This document discusses challenges in designing pharmacogenomics clinical trials. It provides an overview of pharmacogenomics and different types of pharmacogenomics studies. It then discusses three common clinical trial designs - subgroups analysis design, enrichment design, and genotype-guided design - and their advantages and disadvantages. Key challenges in pharmacogenomics clinical trials include small sample sizes for subgroups, possible confounding and selection biases, and statistical power issues. Prospective clinical trials are needed to validate predictive biomarkers and assess clinical utility of genotype-guided treatments.

1. 11
Challenges in Designing
Pharmacogenomics Clinical Trials
Joseph Levy, PhD
Teva Pharmaceutical Industries
MedicReS 2013 Istanbul

2. 2
The present
One Size fits all

3. 3
The Future

4. 4
Outline
 PGx short overview
 Response related research questions
 Clinical trial designs
 Challenges

5. 5
What is Pharmacogenomics?
 Pharmacogenomics (PGx) is the study of how genes
affect the way our bodies respond to a medicine.
 An inter-disciplinary science involving
 Biology and Genetics
 Medicine
 Pharmacology
 Statistics

6. 6
Goals of PGx research
 Achieving better safety profile and mitigating potential
risks
 Better understanding of MoA
 Identify disease genetics
 Investigating the contribution of genetic variation to inter-
individual variability in response to drug treatment
 Promoting an optimal drug response for the individual
patient

7. 7
The Human Genome
 46 chromosomes – 23 pairs
 ~2 meters of DNA
 ~3.4B DNA bases
 ~25000 genes
 ~26M variable sites (0.7%)
 ~10M SNPs

8. 8
Types of PGx studies
Studies to investigate genotype-trait association within a
population of unrelated individuals:
 Exploratory studies
 Gnome-wide association studies (GWA/GWAS)
 Whole Genome Sequencing (WGS) studies
 Prospective studies
 Candidate polymorphism studies
 Candidate gene studies

9. 9
GWA and WGS studies
 Consider large set of SNP and possibly other variants
 Aim to identify association between variants and phenotype
 Less hypothesis driven
 Nee to handle multiplicity testing issues, but…
 Type II error (False Negative) is more important than Type I
error (False Positive)

10. 10
Candidate gene/polymorphism studies
 Consider polymorphism(s) within a gene or multiple genes
 There may be a priori hypothesis about functionality
 Functionality: the given variants influence the disease trait
directly
 Non-functionality: the given variants may be associated to a
functional variant within the gene(s)
 This association is called Linkage Disequilibrium
 In any case, the candidate gene/polymorphism is pre-
specified

11. 11
Response related research questions
1. Does the treatment effect of the drug vary between
subjects with different genotypes?
2. What is the benefit of the drug over placebo or an standard
treatment for patients with a particular genotype?
3. What is the risk-benefit ratio of genotype-guided treatment
over standard care?

12. 12
Subgroups analysis design
Randomi
zation
O
U
T
C
O
M
E
Drug
Pbo/Std
Genoty
pying
Wildtype
Variant
Genoty
pying
Wildtype
Variant
O
U
T
C
O
M
E
Double
Blind
Follow-up
Answers first research question

13. 13
FORTE GWA study
 The FORTE study compared Glatiramer Acetate 20 mg and
40 mg doses in RRMS patients.
 731 subjects consented to provide DBA samples
 Response classification was available for 599 Caucasian
subjects
 Extreme phenotype approach: 52 super-responders, 61
non-responders
 1M SNPs considered
 31 SNPs were identified significantly associated with
SR/NR
 2 relevant pathways that merge into a unique functional
gene network were identified
 Predictive model for response: a 87.9 % sensitivity, 92.7 %
specificity

14. 14
Tamoxifen therapy for breast cancer
 Clinical trial performed in Sweden, 1996
 Four treatment groups, following a modified radical
mastectomy :
 Adjuvant chemotherapy
 Adjuvant chemotherapy + tamoxifen
 Radiotherapy
 Radiotherapy + tamoxifen
 679 postmenopausal breast cancer patients randomized
 226 fresh frozen tumor tissues were available 7 years later
 Relationship of CYP2D6 and SULT1A1 genotypes and
benefit from tamoxifen therapy was observed

