Implementation of a Clinical Data Repository March, 2009 Pat Majcher, Standards patricia.majcher@novartis.com
Agenda Overview of the New Landscape High-level View Objectives Implementing LSH High-level Data Flow Interfaces Some of the Questions and Issues Next Steps 2 Clinical Data Repository March, 2009
Overview of the New Landscape Electronic Protocol Data Capture Modeling& Simulation Projectspecific Metadata Library Database Data Warehouse Global Metadata Repository Define.xml Derived Data Analysis & Reporting System Programming Environment Analysis Datasets Document Management System Tables, Listings, and Graphs Submission Datasets SDTM & ADaM Submission Package (FDA) 3 Clinical Data Repository March, 2009
High-level View Clinical data repository Data warehouse environment, source of all collected data and relevant analysis data Enable end users to focus more on business intelligence and analytics, rather than technical details Functionality includes - Ability to create pooled data specific to user needs (data marts on demand) - Versioning of data, models; audit trail - Different work areas, specific to each business need Oracle s Life Science Data Hub (LSH) selected 4 Clinical Data Repository March, 2009
Objectives Create a data repository that can Integrate various sources of clinical data Store the data in a standardized structure Perform simple standard derivations, imputations, merges, etc. Provide simplified access for users, eliminating the need to use multiple systems to access all data Reduce manual intervention and facilitates data pooling Primary users Programmers and statisticians working on submissions, health authority requests, etc. Modeling and Simulation personnel performing secondary analyses 5 Clinical Data Repository March, 2009
Implementing LSH (1/2) Working with external consultants to provide expertise Internal resources expert at SAS and SQL, but needed additional time/training for Oracle and LSH conventions Product installed successfully in test environment Initial difficulties, installation not straightforward E.g., user laptops needed additional software utilizing Citrix solution to avoid this, adding software only to the Citrix server Subteams established to address different aspects of the project (e.g., process, technical, data) How to work with LSH? how can it be set up to meet Novartis requirements in the most efficient way? 6 Clinical Data Repository March, 2009
Implementing LSH (2/2) Define how to structure the contents of LSH Define domain structure: Hierarchy starting with global data and metadata, down to the level of projects, pools, and studies Execute Proof of Concepts (PoCs) Goals - Confirm the usability of LSH - Acquire additional information and experience to design LSH environment What is the business process? the technical steps? the architecture design? - Involve the stakeholders and gain their buy-in Each PoC tests a different function or process - Stage data, conform data, pool data, execute imputations, load from Oracle Clinical, load SAS data, create data marts, perform ad hoc queries - All successfully completed 7 Clinical Data Repository March, 2009
High-level Data Flow Stage the collected data, global metadata, and trial-level information in LSH Conform data, as needed, to company standard Create value-added data E.g., basic derivations, imputations, etc. Transfer data to statistical and programming environment Analysis and reporting system will create SDTM- and ADAMcompliant data sets from the input data - ADaM in Release 1, SDTM in Release 2 ADaM data will be transferred to LSH Create project-level data pools, other custom pools/data 8 Clinical Data Repository March, 2009 marts
Interfaces Input sources/systems Clinical database management system, Oracle Clinical Remote Data Capture Data from external sources (e.g., Contract Research Organizations) Derived data (e.g., ADaM) Metadata repository Electronic protocol/trial-level information Output sources/systems Statistical and programming environment Analysis and reporting system (creating data sets, tables, etc.) Querying and reporting tools 9 Clinical Data Repository March, 2009
Some of the Questions and Issues (1/3) How to handle special trial situations? Incremental loads for ongoing studies, blinded data, extensions How is a transfer initiated? push or pull model? owner? Where to place transformations that could be done in either database management system or LSH? How to reuse work done in other systems? Some derivations/imputations implemented in analysis and reporting system, since its schedule was ahead of LSH s 10 Clinical Data Repository March, 2009
Some of the Questions and Issues (2/3) Working with SAS data sets in an Oracle environment Initially, expected to use SAS programs Testing showed that PL/SQL was more efficient to run within LSH Network can be an issue Server is in Switzerland, requests initiated at user s computer Some interfaces cannot yet be tested as solutions are pending (metadata repository, electronic protocol) Warehouse data model Novartis internal standard? Janus? SDTM? Version control of data standards 11 Clinical Data Repository March, 2009
Some of the Questions and Issues (3/3) Many of the functions Novartis needed have to be created or customized E.g., link to metadata repository, conformance checks, data pooling Query tools Security and access management LSH is an empty container that can be configured in different ways, with separate areas for different groups Validation checks for loading etc 12 Clinical Data Repository March, 2009
Next Steps Project subteams continue to address open issues Project team coordinates with related projects through a higher-level program forum Integration / interface issues brought to forum for discuss and decision Multiple releases planned Release 1: covers the submission process, target IVQ2009 Releases 2 and 3: enhanced functionality for secondary analyses and additional users like Modeling and Simulation 13 Clinical Data Repository March, 2009
Questions? Comments? Thank you for your time and attention 14 Clinical Data Repository March, 2009
Backup Slides 15 Clinical Data Repository March, 2009
Acronyms ADaM = Analysis Data Model CDISC = Clinical Data Interchange Standards Consortium LSH = Life Sciences Data Hub MDR = Metadata Repository SDTM = Study Data Tabulation Model 16 Clinical Data Repository March, 2009