1. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Review Experience in Evaluating Predictive Biomarkers – Design and Analysis Considerations
Yuan-Li Shen, Dr. P.H.
U.S. Food and Drug Administration
Center for Drug Evaluation and Research
Office of Biostatistics
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

2. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Acknowledgement
Somesh Chattopadhyay
Kun He
Rajeshwari Sridhara
Qiang (Casey) Xu
Sue Jane Wang
The views expressed in this talk are those of the author, not of the FDA
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

3. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Outline
Motivation : HER2, EGFR
Types of designs
Prospective (Adaptive, enrichment, all-comer)
Prospective-Retrospective
Phase II – biomarker subgroup selection
Issues in designs and analysis
- Example of HER-2 for MBC (metastatic breast cancer)
- Example of EGFR for NSCLC (non-small cell lung cancer)
- Example of K-ras for mCRC (metastatic colorectal cancer)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

4. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Why targeted in Cancer trials
Limit drug exposure to those who benefit
Avoid drug use in those who will be harmed
Optimize drug dosing
• Improve efficiency of trial design
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

5. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Terminology
Targeted : selected; enrichment
Untargeted: unselected; all-comers
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

6. Spectrum in study designs incorporating biomarkers
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Spectrum in study designs incorporating biomarkers
Targeted
e.g. Hormonal therapy in ER/PR positive breast cancer women;
e.g. Herceptin for HER2+ metastatic breast cancer; Note: HER2–amplified
in 25–30% of all breast cancer patients (Cobleigh et al, 1999)
e.g.Gleevec for KIT+ Gastrointestinal Stromal Tumor (GIST);
e.g. products underdevelopment: BRAF inhibition in melanoma; ALK
inhibition in NSCLC
e.g. K-Ras WT for metastatic colorectal cancer (mCRC)
e.g. EGFR for non-small cell lung cancer (NSCLC) (???)
e.g. EGFR in squamous cell head and neck cancer.
Note: EGFR is overexpressed in 80% of squamous cell carcinoma
(Kumar, et al,2003)
Untargeted
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

7. Design – Issues to be considered
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Design – Issues to be considered
Strength of the biological evidence
(e.g. relationship between biomarker and treatment; treatment effectiveness)
Prevalence of the biomarker
Validity and reliability of the testing tool
Feasibility of the assessment
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

8. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Type of Trial Designs
Prospective design
--- Enrichment (e.g. Herceptin in HER2 positive for mBC)
--- Biomarker stratified (with subgroup hypothesis)
--- Planned Adaptation
Drop a biomarker subgroup, based on a futility analysis
Adaptive signature design
Prospective-retrospective
(e.g. anti-EGFR mAb in K-ras WT for patients with mCRC)
Phase II : Exploratory biomarker subgroup selection (e.g. I-SPY 2)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

9. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective design –
Enrichment
Randomize
New Rx
Biomarker +
Control
Screen
Biomarker status
Biomarker -
Off study
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

10. Issues in Enrichment Design
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Issues in Enrichment Design
Requires smaller sample size (relative to all comers design), however…..
Can not answer if the biomarker is predictive
Testing assay may not correctly identify the subgroup
(misclassification)
Can not be used for clinical validation of the diagnostic tool
Slow accrual if only a small proportion of the biomarker positive subgroup is in the population (low prevalence)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

11. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Example of limitation in assay precision from an enrichment design (Herceptin)
DFS results in Herceptin for adjuvant breast cancer trial by HER2 overexpression or amplification
Reference: Herceptin package insert
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

12. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective design –
Biomarker Stratified
New Rx
Biomarker +
Control
Screen
Biomarker status
New Rx
Biomarker -
Control
Randomize
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

13. Analysis Issues in the Biomarker Stratified Design
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Analysis Issues in the Biomarker Stratified Design
Potential Analysis Plan:
Treatment by biomarker interaction
Issues: May not have enough power for interaction and failure to show interaction does not mean no differential subgroup effect
Hierarchical testing:
Issues: more complicated when interim analyses for efficacy and futility will be performed and more endpoints (e.g. PFS, OS) will be involved.
α-splitting:
Issues: Overall treatment effect may be driven by one of the Biomarker subgroup; when to perform the event driven analysis
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

