1. From PDF to RDF – Representing the CDISC Foundational Standards
Geoff Low
Frederik Malfait, IMOS Consulting
Scott Bahlavooni, d-Wise
Mitra Rocca, FDA
CTS Ontology Workshop, 2015

2. Agenda Regulatory Submissions Overview CDISC Foundational Standards
PhUSE Computational Science Symposium Working Groups
CDISC Foundational Standards in RDF
Further PhUSE Semantic Technologies Projects

3. Regulatory Submissions, Current State
DM:
- USUBJID: 101
- SEX: F
- BRTHDTC:
Demog:
- SUBJ: 101
- GENDER: FEMALE
- DOB: 02/22/1975
Pharma Co.
BioTech Inc.
Current state of regulatory submissions
- Standards are encouraged, but not required
- Data can be provided electronically or submission can be all paper-based.

4. Regulatory Submissions, Future State
06-Feb-2014, FDA issues (draft) Guidance for Industry: Providing Regulatory Submissions in Electronic Format
06-Feb-2014, FDA issues (draft) Guidance for Industry: Providing Regulatory Submissions in Electronic Format – Standardized Study Data
Standards requirements per FDA Data Standards Catalog
CDISC tabulation, analysis, and terminology standards are the current supported standards for study data
Required 24 months after final guidance (~2017/2018)
Discuss guidance.
- Note that industry has been expecting this for YEARS.
- Note that majority of industry provide subject data in CDISC format; however, non-clinical data is currently not provided in standard format.
- Highlight that supported standards are per FDA Data Standards Catalog and that catalog is subject to change.
- Note that in addition to guidance, FDA issued a Study Data Technical Conformance Guide that provides sponsors additional information relating to the implementation of CDISC standards.

5. CDISC Foundational Standards
Clinical Data Interchange Standards Consortium (CDISC) Mission: “…to develop and support global, platform-independent data standards that enable information system interoperability to improve medical research…”
Provide an overview of CDISC and the different standards
- Mention SDTM, SEND, ADaM, and CT are the standards in the referenced in the FDA Data Standards Catalog
At the end, say…but…

6. CDISC Foundational Standards
Provide an overview of CDISC and the different standards
- Mention SDTM, SEND, ADaM, and CT are the standards in the referenced in the FDA Data Standards Catalog
- Metadata Registry
At the end, say…but…

7. Machine-Processable CDISC Foundational Standards?
CDISC Foundational Standards are not machine processable
Note the types of docs CDISC publishes…model docs, IGs, supplements, Excel workbooks of metadata (not always published)
CPATH initiative to develop 55 TA standards in 5 years
This requires industry to separately develop their own “machine-processable” representations of CDISC Foundational Standards.
CDISC SHARE is supposed to provide a machine-processable representation of CDISC FS, but this is still 2 to 4 years away from a full-production implementation and initial output of SHARE R1 will be CSV.
This is where PhUSE CSS comes in

8. PhUSE CSS Working Groups
PhUSE Computational Science Symposium (CSS) Mission:
“…bring together academia, industry, technology providers and the FDA to collaborate on projects to address unmet computational science needs.”
Overview of CSS collaborations noting “unmet computational science needs”
4 WGs
PhUSE website and dashboard
Overview of ST project
- Note how representation of CDISC standards in RDF helps to address unmet computational science need for machine-processable representation of CDISC standards.

9. PhUSE CSS Working Groups
Semantic Technologies Project:
Investigate the application of W3C semantic standards to support the clinical and non-clinical data life-cycle from protocol development to submission to regulatory agencies
Overview of CSS collaborations noting “unmet computational science needs”
4 WGs
PhUSE website and dashboard
Overview of ST project
- Note how representation of CDISC standards in RDF helps to address unmet computational science need for machine-processable representation of CDISC standards.

