1. Stratified medicine in rheumatoid arthritis: progress and the challenge of implementation
Anne Barton
Professor of Rheumatology
Director, Centre for Musculoskeletal Research 1

2. Nomenclature Personalized – to the individual
Stratified – by groups of patients
Stratified disease
Anti-CCP + vs Anti-CCP –
Stratified medicine
By response to treatment

3. Rheumatoid Arthritis >400,000 patients in UK
Direct medical costs to NHS ~£560M/year
28% patients stop work within 2 years

4. Early effective therapy prevents damage and disability

5. Treatment of Rheumatoid Arthritis
DMARDs - methotrexate
Biologics
Anti-TNFs IL6 inhibitors Anti-CD20
Etanercept Tocilizumab Rituximab
Infliximab
Adalimumab T cell co-stimulation
Certolizumab Abatacept
Golimumab
Biosimilars
The introduction of biologic drug therapies targeting specific components of the inflammatory response represents a huge advance in the treatment of rheumatoid arthritis (RA).
Despite this, up to 40% of patients fail to respond well to these therapies.
Ideally, clinicians would like to identify patients who are likely to respond to therapy as early as possible in the disease course, making identification of reliable biomarkers of response an important area of research.
(We are initially investigating Etanercept due to its’ widespread use and the availability of samples in the cohort we are using.)

6. Rheumatoid arthritis treatment pathway
40% failure
20% failure
Methotrexate + 1 other DMARD
Rituximab
Anti-TNF
time
Toxicity, disability
Quality of life 6

7. Predicting treatment response
Anti-TNFs in top 5 spend in NHS Trusts
£5-10K per patient per year
Biosimilars emerging
~1/3rd cheaper
Methotrexate
Remains cornerstone
Inexpensive (<£100/yr)

8. Hypothesis Transcriptome Epigenetic Adherence Genetic
Treatment Response
Clinical

9. Biologics in Rheumatoid Arthritis Genetics and Genomics Study Syndicate
Large nationwide multi-centre collaboration
Recruited patients registered with BSRBR-RA
Prospective observational study, not RCT
Real world patients
DNA, RNA, serum, clinical longitudinal data collection
Pre-treatment,
3, 6, 12 months post-treatment
So, in order to identify such genetic factors, the Biologics in Rheumatoid Arthritis Genetics and Genomics Study Syndicate was set up.
The aim of BRAGGSS is to investigate genetic predictors of response to anti-TNF therapy.
BRAGGSS is a large nationwide multi-centre collaboration, currently recruiting from 25 centres across the country.
We have established collaborations with individual consultants around the country who allowed us to have access to information from the BSR Biologics Register for their patients. We then invited those patients to participate in this separate study.
The target of BRAGGSS is to recruit 4,000 RA patients treated with Etanercept, Infliximab or Adalimumab. We are still actively recruiting participants, and to date, >1700 patients have agreed to take part. 9

10. Genetic factors

11. Genetic predictors of anti-TNF response
Why genetics?
Genes don’t change – can test once
Cause, not effect of treatment
Cheap to test
Very robust, reliable and consistent testing methods
In other conditions, genetic predictors found
BUT…..
None consistently found to predict anti-TNF response

12. Challenges of stratified medicine
1. Identifying predictors of response
The outcome measure of response
Adherence as a confounder

13. Outcome measure

14. Rheumatoid arthritis: DAS28
28 joints: swollen joint count, tender joint count
ESR / CRP
Patient overall assessment (VAS)
Validated measure used widely in Europe
NICE guidance

15. Psychological factors
Cordingley et al (2012):
TJC and VAS correlate with psychological factors more than SJC or ESR/CRP
Depression and anxiety scores

16. Gaming the DAS28 In UK, need DAS28 >5.1 to get on biologic
“…I think most people lie actually [about DAS28 scores]…most people make it up…the problem for…the [biologics] registry is that people make up the numbers…to keep the CCG happy…but then give those spurious numbers to the registry”. (Participant A, p.12).

