HBN_Medimmune_Biomarker_Higgs

PowerPoint Presentation: 

Biomarkers that Facilitate Clinical Trials for Autoimmune Disease Research Frontiers of Personalized Medicine: A Fresh Look Ahead, Johns Hopkins University Oct 29, 2011 Brandon Higgs, PhD MedImmune

Overview: 

Overview Objectives of PHC Characteristics of a robust predictive biomarker Early stage biomarkers predictive of treatment response Rheumatoid Arthritis (RA) Autoantibodies, cytokines, type I IFN activity, and genetic variants predict TNF α inhibitor or CD20 inhibitor response Polymyositis (PM) Granulysin expression in CD8+ lymphocytes predicts prednisolone response/resistance Psoriasis IL-17, IL-1, MMP-12 genes predict TNF α inhibitor response Relapsing remitting multiple sclerosis (RRMS) Gene signatures predict response to IFN β Systemic Lupus Erythematosus (SLE) Autoantibodies or CD19+ cell counts predict flares or SLEDAI score drop following B cell depletion therapy Later stage biomarker development Type I IFN-inducible genes may predict IFN α inhibitor response A Th2 signature in asthma patients may predict IL-13 inhibitor response Challenges Summary

Why do we need personalized healthcare?: 

Why do we need personalized healthcare? Perfect Medicine Effective in all patients The same dose for every patient No adverse events Real Medicines Effective only in some patients Dose varies for different patients Some patients may develop adverse events Courtesy of Dr. Ruth March, AZ

Goal of personalized medicine: 

Goal of personalized medicine

Potential roles of biomarkers in targeted therapy: 

Potential roles of biomarkers in targeted therapy Apply evolving technologies to explore biomarkers to identify patients likely to respond to an indicated therapy Enhance entry and revenue potential in crowded disease indications Prevent patients who are more likely to have a tox response from receiving treatment Enhance the chances of a drug approval by increasing likelihood of response Downing GD., Biomarkers and Surrogate endpoints: clinical research and applications, Elsevier 2000. Devanarayan V., Biomarker World Congress 2007, Philadelphia

PowerPoint Presentation: 

Phase III Pre-Clinical Phase I Phase II Post Approval Validate predictive biomarker hypothesis Integrate regulatory, commercial, and business development feedback to execute diagnostic strategy and ensure coordinated drug and diagnostic launch Apply technologies to identify lead indication as well as product life cycle management Contribute to predicting therapeutic dosing range by developing PD markers Utilize animal models and cell based assays to gain understanding of target pathways and develop PD biomarkers Establish predictive biomarker hypothesis for PHC approach Provide PD data to enable PK/PD modeling and future dose selection Establish proof of biological activity and mechanism of action Initial testing of predictive biomarker hypothesis Continue execution of biomarker strategy and confirm predictive biomarker hypothesis Generate PD data to guide selection of optimal dose Support Phase III study design Monitor available technologies and incorporate into diagnostic device life cycle management Continue to provide input into product life cycle management through expansion of indications Biomarker development across all stages of R&D

Characteristics of a robust predictive biomarker: 

Characteristics of a robust predictive biomarker Mechanism relevant Reflects the current status of disease Easily accessible and reproducibly measured Sensitive and specific for outcome Capable of transitioning from a research setting (e.g. RUO) to a regulatory agency-approved platform and capacity

PHC is advancing clinical practice: 

PHC is advancing clinical practice Translational science is pushing implementation of PHC approaches into different therapeutics and disease indications Identifying the appropriate patient population for the appropriate treatment New technologies are leading these discovery efforts to identify putative predictive biomarkers of treatment response in various indications This talk will present multiple examples in autoimmunity of preclinical and early stage discoveries Very few biomarkers predictive of treatment response have moved into later stage development Some work demonstrates promising hypotheses that require rigorous testing in clinical trials Qualifier: This is not an exhaustive survey of all treatment response-predictive biomarker studies in autoimmune diseases, rather a brief summary of recent work

A blood autoantibody and cytokine panel predicts etanercept response in RA patients: 