10. Representation of CDISC Foundational Standards in RDF
Project Overview:
Representing existing CDISC standards in RDF provides a foundation for interoperable end-to-end data standards in clinical and non-clinical research
Separate sub-teams represented individual standards and leveraged CDISC community standards SMEs as needed
Separate models were consolidated, vetted within the sub-team co-leadership, and a draft posted to GitHub
Collaborate with CDISC to have the RDF representations adopted as CDISC standards

11. Representation of CDISC Foundational Standards in RDF
Data Capture
Data Tabulations
Data Analysis
CDASH v1.1
SEND IG v3.0
SDTM v1.2, v1.3
SDTM IG v3.1.2, v3.1.3
ADaM v2.1
ADaM IG v1.0
CDISC Controlled Terminology
Note that NCI publishes CT in RDF already.

12. CDISC Foundational Standards: RDF Walkthrough
Based on ISO standard for metadata registries
Schemas define classes and predicates for ontology to represent CDISC standards
RDF datasets in Turtle (ttl) for each CDISC data standard and terminology standard
The core element of the Metadata Model is that of an Administered Item, which may represent any item that is managed in an MDR.
An administered item must be named and defined within a Context that defines the scope within which the administered item has meaning.
A key administered item is that of Data Element that represents a generic unit of data.
A Classifier or a defined subclass thereof offers a generic way to classify administered items.
This applies in particular to data elements, e.g. an SDTM data element may be classified as Character or Numeric, or also according to compliance classifiers such as Required, Expected, or Permissible. A term of the controlled terminology of a data element is defined by a Permissible Value that is part of an Enumerated Value Domain.
Walk through this on the next slide using SDTM.

13. CDISC Foundational Standards: RDF SDTM IG v3.1.2 Walkthrough

14. CDISC Foundational Standards: RDF SDTM IG v3.1.2 Walkthrough

15. CDISC Foundational Standards: RDF SDTM IG v3.1.2 Walkthrough
Explain attributes and context

16. CDISC Foundational Standards: RDF SDTM IG v3.1.2 Walkthrough
Explain…relationship btw. Item and controlled terminology
At the end explain that other CDISC standards are modeled similarly to SDTM and segue into PRM/SDM

17. Step 1: Publication Standards available on Github

18. Step 2: Documentation

19. Step 3: Acceptance
The CDISC SHARE Metadata Registry will support export of RDF representations matching those developed within the PhUSE ST Foundational Standards workgroup.
The NCI controlled terminology representation was updated to match the schema developed.

20. PhUSE ST Working Groups Activities
Representation of Analysis Results Metadata to Support Clinical and Non-Clinical Applications
Representing a Clinical Development Plan in RDF
Representation of Regulations and Guidance in RDF
Representation of the Protocol using RDF
Schedule of Activities model in RDF (using the SDM)
Representation of Eligibility Criteria in RDF
Structured Document Representation using RDF
Document Use Cases for Linked Data in the Clinical Trial Life-Cycle
keyCRF – semantically rich CRFs for EHR integration
determining the logical model for the representation of analysis results and their associated metadata for clinical and non-clinical applications – using RDF Data Cube
Develop an RDF information model to capture, retain, reuse and share the design of clinical programs:
Design inputs such as claims, target product profile, development plans etc.
Design elements such as objectives, endpoints, treatments, patient population, etc.
Options being considered for each design decision
Rationale for each design decision, including related information, guidance, best practice
Various sources (e.g. CDISC standards) refer to regulations and guidance documents. Representing the regulations and guidance documents in RDF enables the linking of said documents to RDF sources
Protocol
Study Design Model – data and metadata about a study design (includes activities, trial parameter, transitions, etc)
Evaluate existing models for Eligibilty Criteria - Eligibility Criteria Model (ECM) described in the TRANSFoRm Project "D6.2 Clinical Research Information Model”
Evaluate structured document representations (eg DITA, DocText) – specifically looking at the representation of the common protocol template – developed by TBI
Flat tables are boring, and mostly ineffective – how can we use linked data to improve the way we capture, represent and analyse clincal trial data
Using the SALUS RDF Metadata repository

21. Conclusions
We have been able to bring industry and standards development organizations around to the value of Semantic Technologies.
There are a number of interesting projects (with active communities) ongoing looking at how we can use Semantic Technologies to improve the way clinical research is done.
Better trials, better drugs and better Patient Outcomes