Quantitative signal detection for the mid sized pharma - webcast

Copyright © 2014, Oracle and/or its affiliates. All rights reserved.1
Quantitative Signal Detection for Mid-
sized Biopharmaceutical Companies
Robert Weber, Senior Product Strategy Manager
Oracle Health Sciences
Dr. Marc A. Zittartz, Chief Quality Officer
PharmaSOL

The following is intended to outline our general product direction. It is
intended for information purposes only, and may not be incorporated into
any contract. It is not a commitment to deliver any material, code, or
functionality, and should not be relied upon in making purchasing decisions.
The development, release, and timing of any features or functionality
described for Oracle’s products remains at the sole discretion of Oracle.

Signal Detection Overview
 Signal Detection and Management has moved into the focus of
Pharmacovigilance activities
 Signals can be detected multiple ways
– Single Case Review
– During PSUR or DSUR creation
– Literature Review
– Authority inquiries
– (Automated) Signal Detection
 Signal can be found from multiple sources
– Spontaneous Reporting Databases
– Clinical Trials
– Electronic Health Records

Signal Detection Approaches
 Identification of new risks: New Signals
– Quantitative Signal Detection – Disproportionality Statistics
– Tracking of Designated Medical Events (DMEs)
– Case Scoring
– Temporal Pattern (Aberration) Detection
 Monitoring of known risks: “Re-Signaling”
– “Keep under Review” – Targeted Medical Events (TMEs)
– Increased Frequency
– Increased Severity (Seriousness, Fatalities…)

History of Signal Detection
 Modern Pharmacovigilance started in the 1960‘s – Thalidomide being one of the main
triggers at the time
 Spontaneous reporting systems were established, intially nationally, from 1968 also as
international collaboration (AU, CA, DE, NL, NZ, SE, UK, US)
 With growing numbers of reports, regulators looked for ways to systematically identify
signals. Napke‘s Pigeon Holes“ (CA 1966) are a famous example of an early manual system
 Computerized signal detection using disproportionality measures began at several centers
during the 80s/90s
 At the end of the century, serveral methods were published:
A group at the WHO (Bate) BCPNN in 1998;
GPS/EBGM (DuMouchel) using FDA data in 1999;
PRR (Evans) using UK MHRA data in 2000

Disproportionality Methods
 Quantitative Signal Detection refers to the identification of
drug-event-combinations within a dataset that appear more often than expected  Signal
(Statistic) of Disproportionate Reporting
 Non-Bayesian Methods
– PRR – Proportional Reporting Ratio
– ROR – Reporting Odds Ratio
 Bayesian Methods
– IC (BCPNN) – Information Component
– EBGM (MGPS) – Empirical Bayes Geometric Mean
 Logistic Regression-based Methods
– ELR – Extended Logistic Regression
– RGPS – Regression-enhanced Gamma-Poisson Shrinker

Issues / Challenges
 “Noise” – False Positives
– Many adverse events are rare (especially if drugs are new) – low counts lead to great
fluctuations of PRR or ROR
– Bayesian methods can reduce false positives
 Detecting Interactions (Drug-Drug-Event Signals)
– Multi-item disproportionality analysis (MGPS -> INTSS)
– LR analysis for computing interaction scores
 Signal Leakage and Masking
– Bias in the database can suppress or falsely elevate signals
– “Innocent bystanders” in polypharmacy situations
– LR can identify the contribution of individual drugs and other factors
– RGPS combines LR with Bayesian shrinkage

 For a certain product 5,4% of all drug-event-combinations are related to a
specific event.
 However only 1,4% of all drug-event combinations are related to this event.
 This drug-event combination appears 3,8 times more than would be expected.
pharmaSOL All Rights Reserved Slide 8
Event All other
events
Percentage PRR
Medicinal Product 52 958 5,4%
3,8
All other medicinal products 691 50.000 1,4%

 Strengths
o Observation in Real-Time
o Case Details
o Availability of Source Data
o Full narrative
 Possible Weaknesses
o Product specific volume
o Total case volume (background)
o Non-diverse product portfolio
o Mix of new and mature products
o Different indications

 FDA AERS (USA)
o since 1968, focus on US data, released quarterly
 WHO Vigibase
o since 1968, data from regulatory agencies worldwide, released quarterly
 PMDA (Japan)
o Recently released, focus on Japan
 Eudravigilance
o EMA intends to publish data in the future
o Focus on European Economic Area (EEA)

 Strengths
o Size and Diversity of public databases
o Information on generic competition
o Ability to detect Class Effects
 Possible Weaknesses
o Case details
o Duplicates
o Time delay (ca. 6 months)

Big
Pharma
Mid-sized
Pharma
Small
Pharma
Need High Medium Medium
Suitable company
dataset
Yes Maybe No
Quantitative Signal
Detection
In use Partly No

Company
dataset
too small
Likely to be
dominated
by few products
or indication areas
Public Data
available with
a time delay

 Spiking: Merge of Company Data with Public Data
o Information related to company product is removed from public
dataset
o Identifier used: compound name
 Company data is injected.
Public Data
Company
Data
Company
Data within
Public Data

Risks
 If product has generic
competition, information is
lost.
Benefits
 Up-to-date with company
data
 Case Details from
company data
 Broad background from
public dataset

Customer Experience
 Mid-sized pharma in Germany
 Employed Argus Safety and Empirica Signal separately
 Developed custom ETL from Argus Safety to Empirica Signal
 Spiked Dataset: Company data was merged with WHO Vigibase
 Different datasets used for different products
Drug Data Interval
Up to 2 years after Launch Spiked Dataset Monthly
Risk Management Plan Spiked Dataset Monthly to Annual
Generic WHO Vigibase 12 months

Quantitative signal detection for the mid sized pharma - webcast

More Related Content

What's hot

Similar to Quantitative signal detection for the mid sized pharma - webcast

Recently uploaded

Quantitative signal detection for the mid sized pharma - webcast