Structured eligibility criteria of breast cancer clinical trials

This data set contains semi structured eligibility criteria of nearly 300 breast cancer trials from ClinicalTrials.gov. The structured representation consists of the detected in text: patterns (e.g. "No prior ()"), UMLS concepts and normalized measurements. The criteria are classified according to several dimensions which describe the meta information of temporal status, time independent status, specification type and medical content. The metadata can be used to define fine grained queries. Additionally, some criteria are interlinked witch each other by a transitive relation "isMoreStrict/isMoreRelaxed" - indicating which criterion can in theory be satisfied by a larger number of patients. The method for criteria structuring has been described in [1]. The method for creating a library and comparing the restrictiveness of criteria has been described in [2]. An interactive visualization of the library, prepared by Rinke Hoekstra can be accessed at http://semweb.cs.vu.nl/eligibility [3]. The library can be queried using the following SPARQL endpoint http://semweb.cs.vu.nl:8080/openrdf-sesame/repositories/ct [4] Below are listed some examples of SPARQL queries. They can be used to: 1. Find criteria which mention a concept with a specific UMLS CUI SELECT DISTINCT ?text WHERE {<br>?trial a etv:Trial .<br>?trial etv:hasCriterion ?criterion .<br>?criterion etv:hasOriginalText ?text .<br>?criterion etv:hasContent ?PatternInstance .<br>?PatternInstance etv:hasContent/etv:hasConceptId ?id .<br>filter regex(?id, "C0029925")<br>} 2. Find criteria corresponding to a specific pattern e.g. "At least () since prior ()".<br>SELECT DISTINCT ?trial ?text WHERE {<br>?x rdf:type .<br>?x etv:occurrsIn ?criterion.<br>?criterion etv:hasOriginalText ?text .<br>?criterion etv:isUsedInTrial ?trial<br>} 3. Get the complete list of patterns: SELECT DISTINCT ?x WHERE {<br>?x rdfs:subClassOf etv:PatternInstance<br>} 4. Find criteria which mention a drug in a specific context e.g. as forbidden prior treatment. SELECT DISTINCT ?trial ?criterion ?text WHERE {<br>?trial a etv:Trial .<br>?trial etv:hasCriterion ?criterion .<br>?criterion etv:hasOriginalText ?text .<br>?criterion etv:hasContent ?PatternInstance .<br>?PatternInstance etv:hasTimeIndependentStatus etv:absent .<br>?PatternInstance etv:hasTemporalStatus etv:current . ?PatternInstance etv:hasContent/etv:hasConceptId ?id .<br>filter regex(?id, "C0039286")<br>} 5. Find criteria of specific semantic type (as defined by the type of a mentioned concpets by UMLS).<br>For instance find criteria related to lab results.<br>A list of semantic type is available at: http://metamap.nlm.nih.gov/Docs/SemanticTypeMappings_2011AA.txt SELECT DISTINCT ?trial ?text WHERE {<br>?trial etv:hasCriterion ?criterion .<br>?criterion etv:hasOriginalText ?text .<br>?criterion etv:hasContent/etv:hasConcept/etv:hasSemanticType r:lbpr<br>} 6. Find criteria that are less strict than a provided one e.g. "bilirubin less than 1.5". SELECT DISTINCT ?trial ?text WHERE {<br>?x etv:isMoreRelaxed ?y .<br>?y etv:occurrsIn/etv:hasOriginalText ?textS .<br>?x etv:occurrsIn ?c .<br>?c etv:isUsedInTrial ?trial .<br>?c etv:hasOriginalText ?text .<br>filter regex(?textS, "bilirubin less than 1.5","i")<br>}