CDISC MDR Definitions & examples on concepts & variables

1. CDISC MDRDefinitions & examples on concepts & variables December 2008

2. Foreword All examples taken in this presentation are extracted from BRIDG 1.1. For the purpose of the presentation, the release is not important. But while looking a the examples must take into account that there might be differences with the last release of BRIDG

3. Table of Content • What is CMDR – architecture & definition • Examples • Open questions

4. What is CMDR • Value proposition • Architecture • Definitions

5. What is CMDR: value proposition The CDISC Meta Data Repository (CMDR) is a “cross industry enriched data dictionary”, supporting full semantic interoperability. It will allow cost avoidance in each pharma companies (decrease cost for standards maintenance and data mapping) while improving data quality, critical for effective scientific decision making. In addition it will enable new capabilities required by authorities – such as comparative profile for safety and cost-efficacy – and will support EHR integration.

6. What is CMDR: content • CMDR is composed by concepts and variables • Concepts are • Abstract ideas associated to a representation i.e. a variable • Related to each others in an information model/ontology (based on BRIDG and on HL7/ISO abstract data type)Note: BRIDG fully bound to controlled terminologies for all the attributes specified as CD can be considered – at least partially – as an ontology for clinical research • Unambiguous as their semantic is fully defined • Used and defined at scientific level (generally in protocol) • Variables • are linked to concepts and • are used to collect, derived and store all information related to a concept in a computer • Note: variables can be considered as the equivalent of what company currently have in terms of data dictionary • CMDR content must be controlled through strict Governance

7. code code code Terms(meaning) Terms(meaning) Terms(meaning) Sponsor OwnMDR Sponsor OwnMDR Sponsor/CRO Regulators / HealthCare Own MDR What is CMDR: architecture Governance Scientific View(clinical & non-clinical concept) CDASH CDISC MDR CDISC MDR Terminologies Information Model Layer (BRIDG) SNOMED CDISC Vocab Data Manager & Biostat view(variables) SEND Data Layer (concepts & variables) SDTM Data Type Layer ADaM MEDDRA LOINC Data Integrator view

8. CDISC MDR What is CMDR: definitionsInformation Model Layer An information model is composed by objects, further described by properties and relationship with other objects. Properties have an abstract data type. CDISC MDR CDISC MDR Information Model Layer (BRIDG) Data Layer (concepts & variables) Data Type Layer

9. What is CMDR: definitionsInformation Model Layer - BRIDG CDISC MDR Concepts are totally defined (e.g SubjectId) => direct linking to variables • Concepts are defined generically (e.g. PerformedObservationResult) • => need to specify each “instance” of concept • => indirect linking to variable

10. CDISC MDR What is CMDR: definitionsData type Layer ISO/HL7 abstract data types are data type definitions for representing and exchanging basic concepts that are commonly encountered in healthcare. Data types used in BRIDG CDISC MDR CDISC MDR Information Model Layer (BRIDG) Data Layer (concepts & variables) Data Type Layer

11. CDISC MDR What is CMDR: definitionsData Layer • The data layer is composed by concepts and their associated variables • Concept can be complex or simple • a complex concept is composed by other concepts (complex or simple) through a structured organizational framework or ontology – i.e. the information model with controlled terminology • VScontainsWeight, Height, systolic blood pressure… • Blood pressure: 2 possibilities • Standing systolic blood pressure upper armcontainssystolic blood pressure, position = standing, site = upper arm • Systolic Blood pressure with conditional concepts & variable such as position and site • a simple concept may be related to many different complex concepts and must be linked to all variables capturing the meaning related to the concept • Systolic blood pressureis linked toSYS_BP_val, SYS_BP_unit, SYS_BP_precision • Person Gender is linked to Person_Gender CDISC MDR CDISC MDR Information Model Layer (BRIDG) Data Layer (concepts & variables) Data Type Layer Semantic of concepts must be unique and captured in structured way, following ISO11179

12. CDISC MDR What is CMDR: definitionsData Layer • Observation • Name: CD • Position: CD • Site: CD • Value: any • Time: TS Information Layer Qualifier = VS domain (explicit context) When qualifier is VS, then synonym of position = VSPOS and site = VSLOC VSPOS and VSLOC would point to “SYSBPLOC and TMPLOC”…. Based of the VSTestname value In caDSR = Observation.name.SYSBP In caDSR = Observation.name.SYSBP.position.list of restricted CD In caDSR = Observation.name.SYSBP.site.list of restricted CD In caDSR = different representation (integer & mmHG, real& cmHG) SYSBPLOC SYSBPSITESYSBPVAL, SYSBPU SYSBPTimeofMesurement SYSBPLOC SYSBPSITESYSBPVAL, SYSBPU Data Layer • ID1_SYSBP • Name = SYSBP • Position: restricted CD from observation • Site: restricted CD from observation • Value: PQ • Time: TS TMPLOC TMPSITETMPVAL, TMPU • ID2_TMP • Name = TMP • Position: restrictedCD from observation • Site: restrictedCD from observation • Value: PQ Data Type Layer PQ = integer, unit, range TS = …. CD = code list, source, ….

