Study Data Tabulation Model (SDTM) - Digital Infuzion, Inc.cdiscportal.digitalinfuzion.com/CDISC User Networks... ·  · 2008-09-23SDTM Basics - Structures Based Upon General Observation

CDISC Italian User Group 16 November 2007

Study Data Tabulation Model (SDTM)

Annamaria MuraroHelsinn Healthcare

2

Regulatory Background

• 1999, FDA guidance “Providing Regulatory Submissions in Electronic Format – NDAs”– SAS XPT datasets– Pdf documents: Define.pdf (data contents), blankcrf.pdf (annotated

CRF)– Case Report Tabulation (Patient Profiles) as pdf files

• October 2005: FDA issued the guidance “Providing Regulatory submissions in Electronic format using the eCTD specifications”

• September 2006: FDA published in the Federal Register a notice of withdrawal of electronic submission guidance for eNDA. – This notice designates eCTD as preferred format for electronic

submissions and notes that beginning Jan 2008 any electronicsubmission must be eCTD

3

What is SDTM: Study Data Tabulation Model

• This model describes the contents and structure of data collected during a clinical trial

• The purpose is to provideregulatory authorityreviewers (FDA) a cleardescription of the structure, attributes and contents of each datasetand variables submitted as part of a product application

Animals (SEND) Humans

(SDTMIG)

SDTM

4

Implementation Guidelineshttp://www.cdisc.org/standards/index.htmlhttp://www.fda.gov/oc/datacouncil/cdisc.html

• Study Data Tabulation Model, Version 1.1, April 2005– which represents the underlying conceptual model behind the

submission data standards (SDS) • SDTM Implementation Guide, Version 3.1.1, August 2005

– Detailed Domains description, Variables to be included and theirattributes, assumptions and convention, Examples

– Version 3.1.2 posted for Comment July 2007• New standard domains• More details

• Controlled Terminology: SDTM Package-1 & Lab Test Controlled Terminology, Package 2A, Package 2B (under development)

• SEND, Standard for Exchange of Nonclinical Data Implementation Guide for Animal Toxicology Studies, Version 2.3

5

Fundamentals of SDTM - Domains

Domain: collection of observations with common topic

• Case Report Form à SDTM domainà Dataset– A domain may collect data from more than one CRF form– Generally each domain is represented by a dataset

• Each domain has a unique two-character domain name (e.g., AE, CM, VS)

• Variables in domain begin with the domain prefix: (e.g., VSTESTCD)• Domain structure: vertical• Two categories of domains:

– CDISC Standard Domains (spelled out in detail in the Implementation Guide).– Custom Domains

• Based on one of the General Observation Classes (findings, events, interventions)• Basic variables are outlined in the SDTM• Rules for creating these are in Section 2 of the IG

6

SDTM Basics - Structures Based Upon General Observation Classes

• Interventions: • Investigational treatments, therapeutic treatments, and procedures

administered to or taken by the subject• One record per constant dosing/treatment interval• Examples: study medications(EX), concomitant medications(CM)

• Events:• Occurrences or incidents independent of planned study evaluations

occurring during the trial or prior to the trial • One record per event• Examples: medical history(MH), adverse events(AE)

• Findings: • Observations resulting from planned evaluations • One record per finding result or measurement• Examples: lab data(LB), vital signs(VS)

Each of these has defined structures and variables. No new variables may be added to these domain classes

7

SDTM Basics - Special-Purpose Datasets

• Not Classified as Interventions, Events, or Findings• They Have Special Rules

– Demographics (DM)– Comments (CO): free text comments– Trial domains: to describe the design of a trial– RELREC dataset: represent the relationship between

datasets and records– SUPPQUAL: used for data items not included in the

SDTM standard

8

SDTM Standard Domains

Interventions Events

ConMeds(CM)

Exposure (EX)

Adverse Event (AE)

MedHist(MH)

Disposition(DS)

Findings

ECG(EG)

PhysExam(PE)

Labs(LB)

Vitals(VS)

Demog(DM)

