Day 11: Disposition (DS) & Trial Design Domains

Screen Failures, Completers, and Study Structure

← Back to Roadmap

1 Learning Objectives

By the end of Day 11, you will be able to:

Build the DS (Disposition) domain - tracking every subject’s journey through the trial
Understand EPOCH and how it relates to study periods (SCREENING, TREATMENT, FOLLOW-UP)
Derive key disposition variables: DSDECOD, DSSTDTC, DSCAT, and completion flags
Survey the Trial Design domains (TA, TE, TV, TI, TS) and understand what each contains
Create a subject disposition flow - the essential “patient accounting” table
Understand how DS feeds into the ADaM ADSL dataset

2 Why Disposition Matters

The DS domain answers the most fundamental question in a clinical trial: What happened to each subject?

Regulatory Requirement

The FDA requires a complete accounting of every subject who entered the study:

Did they complete the study? → COMPLETED
Did they withdraw early? → Why? (adverse event, lost to follow-up, withdrawal by subject, etc.)
Were they a screen failure? → Why? (did not meet inclusion criteria, etc.)

This information feeds into CONSORT diagrams, disposition tables, and the ADaM ADSL dataset - all critical for regulatory submissions.

3 Package Installation & Loading

if (!requireNamespace("dplyr", quietly = TRUE)) suppressMessages(install.packages("dplyr"))
if (!requireNamespace("lubridate", quietly = TRUE)) suppressMessages(install.packages("lubridate"))
if (!requireNamespace("tidyr", quietly = TRUE)) suppressMessages(install.packages("tidyr"))
if (!requireNamespace("pharmaversesdtm", quietly = TRUE)) suppressMessages(install.packages("pharmaversesdtm"))

library(dplyr)
library(lubridate)
library(tidyr)
library(pharmaversesdtm)

4 The DS Domain Structure

4.1 Key Variables

┌──────────────────────────────────────────────────────────────────────────────┐
│                        DS DOMAIN - KEY VARIABLES                             │
├──────────────────────────────────────────────────────────────────────────────┤
│  IDENTIFIER VARIABLES                                                        │
│    STUDYID    = Study identifier                                             │
│    USUBJID    = Unique subject identifier                                    │
│    DSSEQ      = Sequence number within subject                               │
├──────────────────────────────────────────────────────────────────────────────┤
│  TOPIC / RESULT VARIABLES                                                    │
│    DSTERM     = Reported disposition term (as collected)                      │
│    DSDECOD    = Standardized disposition term (CT)                            │
│    DSCAT      = Category: "DISPOSITION EVENT" or "PROTOCOL MILESTONE"        │
│    DSSCAT     = Subcategory: e.g., "STUDY PARTICIPATION", "TREATMENT"        │
├──────────────────────────────────────────────────────────────────────────────┤
│  TIMING VARIABLES                                                            │
│    DSSTDTC    = Date/time of disposition event (ISO 8601)                     │
│    DSDY       = Study day of disposition event                               │
│    EPOCH      = Epoch (SCREENING, TREATMENT, FOLLOW-UP)                      │
│    VISITNUM   = Visit number                                                 │
│    VISIT      = Visit name                                                   │
└──────────────────────────────────────────────────────────────────────────────┘

4.2 Understanding DSCAT and DSSCAT

These categorization variables are critical:

DSCAT	DSSCAT	DSDECOD Examples
`DISPOSITION EVENT`	`STUDY PARTICIPATION`	COMPLETED, SCREEN FAILURE
`DISPOSITION EVENT`	`TREATMENT`	COMPLETED, ADVERSE EVENT, WITHDRAWAL BY SUBJECT
`PROTOCOL MILESTONE`	`INFORMED CONSENT`	INFORMED CONSENT OBTAINED
`PROTOCOL MILESTONE`	`RANDOMIZATION`	RANDOMIZED

DSCAT vs DSSCAT

Think of it this way:

DSCAT = The broad type of event (disposition event vs. protocol milestone)
DSSCAT = Which part of the study this relates to (study participation vs. treatment)

A subject can have multiple DS records: one for each protocol milestone and each disposition decision point.

