CDISC - SDTMs and ADaMs

Quick Links, CDISC References, SAS e-Guide	Mind Map
CDISC Mapping Videos CDISC Video Portal CDISC Training Sessions	Getting Started with SDTMs Medical Devices
Outline	SDTMIG Versions: 3.1, 3.2, 3.1.3, 3.1.1

Mind map SDTM Certification Pilot Sample Data: Harvard Data

SDTM OnLine Specification Control Terminology (Online) New CDISC Mapping e-Guide

Newly Released FDA Guidelines - Electronic Format Submission, Technical Conformance, Search CDISC SAS Papers

FDA Presentation on Data Standards, Drug Review Process SAS, CDSIC Q&A Blog, PhUSE SENDS, CDISC SENDS, CDISC Open Rules Engine (CORE, Blog)

PharmaSUG Single Day Event Presentations PMDA (Pharmaceuticals and Medical Devices Agency)

CDISC Primer (Introduction Videos on SEND, CDASH, SDTMIG, ADaMs)

Clinical_Data_Interchange_Standards_Consortium_CDISC_Warehouse
Animated_Guides
PROC_CDISC_Clinical_Toolkit
CONTROL_TERMINOLOGY
ISO8601_Dates
SDTM
SMQs
SUPPQUAL
RELREC
Trial_Design
ADaM_and_Beyond
SDTM_and_ADaM_Integration
SEND

Clinical Data Interchange Standards Consortium (CDISC) Warehouse

	SDTM	ADaM
Source	Raw datasets with all records (1:M CRFs) PLUS required variables	1:M SDTMs PLUS derived variables/records
Dates	ISO8601 character (YYYY-MM-DD)	Datetime variable (YYMMDD10.), Durations
Structure	Most all except DM are vertical with generic variable names (ideal for by-group nested processing) BY Options: TESTCD (ex. IETESTCD), DTC (ex. AESTDTC), LBTEST, COD (ex. AEDECOD) VSTESTCD = new variable name VSTEST = new variable label VSSTRESN = numeric value with format VSSTRESC = character value with format and length, maybe character version of VSSTRESN	Generally summary record of one record per patient per param (ex. lbtestcd) per visit Horizontal (ideal for modeling, by subjid and time, content-based variable name, transposed from SDTM vertical structure) Lab visit windows are applied to summarize to one record per patient per visit General options for AVAL and AVALC 1. Same as VSSTRESN and VSSTRESC 2. Same as VSSTRESN and VSORRES if VSORRES ^= VSSTRESC 3. Imputed value if VSSTRESC contains '<' 4. Average value of AVAL with DTYPE='AVERAGE'
Datasets	VS LB QC CM, AE	Four Methods to Construct – May need transpose (TEMP, SBP, DBP, etc.) to vertical structure (ADVS) – May need to transpose by visit, id test to vertical structure by visit and test variable names (ADLB) – May need to transpose by visit, id question to vertical structure by visit and question variable names (ADQC) – Not transposed, kept as (ADCM, ADAE)

SDTM

ADaM

Source

Raw datasets with all records (1:M CRFs) PLUS required variables

1:M SDTMs PLUS derived variables/records

Dates

ISO8601 character (YYYY-MM-DD)

Datetime variable (YYMMDD10.), Durations

Structure

Most all except DM are vertical with generic variable names (ideal for by-group nested processing)

BY Options: TESTCD (ex. IETESTCD), DTC (ex. AESTDTC), LBTEST, COD (ex. AEDECOD)

VSTESTCD = new variable name

VSTEST = new variable label

VSSTRESN = numeric value with format

VSSTRESC = character value with format and length, maybe character version of VSSTRESN

Generally summary record of one record per patient per param (ex. lbtestcd) per visit

Horizontal (ideal for modeling, by subjid and time, content-based variable name, transposed from SDTM vertical structure)

Lab visit windows are applied to summarize to one record per patient per visit

General options for AVAL and AVALC

1. Same as VSSTRESN and VSSTRESC

2. Same as VSSTRESN and VSORRES if VSORRES ^= VSSTRESC

3. Imputed value if VSSTRESC contains '<'

4. Average value of AVAL with DTYPE='AVERAGE'

Datasets

CM, AE

Four Methods to Construct

– May need transpose (TEMP, SBP, DBP, etc.) to vertical structure (ADVS)

– May need to transpose by visit, id test to vertical structure by visit and test variable names (ADLB)

– May need to transpose by visit, id question to vertical structure by visit and question variable names (ADQC)