14. Adaptation for selecting biomarker subgroup
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Adaptation for selecting biomarker subgroup
Planned adaptation:
-- Futility analysis
-- Combining Phase II and III data
-- Adaptive signature design :
Stage 1 data will be used to define a biomarker “sensitive” subgroup and the rule will be applied to stage 2 to select a biomarker sensitive subgroup; Analysis will be performed to evaluate treatment effect in overall population and biomarker subgroup selected at stage 2
Unplanned adaptation based on independent information – include subgroup hypothesis
e.g. HER2+ in Lapatinib+Letrozole trial (studied in all comers;
change analysis population to HER2+; not stratified by HER2 status);
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

15. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective - Retrospective
• Working Definition : In completed or post-interim-analysis trial where genomic samples were collected prior to treatment initiation, whether or not full ascertainment, the genomic hypothesis is ‘prospectively specified’ prior to diagnostic assay testing. However, the clinical outcome data without genomic information have already been (partially) collected, unblinded, and analyzed. The genomic data analysis might be arguably ‘prospectively performed, which is a retrospective analysis.
* Wang et al (2006 TPJ) ; Wang, et. al. (2010, Clinical Trials)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

16. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective-retrospective –
Criteria for Consideration
• A pragmatic approach may be considered if
Adequate, well-conducted and well-controlled trial
Large sample size if no stratification by biomarker status for randomization (??)
Biomarker ascertainment in a large portion of randomized subjects
Assay – acceptable analytical performance
Acceptable statistical analysis plan
O’Neill, Dec ODAC
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

17. Retrospective Analysis Issues
Retrospective analysis is for hypothesis generating
and should not be used to salvage a failed trial
Issues:  type I error, if the primary endpoint from the ITT population is not significant; possible selection bias; imbalance between treatment arms.
Convenience sample – Issues
Studies with smaller the sample size, more likely to have imbalance of baseline characteristics
Imbalance in biomarker distribution and baseline characteristics leads to biased treatment effect
Results confounded – depending on ratio of biomarker + vs. –
and distribution of the baseline characteristics
To confirm the results requires replication

18. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Trial Designs-
Phase II dose/biomarker screening
I-SPY (2) : (Investigation of Serial Studies to Predict Your therapeutic response with imaging and molecular analysis 2) – a process targeting the rapid, focused clinical development of paired oncological therapies and biomarkers.
Framework : Adaptive phase II clinical trial design in the
neoadjuvant setting for women with high risk, locally advanced
breast cancer
Rationale:
-- identify a biomarker subgroup which is more likely to respond to a
chemotherapy
-- test, analytically validate and qualify biomarkers
-- employ an adaptive trial design to enable efficient learning
-- utilize organizational management principles and bioinformatics to
eliminate current inefficiencies in clinical trial
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

19. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
I-SPY 2
Biomarkers: ER (+,-), PR(+,-), HER2 (IHC/FISH, gene expression, protein microarray), MammaPrint score (higher MP2, other MP1)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

20. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
I-SPY 2 :
Standard therapy: weekly paclitaxel(+trastuzumab for HER2 positive)
doxorubicin+cyclophosphamide (AC)
Drugs: 5 new chemotherapies are tested simultaneously
(each will be tested in a min. of 20 and a max. of 120 patients)
Biomarkers: 14 out of 256 possible biomarker combinations are of interest
Endpoint: pathologic complete response
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

21. I-SPY 2 : Issues More for hypothesis generating
Difficulties in interpreting the results
– comparative analysis not interpretable
Not clear about the robustness of Baysian adaptation
Products with toxicity, clinical trial hold issues, differential toxicity and benefit risk can not be identified

22. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective - Retrospective
K-ras in EGFR inhibitor for mCRC
Background
The efficacy of EGFR inhibitors is restricted to patients with wild-type
(WT) K-ras status (Amado, JCO 2008, Van Cutsem ASCO 2008 and
others)
June 2008, NIH/NCI issued an action letter to exclude enrollment to
mutant K-ras subgroup
Dec. 16, 2008 Oncologic Advisory Committee was held to discuss
the considerations based on retrospective analyses for a biomarker
subgroup
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