5 Exploring DS Data from pharmaversesdtm

# Load DS domain
data("ds", package = "pharmaversesdtm")
data("dm", package = "pharmaversesdtm")

cat("DS domain dimensions:", nrow(ds), "rows x", ncol(ds), "columns\n")

DS domain dimensions: 850 rows x 13 columns

cat("Number of unique subjects:", n_distinct(ds$USUBJID), "\n\n")

Number of unique subjects: 306

cat("Variables available:\n")

Variables available:

cat(paste(names(ds), collapse = ", "), "\n")

STUDYID, DOMAIN, USUBJID, DSSEQ, DSSPID, DSTERM, DSDECOD, DSCAT, VISITNUM, VISIT, DSDTC, DSSTDTC, DSSTDY

dplyr::glimpse(ds)

Rows: 850
Columns: 13
$ STUDYID  <chr> "CDISCPILOT01", "CDISCPILOT01", "CDISCPILOT01", "CDISCPILOT01…
$ DOMAIN   <chr> "DS", "DS", "DS", "DS", "DS", "DS", "DS", "DS", "DS", "DS", "…
$ USUBJID  <chr> "01-701-1015", "01-701-1015", "01-701-1015", "01-701-1023", "…
$ DSSEQ    <int> 1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 4…
$ DSSPID   <chr> NA, NA, NA, NA, "24", NA, NA, NA, NA, NA, NA, NA, NA, NA, NA,…
$ DSTERM   <chr> "RANDOMIZED", "PROTOCOL COMPLETED", "FINAL LAB VISIT", "RANDO…
$ DSDECOD  <chr> "RANDOMIZED", "COMPLETED", "FINAL LAB VISIT", "RANDOMIZED", "…
$ DSCAT    <chr> "PROTOCOL MILESTONE", "DISPOSITION EVENT", "OTHER EVENT", "PR…
$ VISITNUM <dbl> 3.0, 13.0, 13.0, 3.0, 5.0, 5.0, 201.0, 3.0, 13.0, 13.0, 3.0, …
$ VISIT    <chr> "BASELINE", "WEEK 26", "WEEK 26", "BASELINE", "WEEK 4", "WEEK…
$ DSDTC    <chr> "2014-01-02", "2014-07-02", "2014-07-02T11:45", "2012-08-05",…
$ DSSTDTC  <chr> "2014-01-02", "2014-07-02", "2014-07-02", "2012-08-05", "2012…
$ DSSTDY   <dbl> 1, 182, 182, 1, 29, 29, 198, 1, 180, 180, 1, 28, 28, 182, 1, …

5.1 Understanding the DS Records

# What categories exist?
cat("DS Categories:\n")

DS Categories:

# Use only columns that exist in the dataset
cat_vars <- intersect(c("DSCAT", "DSSCAT"), names(ds))
ds %>%
  count(across(all_of(cat_vars)), name = "Count") %>%
  arrange(across(all_of(cat_vars)))

# A tibble: 3 × 2
  DSCAT              Count
  <chr>              <int>
1 DISPOSITION EVENT    306
2 OTHER EVENT          290
3 PROTOCOL MILESTONE   254

# What disposition terms exist?
cat("Disposition terms (DSDECOD):\n")

Disposition terms (DSDECOD):

term_vars <- intersect(c("DSCAT", "DSSCAT", "DSDECOD"), names(ds))
ds %>%
  count(across(all_of(term_vars)), name = "Count") %>%
  arrange(across(all_of(term_vars[term_vars != "DSDECOD"])), desc(Count))

# A tibble: 13 × 3
   DSCAT              DSDECOD                     Count
   <chr>              <chr>                       <int>
 1 DISPOSITION EVENT  COMPLETED                     110
 2 DISPOSITION EVENT  ADVERSE EVENT                  92
 3 DISPOSITION EVENT  SCREEN FAILURE                 52
 4 DISPOSITION EVENT  WITHDRAWAL BY SUBJECT          27
 5 DISPOSITION EVENT  STUDY TERMINATED BY SPONSOR     7
 6 DISPOSITION EVENT  PROTOCOL VIOLATION              6
 7 DISPOSITION EVENT  LACK OF EFFICACY                4
 8 DISPOSITION EVENT  DEATH                           3
 9 DISPOSITION EVENT  PHYSICIAN DECISION              3
10 DISPOSITION EVENT  LOST TO FOLLOW-UP               2
11 OTHER EVENT        FINAL LAB VISIT               254
12 OTHER EVENT        FINAL RETRIEVAL VISIT          36
13 PROTOCOL MILESTONE RANDOMIZED                    254

