Cancer Pharmacogenomics: Setting a Research Agenda to Accelerate Translation
Slide 1 of 37: Challenges in the Design and Conduct of Clinical Pharmacogenomic Studies
Mark J. Ratain, M.D.
University of Chicago
July 21, 2009
Slide 2 of 37: Goals of Clinical Pharmacogenomic Studies
- Evaluation of association of candidate polymorphism(s) with phenotype(s)
- Discovery of novel polymorphisms associated with phenotype(s)
- Replication of positive findings from another clinical study
Slide 3 of 37: Considerations in Study Design
- Prospective vs. retrospective
- Controlled vs. observational
- Sample size
- Planned vs. convenience
- DNA source
- Blood vs. saliva/buccal vs. tissue samples
- Gene selection
- Candidates vs. pathway vs. genome-wide
- Variant selection
- Functional vs. tag SNPs
- Control of type I error
- Exploratory vs. corrected
Slide 4 of 37: Considerations in Study Design - Phenotyping
- Continuous vs. discrete endpoints
- ANC nadir vs. grade 4 neutropenia (Y/N)
- Incorporation of endophenotypes (e.g., plasma concentrations or pharmacodynamic biomarkers)
- Minimization of phenotyping error
Slide 5 of 37: Iyer, Pharmacogenomics J, 2002 (n = 20 pts at 300 mg/m2)
[Image] showing the correlation between absolute neutrophil count and UGT1A1 promoter genotypes in patients after irinotecan treatment.
Reprinted by permission from Macmillan Publishers Ltd: Iyer et al, The Pharmacogenomics Journal, 2002; 2(1): 43-47 (figure 2), copyright 2002.
Slide 6 of 37: Neutropenia (every 3 week schedule) is Correlated with UGT1A1 Genotype (*28)
[Image] indicating neutropenia is correlated with UGT1A1*28 genotype.
Innocenti F, et al: J Clin Oncol. 2004 Apr 15;22(8):1382-8.
Reprinted with permission. © 2004 American Society of Clinical Oncology. All rights reserved.
Slide 7 of 37: Irinotecan Pharmacokinetics and TA Indel Genotypes
[Image] summarizing irinotecan pharmacokinetics and the TA indel genotype. Concerning the PK parameters, a significant trend was observed for SN-38 AUC and glucuronidation ratios of SN-38G AUC over SN-38. SN-38 AUC significantly increased as the number of TA7 alleles increased from 0 to 2.
Innocenti F, et al: J Clin Oncol. 2004 Apr 15;22(8):1382-8.
Slide 8 of 37: Repeated Measures of Intraocular Pressure Result in a Higher Heritability and Greater Power in Genetic Linkage Studies
[Image] showing screen shot of manuscript by Carbonaro et al, Invest Opthalmol Vis Sci, 2009 Nov 50(11):5115-9.
Using 4 measures versus 1 resulted in
- Increased absolute heritability (10-14%)
- Decreased required sample size (29-48%)
Slide 9 of 37: Power for Association With Error
[Image] showing screenshot of http://linkage.rockefeller.edu/pawe/paweph.htm, a program designed to perform asymptotic power and sample size calculations for genetic case-control studies with a di-allelic marker locus (for example, a SNP) in the presence of errors.
Slide 10 of 37: Considerations in Study Design - Treatment
- Prospective studies minimize confounding
- Treatment is assigned in an unbiased manner
- Not confounded by risk factors apparent to experienced clinicians
- Dose is controlled
- Observational approaches are best utilized in the context of replication of positive findings from prospective studies
Slide 11 of 37: UGT1A1*28 Genotype and Irinotecan-Induced Neutropenia: Dose Matters
[Image] showing probability of grade III-IV hematologic toxicity for *1/*1 and *1/*28 and *28/*28 for doses of irinotecan ranging from 100 to 400 mg/m2.
Hoskins et al, Journal of the National Cancer Institute, 2007; 99(17):1290-5 (figure 2).
Reprinted by permission of Oxford University Press.
Slide 12 of 37: Considerations in Study Design – Sample Size
- Phase III studies will always have sufficient sample size to test pharmacogenomic hypotheses of clinical relevance.
- Discovery studies incorporating genome-wide typing may be fruitful even with relatively small sample sets (e.g., 300 patients).
- For replication studies, the effect size is likely to be smaller than in the discovery study.
Slide 13 of 37: CALGB 40101 – 2 x 2 Factorial Design Adjuvant Therapy for Women with Breast Cancer with 0-3 Positive Nodes
[Image] showing the design of a pharmacogenetic analysis of predictors of paclitaxel and doxorubicin/cyclophosphamide toxicity for women with breast cancer (0 to 3 positive nodes). The patients were stratified by pre- or post-menopausal status and ER/PgR status, adjuvant therapy regimen, and number of cycles of therapy.
