Predicting Safety and Efficacy of Treatment for Colon Cancer Clinical Science Symposium Towards Personalized Medicine: Trials and Technologies That Will.

Predicting Safety and Efficacy of Treatment for Colon Cancer

Clinical Science SymposiumTowards Personalized Medicine: Trials and

Technologies That Will Lead to Individualized Therapy for Cancer

Neal J. Meropol, M.D.Fox Chase Cancer Center

Philadelphia, PAMay 31, 2008

Disclosures

• Consulting or Advisory role– Amgen– Astra Zeneca– Genentech– Genomic Health– Imclone– Saladax– Sanofi Aventis– Zealand Pharma

• Stock Ownership– Saladax

The Context

• Multiple treatment options for patients with colorectal cancer

• No single “correct” treatment algorithm• All available treatments are toxic• All available treatments have modest activity• No obvious new agents on the horizon

The Treatment Discovery Cycle in Oncology: Where are we?

Demonstrate Clinical Activity

Optimize Use

Drug Discovery

Examples• 5FU modulation• Newer cytotoxics• Antibodies

Personalized Medicine Prerequisites: Target, Drug, Classifier

Diagnosis Select Treatment

Diagnosis

Select Treatment

Apply DiagnosticClassifier

Select Treatment

Apply DiagnosticClassifier

Revise Treatment

Old paradigm: Empirical Medicine

New paradigm: Personalized Medicine

It’s all about variability:“Predictive” vs. “Prognostic”

• Predictive: explains variability in response to treatment

• Prognostic: explains variability irrespective of treatment

Variability exists in the host (germline) and tumor (somatic)

Why weren’t validation studies undertaken until recently?

• It wouldn’t affect clinical care– Limited options for alternative therapy– Results not sufficiently discriminating

• Love for new drugs• Technical aspects of assay performance

But

The times have changed

What should we expect from a classifier?

• It must assist in decision making– Must it be perfect as a discriminator?

• Yes - If no competing therapies • No - If competing therapies

• Possible results– Patient will definitely benefit – doesn’t tell

who not to treat– Patient will definitely not benefit – doesn’t

tell who to treat– Patient will be more or less likely to benefit

Potential Predictive Markers for Colon Cancer Treatment

Drug Marker

Fluoropyrimidines TS, DPD*, TP, MSI, MTHFR expression/polymorphisms

Irinotecan UGT polymorphisms*, MSI, transporter polymorphisms

Oxaliplatin ERCC1, GST P1, XPD expression, transporter polymorphisms

EGFR Antibodies gene amplification/polymorphism, RAS mutation, BRAF mutation, ligand expression, PTEN expression, VEGF levels

VEGF inhibitors VEGF polymorphisms, ICAM polymorphisms/levels, E-selectin levels, HIF1, Glut-1, VEGFr gene expression

General Circulating tumor cells

*FDA-recognized Yellow = presented at ASCO 2008

The personalized approach to treatment of colorectal cancer has

arrived

PFS benefit of panitumumab only seen in patient with

wild-type KRAS

R. Amado et el. JCO 2008

Mutant RAS

WT RAS

When added to FOLFIRI, the benefit of cetuximab is restricted to patients with

WT RAS tumors

Van Cutsem et al. ASCO Plenary, 2008

Wild type RAS (N=348)

Mutant RAS (N=192)

Response

FOLFIRI vs. FOLFIRI/Cetuximab

Favor cetuximab

P = 0.0025

No difference

Progression-Free Survival

FOLFIRI vs. FOLFIRI/Cetuximab

Favor cetuximab

HR = 0.68

P = 0.017

No difference

Tumor gene expression and K-Ras mutations in fixed paraffin-embedded tissue predict response to cetuximab in metastatic colon cancer

AuthorsJ.B. Baker1, D. Dutta1, D. Watson1, T. Maddala1, S. Shak1, E.K. Rowinsky2, L. Xu3 , E. Clark3 , D.J. Mauro3 , S. Khambata-Ford3

1Genomic Health, Inc. Redwood City, CA2Imclone Systems, Inc., New York, NY

3Bristol Myers Squibb, Princeton, NJ

Baker et al. Summary

• 226 patients with metastatic colorectal cancer treated with single-agent cetuximab

• Retrospective analysis of banked tissue from 3 studies (~425 patients in parent studies)

• Association of RAS mutation and quantitative gene expression with clinical outcomes

• Key findings:– Gene expression can be reliably measured in

FFPE tissue– RAS mutation associated with lack of response– 4-gene model discriminates outcomes (“disease

control” and PFS) in patients with WT RAS

If validated, is this test “good enough” to assist in treatment decision making?

