Upload
verity-briggs
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
NCICB JamboreeFebruary 25, 2005
NCICB Clinical Trials Modeling Activities
Part 1- CDISC Modeling Activities
Part 2- Clinical Trials Modeling (caBIO extension)
CDISC Modeling Effort - Background
Clinical Data Interchange Standards Consortium (CDISC) is a non-profit organization committed to development of industry standards to support electronic exchange of clinical trials data and metadata.
HL7 provides standards for data exchange to allow interoperability between healthcare information systems. The Clinical Trials technical committee of HL7 (RCRIM) has it’s own models for clinical trials
To achieve true semantic interoperability between these 2 major standards in Clinical Trials domain – CDISC and HL7 have partnered to develop ONE common representation of the Clinical Trials domain across the industry
Shared Objectives – CDISC, HL7 & FDA
To improve the quality of public health
To have ONE overarching standard model for data interchange for healthcare information and clinical trial/clinical research data
Ensure interoperability among the various CDISC standards
Develop interoperability between CDISC standards and the HL7 Reference Information Model (RIM)
CDISC – Domain Analysis Model
Being developed as a UML class diagram and documented in Rational Rose XDE
Main subject areas (packages) are as follows:– Entities and Roles– Interventions and Observations– Structured Protocol– Meta-data (ODM)– Statistics
Modeling team represents various stakeholders:– Pharmaceutical companies– Federal Agencies (FDA, NCICB)– Cancer Centers and Academic Institutions (caBIG CTMS developers)– CDISC and HL7
Modeling effort lead by Dr. Charlie Mead
CDSIC Domain Analysis Model – Entities and Roles
FunctionalRole (HL7 Participation)
+ name+ responsibilities+ GPSAddress+ electronicCommunicationAddress
ProtocolAuthor
IRB
RegulatoryAgency
StudyAnalyst
LivingSubject
+ dateofBirth+ sex+ deceasedIndicator(Y/N)+ dateTimeOfDeath
Person
+ geographicAddress+ electronicCommAddr+ race+ ethnicity
Investigator
+ type+ id
DataManager
ProjectManager
Laboratory
SafetyMonitor
Organization
+ geographicAddress+ electronicCommAddr+ standardIndustryClassCode
CRO
BioPharmaceuticalCompany
Entities and Roles View of Problem Space Model
Entity
+ name+ dateCreated+ status
Non-personLivingSubject
+ strain
Place
+ gpsCoordinatesMaterial
+ formCode
Role
+ status+ effectiveStartDate+ effectiveEndDate+ name+ geographicAddress+ electronicCommAddr+ id+ certificate/licenseText
plays/is played by
0..*1..1
plays/is played by
0..*
1..1
PopulationMember
InvestigativeSite
+ id
StudySponsor
HealthCareFacility
+ type
Subject
+ status+ id+ type
RegulatedProduct
Specimen
FundingOrganization
RegulatoryOverseer
StudyDirector
Patient
+ confidentialityCode
Signatory
See Subject State Diagram
Explain role link in Raleigh-Durham
HealthcareProvider
OVERVIEW: In the clinical trial space, the concept "Study" flows through a business process roughly described as "Propose, Plan, Execute, Report." These phases of a given Study (the more general term for Clinical Trial) can be at least approximately mapped to HL7 mood codes (INTENT, REQUEST, EVENT, ???). The overall goal of the Problem Space Model is to present a logcial view of the concept "Study" at each of the four identified business process stages with a realization that such a construct will result in a certain amount of redundancy beetween views, as well as pointing out structures that are unique to a particular phase and/or play a critical 'hand-off' role. Note that in the real world, not all structures defined for a given phase may be in place when the next stage begins.
