What is known and what is missing:
Safety
1st EULAR RODS Meeting
Prague
October 14-15, 2013
Angela Zink
Programmbereich Epidemiologie
Deutsches Rheuma-Forschungszentrum Berlin
&
Klinik für Rheumatologie und Klinische Immunologie
Charité Universitätsmedizin Berlin
Uses of registers for safety of medications
Key elements
Choice of comparator
Events of interest
Malignancies
Infections
MI, HF and stroke
Mortality
Conclusion
Outline
Strengths:
long-term safety
unselected patients
real world use, e.g. off-label use
assessment of rare events
risk profiles (e.g. age, co-morbidity, co-medication)
balancing benefits and harms (disease activity vs. toxicity)
Limitations:
Confounding by indication may also affect safety
Limited information in routine setting
Uses of biologics registers for assessment of safety
population controls for hard outcomes (death, cancer,
hospitalization)
disease-specific controls: as comparable as possible
ascertainment of events and clinical data collection in
identical manner for cases and controls
application of methods to control for comfounding by
indication (e.g. propensity scores, marginal structural models)
patients may be their own controls if different treatment
episodes are available
Key is the choice of the comparator!
Selected results on safety
Cardiovascular events
Lymphomas and
solid malignancies
Serious
infections
Death
ARTIS: Lymphoma risk
26 incident lymphomas in ARTIS = 96/100.000 person-years
Cross-linkage with cancer and cause-of-death registers
0 1 2 3 4
vs. General
population
vs. anti-TNF
naive RA
population
2.72 (1.82 – 4.08)
1.35 (0.82 – 2.11)
Relative risk for lymphoma
in anti-TNF treated RA
patients
1998-2001: RR=1.62
2002-2006: RR=0.90
Askling et al., Ann Rheum Dis 2009;68(5):648-53
56
40
83 46
15 (number of cancers)
Incidence rate of cancer among 6,366 Swedish RA patients
starting TNFα therapy from 1999 through 2006 (IR per 1,000 PY)
Askling J et al, Arthritis Rheum 2009;60:3180-89
4
6
8
10
12
14
16
18
<1 1-2 2-4 4-6 6+
Time since start of anti-TNF therapy (years)
Inc
ide
nc
e p
er
10
00
ARTIS: Solid Malignancies
BSRBR: Incident malignancies in non-biologic
DMARD patients compared to the population
Mercer L et al., Rheumatology 2013;52:91-98
Total number Overall SIR
All sites 182 1.3*
Solid cancers 156 1.2*
Oesophagus 5 1.5*
Stomach 6 1.9*
Colorectal 17 1.0
Lung 46 2.4*
Melanoma 9 2.1
Lymphoma 5 3.8*
NHL 16 3.1*
* Significantly different
from population
DANBIO: Cancer incidence
Dreyer L et al. Ann Rheum Dis 2013;72:79-82
7,159 RA patients, follow-up 2000 to 2008
linkage with Central Population Register and Danish Cancer Registry
Cases SIR TNF-i vs. population
SIR nbDMARD vs. population
All cancer sites combined 280 1.3 (1.1 – 1.5) 1.3 (1.1 - 1.5)
Digestive organs 41 1.2 (0.8 - 1.8) 1.0 (0.6 - 1.6)
Colon 17 1.6 (0.9 - 2.8) 0.5 (0.2 - 1.4)
Respiratory organs 41 1.3 (0.9 - 2.1) 1.6 (1.1 - 2.5)
Melanoma 9 1.6 (0.7 - 3.5) 1.0 (0.3 - 3.1)
non-melanoma skin cancer
76 1.9 (1.4 - 2.6) 1.8 (1.3 - 2.5)
Breast 28 0.7 (0.4 - 1.2) 0.9 (0.5 - 1.5)
Non-Hodgkin lymphoma 10 1.9 (0.8 - 4.5) 2.3 (0.9 - 5.5)
Cases SIR TNF-i vs. population
SIR nbDMARD vs. population
HR TNF-i vs. nbDMARD
All cancer sites combined 280 1.3 (1.1 – 1.5) 1.3 (1.1 - 1.5) 1.0 (0.8 - 1.3)
Digestive organs 41 1.2 (0.8 - 1.8) 1.0 (0.6 - 1.6) 1.1 (0.6 - 2.1)