Slide 14 of 37: CALGB 40101 Accrual as of 9/9/08
[Image] showing number of patients enrolled to the various treatment arms of the CALGB parent protocol and the pharmacogenetics protocol. The pharmacogenetics companion was added at the time of a treatment modification. Treatment has remained the same throughout the study until last spring when the 6 cycle arms were dropped.
Slide 15 of 37: CALGB 80303 Trial Design
[Image] showing another CALGB trial design for a study of 580 patients with advanced pancreatic cancer randomized to receive either gemcitabine and bevacizumab or gemcitabine and a placebo. The stratification factors were performance status, extent of disease, and whether or not the patient previously received radiation.
Source: Kindler et al. 2007 ASCO Annual Meeting Proceedings (post-meeting edition) 25(18S), 2007:4508. Reprinted by permission of author.
Slide 16 of 37: GWAS in Pancreatic Cancer Patients Treated With Chemotherapy
CALGB #80303, Federico Innocenti, M.D., Ph.D., University of Chicago
Slide 17 of 37: Overall Survival
[Image] showing overall survival and GWAS results.
Slide 18 of 37: IL17F, ch 1, 10-8, q 0.04, HR 3.27 Coding Nonsynonymous
[Image] showing Cox’s proportional hazard model for A/A and A/G variants of rs763780 (Caucasian subset) plotting time since randomization in months.
Slide 19 of 37: SLCO1B1 Variants and Statin-Induced Myopathy – A Genomewide Study
[Image] showing results of tests for a trend in the association between myopathy and each SNP measured in a genome-wide association study. P values are shown for each SNP measured among 85 participants with myopathy and 90 matched controls who were taking 80 mg of simvastatin daily. Analyses were based on 316,184 of the 318,237 SNPs (99.4%) on the Sentrix HumanHap300-Duo BeadChip (Illumina). A horizontal red line represents P<5x10-7 and any SNP above this line may be associated with myopathy. In this figure, it appears there may be an association with rs4363657 (P = 4 x 10-9). This manuscript was published by the SEARCH Collaborative Group in the New England Journal of Medicine.
The SEARCH Collaborative Group. SLCO1B1 variants and statin-induced myopathy – a genomewide study. New England Journal of Medicine, 2008; 359(8): 789-799. Copyright © 2008 Massachusetts Medical Society. All rights reserved.
Slide 20 of 37: Considerations in Study Design – DNA Source
- Blood samples are preferable to other sources of germline DNA
- Tissue samples can be utilized, but are complicated by the potential for somatic mutations
- Genome-wide typing is difficult from paraffin-embedded samples
Slide 21 of 37: Collection of Blood, Saliva, and Buccal Cell Samples in a Pilot Study on the Danish Nurse Cohort: Comparison of the Response Rate and Quality of Genomic Data
[Image] showing chart with information about (not sure what scale on y axis is) for blood, saliva, buccal swabs, and buccal FTA card for genotyping and PCR amplification.
Source: Hansen et al. Collection of blood, saliva, and buccal cell samples in a pilot study on the Danish Nurse Cohort: comparison of the response rate and quality of genomic DNA. Cancer Epidemiology Biomarkers and Prevention. 2007; 16(10): 2072-6.
Slide 22 of 37: Considerations in Study Design – Gene Selection
- Candidate genes should be prioritized
- Endophenotypes increase the value of candidate and pathway gene studies
- Genome-wide discovery studies should always be considered as a secondary objective
Slide 23 of 37: Graphs of Comprehensive Pharmacogenetic Analysis of Irinotecan Neutropenia
[Image] showing association of UGT1A1*93, SLCO1B1*1b, and ABCC1 IVS11 -48C>T with absolute neutrophil count (ANC) nadir normalized to the baseline (pretreatment) ANCs. Data are expressed as medians and 25th and 75th percentiles. [Image] also includes results of the multivariate analysis for the final genetic model.
Innocenti F, et al: J Clin Oncol. 2009 Jun 1;27(16):2604-14.
Reprinted with permission. © 2009 American Society of Clinical Oncology. All rights reserved.
Slide 24 of 37: Rosner et al, CPT, 2008
[Image] showing pharmacokinetic model with enterohepatic recirculation.
Rosner et al, Clinical Pharmacology & Therapeutics, 2008; 84(3):393-402.
Reprinted by permission from Macmillan Publishers Ltd: Clinical Pharmacology & Therapeutics, copyright 2008.
