Principles of Signal Detection & Risk
Management in Pharmacovigilance
Dr Pipasha Biswas
MD MFPM DM MRQA
Principal Consultant & Director
Symogen Limited, UK.
ISPRT & SOPI Conference, Lady Hardinge Medical College, New Delhi
26 – 28 November 2010
Topics Covered in This Session
• What is Signal Detection?
• Methodologies in Signal Detection.
• Usefulness of Signal Detection in
Pharmacovigilance
• How is a Signal Identified and what is done
next
• What is Risk management
• Components of an RMP
Introduction
• Populations exposed to any drug during post marketing
period vary vastly than those studied during the
development of the compound
• In daily practice – real world situation
- Patients are not selected
- Diverse patient populations
- large patient population
- Polypharmacy
- Several underlying disease factors and concomitant
medications
Introduction
• New information on the benefits and risks of
any drug may be generated at any time after
marketing
• Continuously monitor the safety of the
compound/drug throughout the life cycle
product
• Continuously assess the benefit risk profile in
order to guarantee patient safety
What is a Signal?
• Reported information on a possible causal
relationship between an adverse event and a
drug, the relationship being unknown or
incompletely documented previously.
WHO Definition
• Hypothesis generating
• Does not establish any causal relationship
between the drug and the event
• Suggests further studies
• Should be quick and credible
Other Definitions of Signal
• Early identification of suspected signals with
subsequent generation of hypothesis
• Hypothesis of a signal can be generated by
identification of:
- unexpected SAE (also non-serious) or change in its
severity
- Increases in reporting frequency of an expected event
- AEs experienced in special population groups (i.e.
Paediatric, elderly, hepato-renal compromised patients
etc)
SIGNAL:
New Safety Information
• A new Signal
- Unidentified
- Unlabelled
• New information from an existing signal
- Change in frequency
- Change in severity
• Information on risk factors
Sources of Signals
• Premarketing Trials Database
• Spontaneous Reports of ADRs
- Individual company database
- AERS database
- WHO-UMC
- EMEA
- National & Regional RA database
• Published Literature
• Epidemiology/Registries
• Automated Databases
- GPRD
- IMS
- Medicaid
- Kaiser Permanante
• Others
- PEM
- MEMO
Signal Detection Process Flow
Detection of a Signal
Generation of Hypothesis
Verification
Decision to be taken
Information
Factors to Consider in Signal Detection
Considerations for when a Drug-Event Pair is a Signal
Patient
Characteristics
Trends
Pharmacological
Plausibility
SIGNAL
Drug Event Pair
Related Similar
Events
Class
Effect
DPA
Over
Time
DPA
Score
Commonly drugRelated event
Evaluating Signals Using Evidence
Hierarchy
Signal Detection Process
Signal Identified
Clinical
Trial
Data
(ISS, ISO)
Epidemiology
Information Gathering
Literature
Spontaneous
Data
Including
DPA Info
Other
Databases
(AERS,
WHO)
Pre-Clinical
animal
Toxicology &
Pharmacology
Methodologies in Signal Detection
• Several methodologies are used and available for the
purpose of signal detection activities
• Quantitative
• Qualitative
• RA, Drug monitoring Centres and Pharmaceutical
Companies have developed computerised data mining
methods for the purpose of early identification of
safety signals in spontaneous reporting databases
• To date there is no guideline or standard method for
performing signal detection activities
Data Quality
• Data Quality of all case reports entered into
the safety database is essential in order to
retrieve and extract the correct safety
information
• Completeness and correctness of the ICSR
information is the key to good signal detection
• Quality check of entered data
How Signal detection is done?
• Qualitative Analysis:
- Signal case/multiple report evaluation
Combined with or without
• Qualitative Analysis
- Automated identification of signals
Qualitative Analysis
• Case by case review and thorough assessment by
medically qualified persons
• Systematic review of multiple case reports
- review of cummulative data
- review of frequency trends over time and frequency
rates to specific PTs and/or SMQs and/or HLTs or even
SOCs with combined retrospective analysis using
computerised tools
• Pre-clinical data, scientific support documentation,
characteristics of patient population exposed,
pharmacological plausibility etc are evaluated in detail
Quantitative Analysis
• Various automated statistical methods used in
analysis of safety data
• Most commonly used are;
- Proportional reporting Ratio (PRR): MHRA
- Reporting Odd’s Ratio (ROR): LAREB
- Multi-item Gama Poisson Shrinker (MGPS): FDA
- Bayesian Confidence Propogation Neural
Network (BCPNN): UMC
Quantitative analysis
• The key main concept of such statistical
method is “Disproportion or more than what
is Expected”.
