1. A Coherent and Practical End-to-End Metadata Strategy using Existing Standards and Tools for Clinical Research
Stephane AUGER
Danone Research, FRANCE

2. Clinical studies performed in 4 divisions Historically
CONTEXT
Clinical studies performed in 4 divisions
Dairy, Waters, Baby and Medical Nutrition
Historically
Mostly paper-based studies until very recently
DM Outsourced to CROs
Few internal data / metadata standards
Result
Few controls on data / metadata received from CROs
Lack of homogeneity between studies
Resource intensive to pool for meta analysis or
format for a DR/DW
Today i’d like to share with you the results of a project that we’ve had ongoing now for almost 3 Danone Research on implementing a metadata strategy for our clinical studies.
…and today i’m going to focus on the metadata strategy that we’ve put in place in our Dariy and Waters divisions.

3. “First, we need to get more control over our clinical study metadata”
CONTEXT
2010
Upper management request for a “Data Warehouse”
“First, we need to get more control over our clinical study metadata”
Thought process to address the end goal:
Use existing tools/software as much as possible
Standardize up front… CDISC SDTM, CDASH, ODM
A bonus if we could make metadata management and study set-up more efficient
And so this metadata strategy project began in 2010 when I received a request from my….
…and what I told them at the time was that we were nowhere near ready to implement a DW. And that a DW was the End-Result of a project that begins with getting ……
…and so we launched this project with a few things in mind…..

4. Key Decision #1: Metadata Standards Management (1)
Implementation of a Study Metadata Modeler (SMM) … with a duel use
Manage Study Metadata Standards
CRF Forms, Item Groups, Items, Codelists
Item Characteristics (type, formats, range values etc.)
CRF completion guidelines / help text
Database structure
Based on SDTM/CDASH
The first key decision that we made is this project was to implement a SMM,
…..and the 1st thing that this tool allows us to do is…
The tool really provides us with an environment where we can define and manage all of our metadata standards for our CRF moduels
Not so easy to do – need to plan for multiple EDCs and limitations of those EDCs – vertical vs. horizontal structure. Eg of Vital Signs

5. Key Decision #1: Metadata Standards Management (2)
Implementation of a Study Metadata Modeler (SMM) … with a duel use
Create Study-Specific Metadata
Use Metadata Library to create a study-specific “CRF”
Export study-specific “CRF” in ODM format to an EDC system
→Requires ODM-compatible EDC system
The 2nd thing that this tool allows us to do, is to capitalize on our metadata standards library to create a study-specific CRF in a non-programming environment.
…and really, the key point here is that..
…we are able to export, from the SMM, the study-specific CRF metadata in ODM format
…..which we can later exploit by an EDC tool….which of course needs to be ODM compatible

6. SMM EDC SMM Functionality CRF DB Structure Simple Edit Checks
Workflow Config.
ODM
DB Structure
Complex
Edit Checks
Simple
Edit Checks
Database Admin.
ODM
Visit Layout
IWRS
Coding
TSDV
Custom Attributes
Custom Configuration

7. Key Decision #1: Metadata Standards Management (3)
SMM - Implementation Experience
Define company standard CRF forms and variables
Workgroups representing Clin Ops, QA, Statistics, Data Management
1 year – 80 man/days
Resulted in a Data Standards Committee and change management process
Define standard DB structure in the SMM
Each EDC has different production database specificities
One library for multiple EDC systems
SMM provider an essential resource
3 months – 60 man/days
..Some of the return-on-experience from this phase of the project:
We created workgroups in order to define our company-standard CRF forms and variables… with representation from each group of internal stakeholders.
With respect to defining the standard DB structure, this part was a little more tricky, and we really relied on our SMM vendor and their experience with different EDC systems, and each of their unique ways of handling data.
and this was key, because we wanted ONE standards library to be as flexible as possible

8. Key Decision #2: ODM-Compliant EDC (1)
Implementation of ODM-compliant
EDC tool(s)
ODM-in
Import Study-Specific CRF metadata - ODM file
ODM-out
Export study metadata & study data in ODM
The 2nd Key Decision that we made in this project (as you can guess)– was to select an EDC tool that was ODM compliant.
Now i’m not going to get into the debate today of what ODM compliance means - as they only gave me 20 minutes to talk
But, generally, the two main things that were looking for was that the EDC could:
import our study specific CRF metadata to facilitate the study build
…and that we could also Export the study metadata as well as the actual clinical study data