15. 15
Subgroups analysis design
advantages
 Can piggyback on Phase II or III trial
 Appropriate for GWA/WGS exploratory studies
 Can be performed in prospective/retrospective fashion
 Simple, fast, relatively inexpensive
 Can test many genetic markers in one study
 Small chance of bias even where there are many genetic
subgroups
 Provides efficient assessment of relative treatment efficacy
in each genotype subgroup and in the whole group

16. 16
Subgroups analysis design
disadvantages
 Possible confounding bias
 Possible selection bias:
 Some subjects may not consent
 If DNA is not collected at baseline, clinical outcome may affect
decision to consent
 Dependency of genotype distribution in the original study
population
 Size of subgroups can’t be controlled
 Possible imbalance in baseline characteristics/ prognostic factors
 Statistical power issues

17. 17
Enrichment design
Genoty
pying
Wildtype
Variant
Excess of
Wildtype
Randomi
zation
Drug
Pbo/Std
Randomi
zation
Drug
Pbo/Std
O
U
T
C
O
M
E
Double
Blind
Follow-up
Answers second research question
Sometimes
off-study

18. 18
Impact of CYP2D6 on venlafaxine and
desvenlafaxine PK
 14 healthy volunteers
 7 CYP2D6 poor metabolizers
 7 CYP2D6 extensive metabolizers
 CYP2D6 intermediate and ultra-rapid metabolizers were
excluded
 An excess of extensive metabolizers was also excluded
 Randomized sequence crossover study
 Endpoints – AUC and Cmax
 Conclusions
 No impact of CYP2D6 polymorphisms on exposure to
desvenlafaxine
 CYP2D6 polymorphisms impacts exposure to venlafaxine

19. 19
Anti-HER2 monoclonal antibody plus chemotherapy
for metastatic breast cancer
 Genotyping at screening – 469 women that that over-
expressed HER2 were randomized
 HER2 is over-rexpressed in 25-30% percent of breast
cancers ⇒ sample size without enrichment would be 1550-
1900
 Randomization stratified by past treatmenat by an
anthracycline:
 Standard chemotherapy (an anthracycline and
cyclophosphamide or paclitaxel)
 Standard chemotherapy + tratuzumab
 Conclusion: Trastuzumab increases the clinical benefit of
first-line chemotherapy in metastatic breast cancer that
overexpresses HER2

20. 20
Enrichment design advantages
 Genotype strata can be balanced
 Can select subjects with genotypes between which the
largest difference in treatment effect is expected (instead of
all genotypes)
 Sample size and power calculation takes final analysis into
account
 Selective consenting bias is avoided
 Opens door for possible adaptations

21. 21
Enrichment design disadvantages
 “Tailored” study for particular genetic hypothesis
 Appropriate for contexts where there is such a strong
biological basis for believing that “wildtype subjects” will not
benefit from the new drug
 Efficient only when prevalence of variant is high and the
effectiveness in variant population is high compared to the
wildtype population

22. 22
Genotype guided design
Randomi
zation
Drug
Pbo/Std
Genoty
pying
Wildtype
Variant
O
U
T
C
O
M
E
Double
Blind
Follow-up
Answers third research question
Exclusion or
alternative treatment
Drug

23. 23
Sensitivity to primary chemotherapy with
paclitaxel in breast cancer

24. 24
The Tarceva Italian Lung
Optimization study

25. 25
Genotype guided design advantages
 Assesses the added value of the PGx-based treatment over
the current use of the drug and the corresponding costs
 Economic advantage of limiting the number of potentially
expensive DNA genotypings
 Can be mimicked by a “regular” clinical trial, by comparing
the whole treatment arm to its subgroup of “variant
subjects” (statistical analysis is computationally intensive)

26. 26
Genotype guided design
disadvantages
 Inefficient in terms of sample size
 A positive study cannot distinguish between a successful
treatment selection strategy and a situation in which the
experimental drug is better than the control therapy for all
patients