Document
+ version+ author : SET+ ID : SET PSMID+ documentID+ type : ENUMERATED = formal plus non-formal subtypes+ description : PSMDescription+ title+ status : PSMStatus
CDISC Domain Analysis Model – Structured Protocol Package (draft)
Although it is sometimes the case that a protocol specifies the plan for both primary and correlative studies, best practice is that each instance of a study (correlative or primary) should have its own protocol
StudyMethodology
StudyStatisticalConsideration
StudyOrganization
StudyObjective (the "what")
+ description : PSMDescription+ intentCode : SET ENUMERATED+ primaryObjective : TEXT+ secondaryObjective : SET TEXT+ statementOfPurpose : TEXT
StudyBackground (the "why")
+ Attribute1+ description : PSMDescription+ justificationOfApproach : PSMDescription+ justificationOfObjectives : PSMDescription+ populationDescription : PSMDescription+ rationaleForAnalysisApproach : PSMDescript...+ rationaleForControl : PSMDescription+ rationaleForDesign : PSMDescription+ rationaleForEndpoints : PSMDescription+ rationaleForMasking : PSMDescription+ summaryOfPreviousFindings : PSMDescription
...
StudyPlan (the "how","where", "when", "who")
+ description : PSMDescription
«document»StudyProtocol
+ shortTitle : TEXT
SupplementalMaterial
+ description : PSMDescr...+ ID : SET PSMID+ type+ version
SponsorStudyManagementProjectPlan
SiteStudyManagementProjectPlan
SiteSubjectManagementProjectPlan
1..*
1
StudyObligation
+ commissioningParty+ description : PSMDescription+ responsibleParty+ type : ENUMERATED
ActivityManagementPlan
StudyEvent
+ code+ description+ endDateTime+ name+ result+ startDateTime+ type
*
Resource
Milestone
0..*
1 1
1..* 1..*1
Best practice and/or business process conventions may dictate that one of the set of ID #s be designated as the 'primary' or 'global' ID with all other elements of the set linked to the primary #.
Constraint
1
0..*
BusinessRule
1
0..*
Waiver
0..*
1
StudyEventRelationship
+ type
EligibilityCriterion
A Clinical Trial is a Study with type= "intervention" with subjects of type="human".
Subtype specifies phase for clinical trials, cohort for studies.
Variance
1
0..*
Endpoint
+ code+ description : TEXT+ type : ENUMERATED = primary, secondary+ threshold
*
Measure
1
1..*
assesses / is assessed by
Randomization
+ minimumBlockSize+ maximumBlockSize
DesignCharacteristic
+ detailedMethodCode+ detailedMethodDescription+ summaryCode+ summaryDescription+ synopsis+ type : test value domain = a,d,f,g
0..*
0..1
informs / is informed by
Control
The value set for "type" is "Randomization", "Masking", "Control", "Configuration", "Concurrency", "Scope" and "Bias".
We may need a complex datatype, e.g., "characteristic" with attributes "synopsis", "type", "description" and "code" to be used across multiple classes
Masking
+ level+ objectOfMasking (set)+ procedureToBreak+ unmaskTriggerEvent (set)
«document»Amendment
ResearchProgram
+ type
+ target0..*
+ source
1
*
*
1..1
*
1..*
«Placeholder»StudyParticipant
Document
+ author : SET+ description : PSMDescription+ documentID+ ID : SET PSMID+ status : PSMStatus+ title+ type : ENUMERATED = formal plus non-formal subtypes+ version
1
0..*
modifies / is modified by
Study
+ startDate+ endDate+ studyID : SET PSMID+ name+ type+ subtype+ status+ randomizedIndicator+ otherStudyCharacteristics+ description : PSMDescription
is executed according to the plan contained in / contains the plan for
1
1can have / is associated with
+ primaryStudy1
+ correlativeStudy
0..*
PSMID
+ source+ date+ value
PSMDescription
+ synopsis : TEXT+ summaryDescription : TEXT+ detailedDescription : TEXT+ externalReference : SET+ graphics : ED
Protocol View of Domain Space Analysis Model
Note: Going forward, we need to expand this datatype to include graphics (e.g., using HL7's ED datatype.