– Not transposed, kept as (ADCM, ADAE)

Dataset/Item	Version	Description (Data Standards, Specifications - Can use template to create SDTM metadata files, ADaM Specification, CRF Example, Edit Checks)
SDTM 'Raw datasets represented as Case Report Forms' (Useful for listings) 3.1.2 Guide, Compliance Check Macros, SAS DM, AE, LB Questionnaire SDTM SAS Programs SDTM SAS Macros	3.1.3	Study Data Tabulation Model Mapping process from Raw data (See %make_empty_dataset macro) 1. Raw dataset variables (mapped with Domain specification) PLUS all required/expected Domain specific variables (create if needed, such as FDOSEDTN or LBBLFL), values, and records. 2. Using control terminology, transform data to standard names and values. 3. Keep any unmapped raw dataset variables in SUPPQUAL corresponding datasets such as SUPPDM. 4. For unscheduled visits, assign visits if possible. 5. May be many to 1 relationship with CRF page. 6. Variable order is important, 8 char variable names, 40 char variable labels. 7. Create --DTC suffix for ISO8601 character date value. (See %make_dtc_date macro) 8. Create --DY suffix for study day variable = (date portion of --DTC) - (date portion of RFSTDTC) + 1 if --DTC is on or after RFSTDTC, OR study day variable = (date portion of --DTC) - (date portion of RFSTDTC) if --DTC precedes RFSTDTC, where RFSTDTC is a discrete starting point and RFRNDTC is a discrete ending point contained in the DM Domain. Note that AESTDY may be missing for partial AESTDTC. (See %make_sdtm_dy macro)
ADaM 'Analysis datasets for modeling and reporting - One-Proc Away' (Useful for summary tables, statistical modeling and p-values) 2.0, 2.1 Reference, 2.1 presentation, 1.0 Guide, Common Stats, Tool, SAS, ADSL Checklist Guide, XLS ADaM SAS Programs ADaM SAS Macros	2.1	Analysis Data Model Three Models: Subject-level (ADSL), Change from baseline (ADLB) and Categorical analysis Traceability is important to preserve to trace back to SDTMs. For example, keep SRCDOM and SRCVAR variables only if more than one value exists. Also, use caution when combining similar variables such as SDTM.EX.EXTRT and SDTM.CM.CMTRT into ADAM.ADCM.CMTRT. (See %make_empty_dataset macro) Mapping process from SDTM 1. Summary of SDTM datasets PLUS derived variables such as AGE, CHANGE IN BASELINE, SEX/SEXN, RACE/RACEN, etc. for TLGs. 2. Generally one record per patient or one record per patient per visit. May apply visit windows. 3. Identify BLFL, baseline flag and calculate CHANGE FROM BASELINE. 4. Using control terminology, transform data to standard names and values such as AVISIT, analysis visit. For traceability, it is good practice to keep the associated VISIT variable. 5. May create new records or visits such as DTYPE='AVERAGE' or 'LOCF'. (See 1. Proc Means, 2. Proc Sort, 3. Data Step Merge for programming techniques), impute with mean values or calculate total scores. 6. May be one to many relationship with SDTM. 7. Convert --DTC to --DT datetime variables. Use --DTF for date impute flag (D for day, M for month and day, and Y for year, month and day). (See %dtc2dt macro) 8. Create indicator or population flag variables --FN (1, 0) or --FL (Y, N). Other options include AE level flag variables such as AE1FL (patient level), AE2FL (organ class level) and AE3FL (preferred term level). 9. Metadata documentation (pseudo-code specifications) with hyperlinks - similar to define.xml.
LAB		Laboratory Data Model
ODM 1.3.2 Guide		Operational Data Model (Contains both variable attributes and data values)
DEFINE.XML Dictionary Tables, vardef.sas (See diagram below)		Required for each SDTM and ADam dataset. SAS Paper, SAS Paper 2, SAS Paper 3 Clinical Standards Toolkit Doc XML Online Tutorial, XML Online reference
ISO 8601 Date Format SAS Paper, SAS Paper 2, SAS Paper 3 XXDTC for date of XX XXSTDTC for start date XXENDTC for end date See %make_dtc_date in SDTM SAS macros See %dtc2dt in ADaM SAS macro		ISODATE9 character, ex. YYYY-MM-DD or YYYY-MM-DDThh:mm:ss, 2009-01-01T12:00:00 Three step process 1. Collected as text string, not as a date variable. May use informats such as b8601da8. for basic or e8601da10. for extended (XXXX-XX-XX) to store as datetime variable. 2. Use the same informats to display as text strings. 3. Can use CALL routine to calculate duration - CALL IS8601_CONVERT(convert-from, convert-to, <from-variables> , <to-variables>, <date-time-replacements>). a. Create time variable from duration b. Create interval from duration and datetime c. Create duration from two datetimes Note: Alternative to storing dates is to collect duration - PnYnMnDTnHnMnS or PnW (e.g., P2Y : 2 year ; PT42M18S : 42 Minutes 10 Seconds) Note also that start/end reference dates often refer to first/last date of drug exposure.
Control Terminology 'Metadata Documentation' Reference Files, List, Submissions Guide (See %make_codelist_formats macro)		1. Sponsor defined - List of constant uppercase values ex. treatment, race, sex, etc. 2. Dataset used with PROC FORMAT to create user defined formats. 3. From raw to SDTM or SDTM to ADaM, convert variables to standard labels or transformed values. For example convert visits to analysis visits. 4. MEDRA (version 16.1) is another example. 5. Can be used to create macro variables for do loop processing. 6. None or minimized user-defined formats.