Other

SubjChar(SC)

Subst Use(SU)

Incl/Excl(IE)

Relationships

Supp Qualifiers

Trial DesignQuestionnaire (QS)

Comments (CO)

Deviations(DV)

Concentrations (PC)

Drug Accountability (DA)

Microbiology (MB)

PK Param (PP)Draft 3.1.2

9

Fundamentals of SDTM - Variables

• CDISC categorizes variables as being– Required: variables need to be in the domains

• Their values cannot be null– Expected: variables need to be in the domain

• Some values may be null– Permissible: variables included in the domain as needed

• CDISC categorizes variables into five roles– Identifier: identify the study, subject of the observation, the sequence

number– Topic: specify the focus of the observation (such as the name of the lab

test)– Timing: describe the timing of the observation (Visit, Start/End date,

Days, Time Points, Duration)– Qualifier: additional text or numeric values– Rule: express an algorithm or method to define start, end or looping

conditions in the Trial Design model

10

Fundamentals of SDTM - Variables• Variables Attributes

– Variable Name: limited to 8-chars– Variable Label: <=40 chars– Variable Type: mainly char– Variable Lenght: <=200 chars

• Date/time format– ISO 8601 is a text string YYYY-MM-DDT hh:mm:ss (not a SAS format)– Has ability to handle incomplete date– Example: December 15, 2003 13:14:17 à 2003-12-15T13:14:17

December 15, 2003 à 2003-12-15

• Controlled Terms or Format– Controlled terminology or text should be used instead of, or in addition to

arbitrary number codes (à No SAS format!!)– CT (**) published externally (ex.MedDRA or follow CDISC-specific terminology)– CT (*) value from a sponsor-defined codelist

• Raw data (as originally collected) and derived values(e.g. conversion to standard units, age)

11

Example: DMRequiredExpected

Permissible

CRFDerived

Sponsor DefinedExternal Lab Details

YYYY-MM-DDThh:mm:ss

12

Example: DM

Reference standards

13

Example: DS

14

Example: DS

DISPOSITION EVENTPROTOCOL MILESTONE

OTHER EVENT

15

Example (LB)

ORIGINAL RESULT

16

Example (LB)

STANDARD FORMAT,

char

STANDARD FORMAT,

num

17

Trial Design Model

• Defines a standard structure for representing the planned sequence of events and the treatment plan forthe trial

• Trial Arms (Planned): described each planned arm in the trial (ordered sequence of elements)

• Trial Element (Planned): describes the Element, rules forstarting/ending the element

• Trial Visits (Planned): describes the planned order and numberof visits in the study

IG Section

7.1, page 87

18

Trial Design Model

• Subject Element (Actual), Subject Visit (Actual)– Subject Element and Subject Visit are special-

purpose domain in the draft version SDTMIG 3.1.2

• Trial Inclusion/Exclusion Criteria: lookup table: one record for each inclusion/exclusion criterion. Not subject oriented

• Trial Summary: descriptive attributes of trial like trial phase, title, objective etc.)

19

SDTM adoption by FDA• July 2004: FDA has endorsed the CDISC standards

• October 2005, FDA Guidance for ICH eCommon TechnicalDocument (eCTD) is updated. CDISC SDTM format isrecommended for clinical data and define.xml for metadata

20

Dec 2006: Proposed Rule

The Food and Drug Administration is proposing to amend the regulations governing the format in which clinical study data and bioequivalence data are required to be submitted for new drug applications (NDAs), biological license applications (BLAs), and abbreviated new drug applications (ANDAs). The proposal would revise our regulations to require that data submitted for NDAs, BLAs, and ANDAs, and their supplements and amendments be provided in an electronic format that FDA can process, review, and archive. The proposal would also require the use of standardized data structure, terminology, and code sets contained in current FDA guidance (the Study Data Tabulation Model (SDTM) developed by the Clinical Data Interchange Standards Consortium) to allow for more efficient and comprehensive data review.