Activity
+ name : TEXT+ description : TEXT+ type = {TEMPORAL/NON-TEMPORAL}+ anchorPoint : TIMINGSPECIFICATION+ relativeTime : TIMINGSPECIFICATION
ActivityRelationship
is managed by / manages
Note: The various plans (PlannedAnalysisEvents, DataSafetyMonitoringBoardPlan, DosingPlan, SiteMonitoringPlan, AccrualPlan, DataManagementPlan, InvestigationalPharmacyPlan etc) draw upon the activities in Activity and perform some kind of management on it. A given plan may have both scheduled/determined activities and contingent activities with associated triggers.
Note: We should consider the perspective of an event as the trigger for the execution of one or more activities.
StudyDesign
+ description : PSMDescription
*
1..1
*
ClinicalDevelopmentPlan
describes/is described by
*
1
IntegratedDevelopmentPlan
0..*
*
Note: IS the definition correct? Is research program a subtype of document? There are at least 3 other types of programs that supply context to functional roles including Business Portfolio, Therapeutic Program and Clinical Program with as yet unexplained relationships to study. CommunicationRecord
*
1..*
references/is referenced by
Communication
+ dateTime
Note: This is an act which must be attributed appropriately.
PSMStatus
+ effectiveEndDate+ effectiveStartDate+ statusValue
Note: This class represents a logical not a physical document. Specificially it is not meant to represent a report which is a logical representation of data gathered from across the model.
CDISC Model – Next Steps
Once the model is complete, it will be mapped to the Clinical Trials DMIM and essentially to the HL7 Reference Information Model (RIM)
Some of the subject areas of the CDISC Model are further ahead than others: – Meta-data (ODM) to RIM mapping is already in progress– Structured Protocol Representation is very actively modeling at
present and have begun early mapping to RIM.
Both these areas hope to work towards developing HL7 v3 message specifications in late 2005 - 2006
HL7 Development Framework (HDF) – Model Flow
Technical CommitteeDomain AnalysisModel(s)(integrated models ofbusiness requirementsstated in businesslanguage)
Technical CommitteeProactive1 Domain MessageInformation Models
Public HealthReporting DMIM
Clinical TrialsDMIM
RCRIM DMIM3
RCRIM DMIMn
Technical CommitteeSponsored Standards
RMIM, HD/HMD, MessageTypes, Schemas
Annotated ECG
Clinical Genomics
Clinical Trial Registries
OCSR
ICSR
CT Laboratory
Business Inputs
Regulatory AgencyProjects(e.g. aECG,Product Label,Case Investigation,ICSR,Janus, ...)
Other RCRIMConstituents'Projects (e.g., CTRegistries, caBIG)
CDISC Standards
Mapping/ IntegrationBoth input and outputrevised as necessary
TechnicalSpecificationBoth input and outputrevised as necessaryStandards mustderive from DMIMs
DM
IMs
mus
t
deriv
e fr
om R
IM
RIMincluding bounddatatypes and
vocabulary
Stability
SPL
Future Standards
1 Proactive indicates models inform standards, rather than being revised after the fact to conform to standards
DM
IMs
inco
rpor
ate
thes
efo
r in
tero
pera
bilit
y
CMETsPatterns (e.g., Clinical Statement,Medication Model)Other TC's DMIMs (e.g., StructuredDocuments, Laboratory, Pharmacy,Clinical Genomics)Other supporting HL7 standards(e.g., GELLO)
ODM
special mapping projectfor GAP Analysis
MappingBoth input and outputrevised as necessary
Lab
SDTM
Case InvestigationDomain SpaceAnalysis Model(DSAM)
special mappimg projectdue to scope of DSAM
Process Flow - Requirements to Message specification
Requirements
(CDISC, FDA, NCICB, Pharma,
vendors, etc.)
