Phase I Trial Designs Jud Blatchford, PhD BIOM 6649 Clinical Trials April 9th, 2015 Table of Contents 1. 2. 3. 4. Orientation Introduction Components of a Phase I Trial Phase I Trial Designs A. Rule-Based Designs B. Statistical Designs 5. References

Jud Blatchford, PhD Phase I Trial Designs 2 ORIENTATION Jud Blatchford, PhD Phase I Trial Designs 3 Orientation Features of a Clinical Trial (CT) Study of human beings Prospective Uses an intervention (i.e. changes some aspect of the

subjects) Protects the safety of the subjects Follows an approved protocol Jud Blatchford, PhD Phase I Trial Designs 4 Orientation Phases of Clinical Trials Phase I First time an experimental drug or treatment is tested in humans to examine how well the drug is tolerated Phase II Trials designed to examine if the drug or treatment has a biological treatment effect Phase III Trials designed to assess the treatment effect on a clinically meaningful endpoint Phase IV Post-marketing studies to gain additional information regarding the safety of the drug or treatment

Jud Blatchford, PhD Phase I Trial Designs 5 Orientation Components of Study Design Rationale Establishing a legitimate reason for the study Design Detailed description of what treatments will be administered, how they are administered, including a timeline of administration Subjects Determining the group to be studied and how they will be assigned to treatment groups Data Endpoint(s), obtaining data, and QA Sample Size Justification Ensuring the study will be able to answer the scientific question with adequate power Study Closure Archiving study data, analysis files Jud Blatchford, PhD

Phase I Trial Designs 6 INTRODUCTION Jud Blatchford, PhD Phase I Trial Designs 7 Introduction Phase I Clinical Trials An experimental drug, treatment, chemotherapeutic agent, cytotoxic agent, is studiedhereafter referred to as drug Primary Goal: Safety Investigate whether the new drug or combination of drugs can be administered safely to subjects

Investigate optimal dosing and administration of drug Secondary Goal: Efficacy Offer a treatment option to subjects who have failed other treatment regimens Jud Blatchford, PhD Phase I Trial Designs 8 Introduction Underlying Assumptions The drug kills both cancer cells and other cells The effect is dose-dependent, therefore: 1. The efficacy of the drug increases with the dose 2. The toxicity of the drug increases with the dose Logically, it would be optimal to give the subjects the highest dose of a drug that can be administered without unacceptable toxicity Fundamental Question: What is this dose?

Jud Blatchford, PhD Phase I Trial Designs 9 Maximum Tolerated Dose (MTD) Definition of MTD: The highest dose without observing an unacceptable rate of toxicity Aliases: Recommended Phase 2 Dose (RP2D) Phase 2 Recommended Dose (P2RD) Jud Blatchford, PhD Phase I Trial Designs

10 COMPONENTS OF A PHASE I TRIAL Jud Blatchford, PhD Phase I Trial Designs 11 Components of a Phase I Trial Definition of a Dose-Limiting Toxicity (DLT) Clarify time-frame for experiencing a DLT Dose Levels How many dose levels will be tested? What will the smallest dose be? What will the starting dose be? Subjects

How many subjects will be tested? Will single subjects or cohorts be tested at each dose? What dose-escalation scheme will be employed? Jud Blatchford, PhD Phase I Trial Designs 12 Definition of a DLT DLTs are typically defined using the National Cancer Institutes (NCI) Common Terminology Criteria for Adverse Events (CTCAE). DLTs are often grade 3 nonhematological and grade 4 hematological toxicities, which are definitely, probably, or possibly related

to the drug. Jud Blatchford, PhD CTCAE Grades 0 No AE 1 Mild 2 Moderate 3 Severe 4 Life threatening 5 Death Degrees of Related

Phase I Trial Designs Unrelated Unlikely Possibly Probably Definitely 13 Definition of a DLT The length of observation within which a DLT occurrence is counted should be explicitly stated in the protocol Typical lengths used are the first cycle of

chemotherapy (often 3 weeks) Weight the trade-off between observation time for a DLT and efficiency in enrolling subjects Jud Blatchford, PhD Phase I Trial Designs 14 Choosing the Starting Dose Goals: Dose high enough to have chance of efficacy Dose low enough to avoid a DLT Use data from animal pre-clinical studies Scale dose by body surface area (mg/m2) Studies that arent first-in-human studies may be informed from previous studies using the same drug