23. Retrospective analysis results (Dec. 2008 ODAC)
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Retrospective analysis results (Dec ODAC)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

24. Retrospective analysis results (Dec. 2008 ODAC)
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Retrospective analysis results (Dec ODAC)
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

25. Issues in retrospective analysis (Dec. 2008 ODAC)
If no statistical significance in the primary efficacy endpoint, all other analyses are exploratory (i.e. can not be used to salvage a failed trial) :
e.g. EPIC, OPUS and PACCE
Convenience sample, e.g. EPIC
Interim results may not be robust and reproducible
No documented evidence that the treatment groups with ascertained biomarker status are comparable in baseline characteristics

26. Prospective Designed, with subgroup hypothesis Study BO18192 (SATURN)
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Prospective Designed, with subgroup hypothesis Study BO18192 (SATURN)
Multicenter, International, Randomized, Double-blind, Placebo-controlled, Phase III – maintenance therapy after platinum-based doublet chemotherapy in patients with advanced or recurrent or metastatic NSCLC
Patients who had CR, PR or SD were randomized to receive Tarceva or placebo using adaptive randomization method (Pocock and Simon, 1975) based on the following stratification factors :
EGFR protein expression, stage of disease, ECOG performance status, chemotherapy regimen, smoking status and region.
Tarceva (Erlotinib) : EGFR tyrosine kinase inhibitor (TKI)
Co-primary hypothesis : PFS in all randomized patients
PFS in EGFR IHC positive subset
Alpha – allocation : 0.03 for PFS in all randomized patients;
0.02 for PFS in EGFR IHC positive subset
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

27. Study BO18192 (SATURN) Design (continue…)

28. Study BO18192 (SATURN): Biomarker Characteristics
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Study BO18192 (SATURN): Biomarker Characteristics
Note: Protocol implemented a hierarchy of biomarker ascertainment
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

29. Study BO18192 (SATURN): Issues of the EGFR IHC status
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Study BO18192 (SATURN): Issues of the EGFR IHC status
~16% unevaluable EGFR IHC status
EGFR (IHC) test was not reliable
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

30. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
SATURN : Statistical Issues in biomarker subgroup analysis
More than 16% had missing or indeterminate EGFR IHC status which
may create problems in interpreting the biomarker subgroup results;
(2) The test used for EGFR IHC status is not approved by FDA for NSCLC.
(3) Whether EGFR IHC testing is more likely to accurately assess EGFR
status for NSCLC is uncertain –
Other EGFR status (mutation, FISH) had close to 50% patients with
missing or indeterminate status, so the analyses based on the other
biomarker testing are considered only exploratory.
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

31. Prospective designed, unplanned adaptation Study EGF 3008:
Randomized, double-blind, placebo-controlled, parallel-group, multi-center, study to evaluate the efficacy and tolerability of the combined treatment of lapatinib and letrozole versus the treatment of placebo and letrozole in 1286 post-menopausal women with hormone receptor positive (ER positive and/or PgR positive) advanced or metastatic breast cancer and had not received prior therapy for advanced or metastatic disease.
Stratified by site of disease and prior adjuvant endocrine therapy
Adaptation -- Sample size increased from 760 to 1280
-- Change primary population (from ITT to HER2+)
(after 760 patients were randomized)
-- Hierarchical testing
10 : PFS in HER2+
20 : PFS in ITT

32. Study EGF 3008 design 10 objective : PFS in HER2+
20 objective : PFS in ITT

33. Study EGF 3008: Patient populations
Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Study EGF 3008: Patient populations
HER2 : FISH + or IHC : 3+
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

34. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Study EGF 3008 : Kaplan-Meier Estimate of Investigator-Evaluated PFS
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

35. Review Experience in Evaluating Predictive Biomarkers
April 15, 2017
Study EGF 3008 : Statistical Issues
(1) The primary efficacy population was changed from ITT to HER2-positive after 760 patients were randomized, and sample size was increased to 1286 to have adequate HER2-positive patients;
(2) Study was not stratified at randomization by HER2 status;
(3) Demonstration of OS benefit has generally been the standard for consideration of regular approval in first-line setting of breast cancer.
Currently, the OS benefit of Lapatinib has not been demonstrated.
Note: Lapatinib was approved under accelerated approval.
Yuan-Li Shen, Dr. P.H. FDA/CDER/OTS/OB/DBV

36. Summary The prospective designed trial is preferred.
-- if there is no clear biological evidence of the
biomarker subgroup and diagnostic tool can
reliably identify the biomarker subgroup,
biomarker stratified design may be preferred.
If the retrospective analysis would be performed, high biomarker ascertainment rate should be required. It can not be used to salvage a failed study. Robustness of the results should be demonstrated and an additional study that can replicate the results is required as the supporting information.