DomainAnalysis
Model
(UML classdiagram)
HL7 Clinical Trials
DMIM
RMIM &Message Types
will be mapped to the DMIM/RIM
Req. will be Modeled intothe DAM
specificationswill be derivedFrom DMIM
DAM = Domain Analysis ModelDMIM= Domain Message Information ModelRMIM=Refined Message Information Model
NCICB Internal Clinical Trials Modeling Effort
Part 2
Design Clinical Trials related classes as an extension to caBIO model
Focus on Results and Outcomes reporting requirements. Use the areas identified during Outcomes KA analysis
Identify and develop Use Cases to support this effort and scope the model in this Phase
Focus on Intervention Trials ONLY for the first iteration of the model
The data source for these objects would be C3D, therefore limit the scope of the model based on what is available in Oracle Clinical
Since C3D is based on CDEs, this Clinical Trials UML model will be CDE based
The Clinical Trials UML classes/objects developed here MAY serve as Clinical Trials API’s to C3D (Oracle Clinical)
Clinical Trials Modeling - Objectives
Clinical Trials Modeling – Objectives (cont’d.)
The CT APIs will eventually be used to retrieve data and build Decision Support Applications
Ensure integration with other NCICB Clinical Trials related UML models (classes) – thru interfaces – caBIO, SPORES, etc.
Long Term Objectives
The model should be reflective of the CDISC Domain Analysis Model in future phases
The model should also be reflective of the HL7 CT DMIM – ensure placeholders and structure to be able to receive data from HL7 v3 messages (some of this will happen naturally since the CDISC DSAM is being adopted by HL7 as the Domain model for HL7 Clinical Trials committee and will be mapped to the HL7 CT DMIM)
Data Areas Identified by Outcomes Requirements
Response
Biomarkers Modulation
Survival
Disease Recurrence
Adverse Events
Quality of Life
Baseline Data
Clinical Trials Modeling – Sample Use Cases
RESPONSE– Do men with early stage prostate cancer suffer complications after
treatment with Proscar has ended? (Show patient data where Disease Stage Numeric = I and where the Adverse Event Onset Date is after the Treatment End Date)
DISEASE RECURRENCE– What is the average time to disease progression (in days) for all
patients on breast cancer trials? What is the most common reason for progression?
SURVIVAL– Is a high BMI associated with lower survival among breast cancer
patients? Is there a correlation between BMI and disease stage? Metadata Loader – Implementation of HMD meta loader
Example Use Case Documentation
Name/Query:
Is a high BMI associated with lower survival among breast cancer patients? Is there a correlation between BMI and Disease Stage?
Source: Center for Cancer Research KA Report
Detail: The user is able to view survival status of breast cancer patients and determine if a high BMI is associated with a lower survival status (1 or 5).
Background:
BMI is calculated from a patient’s height and weight. Survival Status is based on caDSR DEs.
Purpose: This will provide the survival rate and disease stage values for breast cancer patients with various BMI measurements.
Potential CDEs used for this Query:
The user selects to generate a report of Survival status for breast cancer patients. The system responds by requesting the user to select a set of breast cancer trials. Patient IdentifierProtocol NumberPatient Survival Status Patient Body Mass Index Measurement (can also be calculated by the system using height and weight CDEs)Diagnosis DateDisease Stage NumericPrimary Site of Disease Name
Valid Values/ Meanings Associated with Potential CDEs:
Patient Survival Status :
3 ALIVE DISEASE STATUS UNKNOWN1ALIVE WITH DISEASE2ALIVE WITH NO EVIDENCE OF DISEASE5DIED4UNKNOWN
Use Case #3 – Survival – Across Trials
Key Features of the Clinical Trials Model
Completely based on CDEs– CDEs from CCR Context with CS/CSI of C3D– all the class attributes are existing CDEs– only CDEs with a workflow status of “Released” have been used– actual CDE Long Name captured in the EA as Alias– experimenting with adding a Tag in EA for capturing the PublicID– currently using 210 CDEs from a total set of 576 from CCR context
Although currently focused on C3D, the model is a generic Clinical Trials model to support other data sources in future iterations
cd Data Model
Study
Agent
StudyAgent
Subject
Event Assessment
DrugAdministration
Protocol
DiseaseStudyDisease
Histopathology
StudySite QualitativeAdverseEvent
SpecimenCollection
DiseaseResponse
Organ
StudyObjective StudyBackground
Endpoint
LesionEvaluation
Site
AgentSynonym
Procedure
Imaging
Radiation
Surgery
Sample
Observation
QuantitativeResults
Genomic MicroSpecimen Base
EligibilityCriteria
Hook to existing caBIO class.