A blood autoantibody and cytokine panel predicts etanercept response in RA patients Panel of 24 protein biomarkers predict response to etanercept in RA patients at least 3 months post treatment Arthritis antigen array and cytokine array measuring patient sera (>540 proteins) Use ≥ ACR50 and <ACR20 as responder and non-responder criteria, respectively Included 3 independent cohorts US-based, Japanese, and Swedish Biomarker prediction accuracy based on training set and cross-validation (no validation set) PPV=58-72% NPV=63-78% Hueber et al., Arthritis Res Ther, 2009

Type I IFN activity in plasma predicts TNF-α antagonist response in RA patients: 

Type I IFN activity in plasma predicts TNF- α antagonist response in RA patients Plasma type I IFN activity (reporter cell assay), IFN β / α ratio (using mAbs), and IL-1Ra (ELISA) levels predict RA patient (n=35) response to etanercept, infliximab, or adalimumab Using DAS28 evaluated from 3-9 months Agnostic to specific TNF α antagonist Stratified patients into different combinations of No, Moderate, or Good response Mavragani CP et al., Arthritis and Rheum, 2010

Type I IFN activity in blood associates with infliximab response in RA patients: 

Type I IFN activity in blood associates with infliximab response in RA patients Data suggest TNF- α antagonists modulate expression levels of IFN -inducible genes Increased type I IFN gene signature in peripheral blood cells at 1 month post infliximab treatment predicts poor clinical response in RA patients (n=33) Using EULAR response criteria at 16 weeks van Baarsen LG et al., Arthritis Res and Ther, 2010 Not a baseline predictor; requires 1 month post treatment

Type I IFN activity in PBMCs predicts CD20 inhibitor response in RA patients: 

Type I IFN activity in PBMCs predicts CD20 inhibitor response in RA patients A type I IFN gene signature negatively predicts response to rituximab in RA patients (n=51) Using DAS28 evaluated at 12 and 24 weeks Lack of consistency between independent cohorts (c 1 n=20; c 2 n=31) †=p<0.10; *=p<0.05; **=p<0.01 Thurlings RM et al., Arthritis Rheum, 2010

Genetic variants associated with treatment response in RA: 

Genetic variants associated with treatment response in RA Numerous genetics studies have reported polymorphisms associated with TNF antagonists such as etanercept, infliximab, and adalimumab SNPs identified within regions of TNF- α , MHC, FC γ receptor, etc Typically use candidate gene approaches involved with TNF- α signaling Meta-analysis methods can better elucidate consensus effects The next few slides demonstrate results from studies ranging from modest size to the largest to date, and a recent meta-analysis study Genetics studies selected required at least 100 patients Plenge RM et al., Current Opinions in Rheumatology, 2008

Three genetic studies of anti-TNFα response in RA: 

Three genetic studies of anti-TNF α response in RA Plenge RM et al., Curr Opin Rheumatol, 2008 1 Criswell LA et al., Arthritis Rheum, 2004 2 Marotte H et al., Ann Rheum Dis, 2008 3 Padyukov L et al., Ann Rheum Dis, 2003 Anti-TNF α response in RA studies (100<n<200 in each) Etanercept example: 2 SNPs identified 1 Infliximab example: No significant DNA variants found 2 Etanercept example: No significant DNA variants found 3 Small studies have limited power to detect adequate effect sizes; no replication demonstrated

Genetic study of anti-TNFα response in RA: largest GWAS to date: 

Genetic study of anti-TNF α response in RA: largest GWAS to date Included 1,286 patients A discovery cohort of 566 patients and independent cohorts of 379 and 341 for combined analysis in validation Assessment: 6 months using DAS28 Disease duration: ~13 years Findings 5 SNPs within intergenic regions, 2 SNPs within genes PDZD2 and EYA4 Summary 2 SNPs showed improved and reduced response to treatment, agnostic to the actual therapy No SNPs identified were correlated with loci previously published Plant D et al., Arthritis Rheum, 2011

Meta-analysis of 9 genetic studies of anti-TNFα response in RA (using DAS28 or ACR20): 

