eduKateSG · Why Science?
Organise clinical-study observations into reviewable standard domains—while preserving source traceability, versions, terminology, timing and uncertainty
Choose the closest route, then return to the full index whenever you need the wider evidence chain.
Science learning becomes powerful when students can tell a measured pattern from a model, a statistical decision from a mechanism, and a useful lead from a finished conclusion. The Clinical Data Interchange Standards Consortium Study Data Tabulation Model—CDISC SDTM—provides a standard model for organising and formatting study data. CDISC published SDTM v2.1 on 10 June 2024; SDTM v3.0 and SDTMIG v4.0 remain in development with projected publication in 2027 as of 12 October 2026. For human clinical trials, SDTM is used with the SDTM Implementation Guide, controlled terminology and metadata. Standard tables improve review and exchange, but they do not prove that a trial was well designed, that source data are correct, that transformations were lossless or that an intervention is safe or effective. This guide supports Primary Science, PSLE Science, Secondary Science, O-Level Science and STEM education as a longform evidence-reading route; it is not medical advice, a laboratory protocol, an admission promise or a career guarantee.
Related eduKate reading: SDRF-Proteomics and sample metadata; PSI-MS and controlled scientific vocabulary; GA4GH Data Connect and federated dataset queries; GA4GH Phenopackets and computable case representation; Science Learning Hub; Education Hub; STEM Education; Careers by Subject Capability.
Primary and official sources checked on 12 October 2026: CDISC official SDTM v2.1 page, published 10 June 2024, checked 12 October 2026; CDISC official SDTMIG v3.4 page, checked 12 October 2026; CDISC official standards-in-development page, SDTM v3.0 projected for 2027, checked 12 October 2026; CDISC official controlled terminology update dated 25 September 2026, checked 12 October 2026; 2026 Singapore–Cambridge O-Level Biology syllabus; MOE G2/G3 Lower Secondary Science teaching and learning syllabus; 2026 MOE G2 Computing syllabus. Standards, terminologies, implementation guides, tools and recommendations can change, so preserve dated documentation, exact source data and software versions.
A deeper reading route
Clinical studies produce participant characteristics, exposures, laboratory results, adverse events, vital signs, questionnaires, outcomes and operational records. If every sponsor invents unrelated tables, reviewers spend time decoding structure instead of evaluating evidence. SDTM supplies a common conceptual model and domain organisation for tabulation datasets.
Versioning is central. The official CDISC page records SDTM v2.1 as published on 10 June 2024. SDTMIG v3.4 is a published implementation guide for human clinical trials. CDISC also lists SDTM v3.0 and SDTMIG v4.0 as in development with projected publication in 2027. A public-review artifact must never be represented as the current published production standard.
Domains group related observations. Two-character domain codes, variables and observation classes provide regularity across demographics, interventions, events, findings, trial design and relationships. The same-looking variable can have domain-specific rules, and not every collected item belongs in a new custom variable.
Identifiers and sequence numbers preserve context. Study, domain, subject and record keys let rows be traced and related. Timing variables can represent study day, date-time, duration, epoch or reference points. Converting partial dates or local timestamps into false precision can make a clean table scientifically misleading.
Controlled terminology provides codelists and valid values for defined items. CDISC and NCI Enterprise Vocabulary Services publish updates separately from the foundational model. The 25 September 2026 terminology release replaced older files for several CDISC standards. A dataset should state the terminology package version instead of saying simply “current terminology.”
Original and standardised results answer different needs. Laboratory domains can retain an original result and unit while also carrying a standard result and unit. Conversion formulas, significant figures and handling of values below or above quantification limits need documented rules. A standard unit is useful only if the transformation is correct.
Supplemental qualifiers can carry non-standard variables, but they are not a licence to avoid modelling. Relationship datasets connect records across domains when the link is not otherwise represented. Reviewers need metadata explaining origin, derivation, controlled terminology, value-level rules and dataset purpose.
Source-to-SDTM transformation is an evidence chain. Electronic data capture fields, central laboratory transfers, adjudication results and external devices can have different identities and timestamps. Mapping specifications should show how each target variable was populated and how discrepancies, missingness and updates were handled.
