Ontology and Controlled Vocabulary in Clinical Trials Javed

Ontology and Controlled Vocabulary in Clinical Trials Javed Mostafa Jane Greenberg Rahul Deshmukh Lina Huang

Outline � Clinical Trials � SPIROMICS � Application of Ontologies and Controlled Vocabularies � Use Cases - Ontologies in Clinical Trials � SPIRO-V : Role of Ontology and Controlled Vocabularies in SPIROMICS � SPIRO-V: Where We Are Now � Controlled Vocabulary Management - The Road Ahead � Demo � Questions?

Clinical Trials � Conducted by Government Organizations, Pharmaceutical Companies, Academic Research Centers etc. � Mostly to assess safety and effectiveness of new medication or device � Types � Treatments - Combination of drugs � Diagnostics � Quality of Life - For patients with chronic illness

Clinical Trials (Contd. ) � Phases � Phase 0 - Protocol , Patient Identification � Phase 1 – Small Group (20 -80) safety & side – effects of drug/treatment � Phase II – Larger Group (100 -300) � Phase III – Large Group (1000 -3000) � Phase IV – Drug’s Risks, Benefits and Optimal Uses � Duration – 6 to 8 years � Cost for pharmaceutical companies between $100 - $800 Million � In 2005, 8000 Clinical Trials, $24 Billion Invested

SPIROMICS Subpopulations and intermediate outcome measures in COPD study (SPIROMICS) � Primary Goals � Identify and validate markers of disease severity � Identify disease subpopulations � Secondary Goals � Clarify the natural history of COPD � Develop bioinformatics infrastructure � Generate clinical, radiographic and genetic data that can be used for future multisite clinical trials

Application of Ontologies and Controlled Vocabularies Ontology , Controlled Vocabulary Communication For people to talk the same language Indexing To improve retrieval and analysis of data Functions Retrieval Browsing Visualization

Use Cases - Ontologies in Clinical Trials � � Locating eligible patients for clinical trials – IBM, Columbia University � Matched patient data to SNOMED-CT ontology � Semantic gulf between raw data and clinician’s interpretation Structural representation of a disease ontology – Influenza Infectious Disease Ontology � � Coverage of Infectious Disease Domain Clinical Trial Data Management System – Cancer. Grid � Model of study OR Dataset Forms, Services, Metadata Registry etc

SPIRO-V: Vision SPIRO-V Clinical Trial Application Ontology Visualization SPIROMICS Knowledge Base Ontology SPIROMICS Controlled Vocabularies Patient Identification Mapping Specimen Tracking Clinical Trial DBMS Controlled Vocabularies Editing/ Management

Where We Are Now Controlled Vocabularies Harvesting Goal - A COPD Vocabulary set to accurately describe all the SPIROMICS cohorts, phenotypes, and outcome measures. Two Approaches � Manual � Automatic

Manual Approach � Manual approach to collect vocabularies from authoritative sources on COPD � Domain experts conduct quality control � Downside: � Low efficiency � communication, coordination takes time

Consolidated Excel Spreadsheet

Automatic Approach �A relational database back end to store the terms, definitions and associations � Incorporation of VCGS automatic metadata generation system for rapid harvesting � Provide human review functionality to control the quality of terms and associations generated. � Manage controlled vocabularies development process

VCGS

Manage Controlled Vocabulary � Visualize vocabulary set and make it browsable, searchable, and editable. � Vocabulary gathering workflow: Suggest candidates -> review -> release/reject � Collaborative discussion initiatives: co-authoring and

Demo �Database • Physical database in My. SQL �VCGS �Tema. Tres �SPIRO-V

Questions?

References � http: //clinicaltrials. gov/ct 2/home � http: //www. cscc. unc. edu/spir/ � http: //iswc 2007. semanticweb. org/papers/809. pdf � http: //influenzaontologywiki. igs. umaryland. edu/wiki/index. php/Main_Page � http: //www. cancergrid. org/

Collaborative thesaurus editing �Level of user privileges • Common users can suggest a candidate term, an association or a definition and provide feedback • Authorized users can reject/accept the suggestions. • Document changes and comments from different users.

Ontology, Thesaurus, Controlled Vocabularies Ontology Thesaurus Where We Are Controlled Vocabularies