Reading DS Data

Each subject typically has multiple records in DS:

Informed consent record - When they signed the ICF
Randomization record - When they were randomized (if applicable)
Treatment disposition - Did they complete treatment? If not, why?
Study disposition - Did they complete the study? If not, why?

These records tell the full story of each subject’s journey.

6 Understanding EPOCH

6.1 What is EPOCH?

EPOCH is a variable that indicates which period of the study the data belongs to. It divides the trial into distinct phases:

Timeline:
────────────────────────────────────────────────────────────────────
  SCREENING    │   TREATMENT (Run-in)   │   TREATMENT    │   FOLLOW-UP
  (Pre-study)  │   (Placebo lead-in)    │   (Active)     │   (Post-treatment)
────────────────────────────────────────────────────────────────────

# Check EPOCH values in DM (EPOCH is often in DM as well)
if ("EPOCH" %in% names(ds)) {
  cat("EPOCH values in DS:\n")
  ds %>%
    count(EPOCH, name = "Count") %>%
    arrange(desc(Count))
} else {
  cat("EPOCH not found in DS. Checking DM...\n")
  if ("EPOCH" %in% names(dm)) {
    cat("EPOCH values in DM:\n")
    dm %>% count(EPOCH) %>% print()
  }
}

EPOCH not found in DS. Checking DM...

6.2 EPOCH in the DM Domain

The DM domain has EPOCH-related variables that indicate the last known state of the subject:

# Key epoch/disposition-related variables in DM
dm_disp_vars <- intersect(
  c("USUBJID", "ARM", "ACTARM", "RFSTDTC", "RFENDTC", "DTHDTC", "DTHFL"),
  names(dm)
)
dm %>%
  select(all_of(dm_disp_vars)) %>%
  head(10)

# A tibble: 10 × 7
   USUBJID     ARM                  ACTARM          RFSTDTC RFENDTC DTHDTC DTHFL
   <chr>       <chr>                <chr>           <chr>   <chr>   <chr>  <chr>
 1 01-701-1015 Placebo              Placebo         2014-0… 2014-0… <NA>   <NA> 
 2 01-701-1023 Placebo              Placebo         2012-0… 2012-0… <NA>   <NA> 
 3 01-701-1028 Xanomeline High Dose Xanomeline Hig… 2013-0… 2014-0… <NA>   <NA> 
 4 01-701-1033 Xanomeline Low Dose  Xanomeline Low… 2014-0… 2014-0… <NA>   <NA> 
 5 01-701-1034 Xanomeline High Dose Xanomeline Hig… 2014-0… 2014-1… <NA>   <NA> 
 6 01-701-1047 Placebo              Placebo         2013-0… 2013-0… <NA>   <NA> 
 7 01-701-1057 Screen Failure       Screen Failure  <NA>    <NA>    <NA>   <NA> 
 8 01-701-1097 Xanomeline Low Dose  Xanomeline Low… 2014-0… 2014-0… <NA>   <NA> 
 9 01-701-1111 Xanomeline Low Dose  Xanomeline Low… 2012-0… 2012-0… <NA>   <NA> 
10 01-701-1115 Xanomeline Low Dose  Xanomeline Low… 2012-1… 2013-0… <NA>   <NA>

Important DM Variables Related to DS

Variable	Description
`RFSTDTC`	Reference start date (first exposure)
`RFENDTC`	Reference end date (last exposure)
`RFXSTDTC`	Date/time of first study treatment
`RFXENDTC`	Date/time of last study treatment
`DTHDTC`	Date of death
`DTHFL`	Death flag (Y/N)
`DSDECOD`	Disposition decoded term (in some implementations)

These DM variables are derived from the DS and EX domain data.

