SDTM Domain Basics

Picture your first morning on a study. You receive a raw electronic data capture export from the data-entry screens, plus the Study Data Tabulation Model Implementation Guide, the SDTMIG, hundreds of pages long. Your job is to turn those raw rows into submission datasets. Raw rows give no clue about the target shape; the Study Data Tabulation Model, SDTM, model does, because it is a fixed table grammar that makes every dataset predictable, so a reviewer who has never seen your case report form, your CRF, can still read it. Today we cover domain classes, the topic–timing–qualifier skeleton, the Unique Subject Identifier, USUBJID, controlled terminology, Supplemental Qualifiers, SUPPQUAL, and a first-day reading order. All examples come from a simulated study.

In the Study Data Tabulation Model, or SDTM, a domain is one table holding one kind of observation. Adverse events, or AE, are rows in the AE domain; medications, or CM, are rows in CM; and vital signs, or VS, are rows in VS. Rule one: one row per observation per subject, so three reported adverse events generate three AE rows, while Demographics, or DM, carries exactly one row per subject. Rule two: data as collected, meaning no imputation and no analysis logic. For example, DM stores the collected Reference Start Date/Time, or RFSTDTC, as 2019-05-08T09:23, exactly as the electronic data capture, or EDC, export supplied it. Derived, analysis-ready values live one step downstream in the Analysis Data Model, or ADaM, not in SDTM.

The Study Data Tabulation Model Implementation Guide (SDTMIG) groups domains into five classes. Interventions answer what was administered, like Concomitant Medications (CM) and Exposure (EX); Events answer what happened, like Adverse Events (AE), Medical History (MH), and Disposition (DS). Findings answer what was measured or observed, such as Laboratory (LB), Vital Signs (VS), Electrocardiogram (EG), and Physical Examination (PE). Special-Purpose covers Demographics (DM) and Subject Visits (SV); Trial Design covers Trial Arms (TA), Trial Visits (TV), and Trial Summary (TS). Class predicts the variable list: identify it, then confirm in SDTMIG; DM has one row per subject, SV records actual visits, and DS and EX still belong to Events and Interventions. Trial Design is protocol data: TA carries ARMCD AIRIS1 and ETCD CYCLE1, and the TV row for CYCLE 1 DAY 5 begins '4 days after start of Treatment Epoch.'

In the Study Data Tabulation Model, or SDTM, every domain repeats a skeleton: identifiers, a topic variable, timing, qualifiers, and a standardized topic. Identifiers include STUDYID, the study identifier, and USUBJID, the unique subject identifier, plus --SEQ, the sequence number. The prefix changes by class: --TRT for interventions, --TERM for events, --TESTCD for findings, the same skeleton fits Concomitant Medications, Medical History, or Vital Signs. --DTC is a character International Organization for Standardization 8601 string, because partial dates ship as collected. --DY is study day relative to the reference start date: day one is on or after treatment, negative down to minus one, and no day zero. In the Subject Visits extract, the demo subject shows the subject visit study day, SVSTDY, equal to minus six on 2019-05-02, and --SEQ must be deterministic because later tables link back to it.

The Unique Subject Identifier, USUBJID, is built once in Demographics, DM, and reused verbatim. Each USUBJID is the study-site-subject key: STUDYID, the study identifier, a hyphen, then SUBJID, the subject identifier. DM holds one row per subject; Subject Visits, SV, holds five rows for that same demo subject and carries the identical string, so every domain stays consistent. Study day comes from DM's reference start date, RFSTDTC: Cycle 1 Day 1 on 2019-05-08 gives SVSTDY, the subject visit study day, equal to 1; Day 5 gives 5; Day 8 gives 8. A missing value is a data query, not a programming problem: one demo subject has RFSTDTC 2019-12-19T10:25 and an empty RFENDTC, the reference end date. Where no matching rows exist, we state the rule; we never fabricate patient records.

This is your checkpoint quiz on reading the Study Data Tabulation Model, or SDTM, and its domain structure. For a domain, the one-row-per-observation rule means the correct answer is B: each row is one recorded observation, such as one event, one finding, or one dose administration, which sets row granularity so a subject can have many rows in a domain. For the adverse event, or AE, dataset, which contains one row per reported adverse event, the correct answer is B: what happened to the subject during the study, because an adverse event is something that happens to a subject and AE belongs to the Events class. Which statements correctly identify identifier, topic, timing, and qualifier roles in the standard SDTM domain skeleton? The correct answers are A, B, and C: Unique Subject Identifier, or USUBJID, is an identifier; --DTC is a timing variable for when the observation occurred; and in an Events domain, --TERM is the topic variable that names the specific event, following the standard role pattern of identifiers, topic variable, timing variables, then qualifier variables.

