Improving Women's Health Through Modernization of Our Bioinformatics Infrastructure

Abstract

Our nationwide bioinformatics infrastructure used to detect important sex differences associated with medical product use is antiquated. The Food and Drug Administration (FDA) has embarked on an ambitious bioinformatics modernization effort that will improve our ability to assess the safety and effectiveness of new medical products. This, in turn, will improve our ability to detect important sex differences.

Full text

Improving Women’s Health Through
Modernization of Our Bioinformatics
Infrastructure
A Oliva
1
, E Pinnow
2
, R Levin
1
and K Uhl
2
Our nationwide bioinformatics infrastructure used
to detect important sex differences associated with
medical product use is antiquated. The Food and Drug
Administration (FDA) has embarked on an ambitious
bioinformatics modernization effort that will improve our
ability to assess the safety and effectiveness of new medical
products. This, in turn, will improve our ability to detect
important sex differences.
THE BIOINFORMATICS INFRASTRUCTURE PROBLEM
Bioinformatics is the design and development of computer-
based technology to support the life sciences.
1
At the FDA,
bioinformatics means having modern computer systems to
efficiently and effectively manage information regarding
FDA-regulated medical products. Ensuring the efficient
exchange of medical product information among the FDA
and its stakeholders is critical to the FDA’s mission to protect
and promote public health (see Figure 1).
Through its Critical Path Initiative, the FDA is seeking to
facilitate a national effort to modernize the scientific process
through which a potential human drug, biological product,
or medical device is transformed from a discovery or ‘‘proof
of concept’’ into a marketed medical product.
2
The FDA has
identified bioinformatics as an area where improvements can
increase efficiency, predictability, and productivity in the
development process. Effective bioinformatics modernization
requires an understanding of the existing problems and
addressing them.
To understand the current inefficiencies in our bioinfor-
matics infrastructure, one need only consider a time before
we developed our modern financial information infrastruc-
ture. Back then, we lacked a standard currency, and wealth
was maintained in a variety of different forms (e.g., gold,
silver, beads). Banks were few and wealth was commonly
stored under a mattress or hidden away under floorboards.
Money was counted by hand and transported in sacks. It is
easy to imagine the inherent inefficiencies of such a system.
Today, the US financial industry has a standard currency
(US dollars), has improved access to money and transaction
methods (e.g., banks, automatic teller machines, credit
cards), and has developed user-friendly interfaces to financial
information (e.g., automated money counting machines,
calculators). Our modern financial infrastructure allows
money to travel electronically both quickly and securely
wherever needed, and financial information is accessible
virtually anywhere. This modern financial infrastructure has
enabled modern financial management tools that were
previously unthinkable. For example, the Federal Reserve
now has tools to monitor billions of financial transactions
almost in real time to assess the health of the economy and to
make better monetary policy decisions. For the individual
user, some banks now provide account balances through
instant messages to an account holder’s cell phone.
EDUCATION
nature publishing group
Figure 1 The regulatory product information supply chain. Arrows
represent points at which product information is exchanged. The regulated
industry gives the FDA certain information about a product (e.g., clinical trial
data). The FDA evaluates that information and, if the product is approved, in
turn, makes certain product information available to the public (e.g., drug
labeling). The public in turn provides product information to the Agency
(e.g., adverse event reports), which the FDA processes, providing updates to
the regulated industry (e.g., recall) and the public if needed (e.g., new
labeling, warning box). PIE, Product Information Exchange.
1
Office of Critical Path Programs, Food and Drug Administration, Rockville, Maryland, USA;
2
Office of Women’s Health, Food and Drug Administration, Rockville,
Maryland, USA. Correspondence: A Oliva (armando.oliva@fda.hhs.gov)
Published online 7 November 2007. doi:10.1038/sj.clpt.6100437
192 VOLUME 83 NUMBER 1 | JANUARY 2008 | www.nature.com/cpt

