24 Exercise 5: ADAE

24.1 Goal

Extend the ADAE from the ADAE session with a severity grouping and a second treatment-emergent window, then build the non-emergent event GLPX-1 never gives you, and see what it does to the safety table.

24.2 Setup

Work from the project root. Both inputs are persisted:

library(tidyverse)
library(sdtm.terminology)

ae <- read_csv("data/sdtm/ae.csv", show_col_types = FALSE,
               col_types = cols(.default = col_character()))

adsl <- read_csv("data/adam/adsl.csv", show_col_types = FALSE)

Build the ADAE spine as in Example 1 before starting, or read the persisted copy from data/adam/adae.csv.

24.3 Task

(Guided) Pooled severity. Add SEVGR1 and SEVGR1N, pooling severity into two groups: Mild/Moderate and Severe. Look up SEVGRy in OCCDS v1.1 §3.2 (Table 3.2.8.1) and note its Core designation before you start.

Then answer: is SEVGR1 checkable against a CDISC codelist? Give the same reasoning you would give for ASEV, and name the marker in the OCCDS table that tells you.
(The trap) A window that changes the answer. The session used x = 0 in TRTSDT <= ASTDT <= TRTEDT + x. Suppose the SAP asks for a second flag using a 30-day follow-up window, alongside the first.

Start by finding the variable OCCDS provides for exactly this situation (a second treatment-emergent flag with a different cut-off) in §3.2, Table 3.2.5.3. Two similar-looking variables sit just below TRTEMFL; only one is for different cut-offs rather than for multiple periods. Derive the flag under that name, leaving TRTEMFL untouched.

Then compare the two flags and explain the result, carefully, because it is not the result the exercise appears to be fishing for. What does it tell you about validating a window choice against this dataset?
(Semi-guided) The event that came first. Take subject GLPX1-101-147, whose four events you saw in the session. Construct the scenario where their first event (the diarrhoea at day 51) had instead started 10 days before first dose.

Re-derive TRTEMFL, AOCCFL and AOCCIFL for that subject and show all three columns. Then answer in one sentence each:
- Which row now carries AOCCFL, and why did it move?
- Does AOCCIFL move, and why or why not?
- The subject still had four adverse events. Does the count of subjects in a treatment-emergent incidence table change?

24.4 Acceptance criteria

SEVGR1/SEVGR1N one-to-one, with the no-codelist reasoning stated and the OCCDS marker named
The correct OCCDS variable for a second cut-off identified, and the derivation written under that name with TRTEMFL unmodified
The comparison result correctly explained rather than assumed
The modified subject yields TRTEMFL = "N" on the pre-treatment row, with AOCCFL moved and AOCCIFL unchanged
All three follow-up questions answered

Hint

Task 1: the structural rule for a grouping and its numeric counterpart is the same one AGEGR1/AGEGR1N followed in ADSL. For the codelist question, look at the footnote attached to the asterisk in the OCCDS table.

Task 2: derive it, then count() both flags together before writing any explanation. Every GLPX-1 event sits well inside the treatment window - the session says so explicitly. For the variable name, read the rows immediately below TRTEMFL in Table 3.2.5.3.

Task 3: you are changing one date, not three flags. Shift ASTDT for that one row, then re-run the derivations you already wrote and let them produce the answers. AOCCFL keys on the first treatment-emergent record; ask what the flags are sorted and filtered by before they are assigned.

Solutions are in the paid tier.