Statistical Tools Used for Signal Detection
• All measures are calculated from a 2x2 table
- Proportional Rate Ratio (PRR)
- Reporting Odds Ratio (ROR)
- Relative Reporting Ratio (RRR)
- Information Component (IC; Bayesian)
Statistical Tools Used for Signal Detection
• All measures are calculated from a 2x2 table
- Proportional Rate Ratio (PRR)
- Reporting Odds Ratio (ROR)
- Relative Reporting Ratio (RRR)
- Information Component (IC; Bayesian)
Event
(R)
All Other
Events
TOTAL
Medicinal Product (P)
A
B
A+B
All other medicinal
Products
C
D
C+D
TOTAL
A+C
B+D
N=A+B+
C+D
PRR
• PRR is the ratio of the number of reports of:
• Event of interest for drug of interest Event of
interest for all other drugs/All events for drug
of interest All events for all other drugs
• PRR > 1 - positive quantitative association
between the drug and the event of interest.
Multi-Item Gamma Poison Shrinker (MGPS)
Adjusted relative
Reporting ratio
(After modelling)
Modified
Reporting
Ratio
Reporting Ratio
Stratification
(e.g. gender,
Age, year)
EB05
Bayesian shrinkage
To address small cell sizes
Empiric Bayesian
Geometric Mean
EBGM
EB95
DuMouchel W, Pregibon D. Emperical bayes screening for multi-item associations. Proceedings of the
Conference on knowledge discovery and data; 2001 Aug 26-29; San Diego (CA): ACM Press: 67-76.
Empirical Bayes Geometric Mean (EBGM)
• EBGM is an Observed/Expected score that is the output of
the MGPS method
• EB05 and EB95 are the lower and upper limits of the 2-sided
90% CI around EBGM
• Interpretation:
- If EBGM = 7.2 for paracetamol-hepatic failure, then this
drug-event combination occurred in the data set 7.2 times
more frequently than expected
- If EB05 + 4, then the drug-event occurred at least 4 times
more frequently in the data set than expected
• Thresholds used for Data Mining:
- EB05 ≥ 2 will flag drug-event combinations that occur at
least twice as often expected
Measures Used for Signal Detection
• All measures of SDR are basically calculations of
Observed/Expected event/drug reports
• Since Expected Data originates from the same pool
as the Observed Data – cannot use a PRR as an RR
nor a ROR as a OR
• Expected Data in epidemiology comes from sources
other than the Observed
• Expected Data in PV also referred to as “Background”
Measures Used for Signal Detection
• Since the calculation is O/E, the relationship
between background and statistic of interest is
inversely related
- As background increases resulting statistic
decreases (large E results in small PRR)
- As background decreases resulting statistic
increases (Small E results in large PRR)
What does the MAH do when a signal
has be confirmed and strengthened?
Variation of CCSI/SPC/PIL
PASS
Further to validate a signal
Update of Risk management Plan
Presentation of the Signal in the
PSUR with planned future actions
Provision of the safety
information to HCP and/or
patients
Principles of Risk Management in
Pharmacovigilance
28
29
Hrithik Roshan Suffered Drug Allergy
• By SUNIL Sonkar November 22 — Bollywood actor Hrithik Roshan was
hospitalised on Sunday to Kokilaben Dhirubhai Ambani Hospital in
Mumbai after his lungs practically collapsed. Hrithik was unable to breathe
and his lips became ten times bigger and face was also swollen. The
problem started when he took an antibiotic after complain of chest
infection, which led to strong allergy. He took the tablet on Saturday.
Though he was discharged in the evening, but was kept under
observation. He said, “It was an allergic reaction to the antibiotics I was
taking. Lungs and throat went into spasm. I reached hospital just in
time…All cool now. I’m under observationIt was an allergic reaction to the
antibiotics I was taking. Lungs and throat went into spasm. I reached
hospital just in time…All cool now. I’m under observation.”