27. 27
More to consider
 Adaptations: randomization, patient enrollment, enrichment,
sample size re-estimation, group sequential design
 Combine general efficacy endpoint and PGx related
endpoint in a single study – need to consider multiplicity
issues

28. 28
References
 van der Baan, F. H., Klungel, O. H., Egberts, A. C., Leufkens, H. G., Grobbee, D. E., Roes, K.
C., & Knol, M. J. (2011). Pharmacogenetics in randomized controlled trials: considerations for
trial design. Pharmacogenomics, 12(10), 1485-1492.
 Farina, G., Longo, F., Martelli, O., Pavese, I., Mancuso, A., Moscetti, L., ... & Scanni, A. (2011).
Rationale for Treatment and Study Design of TAILOR: A Randomized Phase III Trial of Second-
line Erlotinib Versus Docetaxel in the Treatment of Patients Affected by Advanced Non–Small-
Cell Lung Cancer With the Absence of Epidermal Growth Factor Receptor Mutations. Clinical
lung cancer, 12(2), 138-141.
 Freidlin, B., McShane, L. M., & Korn, E. L. (2010). Randomized clinical trials with biomarkers:
design issues. Journal of the National Cancer Institute, 102(3), 152-160.
 Ito, Y., Nagasaki, K., Miki, Y., Iwase, T., Akiyama, F., Matsuura, M., ... & Hatake, K. (2011).
Prospective randomized phase II study determines the clinical usefulness of genetic biomarkers
for sensitivity to primary chemotherapy with paclitaxel in breast cancer. Cancer science, 102(1),
130-136.
 Macciardi, F., Cohen, J., Comabella Lopez, M., Comi, G., Cutter, G., Eyal, E., ... & Wolinsky, J.
(2012, April). A Genetic Model To Predict Response to Glatiramer Acetate Developed from a
Genome Wide Association Study (GWAS). In NEUROLOGY (Vol. 78).

29. 29
References (Cont.)
 Mandrekar, S. J., & Sargent, D. J. (2009). Clinical trial designs for predictive biomarker
validation: one size does not fit all. Journal of biopharmaceutical statistics, 19(3), 530-542.
 Preskorn, S., Patroneva, A., Silman, H., Jiang, Q., Isler, J. A., Burczynski, M. E., Saeeduddin
A., Jeffrey P., & Nichols, A. I. (2009). Comparison of the pharmacokinetics of venlafaxine
extended release and desvenlafaxine in extensive and poor cytochrome P450 2D6
metabolizers. Journal of clinical psychopharmacology, 29(1), 39-43.
 Simon, R. (2012). Clinical trials for predictive medicine. Statistics in Medicine, 31(25), 3031-
3040
 Slamon, D. J., Leyland-Jones, B., Shak, S., Fuchs, H., Paton, V., Bajamonde, A., ... & Norton,
L. (2001). Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic
breast cancer that overexpresses HER2. New England Journal of Medicine, 344(11), 783-792.
 Trepicchio, W. L., Essayan, D., Hall, S. T., Schechter, G., Tezak, Z., Wang, S. J., Weinreich D.,
& Simon, R. (2006). Designing prospective clinical pharmacogenomic (PG) trials: meeting
report on drug development strategies to enhance therapeutic decision making. The
pharmacogenomics journal, 6(2), 89-94.
 Wegman, P., Vainikka, L., Stal, O., Nordenskjold, B., Skoog, L., Rutqvist, L. E., & Wingren, S.
(2005). Genotype of metabolic enzymes and the benefit of tamoxifen in postmenopausal breast
cancer patients. Breast Cancer Res, 7(3), R284-R290.

30. 3030
Backup Slides

31. 31
SNPs
 SNP - Single Nucleotide
Polymorphism
 Most common type of
genetic variation
 Each SNP represent a
difference in a single DNA
base
 The SNP in the picture is
CT or AC or AG or GT –
they are all the same
 SNP can have 3 possible
values: AA, Aa or aa