17. Endotype of treatment response?
Treatments developed to treat synovitis
Poor correlation:
DAS28 r2 0.3
CRP r2 0.5
Need biological endotype
HbA1c for diabetes

18. Adherence

19. Assessment of adherence (n=390)
Adherence to Anti-TNF
When you were last due to take your biologic injection, did you take it:
day agreed with the nurse?
day before or after
within a week
more than a week
not at all
Assessment of adherence (n=390)
Adherent
Non-adherent
Characteristic
β - coefficient (95% CI)
P-value
Disease duration
-0.07 (-0.02 – 0.01)
0.448
Age
0.02 (0.00 – 0.04)
0.012
Female gender
0.34 (-0.08 – 0.76)
0.108
NSAID usage
-0.13 (-0.50 – 0.25)
0.500
Marital status
-0.32 (-0.74 – 0.11)
0.148
Ever non-adherent status
0.53 (0.12 – 0.95)
0.013
Bluett et al, 2015

20. Progress in identifying clinically useful stratifiers

21. MAximising Therapeutic Utility for Rheumatoid Arthritis
MATURA

22. Workstream 1 Workstream 2 9 industry partners
Large scale, blood based screening from observational studies
Synovial tissue sampling:
Pathobiology from RCT
9 industry partners

23. MRC/ARUK-Funded MATURA DMARD Inadequate Response
Workstream 1
MRC/ARUK-Funded MATURA
Stratification of Therapy for RA by Pathobiology (STRAP)
DMARD Inadequate Response
Synovial Bx
Prof Cos Pitzalis
Rituximab
Anti-TNF
Tocilizumab
Pitzalis et al. Curr Op Rheum 2013
Pitzalis et al. Nat Rev Immunol 2013

24. Work Stream 2 Blood based sampling Genetics
Anti TNF samples
Methotrexate samples
Rituximab samples
Genetics
Epigenetics (partial funding)
Proteomics (pilot funding)
Transcriptomics (unfunded)
Integrated Analysis
The work stream 2 programme is based around blood sampling, processing and subsequent analysis integrated with the “omic” cross cutting themes. We have existing cohorts of samples and are planning new collections of samples which will then be analysed and integrated to answer the main scientific questions from our analysis plan, which you will hear about from Paul this afternoon.

25. Drug Levels and
anti-drug antibodies

26. Random drug levels

27. Regression coefficient (95% CI)
Variable
Regression coefficient (95% CI)
P value
Adalimumab patients
Univariate analysis
Adalimumab level
0.08 ( )
<0.0001
Anti-drug antibody status
-0.8 (-1.2 to -0.3)
0.002
Multivariate model*
Adalimumab level
0.06 ( )
0.009
Anti-drug antibody status
-0.2 ( )
0.45
* Adjustment for age, gender, BMI, disease duration and adherence
Etanercept patients
Etanercept drug level
0.008 ( )
0.5

28. Predictors of drug levels
Jani et al, 2015

29. Challenges of stratified medicine
2. Proving clinical usefulness to NHS
Efficacy:
So far, have shown correlation, not causation
Will testing drug levels / anti-drug antibodies and changing treatments according to results benefit patients?
Cost effectiveness

30. Efficacy: Results fed back to clinicians, n = 45 Standard care, n = 45
Switch drug
Increase MTX
Continue
Efficacy:
Results fed back to clinicians, n = 45
Patients about to start Adalimumab, recruited
Standard care, n = 45
Assess eligibility,
PIS
USS – confirm synovitis
Randomisation.
Baseline clinical Q, blood samples
Clinical Q, blood samples
Clinical Q, blood samples
Clinical Q, blood samples
Outcome assessed: DAS28
EQ5D, HAQ
Time (months)

31. Cost-effectiveness Test needs to either:
Result in better outcome for same cost OR
Achieve cost savings with minimal adverse impact on outcome
Cost of tests ~£150 / patient
Jani et al, 2016

32. Health Economic modelling
Direct health care costs – NHS payer perspective
Model relies on assumptions
Sensitivity / specificity of test used
Percentage who will flare if treatment continues when tests suggests should switch
Impact of uncontrolled disease activity on long-term disability
Testing strategy / frequency
Comparator strategy

33. Is it cost-effective to test drug levels and anti-drug antibodies?
Probably in Crohn’s disease
Avoid surgery
Less certain in RA
Depends on assumptions made in the model
May need formal testing – clinical trial
Biosimilars
Adalimumab soon to come off patent
Biosimilars should have same immunogenicity profile
Model should still be valid
Gavan et al, unpublished

34. Acknowledgements Ann Morgan John Isaacs Gerry Wilson Kimme Hyrich
Cos Pitzalis and co-investigators
Darren Plant Jane Worthington
Katherine Payne Wendy Thomson
Sean Gavan Steve Eyre
Meghna Jani James Bluett