Jud Blatchford, PhD Phase I Trial Designs 15 Choosing the Starting Dose Choices Used: First find dose that is lethal in 10% of mice (LD10) Standard starting dose was 10% of this dose (MELD10), if no grade 4+ AEs observed in other species (rats, dogs, etc.) Find the highest dose for which the most sensitive animals investigated had no AEs Starting dose is 1/3 of this level (scaled) Find the minimal dose for which any toxicity is seen (TDL) Starting dose is 1/3 of the TDL Jud Blatchford, PhD Phase I Trial Designs

16 Choosing the Number of Dose Levels Testing more dose levels to accurately estimate the MTD creates a more cumbersome trial, and may require more subjects Common number of levels is 4 to 7 Observed number has ranged from 3 to 14 Jud Blatchford, PhD Phase I Trial Designs 17 Choosing the Dose Levels Desire to progress through possible doses in a quick (e.g. exponential) manner Ethical considerations should guide the dose

escalation scheme used Linear sequence of numbers may be inefficient 20, 40, 60, 80, 100, 120, 140, Famous sequence of increasing numbers: 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987, Jud Blatchford, PhD Phase I Trial Designs 18 The Fibonacci Sequence Term (n) Value (fn) Ratio (fn / fn-1)

1 1 - 2 1 1.000 3 2 2.000 4

3 1.500 5 5 1.667 6 8 1.600 7 13

1.625 8 21 1.615 9 34 1.619 10 55 1.618

11 89 1.618 Jud Blatchford, PhD Phase I Trial Designs 19 The Golden Ratio Jud Blatchford, PhD Phase I Trial Designs 20

The Golden Ratio Jud Blatchford, PhD Phase I Trial Designs 21 Fibonacci Sequence in Nature Jud Blatchford, PhD Phase I Trial Designs 22 Spirals in a Pine Cone Clockwise from Center Jud Blatchford, PhD

Counter-clockwise from Center Phase I Trial Designs 23 Modified Fibonacci Dose Escalation (MFDE) Ratio (fn / f1) Fib. Seq. 1 - 1.00 1

2 2.00 2.00 1 3 1.67 3.33 2 4 1.50

5.00 3 5 1.40 7.00 5 6 1.33 9.33 8

7 1.33 12.44 13 8 1.33 16.59 21 9 1.33

22.12 34 10 1.33 29.50 55 11 1.33 39.33 89

Jud Blatchford, PhD Phase I Trial Designs Comparison Conservative Ratio (fn / fn-1) Similar Term (n) 24 Ethical Considerations Approach the MTD from below (under-estimates MTD) Bracketing the MTD is unbiased and more efficient

Expected efficacy is minimal Historical response rate is 11%; temp. stable rate is 34% 40% expect a cure Subjects suffer significant toxicity Rate of grade 4 toxicity is 14%; death rate is 0.5% What subjects are told is very important Jud Blatchford, PhD Phase I Trial Designs 25 PHASE I TRIAL DESIGNS Jud Blatchford, PhD Phase I Trial Designs

26 Rule-Based Designs 1. 2. 3. 4. 5. Traditional Escalation Rule Variations of the Traditional Escalation Rule Best of 5 Rule Up-and-Down Designs 2-Stage Designs Jud Blatchford, PhD Phase I Trial Designs

27 Traditional Escalation Rule (TER) 3 subjects receive dose di 0 DLTs 1 DLT 2 or 3 DLTs Escalate - 3 subjects receive dose di+1 3 more subjects at dose di Stop escalation De-escalate to di-1

0 DLTs (1/6 with DLT) 13 DLTs ( 2/6 with DLT) Escalate - 3 subjects receive dose di+1 Stop escalation De-escalate to di-1 De-escalate until a level is reached where at least 6 subjects are treated and at most 1 DLT occurs. MTD is the highest dose where at least 6 subjects were treated with at most 1 DLT. Jud Blatchford, PhD Phase I Trial Designs 28

Evaluating the TER Benefits Criticisms Conservative escalation Ease of implementation Many subjects treated at low, ineffective doses At least 2 subjects treated at level above MTD The true MTD is underestimated Rules regarding dose assignment are clear Statistical models not fit after each subject

Design is robust Will arrive at reasonable estimate of MTD Jud Blatchford, PhD Phase I Trial Designs 29 Variations to TER After escalation stops, fill out all lower levels until at least 6 subjects are treated at each level Treat subjects at a dose level between the level where escalation stopped and the next lower level Jud Blatchford, PhD Phase I Trial Designs

30 Best of 5 Rule 3 subjects receive dose di 0 DLTs 1 or 2 DLTs 3 DLTs Escalate to dose di+1 1 more at dose di Stop escalation 1/4 with DLT