Meta-analysis of 9 genetic studies of anti-TNF α response in RA (using DAS28 or ACR20) Plant D et al., Arthritis Rheum, 2011 Included a total of 692 patients Obvious target is the promoter region of the TNF- α gene TNF- α variant -308 (G/A) appears most promising among other SNPs identified Summary A allele : 22% in responders and 37% in non-responders (p=2.5e-4) Agnostic to TNF α inhibitor therapy O’Rielly DD et al., Pharmacogenomics J, 2009

Granulysin expression in CD8+ lymphocytes predicts prednisolone response in polymyositis : 

Granulysin expression in CD8+ lymphocytes predicts prednisolone response in polymyositis Granulysin levels in CD8+ cells are elevated in pre-treatment muscle biopsies for PM patients found to be steroid resistant ( n=6 ), compared to steroid PM patient responders ( n=11 ) Ratio of the number of double-positive cells for granulysin and CD8 to all CD8 cells Perforin levels were not considerably different Ikezoe K et al., J Neurol Neurosurg Psychiatry, 2006

Preliminary studies of genomic predictors of TNF-blocking treatment response in psoriasis: 

Preliminary studies of genomic predictors of TNF-blocking treatment response in psoriasis Small study (n=10) in moderate to severe psoriasis patients treated with etanercept shows down-regulation of multiple proinflammatory pathways 1 Skin biopsies showed genes IL-1, MMP-12, and type I signaling correlated with patient improvement (using PASI) at 1 month Small study (n=15) in psoriasis vulgaris patients treated with etanercept 2 Lesion skin biopsies showed down-regulation of IL-17 pathway genes in responders 1 Gottlieb AB et al., J Immunol, 2005 2 Zaba LC et al., Clin Immunol, 2009 Not baseline predictors; mechanism observed post treatment

Nine sets of gene triplets in PBMCs predict response to IFNβ in RRMS patients: 

Nine sets of gene triplets in PBMCs predict response to IFN β in RRMS patients Baseline gene triplets predict response to IFN β 2 years after initiation of treatment in RRMS patients Responders defined as total suppression of relapses and no increase in the expanded disability status scale (EDSS) Poor responders defined as having suffered 2 or more relapses or having a confirmed increase of one point in the EDSS Patients with intermediate phenotypes were excluded Top scoring triplet: Casp2, Casp10, and FLIP Study n=52 patients with 64-100% accuracy Baranzini SE et al., PLoS Biology, 2005

Type I IFN-inducible genes in PBMCs predict response to IFNβ in RRMS patients: 

Type I IFN-inducible genes in PBMCs predict response to IFN β in RRMS patients Baseline type I IFN gene signature predicts RRMS patient response to IFN β after 2 years Responders defined as total suppression of relapses and no increase in the EDSS Poor responders defined as having suffered 1 or more relapses or having a confirmed increase of one point in the EDSS persisting for a minimum of 2 consecutive visits separated by a 6-month interval Different gene sets were selected in the discovery and validation patient cohorts Discovery set n=6 Validation set n=8 Study discovery set n=47 patients with 78% accuracy and validation set n=30 with 63% accuracy Comabella M et al., Brain, 2009

Serum autoantibodies predict flares post B cell depletion therapy in SLE: 

Serum autoantibodies predict flares post B cell depletion therapy in SLE Baseline anti-extractable nuclear antigens (ENA) predict flares post rituximab and cyclophosphamide treatment in SLE patients Low baseline serum complement C3 in SLE patients showed a shorter time interval to flare post treatment Study n=32 patients Ng KP et al., Ann Rheum Dis, 2007

Serum autoantibodies and CD19+ cell counts predict response to B cell depletion therapy in SLE: 

Serum autoantibodies and CD19+ cell counts predict response to B cell depletion therapy in SLE Anti-dsDNA antibodies showed decreases in IgG and IgA, but not IgM post rituximab and cyclophosphamide treatment in SLE patients 1 Higher CD19+ lymphocyte counts at baseline were associated with SLE patients not achieving remission ( SLEDAI<3 ) at 6 months 1 Study n=16 patients Two large clinical trials in SLE did not report any correlation between either autoantibody levels or CD19+ cell counts and rituximab response 2,3 Though difference have been observed between drug treatment and placebo treatment – proof of biological activity 2,3 1 Jonsdottir T et al., Ann Rheum Dis, 2008 2 Merril et al., Arthritis Rheum, 2010 3 Tew et al., Lupus, 2010