13. Examples of simple concepts and related variables • One simple concept, one variable • One simple concept, several variables

14. Example within ISO11179 (1 simple concept = 1 variable, concept fully defined in BRIDG) Conceptual Domain GENDER C1913 Valid Values Female C246 Male C1323 Unknown C2691 … … Object Class Person C1892 Value Domain GENDER value Information Model BRIDG Data Element Concept Person.gender There is a black box behind this (see next slide). CDISC MDR data standard stewards need to understand it in details. End users are not interested Property gender C42614 Representation Gender code C42614 Data Element (= CONCEPT in sense of CMDR): Person.Gender Data type: CDSemantic Signature: C1913.C1892.C42614.C42614.C246.C1323…. (ConceptualDomain.ObjectClass.Property.Representation.Values) Code lists F, Female ,C246 M. Male, C1323 U, Unknown, C2691 … … VARIABLE: personGenderData type: CD

15. Example within ISO11179 (1 simple concept = 1 variable, concept fully defined in BRIDG) Conceptual Domain GENDER C1913 Valid Values Female C246 Male C1323 Unknown C2691 … … Object Class Person C1892 Value Domain GENDER value Information Model BRIDG Data Element Concept Person.gender Property gender C42614 Representation Gender code C42614 Data Element (= CONCEPT in sense of CMDR): Person.Gender Data type: CDSemantic Signature: C1913.C1892.C42614.C42614.C246.C1323…. (ConceptualDomain.ObjectClass.Property.Representation.Values) Code lists F, Female ,C246 M. Male, C1323 U, Unknown, C2691 … … VARIABLE: personGenderData type: CD

16. What is CMDR: example within 11179 (1 simple concept = many variables, concept fully defined in BRIDG) QUESTIONSWhat is the conceptual domain for data type PN ??? For for type PQ ??? Can we establish a formal relationship between a ADT and the related variables, based on the ITS of the datatype ? For instance PQ always associated with 4 variables: value, unit, range, precision… Conceptual Domain Person Name C1913 Valid Values …….?????? Object Class Person C1892 Value Domain Person Information Model BRIDG Data Element Concept Person.gender Property name C42614 Need to specify content of the different boxes Representation Gender code C42614 Data Element (= CONCEPT in sense of CMDR): Person.Gender Data type: CDSemantic Signature: C1913.C1892.C42614.C42614.C246.C1323…. (ConceptualDomain.ObjectClass.Property.Repsentation.Values) Code lists F, Female ,C246 M. Male, C1323 U, Unknown, C2691 … … VARIABLE: personGenderData type: CD

17. PerformedObservationResult.name What is CMDR: example within 11179 (1 simple concept = 1 variable, concept defined generically in BRIDG) QUESTIONSHow do we link a Generic object like PerformedObservation result with a specific concept –e.g. systolic blood pressure How/where do we link all code value that are contrained toegther (e.g. Bodyposition for Sys BP has only a limited set of value)=> do we need to create a fully instantiated version of BRIDG (for all PerfomedObservationResults – which would constitute the concepts – out of which we can then create the variables ???/(this can be stored in ISO11170 but should be displayed in a user friendly way – with graphics If we do this – we also need to create some kind of structure (e.g. BP is subspecialized in SYSBP and DIAS BP) Conceptual Domain GENDER C1913 Valid Values Female C246 Male C1323 Unknown C2691 … … C1818 Object Class Person C1892 Value Domain GENDER value Information Model BRIDG Data Element Concept Person.gender Property gender C42614 Need to specify content of the different boxes Representation Gender code C42614 Data Element (= CONCEPT in sense of CMDR): Person.Gender Data type: CDSemantic Signature: C1913.C1892.C42614.C42614.C246.C1323…. (ConceptualDomain.ObjectClass.Property.Repsentation.Values) Code lists F, Female ,C246 M. Male, C1323 U, Unknown, C2691 … … VARIABLE: personGenderData type: CD

18. PerformedClinicalObservationResult. numericResult What is CMDR: example within 11179 (1 simple concept = many variables, concept defined generically in BRIDG) Conceptual Domain GENDER C1913 Valid Values Female C246 Male C1323 Unknown C2691 … … C1818 Object Class Person C1892 Value Domain GENDER value Information Model BRIDG Data Element Concept Person.gender Property gender C42614 Need to specify content of the different boxes Representation Gender code C42614 Data Element (= CONCEPT in sense of CMDR): Person.Gender Data type: CDSemantic Signature: C1913.C1892.C42614.C42614.C246.C1323…. (ConceptualDomain.ObjectClass.Property.Repsentation.Values) Code lists F, Female ,C246 M. Male, C1323 U, Unknown, C2691 … … VARIABLE: personGenderData type: CD

19. Examples of complex concept within “ontology” • Inclusion/exclusion criteria • Procedures • Combination of procedures around disease • eCRF Forms • Standard Data Sets (STDM, ADaM, …)

