CDISC: Why SAS Programmers Need to Know

Victor Sun, Novartis, East Hanover, NJ

ABSTRACT OVERVIEW

As CDISC advances global clinical data standards, it With the recently publications of the FDA guidelines
has been widely and rapidly expanded in for electric submissions (1-4), they require more data
pharmaceutical industry both domestically and globally. standards than ever before so the new required
The pilot studies with CDISC based electronic deliverables, such as define.pdf and blanckcrf.pdf, etc,
submissions to FDA have been conducted from about can be clearly and conveniently provided to agency for
ten companies. CDISC attention, however, has not reviewing. To facilitate this process, CDISC was
been addressed enough in STAT group even though its founded to develop data standards that could be adopted
implementation is bringing up a significant impact, by reviewing agencies, sponsors and vendors to support
especially to SAS programmers in terms of work skills the electrical aquisition, exchange, submission and
and job security. After briefly introducing CDISC archiving of clinical data and metadata for drug
development and outlining its implementation development.
approaches, future expectations of challenge and
opportunity facing to SAS programmers are discussed. CDISC is an open, multidisciplinary and non-profit
The suggestions provided are intended as a useful organization. The mission of CDISC is to develop of
reference for SAS users in understanding a potential global, vendor-neutral, platform independent standards
trend of electronic submission with CDISC and to improve data quality and accelerate product develop
leveraging job skills to satisfy this new change. in our industry. During last six years, CDISC has
established a number of working teams each with a
INTRODUCTION specific area of focus below.

CDSIC (Clinical Data Interchange Standards ODM Operational Data Modeling team
Consortium), as one of new data standards for clinical to build the standard data exchange
development has been developing for last six years. models that are being developed to
Due to the core concept of open, multidisciplinary support the data acquisition, exchange and
nature, its sponsorship, membership and participation archiving of operational data by using
have rapidly expanded across various organizations in XML.
the US, Europe and Japan. FDA has appointed Liaisons
to CDISC, and completed a pilot study on patient LAB Laboratory Data Modeling team
profiler reviewing for electronic submission by using to build a superset of data items that fully
CDISC standards. describes a clinical trial to the satisfaction
of all the stakeholders included. The
From many one though impossible to the concept superset of fields constitutes the contents
has become a reality, CDSIC is no longer the concern of the model. It will be operated by
of only IT and DM, but of the Statisticians and SAS ODM.
programmers as well. The impact, however, of CDSIC
implementation from top of company level to bottom of SDS Submission Data Standards team
a trial team is unavoidable per current exercises. After to build the standard metadata models for
a brief overview of CDISC, its rapid global expansion supporting the data flow from the
and implementing approaches are introduced. Then, the operational database to regulatory
essential reasons to understand CDISC and new submission.
challenges facing SAS programmers are addressed.
Those comments are written to provide an insight to ADaM Analysis Dataset Modeling team
suggest what one should focus on during and after (a) to build on metadata CRT models
CDSIC is essential to be adopted in your organization developed for safety domains, adding
soon so that one can take a competitive advantage by attributes and examples specific to
following up new technology, expanding job skills, statistical analysis;
heightening productivity and certifying the acquiring (b) to use sample statistical results as
skills along career path even under todays tough job guide for developing data models; and
market. (c) to initially focus on primary and
secondary efficacy variables.
There are also a number task forces associated with
There are also a number of the task forces associated
In Japan
with the four operational teams mentioned above. They
include the strategy task force, the public relations and
communications task force, and the membership task 2002 (Jan) - First CDISC meeting
force. The CDSIC group also cooperates with HL7 DIA Clinical Data Management
(Health Level 7) to harmonize existing CDISC work meeting in Tokyo
with the HL7 reference information model (RIM). JPMA
2002 (May) - CDISC Meetings in Tokyo
EXPANSION 50 Pharma, CROs, Vendors
2002 (Jul) - CDISC Presentations by
CDISC participants represent all stakeholders in the Japanese representatives to MHLW (Japan
pharmaceutical industry. Within a total of 85 CDISC Regulatory Agency) and data manager
member companies, there are 47 sponsors member group
companies, 12 member companies and 26 associate Currently planning Japan CDISC
member companies who have joined in so far and Coordinating Committee(j3C)
helped CDISC to achieve its mission.