Dataset/Item

Version

Description

(Data Standards, Specifications - Can use template to create SDTM metadata files, ADaM Specification, CRF Example, Edit Checks)

SDTM

'Raw datasets represented as Case Report Forms'

(Useful for listings)

3.1.2 Guide,

Compliance Check Macros, SAS

DM, AE, LB

Questionnaire

SDTM SAS Programs

SDTM SAS Macros

3.1.3

Study Data Tabulation Model

Mapping process from Raw data

(See %make_empty_dataset macro)

1. Raw dataset variables (mapped with Domain specification) PLUS all required/expected Domain specific variables (create if needed, such as FDOSEDTN or LBBLFL), values, and records.

2. Using control terminology, transform data to standard names and values.

3. Keep any unmapped raw dataset variables in SUPPQUAL corresponding datasets such as SUPPDM.

4. For unscheduled visits, assign visits if possible.

5. May be many to 1 relationship with CRF page.

6. Variable order is important, 8 char variable names, 40 char variable labels.

7. Create --DTC suffix for ISO8601 character date value. (See %make_dtc_date macro)

8. Create --DY suffix for study day variable = (date portion of --DTC) - (date portion of RFSTDTC) + 1 if --DTC is on or after RFSTDTC,

OR study day variable = (date portion of --DTC) - (date portion of RFSTDTC) if --DTC precedes RFSTDTC,

where RFSTDTC is a discrete starting point and RFRNDTC is a discrete ending point contained in the DM Domain.

Note that AESTDY may be missing for partial AESTDTC.

(See %make_sdtm_dy macro)

ADaM

'Analysis datasets for modeling and reporting - One-Proc Away'

(Useful for summary tables, statistical modeling and p-values)

2.0, 2.1 Reference, 2.1 presentation, 1.0 Guide,

Common Stats,

SAS,

2.1

Analysis Data Model

Three Models: Subject-level (ADSL), Change from baseline (ADLB) and Categorical analysis

Traceability is important to preserve to trace back to SDTMs. For example, keep SRCDOM and SRCVAR variables only if more than one value exists. Also, use caution when combining similar variables such as SDTM.EX.EXTRT and SDTM.CM.CMTRT into ADAM.ADCM.CMTRT.

(See %make_empty_dataset macro)

Mapping process from SDTM

1. Summary of SDTM datasets PLUS derived variables such as AGE, CHANGE IN BASELINE, SEX/SEXN, RACE/RACEN, etc. for TLGs.

2. Generally one record per patient or one record per patient per visit. May apply visit windows.

3. Identify BLFL, baseline flag and calculate CHANGE FROM BASELINE.

4. Using control terminology, transform data to standard names and values such as AVISIT, analysis visit. For traceability, it is good practice to keep the associated VISIT variable.

5. May create new records or visits such as DTYPE='AVERAGE' or 'LOCF'. (See 1. Proc Means, 2. Proc Sort, 3. Data Step Merge for programming techniques), impute with mean values or calculate total scores.

6. May be one to many relationship with SDTM.

7. Convert --DTC to --DT datetime variables. Use --DTF for date impute flag (D for day, M for month and day, and Y for year, month and day). (See %dtc2dt macro)

8. Create indicator or population flag variables --FN (1, 0) or --FL (Y, N). Other options include AE level flag variables such as AE1FL (patient level), AE2FL (organ class level) and AE3FL (preferred term level).

