SDTM and ADaM are two CDISC data standards used in clinical trials. SDTM (Study Data Tabulation Model) organises the data collected in a trial into standard domains, such as DM for demographics and AE for adverse events. ADaM (Analysis Data Model) builds analysis-ready datasets from SDTM, such as ADSL with one row per subject, so statisticians can produce results. From ADaM, programmers create TLFs — the tables, listings and figures in the clinical study report. The US FDA expects SDTM and ADaM data, with Define-XML, for NDA, BLA and ANDA studies that started after 17 December 2016.
Key takeaways
- SDTM = collected data in standard domains; ADaM = analysis-ready data built from SDTM.
- The flow is raw data → SDTM → ADaM → TLFs, with traceability at every step.
- Current versions: SDTMIG v3.4 with SDTM v2.0; ADaMIG v1.3 with ADaM v2.1.
- Define-XML describes every dataset and variable; validators check conformance.
- TLFs follow the statistical analysis plan and are checked by independent programming.
On this page ⌄
How SDTM and ADaM fit into a clinical trial
A trial collects data from many sites through case report forms, lab vendors and devices. Each source looks different. To make the data easy to review, the Clinical Data Interchange Standards Consortium (CDISC) publishes standards that every sponsor can follow. A clinical SAS programmer moves the data through four steps:
- 1
Raw data
Data exported from the EDC system and external vendors, in each study’s own format.
- 2
SDTM
The same data, reorganised into standard CDISC domains with standard variable names and controlled terminology.
- 3
ADaM
Analysis-ready datasets derived from SDTM: treatment groups, baselines, changes and flags added.
- 4
TLFs
Tables, listings and figures produced from ADaM for the clinical study report and the submission.
New to the field? Start with what is clinical SAS, then come back here.
What is SDTM?
SDTM (Study Data Tabulation Model) is the CDISC standard for organising the data collected in a human clinical trial. It does not analyse anything: it puts each type of data into a fixed domain — a dataset with a two-letter code and standard variables. The current implementation guide is SDTMIG v3.4 (November 2021), based on SDTM v2.0.
Holds
- DM
- Demographics: age, sex, race, treatment arm, one row per subject
- AE
- Adverse events
- MH
- Medical history
- CM
- Concomitant medications
- EX
- Exposure to study treatment
- LB
- Laboratory test results
- VS
- Vital signs
- TS
- Trial summary, such as the study start date
Class
- DM
- Special-purpose
- AE
- Events
- MH
- Events
- CM
- Interventions
- EX
- Interventions
- LB
- Findings
- VS
- Findings
- TS
- Trial design
Every SDTM record carries identifiers such as STUDYID, DOMAIN and USUBJID (unique subject ID). Dates use the ISO 8601 format, for example AESTDTC for the start date of an adverse event. Many values must come from CDISC Controlled Terminology, and adverse event terms are coded with MedDRA: AETERM holds what the investigator wrote, AEDECOD the coded term.
What is ADaM?
ADaM (Analysis Data Model) is the CDISC standard for analysis datasets. CDISC’s principles are that ADaM datasets should be analysis-ready — a statistician can produce a result with little extra work — and traceable back to SDTM, so a reviewer can see where every value came from. The current versions are ADaMIG v1.3 (November 2021) with the ADaM model v2.1.
What it is
- ADSL
- Subject-level dataset: one row per subject, with treatment, population flags and key dates
- BDS (Basic Data Structure)
- One row per subject, parameter and time point — for labs, vitals, efficacy scores
- OCCDS (Occurrence Data Structure)
- For events and medications that occur, counted rather than measured
Example
- ADSL
- TRT01P (planned treatment), TRT01A (actual), SAFFL (safety population flag)
- BDS (Basic Data Structure)
- ADLB with PARAMCD, AVAL (value), BASE (baseline), CHG (change), ABLFL (baseline flag)
- OCCDS (Occurrence Data Structure)
- ADAE with TRTEMFL (treatment-emergent flag)
ADSL is built first, because almost every other ADaM dataset takes its treatment and population variables from it.
SDTM and ADaM: the key differences
SDTM
- Purpose
- Tabulate what was collected
- Source
- Raw EDC and vendor data
- Derived values
- Very few — mostly as collected
- Structure
- Domains such as DM, AE, LB
- Driven by
- The SDTM implementation guide
- Example
- LB: each lab result as reported
ADaM
- Purpose
- Support the statistical analysis
- Source
- SDTM datasets
- Derived values
- Many: baselines, changes, flags, analysis visits
- Structure
- ADSL, BDS and OCCDS datasets
- Driven by
- The statistical analysis plan (SAP), within ADaM rules
- Example
- ADLB: result, baseline and change from baseline per visit
A simple way to remember it: SDTM shows what happened; ADaM gets the data ready to answer the study’s questions.