7 Building a DS Domain from Scratch

Let’s simulate raw disposition data and build a complete DS domain:

7.1 Step 1: Create Raw Disposition Data

set.seed(42)

n_subjects <- 30

# Generate subject disposition decisions
raw_disp <- tibble(
  STUDYID = "CDISC01",
  USUBJID = paste0("CDISC01-001-", sprintf("%03d", 1:n_subjects)),
  
  # Informed consent date
  CONSENT_DATE = as.Date("2024-01-01") + sample(0:30, n_subjects, replace = TRUE),
  
  # Screening result
  SCREEN_RESULT = sample(
    c("PASSED", "FAILED"),
    n_subjects, replace = TRUE, prob = c(0.85, 0.15)
  ),
  SCREEN_REASON = if_else(
    SCREEN_RESULT == "FAILED",
    sample(c("DID NOT MEET INCLUSION CRITERIA",
             "EXCLUSION CRITERIA MET",
             "WITHDREW CONSENT"),
           n_subjects, replace = TRUE, prob = c(0.5, 0.3, 0.2)),
    NA_character_
  ),
  
  # Randomization date (only for those who passed screening)
  RAND_DATE = if_else(
    SCREEN_RESULT == "PASSED",
    CONSENT_DATE + sample(7:21, n_subjects, replace = TRUE),
    as.Date(NA)
  ),
  
  # Treatment arm (only for randomized subjects)
  ARMCD = if_else(
    SCREEN_RESULT == "PASSED",
    sample(c("PBO", "ACT10", "ACT20"), n_subjects, replace = TRUE),
    NA_character_
  )
) %>%
  # Add first dose date
  mutate(
    FIRST_DOSE = if_else(!is.na(RAND_DATE), RAND_DATE + sample(0:3, n(), replace = TRUE), as.Date(NA))
  )

# Determine treatment disposition for randomized subjects
raw_disp <- raw_disp %>%
  mutate(
    TRT_RESULT = case_when(
      SCREEN_RESULT == "FAILED" ~ NA_character_,
      TRUE ~ sample(c("COMPLETED", "ADVERSE EVENT", "WITHDRAWAL BY SUBJECT",
                       "LOST TO FOLLOW-UP", "PHYSICIAN DECISION"),
                    n(), replace = TRUE,
                    prob = c(0.65, 0.12, 0.10, 0.08, 0.05))
    ),
    # End of treatment date
    TRT_END_DATE = case_when(
      is.na(FIRST_DOSE) ~ as.Date(NA),
      TRT_RESULT == "COMPLETED" ~ FIRST_DOSE + 84,  # 12-week study
      TRUE ~ FIRST_DOSE + sample(7:60, n(), replace = TRUE)  # Early termination
    )
  )

cat("Raw disposition data:\n")

Raw disposition data:

cat("Total subjects:", nrow(raw_disp), "\n")

Total subjects: 30

cat("Screen failures:", sum(raw_disp$SCREEN_RESULT == "FAILED"), "\n")

Screen failures: 5

cat("Randomized:", sum(raw_disp$SCREEN_RESULT == "PASSED"), "\n\n")

Randomized: 25

raw_disp %>%
  select(USUBJID, SCREEN_RESULT, TRT_RESULT) %>%
  head(10)

# A tibble: 10 × 3
   USUBJID         SCREEN_RESULT TRT_RESULT           
   <chr>           <chr>         <chr>                
 1 CDISC01-001-001 PASSED        WITHDRAWAL BY SUBJECT
 2 CDISC01-001-002 PASSED        COMPLETED            
 3 CDISC01-001-003 PASSED        COMPLETED            
 4 CDISC01-001-004 PASSED        COMPLETED            
 5 CDISC01-001-005 PASSED        COMPLETED            
 6 CDISC01-001-006 PASSED        COMPLETED            
 7 CDISC01-001-007 PASSED        COMPLETED            
 8 CDISC01-001-008 FAILED        <NA>                 
 9 CDISC01-001-009 PASSED        ADVERSE EVENT        
10 CDISC01-001-010 PASSED        LOST TO FOLLOW-UP