Type I IFN gene signature in the blood divides SLE patients into high and low signature groups: 

Type I IFN gene signature in the blood divides SLE patients into high and low signature groups A type I IFN gene signature is suppressed in a dose-dependent manner in SLE patients 3 months post single dose of sifalimumab, an investigational drug Gene signature was developed from >300 SLE patients Expression of the type I interferon signature in blood reflects involved tissue in SLE Suggests potential predictive biomarker role to identify SLE patient subsets who may preferentially respond to anti-IFNα treatment Yao Y et al., Arthritis Res Ther, 2010 Proprietary to MedImmune

IL-13 inducible genes in bronchial epithelium divides moderate and severe asthmatic patients into two groups: 

IL-13 inducible genes in bronchial epithelium divides moderate and severe asthmatic patients into two groups An IL-13-inducible gene signature in airway epithelial brushings from 42 asthmatics and 28 healthy controls provides a Th2-high and Th2-low signature group 1 ~50% of the patients have the Th2-high signature 1 Clinical features correlate with patients in the Th2-high and low groups 1 Provides hypothesis for Th2 cytokines as targets for a subset of asthma patients with this high signature Recent clinical trial shows Th2 high/periostin high subgroup has higher FEV1 increase from baseline from placebo at week 12 compared to all comers 2 Woodruff PG et al., Am J Respir Crit Care Med , 2009 Corren et al., NEJM, 2011

Challenges facing predictive biomarker studies in autoimmune diseases: 

Challenges facing predictive biomarker studies in autoimmune diseases Hypothesis-driven biomarker versus best variable correlates Few or no studies include a placebo-treated group Replication of findings remains a challenge due to various factors of clinical heterogeneity Baseline characteristics Demographic variability Concurrent medications (NSAIDs, methotrexate, etc) Disease duration Disease subtypes Sample sizes (most studies presented here had <100 patients) Trial design and patient inclusion/exclusion criteria Clinical response criteria RA: ACR20, ACR50, DAS28 SLE: BILAG, flares, SLEDAI, steroid minimization Polymyositis: MMT8, MITAX, steroid minimization Etc. Plenge RM et al., Current Opinions in Rheumatology, 2008

Factors that can confound the perceived analyte pattern: 

Factors that can confound the perceived analyte pattern p<0.01; FC>1.5 On NSAIDs Not on NSAIDs p<0.01; FC>1.5 Dummy data example

No clear identification of analyte cut point: 

No clear identification of analyte cut point p<0.01; FC>1.5 p<0.01; FC>1.5 Cut point PPV=52% NPV=0% Cut point PPV=100% NPV=59% Cut point PPV=65% NPV=63% Dummy data example

Lack of reproducibility in an independent patient population: 

Lack of reproducibility in an independent patient population p<0.01; FC>1.5 P=0.86; FC>1.5 New drug-treated population Dummy data example

Analyte difference between drug- and placebo-treated patient cohorts does not necessarily correlate with patient response: 

Analyte difference between drug- and placebo-treated patient cohorts does not necessarily correlate with patient response p<1x10 -15 ; FC>1.5 p<1x10 -15 ; FC>1.5 Responders Non-Responders Dummy data example

Summary: 

Summary Personalized Healthcare has the potential to benefit many in the healthcare system patients, partners, physicians, payers, pharma/biotech Predictive biomarkers for efficacy and safety have been adopted in several therapeutic areas, particularly in oncology Autoimmune diseases are complicated and heterogeneous diseases Robust predictive biomarkers for efficacy and safety for targeted therapies will benefit patients and also likely to improve the probabilities of success of pivotal trials Public resources Biomarker Commons: http://biomarkercommons.org/biomarker-research/category/autoimmune-system-disease

Acknowledgements: 

Acknowledgements MedImmune Yihong Yao Koustubh Ranade Wei Zhu Chris Morehouse Roger Liu Zach Brohawn Wendy White Laura Richman Bing Yao Bahija Jallal