Transcript The role of Signal Detection and Signal Management x

The role of
Signal Detection
and Signal Management
November 2015 Belgrade
Dr Sabine Straus, Medicine Evaluation Board,
The Netherlands
0
19-11-2015 Belgrade
Regulatory framework
Regulation (EU) 1235/2010
National legislation
Directive 2010/84/EU
EC Implementing Regulations (binding)
Good Pharmacovigilance Practice
Questions & Answers
1
19-11-2015 Belgrade
1
Historically, spontaneous reports have been
the main source of postmarketing surveillance.
•
Arnaiz JA, Carne X, Riba N, et al. The use of evidence in pharmacovigilance: case reports
as th reference source for drug withdrawals.
Eur J Clin Pharmacol 2001; 57 (1): 89-91
• Lasser KE, Allen PD, Woolhandler SJ, et al. Timing of new black box warnings and
withdrawals for prescription medications.
JAMA 2002; 287 (17): 2215-20
• Clarke A, Deeks JJ, Shakir SA. An assessment of the publicly disseminated evidence of
safety used in decisions to withdraw medicinal products from the UK and US markets.
Drug Saf 2006; 29 (2): 175-81
• Olivier P, Montastruc JL. The nature of the scientific evi- dence leading to drug withdrawals
for pharmacovigilance reasons in France.
Pharmacoepidemiol Drug Saf 2006; 15 (11): 808-12
2
19-11-2015 Belgrade
3
Archives of Internal Medicine 2012
19-11-2015 Belgrade
•
‘signal detection’ is recognised in law
Dir. Section 3: Signal Detection
• Responsibility for EMA and Member States (D107h)
monitor the data in the EudraVigilance database to determine
whether there are new risks or whether risks have changed and
whether those risks impact on the risk benefit balance’
• Responsibility for Marketing Authorisation Holder (D104.3)
monitor pharmacovigilance data to determine whether there are
new risks or whether risks have changed or whether there are
changes to the benefit risk balance
• Elaborated in ‘Implementing Measures’
4
19-11-2015 Belgrade
Spontaneous reporting databases - Strengths
- Inexpensive and simple to operate
– Large scale
• Cover all drugs during their whole life cycle
• Cover the whole patient population, including special subgroups
– Clinical evaluation of reporting healthcare professional
– Does not interfere with prescribing habits
– Can be used for follow-up studies of patients with severe ADRs, to study
mechanisms
– Generation of hypotheses and signals
Spontaneous reporting databases - Weaknesses
– Limited clinical information to permit thorough case evaluation / report
quality
– Underreporting decreases sensitivity and makes the systems prone to
selective reporting
– The reporting rate is seldom stable over time (e.g. Weber effect)
– No direct information on incidence rates
Brian L. Strom and Stephen E. Kimmel.
Textbook of Pharmacoepidemiology. 2006. Chapters 7 & 8.
5
19-11-2015 Belgrade
Signal detection
•
Qualitative
– Manual
– Case by Case evaluation
•
Quantitative
– Automatic method – Data mining
– Measures of disproportionality
6
19-11-2015 Belgrade
Quantitative signal detection
•
•
•
“Do we observe what we expect?”
Key = disproportionality
Ed Napke (1968)
– Pigeon hole system
•
Computerised (dynamic field many dev )
– Data mining
– First methods proposed in 70s
– More and more used from 90s on
• Large databases EHR screening :
– Sentinel
– EU-ADR
– OMOP
7
19-11-2015 Belgrade
Assumptions of disproportionality analysis
• If drug causes a certain ADR, this ADR is reported more
frequently for this drug compared to drugs that do not
cause this ADR
• Non-differential reporting for an ADR
• Is the overall reporting rate a valid reference to compare
to?
• Background data set impacts the disproportionality results
– Size
– Heterogeneity
– Type of adverse events represented
CIOMS VIII: Practical Aspects of Signal
2010
Detection
in Pharmacovigilance.
8
19-11-2015 Belgrade
Event +
Measures of Disproportionality
•
ROR – Reporting Odds Ratio
– analogous to an odds ratio
•
PRR – Proportional Reporting Ratio
Event -
Drug
+
Drug -
– analogous to rate ratio
•
IC – Information Component
– Bayesian statistics
• IC025 = lower limit 95% confidence interval IC
• Be aware of logarithmic scale when interpretating the information
component!
•
All methods based on 2 x 2 table
9
19-11-2015 Belgrade
Formulae in two-by-two table
Event +
Event -
Drug +
A
B
A+B
Drug -
C
D
C+D
A+C
B+D
A+B+C+D
ROR = (A*D)/ (C*B)
PRR = (A/(A+B))
/ (C/(C+D))
IC= log 2 [ (A+½)
/ ((((A+B)*(A+C))/(A+B+C+D)+½) ]
10
Hypothesis:
• I have some kind of feeling that the
combination of the slogan “Yes, we can!”
is highly connected to Barack Obama
Study:
• Simple research on the internet – and
apply the formulas to substantiate the
hypothesis – or refute it
* adapted from http://www.signaldetection.net/
11
19-11-2015 Belgrade
Webpages containing
‘Obama’
Webpages containing
‘Yes we can’
Other webpages
Other webpages
A
B
2,140,000
2,310,000
C
D
147,860,000
21,727,690,000
ROR
= (ad)/(bc)
PRR
IC
ROR
=
=[A/(A+B)]
136.1
lower limit/ of
[C/(C+D)]
confidence
interval of IC
025=
ROR
= (2,140,000*21,727,690,000)
/
IC ==134.0
PRR
(2,140,000/4,450,000)
6.13 (6.13all
– 6.13)
3 highly
significant
/ (147,860,000/21,875,550,000)
(2,310,000*147,860,000)
PRR
IC
=134.2
6.13 (134.0 – 134.5)
025=
ROR = 136.1 (135.9 – 136.4)
Adapted from: http://signaldetection.net/2009/08/19/the-prr-ofobama-and-yes-we-can/
12
19-11-2015 Belgrade
Purposes of statistical methods in Pharmacovigilance
– Detecting signals
– Triage among signals in an existing database
– Triage among signals in newly arrived data
• avoiding expensive assessment of all reports
– Preliminary evaluation of signals?
• but NOT– Inferring causality based on PRR etc
– Unthinking application of extreme PRR etc
– Failing to allow for uncertainties & biases
13
19-11-2015 Belgrade
Is reporting unbiased?
•
•
•
•
•
•
•
•
14
Severe ADRs more likely reported