WT RAS

Response + SD 87 (60%)

Disease Progression

57 (40%)

Total 144 (100%)

If validated, is this test “good enough” to assist in treatment decision making?

WT RASLow Response

Gene ScoreHigh Response

Gene Score

Response + SD 87 (60%) 16 (27%) 71 (85%)

Disease Progression

57 (40%) 44 (73%) 13 (15%)

Total 144 (100%) 60 (100%) 84 (100%)

Clinical utility is dependent on other available options

Things I’d like to know more about

• Platform characteristics– e.g. how frequent are indeterminate results?

• Prediction vs. Prognosis– Is gene expression profile associated with

response or only “disease control”?– “Disease control” may be heavily influenced by

natural history rather than treatment• Will equivalent results be obtained with other

EGFR inhibitors?• Will the use of this test result in improved patient

outcomes for patients?• These data are worthy of validation in an

independent patient sample

H. L. McLeod, K. Owzar, D. Kroetz, F. Innocenti, S. Das, P. Friedman, K. Giacomini, R. Goldberg, A. Venook, M. J. Ratain

Univ of North Carolina-Chapel Hill, Chapel Hill, NC; Duke, Durham, NC; UCSF, San Francisco, CA; University of Chicago, Chicago, IL; CALGB, Chicago, IL

Cellular transporter pharmacogenetics in metastatic colorectal cancer: initial analysis of

C80203

McLeod et al. Summary

• Genomic DNA from 180 of 238 patients on C80203 (FOLFOX vs. FOLFIRI +/- cetuximab)

• Genotyping of transporter genes involved in irinotecan and oxaliplatin clearance:– ABCC2, ABCC4, ABCG2, SLCO1B1, SLC22A1,

and SLC22A2• Association of genotype with response and toxicity

• Key findings:– ABCG2 34 G>A associated with response to

FOLFOX, resistance to FOLFIRI– No associations with toxicity

Pharmacogenetics (Genetic Variation) Impacts Pharmacokinetics and Pharmacodynamics

Pharmacokinetics-Absorption-Distribution-Metabolism-Excretion

PharmacodynamicsTumor Normal Tissue

Response Toxicity

Dose andCompliance

Drug focused

Target focusedTarget focused

The promise and challenge of pharmacogenetics

• The promise– Mechanism-based– Non-invasive– Response and toxicity prediction

• The challenge– Low-frequency alleles– Multiple interrelated systems– Large sample sizes required to develop

and validate models

What have we learned?

• Germline DNA collection is possible in an Intergroup clinical trial

• ABCG2 34 G>A polymorphisms are uncommon• Association with treatment effect requires clinical

validation and mechanistic support (previous published work suggests no impact on irinotecan PK and increased in vitro sensitivity)

• Large sample sizes will be required to identify predictive associations with low frequency SNPs

• Individual SNPs as predictive markers will likely be rare given the complexity of drug metabolism and clearance, target expression and function, and mutidrug regimens

• Candidate gene selection based on pathway understanding complements genome wide screening efforts

We must be prepared to integrate new findings

• Patient care– Recognize that predictive markers will generally

not provide absolute guidance– Assess and communicate value and comparative

effectiveness of personalized approaches– Develop streamlined systems for tumor and

germline marker assessment • Research

– Emphasize prospective tissue acquisition– Anticipate and react promptly to new data that

impact ongoing research studies and patient care

The impact of personalized medicine for pharma is uncertain

Advantage?• Costs

– Development time ?– Production =

• Risks– Success rate ?

• Returns– Market size -– Duration of treatment +– New treatment market +– New diagnostic market ?– Competition +– Pricing (value, novelty, need) ?

Conclusions

• We can successfully personalize the therapy of patients with colorectal cancer

• We must continue to build well-annotated tissue repositories in prospective randomized clinical trials

• Now more than ever, industry and academia must identify shared goals

• With more effective personalized treatment approaches we have an opportunity to shift emphasis from the traditional focus on p-values (and marginal benefit) to focus on (and demand for higher) value of new innovation