Tools to Reduce Phase III Tools to Reduce Phase III Trial FailuresTrial Failures

Lawrence J. Lesko, Ph.D., FCPDirector of the Office of Clinical Pharmacologyand BiopharmaceuticsCenter for Drug Evaluation and ResearchFood and Drug AdministrationSilver Spring, Maryland

Session VII: Innovation or Stagnation:The Critical Path InitiativeAGAH Annual Meeting 2006February 21, 2006Dusseldorf, Germany

OverviewOverview

The productivity problem to be solved by critical path initiative

Critical path opportunities that can influence early drug development and regulatory decisions

General Agreement on the General Agreement on the Problem to Fix: Rising CostsProblem to Fix: Rising Costs

0

20

40

60

80

100

Bil

lio

ns

of

Do

llar

s

1996 '97 '98 '99 2000 '01 '02 '03 '04

US Funding for Medical Research

Total

Pharm Ind

Data from JAMA, Sept 21, 2005; NIH, and PhRMA Annual Surveys

But New Drug Applications Are But New Drug Applications Are Not Rising at the Same RateNot Rising at the Same Rate

0

10

20

30

40

50

60

1996 '97 '98 '99 2000 '01 '02 '03 '04 '05

Total Number of NDAs Filed with CDER

Data from FDA; beginning in 2004, numbers include BLAs transferred from CBER to CDER

Barrier to Improving Productivity Barrier to Improving Productivity is the High Attrition Rateis the High Attrition Rate

0102030405060708090

100

Su

cces

s R

ate

(%)

Phase I Phase II Phase III NDA Approval

Stage of Development

Kola and Landis, Nature Review Drug Discovery, 2004 (3):711-715

“We are an industry with a 98% failure rate…..The only thing we have to do to double our success rate is to drop our failure rate by 2%”

Hank McKinnell, Pfizer CEO, at http://www.bio-itworld.com, 2/14/06

Driver for Industry to Seriously Driver for Industry to Seriously Commit to Critical Path ConceptsCommit to Critical Path Concepts

Why Drugs Fail in Development: Why Drugs Fail in Development: Root Cause Analysis is NeededRoot Cause Analysis is Needed

0%

5%

10%

15%

20%

25%

30%

35%

40%

Efficacy Safety Toxicology Commercial Costs Formulation BA/PK Other

1991

2001

Kola and Landis, Nature Review Drug Discovery, 2004 (3):711-715

Shift Failures Earlier: Shift Failures Earlier: Quick Win – Quick Kill ParadigmQuick Win – Quick Kill Paradigm

Predicting phase 3 clinical outcomes from phase 2 study results is no better than a

coin flip

“50% of phase 3 studies fail in 2005 as compared to 35% in 1997”*

* From PhRMA at http://www.pharma.org

Phase 3 Trials Have Become Phase 3 Trials Have Become Larger and More Costly Larger and More Costly

11%19%

70%

0%10%20%30%40%50%60%70%

Ave

Dir

ect

Co

sts

(%)

Phase I Phase II Phase III

Stage of Development

Distribution of Total Costs of Clinical Trials

Dimasi et al, J Health Economics, 2003 (22): 151-185

Paradox of Decreased Productivity: Paradox of Decreased Productivity: Sustained Profitability (Inertia to Change)Sustained Profitability (Inertia to Change)

From Federal Government API Calculations and Price Waterhouse-Coopers LLP, Reported February 8, 2006

Earnings of Major Industries For 2000-2005

5.8

7.6

7.7

10.8

16.2

17.3

Oil and Gas

Software Services

Health Care

Real Estate

Pharmaceuticals

Banks

Cents / Dollars of Sales

Pillars of Industry Profitability: Pillars of Industry Profitability: Changing FundamentalsChanging Fundamentals

Product Life Cycles

Flexibility Pricing

Blockbuster Market

Patent Expirations

R&D Productivity

Shrinking

Fixed Pricing

Segmented Market

Increasingly Important

Absolutely Essential

Adapted in Part From a Presentation by Dr. Eiry W. Roberts, Lilly

The FDA Critical Path Initiative: An The FDA Critical Path Initiative: An Opportunity to ChangeOpportunity to Change

http://www.fda.gov/oc/initiatives/criticalpath/whitepaper.html

Goals

1. To develop new predictive “tools” and bring innovation into the drug development process

2. To improve the productivity and success of drug development

3. To speed approval of innovative products to improve public health

Progress Is Steady But Slow: Progress Is Steady But Slow: Widespread Recognition of BarriersWidespread Recognition of Barriers

FDA role is largely to act as an enabler, convener or stimulator of critical path