Colon 17 1.6 (0.9 - 2.8) 0.5 (0.2 - 1.4) 3.5 (1.1 - 11.2)
Respiratory organs 41 1.3 (0.9 - 2.1) 1.6 (1.1 - 2.5) 0.8 (0.5 - 1.6)
Melanoma 9 1.6 (0.7 - 3.5) 1.0 (0.3 - 3.1) 1.5 (0.4 - 6.3)
non-melanoma skin cancer
76 1.9 (1.4 - 2.6) 1.8 (1.3 - 2.5) 1.1 (0.7 - 1.8)
Breast 28 0.7 (0.4 - 1.2) 0.9 (0.5 - 1.5) 0.7 (0.4 - 1.6)
Non-Hodgkin lymphoma 10 1.9 (0.8 - 4.5) 2.3 (0.9 - 5.5) 0.6 (0.2 - 2.2)
Dreyer L et al. Ann Rheum Dis 2013;72:79-82
7,159 RA patients, follow-up 2000 to 2008
linkage with Central Population Register and Danish Cancer Registry
DANBIO: Cancer incidence
BSRBR: Non-melanoma skin cancer
3.771 RA patients in DMARD group
15,921 RA patients in biologics group
Mercer L et al., Ann Rheum Dis 2012; 71:869-74
Observed Expected SIR 95%CI
nbDMARD 39 21.3 1.8 1.3-2.5
Men 15 9.0 1.7 0.9-2.8
Women 24 12.4 1.9 1.2-2.9
Anti-TNF 126 73.4 1.7 1.4-2.0
Men 48 23.8 2.0 1.5-2.7
Women 78 49.7 1.6 1.2-2.0
Standardised incidence rate ratio of NHS-IC reported skin cancer (exept melanoma) compared to the general population of 2003-8 (incl. patients with prior skin cancer)
Linkage of ARTIS with registers on RA outpatients, cancer and general
population
10 878 RA Patients exposed to TNF-i
42 198 RA Patients exposed to nbDMARDs
38 malignant melanomas on TNF-i (68/100,000)
113 malignant melanomas on nbDMARD (56/100,000)
162 743 matched
population controls
Raaschou P et al., BMJ 2013;346:f1939
ARTIS: Incidence of malignant melanoma
HR nbDMARD vs.
population
HR anti TNF vs. nbDMARD
Invasive malignant melanoma 1.2 (0.9-1.5) 1.5 (1.0-2.2)
In situ melanoma 1.2 (0.9-1.7) 1.1 (0.5-2.1)
Raaschou P et al., BMJ 2013;346:f1939
Adjusted for sex, age, country of birth, history of NMSC in situ, family history, education, co-morbidities
Increased HR TNF-i vs. nbDMARD restricted to men:
men 2.7 (1.6-4.6)
women 1.2 (0.7-1.9)
ARTIS: Incidence of malignant melanoma
ARTIS: Recurrent malignant melanoma
Increased risk of second melanoma on TNFi
in patients with a history of melanoma:
n=3/54 TNF-i vs. 10/295 nbDMARDs
HR 3.2 (0.8-13.1)
Raaschou P et al., BMJ 2013;346:f1939
BSRBR (1) RABBIT (2)
TNFi DMARDs TNFi DMARDs
N of patients 10,735 3,235 3,260 1,774
N with prior malignancies 177 117 58 55
% with prior malignancies 1.6% 3.6% 1.8% 3.1%
Median years since prior malignancy (t0) 11.5 8.5 4 5
≤ 10 years after last malignancy 42% 61% 77% 73%
Median years of follow-up 3.1 1.9 2.1 2.5
Recurrent malignancies 13 9 8 5
Recurrence rate (per 1.000 PY) 25.3 38.3 45.5 31.4
Recurrence rate ratio TNFi-DMARD 0.58 [0.23-1.43] 1.4 (p=0.6)
(1) Dixon W et al., Arthritis Care Res 2010;62(6):755-63
(2) Strangfeld A et al., Arthritis Res Ther 2010;12(1):R5
Recurrent malignancies in BSRBR and RABBIT
ARTIS: Survival after cancer diagnosis
Raaschou P et al., Arthritis Rheum 2011; 63 (7):1812-22
No difference in stage at presentation or in post-cancer
survival rates between RA patients exposed or unexposed to
TNF inhibitors
What is known and what is missing: Malignancies
No hint on increased risk of
lymphoma from TNF-i in
individual studies
No overall increased risk of
solid tumours under TNF-i
No difference in survival after
cancer diagnosis
Increased risk of NMSC in RA
Increased risk of malignant
melanoma under TNF-i
Limited power of individual
lymphoma studies
Melanoma risk has to be validated
with other data (difference
men/women?)