Conformance rules and validators can detect structural problems, invalid terms and inconsistent values. They cannot determine whether a participant had the reported event or whether the study conclusion is causal. Structural validation, medical review, statistical review and source verification remain different layers.
The optimistic value is reviewable evidence. Standard domains and terminology let teams compare like with like, automate checks and focus human attention on unusual patterns and scientific questions. Clear provenance and conservative interpretation keep a tidy submission from being mistaken for guaranteed truth.
Protocol amendments can change visits, assessments, endpoints or eligibility while subjects are enrolled. SDTM datasets should preserve actual timing and arm or element context, and metadata should explain amendment-dependent fields. Forcing every subject into the final schedule can erase the study that was actually conducted.
Partial dates are scientific information. A month known without a day is not the same as a complete date. Imputation may be appropriate later for analysis under a documented rule, but tabulation should not invent precision merely to satisfy a software datatype or sort order.
Unscheduled visits and repeated assessments need stable rules. Visit labels, numbers and timing should support review without pretending that unplanned events occurred at scheduled timepoints. Test subjects with missed, repeated and out-of-window visits, not only the ideal schedule.
Adverse-event records require careful distinction among occurrence, seriousness, severity, action, outcome and causality assessment. These variables answer different questions. A severe event is not automatically serious, and a reported relationship does not by itself establish that the study treatment caused the event.
Laboratory normal ranges can depend on laboratory, method, unit, age, sex and time. Standardising a result without its range and method can create misleading flags. Preserve the original range, unit and source laboratory while documenting any standard-unit conversion.
Coding dictionaries and terminology updates can change preferred terms or hierarchy. Store the dictionary name and version used for each coded result. Recode into a new copy, compare changes and obtain medical review rather than silently refreshing terminology shortly before submission.
Dataset metadata and reviewer guides should explain known validation findings, sponsor-defined extensions and important derivations. An unexplained warning creates uncertainty; a reasoned explanation makes the decision auditable. Explanations cannot legitimise a rule violation that changes meaning, so remediation remains the first choice.
A publication dataset may differ from a regulatory submission because privacy, licensing and purpose differ. Document removed variables, date transformations, coding changes and subject-ID remapping. Re-run structural checks and key scientific summaries after disclosure control.
Record-level lineage should survive dataset assembly and updates. If a source correction changes one result, the team should identify the affected SDTM row, related records, metadata and downstream analyses without rebuilding the story from memory. Stable source keys, controlled change logs and repeatable mappings make a corrected submission more trustworthy than an unexplained clean replacement.
Cross-domain checks catch contradictions that single-table validators miss. Compare death dates with later assessments, treatment timing with study participation, laboratory specimens with collection events and adverse-event dates with exposure windows. An inconsistency can reflect source error, mapping error or a legitimate exception; the review process should preserve that distinction.
| Stage | Question | Evidence | Responsible output |
|---|---|---|---|
| Collect | What happened in the study? | Source forms, devices, laboratories and operational records | A traceable source observation |
| Tabulate | Where does the observation belong? | SDTM version, domain, identifiers, timing and terminology | A standard row |
| Describe | How was the row created? | Define metadata, mapping rule, origin, derivation and codelist version | A reviewable transformation |
| Interpret | What conclusion is supported? | Protocol, analysis, missingness, validation and qualified review | A bounded study claim |
Inside this guide
1–12 · Foundations and design
13–24 · Models and estimation
- 13. Read trial-design domains
- 14. Read timing variables
- 15. Read original results
- 16. Read standard results
- 17. Use controlled terminology
- 18. Preserve source terms
- 19. Use supplemental qualifiers cautiously
- 20. Use relationships explicitly
- 21. Publish metadata
- 22. Trace every variable
- 23. Validate conformance
- 24. Review missingness
25–36 · Diagnostics and interpretation
- 25. Protect participant privacy
- 26. Separate tabulation from analysis
- 27. Separate conformance from evidence quality
- 28. Connect to Primary Science
- 29. Connect to Secondary Science
- 30. Connect to mathematics
- 31. Connect to computing
- 32. Connect to pathways
- 33. Did you know?
