Summary: CDISC SDTM vs. ADaM at a Glance

In regulated clinical trials submitted to the FDA, EMA, and PMDA, CDISC SDTM (Study Data Tabulation Model) organizes raw Case Report Form (CRF) observations into standardized tabulation domains (DM, AE, EX, LB, VS) with minimal alteration. CDISC ADaM (Analysis Data Model) transforms those SDTM datasets into analysis-ready structures—anchored by the subject-level ADSL dataset and Basic Data Structure (BDS) datasets—containing derived baseline flags, imputed dates, and population indicators so statistical procedures (PROC FREQ, PROC MEANS, PROC LIFETEST) run with one-step traceability.

CDISC CLINICAL DATA ARCHITECTURE: SDTM TABULATION VS. ADaM ANALYSIS 1. CDISC SDTM DATASETS (Tabulation) Purpose: Standardized Raw CRF Observations • Special Purpose: DM (Demographics), SE, CO • Interventions: EX (Exposure), CM (ConMed) • Events: AE (Adverse Events), DS, MH • Findings: LB (Labs), VS (Vitals), EG Rule: Zero Imputation • ISO 8601 Dates (--DTC) 2. CDISC ADaM DATASETS (Analysis-Ready) Purpose: One-Step Statistical TLF Generation • Master Subject Key: ADSL (1 Row per Subject) • Population Flags: SAFFL, ITTFL, PPROTFL • BDS Longitudinal: ADLB, ADVS, ADTTE (AVAL, CHG) • OCCDS Safety: ADAE (Derived TRTEMFL='Y') Rule: 100% Traceable to SDTM via define.xml

If you are transitioning into clinical SAS programming or pharmaceutical biostatistics, fluency in CDISC standards is mandatory. Every electronic New Drug Application (eCTD) submitted to the US FDA or Japan PMDA relies on two distinct dataset layers: SDTM and ADaM.

1. What is CDISC SDTM (Study Data Tabulation Model)? #

SDTM is the regulatory standard for organizing raw clinical trial data collected on electronic Case Report Forms (eCRFs) and external vendor feeds. SDTM datasets are observation-level and grouped into four primary domain classes:

  • Special-Purpose Domains: DM (Demographics — one record per subject), CO (Comments), and SE (Subject Elements).
  • Interventions Domains: EX (Exposure to study drug) and CM (Concomitant Medications).
  • Events Domains: AE (Adverse Events — one record per event occurrence), DS (Disposition), and MH (Medical History).
  • Findings Domains: VS (Vital Signs), LB (Laboratory Test Results), and EG (ECG Measurements).

2. What is CDISC ADaM (Analysis Data Model)? #

While SDTM answers "What raw observations were collected?", regulatory reviewers and biostatisticians cannot run summary procedures directly on SDTM datasets. Why? Because raw SDTM lacks imputed numeric dates, baseline flags, change-from-baseline calculations, and treatment-emergent flags.

ADaM datasets are derived directly from SDTM domains so that statistical procedures (PROC MEANS, PROC FREQ, PROC LIFETEST) can generate Clinical Study Report (CSR) Tables, Listings, and Figures (TLFs) in a single step.

"SDTM captures what happened to the patient on the CRF. ADaM structures and derives the data so statisticians and FDA reviewers can verify clinical efficacy and safety in one step."

3. Core Differences: SDTM vs. ADaM Datasets #

Direct Summary: CDISC SDTM organizes raw Case Report Form clinical observations with zero imputation and ISO 8601 character dates, whereas CDISC ADaM derives analysis-ready datasets with numeric SAS dates, baseline flags, and subject denominators required for one-step statistical TLF generation.
Feature CDISC SDTM Datasets CDISC ADaM Datasets
Primary Purpose Standardized raw clinical observation tabulation One-step statistical analysis & TLF generation
Target Audience Data Managers & Clinical Operations Biostatisticians & Regulatory Reviewers (FDA, EMA)
Subject-Level Dataset DM (Demographics) ADSL (Subject-Level Analysis Dataset)
Safety Event Dataset AE (Raw Adverse Events) ADAE (Derived TRTEMFL, AESTDT, ASTDY)
Findings / Labs LB, VS (STRESN, --DTC) ADLB, ADVS (AVAL, BASE, CHG, ABLFL)
Date Storing Format ISO 8601 Character strings (--DTC) Numeric SAS dates (--DT) + SAP-driven imputation
Derivation Rules Minimal alteration; strictly aligned to raw CRF SAP-driven population flags, baseline changes & endpoints

4. How the ADSL Dataset Anchors Every Clinical Analysis #

Every clinical programming pipeline starts by building the ADSL (Subject-Level Analysis) dataset. Containing exactly one record per USUBJID, ADSL houses the master population flags (SAFFL for Safety, ITTFL for Intent-to-Treat), planned vs. actual treatment arms (TRT01P / TRT01A), and treatment start/end dates (TRTSDT / TRTEDT). Every downstream Basic Data Structure (BDS) dataset merges with ADSL to inherit these denominators.

🔬 Authoritative Knowledge Hub Blueprint

Deep Dive: CDISC SDTM vs. ADaM Architecture & ADSL Derivation Code

Explore complete SAS DATA step code examples for transforming SDTM.DM and SDTM.EX into ADaM.ADSL, along with FDA define.xml traceability rules in our Clinical SAS Knowledge Hub.