Our bioinformatics infrastructure today is comparable to
the financial infrastructure of years past. If one compared
clinical study data to money, we have no widely implemented
standard currency for study data. Furthermore, companies
store their study data locally and in different formats,
limiting access and making it extremely difficult to exchange
data quickly and securely between authorized entities. The
available tools to analyze study data are outdated and difficult
to use, making the assessment of a medical product’s safety
and effectiveness time-consuming. The resulting inefficien-
cies make it difficult for the FDA to carry out its mission to
ensure that medical products are safe and effective. Unlike the
Federal Reserve’s ability to leverage our modern financial
information infrastructure to inform monetary policy, the
lack of a modern bioinformatics infrastructure hinders the
FDA’s ability to make better informed health policy decisions
about new medical products.
We must maximize the usefulness of the information we
collect from clinical studies that are submitted to the FDA to
support marketing approval. To do this, the FDA and its
regulated industry must undergo a major transformation.
By implementing a standard currency for study data,
improving access to the data, and developing user-friendly
tools (i.e., interfaces) that can help us efficiently convert data
into knowledge, we can lower development costs, shorten the
time to market for new products, and enhance our ability to
communicate information about their safe and effective use,
thus promoting the public health. The FDA recognizes its
leadership role in this transformation.
THE SOLUTION
A modern bioinformatics environment requires enhancing
three key information management domains: (1) informa-
tion standards, (2) information access, and (3) interface or
tools that can efficiently analyze information. Standards,
access, and interface tools combine to influence the way we
process information and improve regulatory decision making
(see Figure 2).
Standards
By using data standards, we establish a common language for
managing health-related information. Data standards enable
both people and computers to communicate effectively and
efficiently. Although all communication with the agency must
be in Standard English, we must use a standard language that
computer systems can understand to automate the processing
of information. We must also use standard terminology:
formalandconsistentdefinitionsaswellasexamplessothatwe
can effectively share concepts and ideas. Computer systems have
their own language with precise terminologies and dictionaries
(i.e., semantics) and grammar (i.e., syntax). Using precise,
standard machine-readable language is a necessary component
of an effective bioinformatics environment.
Access
Improving access means making it easy to send, receive, and
share information. Eliminating paper and exchanging informa-
tion electronically greatly improve access. Improved access also
means being able to quickly locate specific information contained
within electronic files and documents. It is not sufficient to easily
access a review of a new marketing application; we also need
quick access to detailed information within a review, such as
information about an unusual adverse event.
Interface
The third key information management domain is interface.
We must develop user-friendly interfaces or tools for
analyzing information that efficiently and effectively help us
convert information into knowledge. We must ensure that
users are properly trained to use these tools. Increasingly, the
modern interface tools to information are software programs
on a computer.
Improvements in all three key information management
domains are needed to enhance decision making and
improve the public health. Creation of a national modern
bioinformatics infrastructure is not something industry and
other relevant stakeholders can achieve without Federal
leadership—the government has already taken the lead by
launching its broad health information technology (Health
IT) effort.
3
In its Critical Path Initiative, the FDA identified
harnessing bioinformatics
4
as a key area needing the most
work. By harnessing bioinformatics to modernize our own
information management environment, the FDA will help
pave the road for its stakeholders to do the same.
It is the FDA’s goal to achieve a modern bioinformatics
environment that enhances its decision-making capability.
Ideally, all information flowing through the regulatory
product information supply chain is standardized, both
information and knowledge are accessible electronically, and
the tools for analyzing information are both effective and
user friendly.
EXAMPLE: TRACKING WOMEN IN CLINICAL TRIALS
The inclusion of women in clinical trials is necessary to
identify sex-related safety and efficacy differences. A 2001
Figure 2 Three information management domains.
CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 83 NUMBER 1 | JANUARY 2008 193
EDUCATION