- 34. Write a reader-ready methods note
- 35. Build a validation ladder
- 36. Finish with a reproducible bundle
Section 1 of 36
1. Start with the study question
Name the protocol, population, observation and review purpose before designing datasets.
Working checkpoint — CDISC SDTM: Write the intended evidence chain. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 2 of 36
2. Pin the SDTM version
SDTM v2.1 is published, while SDTM v3.0 remains in development with projected publication in 2027.
Working checkpoint — clinical study data: Record model, guide and terminology versions. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 3 of 36
3. Pair model with implementation guide
SDTM supplies the general model and SDTMIG guides human clinical-trial tabulation.
Working checkpoint — SDTM domain: Use compatible published versions. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 4 of 36
4. Identify the study
STUDYID anchors records to the correct study and should agree across datasets.
Working checkpoint — SDTMIG: Reconcile against protocol and metadata. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 5 of 36
5. Identify the domain
DOMAIN states the two-character domain code for a tabulation dataset.
Working checkpoint — controlled terminology: Reject rows in the wrong domain. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 6 of 36
6. Identify the subject
USUBJID should uniquely identify a subject across the study submission context.
Working checkpoint — regulatory submission: Never expose source personal identifiers. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 7 of 36
7. Sequence records
Sequence variables distinguish records when business keys are not sufficient.
Working checkpoint — CDISC SDTM: Keep stable rules and document regeneration. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 8 of 36
8. Read observation classes
Interventions, events and findings have different structures and scientific meanings.
Working checkpoint — clinical study data: Do not move data merely to simplify programming. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 9 of 36
9. Read demographics
Demographic rows support population description but can carry sensitive or incomplete attributes.
Working checkpoint — SDTM domain: Apply protocol definitions and privacy controls. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 10 of 36
10. Read interventions
Exposure and concomitant-medication records need product, dose, route, timing and occurrence context.
Working checkpoint — SDTMIG: Separate planned from actual administration. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 11 of 36
11. Read events
Adverse events and other events require term, timing, seriousness and outcome fields under defined rules.
Working checkpoint — controlled terminology: Do not infer causality from occurrence alone. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 12 of 36
12. Read findings
Laboratory, vital-sign and test results require test identity, result, unit, method and specimen context.
Working checkpoint — regulatory submission: Keep original and standard values distinct. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 13 of 36
13. Read trial-design domains
Arms, elements, visits and disease milestones describe planned structure.
Working checkpoint — CDISC SDTM: Compare planned structure with actual subject data. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 14 of 36
14. Read timing variables
Dates, study days, durations, epochs and reference points each answer different questions.
Working checkpoint — clinical study data: Preserve partial precision and timezone context. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 15 of 36
15. Read original results
Original values and units preserve how a result arrived from its source.
Working checkpoint — SDTM domain: Do not overwrite with standardised display values. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 16 of 36
16. Read standard results
Standard values enable comparison only after a documented, scientifically valid conversion.
Working checkpoint — SDTMIG: Test formulas, units and rounding. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 17 of 36
17. Use controlled terminology
Codelists supply permitted or expected values for defined variables and change on their own release schedule.
Working checkpoint — controlled terminology: Pin the exact terminology package. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 18 of 36
18. Preserve source terms
Reported or verbatim terms keep source wording beside coded or normalised concepts.
Working checkpoint — regulatory submission: Document coding dictionary and version. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 19 of 36
19. Use supplemental qualifiers cautiously
SUPPQUAL can carry permitted non-standard qualifiers but can obscure an avoidable modelling problem.
Working checkpoint — CDISC SDTM: Justify each qualifier and parent link. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 20 of 36
20. Use relationships explicitly
RELREC and relationship variables can connect records whose links are not otherwise represented.
Working checkpoint — clinical study data: Test both directions and missing targets. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 21 of 36
21. Publish metadata
Dataset and variable metadata should describe labels, roles, origins, derivations, codelists and value-level conditions.