37. References
M. Cobleigh, CL Vogel, D. Tripathy, et al.; Multination study of the efficacy and safety of humanized anti-HER2 monoclonal antibody in women who have HER2-overexpressing metastatic breast cancer that has women who have HER2-overexpressing metastatic breast cancer that has progressed after chemotherapy for metastatic disease. J. Clin Oncol. 1999;17(9):
AD Barker, CC Sigman, et. al.; I-SPY 2: An Adaptive Breast Cancer Trial Design
in the Setting of Neoadjuvant Chemotherapy; Clinical pharmacology & Therapeutics, vol. 86, No. 1, July 2009
B. Johnson and P. Ja¨nne, Selecting Patients for Epidermal Growth Factor Receptor Inhibitor Treatment: A FISH Story or a Tale of Mutations? J. of Clin. Oncology, Vol. 23, No. 28, 2005,
V. Kumar, RS Cotran, SL Robbins, et. al, editors: Robbins Basic Pathology. Seventh edition. Saunders,
Philadelphia; 2003.
B. Freidlin, L. M. McShane, E. L. Korn; Randomized Clinical Trials With Biomarkers: Design Issues; J Natl Cancer Inst 2010;102:152–160
B. Freidlin, R. Simon; Adaptive Signature Design: An Adaptive Clinical Trial Design for Generating and Prospectively Testing A Gene Expression Signature for Sensitive Patients; Clin Cancer Res, 2005:11(21), 2005,
S. J. Mandrekar and D. J. Sargent; Predictive biomarker validation in practice: lessons from real trials; Clinical Trials 2010; 7: 567–573
V. Cutsem, I. Lang, et. al., KRAS status and efficacy in the first-line treatment of patients with metastatic colorectal cancer (mCRC) treated with FOLFIRI with or without cetuximab: The CRYSTAL experience., J. Clin Oncol 26, No. 15S (May 20 Supplement), 2008:2
R. Simon, S. Wang, Use of genomic signatures in therapeutics development in oncology and other diseases, The Pharmacogenomics Journal 2006; 6, 166–173
R. Simon, The use of Genomics in Clinical Trial Design, Clin. Cancer Res., 2008; 14(19), 2008,
R. Simon, S. Paik, D. Hayes, Use of Archived Specimens in Evaluation of Prognostic and Predictive Biomarkers, Commentaries, J. of National Cancer Ins., V. 101, Issue 21, 2009,

38. References
12. S. Wang, R. O’Neill and H. M. J.Hung, Approaches to evaluation of treatment effect in randomized clinical trials with
genomic subsets, Pharmaceut. Statist. 2007; 6: 227–244
13. S. Wang, R. O’Neill, J. Hung, Statistical considerations in evaluating pharmacogenomics-based clinical effect for confirmatory trials, Clinical Trials 2010; 7: 525–536
14. K-Ras Briefing document from Oncologic Drugs Advisory Committee, 12/16/08:
15. Tarceva Briefing document from Oncologic Drugs Advisory Committee, 12/16/09:
16. Gleevec labeling:
Herceptin labeling:
Lapatinib labeling:
19. G. Amado, M. Wolf, et.al., Wild-Type KRAS Is Required for Panitumumab Efficacy in patients With Metastatic Colorectal Cancer, J. of Clin. Oncology, Vol. 26, No. 10, 2008,
20. A. S. Crystal, and A. T. Shaw, New Targets in Advanced NSCLC: EML4-ALK, Clinical Advances in Hematology & Oncology Volume 9, Issue 3, 2011,
A. Vultur, J. Villanueva and M. Herln, Targeting BRAF in Advanced Melanoma: A First Step toward Manageable Disease Clin Cancer Res April 1, :1658