Other individual tumours, e.g. colon?
Revalidation of the risk of recurrent
tumours under different biologic
agents
In general: Risk in newer agents
Selected results on safety
Cardiovascular events
Lymphomas and
solid malignancies
Serious
infections
Death
RABBIT: Risk of incident or worsening of prevalent
congestive heart failure
After adjustment for traditional CV risk factors:
Increased risk for higher DAS28 at follow-up: HRadj=1.5 [1.1-2.0]
Increased risk for higher doses of GCs (≥10 mg vs. <10 mg): HRadj=3.3 [1.0-11.0]
No increased risk for TNF-i: HRadj=1.2 [0.3-4.7]
Risk for developing or worsening of CHF by status at baseline
Listing J et al. Arthritis Rheum 2008;58:667-77
15%
10%
5%
0% 0 6 12 18 24 30 36
12.5%
2.2%
0.4%
Time (months)
CHF at t0, worsening
CVD disease at t0, new CHF
No CVD disease at t0, new CHF
BSRBR: Myocardial infarction
63 myocardial infarctions per 13,233 PYs in patients treated with
anti-TNF = 4.8/1,000 PYs
17 myocardial infarctions per 2,893 PYs in patients treated with
DMARDs = 5.9/1,000 PYs
No difference between patients on TNF-I and controls
Risk in responders to TNF-I only about one third of the
risk in non-responders.
Dixon W et al., Arthritis Rheum 2007; 56:2905-12
RABBIT: Stroke and disease activity
RABBIT: 10,197 patients, 114 strokes 3,405 patients nbDMARDs, 5,293 TNF-i, 1,499 other biologics Mean follow-up 31-42 months
Strangfeld A. et al., ACR Congress 2013, Abstract # 377
Multivariate Hazard Ratio
Male sex 1.5
Age (by 10 years) 1.6
Smoking (ever) 1.5
Hypertension 1.7
Atrial fibrillation 3.5
CHD 1.3
Diabetes 1.2
DAS28 (per unit) 1.5
TNF-i 1.0
other biologics 0.9
Time to first stroke
What is known and what is missing:
Cardiovascular and cerebrovascular diseases
No increased risk for incident or
worsening of prevalent HF from
TNF-i
No increased risk of MI from TNF-i
Decreased risk of MI in responders
to TNF-i
Risk of stroke depends on disease
activity, irrespective of treatment
Risk in patients with severe
heart failure
Replication of connection
between disease activity
and MI or stroke
CV risk on non-TNF-i
Selected results on safety
Cardiovascular events
Lymphomas and
solid malignancies
Serious
infections
Death
BIOBADASER: Effectiveness of Screening
IRR vs. RA cohort IRR vs. popul.
