Implications for SDTM

CDER Common Data Standards
Issues Document:
Implications for SDTM
John Brega, Jane Diefenbach,
Micaela Salgado-Gomez
PharmaStat LLC
Bay Area CDISC Implementation Network
28 September 2011

Disclaimers
• The views expressed are those of the presenter
and not necessarily of the FDA or CDISC
• With the FDA, as with the stock market, past
performance is no guarantee of future
direction…
• Nothing in this presentation should be taken as
regulatory advice!
2

The Documents
• CDER published the CDER Common Data
Standards Issues Document in May, 2011
– Find the document by Googling its title. It is the first link.
– I’ll call it the CCDSID because we love acronyms and need
more of them.
• At the same time, CDISC published Amendment 1
to SDTM v1.2 and SDTM IG v3.1.2
– These actions were coordinated. The CCDSID references
Amendment 1, which supports CDER’s preferences
– This document is temporarily not available on the CDISC
website because the comment period has closed.
3

The Highlights
• CCDSID Contains advice on
– SDTM, ADaM and SEND
– Controlled terminologies, MedDRA and drug
dictionaries
– Use of variables and domains
– Traceability and documentation
• Amendment 1 contains
– New variables for DM and the Events class domains
– Instructions for their use
4

CDER’s Rules of Thumb
• Submit SDTM
• Document it with define.xml and define.pdf
• If you submit SDTM, also submit ADSL
• Derive ADaM datasets from the submitted SDTM
• Ensure traceability from CRF to SDTM to ADaM
• Use CDISC controlled terminology
• Use MedDRA as it comes from the dictionary, use a
single version for the ISS, don’t include numeric
codes
• Don’t use Suppquals as a wastebasket
5

The SDTM Specifics: AE
• MedDRA terms other than SOC and preferred term
were suppquals. Now they are variables.
– Amendment 1 defines 10 new variables for MedDRA
terms and codes
• A Treatment Emergent Flag (--TRTEM) variable has
been added. FDA wants us to use it.
• These new variables can be used in any Events
class domain with coded terms, such as MH
(Medical History)
• All are Expected, which contradicts the CCDSID
6

New Events Variables
Variable Name Variable Label Type
Controlled Terms,
Codelist or Format
--TRTEM Treatment Emergent Flag Char (NY)
--LLT Lowest Level Term Char MedDRA
--LLTCD Lowest Level Term Code Num MedDRA
--PTCD Preferred Term Code Num MedDRA
--HLT High Level Term Char MedDRA
--HLTCD High Level Term Code Num MedDRA
--HLGT High Level Group Term Char MedDRA
--HLGTCD High Level Group Term Code Num MedDRA
--SOC Primary System Organ Class Char MedDRA
--SOCCD Primary System Organ Class Code Num MedDRA
--BDSYCD Body System Code Num MedDRA
7

The SDTM Specifics: DM
• 8 new variables are added:
– Actual Arm and Arm Code
– Dates of First and Last Study Treatment Exposure
– Dates of Informed Consent and End of Participation
– Date of Death and Subject Death Flag
• Some may be redundant with RFSTDTC or
RFENDTC
• All are Required or Expected
8

New DM Variables
Variable Name Variable Label Type
Controlled Terms,
Codelist or Format
ACTARMCD Actual Arm Code Char *
ACTARM D Actual Arm Char *
RFXSTDTC Date/Time of First Study Treatment Exposure Char ISO 8601
RFXENDTC Date/Time of Last Study Treatment Exposure Char ISO 8601
RFICDTC Date/Time of Informed Consent Char ISO 8601
RFPENDTC Date/Time of End of Participation Char ISO 8601
DTHDTC Date of Death Char ISO 8601
DTHFL Subject Death Flag Char (NY)
*Note that labels are too long for two of the variables…
9

The SDTM Specifics: DS
• Use EPOCH in DS to distinguish multiple
disposition events
• If DEATH occurs it’s the last record and has an
EPOCH
10

More SDTM Specifics
• Include EPOCH, ELEMENT and ETCD in every
“subject-level” dataset (e.g., AE, LB, CM, EX, VS…)
• Split large datasets by --CAT, include both the
split and the complete datasets
– How big is “large”? Up to 400mb is OK, but ask your
review division
11

The Larger Implications
• Some new variables are very analysis-like, some
old variables will be used in an analysis-like
way, which may involve complex derivations that
depend on other derivations.
• For example,
– Flagging Treatment Emergent AEs requires
comparison to exposure data, which may in turn be
derived from drug accountability data
– Determining EPOCH may be difficult if defined by
events rather than visits, and involves date windowing
for log data like AE and CM
12

Questions to Ponder
• Prior to these documents, a case could be made
for using “pure” SDTM as an internal data
standard. Is that still a valid strategy? If not,
what should the new strategy be?
• How will our processes need to change to
support production of the new SDTM?
14

Thank you!
Questions?
John Brega: JBrega@PharmaStat.com
15