinical Trials 2011; 8: 380-89.

Using 1-dimensional CRM with partial ordering If a prior ordering of the dose combinations can be assumed then a 1-dimensional CRM can be applied. 2 Alternatively, the likelihood of each possible ordering can be assessed after each patient is recruited. 3 Can become problematic if many dose levels are used (many possible orderings). Drug B 2 1 1 2 3 Drug A 2 Kramar, A. et al. Statistics in Medicine 1999; 18: 1849-64. 3 Wages, N. A. et al. Clinical Trials 2011; 8: 380-89.

Using 1-dimensional CRM with partial ordering If a prior ordering of the dose combinations can be assumed then a 1-dimensional CRM can be applied. 2 Alternatively, the likelihood of each possible ordering can be assessed after each patient is recruited. 3 Can become problematic if many dose levels are used (many possible orderings). Drug B 2 1 1 2 3 Drug A 2 Kramar, A. et al. Statistics in Medicine 1999; 18: 1849-64. 3 Wages, N. A. et al. Clinical Trials 2011; 8: 380-89.

Using 1-dimensional CRM with partial ordering If a prior ordering of the dose combinations can be assumed then a 1-dimensional CRM can be applied. 2 Alternatively, the likelihood of each possible ordering can be assessed after each patient is recruited. 3 Can become problematic if many dose levels are used (many possible orderings). Drug B 2 1 1 2 3 Drug A 2 Kramar, A. et al. Statistics in Medicine 1999; 18: 1849-64. 3 Wages, N. A. et al. Clinical Trials 2011; 8: 380-89.

Escalation strategies for 2-dimensional models: Admissible next doses 4 Diagonal (or Fast ) escalation. Non-diagonal (or Slow ) escalation. 6 6 5 5 Drug B 4 3 Drug B 4 3 2 2 1 1 1 2 3 4 5 6 Drug A 1 2 3 4 5 6 Drug A A third strategy: Allow any previously experimented dose combination to be admissible. 4 Sweeting, MJ. and Mander, AP. Pharm. Stats. 2012;11(3):258-266.

Escalation strategies: Decision rules Patient gain The next cohort is treated at our best estimate of the MTD. Patient gain used in most CRM methodology. Variance (population) gain The next cohort is treated at the dose that will provide the most information regarding the dose-toxicity surface. This dose is the one that will provide most benefit to the population (by giving more precise estimates at the end of the trial). Hybrid patient/variance gain In dual-agent trials, we want doses close to the TL, but also allow experimentation across the dose-toxicity surface. Hence this requires a trade-off between Patient and Population Gain.

Escalation strategies: Decision rules Patient gain The next cohort is treated at our best estimate of the MTD. Patient gain used in most CRM methodology. Variance (population) gain The next cohort is treated at the dose that will provide the most information regarding the dose-toxicity surface. This dose is the one that will provide most benefit to the population (by giving more precise estimates at the end of the trial). Hybrid patient/variance gain In dual-agent trials, we want doses close to the TL, but also allow experimentation across the dose-toxicity surface. Hence this requires a trade-off between Patient and Population Gain.

Escalation strategies: Decision rules Patient gain The next cohort is treated at our best estimate of the MTD. Patient gain used in most CRM methodology. Variance (population) gain The next cohort is treated at the dose that will provide the most information regarding the dose-toxicity surface. This dose is the one that will provide most benefit to the population (by giving more precise estimates at the end of the trial). Hybrid patient/variance gain In dual-agent trials, we want doses close to the TL, but also allow experimentation across the dose-toxicity surface. Hence this requires a trade-off between Patient and Population Gain.

Escalation strategies: Decision rules Patient gain The next cohort is treated at our best estimate of the MTD. Patient gain used in most CRM methodology. Variance (population) gain The next cohort is treated at the dose that will provide the most information regarding the dose-toxicity surface. This dose is the one that will provide most benefit to the population (by giving more precise estimates at the end of the trial). Hybrid patient/variance gain In dual-agent trials, we want doses close to the TL, but also allow experimentation across the dose-toxicity surface. Hence this requires a trade-off between Patient and Population Gain.

An example trial in practice: using non-diagonal escalation n = 0 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 1 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 2 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 3 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 4 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 5 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 6 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 7 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 8 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 9 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 10 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 11 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 12 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 13 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 14 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 15 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 20 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: using non-diagonal escalation n = 30 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A Previous slide Next slide

An example trial in practice: diagonal escalation with population gain function n = 30 1.0 Drug B Probability DLT 0.1 0.3 0.5 0.7 1.0 Drug A

Incorporating measures of efficacy There is increasing interest in developing dose finding methods that incorporate both toxicity and efficacy endpoints. Considered as a phase I-II trial. A trade-off must be made between efficacy and toxicity. Response Toxicity Efficacy MED MTD Dose

Incorporating measures of efficacy There is increasing interest in developing dose finding methods that incorporate both toxicity and efficacy endpoints. Considered as a phase I-II trial. A trade-off must be made between efficacy and toxicity. Response Toxicity Efficacy MED MTD Dose

Incorporating measures of efficacy There is increasing interest in developing dose finding methods that incorporate both toxicity and efficacy endpoints. Considered as a phase I-II trial. A trade-off must be made between efficacy and toxicity. Response Toxicity Efficacy MED MTD Dose

Incorporating measures of efficacy There is increasing interest in developing dose finding methods that incorporate both toxicity and efficacy endpoints. Considered as a phase I-II trial. A trade-off must be made between efficacy and toxicity. Response Toxicity Efficacy MED MTD Dose

EffTox contours of desirability 5 Dose-Finding Based on Efficacy Toxicity Trade-Offs Prob(TOXICITY) 1.0 π * 3 q π * 1 1.0 p Prob(EFFICACY) Figure 1. Efficacy toxicity trade-off contours for the 5 Pentostatin trial. The target contour C is given by the solid Thall, PF. and Cook, line, and JD. the Biometrics three elicited2004; target points 60: 684-693. that determine C are C L(q) π * 2 This may be used to construct a that partition Π, with the points desirable. Given any p Cand z> hz (p) =q if q L(p) and ρ(p)/ρ hz (p) =(1 (1 pe )/z, pt /z). W not be a subset of Π for some z, sin contours inside Π we define Cz = { contour in Π obtained by shifting e point q in Π such that ρ(p)/ρ(q) = ordered by their desirabilities and contours is a partition of Π, the co on Π. We require f(πe) to be strictly given (πe,πt ) C and ɛ>0, prov Π, it must be the case that (πe + ɛ C and hence is more desirable than (πe, πt + ɛ) Π must be on a con desirable than (πe, πt ). In partic binary outcome case, the rectang line segments from (π E, πt )to(π (1, πt ) does not satisfy this admis the convenient form πt = f(πe) the applications described here, fit pairs subject to the constraint th πe such that {πe,f(πe)} C. O elliptical contour, should work as w 4.2 The Trade-Off-Based Algorithm Initially, the physician must provid dose for the first cohort, N, c, a used in the acceptability criteria

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.

Summary Escalation strategies more complex in two-agent trials. Generally require twice as many patients as single-agent trials. May wish for varied experimentation around MTD contour:- Allows more drug combinations to be recommended for phase II experimentation. Such an objective requires sequential learning about MTD contour to choose next optimal dose. Non-diagonal escalation rarely behave in a step-like manner, and may get stuck in regions where one drug is given at a low dose. Need to consider carefully trade-off between:- Patient and Population gain. Toxicity and Efficacy. In summary, more consideration needs to be given to these design aspects in order to gain optimal information from the available patients. Most of these designs are disappointingly under used in practice.