Gomez-Reino JJ et al., Arthritis Rheum 2003 48: 2122
Carmona L et al., Arthritis Rheum 2005;52:1766-72
1,0 5,0 10,0 15,0 20,0 25,0
Before
screening
recommen-
dation
after
screening
recommen-
dation
34 cases of TB in 7,825 pat-years of exposure (= 0.4 per 100 patient-years)
rate dropped by 83% and reached the EMECAR rate
BSRBR:
Lowest risk for TB on treatment with etanercept
DMARD TNF-i ETA INF ADA
n 3232 10,712 5521 3718 4857
Patient years 7345 28,447 12,744 8069 7634
No. of cases with TB 0 27 5 11 11
Rate / 100,000 (95% CI) 95 (63-138) 39 (13-92) 136 (68-244 ) 144 (72-258)
IRR (adj. by age, sex,
year of entry) Referent 3.1 (1.0-9.5) 4.2 (1.4-12.4)
Dixon W et al., Ann Rheum Dis 2010 Mar; 69(3):522-8
Only patients with at least one treatment switch under observation
(n = 1344 with 41 HZ events)
Adjusted for age, propensity score and treatment with glucocorticoids
Strangfeld et al. JAMA 2009;301:737-44
RABBIT: Different risk for Herpes zoster
Relative risk of serious infection by observation
period
modified from: Askling & Dixon, Curr Opin Rheumatol, 2008;20:138-44
1 2 3
1,0
2,0
3,0
4,0
5,0
= Dixon 2007
= Listing 2005
= Dixon 2006
= Askling 2007
0,8
= Strangfeld 2011
years 0.5
▼
▼ = Galloway 2011 ▼
▼
▼
Cohort level
Decreasing disease activity and improving function in
responders
Decreasing need for glucocorticoids
Patient level
Individual risk
Caused by
Therapy stop in high risk patients
Loss-to follow-up or death in patients at increased risk
Composition of the cohort
The “healthy drug survivor” effect
RABBIT: Risk factors for serious infections
Adjustment for
baseline
characteristics
IRR
Age > 60 1.7
Chronic lung disease 2.6
Chronic renal disease 2.3
FFbH per 10% improvement 0.9
GC 7.5-14 mg/d 1.0
GC >= 15 mg/d 1.5
Treatment with TNF inhibitors 1.6
Trend (IRR TNF-i year 2/year 1) 0.7
Strangfeld et al. Ann Rheum Dis
2011;70(11):1914-20
RABBIT: Risk factors for serious infections
Adjustment for
baseline
characteristics
Fully adjusted
model
IRR IRR
Age > 60 1.7 1.6
Chronic lung disease 2.6 1.7
Chronic renal disease 2.3 1.6
FFbH per 10% improvement 0.9 0.9
GC 7.5-14 mg/d 1.0 2.1
GC >= 15 mg/d 1.5 4.7
Treatment with TNF inhibitors 1.6 1.8
Trend (IRR TNF-i year 2/year 1) 0.7 1.0
Adjusted for baseline characteristics
AND time varying risks
AND dropout processes Strangfeld et al. Ann Rheum Dis
2011;70(11):1914-20
Strangfeld A et al. Ann Rheum Dis 2011;70(11):1914-20
Zink A et al. Ann Rheum Dis 2013, Jun 28 [epub ahead of print]
RABBIT Risk Score: Estimated incidences of serious infection
in 100 patients per year by treatment and risk profile
Risk factors:
• age > 60
• chronic renal or lung disease
• >5 previous treatment failures
• previous serious infection
Calculating the individual infection risks using the
RABBIT Risiko Score: www.biologika-register.de
What is known and what is missing: Infections
Increased risk of serious
infection under TNF-i
Decreasing risk over time is
partly a cohort effect
Risk has to be balanced
against competing risks
(high disease activity, GC)
Differences in risk for Tbc
and HZ for different TNF-i
Risk for other specific infections
in different TNF-i
Infection risk in non-TNF-i
Infection risk in populations
with different baseline risks
Replication of risk score in
different populations
Selected results on safety
Cardiovascular events
Lymphomas and
solid malignancies
Serious
infections
Death
BSRBR: No difference in mortality risk between
TNF-i and nbDMARDs
12,672 patients exposed to TNF-i (50,026 patient-years of exposure) 3,522 patients exposed to nbDMARDs (9,836 patient-years of exposure)
Survival curves, adjusted for baseline differences
