Content uploaded by A. Ali
Author content
All content in this area was uploaded by A. Ali on Nov 16, 2018
Content may be subject to copyright.
associations on this list received a short assessment in order of descend-
ing score. After short assessments, based on internal criteria, drug‐ADR
associations can be selected for more detailed analysis and potentially
lead to Signals.2 We tested the performance of the new model on the
proportion of initial assessments that were selected for detailed analysis
by trained pharmacovigilance assessors. As comparator the proportion
selected during use of the old screening model in 2014‐2016 was used.
A2‐sided Pearson X
2
‐test was used to test the difference in proportion
of associations selected for detailed analyses.
Results: 626 initial assessments suggested by the new model were per-
formed,resulted in 53 more detailed analyses (8.5%).In 2016 the propor-
tion was 6.6% (79 detailed/1203 initial), in 2015 2.5% (14 detailed/520
initial), in 2014 4.2% (13 detailed/312 initial). For the new model (2017)
vsoldmethod(2014‐2016) X
2
‐test p=.003.
Conclusions: The real‐world performance of the model in its first year of
use and the comparison with the old method during three earlier years
showed that the prediction model‐basedapproachleadstoabetterpre‐
selection of associations and thus more detailed analysis. Literature 1.
Scholl JHG, van Hunsel FPAM, Hak E, van Puijenbroek EP. A prediction
model‐based algorithm for computer‐assisted database screening of
adverse drug reactions in the Netherlands. Pharmacoepidemiol Drug
Saf. 2018 Feb;27(2):199‐205.2. van Hunsel F, Ekhart C. Experiences with
acomputer‐assisted database screening tool at The Netherlands
Pharmacovigilance Centre Lareb. Pharmacoepidemiol Drug Saf.
2015;24(S1):442.
561 | A model for data monitoring
committees for retrospective study designs in
the real‐world setting
Atheline Major‐Pedersen
1
; Mary Kate McCullen
2
; Mary Beth Sabol
3
;
Omolara Adetunji
4
; Joseph Massaro
5
; Alfred Neugut
6
;
Julie Ann Sosa
7
; Anthony Hollenberg
8
1
Novo Nordisk A/S, Copenhagen, Denmark;
2
AstraZeneca, Wilmington,
Delaware;
3
GlaxoSmithKline, Collegeville, Pennsylvania;
4
Eli Lilly and Co,
Windlesham, Surrey, UK;
5
Boston University School of Public Health, Boston,
Massachusetts;
6
Columbia University Medical Center, New York, New York;
7
Duke University Medical Center, Durham, North Carolina;
8
New York‐
Presbyterian Hospital‐Weill Cornell Medical Center, New York, New York
Background: Non‐interventional post‐authorisation safety studies (NI
PASS) are increasingly performed for assessing post‐marketing drug
safety. They are an important tool for detecting rare risks that are hard
to measure during drug development programmes. Regulatory Author-
ities encourage sponsors with marketing authorisations within the
same drug class to collaborate in NI PASS. Data monitoring commit-
tees (DMCs) are traditionally set up for randomised clinical trials
(RCTs), but may have a central role in the validation and ongoing inter-
pretation of the large amount of data emerging from NI PASS. To our
knowledge, DMC guidelines for NI PASS are non‐existent.
Objectives: To share our experience with target audience (pharma,
academia, regulatory authorities) in view of the expected increase in
such joint pharma NI PASS.
Methods: Medullary thyroid cancer (MTC) is an important potential
risk for long acting glucagon like peptide 1 receptor agonists (LA
GLP‐1 RAs). Since MTC is rare, RCTs cannot easily characterise this
potential risk. FDA imposed a joint pharma‐sponsored, 15‐year case‐
series NI PASS investigating the incidence of MTC across the US state
cancer registries in relation to exposure to LA GLP‐1 RAs. Because of
the study´s anticipated long duration and large amount of safety data,
sponsors jointly established a DMC. We performed descriptive analy-
sis (hypothesis testing and power calculations were not applicable) of
existing literature on safety data monitoring in RCTs, and assessed
their applicability and needed modifications, for ongoing interpreta-
tion and validation of data from retrospective NI PASS.
Results: We highlight features identified in the literature for data mon-
itoring in RCTs and their applicability to NI PASS, and how we incor-
porated/adjusted these into a DMC model. Challenges are described
in collaborating with multiple sponsors to reach a common DMC
model that would balance the individual sponsors´ high level of confi-
dentiality and the requirement for all participating sponsors to be
alerted of safety concerns raised by the DMC, triggered by any of
the participating sponsors´ products.
We present an effective DMC meeting structure and communication
flow between sponsors, the DMC, other study specific committees
and the FDA. Sponsors and DMC members have signed a DMC
charter; a kick‐off and 3 successful DMC meetings have been held.
Conclusions: This DMC model may be of interest to others involved in
this evolving real‐world pharmacovigilance area.
562 | Factors distinguishing identified risks
from potential risks: Analysis of safety
specification of Japan and EU risk
management plan
Saeko Hirota
1,2
; Takuhiro Yamaguchi
1
1
Tohoku University Graduate School of Medicine, Sendai, Japan;
2
EPS
Corporation, Tokyo, Japan
Background: Based on ICH‐E2E agreed in 2004, European Medicines
Agency and Japan Pharmaceuticals and Medical Devices Agency
(PMDA) introduced Risk Management Plan (RMP) in 2005 and 2013,
respectively. Safety specification, the core component of RMP, is
divided into 3 categories: identified risk, potential risk, and missing
information, but it is unknown what factor is emphasized when decid-
ing categories of safety specification in Japan and EU.
Objectives: To compare the safety specification between Japan and
EU, and to identify the factors which strongly affect decision‐making
to categorize safety specification as identified risk.
Methods: This study included non‐orphan drugs approved both in Japan
and EU as of Dec 31, 2016 with available RMP. Initial safety specifications
and clinical trial data were obtained from RMP, review report, European
Public Assessment Report, and CommonTechnical Document. We ana-
lyzed contributing factors to the categorization of identified risk using
mixed‐effects logistic regression model, taking random effects for drug
into account. Outcome was defined dichotomously as (1) listed as
ABSTRACTS 259
for the MTC Registry Consortium