Known reactions less likely
Weber effect
‘Black triangle’ in UK additional monitoring EU
Underreporting: 10% (Inman) UK ADRs reported media attention
Media/ law suits
….
19-11-2015 Belgrade
What should be done?
•
Remember the purposes of data-mining
– focus resources on the really important areas
– Don’t waste time on each and every report
•
Use data-mining as a filtration process
– Simple and rapid filter that focuses attention on only a few of the
important reports
•
•
•
Improve by grouping terms in a sensible way
Improve detection of interactions
Ensure known, serious, problems monitored
•
The methods used in signal detection are dependent on the
databases ( heterogeneity , background incidence etc
15
19-11-2015 Belgrade
Responsibilities
• EMA and National Competent Authorities (NCAs)
shall collaborate to monitor the data available in the
EudraVigilance database
within the terms of the marketing authorisation
as well as outside the terms of the marketing authorisation,
and for medicinal products authorised in the Union that may
induce an occupational exposure.
• The Marketing Authorisation Holders (MAHs)
shall monitor the data
to the extent of their accessibility to the EudraVigilance database,
as part of their broader monitoring of all emerging data and
global signal detection activities.
• MAH shall inform EMA and NCAs in the event of new risks or
risks that have changed or when changes to the risk-benefit
balance have been detected.
16
19-11-2015 Belgrade
Implementing measures
The identification of new risks or changed risks is based
on the detection and analysis of the relevant signals.
A signal
consists of information that arises from one or multiple
sources (including observations and experiments), which
suggest:
•
•
a new potentially causal association, or
a new aspect of a known association,
between an intervention and an event or a set of related
events, either adverse or beneficial that is judged to be of
sufficient likelihood to justify verificatory action.
17
19-11-2015 Belgrade
• NCAs shall specifically monitor data originated in their territory.
• Any signal
 communicated to EMA or NCAs by the MAH
 Validated* to determine whether further analysis is required
Based on multidisciplinary approach
Supported by EV
List of Medical Events to be taken into account
(may be published following
consultation of the PRAC)
– For products authorised in accordance with Directive
2001/83/EC validation shall be done by the NCAs.
– For products authorised in accordance with
No 726/2004 validation shall be done by EMA.
18
19-11-2015 Belgrade
Work sharing of signal management
For medicinal products
 authorised more than one Member State and
 for active substances contained in several medicinal
products where at least one marketing authorisation has
been granted in accordance with Directive 2001/83/EC:
– A lead Member State for the monitoring of data in the
EudraVigilance database will be appointed
All Member States retain their responsibility pursuant to
Article 107h(1)(c) and (3) of Directive 2001/83/EC.
19
19-11-2015 Belgrade
Work sharing of signal management
Lead member state appointment
the coordination group shall take into account
Reference Member State (art 28(1))
lead Member State for PSUR assessment
The EMA shall make public as list of active substances
and of the lead Member State
Risk in cumulating all tasks in one Member State (for the
authorisation, PSUR scrutiny and EudraVigilance
monitoring)?
 Benefits of parallel monitoring lost “ peer review”
 Work sharing also for CAPs
20
19-11-2015 Belgrade
Signal detection support
EMA shall support the monitoring of the EudraVigilance
database by providing access to:
– data outputs and statistical reports allowing a review of
new adverse reactions and of all adverse reactions
reported to EudraVigilance in relation with an active
substance or a medicinal product;
– customised queries supporting the evaluation of
individual case safety reports and case series;
– customised grouping and stratification of data enabling
the identification of patient groups with a higher risk of
occurrence of adverse reactions or with a risk of a more
severe adverse reaction;
– statistical signal detection methods.
21
19-11-2015 Belgrade
Signal detection audit
The Signal Detection process should be fully auditable:
• The NCAs and EMA shall keep an audit trail of their signal detection
activities in EudraVigilance and of the relevant queries and their
outcomes
• The audit trail shall allow traceability of how signals have been
detected and how validated signals have been investigated
• All validated signals shall be entered into a tracking system
administered by EMA
22
19-11-2015 Belgrade
PRAC Mandate
R 61
The Pharmacovigilance Risk Assessment Committee:
• Any validated signal that requires further analysis shall be
transmitted to the PRAC in order to perform the initial
analysis and prioritisation of signals and to consider followup action (R 28, D 107h)
• The PRAC shall perform a regular review of the
methodology to be used and publish recommendations, if
appropriate (IM)
• Following consultation of the PRAC, EMA may publish a list
of medical events that have to be taken into account for the
detection of a signal (IM)
23
19-11-2015 Belgrade
The newly defined role of signal detection
The non serious reports/patient reporting
D 107
• Eudravigilance database
– Single point of entry
• ADRs not only from the authorised use of the product at normal
dose, but also use outside the terms of the MA
– Medication error
– Overdose
– Misuse
– Off label
• Serious and non serious
• Patient reports
Background data set impacts the disproportionality results
Size
Heterogeneity
Type of adverse events represented
24
19-11-2015 Belgrade
Additional monitoring R23