Agency does not have staff exclusively dedicated to critical path initiatives

Research must be spearheaded by outside non-profit consortium (few academic rewards)

2006 budget is supposed to have $10 million dollars allotted to critical path

Drug companies must be persuaded to share their data and pool information (concerns about IP)

FDA has been distracted with safety issues

Need for New Organizational Paradigms: Need for New Organizational Paradigms: Formation of New FDA “Super Office”Formation of New FDA “Super Office”

Office of Clinical

Pharmacology

Office of Biostatistics

Virtual Officeof

Critical Path Initiatives

Office of the Commissioner

Office of New Drugs

To be completed by June 2006

Other Changes in FDA Infrastructure Other Changes in FDA Infrastructure to Achieve Critical Path Goalsto Achieve Critical Path Goals

CDER-wide centralized consulting groups– Pharmacometrics (applying quantitative methods)– QT protocols, analysis of thorough QT studies – Pharmacogenomics, diagnostics and VGDS– Pediatric written requests, data analysis, and exclusivity

New interface opportunities with industry– End-of-phase 2A meetings

New information management system using CDISC standards and data warehousing

Fellowship and sabbatical opportunities “Soft skill” training in negotiation and

communication

One of the First Products of Critical One of the First Products of Critical Path: Exploratory IND GuidancePath: Exploratory IND Guidance

Exploratory IND precedes traditional IND to reduce time/resources on molecule unlikely to succeed (“quick kill” concept)– Conduct early in phase 1– Very limited human exposure (e.g., < 7 days)– No therapeutic intent– Preclinical toxicology and CMC requirements

scaled to type of study (e.g., microdosing)– Flexible clinical stop doses

January 6, 2006; http://www.fda.gov/cder/Guidance/7086fnl.htm

Focus on Clinical Pharmacology Focus on Clinical Pharmacology Efforts in Critical Path InitiativeEfforts in Critical Path Initiative

Areas of greatest potential gain– Improve predictions of efficacy and safety in

early drug development Biomarkers ~ better evaluation tools

– General biomarker qualification, qualifying disease specific biomarkers

M&S ~ better harnessing of bioinformatics– Disease state models, clinical trial simulation

Clinical trials ~ improving efficiency– Enrichment designs, adaptive trial designs

Biomarkers: Classic Thinking Biomarkers: Classic Thinking Inhibits Their DevelopmentInhibits Their Development

Overemphasis on surrogate endpoints as an objective confounds biomarker development– Uncertainty over what is needed for “validation” and difficulty

in getting “validation” data frustrates progress– Need to reassess the idea of “validation” perhaps to

“qualification”

Regulatory agencies have focused to much on empirical testing of treatment vs placebo– Skewed research away from mechanistic biomarkers that

would provide a better understanding of clinical evaluation– Provide incentives to use biomarkers throughout preclinical

and clinical development

“Pharmacometrics Can Guide Future Trials, Minimize Risk -- FDA Analysis”

• 244 ~ number of NDAs surveyed in cardio-renal, oncology and neuropharmacology

• 42 ~ NDAs with pharmacometric (PM) analysis**

• 26 ~ PM pivotal or supportive of NDA approval

• 32 ~ PM provided evidence for label language

One Incentive: Show How Biomarkers One Incentive: Show How Biomarkers Benefit in Regulatory Decision-MakingBenefit in Regulatory Decision-Making

October 3, 2005, Volume 67, Number 40, Page 15

** Number not higher because sponsor application lacked necessary data

Re-emphasize 5 Fundamental Principles to Re-emphasize 5 Fundamental Principles to Greatly Improve Biomarker PredictionsGreatly Improve Biomarker Predictions

Develop reliable standards for the technology and analyte being measured

Clearly state the intended use of the biomarker, i.e., what is the question?

Define the necessary performance expectations and assumptions to make a binary decision

Express biomarker predictions in terms of probabilities of seeing clinical outcome of interest, i.e., inform decisions

Evaluate the cost and benefit of biomarker development vs alternative approaches, i.e., when does it really make a difference

Example: Can EGFR Expression Distinguish Example: Can EGFR Expression Distinguish Between Aggressive and Non-Aggressive Between Aggressive and Non-Aggressive Pancreatic Tumors?Pancreatic Tumors?

What is the definition of overexpression and how is this related to the technology platform used (quality)?

What is the definition of aggressive? Locally advanced or metastatic? Survival of 3 months or 6 months?

What kind of performance attributes are required? Is a PPV ~ 90% to distinguish between aggressive and non-aggressive acceptable? How about 75%?

Is it necessary to predict aggressiveness for patients that received combination therapy with gemcitabine or not?