7.2 Step 2: Transform to SDTM DS

# Create DS records for each disposition event
# A subject may have MULTIPLE DS records

# Record 1: Informed Consent
ds_consent <- raw_disp %>%
  mutate(
    DOMAIN   = "DS",
    DSSEQ    = 1L,
    DSTERM   = "INFORMED CONSENT OBTAINED",
    DSDECOD  = "INFORMED CONSENT OBTAINED",
    DSCAT    = "PROTOCOL MILESTONE",
    DSSCAT   = "INFORMED CONSENT",
    DSSTDTC  = as.character(CONSENT_DATE),
    EPOCH    = "SCREENING"
  ) %>%
  select(STUDYID, DOMAIN, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC, EPOCH)

# Record 2: Randomization (only for those who passed screening)
ds_random <- raw_disp %>%
  filter(SCREEN_RESULT == "PASSED") %>%
  mutate(
    DOMAIN   = "DS",
    DSSEQ    = 2L,
    DSTERM   = "RANDOMIZED",
    DSDECOD  = "RANDOMIZED",
    DSCAT    = "PROTOCOL MILESTONE",
    DSSCAT   = "RANDOMIZATION",
    DSSTDTC  = as.character(RAND_DATE),
    EPOCH    = "SCREENING"
  ) %>%
  select(STUDYID, DOMAIN, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC, EPOCH)

# Record 3: Treatment Disposition
ds_treatment <- raw_disp %>%
  filter(!is.na(TRT_RESULT)) %>%
  mutate(
    DOMAIN   = "DS",
    DSSEQ    = 3L,
    DSTERM   = TRT_RESULT,
    DSDECOD  = TRT_RESULT,
    DSCAT    = "DISPOSITION EVENT",
    DSSCAT   = "TREATMENT",
    DSSTDTC  = as.character(TRT_END_DATE),
    EPOCH    = "TREATMENT"
  ) %>%
  select(STUDYID, DOMAIN, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC, EPOCH)

# Record 4: Screen Failure disposition
ds_screenfail <- raw_disp %>%
  filter(SCREEN_RESULT == "FAILED") %>%
  mutate(
    DOMAIN   = "DS",
    DSSEQ    = 2L,
    DSTERM   = SCREEN_REASON,
    DSDECOD  = "SCREEN FAILURE",
    DSCAT    = "DISPOSITION EVENT",
    DSSCAT   = "STUDY PARTICIPATION",
    DSSTDTC  = as.character(CONSENT_DATE + sample(1:14, n(), replace = TRUE)),
    EPOCH    = "SCREENING"
  ) %>%
  select(STUDYID, DOMAIN, USUBJID, DSSEQ, DSTERM, DSDECOD, DSCAT, DSSCAT, DSSTDTC, EPOCH)

# Combine all DS records
sdtm_ds <- bind_rows(ds_consent, ds_random, ds_treatment, ds_screenfail) %>%
  # Re-sequence within each subject
  arrange(STUDYID, USUBJID, DSSTDTC) %>%
  group_by(USUBJID) %>%
  mutate(DSSEQ = row_number()) %>%
  ungroup()

cat("SDTM DS domain created:\n")

SDTM DS domain created:

cat("Total records:", nrow(sdtm_ds), "\n")

Total records: 85

cat("Unique subjects:", n_distinct(sdtm_ds$USUBJID), "\n\n")

Unique subjects: 30

sdtm_ds %>%
  filter(USUBJID == first(sdtm_ds$USUBJID)) %>%
  print()

# A tibble: 3 × 10
  STUDYID DOMAIN USUBJID         DSSEQ DSTERM DSDECOD DSCAT DSSCAT DSSTDTC EPOCH
  <chr>   <chr>  <chr>           <int> <chr>  <chr>   <chr> <chr>  <chr>   <chr>
1 CDISC01 DS     CDISC01-001-001     1 INFOR… INFORM… PROT… INFOR… 2024-0… SCRE…
2 CDISC01 DS     CDISC01-001-001     2 RANDO… RANDOM… PROT… RANDO… 2024-0… SCRE…
3 CDISC01 DS     CDISC01-001-001     3 WITHD… WITHDR… DISP… TREAT… 2024-0… TREA…

7.3 Step 3: Validate DS Domain

cat("=== DS Domain Validation ===\n\n")