2/4 with DLTs 3/4 with DLTs Escalate to dose di+1 1 more at dose di Stop escalation 2/5 with DLTs 3/5 with DLTs Escalate to dose di+1 Stop escalation MTD is the dose prior to the dose on which escalation stopped. Jud Blatchford, PhD

Phase I Trial Designs 31 Up-and-Down Design (UaD) 1 subject receives dose di 0 DLT 1 DLT Escalate to dose di+1 De-escalate to di-1 Perform UaD for a pre-specified number of subjects (j). MTD is the dose that would be assigned to the j+1st subject. Jud Blatchford, PhD

Phase I Trial Designs 32 Storers C Design (UaD-C) 1 subject receives dose di 0 DLT 1 DLT If 2 consecutive subjects with 0 DLT, escalate to dose di+1; else give dose di De-escalate to di-1

Perform UaD for a pre-specified number of subjects (j). MTD is the dose that would be assigned to the j+1st subject. Jud Blatchford, PhD Phase I Trial Designs 33 Stage 1 Storers Two-Stage BC Design (UaD-BC) 1 subject receives dose di 0 DLT 1 DLT Escalate to dose di+1 De-escalate to di-1

Stage 2 1 subject receives dose di-1 0 DLT 1 DLT If 2 consecutive subjects with 0 DLT, escalate to dose di; else give dose di-1 De-escalate to di-2 Perform UaD for a pre-specified number of subjects (j). MTD is the dose that would be assigned to the j+1st subject. Jud Blatchford, PhD

Phase I Trial Designs 34 Accelerated Titration Designs Extension by Simon of Storers work Design 1: TER Designs 24 Stage 1: Single subjects until first DLT or second grade 2 AE Stage 2: TER Design 2: Toxicities observed in first cycle only Design 3: Toxicities may be observed in any cycle Design 4: Same as 3 except escalation factor is 2.0 Jud Blatchford, PhD Phase I Trial Designs 35

Statistical Designs Dose escalation guided by a statistical model of the relationship between dose and toxic response 1. Continual Reassessment Method 2. Modifications to the CRM 3. 2-Stage CRM Designs 4. TITE-CRM Jud Blatchford, PhD Phase I Trial Designs 36 Continual Reassessment Method (CRM) First proposed by OQuigley in 1990 Subjects are enrolled individually A dose-toxicity function is assumed f(d | ) = Pr{DLT | }

After each patient completes observation, the estimate of is updated Strategy is to assign the dose closest to the estimated MTD to each subject Jud Blatchford, PhD Phase I Trial Designs 37 Considerations for the CRM Number of dose levels Initial estimates of toxicity rates at each dose level Target rate of DLT () Dose-toxicity function Escalation restrictions Number of subjects to be treated Jud Blatchford, PhD

Phase I Trial Designs 38 Considerations Number of Dose Levels Initial Estimates of Toxicity Typically between 3 and 8 In general, as the number of dose levels in the trial increases, the number of subjects needed to accurately estimate the MTD will increase The estimates should bound the target rate ()

Jud Blatchford, PhD Phase I Trial Designs The CRM is not robust when doses tested do not induce toxicity 39 Choosing a Dose-Response Function Logistic Function Logistic Regression Let p = Pr{DLT} Solving for p we have: One-parameter model: Jud Blatchford, PhD

Phase I Trial Designs 40 Choosing a Dose-Response Function Hyperbolic Tangent Function Scaled Tanh Function = Pr{DLT} Jud Blatchford, PhD Phase I Trial Designs 41 Choosing a Dose-Response Function

CDF of Normal Distribution P{DLT} Jud Blatchford, PhD Phase I Trial Designs 42 The Method of CRM Dose-toxicity function and are chosen a-priori Function is re-fit (i.e. new estimate of is obtained) after each subjects observed toxicity New function is determined from the a-priori function and the vector of observed toxicities Curve shifts to the right without toxicity; left with toxicity Next subject is treated at the dose level whose Pr{DLT} is closest to

Jud Blatchford, PhD Phase I Trial Designs 43 Distributions of DLT Occurrence By Dose Priors for Subject 1 Priors for Subject 26 High degree of overlap of probabilities between doses Separation between dose levels becoming clearer Jud Blatchford, PhD Phase I Trial Designs

44 Evaluating the CRM Benefits Criticisms Few subjects are treated at low, ineffective doses Subjects are treated at doses believed at the time to be the most efficacious, yet safe Starting dose is too high Dose escalation is too aggressive Trial length is too long Jud Blatchford, PhD