What endpoint will I use to link clinical outcome to EGFR overexpression? Tumor size? Progression-free survival?

FDA-NCI Collaboration: Develop Such a FDA-NCI Collaboration: Develop Such a Grid for Biomarkers Used in Cancer Drug Grid for Biomarkers Used in Cancer Drug DevelopmentDevelopment

Defined most important primary and secondary oncology biomarkers and how they are used

Primary list– 4 kinases (VGEF, EGFR, PISK/Akt and Src) – 1 cell surface antigen (CD20)

Secondary list– 3 kinases (JaK, ILK, cell cycle checkpoints)– 2 cell surface antigens (CD30 and CTLA-4)

Developing detailed performance specifications and plan conduct “gap” research– Couple with complimentary biomarkers, e.g., imaging to

improve predictability of outcomes

Define Regulatory Framework for Technical Define Regulatory Framework for Technical Qualification of Biomarkers as SurrogatesQualification of Biomarkers as Surrogates

Develop inventory of biomarkers used as surrogate endpoints for full approval, accelerated approval, supplements and for support of one-clinical-study approvals in each of CDER review divisions

1. What surrogate endpoint is being used and what is the required effect size, if there is any?

2. Which category of approval was it used for?

3. When was it first used, what was the exact claim that was granted, and what did the label say?

4. What was the evidence basis for reliance on a surrogate?

5. What other surrogate endpoints are under consideration?

Model-Based Drug Development: An Model-Based Drug Development: An Extension of Dose-ResponseExtension of Dose-Response

A mathematical, model-based approach to integrating information and improving the quality of decision making in drug development– Preclinical and clinical biomarkers– Dose-response and/or PK-PD relationships– Mechanistic or empirical disease models– Clinical trial simulations and probabilities of success– Baseline-, placebo- and dropout-modified models

Ten disease models created internally including HIV-AIDS, osteoarthritis, alzheimers, parkinsons and pain– Exploring feasibility of creating a public space where

models can be shared and grown

Build a Drug Disease Model: Build a Drug Disease Model: Example of HIV/AIDSExample of HIV/AIDS

Mechanistic Model of Disease

Ex: HIV/AIDS

Mathematical Model of Dose – Conc. (PK)

Ex: HIV, viral load vs. time

Biomarkers of EfficacyEx: viral RNA over time

Biomarkers of SafetyEx: GIT events over time

D/R and/or PK/PDEx: viral RNA and GIT

events as f ( E, t)

Biomarkers (clinical outcome) Over Time

Patient Co-Variates

Placebo Response

Example: New CCR5 InhibitorExample: New CCR5 Inhibitor

D/R for efficacy from 0.5 to 6 mg BID– Co-administered with Kaletra 400 mg/100 mg

Risk– Severe GI events increased at higher doses

Benefit– Patient co-variates, resistance, drop-outs, non-

compliance Question to be asked

– How can optimal dosing and study design be determined after 4 weeks in order to predict phase 2B trial outcome at 48 weeks?

Built Dynamic Viral Disease Model Using Built Dynamic Viral Disease Model Using Literature, In-House Data, Information Literature, In-House Data, Information Provided Voluntarily by CompaniesProvided Voluntarily by Companies

J Acquir Immun Defic Syndr 26:397, 2001

CD4+ Cells Virus

LatentInfected

Active Infected

+

d2

d3

d1 c

a

p

fAVT

fLVT

PI

(N)NRTI

(N)NRTI

production rate of target celld1: dying rate of target cellc: dying rate of virus: infection rate constantd2: dying rate of active cellsd3: dying rate of latent cells p: production rate of virus

Differentiated Dosing and Study Designs Differentiated Dosing and Study Designs by Simulating Viral Load Over Timeby Simulating Viral Load Over Time

2 mg QD

4 mg QD

2 mg BID

6 mg BID

Time in day

HIV

RN

A c

hang

e fr

om B

L lo

g (c

opie

s/m

L)

0 5 10 15 20

-1.5

-1.0

-0.5

0.0

0.5

Simulating 20 Clinical Trials with 50 Simulating 20 Clinical Trials with 50 Patients per Group to Estimate Probability Patients per Group to Estimate Probability of “Picking the “Winner”*of “Picking the “Winner”*

* 2 log drop in viral RNA

% of Simulated Trials Achieving Target Efficacy Outcome

20%

59%

21%

1 mg BID

2 mg BID

4 mg OD

Tipranavir: Good Biomarker Work Informs Tipranavir: Good Biomarker Work Informs Drug Development and TherapeuticsDrug Development and Therapeutics

Non-peptidic protease inhibitor for experienced patients or patients with virus resistance to other PIs