CancerDiagnosis
SubjectAssignment
EventRelationshipSubjectDisease
1..*
0..*
1..*0..*
0..*
1..*
1..*
0..*
1
participate
1..*
0..*
1..*
1..*
0..*
0..1
0..*
1
0..*
1..* 1
1..*0..*
0..*
1..*
0..* 1..*
1..* 0..*
0..* 0..*
0..*
1..*
Clinical Trials UML Class Diagram (work-in-progress)
Clinical Trials UML Diagram
cd Data Model
Study
Agent
- name: string- commercialVsINDindicator--NEW:
StudyAgent
Subject
- id: - initialsName: - medicalRecordNumber: int- age: - addressPostalCode: - subjectID: int- genderCategory: string- raceCategory: string- ethnicityCategory: string- eligibilityMetInd: - effectiveDate: date- eligibilityWaiverReason: string- firstLiveBirthAgeInYears: int- pregnancyLiveBirthCount: int- menopauseAgeInYears: int- ICFSignedDate: date
Event
- id: - epochName--NEW: string- visitName??: string- visitDescription??: string- courseDayNumber: number- startDate: date- endDate: date- durationUOM: string- duration: int- courseNumber: int- courseStartDate: date- courseStopDate: date- courseDisposition: char- totalDoseValue??: string
Assessment
- onsetDate: date- ??neurologicCriteria??: string- conditionResult: string- changeFromPreviousInd: string- imageEvaluationResult: string- ??abnormalConditionDetectionTextType??: string
DrugAdministration
- actualDose: int- actualDoseUnitOfMeasure: string- routeName: string- medicationSchedule-movetoRegimen?: string- frequency: string- totalDailyDoseValue: int
Protocol
- id: - number: string- longTitleText: string- shortTitleText: string- navyNCINumber: - activationDate: date- amendementDate: date- amendmentIdentifier: int- closureDate: date- suspensionDate: date- phaseCode: string- statusCode: string- typeCode: string- monitorCode: string- blindingIndicator: boolean- diseaseCode: strinbg- sponsorCode: string- multiInstitutionInd: boolean- targetAccrualNumber: int- precisText: string
Disease
- CDUScode:
StudyDisease
Histopathology
- breastCancerInvasiveHistology: string- breastCancerInSituHistology: string- description: string- grade: string- confirmationDate: date- lungScore: string- findingsGrossExamResult: string- reportDescriptiveText: string- reportIdentifier: int- responseSpecimenCriteria: string- ?responseAssessmentInd-Delete?MoveToAssessment?: string- benignTissueChange: string
StudySite
- id: - role--NEW: string- piName: string
Qualitative
- survivalStatus: string- deathDate: date- deathCause: string- deathCauseText: string- performanceStatus: string- painIndex: int
AdverseEvent
- onsetDate: date- resolvedDate: date- ctcCategory: string- ctcTermType: string- ctcAttributionCode: string- ctcGrade: int- ctcVersionNumber: string- seriousReason: string- outcome: string- action: string- therapy: string- therapyDate: date- therapyDelayDuration: string- therapyIntensity: string- conditionPattern: int- courseInclusionInd: boolean- descriptionText: string- reporterFirstName??: string- reporterLastName??: string- similarAEDescription??: string- similarAENumber??: int- directedMeasureName: string
SpecimenCollection
- collectionDate: date- type: string- siteType-DELETE: string- collectionMethod: string- criteriaType: string- condition???: string
DiseaseResponse
- assessmentType: string- bestAssessmentType: string- bestResponseDate: date- progressionDate: date- diseaseProgressionInMonths: int- noteText: string
Organ
- id: int- name: string
StudyObjectiveStudyBackground
Endpoint
LesionEvaluation