• Hrithik also revealed later that a delay of about 15-minute would make
the condition fatal.
• He complained of temperature, headache, backache, muslce pain and
tiredness earlier.
• Family members of the star are quite shocked and also scared over his
condition
The Bar is Being Raised Across the Industry for
Formal Risk Management Planning
• While approval times are decreasing, industry has experienced high
profile withdrawals in recent times
• Product Safety is under increasing scrutiny
- Patients
- Prescribers
- Regulators
- Auditors
- Media
- Legal
• Industry is moving quickly to design and implement Risk Management
processes
• Risk Management Programs can enable challenging products to stay
on the market (by supporting the use of appropriate products by
appropriate patients)
What is Risk Management?
• The activities and interventions deployed to a drug,
in order to manage and mitigate known and possible
risks, with the aim of protecting the individual
• Identification and implementation of strategies to
reduce risk to individuals & populations
• A continuous process of minimising a product’s risks
throughout its life cycle in order to optimise that
product’s risk/benefit balance
Why Manage Risk Proactively?
• Regulatory Expectation
- US, Europe, ICH E2E
• Company Perspective
- to understand the risk profile
- to protect the company’s asset
• Patient perception
- expect safe and effective drugs
- do not fully understand risks
• Need to change prescribing behaviour: labelling
not always sufficient
Risk Management – A Shift in Emphasis
“New Model”
Modify in the light of new safety data
Pre-marketing
Risk Assessment





Traditional analyses
plus
Anticipated conditions
of use
Intrinsic/extrinsic risks
(identified and
potential)
Epidemiology of
disease
Benefit : risk
assessment
ISS
Safety Specification
Approval
Pharmacovigilance
Plan
Risk Management
Implementation

± Risk Minimization
Plan/ Risk Map


Enhanced PMS/
Communication
activities
Active influence on
safe use in the
market place
Assessment of RM
programme
effectiveness
New data
34
Overall Objectives of Risk Management Planning
Benefit - Risk Optimization
35
Optimizing Benefit Risk
High
Unacceptable Risk
Risk
Acceptable Risk
Low
Low
Benefit
High
36
Risk Management Strategy
• Product Risk Management Plan
Plan identifying the risks associated with a medicinal product,
methods to further clarify the safety profile and ways to
minimise risk to individual patients in clinical use
• Three elements
Pharmacovigilance specification
Pharmacovigilance Plan
Risk Minimisation “toolkit”
Risk Management Definition
Risk Management
=
Risk Assessment
+
Risk Minimization
Basic Components of a Risk Management Plan
Risk Management Plan
Safety Specification
Summary of important identified risks,
important potential risks and missing
information
(ICH E2E)
Pharmacovigilance Plan
b
Based on safety specification; Routine PV practices and
action plan to investigate specific safety concerns
(ICH E2E)
Risk Minimization
Activities to be taken to minimize the impact of
specific safety concerns on the benefit-risk
balance
39
RMP Elements
• Identified pre-clinical concerns
• Missing Pre-clinical data
• ADRs in clinical trials (including seriousness and
predictability)
• Potential ADRs requiring further evaluation to clarify
a risk hypothesis
• Population not studied in the pre-approval phase
• Documented interactions
• Potential for unidentified interactions that may occur
post-approval
• Disease epidemiology
• Class effects
Pharmacovigilance Specification
• A structured method of documenting the
established risks of a drug and the
potential for unidentified risks at the
time of marketing authorisation
Safety Specification Purpose
The Safety Specification should be a summary of the
-
[important] identified risks of a medicinal product,
-
[important] potential risks,
-
important missing information.
-
populations potentially at risk
-
outstanding safety questions which warrant further investigation
-
identifies any need for specific data collection
to refine understanding of the benefit-risk profile during the postauthorisation period.