Phase I Trial Designs 45 Modified CRM Start at the lowest dose level under consideration Enroll two or three subjects at each cohort Constrain dose escalation to increase by at most one dose level Jud Blatchford, PhD Phase I Trial Designs 46 Practical CRM Proposed by Piantadosi Based on pre-clinical toxicity data:

Choose dose that would produce low (10%) rate of DLT Choose dose that would produce high (90%) rate of DLT Estimate dose/toxicity curve that fits these 2 points Use the dose/toxicity curve to find dose for Treat three subjects at this level, then re-estimate the dose-toxicity curve, dose for , and tx 3 more Repeat until target dose changes by < 10% Jud Blatchford, PhD Phase I Trial Designs 47 2-Stage CRM Designs Stage 1: TER 2 + 2 is a more common first stage than 3 + 3 Continue until first toxicity is observed Stage 2: CRM

After first toxicity, fit the dose-response curve using the toxicity data accrued thus far Choose dose for next cohort of 2 as dose with estimated rate of DLT closest to Jud Blatchford, PhD Phase I Trial Designs 48 Time-to-Event CRM (TITE-CRM) Builds on the CRM described thus far Uses information from subjects accrued, even if they havent finished observation period Subjects with DLT are given full weight Subjects without DLT are given weight t/T. Allows subjects to be enrolled without waiting for prior cohorts to finish

Benefits studies with delayed toxicity (e.g. radiation studies) Jud Blatchford, PhD Phase I Trial Designs 49 Example of a TITE-CRM Trial Subject accrual is instantaneous The majority of doses administered are near MTD Jud Blatchford, PhD Phase I Trial Designs 50 Additional TITE-CRM Considerations Choice of weight function

Uniform toxicities may use a linear function Expecting late toxicities may use a convex function Expecting early toxicities may use a concave function Setting a Margin (i.e. upper limit) on toxicity If = 0.20 and Margin = 0.05, dose for next subject will be dose closest to 0.20 and not greater than 0.25 Determine cumulative time exposure (B) before allowing escalation (e.g. B = 2) Jud Blatchford, PhD Phase I Trial Designs 51 Design Comparisons Fitting a model to the data will improve the accuracy of the MTD found by rule-based designs Model-guided designs only perform well if

assumptions are met ( in range of doses tested) Conflicting results when designs compared Few comparisons made on level playing field Both rule-based and model-guided designs are in common use, for good reason Jud Blatchford, PhD Phase I Trial Designs 52 Important Future Work Individualized Designs Development of designs allowing for within-subject dose escalation Development of designs for targeted agents Designs for trials expecting minimal toxicity Jud Blatchford, PhD

Phase I Trial Designs 53 Pharmacokinetic Profiles Often of interest in a phase 1 study Studies how the body processes the drug Parameters of interest are typically: Cmax Maximum concentration of drug Tmax Time until the maximum concentration of drug Elimination constant describes rate of loss from body T1/2 Half-life of the drug AUC Area under the concentration curve Jud Blatchford, PhD Phase I Trial Designs 54

Calculation of PK Parameters Cmax & Tmax Directly from table of results = Taken from regression of ln(C) on time T1/2 Create exponential decay equation from above regression Solve for T when the concentration is half of reference amt AUC = AUC0-t + AUCt- AUC0-t Use trapezoidal rule AUCt- = Ct / Jud Blatchford, PhD Phase I Trial Designs 55 Example Calculations From regression of ln(C) on time we get:

ln (C) = 2.25 0.58 T, so = 0.58 (not -0.58) C = exp(2.25) * exp(-0.58)T C = 9.5(-.56)T 4.75 = 9.5(-0.56)T to solve for T1/2 = -0.56T ln(1/2) = T*ln(-0.56) T = 1.19 so T1/2 = 1.19 hours AUC = 12.03 + 0.06/0.58 = 12.18 Jud Blatchford, PhD Phase I Trial Designs

56 PK Design Considerations Sampling Times Important to have accurate estimates of Cmax Cluster several times around expected Tmax Sampling Period Important to have accurate estimate of US FDA requires times to capture at least 3 half-lives after Tmax Bioequivalence Trials Create 90% CIs for ln[(A)] - ln[(B)], = PK parameter Check if all CIs are within 0.80 to 1.25 range Jud Blatchford, PhD Phase I Trial Designs