- number: string- targetNontargetIdentificationText: string- evaluationNumber: int- measurementX: int- measurementY: int- measurementZ: int- measurementProduct: int- measurementMethod: string- bodySiteName: string- appearanceTextType: string- nodularQuantityNumber: int- flatQuantityNumber: int
Site
- nationalCancerInstituteInstituteCode:
AgentSynonym
Procedure
- name: string- bodySiteName: string
Imaging
- identifier: - contrastAgentEnhanceInd: boolean- enhancementDescriptiveText: string- rateOfEnhancementValueNumber: int
Radiation
- type: string- schedule: string- doseUOM: string- dose--NEW: string
Surgery
Sample
- IDnumber: int
Observation
- reportingDate--NEW: date
QuantitativeResults
- testName: string- UnitOfMeasure: string- highNormalValue: int- lowNormalValue: int- assayMethod--DraftNew: string- technique: string- assessmentMethod--NEW: string- patientPositionType: string- panelName: string- characterValue: string- numericValue: int- biomarkerInd--NEW: boolean- meansVitalStatusObtained: string
Genomic
MicroSpecimen
Base
EligibilityCriteria
- text: string- expectedAnswer: string
QOLIndicators
- smokingChangeFromBaselineInd: boolean- smokingQuitDurationMonthsNumber: int- alcoholConsumptionPastYearInd: boolean- alcoholUseChangeStatusInd: string- alcoholUseLifetimeInd: string- alcoholUseCurrentStatusInd: string- alcoholCurrentUseRange: string- alcoholType: string- alcoholStopUseAgeInYears: int- alcoholUseStopDate: date- alcoholUseStartYear: date- alcoholUseStartAgeInYears: int- alcoholUseDescriptiveText: string- moderateExerciseHoursPerWeek: int- moderateExerciseMonthsPerYearNumber: int- sedentaryExerciseDaysPerWeekNumber: int- sedentaryExerciseHoursPerDayNumber: int- strenuousExerciseHoursPerWeekNumber: int- strenuousExerciseMonthsPerYearNumber: int- aNAMResultAccruarcyInPercentage: int
Hook to existing caBIO class.
??attributeName?? = DEs that need reviewed before including as attributes.attributeName?? = remove due to personnel identification reasons?attributeName--NEW = attributes that do not have a supporting DE from C3D.
Bullpen
CancerDiagnosis
- diseaseStageTNM: string
Denotes model skeleton
SubjectAssignment
- id: int- arm: int- armTargetedAccrualNumber: int- CDUSSubgroup: string- registeringGroup: - registeringInstitutionCode: int
EventRelationship
SubjectDisease
- diagnosisCode: int- diagnosisName:
1..*
0..*
1..* 0..*
0..*
1..*
1..*
0..*
1
participate
1..*
0..*1..*
1..*
0..*
0..1
?Delete due to collectionMethod 2/2/05
0..*
1 0..*
1..* 1
1..*0..*
0..*
1..*
0..* 1..*
1..* 0..*
0..*0..*
0..*
1..*
Example – CDE Long Name as Alias
cd Data Model
Protocol
- id: - number: string- longTitleText: string- shortTitleText: string- navyNCINumber: - activationDate: date- amendementDate: date- amendmentIdentifier: int- closureDate: date- suspensionDate: date- phaseCode: string- statusCode: string- typeCode: string- monitorCode: string- blindingIndicator: boolean- diseaseCode: strinbg- sponsorCode: string- multiInstitutionInd: boolean- targetAccrualNumber: int- precisText: string
CDELong Name
Status and Next Steps
The CT model is scheduled to be complete in first week on April
Would like to begin meeting with other NCICB development teams by mid-March for model reviews
Will incorporate as much as possible of the CDISC Domain Analysis Model into this iteration
This is just the first Phase of the effort – additional requirements will be accommodated in future iterations
Implementation of this model will be done with other NCICB developers/members
Acknowledgements
Sue DubmanChristo AndonyadisDianne ReevesPeter Covitz
Bill McCurrySmita HastakBecky RiggleVal BraggCharles YaghmourJane Harrell