Transcript Click to open - the Population Approach Group in Europe

Literature databases: integrating information on
diseases and their treatments
Vandemeulebroecke M, Demin I, Luttringer O, McDevitt H, Ramakrishna R, Sander O
Advanced Quantitative Sciences
Novartis Pharma AG, Basel, Switzerland
Background
Objectives
Conclusion
Drug development relies on thorough knowledge of
the disease in focus, in particular with respect to its
natural course and the efficacy and safety of already
existing treatment options1,2. Such knowledge allows
to put a therapy under development into context and
to inform decision making throughout development.
A great wealth of information is available for this
purpose in published literature. However, it is rarely
compiled systematically into one integrated, welldocumented and easy-to-search source database.
• To build comprehensive literature-based summary
Significant lead time and effort is required to
assemble the relevant data for comprehensive
literature databases, as well as for the
implementation of a generic IT solution to host
such databases. However, great value can be
derived from this investment.
level databases on selected indications and their
major treatments, including longitudinal outcome
data and covariates to facilitate dynamic
modeling.
• To implement a relational database infrastructure
as a generic IT solution that allows state-of-the art
creation and maintenance of such literature
databases.
Relational database solution
Methods
Results
The development of infrastructure for building and
maintaining indication-specific literature databases
has been described previously3. Building upon this,
literature sources for each database are
• systematically identified,
• curated according to a standard process,
• condensed into a standard format database.
Current range of applications
Table 1 provides an overview of the literature
databases that have been built so far, along with
their major applications. For two databases currently
under construction, the intended application is given.
Each database assembles
• comprehensive longitudinal (group-level) data
• on a wide range of endpoints
• for all major drugs of the respective indication
• also including demographic and background
characteristics
• based on all relevant publications in the field.
The methods are illustrated by the following
motivating example4.
Summary-level longitudinal data on the clinical
efficacy of biologics for Rheumatoid Arthritis (RA)
are available in the literature, including patient
characteristics and concomitant medications. We
carried out a systematic review of the published
literature on clinical trials of biological treatments in
RA. The data were extracted using robust processes
into a literature database, which contained all the
relevant information necessary to build a drug–
disease model.
The aim of this work was twofold: first, to quantify
the time course of the ACR20 score across
approved drugs and patient populations, and
second, to apply this knowledge in the decisionmaking process for an internal compound. The
integrated analysis included data from 37 phase II–
III studies describing 13,474 patients (Figure 1).
With this, the efficacy of the internal compound
could be put into perspective, and decision-making
processes could be effectively supported. The
framework can be applied to any other compound
targeting RA, thereby supporting internal and
external decision making at all clinical development
stages
Figure 1: Model-based predictions of median ACR20 responder
rates together with their 90% Bayesian confidence intervals for
approved biologics and placebo. All treatments including placebo are
given in combination with methotrexate (MTX) to patients with previous
exposure to MTX. The blue stripe represents the 90% Bayesian
confidence interval for certolizumab.
Table 1. Novartis AQS literature databases with major applications
Database
COPD
Type 2
Diabetes
Rheumatoid
Arthritis
Hepatitis C
Virus
Multiple
Sclerosis
Chronic
Kidney
Disease
Psoriasis
Number of
Major applications
publications
included
39
• Comparison of in vivo
performance of different
bronchodilators / devices
45
• Benchmarking against
competition and
standard of care
128
• Support of Go/Nogo
milestone
• Optimizing clinical trial
design
21
• Benchmarking against
competition
• Predicting sustained viral
response from early viral
response
40
• Benchmarking against
competition
21
• Quantifying the
competitive landscape
23
Osteoporosis
20
Dyslipidemia*
-
Heart Failure*
-
• Competitive profiling
including benefit/risk
assessment
• Supporting dose
selection
• Assessment of
compound‘s efficacy
• Explore lipid lowering
potential of own
compounds on top of
standard of care
• Calibration of a
simulation platform for
the cardiovascular/renal
system
* currently being built
References:
Managing the literature data presents some unique
challenges:
• Data need to be kept up-to-date, and updates should
be traceable
• Data need to be traceable back to the source
references
• Data need to be queried and searched in a
meaningful way
Our literature databases were implemented as
comprehensive Excel spreadsheets in the past.
However, Excel spreadsheets do not provide an ideal
way of addressing these needs. A relational database
system has therefore been developed in collaboration
with an external vendor (GVKBio) that addresses these
requirements. Existing Excel spreadsheets can be
uploaded, and new data can be entered directly. Once
the data have been verified, they can be queried and
searched through a simple user interface (Figure 2).
The system can be developed further based on growing
practical experience.
Figure 2: User interface of the relational database application
Discussion
A great wealth of information is available in
published literature on the natural course of
diseases and the effect of available treatments
options. However, it is rarely compiled into
comprehensive quantitative databases. If available,
such databases allow to assess the effect of a new
compound quickly and accurately in the context of
the competitive landscape.
For this reason, Novartis has been building several
literature databases on selected medical indications.
This requires significant lead time and effort, but
great value can be derived from this investment The
range of successful applications spans from dose
selection to supporting Go/Nogo milestones.
Building on this experience, we have recently built a
more generic IT infrastructure to capture these and
future databases. This solution has an intuitive webbased interface, while still retaining the full flexibility
of a relational database in the background.
1 Mandema
et al.: „Model-based development of gemcabene, a new lipid-altering agent“, AAPS J 2005
2 Ito et al.: „Disease progression meta-analysis model in Alzheimer‘s disease“, Alzheimer‘s & Dementia 2010
3 McDevitt et al.: „Infrastructure development for building, maintaining and modeling indication-specific summary-level literature databases to support model-based drug development”, PAGE 2009
4 Demin et al.: „Longitudinal model-based meta-analysis in rheumatoid arthritis: an application toward model-based drug development”, Clin Pharmacol Ther 2012
This study was supported by Novartis Pharma AG, Basel, Switzerland.
Copyright © 2013 Novartis Pharma AG, Basel, Switzerland.
All rights reserved.
Poster presented at the Population Approach Group Europe (PAGE), June 11-14, 2013, Glasgow, UK.