Activities associated with electronic submission based IMPLEMENTATION

on CDISC are briefly summarized below.
Based on the current exercises or pilot studies from
some companies, various approaches at different levels
Included 9 sponsors who submitted
are presented for implementing CDISC. DM and IT
electric data and metadata based on groups usually initiate the adopting activities at small
CDISC SDS V2.0 scale. Only few know about CDISC from STAT group.
Facilitating FDA adoption of Patient In general, there are three approaches to implement
Profiles tool and CDISC SDS Standards CDISC. Each method with its major tasks is briefly
Demonstrated value of standard domain described as follows.
formats
Evaluated the pilot data, and second phase Forward
pilot in planning (with additional data)
Start from CRF, work toward submission
Decide what data groups are required
Due to open, multidisciplinary and cross-functional Decide what data items to be collected
composition of CDISC, it not only widely recognizes in Decide what data items to be collected
domestic, but also rapidly expands in global base, Develop database standards
especially in year 2002. The milestones of CDISC Develop reporting dataset and display
development in Europe and Japan are highlighted standards
below.
Backward
In Europe
Start from regulatory requirements and
1999- First CDISC presentation at DIA CDSIC standards for submission
CDM Brussels Create continuity by using CDSIC variable
2002 (Jan) - European Repres. to CDISC names in clinical data management system
BOD Make sure CRFs collect data consistent
2002 (Apr) - Regional CDISC Group with CDISC submission data modes
proposed
2002 (May) - European CDISC Launch, Forward and Backward
Brussels
10 participants from Pharma and
Work on principles for CRFs, and
CROs database design, and analysis and database
2002 (Jun) - CDISC Kick-off meeting design, and analysis and statistical displays
50 participants from 7 counties started simultaneously
Initiation of EU CDISC Coordinating Forward: Specific standards developed
Committee(e3C) from standards CRFs to database to
2002 (Nov) - Next action at DIA Clinical reporting
Data Management Conference in Backward: CDSIC submission data models
Edinburgh used to develop
CHALLEANGE standard CRFs
VS. OPPORTUNITY
CHALLEANGE VS. OPPORTUNITY understanding metadata concept to SAS programmer
may vary. Right now, the way we are applying SAS
It is estimated that about 65% - 75% submission is to use programming-intensive activity for creation,
information is associated with safety data. A big volume revision and reuse of reports without information about
of listings (e.g. for CRT dataset and patient profile, etc.) tables content or style that can go back to SAS in an
is always included even in electronic submissions. Only automated way. SAS is now integrating with XML for
about 30% of programming time is used to generate more convenient reporting process(7). Four
statistical results with SAS, and the rest of components of this underlying technology such as table
programming time to familiarize data structure, check content, table style, database map and publishable XML
data accuracy, and tabulate/list raw data and statistical are mainly based on metadata concept.
results into certain formats(5). This non-statistical
programming time will be significantly reduced after Metadata as one of three data groups (CRT datasets,
implementing the CDISC because of CDISCs uniform analysis datasets and metadata) are also required under
data structure, its useful functions of four models and CDISC submission or reviewing convenience.
some ready-to-be-used CDISC based tools (e.g. PPV
Patient Profile Viewer, etc) being completed or Knowing XML
ongoing. Some listings are expected to be dropped and
some certain tables are not necessary to be generated. It The language of the four models(e.g. ODM, LAB, SDS
is reasonable to consider production of integrated, and ADaM) in CDISC is based on XML(eXtensive
bookmarked and hyperlinked documents to be the final Markup Language), which is new emerging standard
deliverables without too much redundancy. Those marketed as the ideal tool to help industries streamline
could be addressed in the FDA guidance on submission business processes in terms of data management,
of CRT datasets in 2003, which the FDA recently electronic communications and contents managements.
announced it plans to be issued. Both data and metadata are key components in XML.
Therefore, SAS and XML are now cooperating.
Because of these changes, new challenge to SAS XML Engine in SAS v9.0 is built up so one can
programmers is obvious and inevitable(6). At the same import a wide variety of XML documentation. SAS
time, the changes also provide a great opportunity to does what is does best - statistics, and XML does what
people who are willing to reshape their job skills. A it does best - creating report-quality tables by taking
few suggestions are listed below. advantage of the full feature set of the publishing
software. This conversation can produce report-quality
Familiarizing with CDISC tables in an automated hands-off/light out process.