Controlled Terminology, or CT, is the dated, versioned set of codelists published by CDISC, the Clinical Data Interchange Standards Consortium, and tied to your Implementation Guide version. Your study locks one CT package at the specification stage. Raw collected values are never moved into SDTM, the Study Data Tabulation Model, as-is. In the raw Electronic Data Capture extract, one column pair holds code M with display label Male; your mapping specification needs decision rows that map collected value to submission value. For SEX that means M and Male map to M, and the submitted DM, Demographics, dataset ships M from the codelist. Remember: no decision row, no mapping; SEX codelist values are M and F, and ETHNIC codelist values are HISPANIC OR LATINO, NOT HISPANIC OR LATINO, and NOT REPORTED.

Some case report form (CRF) fields have no variable in the Study Data Tabulation Model Implementation Guide (SDTMIG). Those values go vertical into a supplemental qualifier dataset, such as SUPPAE, SUPPCM, or SUPPVS — one row per stored value, not an extra column. RDOMAIN and the unique subject identifier, USUBJID, point to the parent domain and subject; IDVAR and IDVARVAL route back to the parent record, usually through --SEQ. QNAM, at most eight characters, and QLABEL name the qualifier; QVAL, QORIG, and QEVAL carry the value, origin, and evaluator. In SAS, sort by usubjid idvarval qnam, transpose with id=qnam and var=qval, then merge the wide result back onto the adverse event (AE) domain by USUBJID. Then compare row counts before and after that merge, and convert IDVARVAL from character to numeric before comparing it with --SEQ.

SDTMIG, the Study Data Tabulation Model Implementation Guide, is your human-readable rulebook: per-domain chapters, required, expected, and permissible variables, plus assumptions. define-XML is the machine-readable profile of your actual datasets: variables, origins, codelist bindings, and computational methods. So SDTMIG says what the Concomitant Medications domain, or CM, should look like in general; define-XML says what your CM looks like in particular. First-day reading order: protocol synopsis and schedule of activities, annotated CRF, SDTMIG chapter for each domain, then study mapping specification. Modern flow keeps the mapping spec structured and versioned, generates define-XML from that same metadata, and runs CDISC CORE validation on every build; reviewers and validation engines read it alongside the data. Agentic drafting is fast but risky: mapped variables must trace to a spec row, and --DTC strings must match the collected string byte for byte.

This exercise is hands-on on the website at jaimeyan.com/learn, where you classify twelve domain codes into Interventions, Events, Findings, Special-Purpose, and Trial Design buckets, then match variable-role cards to a domain prefix, and finish with a join-key mini-round. That means you practice the standard terminology like --TERM for events, --TRT for interventions, and --TESTCD for findings, and you choose the identifier that connects Demographics (DM) to Subject Visits (SV). The correct join key is the Unique Subject Identifier (USUBJID), the Study Identifier-Subject Identifier composite built once and reused verbatim; wrong answers trigger the domain-class or skeleton rule as immediate feedback, and no patient data appears beyond the provided extract, so try it yourself after the video.

This is your knowledge check on the whole Study Data Tabulation Model (SDTM) model. For a draft Study Data Tabulation Model Implementation Guide (SDTMIG) domain with code XX, where the only pattern is --TESTCD and --ORRES and there are no planned-dose variables, the correct answer is A: map XX to the Findings class, because --TESTCD names the measurement or test and --ORRES is the original result, which is the classic observation pattern. For the Unique Subject Identifier (USUBJID) composition and reuse question, the correct answers are A, B, and C: USUBJID is derived once at the subject level and reused unchanged in every SDTM domain, it is the key that links the same subject's records across domains, and in a multi-site study it typically combines site information and a subject number so it stays unique across sites. In the Supplemental Qualifiers for the Adverse Events (AE) domain, called SUPPAE, when IDVAR and IDVARVAL are both empty the QSPID qualifier is subject-level, so you associate it by USUBJID only and carry it down to every AE record for that subject; before merging SUPPAE with AE, run a uniqueness check on the SUPPAE merge key and a record-count or unmatched-record check to confirm no AE records are duplicated or lost.

Study Data Tabulation Model, or SDTM, is a fixed grammar: domains are one row per observation, grouped into Interventions, Events, and Findings, plus special-purpose and trial-design tables. Every domain repeats the same skeleton — identifiers, a topic variable, International Organization for Standardization 8601 timing, and qualifiers. Unique Subject Identifier, or USUBJID, is built once from Study Identifier and Subject Identifier composite and reused verbatim. If a value has no home in the Study Data Tabulation Model Implementation Guide, or SDTMIG, send it vertical into Supplemental Qualifiers, with Qualifier Variable Name of eight characters or fewer, linked to Sequence Number. Read study in order: protocol synopsis, annotated Case Report Form, SDTMIG chapter, then mapping specification. Full article is 'SDTM Domain Basics: The Model Every Clinical Programmer Must Know'; next is Adverse Events domain mapping worked example, and sdtm-mapping-conventions SKILL.md is distilled reference.