TLFs: turning ADaM data into results
TLFs are the tables, listings and figures that present a trial’s results in the clinical study report. Their layout is planned in advance in mock shells, based on the statistical analysis plan, before the data is unblinded.
Example
- Table
- Demographics and baseline characteristics by treatment group
- Table
- Adverse events by system organ class and preferred term
- Table
- Change from baseline in a lab parameter by visit
- Listing
- All serious adverse events, subject by subject
- Figure
- Mean change over time, or a Kaplan-Meier survival curve
ADaM source
- Table
- ADSL
- Table
- ADAE + ADSL
- Table
- ADLB
- Listing
- ADAE
- Figure
- BDS dataset (e.g. ADTTE)
In industry, important outputs are checked by independent double programming: a second programmer writes separate code and the results are compared. SAS procedures such as PROC REPORT, PROC FREQ, PROC MEANS and PROC SGPLOT are the everyday tools.
Define-XML, validation and FDA requirements
- Define-XML is the metadata file that describes every dataset, variable, code list and derivation in an SDTM or ADaM package. Reviewers read it first.
- Validation: FDA checks submitted data against its business and validator rules. Programmers run the same kind of checks before submission, often with tools such as Pinnacle 21.
- FDA requirement: CDISC standards from FDA’s Data Standards Catalog apply to NDA, BLA and ANDA studies that started after 17 December 2016, and commercial IND studies that started after 17 December 2017.
- File format: datasets are sent as SAS Transport (XPT v5) files. FDA is testing CDISC’s Dataset-JSON as a possible future replacement.
- Nonclinical studies use a sister standard, SEND, rather than SDTM.
FDA requires the data standard, not a particular software. SAS remains the most common tool for this work; read clinical SAS certification and SAS vs clinical SAS for what employers look for.
How to learn SDTM and ADaM
- 1
Base SAS first
DATA step, merging, PROC SORT, PROC FREQ and PROC MEANS. You cannot map data without them.
- 2
SDTM mapping
Map sample raw data to DM, AE, CM, LB and VS, using the SDTMIG and controlled terminology.
- 3
ADaM derivation
Build ADSL, then a BDS dataset such as ADLB and an OCCDS dataset such as ADAE.
- 4
TLFs
Produce a demographics table, an AE table, a listing and one figure from your ADaM data.
- 5
Define-XML and validation
Describe your datasets and check them, as you would before a real submission.
You can practise SAS free in SAS OnDemand for Academics; see what is clinical SAS for how to start.
Learn SDTM, ADaM and TLFs with BCRI
BCRI’s Clinical SAS course online is a 3 months live online programme (₹38,000) covering Base and Advanced SAS, CDISC SDTM and ADaM, TLFs, Define-XML and Pinnacle 21, with recordings and placement assistance.
FAQs
What is the difference between SDTM and ADaM?
SDTM organises the data collected in a trial into standard domains, with few derived values. ADaM builds analysis-ready datasets from SDTM — adding baselines, changes and flags — so statisticians can produce results.
Which comes first, SDTM or ADaM?
SDTM. ADaM datasets are derived from SDTM, and CDISC expects them to be traceable back to it. TLFs are then produced from ADaM.
What is ADSL?
ADSL is the ADaM subject-level analysis dataset: one row per subject, with treatment groups, population flags such as SAFFL and key dates. Most other ADaM datasets use it.
What are TLFs in clinical trials?
Tables, listings and figures that present the trial’s results in the clinical study report, such as a demographics table, an adverse events table or a survival curve. They are produced from ADaM datasets.
What are the latest SDTM and ADaM versions?
As of November 2026, the latest released versions are SDTMIG v3.4 (with SDTM v2.0) and ADaMIG v1.3 (with ADaM v2.1). Always check which versions FDA’s Data Standards Catalog supports for your submission.
Is SDTM required by the FDA?
Yes, for studies covered by FDA’s standardised data requirement: NDA, BLA and ANDA studies that started after 17 December 2016 and commercial IND studies after 17 December 2017 must use the standards in the Data Standards Catalog, including SDTM and ADaM.
Can I learn SDTM and ADaM without a programming background?
Yes. Start with Base SAS, then learn SDTM mapping and ADaM derivations on sample data. Many clinical SAS programmers come from pharmacy and life science degrees.