Federal Register / Volume 71, No. 237 /Monday, December 11, 2006Federal Register / Volume 71, No. 237 /Monday, December 11, 2006

The CDISC SDTM is still the preferred data specification. It will become requireddata specification when the proposed rule is approved.

21

• Data standards allow FDA to develop standard tools forreview (http://www.fda.gov/oc/datacouncil/meetings/oliva.pdf)

• WebSDM was developed under a Cooperative Research and Development Agreement (CRADA) between the FDA and Lincoln Technologies with the goal of providing a user-friendlyenvironment for browsing and reviewing CDISC SDTM-compliantclinical trial. This product has beenin use at FDA since 2004

• For each domain, users can access variable summarystatistics, data visualization toolsand drill down into individualsubject data and patient profiles

FDA standard tools

22

SDTM Adoption by FDA

• SDTM / ADaM Pilot projects– Pilot regulatory submission using SDTM/ADaM (data

from Eli Lilly company, lesson learned reported at CDISC Interchange 2006)

– Integrated Safety Data Pilot is now underway

• CDISC SDTM Training at FDA: “In preparation for the increasing volume in SDTM submissions to the FDA, CDISC and FDA have joined forces to ensure that FDA personnel, "speak SDTM". On 9 July 2007, there was an open SDTM training session for FDA at the White Oak facility in Maryland. Approximately 50 people from FDA attended with representation from statistical, data management, and programming functions” (CDISC newsletter, Sept 2007)

23

Implementation: Helsinn experience

• Integrated approach to standardization• Team work and training• Standards applied to any study, any phase• Mapping to CDISC standards

– CDISC general assumptions 100% implemented– Include any Required and Expected variables– Select Permissible variables based on data we collect– Some deviations from CDISC acceptable– Need to develop code lists (lab parameter, pk parameters, etc.), CDISC

controlled terminology not available yet– Clear understanding of the SDTMIG– Are custom domains needed? Where to collect cancer history? How to map data

collected by patient diary?– Coding for CM, intermediate levels were needed– Trial domains mapping for cycle based studies, cross-over studies

• Interacting with CROs and providers– Provide CROs with a clear understanding of CDISC mapping requirements– Central Lab, Bioanalytical provider: collect data CDISC compliant from the

source• Integration of old clinical data

24

Should SDTM be used for submissionsto the FDA only? • Benefit of SDTM as internal data standard

– No need to convert data for esubmission– Reduce data transformation efforts– Efficient (and coste effective) data exchange with providers (CROs,

Central Lab, etc)– Easier interaction with CROs, in-licensing made easier, mergers less

painful– To maximize clarity in the process using a data standard which

everyone will be familiar: integrated process from CRF to submission– To use standards with large consensus with pharma industry instead of

maintain its own standard

But don’t expect it to be easy!Expect resistance e.g. “what’s wrong with my existing standard”, “we don’t have time/resource” …Have answers ready for questions such as How long will it take? How much resource will be required? What is the cost of doing this? What are the cost savings if we implement this?

25

Questions

26

Representing Relationship – Back-up

• Relating groups of records in a domain for a subject– Use the variable --GRPID

• Relating non-standard variables to a parent domain– SDTM does not allow the addition of new variables– Use the SUPPQUAL dataset to capture non-standard variables

and their association to parent records – Suppl variables may be collected by separete SUPP dataset

(convention SUPP– where – is the two letter domain code (example SUPPLB, SUPPDM)

• Relating independent records in separate domains– Use the RELREC dataset

• Relating datasets– Use the RELREC dataset

IG, par. 8.1,

page 121

27

Creating a New Domain: BACK-UP slide

• Ensure that there is a definite need to create a new domain

• Assign a 2-letter domain code

• Choose the general observation class (Interventions, Events or Findings) that best fits the data

• Add variables as follow: Identifier Variables (studyid, domain, usubjid, --seq), Timing Variables, Topic & Qualifiers

• No new sponsor-defined variables can be added toSDTM

• Follow SDTM conventions for variable order, name/label/type

IG, par. 2.6,

page 13