Working checkpoint — SDTM domain: Validate metadata against actual files. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 22 of 36
22. Trace every variable
Each target value should resolve to a source, direct mapping, controlled transformation or documented derivation.
Working checkpoint — SDTMIG: Sample the full path in both directions. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 23 of 36
23. Validate conformance
Version-aware rules can detect structure, terminology and cross-record inconsistencies.
Working checkpoint — controlled terminology: Keep all errors, warnings and explanations. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 24 of 36
24. Review missingness
Blank, unknown, not done and not applicable are not interchangeable states.
Working checkpoint — regulatory submission: Use protocol and domain rules. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 25 of 36
25. Protect participant privacy
Dates, free text, rare conditions and identifiers can re-identify participants.
Working checkpoint — CDISC SDTM: Apply consent, minimisation and governed access. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 26 of 36
26. Separate tabulation from analysis
SDTM is designed for tabulation; analysis datasets and statistical methods answer different questions.
Working checkpoint — clinical study data: Keep analysis derivations and traceability separate. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 27 of 36
27. Separate conformance from evidence quality
A clean validation report cannot prove source accuracy, trial integrity, benefit or safety.
Working checkpoint — SDTM domain: Combine structural, medical and statistical review. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 28 of 36
28. Connect to Primary Science
A fictional fictional clinical-study table helps younger learners separate a real-world observation from the record used to describe it.
Working checkpoint — SDTMIG: Use invented data and define every label. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 29 of 36
29. Connect to Secondary Science
SDTM supports lessons about measurement, models, evidence, systems and uncertainty.
Working checkpoint — controlled terminology: Trace one statement back to its source field. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 30 of 36
30. Connect to mathematics
Counts, dates, units, intervals, codes and probabilities require compatible definitions before comparison.
Working checkpoint — regulatory submission: Recalculate one small example by hand. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 31 of 36
31. Connect to computing
Domains, keys, dates, units, codelists and metadata show how observations become reviewable datasets.
Working checkpoint — CDISC SDTM: Break one field or reference, validate it and repair it transparently. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Section 32 of 36
32. Connect to pathways
Clinical-data standards connect biology, medicine, statistics, regulation and data engineering and can illuminate STEM education and career pathways.
Working checkpoint — clinical study data: Check current official education and qualification routes separately. Read every table row beside subject, domain, timing, origin, unit and source terminology.
For future study, connect biology to statistics and regulation while checking pathways separately.
Section 33 of 36
33. Did you know?
CDISC controlled terminology has its own release cycle, so a dataset needs both a model version and a terminology-package date.
Working checkpoint — SDTM domain: Explain the fact in plain language, then name what it does not prove. Preserve source terms and original results; standardisation is a traceable transformation.
For writing, cite protocol, SDTM version, guide, terminology, mapping and validation.
Section 34 of 36
34. Write a reader-ready methods note
State standard version, source system, transformation, terminology, validation, privacy controls and supported claim.
Working checkpoint — SDTMIG: Keep the note beside the machine-readable record. Separate conformance from source accuracy, study design, analysis and benefit-risk conclusions.
For students, ask how one laboratory result travels from collection to a study table.
Section 35 of 36
35. Build a validation ladder
Check bytes, syntax, identifiers, references, terminology, scientific plausibility, privacy and a known fixture in sequence.
Working checkpoint — controlled terminology: Stop at the earliest unresolved layer. Treat domain, keys, dates, units and codelists as scientific inputs: one mapping can change a review.
For parents, ask whether a neat table still shows missing and transformed values.
Section 36 of 36
36. Finish with a reproducible bundle
Deliver specifications, source receipts, mappings, software versions, tests, hashes and limitations without exposing governed records.
Working checkpoint — regulatory submission: Make every published claim traceable. Keep protocol, source record, mapping specification, SDTM and SDTMIG versions, terminology package and reviewer explanation together.
For a classroom model, tabulate a fictional plant-growth experiment.
Contents · Previous section · Continue to the Science Learning Hub