1. Metadata Management – Our Journey Thus Far
Archana Bhaskaran
Oncology Database Development Operations
28-Jan-2015

2. Disclaimer
The opinions expressed in this presentation and on the following slides are solely those of the presenter and not necessarily those of Novartis. Novartis does not guarantee the accuracy or reliability of the information provided herein.

3. Evolution of metadata management – Beginning, Current and Future
Agenda
Introduction
Evolution of metadata management – Beginning, Current and Future
Key Benefits Achieved and Current Challenges
Governance process

4. Introduction
Present a high level overview of Novartis roadmap on metadata management and associated processes from inception to present.
What is Metadata?
Metadata is information about the data.
Metadata describes the domains and data elements used in clinical trials
Example of metadata: variable names, type, derivation algorithm, codelists, etc.

5. MDR Roadmap
Novartis has defined an end to end process that governs the framework to define, maintain and use of metadata within a single repository to support all upstream and downstream processes and tools.
Beginning:
iMDR - Interim MetaData Repository (Excel spreadsheets)
Current:
MDR - off-the-shelf Oracle based global metadata management tool customized for Novartis
Future:
MDR – off-the-shelf Oracle based ‘Single Repository’ that enables the management of Global standards and study level Metadata

6. Metadata Hub Data Warehouse MDR Like a center screw in a wheel
Data Collection
(CDMS)
Data Warehouse
MDR
Standards Development / Modeling
Submission
(SDTM, ADaM)
Like a center screw in a wheel
the Metadata (Standards)
in the MDR drives all
Implementation programs and
data flow (hence process)

7. Type of metadata maintained
Clinical Data Elements (CDE)
Controlled Terminology (CT)
Derivations and Imputations (DI)
Reference Tables (Lab/Non-Lab/Questionnaires)
Master Data Domains (MDD)
External Partners

8. Metadata Management – Our Beginning(2009)..
iMDR – Interim MetaData Repository
Contained only global level metadata with no version control.
Metadata was defined and managed in excel spreadsheets for each type of metadata. E.g. CDE, CT, etc.
All files were stored and maintained in a central sharepoint location
Give a high level overview of maintenance within SP

9. Metadata Management – Our Beginning (contd) Domain Sheet - Clinical Data Elements Tab
Data Elements were defined at Domain level. One spreadsheet for every domain referred as the domain sheet. E.g. AE, CM, etc.
The domain sheets contained metadata from collection through submission (SDTM)
ADaM metadata was maintained separately in Excel spreadsheets.
Each domain sheet contained two main tabs: Clinical Data Elements and Derivations/Imputations.
Clinical Data Elements – Contained data elements from collection and submission (SDTM). Each element had 53 attributes that were defined
Derivation/imputations – Contained details on the derivations for elements that were derived or imputed. Each derivation had 19 attributes that were defined.

10. Metadata Management – Our Beginning (contd) Domain Sheet - Clinical Data Elements Tab
The Clinical Data Elements Tab

11. Metadata Management – Our Beginning (contd
Metadata Management – Our Beginning (contd.) Domain Sheet – Derivations/Imputations Tab
This sheet provides information regarding the derivations and imputations used within the data domain.
Derivation: a sequence of steps, logical or computational, from one result to another. The result is a new variable. Example of this method is AGE_DRV - If Date of Birth is present and non- missing then the derived variable will be calculated as Visit Date of Visit 1 minus the Date of Birth.

12. Metadata Management – Our Beginning (contd
Metadata Management – Our Beginning (contd.) Domain Sheet – Derivations/Imputations Tab
Imputation: Creating a value when all or part of the original variable is missing. Example of this method is commonly found within dates
Data collected Mar1960
Data imputed 15Mar1960

13. Metadata Management – Our Beginning (contd.) Controlled Terminology
Defines the codelist choices that are allowed in a domain for a given CDE.
Master controlled terminology was maintained in Excel sheet. Subsets (groups) of codelist values were maintained within CDMS system.
This sheet contained 24 attributes for each value defined in a codelist.

14. Metadata Management – Our Beginning (contd.) QC measures
QC checks – SAS based programs to check for the completeness of the domain sheets based on pre- established rules.
Manual Review

15. Metadata Management – Current (2013)
Metadata defined and maintained in an off-the-shelf Oracle based tool customized for Novartis.
A Central global metadata repository that holds all metadata that is easily accessible to all end users
The same concept from iMDR was adapted into the MDR tool.
The views were customized to match the iMDR sheets for ease of review and use by end users
Following key enhancements were made to the tool:
Enabled workflow process for adding and approving new objects
User friendly screens for codelist and lab reference tables were added
Introduced online validation checks and QC reports
Updated interface views for use by downstream systems

16. Metadata Management – Current
Clinical Data Element View
Derivations/Imputation View

17. Key Benefits Achieved Controlled and Audit trailed environment
Better quality gained by having on-line checks and built in QC reports
Efficiency gained in defining and maintenance of metadata. E.g. Codelist, lab reference tables, etc.
User friendly data entry screens for codelists and reference tables
Efficiency gained by providing improved interface views to the downstream systems
Reduced manual review steps
Instant access to all end users to browse metadata

18. Current Challenges Data entry screens for CDE not user friendly
Batch load functionality requires lot of pre-processing steps due to de-normalized structure of MDR
Option to subset elements and codelists does not exist
Dependent objects and attributes need to be manually created. This entails 40% of the objects within the CDE
Unable to manage study level metadata
Transformation challenges – horizontal to vertical, handling blank records, etc.
Other challenges include Versioning, lack of interface with upstream/downstream systems, ADaM metadata housed outside of MDR, etc.

19. Metadata Management - Future
Support both Global Data Standards and Study level metadata needs
Support global & study level governance workflows
Enable integration with CDMS and CDISC - SHARE
Enable integration with upstream and downstream systems
Create views to support Define.xml requirements
Integrate ADaM Metadata and Structures

20. Metadata Management - Future Proposed process flow

21. Governance workflow model
Request
Decision
Feedback
* CSU encompasses:
Clinical Science Unit
Business Unit (i.e. Oncology)
Translational Sciences (TS)
Requestor
(CTH/CTL)
CSU
Data Standards Governance
Standards Extended Team
Onc. Standards - approves (from a scientific & franchise specific standpoint) requests
DSG - final approval
DSG - Standards Experts who approve requests from CSUs, with input from Extended Team, as required (senior representatives with a background in Clinical Science, Biostatistics, Statistical Reporting, Data Management
Ext. Team - Provides adhoc expertise in specific functional area (e.g. Imaging, lab data, biomarker development)
Oncology - Integrated Disease Area strategy team (IDS)
Requester - Identifies new needs for data element /codelist for safety / indication standards

22. Questions?