Slide 25 of 37: Association of ABCC1 SNPs with Principal Components (p value)
- IRN to SN-38 & SN-38 recirculation (0.001)
- IRN compartments (0.004)
- SN-38 to SN-38G & SN-38G elimination
- IRN to APC
- APC elimination
- Enterohepatic recirculation (EHR)
- SN-38 recirculation without EHR
- SN-38 elimination
- IRN elimination
Slide 26 of 37: Considerations in Study Design – Variant Selection
- Linkage disequilibrium must always be considered
- Haplotype tag SNPS should be included whenever feasible
- Genome-wide platforms do not have complete coverage of all genes, particularly for individuals of African descent
Slide 27 of 37: Considerations in Study Design – Control of Type I Error
- Even exploratory studies should attempt to correct for multiple testing!
- Avoid testing of associations that lack plausibility
- Fundamental principles of statistics still apply to clinical pharmacogenomic studies!!
- Sources of false discovery include multiple testing of
- Continuous, subsets
- Genetic models
- Multivariate analyses
Slide 28 of 37: A Comprehensive Review of Genetic Association Studies
- “most reported associations are not robust”
- 603 putative gene-disease associations
- 166 putative associations studied 3+ times
- 6/166 (3.6%) consistently replicated
- 166 putative associations studied 3+ times
Hirschhorn J, et al. Genetic Medicine. 2002; 4(2): 45-61.
Slide 29 of 37: What is TheraGuide 5-FU™?
[Image] showing screenshot of description of TheraGuide 5-FU™ and references downloaded from Myriad Genetics Web site on 7/16/09.
Slide 30 of 37: References Cited by Myriad Genetics Regarding TheraGuide 5-FU™
- Morel, 2006 (DPYD)
- LeComte, 2004 (TYMS)
- Pullarkat, 2001 (TYMS)
- Ichikawa, 2006 (OPRT)
- Schwab, 2008 (DPYD, TYMS, MTHFR)
Slide 31 of 37: Thymidylate Synthase Gene Polymorphism Predicts Toxicity in Colorectal Cancer Patients Receiving 5-Fluorouracil-based Chemotherapy
Thierry Lecomte, Jean-Marc Ferraz, Franck Zinzindohoué, Marie-Anne Loriot, David-Alexandre Tregouet, Bruno Landi, Anne Berger, Paul-Henri Cugnenc, Raymond Jian, Phillipe Beaune, and Pierre Laurent-Puig. Thymidylate Synthase Gene Polymorphism Predicts Toxicity in Colorectal Cancer Patients Receiving 5-Fluorouracil-based Chemotherapy. Clinical Cancer Research, 2004; 10: 5880-8.
Slide 32 of 37: TheraGuide 5-FU™ Analysis Results
[Image] showing sample TheraGuide 5-FU™ test results and interpretation.
Slide 33 of 37: LeComte, CCR, 2004 - Methods
- 90 patients over 6 years (1995-2001)
- 8 different fluoropyrimidine regimens
- Included combinations with CPT-11 and OHP
- Retrospective phenotyping
- Toxicities, response, PFS, survival, etc.
- Tumor samples (frozen/FFPE) used for genotyping
- 1 gene (TYMS)
- 3 variants
Slide 34 of 37: LeComte, CCR, 2004 - Results
- 25 reported tests of association of genotypes/haplotypes and phenotypes
- 0.28 p values/subject
- “Patients with a 2R/2R, a 2R/3R, or a 3R/3R genotype had a grade 3 or 4 toxicity rate of 43, 18, and 3% respectively (P < 0.01).”
Slide 35 of 37: Role of Genetic and Nongenetic Factors for Fluorouracil Treatment-Related Severe Toxicity: A Prospective Clinical Trial by the German 5-FU Toxicity Study Group
[Image] showing screenshot of manuscript published by Schwab et al in the Journal of Clinical Oncology, 2008; 26(13): 2131-8.
- Phenotyping prospective
- Treatment uncontrolled (no other cytotoxics)
Schwab M, et al: J Clin Oncol. 2008; 26(13): 2131-8. Reprinted with permission. © 2008 American Society of Clinical Oncology. All rights reserved.
Slide 36 of 37: Figure 2. Nomogram for Estimating Individual Fluorouracil (FU) Toxicity Risk Based on the Multifactorial Model
[Image] showing nomogram for estimating individual fluorouracil (FU) toxicity risk based on the multifactorial model as published by Schwab et al in the Journal of Clinical Oncology, 2008; 26(13): 2131-8. Reprinted with permission. © 2008 American Society of Clinical Oncology. All rights reserved.
Slide 37 of 37: In Conclusion
- NCI should mandate (and fund) germline DNA collection on all NCI-funded phase III trials, unless acceptable justification is provided by the PI.
- NCI should interact more closely with the PGRN to ensure that its studies are of the highest possible quality.
- NCI also needs to consider how results of its studies will be appropriately replicated and translated into clinical practice.