Plasma TPV levels ~ major driver of efficacy and toxicity, boosted with ritonavir (RTV)

HIV-1 protease mutations ~ major driver of resistance and decreased efficacy

500/200 TPV/RTV dose selected for phase III– Plasma TPV levels > IC50 to suppress viral load and avoid

development of resistance Inhibitory quotient, IQ, predicts responders after 24

weeks– IQ = Cmin / [Wild Type IC50 x 3.75]

See The Pink Sheet, June 30, 2005

Impact of IQ on 24-Week Viral Load Impact of IQ on 24-Week Viral Load Response and CResponse and Cminmin on Liver Toxicity on Liver Toxicity

0 200 400 600 800 1000

Inhibitory Quotient

0%20

%40

%60

%80

%10

0%

Per

cen

t of R

espo

nde

rs a

t We

ek 2

4

phase 3 without T20 (n=200)phase 3 with T20 (n=91)phase 2 (n=160)

10 20 30 40 50Cmin in ug/mL

0%20

%40

%60

%80

%10

0%

Per

cent

of P

atie

nts

with

Gra

de 3

/4 A

LT T

oxic

ity

Benefit: Viral Load Change From Baseline (log10)

Risk: Grade 3-4 ALT, AST or Bilirubin

From Dr. Jenny Zheng (OCPB), FDA Antiviral Drug AC Meeting, May 19, 2005

Translation of Information to Translation of Information to Approved LabelApproved Label

“Among the 206 patients receiving APTIVUS-ritonavir without enfuvirtide…..the response rate was 23% in those with an IQ value < 75 and 55% with an IQ value > 75.”

“Among the 95 patients receiving APTIVUS-ritonavir with enfuvirtide, the response rate in patients with an IQ < 75 vs. those with IQ > 75 was 43% and 84% respectively.”

Pharmacogenomics

Critical Path Opportunity for Critical Path Opportunity for Innovative Adaptive Trial DesignInnovative Adaptive Trial Design

Focus on Phase II/III Randomized Focus on Phase II/III Randomized Controlled Trials of Targeted MedicinesControlled Trials of Targeted Medicines

Several innovative clinical trial designs and statistical methodogies that increase efficiency ~ focus on “right patients”– adaptive

Predictive assay to identify binary outcomes (e.g., response) not available before trial

– enrichment Predictive assay to identify binary outcomes (e.g.,

response) known before trial (a priori)

– stratification Predictive assay to identify a range of outcomes (e.g.,

response) known before trial

Improving Efficiency: Prospective Evaluation of a Improving Efficiency: Prospective Evaluation of a Predictive Biomarker in a Phase 3 RCT Without Predictive Biomarker in a Phase 3 RCT Without Compromising Evaluation of Overall EffectCompromising Evaluation of Overall Effect

All patients (1000)Treatment vs Control

Treatment armStage 1: All-Comers (250)

10% response rate

Treatment armStage 2: Subset (250)

Develop marker in sensitive patients(40% marker +)

Sensitive subsetMarker +

25 % response rate

Nonsensitive subsetMarker -

5 % response rate

Control arm (500)5 % response rate

Prospectively apply testUnrestricted entry

• Compare T vs C using data from all patients from Stage 1 at alpha = 0.04

• Compare T vs C using data from sensitive subset from Stage 2 at alpha = 0.01

• “Win” if either of two tests is positive

• 85% chance of finding overall effect or effect in sensitive subset

Freidlin and Simon, Clin Can Res 2005, 11:7872-7878

Confirmatory Adaptive Design: Confirmatory Adaptive Design: FeaturesFeatures

Prospectively define N in first and second stage

Preserve ability to detect overall effect as well as effect in sensitive subset if overall effect is negative

As efficient as traditional designs to detect overall benefit to all patients

Reduce chance of rejecting an effective medicine if only effective in sensitive subset

More stringent significance level at stage 1 (0.04 vs 0.05)

Context for use is looking at anticancer drugs but applicability to other areas may be limited

Examine timeframe for identifying test at Stage 1 (e.g. vs earlier biomarkers)

Disease pathophysiology less established than tumor behavior

Summary: Integrating Use of Tools Summary: Integrating Use of Tools Along the Critical PathAlong the Critical Path

Continual Reduction in Uncertainty in Benefit/Risk

Toolkit for Improving Success in Drug Development

Biomarkers: Prognostic, PD and Predictive Patient Selection Criteria

Targeted Label Information Optimal Use

Drug and Disease Modeling Dose Response, PK-PD and Dosing

Adaptive Trial Design

Thanks for your attention

lawrence.lesko@cder.hhs.gov