Linking biobanks to registries: Why and how? Anne Barton
Linking biobanks to registries:
Why and how?
Anne Barton
Biobanks – why should we collect
samples?
Anti-TNF treatment in RA
• Cost approx. £8,000/person/year
• 30 - 40% RA patients do not respond
• Rare, serious adverse events
• Response likely to be multi-factorial
Predictors of treatment efficacy
Overall response prediction
• Several disease-related factors are predictive of anti-TNF
response
Concurrent DMARD therapy
Higher baseline HAQ Score Female gender
RF/Anti-CCP
R2 =0.17
Why should we collect samples?
• Clinical factors alone insufficient to predict efficacy
• Better targeting of treatment = more cost-effective
• Avoids potential harm in those unlikely to respond
• Clues as to mechanism of non-response
– ?different sub-phenotypes
– ? Different biological pathways
Treatment Response
Genetic
Epigenetic Transcriptome
Adherence
Overall hypothesis
Disease-related clinical, demographic, serological,
psychological and genetic factors define specific
subgroups of patients who are more or less likely to
respond to anti-TNF therapy
1) Class effect
2) Individual drug effect
Biobanks for pharmacogenetics
Pharmacogenetics
• Hypothesis:
Genetic factors influence treatment response
• Advantage of genetic predictors
– Stable – can collect after treatment has started / finished
– Easy to assess
– Clues about causality
• A national register of patients
with rheumatic diseases in the
UK receiving biologic therapy (up to ceiling of n=4000 per drug)
• All hospitals in the UK
• Commenced 2001
• Primary aim: assess long-term
safety and efficacy
BSRBRBSBSRRBRBRBSRBRBSBSRRBRBRThe British Society for Rheumatology
Biologics Register
Watson et al, 2007
Biologics in Rheumatoid Arthritis Genetics
and Genomics Study Syndicate
• Aim of BRAGGSS
– Investigate genetic predictors of response to anti-
TNF therapy
• Large nationwide multi-centre collaboration
• Recruited patients registered with BSRBR
• Target: recruit 4,000 RA patients treated with
etanercept, infliximab or adalimumab
Patient cohort
• Patients identified from BSRBR register
• Actively involved in the BSRBR
• Consultant based RA status
• Baseline and 6/12 follow-up DAS28 score
• Caucasian
• Recruitment and blood sample collection through mail
correspondence (COREC 04/Q1403/37)
Patient recruitment
* % compared to those initially contacted
Stage Numbers (%*)
Centres recruited 54
Patients contacted 3965
Patients responded 3194 (81%)
Patients participating 2921 (74%)
Bloods received 2590
GWAS of anti-TNF response
• Plant et al 2011: GWAS 566 UK patients
– WTCCC
– 5 loci identified, none replicated
• Krintel et al 2012
– N = 196 anti-TNF treated Danish subjects
– No genome-wide hits
– Replication of PDE3A-SLCO1C1
in Spanish cohort (n ~350) with EULAR response
• Mirkov et al 2012
– GWAS 882 Dutch patients
– 8 loci identified
– None replicated, yet
• Cui et al 2013: GWAS 2,700
– CD84 identified, p = 8 x 10-8
– Etanercept-treated
Candidate genes
• Conflicting evidence for association of TNF -308
– Recent meta-analysis: no association
• PTPRC
– Reported by Cui et al with good/poor response
– Replicated by Plant et al
– Not replicated by CORRONA; Dutch GWAS
Role of genetics?
• Genetic studies have provided little supportive
evidence
– Lack of power to detect modest effects
– Treatment response has little/no genetic
component
– The measure of response (DAS28) is
inappropriate
Predictors of toxicity
Anti-TNF related SAE
• TB – monoclonals, ethnicity
• New-onset SLE
• Serious infection – especially in first 6 months
• New-onset psoriasis
• Septic arthritis
• Non-melanoma skin cancer – Infliximab
• Herpes Zoster
• Transfusion reactions
Why collect samples for toxicity
outcomes?
• To target treatment better
• Single studies unlikely to be able to address rare
SAE outcomes
Toxicity studies
• Flucloxacillin – induced hepatotoxicity
– HLA-DRB*5701, OR > 80
• Carbamazepine-induced Stevens Johnson syndrome
– HLA B*1502, OR ~100
• Azathioprine-induced bone marrow suppression
– TPMT gene polymorphisms
Summary of toxicity predictors
• No studies yet undertaken
• Will require international collaboration to achieve
sample sizes
• ?Targeted collection based on register information
The way forward
• Larger sample sizes
– Collaboration
• Prospective studies
– Account for confounders
– Test other types of predictors
– Anti-drug antibodies
• Combined algorithm
Biobanks – how to collect
When to collect?
• Ideally prospectively
– Prior to biologic administration
– Collect multiple sample types
– Collect detailed clinical data
• Allows development of ‘biological response
signatures’
Epigenetics in treatment response
• Ideal for studies of treatment response
– DNA methylation relatively stable
– Amenable to whole genome approaches
– Baseline status / change in status
Transcriptomic studies
• Expression studies used to identify response
predictors in breast cancer
– Tamoxifen: ER expression
– Herceptin:
• IFN gene expression signature reported as predictive
of response to RTX
11
.52
2.5
33
.5
Norm
alis
ed
CD
11c E
xpre
ssio
n
Response
Relative CD11c Expression in Non-Responders and Responders
Non-Responders Responders
Whole genome microarrays
RNA-seq
Samples
• DNA – EDTA
– Allows genetic and methylation studies
• RNA
– Paxgene vs tempus
• Serum
– Proteomics / autoantibodies / metabolomics
– If postal collection, only stable markers can be
analysed
• ?urine / faecal collection (microbiome)
• Postal collection
• Protocols available (email
Costs
• Tempus - £2.50
• EDTA – 9p
• Serum – 9p
• Plastic holders – 70p
• Blood box - £1
• Labels – 7p
• A4 envelopes – 2p
• Pre-payment for postage – 18p
Acknowledgements
• Dr. Darren Plant
• Biologics in Rheumatoid Arthritis Genetics &
Genomics Study Syndicate:
– Prof. Ann Morgan
– Prof. Anthony G Wilson
– Prof. John Isaacs
– Dr. Kimme Hyrich
www.medicine.manchester.ac.uk/arc/BRAGGSS/