Safety Specification Content
Non-clinical and Clinical Part of the Safety Specification
Limitations of the Human Safety Database
Populations not studied in the Pre-Authorisation Phase
Adverse Events/Adverse Reactions
Identified risks that require further evaluation
Potential risks that require further evaluation
Presentation of risk data
Identified and Potential Interactions
Epidemiology
Pharmacological Class Effects
Additional EU Requirements
Potential for overdose
Potential for transmission of infectious agents
Potential for misuse for illegal purposes
Potential for off-label use
Potential for off-label paediatric use
Risk Management Plan
Purpose
Assessing risks by focused evaluation to close gaps in knowledge
systematically (PM commitments - continued development - targeted
populations)
- looking for potential risks (class effects)
- following observed events
- characterizing outcomes that are mulifactorial
Advance planning and communication of evaluation for new products
Method
Integration of incremental data acquisition starting in development, systemizing
postmarketing commitments and new indication projects for the newly
released compound
Continued integration of all available data requires start at phase 1
Limitations of human safety database
Table x: Exposure by baseline disease
No of patients
Total ( male/female )
Diabetic nephropathy
Hypertensive nephropathy
65 (39/26)
71 ( 47/24)
Glomerulonephritis
207 (143/64)
Other
246 (140/106)
Table y: Special population exposure
Population
Children (<12 years)
Elderly (>75 years)
Pregnant or lactating women
Relevant co-morbidities
•Hepatic impairment
•Cardiac disease
Genetic polymorphism
Ethnic origin
•Caucasian
•other
Number of patients
None
14
None
57
243
….
Not applicable
584
5
45
Toxic Epidermal
Necrolysis (TEN)
Clearly a serious and
important risk
46
Risk Minimisation “Tool Kit”
• Based on Specification
• Focus on practical ways to reduce risk to the
population
- SPC and labelling – population, indication(s),
warnings, contradictions and monitoring
- Communication to healthcare professionals and the
public both pre and post launch, including letters,
advertisements and educational programmes
- Control of distribution or prescribing
- Treatment protocols
Risk Mitigation
How to bring the risk profile to the prescriber and the patient
• Routine label activities
• Outside communication (DDL, Training)
• Tests associated with precriptions
• Market withdrawal
Drugs can safely stay in the market by targeting the right patient
groups through a coordinated safety and marketing strategy
where revenue expectations that are consistent with what the
safety profile supports
Good decision making in drug lifecycle management must be the
objective of strategic pharmacovigilance in implementing the
recognized risk/benefit
Risk Management Strategies
• Reduce drug exposure
- restrict indication
- controlled drug distribution
- optimise dosage regimen
• Modify ADR occurrence
- screen patients at baseline
- monitor precursors (signs & symptoms)
- educate prescribers and patients
- ask for informed consent
- introduce independent patient monitoring
- provide hotlines for medical advice
Safety Communications - A Patient Perspective
“Wonder Pills”
Sir, My wife has been prescribed pills. According to the accompanying leaflet,
possible side-effects are: sickness, diarrhoea, indigestion, loss of appetite,
belching, vertigo, abdominal cramps, dizziness, stomach ulcers, bleeding from
intestine or blood diarrhoea, ulcerative colitis, sore mouth and tongue,
constipation, back pains, inflammation of pancreas, mouth ulcers, skin rashes,
hair loss, sensitivity to sunlight, drowsiness, tiredness, impaired hearing,
difficulty with sleeping, seizures, irritability, anxiety, depression, mood changes,
tremor, memory disturbances, disorientation, changes in vision, ringing in ears,
bad dreams, taste alteration, allergic reactions, swelling due to water retention,
palpitations, impotence or tightness of the chest.
Should she take them?
Yours faithfully,
A D. O,
Hertfordshire.
Letter to the Editor,1996
Information withheld due to
data privacy
50
Conclusion
• Signal detection and Risk Management is a fascinating field,
which is now more often used in PV in majority of the MNCs
• Databases provide a powerful tool for early detection of
safety signals
• ICSR & domain knowledge (epidemiology & natural history of
disease, medical practice, pharmacology of drug) are
essential for assessing causality between a drug & an event
• Risk management should start early during clinical drug
development process
• Risk Management Plans should be for all drugs and submitted
to RA
Risk Management - Conclusion
Embracing Change
“It is not the strongest of the species that survive,
nor the most intelligent, but the one most responsive
to change.”
Charles Darwin, 1859
52
Thank You!
pippa@symogenlimited.com
pipasha.biswas@gmail.com
www.symogenlimited.com