report by the Government Accountability Office found that 8
out of 10 drugs withdrawn since 1 January 1997 posed greater
health risks for women.
5
Congress expressed concern that the
FDA had paid insufficient attention to gender-based research,
mandating the FDA to pilot a database system that collects
demographic information.
6
Tracking the participation of women in clinical trials is a
good example of an existing inefficiency in our bioinfor-
matics infrastructure with a direct impact on women’s health.
We use this information to determine if an adequate number
of women participated in a trial (or a group of trials) to begin
to assess whether an important sex difference exists.
Surprisingly, there is no widely used standard to record the
sex of a subject in a clinical trial. Some call this important
variable ‘‘sex.’’ Others call it ‘‘gender,’’ even though it is the
biological sex that is recorded. Some clinical trials record the
sex as ‘‘male’’ or ‘‘female,’’ others as ‘‘M’’ or ‘‘F,’’ still others as
‘‘0’’ or ‘‘1,’’ or ‘‘1’’ or ‘‘2’’ (leaving the analyzer of the data
sometimes wondering which number corresponds to which
sex!). If one wished to write a computer program to count
the number of women that participated in 10 trials of a new
drug, a manual, time-consuming process is needed. One has
to first see how the sex information is recorded in each trial,
what terms were used to describe each sex, and then write a
computer program to analyze those specific 10 trials. Once the
analysis is complete, one discards the program because the next
10 trials will use different standards to document sex, and
another custom program will be needed. Without a standard
way of describing the sex of a subject, our ability to develop
reusable and computer-automated counting techniques (i.e,a
modern and efficient interface, analogous to money-counting
machines in the financial world) is substantially impaired.
The Clinical Data Interchange Standards Consortium
(CDISC)
7
is a non-profit organization that has developed
standards (called the Study Data Tabulation Model, or
SDTM) for the exchange of clinical trial data. They
recommend that all clinical trials always record sex informa-
tion in a variable called ‘‘SEX’’ and identify the sex of each
subject as M, F, or U (unknown). The FDA views the SDTM
as a positive step forward in standardizing clinical trial data
and recommends its use.
8
CDISC-formatted clinical trial data
are easier to analyze because they support the development of
reusable computer programs (interfaces) that will greatly
facilitate tracking the participation of women in clinical
trials, for example.
Implementation of the SDTM industry-wide is still in its
infancy. A recent FDA study revealed a significant learning
curve in using the standard correctly.
9
But, with time and
experience, we anticipate that wider and more consistent
adoption of the SDTM will yield efficiencies that will make
tracking the participation of women in clinical trials a
relatively trivial computational problem. The same data
standardization strategy will allow ease in tracking any
particular subpopulation in clinical trials (e.g., pediatrics,
elderly), so that clinical trial data become more useful even
beyond women’s health.
This, of course, is just the first step. With standardized
electronic clinical trial data, to include adverse event and
laboratory data, modern computerized analytic tools can be
developed (prototypes of which are currently under evalua-
tion and limited use at the FDA) to easily understand the
effects of medical products in women and to efficiently
identify important sex and gender differences associated with
their use.
A common limitation of clinical trials is that they do not
enroll sufficient members of an important subgroup such as
women, often making it difficult to identify important sex
differences by analyzing a single trial. The availability of
standardized trial data makes it much easier to use powerful
statistical analytic techniques across many trials to increase
our ability to detect important subgroup differences, without
needing to conduct difficult, expensive, and time-consuming
data conversion each time.
NEXT STEPS
A modern bioinformatics infrastructure for clinical trial data
means that all data are available electronically (improved
access) and are standardized and that modern analytic tools
(interfaces) are available to quickly and efficiently analyze the
data to create useful knowledge about medical products. The
key is more widespread adoption of clinical trial data
standards industry-wide. The FDA is collaborating with
CDISC and Health Level 7 (HL7) (an American Standards
Institute (ANSI)-accredited standards development organiza-
tion working in the health-care arena)
10
to create a data
exchange standard that is completely harmonized with
existing HL7 standards currently being used in exchanging
Electronic Health Record (EHR) information. Although also
in their infancy, EHRs provide an exciting opportunity to
monitor the effects of medical products on a much larger
scale, and throughout the entire life cycle of a medical
product, from early development through widespread post-
marketing use. The CDISC-HL7 standard, once developed,
will support seamless integration of medical product
information wherever and whenever it is collected. The
ultimate goal is having high-quality standardized data about
medical products accessible electronically, exchanged quickly
and securely, and analyzed using modern analytic tools to
improve the public health.
Modernization of our bioinformatics infrastructure is
costly, complex, and time-consuming. Nonetheless, it is a
necessary step to improve the benefit–risk assessments of
medical products, which in turn will enhance our ability to
identify important sex differences associated with their use.
These bioinformatics modernization principles, when applied
to all types of non-clinical and clinical research data, e.g.,
pharmacokinetics, pharmacodynamics, genomics, biomar-
kers, create tremendous opportunities to improve our
understanding of the effects of the medical products that
we all use, so that we can truly personalize health care in the
future. No longer should we be satisfied with the simple
knowledge that a medical product is safe and effective in
194 VOLUME 83 NUMBER 1 | JANUARY 2008 | www.nature.com/cpt
EDUCATION

humans. We must strive to answer the more important
question: is this product safe and effective for this individual,
based on her important genetic and phenotypic markers?
Modernizing our bioinformatics infrastructure will move us
one giant step closer to answering that question.
CONFLICT OF INTEREST
The authors declared no conflict of interest.
& 2008 American Society for Clinical Pharmacology and Therapeutics
1. Lacroix, Z. & Critchlow, T. Bioinformatics: Managing Scientific Data.
(Morgan Kaufman Publishers: San Francisco, 2003).
2. US Food and Drug Administration. FDA’s Critical Path Initiative /http://
www.fda.gov/oc/initiatives/criticalpath/S. Accessed 28 August 2007.
3. Health and Human Services. Health information technology
/http://www.hhs.gov/healthit/S. Accessed 28 August 2007.
4. US Food and Drug Administration. Critical path opportunities list and
report. Topic 3: Harnessing Bioinformatics, R-15 /http://www.fda.gov/
oc/initiatives/criticalpath/reports/opp_report.pdfS. Accessed 28 August
2007 (2006).
5. US General Accounting Office. Drug safety: most drugs withdrawn in
recent years had greater health risks for women /http://www.gao.gov/
new.items/d01286r.pdfS. Accessed 28 August 2007.
6. 107th Congress (2001–2002) 1st Session (FY2002). Senate Committee
Report (S. Rep. 107–41) /http://frwebgate.access.gpo.gov/cgi-bin/
getdoc.cgi?dbname=107_cong_reports&docid=f:sr041.pdfS, 146.
Accessed 28 August 2007.
7. Clinical Data Interchange Standards Consortium /http://
www.cdisc.orgS. Accessed 28 August 2007.
8. US Food and Drug Administration. Study Data Specifications, Version 1.3
/http://www.fda.gov/cder/regulatory/ersr/Studydata.pdfS. Accessed
28 August 2007.
9. Pinnow, E. Evaluation of NDAs submitted to the FDA in CDISC SDTM
format: feasibility of using demographic domain data to track the
inclusion of women in trials /http://www.cdisc.org/publications/
interchange2006/session8/EllenPinnowCDISC2006Pinnow.pdfS.
Accessed 28 August 2007.
10. Health Level Seven /http://www.hl7.orgS. Accessed 28 August
2007.
CLINICAL PHARMACOLOGY & THERAPEUTICS | VOLUME 83 NUMBER 1 | JANUARY 2008 195
EDUCATION