9. Metadata documentation (pseudo-code specifications) with hyperlinks - similar to define.xml.

LAB

Laboratory Data Model

ODM

1.3.2 Guide

Operational Data Model

(Contains both variable attributes and data values)

DEFINE.XML

Dictionary Tables,

vardef.sas (See diagram below)

Required for each SDTM and ADam dataset.

SAS Paper, SAS Paper 2, SAS Paper 3

Clinical Standards Toolkit Doc

XML Online Tutorial, XML Online reference

ISO 8601 Date Format

SAS Paper, SAS Paper 2, SAS Paper 3

XXDTC for date of XX

XXSTDTC for start date

XXENDTC for end date

See %make_dtc_date in SDTM SAS macros

See %dtc2dt in ADaM SAS macro

ISODATE9 character, ex. YYYY-MM-DD or YYYY-MM-DDThh:mm:ss, 2009-01-01T12:00:00

Three step process

1. Collected as text string, not as a date variable. May use informats such as b8601da8. for basic or e8601da10. for extended (XXXX-XX-XX) to store as datetime variable.

2. Use the same informats to display as text strings.

3. Can use CALL routine to calculate duration - CALL IS8601_CONVERT(convert-from, convert-to, <from-variables> , <to-variables>, <date-time-replacements>).

6 CDISC Classes based on type of data	SDTM/ADaM Domain
Special Purpose (Patient Attributes)	DM /ADSL, SE, CO, SV (Generally one record per patient per X)
Interventions (Sponsor Controlled, protocol planned treatment)	CM, EX/ADEX, SU (Generally multiple records per patient)
Events (Patient Controlled, Not planned)	AE /ADAE, DS, MH, DV, CE (Generally multiple records per patient) 1. Generally all upper case text 2. Numeric values such as 1, 2 or 3 instead of user defined formats such as MILD, MODERATE, SEVERE. 3. Generally requires relative times variables for partial dates --STRF [BEFORE, DURING, AFTER, U], --ENRF [BEFORE, DURING, AFTER, URING/AFTER, U])
Findings (Patient Measurements - Normalized, Vertical Structure)	LB /ADLB, EG, VS/ADVS, QS/ADQS, PC/ADPC etc. (Generally multiple records per patient) 1. Generally requires VISITNUM and or a sequence number 2. May also contain paired variables such as VSSTRESN and VSSTRESC 3. Almost always requires subsetting to process, ex. WHERE TESTCD = '';
Trial Design (Study Protocol/Compliance)	TE, TA, TV, TI, TS, TX (Generally one record per patient per X)
Special Relationship	SUPPQUAL (Supplemental Qualifer), RELREC (Relating Records) 1. One single dataset with Domain variable or for each Domain containing unmapped raw dataset variables (Domain specification) or non-standard raw variables. These may be extra variables not on the case report form. 2. Store free text values such as for LIST and OTHER in generic character variables - QNAM ($8), QLABEL ($40), QVAL ($200) and QORIG ($10). Example - XYZ123, DM, UNI101, RACEOTH, Race, Other, BRAZILIAN, CRF. 3. For multiple records per patient, there may be extra variables - IDVAR ($8), IDVARVAL ($3). Example - 'DMSEQ', DMSEQ. DMSEQ is an example of a traceability variable. DMSEQ is manually created based on key variable sort order. 4. Often linked with corresponding SDTM domain dataset by STUDYID, RDOMAIN (DM), and USUBJID for one record per patient or STUDYID, RDOMAIN, USUBJID, and IDVARVAL (XXSEQ) for multiple records per patient.

ADaM Datasets	Key Variables
ADSL (Paper)	Subject-Level
ADEX, ADAE, ADCM, ADMH, ADEG and ADLB	Multiple records per subject (Generally summary record of one record per patient per param (ex. lbtestcd) per visit)

Top Benefits of SAS Clinical Standards Toolkit
1. Create a define.xml file
2. Can run up to 143 standard SDTM, WebSDM and Janus checks for compliance on metadata, data values, date format, multiple-record, multiple-table and control term 1st level (Structure) - Domain and variable metadata requirements and unwanted variables 2nd level (Observational) - Data structure and value requirements Details - control terminology, column, column attribute, column value, date, metadata, multiple-record, multiple-tables.
3. Create customized standards
4. Create empty domain tables modeled after registered standard
5. Create custom validation checks

CDISC - SDTMs and ADaMs

Quick Links, CDISC References, SAS e-Guide

Mind Map

CDISC Mapping Videos

Getting Started with SDTMs