Lunt M et al., Arthritis Rheum 2010;62:3145-53
ARTIS: Comparison of mortality risk under
three different TNF inhibitors
6,322 patients exposed to adalimumab, etanercept or infliximab
19,118 person-years of follow-up
211 patient died
Comparison of mortality rates among the three TNFi:
INF vs. ETA: RR = 1.1 [95%CI 0.7-1.7]
ADA vs. ETA: RR = 1.3 [95%CI 0.9-2.0]
Simard JF et al. Arthritis Rheum 2012; 64: 3502-10
BIOBADASER: Mortality risk compared to
external cohort
Carmona L et al, Ann Rheum Dis 2007;66:880-85
Relative Risk BIOBADASER vs EMECAR:
All cause mortality 0.3 [0.2-0.5]
Deaths from:
CV diseases 0.6 [0.2-1.4]
Infection 0.5 [0.2-1.3]
Cancer 0.4 [0.1-1.3]
0 1 2 3 4
DAS28 > 5.1
DAS28 4.1 to 5.1
DAS28 3.2 to 4.1
DAS28 <3.2 Referent
Adjusted hazard ratios and 95%CI
PYRS Adj. HR
6,762 Referent
8,915 1.3
8,793 1.4
7,005 2.4
Mean DAS28
at follow-up
RABBIT: Mortality risk depends on disease activity
Hazard ratios adjusted for general risk factors, glucocorticoid dose, treatment
with DMARDS, or biologics; PYRS: patient-years
Listing J et al. Ann Rheum Dis 2013 (accepted)
PYRS Adj. HR
9,142 Referent
13,592 1.1
7,106 1.5
1,169 2.0
449 3.5
Mean prednisone dose
last 12 months
0 2 4 6 8
> 15mg/d
> 10 to 15mg/d
> 5 to 10mg/d
to 5mg/d
No glucocorticoids Referent
1
Adjusted hazard ratios and 95%CI
RABBIT: Mortality risk depends on prednisone dose
Hazard ratios adjusted for general risk factors , DAS28, functional capacity,
treatment with DMARDS, or biologics; PYRS: patient-years
Listing J et al. Ann Rheum Dis 2013 (accepted)
RABBIT: Lower mortality on biologic DMARDs
Pat-Years Adj. HR
7.012 Reference
16.843 0,64
2.599 0,57
1.654 0,64
0.0 0.5 1.0 1.5 2.0
Other biologics 6 Months
Rituximab 12 Months
TNF-i 6 Months
MTX Reference
Adjusted Hazard Ratios with 95% CI
6 (12) Months risk window
favours biologics favours MTX
Hazard Ratios adjusted for general risk factors, DAS28, glucocorticoids, function
Listing J et al. Ann Rheum Dis 2013 (accepted)
Pat-Years Adj. HR
7.012 Reference
16.843 0,64
2.599 0,57
1.654 0,64
6.469 Reference
22.370 0,77
2.806 0,91
6 (12) Monaths risk window and „ever exposed“
0.0 0.5 1.0 1.5 2.0
Other biologics ever
TNF-i or Rituximab ever
MTX
Other biologics 6 Months
Rituximab 12 Months
TNFi 6 Months
MTX
Adjusted Hazard Ratios with 95% CI
favours biologics favours MTX
Reference
Reference
Hazard Ratios adjusted for general risk factors, DAS28, glucocorticoids, function
Listing J et al. Ann Rheum Dis 2013 (accepted)
RABBIT: Lower mortality on biologic DMARDs
What is known and what is missing: Mortality
Different registers show
equal or lower mortality
under TNF-i and other
biologic agents
There is a beneficial effect
of control of disease activity
There is possibly an
additional beneficial effect
of biologic agents
Explanation of the possible
beneficial effect of biologics
Replication of preliminary
results in other populations
Definition of patient groups
who benefit most from
biologic therapy concerning
mortality
Summary
Registers provide important safety information for patient groups usually
excluded from RCTs (due to comorbidity, age, severe functional
disability ...)
They enable observation of rare events or outcomes with long latency
Registers are able to identify ‚high risk‘ patients and ‚high risk‘
treatment combinations
Registers have impressively shown the importance of disease activity for
several adverse outcomes
Register data enable physicians to balance benefits and risks in individual
patients
Despite a wealth of results from the registers, many open questions remain
Collaboration is required to answer the more complicated questions!