With the new requirements of electronic submission, In addition, it is expected that future FDA standards for
CRT datasets need to conform to a set of standards for delivery of submission components may well specify
facilitating reviewing process. They no longer are XML as the format of choice. Some examples below
created solely for programers convenient. SDS will be indicate this intention. Those include XML-based FDA
treated as specifications of datasets to be submitted, PPV tool, FDA XML standard for exchange of
potentially as reference of CRF design. Therefore, annotated ECG, and the updating guidance on
statistical programing may need to start from this submission of CRT datasets which will be released in
common ground. All existing programs/macros may 2003.
also need to be remapped based on CDISC so one can
take advantage to validate submission information by Intensifying Analysis Skills
using tools which reviewer may use for reviewing and
to accelerate reviewing process without providing Data analysis and result reporting are two major tasks to
unnecessary data, tables and listings. SAS programers. With CDISC, XML and SAS
cooperation, the routine reporting process is expected to
Although CDISC is not rigid, nor does it dictate what be more efficient than before. The task of analysis data,
data a sponsor should collect, it will become a Bible however, will be addressed more for selections of better
which SAS programmers can often refer to for methods on data analysis, tabulation on result reporting
understandings of CRT datasets, programming and and visualisation on data presentation.
preparation of submission.
One way to strength of analysis skill, of course, is to
Understanding Metadata Concept closely follow up new features released from SAS due
to its continuous improvement along four cornerstones
Metadata, by a loosing definition under CDISC is data (e.g. scalability, interpretability, manageability, and
about data. Depending on the strategy of CDISC to be usability). PROC GAM in v8.2 for data analysis and
implemented and functionality of IT, DM and STAT PROC GRADAR in v9.0 for data visualisation are just
groups in different companies, the requirement of two examples for those new features.
Another way of enhancement of analysis skills is to try Useful Web Addresses
to understand clinical data more so more meaningful
analysis can be performed at some aspects such as http://www.cdisc.org/
benefit/risk assessment, simultaneous evaluation http://www.xmltimes.com
adverse evens and laboratory value, etc. on safety
information. Some exploratory analysis with graphic ACKNOLEGEMENTS
skills could be helps to obtain more information for
concerning drug safety. The author wishes to thank Dr. William Stewart and
Mr. Raymond Heimbuch for their support and
CONCLUSION reviewing on this paper.

With the new requirements from updating electronic TRADEMARK INFORMATION

submission and CDISC implementation, understanding
only SAS may not be good enough to fulfill for final SAS and all other SAS Institute Inc. product or Service
deliverables. It is a time to expand and enhance the job
names are registered trademarks or trademarks of SAS
skills from various aspects under new change so that institute Inc. in the USA and other countries.
SAS programmers can take a competitive advantage, indicates USA registration.
and continue to play a main role in both statistical
analysis and reporting for drug development. CONTACT INFROAMTION
REFERENCE
Your comments and suggestions are valued and
encouraged. Contact the author at
1. US Food Drug Administration. Guidance for Victor Sun, Ph.D.
Industry, Providing Regulatory Submission in
Novartis
Electronic Format-General Consideration. IT-2, One Health Plaza
January 1999. East Hanover, NJ 07936
Work Phone: 862-778-6267
2. US Food Drug Administration. Providing Email: [email protected]
Regulatory Submission in Electronic Format-NDAs.
IT-3, January 1999.

3. US Food Drug Administration. Guidance for

Industry, Providing Regulatory Submission to the
Center for Biologics Evaluation in Electronic
Format-Biologic Marketing Applications.
November 1999.

4. US Food Drug Administration. Example of an

Electronic New Drug Application Submission.
2/17/99

5. Victor Sun and Greg Steffens, Comparison

Multiple Database Structures by Using SAS Prior
to Dealing with Clinical Trial Data, Paper
presented in PharmaSUG2000. May 2000.

6. Steven Light, The Changing Nature of SAS

Programming in the Pharmaceuticals Industry,
Conference Proceedings at PharmaSUG2001, May
2001, 255-257.

7. Michael Palmer and Cecillia Hale, Software:

Replace SAS Programmers with XML_Documents
to Help People and Computers Be Happier with
Each Other, Conference Proceedings at
PharmaSUG2001, May 2001, 245-249.