18 Course A: SDTM
Seven evenings, two hours each, one per week. You arrive with a raw clinical trial export and leave with four conformant SDTM domains and the machine-readable document that describes them.
Complete Before You Begin: Setup. It takes about 30 minutes and it must be done before the first evening, not during it. We will triage failures in the first ten minutes of Evening 1, but only for people who have tried.
You need R, RStudio and Quarto working, the repository cloned, and renv::restore() finished.
18.1 The schedule
| Ev | Session | What you build | Homework after |
|---|---|---|---|
| 1 | Why Standards | Nothing yet. You find the mess in the raw export that makes the rest necessary | Read The Trial background |
| 2 | The Trial: GLPX-1 | The trial’s shape: arms, visits, endpoints, and what its data should look like | Read SDTM Concepts background |
| 3 | SDTM Concepts | Observation classes, controlled terminology, the variable naming system | The two classification exercises in the session |
| 4 | The DM Domain | dm: identifiers, treatment dates, the study-day rule |
Exercise 1 |
| 5 | Events (AE) | ae: one row per event, and an event with no date |
Exercise 2 |
| 6 | Findings (LB and VS) | lb and vs: results, units, reference ranges, a duplicate and an outlier |
Exercise 3 |
| 7 | Define-XML | define.xml: the document that describes everything you just built |
Exercise 8 |
Evening 7 (Define-XML) is the only session that needs a file this repository cannot ship. You will need a free CDISC account and one download, done before that evening:
Unzip it into a refs/ folder at the project root, so that the schema sits at refs/DefineV2111_0/schema/cdisc-define-2.1/. It is a free account, not a paid membership, and refs/ is gitignored.
Without it you can still build the document; you will not be able to validate it against the CDISC schema, which is the best part of the evening. Start the account early.
18.2 How an evening runs
Two hours, with a ten-minute break in the middle.
| Time | What |
|---|---|
| 0:00 | Homework review and warm-up |
| 0:10 | Code along, first block |
| 0:45 | Short challenge in small groups |
| 0:55 | Break |
| 1:05 | Code along, second block |
| 1:40 | Discussion |
| 1:52 | Wrap and homework brief |
Bring a working laptop and type along. The sessions are built around “Run It Yourself” boxes: we stop, you run the code, and then we read the output together. Watching someone else type teaches very little.
18.3 What is expected between evenings
Roughly 45 to 60 minutes:
- The homework exercise from the table above
- Reading the Background section of the next session before you arrive
The Background sections are the reading-heavy parts. If you read them beforehand, the live time goes on code and questions instead of on me reading to you.
This is deliberate. Nothing is recorded, so discussion is open and nobody is performing for a camera.
The cost is that missing an evening genuinely costs you. If you have to miss one, work through that session in the book yourself before the next evening: every code chunk runs against data that ships with the repository, and the outputs shown in the book are the outputs you will get. The homework exercise is the check that you got there.
If you get stuck doing that, ask in GitHub Discussions rather than waiting for the next evening.
18.4 What you will have at the end
- Four SDTM domains built from a raw export, not handed to you finished:
dm,ae,lb,vs - A defensible answer for a duplicate subject, a missing unit, a sex value that disagrees between two sources, and an adverse event with no start date
- A schema-valid Define-XML v2.1 document you assembled element by element
- The habit of checking a claim against the Implementation Guide section rather than against your memory
18.5 After Course A
Course B: ADaM and Outputs takes these datasets and builds the analysis layer on top of them. Course A stands on its own, and you can stop there having built a real SDTM.