Medicinal products to be subject to additional monitoring:

medicinal products: a new active substance which, on
1 January 2011 not authorised in the Union,

Biological medicinal products and biosimilars (MA
after 1-1-2011)

Medicinal products subject to the obligation

to conduct PAS

conditions regarding their safe and effective use
(e.g. additional risk minimisation activities).

List is publicly available

5 years minimum period to be on the list

EC or MS may extend that period
25
19-11-2015 Belgrade
GVP Module IX – Signal Management
• signal detection
• signal validation and confirmation
• signal analysis and prioritisation
• signal assessment
• recommendation for action
• exchange of information
by EMA, MSs*
by PRAC
all stakeholders
26
19-11-2015 Belgrade
26
Electronic Reaction Monitoring Report (eRMR)
Excel based summary data
• Contain all Drug Event Combinations for a substance
• Counts against different criteria, eg
- literature
- parent-child
- fatal
- source (Patient, Health Care Professional, Study..)
- route of administration
- pediatric, geriatric
- abuse, medication error
• Highlighted Signals of Disproportionate Reporting
27
19-11-2015 Belgrade
Periodicity EudraVigilance monitoring
Frequency proportionate to:
• identified risk
• potential risks
• need for additional information
Baseline periodicity: once monthly ( under discussion )
2 week frequency:
• products under additional monitoring
• products with need for additional information
More frequent in specific situations e.g.
Pandemic situation,
Targeted specific safety issues
28
19-11-2015 Belgrade
Signal management-validation
29
19-11-2015 Belgrade
Signal validation [GVP IX.B.3.3]
How to validate
signals?
Clinical judgment based on review of
ICSRs is a key element
Aspects to be considered to decide if a
signal is “validated”
Strength of evidence
Clinical relevance
Previous awareness
Other data / set(s)
severity of the reaction and its outcome
novelty of the reaction (e.g. new and
serious adverse reaction)
drug-drug interactions
reactions occurring in special populations
30
19-11-2015 Belgrade
IR art 21
Confirmation ( in EPITT
indicates need for discussion at PRAC
not a scientific confirmation
Responsibility
AGENCY/Raporteur CAPS
(Lead) MEMBER STATE
Signal concerns a product authorised in
accordance with Regulation (EC) No
726/2004
Signal concerns a product authorised in
accordance with Directive 2001/83/EC
31
19-11-2015 Belgrade
Communication MAHs
PRAC signals and recommendations
Summary of current approach to informing MAHs on
signals for NAPs
• Agendas of PRAC with list of confirmed signals published
MAHs informed before the agenda is publicly available
• EMA sends ‘outcome fax’ to MAHs ( AR and recommendation)
• Conclusions of the PRAC following the
assessment/prioristisation of signals in minutes – published
month after the PRAC plenary
• Excel list published at EMA side with all recommendations
32
19-11-2015 Belgrade
32
Communication MAHs-PRAC
• Recommendation leads to SPC/PL update:
how to implement?
Smooth implementation of PRAC recommendations due to
- clear proposals for text in SPC and PL provided in
recommendations
- AR explaining the rationale/evidence
avaliable for MAHs
- Translations provided
- timelines implementation available
- clear agreement on process: all products containing the
same active substance (concurrent)
33
19-11-2015 Belgrade
• EMA publishes translations (in 22* EU languages) of PRAC
recommendations for PI update first publication in Feb 15
Benefits
• Recommended translations of PI wording provided centrally by
the EMA
• Harmonisation of PI wording in all EU languages
• Decrease of workload for MAHs
• Simplification of assessment of variations to update PI, with
decrease in workload for network
• Enhanced quality
• Concurrent submissions
* Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France,
Germany, Greece, Hungary, Italy, Lithuania, Latvia, Malta, Netherlands,
Poland, Portugal, Romania, Slovak Republic, Slovenia, Spain, Sweden
34