20. Complex concepts • Complex concepts are composed by simple or complex concepts. The same simple concept – with potentially a different code list or value – can be used for different concepts • So far the following complex concepts that would be valuable to standardize across the industry have been identified • Inclusion/exclusion criteria • Procedures • Combination of procedures around disease • SDTM domains • eCRF Forms • Standard Analysis Data Sets

21. Complex concept: inclusion/exclusion criteria • Most inclusion/exclusion criteria would be a simple concept such as • Age < or > a certain value • Tumor size < or > a certain value • Note: the actual criteria will be a combination of the concept/variable with a specific value (e.g. age below 20 years) …..CMDR provides the definition of the concept/variable (e.g. age), the data integrator provides the value (e.g. below 20 years) • Some inclusion/exclusion criteria – including several check – are complex concept • ???? Example ???

22. Complex concept: Forms ??? Concept: Depression Concept: Anxiety Concept: Positive, Negative, or Affective Symptoms of Psychotic Disorder • Variables • HAMD Total Score • AETERM: “Depressed/Depression/Depressive Symptoms” • TRT+IFU: “Antidepressant for Depression” • BPRS Item 9 • Variables • BAI Total Score • HAMD-D Item 11 • AETERM: “Anxious/Anxiety/Panic” • TRT+IFU: “Anxiolytic for Anxiety” • BPRS Item 2 • Variables • BPRS Total Score EQ5D Question 5 “Are you Anxious or Depressed?” BAI Items HAMD Items BPRS Items BAI Total Score HAMD-D Total Score BRPS Total Score

23. Complex concepts: procedureConcept interdependencies: concept with testing CRF Study design Concept Sample Finding • CRF template(s) • Visit(s) • Data collector role(s) • Variable(s) with permissible values • Sample type(s) • Collection date/time • Volume/Weight • Quality • Storage condition • Location • Result value • Result unit • Measurement date/time Other concept • Simple or complex concept Mandatory Optional

24. HPVAmplicor HPVAmplicor HPVTherapeuticVaccine Complex concepts: procedureConcept interdependencies: HPV treatment Human papilloma virus Detect Disease Select Treatment Treat Disease Monitor Response • HPV causes cervical cancer • Cervical cancer is2nd biggest cause of female cancermortality • Up to 250,000 deaths / year HPVLinear Array Source: WHO

25. Complex concepts: procedureHPV treatment - Step 1 – Disease detection CRF Study design Concept Sample Finding • CRF template(s) • Visit(s) • Data collector role(s) • HPV detection • Plasma • Collection date/time • Volume/Weight • Quality • Storage condition • Location • HPV Amplicor Test • Result value • Result unit • Measurement date/time Other concept • Demographics • Disease history • Treatment history Mandatory Optional

26. Complex concepts: procedureHPV treatment - Step 2 – Disease classification/treatment selection CRF Study design Concept Sample Finding • CRF template(s) • Visit(s) • Data collector role(s) • HPV disease classification / treatment selection • Plasma • Collection date/time • Volume/Weight • Quality • Storage condition • Location • HPV LR Test • Result value • Result unit • Measurement date/time Mandatory Optional

27. Complex concepts: procedureHPV treatment - Step 3 – Treatment CRF Study design Concept • CRF template(s) • Visit(s) • Data collector role(s) • Disease Treatment by therapeutic vaccination Other concept • Weight measurement Mandatory Optional

28. Complex concepts: procedureHPV treatment - Step 4 – Disease monitoring CRF Study design Concept Sample Finding • CRF template(s) • Visit(s) • Data collector role(s) • Disease monitoring • Plasma • Collection date/time • Volume/Weight • Quality • Storage condition • Location • HPV Amplicor test • Result value • Result unit • Measurement date/time Mandatory Optional

29. Open questions • ….

30. Around Time related variable • SDTM is having one variable for data+time • CDASH is decomposing this in 2 or more variable • Should we have the variables defined only at the most granular level – related to a hierarchy of concepts • Or should we have as well a hierarchy of variable • What is the +/- of both ?

31. Systolic blood pressure – and “concept code list” • What is the conceptual domain for the concept systolic blood pressure – is this the Data type (i.e. PQ) ? • concept = systolic blood pressure – which is a PQ (physical quantity). A PQ is implemented through a value, unit, range, precision • related variables are hence SYSBP_VAL, SYSBP_unit, SYSBP_range, SYSBP_precision; • this does not include position and location which are part of a “complex concept” called “clinical measurement” including indeed position, site but also devices, …. • Measurement unit and systolic blood pressure • the "concept" measurement unit is related to a very long list of code • for systolic blood pressure we want only a subset of this. I believe this is where the "implementation" of BRIDG comes into place i.e. we will specify for each possible measurement like systolic blood pressure and all ... exactly which value applies.... though we need to show how this will practically work !!! • How to represent the different concept (SYS_BP standing UA, SYS_BP_supine…)