October 2004
The Implementation of CDISC
(SDS V3.1) standards within a CRO
Everyone uses them, we all think we understand them and we'd like to think we all keep up to date with the current versions. I'm talking about our industry standards, but what we don't often say is that we're probably all following our own organisation's interpretation of these standards.
In July this year, a guidance document was issued from the FDA (www.fda.gov/cder/regulatory/ersr) outlining their specification for the use of the CDISC (Clinical Data Interchange Standards Consortium) Study Data Tabulation Model (SDTM) for future electronic submissions of clinical data.
How will your company handle this change in the way clinical trial data are collected and what impact will it have on the relationship between the patient, the CRF, the Database and the Regulators?
Scary, don't you think?
Through their data models, CDISC (www.CDISC.org) has been at the forefront of raising awareness of standards over the last few years. Since starting up as a focus group of volunteers in 1997, it has become a thriving non-profit organisation, lead by their president, Rebecca Kush, and an elected board of directors. They now have over 150 members made up of pharmaceutical/biotechnology companies and CRO/Service/IT providers.
They see a large number of benefits to our industry from having an established set of standards:
Increases the efficiencies in performing clinical trials
Facilitates 'business' processes among investigators, pharma companies, CROs, vendors & laboratories
Streamlines high quality data collection at investigator sites due to consistency of requirements
Allows for software development against a common standard
Provides a long-term means for electronic data archiving
Facilitates FDA/EMEA review of submissions
Improves links between healthcare delivery and clinical research
From a Scientific standpoint, the advantages are equally numerous:
Facilitates integration of data across trials, compounds and companies (especially within the FDA)
Enables quick assessment of trends and patterns in data
Results in improved public health and better guidance for studying disease states
Allows better safety monitoring and pharmacovigilence by supporting ad hoc analyses and data review
Ultimately, the aim is to allow our companies to focus more on the science surrounding the test compound and less on our data management and reporting activities.
slide courtesy of CDISC
As a CRO, Quintiles has worked closely with CDISC to ensure that we can offer our customers a seamless reporting solution, the goal being to reduce submission review times for future drug approvals within the major agencies around the world (FDA, EMEA). In June 2004, Quintiles became the first global provider of the full range of drug development and marketing services to be certified as a CDISC Registered Solutions Provider for managing data submitted for the approval of new medicines.
Since the first data models were introduced there have been a number of changes, which have deterred some companies from investing time in implementing them. The largest changes have effected the ODM (Operational Data Model) and SDS (Submission Data Standards). Each model has been developed by a sub-team of people, who also work on future enhancements to the models, through a consensus-based, public review. Major changes are then balloted through HL7 (Health Level 7, www.hl7.org) before eventually being released.
Quintiles © 2004
This screenshot above shows and exple of one of our CDISC - compliant pages
The current models are:
Operational Data Model (ODM)
Production Version 1.2
Refers to the Collection, Interchange and Archiving of Operational Data CRF/Raw data in our various database systems
Uses XML (eXtensible Mark-up Language) to program the ODM code (XML – vendor neutral, platform independent, used to program the World-Wide-Web)
Laboratory Data Model (LAB)
Production Version 1.0.1, includes new Microbiology module and ECG extension
Submissions Data Standards (SDS)
Production Version 3.1
Now in two-parts; Submission Data Tabulations Model (SDTM) and detailed Implementation Guide
Refers to the data flow from the operational database to regulatory submission
As the SDS model is now made up of two parts, future amendments to the SDTM portion should not require the issuance of a completely new version of the model
Analysis Dataset Models (ADaM)
Guidelines and examples of Analysis Data sets
Protocol Representation Group (PRG)
Creating a model to electronically capture protocol information in a machine-readable format
Standard for Exchange of Non-Clinical Data (SEND)
Currently review version 1.6
Focusing on data collected from animal toxicology studies and its representation to the FDA.
slide courtesy of CDISC
Within Quintiles, we have already implemented CDISC Submission Data Standards (SDS V3.1) and Laboratory Data Standards (LAB) for a number of clients and have been providing CDISC with feedback on our practical experiences. In addition, a Global Working Party was created within Quintiles in May 2003 to look into CDISC and to produce an Implementation Plan for this initiative within our business. The plan was broken down into five parts and we meet every month to discuss progress.
Quintiles Implementation Plan Summary:
Interpretation of Data Items – creation of a mock study and test Oracle Clinical and Clintrial builds
Technical Considerations – Global library creation, instance set-up and protocol migration
SOP Changes – naming conventions
Contract/Budget Issues – what savings can be passed onto our clients?
Training Requirements – who needs to know what within our business? (PM, CDC, DBA)
Within our working party are a number of sub-teams, focusing on electronic CRF annotations, business development/costing and analysis structures. We created the mock study 'Hyperdrug' and produced an annotated CRF booklet of visits and the subsequent database builds (Oracle Clinical and Clintrial). We also introduced the concept of QCORE to work out which variables would be collected by DM or derived by Statistics.
The main focus for most companies is how should the data be submitted? The way we design/annotate CRFs and the resulting export of data to statistics will be the areas impacted the most. The changes from V2 to V3.1 of the SDS model largely affect the underlying structure and include additional domains (Questionnaires, Relates, Supplemental Qualifiers, etc).
Horizontal models (V2) were best suited for SAS flatfile formats:
Intuitive presentation with more descriptive variable metadata (browser/analysis ready), but there was too much variation in the structures
There were 10 Domain structures to represent only most common safety and demographic data: DEMO, AE, CONMEDS, DISPOSIT, ECG, EXPOSURE, LAB, MEDHIST, PE and VITALS
The current vertical models (V3.1) are more flexible for storing diverse data in only a few standard structures:
Easier to validate and load in data repository, but less intuitive for current viewing tools and require use of controlled terminology
V3.1 has three new structures to represent everything: Interventions, Events and Findings. There is also a group called 'Other', which covers the areas shown in the diagram at the top of page 23.
The mock studies built at Quintiles are compliant with the newly approved standard, and have been a useful training resource for DBAs and CDCs alike. Moving forward, Quintiles plans to develop standard validation checks and queries based on the V3.1 model to enhance our global data libraries. Based on guidance from the FDA, Quintiles is also looking to facilitate the transition from SAS to XML transfers.
The message from CDISC is a very positive one. Data standards improve quality and speed and their hope is that future clinical trial submissions will be much smoother allowing these savings to be passed on to patients around the world.
Jason Turner, Associate Group Manager
Clinical Data Management, Quintiles, Edinburgh