57 REFERENCES Jud Blatchford, PhD Phase I Trial Designs 58 References 1989. Storer BE. Design and Analysis of Phase I Clinical Trials. Biometrics, 45, 925937. 1990. OQuigley J, Pepe M, and Fisher L. Continual Reassessment Method: A Practical Design for Phase I Clinical Trials in Cancer. Biometrics, 46, 3348. 1993. Korn EL, and Simon R. Using the Tolerable-Dose Diagram in the Design of Phase I Combination Chemotherapy Trials. Journal of Clinical Oncology, 11 (4), 794801. 1993. Mick R, and Ratain MJ. Model-Guided Determination of Maximum Tolerated Dose in Phase I Clinical Trials: Evidence for Increased Precision. Journal of the National Cancer Institute, 85 (3), 217223. 1994. Faries D. Practical Modifications of the Continual Reassessment Method for Phase I Cancer

Clinical Trials. Journal of Biopharmaceutical Statistics, 4 (2), 147164. 1996. Piantadosi S, and Liu G. Improved Designs for Dose Escalation Studies Using Pharmacokinetic Measurements. Statistics in Medicine, 15, 16051618. 1996. Smith TL, Lee JJ, Kantarjian HM, Legha SS, and Raber MN. Design and Results of Phase I Cancer Clinical Trials: Three-Year Experience at M. D. Anderson Cancer Center. Journal of Clinical Oncology, 14 (1), 287295. 1997. Durham SD, Flournoy N, and Rosenberger WF. A Random Walk Rule for Phase I Clinical Trials. Biometrics, 53, 745760. Jud Blatchford, PhD Phase I Trial Designs 59 References (Continued) 1997. Simon R, Freidlin B, Rubinstein L, Arbuck SG, Collins J, and Christian MC. Accelerated Titration Designs for Phase I Clinical Trials in Oncology. Journal of the National Cancer Institute, 89 (15), 11381147. 1998. Friedman LM, Furberg CD, and DeMets DL. Fundamentals of Clinical Trials. Springer. 1998. Whitehead J, and Williamson D. Bayesian Decision Procedures Based on Logistic

Regression Models for Dose-Finding Studies. Journal of Biopharmaceutical Statistics, 8 (3), 445467. 1999. Reiner E, Paoletti X, and OQuigley J. Operating Characteristics of the Standard Phase I Clinical Trial Design. Computational Statistics and Data Analysis, 30, 303315. 2000. Cheung YK, and Chappell R. Sequential Designs for Phase I Clinical Trials with Late-Onset Toxicities. Biometrics, 56, 11771182. 2000. Eisenhauer EA, ODwyer PJ, Christian M, and Humphrey JS. Phase I Clinical Trial Design in Cancer Drug Development. Journal of Clinical Oncology, 18 (3), 684692. 2000. Wang O, and Faries DE. A Two-Stage Dose Selection Strategy in Phase I Trials with Wide Dose Ranges. Journal of Biopharmaceutical Statistics, 10 (3), 319333. 2001. Lin Y, and Shih WJ. Statistical Properties of the Traditional Algorithm-Based Designs for Phase I Cancer Clinical Trials. Biostatistics, 2 (2), 203215. Jud Blatchford, PhD Phase I Trial Designs 60 References (Continued) 2001. Ishizuka N, and Ohashi Y. The Continual Reassessment Method and Its Applications: A

Bayesian Methodology for Phase I Cancer Clinical Trials. Statistics in Medicine, 20, 26612681. 2002. Potter PM. Adaptive Dose Finding for Phase I Clinical Trials of Drugs Used for Chemotherapy of Cancer. Statistics in Medicine, 21, 18051823. 2003. Agrawal M, and Emanuel EJ. Ethics of Phase I Oncology Studies: Reexamining the Arguments and Data. Journal of the American Medical Association, 290 (8), 1075 1082. 2003. Ivanova A, Montazer-Haghighi A, Mohanty SG, and Durham SD. Improved Up-and-Down Designs for Phase I Trials. Statistics in Medicine, 22, 6982. 2004. Stylianou M, and Follmann DA. The Accelerated Biased Coin Up-and-Down Design in Phase I Trials. Journal of Biopharmaceutical Statistics, 14 (1), 249260. 2005. Horstmann E, McCabe MS, Grochow L, Yamamoto S, Rubinstein L, Budd T, Shoemaker D, Emanuel EJ, and Grady C. Risks and Benefits of Phase I Oncology Trials, 1991 Through 1992. New England Journal of Medicine, 352, 895904. 2006. Crowley J, and Ankerst DP. Handbook of Statistics in Clinical Oncology. Chapman and Hall/ CRC. Jud Blatchford, PhD Phase I Trial Designs

61 References (Continued) 2006. Potter DM. Phase I Studies of Chemotherapeutic Agents in Cancer Patients: A Review of the Designs. Journal of Biopharmaceutical Statistics, 16, 579604. DOI: 10.1080/10543400600860295. Jud Blatchford, PhD Phase I Trial Designs 62