9. Study Set-up: What’s left to do?
SMM
EDC
CRF
Workflow Config.
Visit Layout
ODM
Complex
Edit Checks
Simple
Edit Checks
Database Admin.
ODM
DB Structure
So, what parts of the EDC study set-up are still left to configure once we’ve imported our metadata from the SMM?
So, highlighted in yellow here are the tasks that still need to be configured in the EDC tool itself before study launch.
Medical Coding
Custom Attributes
TSDV

10. Key Decision #2: ODM-Compliant EDC (2)
Implementation Experience
Many challenges to overcome and details to solve
SMM & EDC implementation at the same time
Requires tech-savy personnel
60-80% of the EDC study build is done automatically in this process
Main differences between our two EDC tools
Two-step ODM import process
Visit configuration
Complex edit checks need to be re-configured
Our return-on-experience from this phase of the project:
As you can imagine, there were …… especially considering that we were implementing
After our test phase and our first two implementations, we’ve estimated that…..

11. Key Decision #2: ODM-Compliant EDC (3)
Implementation Experience (2)
Impact on EDC budget?
75% decrease in professional service costs
62% decrease in overall EDC budget
Impact on Data Management Resources?
No Change in Study Build FTEs (but more tasks performed internally)
≈50 decrease in study set-up time (Protocol to EDC Go-Live)
≈ 20% increase in overall DM study budget (vs. Paper-based)
Database &
EDC Set-up
CRF = EDC Set-up
Reduced Programing
Reduced Validation
Metadata Control
Standardization
Efficiency
Warehouse-ready data
As we are in our 1st few studies with this implementation – we will get a better with the next few studies exactly how the FTEs will be affected, but for the first few studies – no change
.so not in the # of man days, but in the time interval from Protocol to EDC Go-Live
..and what are we getting for this 20%

12. Roles and Responsibilities
EDC Vendor
CRO
Complex
Edit Checks
Hosting & Support
SMM
CRF & Study Build
Validation
Danone
(Data Manager)
Custom Config. / Reports
Workflow Config.
In-Study DM
Database Admin.
EDC
ODM In
Database Lock
Study-Specific validation is greatly reduced since standards are used from the beginning.

13. Key Decision #3: ODM-Compliant Mapping Tool
SMM
EDC ??
Study Metadata Standards
ODM
ODM
Study Build ODM

14. Key Decision #3: ODM-Compliant Mapping Tool
Implementation of ODM-compliant
Mapping (ETL) tool
Maps ODM (from the EDC) to produce:
SAS transport datasets (SDTM format)
Define.xml
Standardized ODM
Standard Mapping Metadata
Re-Usable mapping scripts
Validation = only once

15. Key Decision #3: ODM-Compliant Mapping Tool
EDC a
ETL
Mapping
Tool
Data Repository?
ODM
SAS Datasets
Define.xml
EDC b
ODM
ODM

16. Clinical Study Data & Metadata Flow: Tools
Build
Conduct
Exploit
METADATA
Metadata Standards & Study Set-up
(Formedix)
METADATA + STUDY DATA
Data Collection (EDC)
(Medidata RAVE
OpenClinica)
Mapping
(XML4Pharma)
Data Repository

17. Clinical Study Data & Metadata Flow: Management
Build
Conduct
Exploit
METADATA
Metadata Standards & Study Set-up
(Formedix)
METADATA + STUDY DATA
Data Collection (EDC)
(Medidata RAVE
OpenClinica)
Mapping
(XML4Pharma)
Metadata Standards & Management
Data Repository

18. Clinical Study Data & Metadata Flow: Standards
Build
Conduct
Exploit
METADATA
Metadata Standards & Study Set-up
(Formedix)
METADATA + STUDY DATA
Data Collection (EDC)
(Medidata RAVE
OpenClinica)
SDTM/CDASH
Mapping
(XML4Pharma)
SDTM
Data Repository
Define.xml
ODM

19. End-to-End metadata standardization and control is possible
Conclusion
End-to-End metadata standardization and control is possible
There are significant cost and time savings to be had!
The tools and standards are ready and available NOW

20. Thank You!