=== DS Domain Validation ===

# Check 1: Every subject has at least one record
cat("Subjects with DS records:", n_distinct(sdtm_ds$USUBJID), "\n")

Subjects with DS records: 30

# Check 2: Every subject has informed consent
consent_check <- sdtm_ds %>%
  filter(DSDECOD == "INFORMED CONSENT OBTAINED") %>%
  n_distinct(.$USUBJID)
cat("Subjects with informed consent:", consent_check, "\n\n")

Subjects with informed consent: 30

# Check 3: Disposition summary
cat("Treatment Disposition Summary:\n")

Treatment Disposition Summary:

sdtm_ds %>%
  filter(DSCAT == "DISPOSITION EVENT", DSSCAT == "TREATMENT") %>%
  count(DSDECOD, name = "Count") %>%
  mutate(Percent = round(100 * Count / sum(Count), 1)) %>%
  arrange(desc(Count))

# A tibble: 4 × 3
  DSDECOD               Count Percent
  <chr>                 <int>   <dbl>
1 COMPLETED                19      76
2 WITHDRAWAL BY SUBJECT     3      12
3 LOST TO FOLLOW-UP         2       8
4 ADVERSE EVENT             1       4

8 Subject Disposition Flow

This is one of the most important outputs in a clinical study report:

# Create a CONSORT-style flow summary
cat("=== SUBJECT DISPOSITION FLOW ===\n\n")

=== SUBJECT DISPOSITION FLOW ===

n_consented <- n_distinct(sdtm_ds$USUBJID)
cat("Subjects consented:         ", n_consented, "\n")

Subjects consented:          30

n_screened_out <- sdtm_ds %>%
  filter(DSDECOD == "SCREEN FAILURE") %>%
  n_distinct(.$USUBJID)
cat("  Screen failures:           ", n_screened_out, "\n")

  Screen failures:            5

# Screen failure reasons
if (n_screened_out > 0) {
  sdtm_ds %>%
    filter(DSDECOD == "SCREEN FAILURE") %>%
    count(DSTERM, name = "N") %>%
    arrange(desc(N)) %>%
    mutate(line = paste0("    - ", DSTERM, ": ", N)) %>%
    pull(line) %>%
    cat(sep = "\n")
  cat("\n")
}

    - EXCLUSION CRITERIA MET: 3
    - DID NOT MEET INCLUSION CRITERIA: 2

n_randomized <- sdtm_ds %>%
  filter(DSDECOD == "RANDOMIZED") %>%
  n_distinct(.$USUBJID)
cat("Subjects randomized:        ", n_randomized, "\n")

Subjects randomized:         25

n_completed <- sdtm_ds %>%
  filter(DSCAT == "DISPOSITION EVENT", DSSCAT == "TREATMENT", DSDECOD == "COMPLETED") %>%
  n_distinct(.$USUBJID)
cat("  Completed treatment:       ", n_completed, "\n")

  Completed treatment:        19

n_discontinued <- n_randomized - n_completed
cat("  Discontinued treatment:    ", n_discontinued, "\n")

  Discontinued treatment:     6

# Discontinuation reasons
if (n_discontinued > 0) {
  sdtm_ds %>%
    filter(DSCAT == "DISPOSITION EVENT", DSSCAT == "TREATMENT", DSDECOD != "COMPLETED") %>%
    count(DSDECOD, name = "N") %>%
    arrange(desc(N)) %>%
    mutate(line = paste0("    - ", DSDECOD, ": ", N)) %>%
    pull(line) %>%
    cat(sep = "\n")
  cat("\n")
}

    - WITHDRAWAL BY SUBJECT: 3
    - LOST TO FOLLOW-UP: 2
    - ADVERSE EVENT: 1

cat("\nCompletion rate:", round(100 * n_completed / n_randomized, 1), "%\n")


Completion rate: 76 %

9 Trial Design Domains Overview

9.1 What Are Trial Design Domains?

Trial Design domains describe the structure of the study itself, not individual subject data. Think of them as the blueprint of the clinical trial.

Domain	Name	Purpose
TA	Trial Arms	Describes each treatment arm and its sequence of elements
TE	Trial Elements	Describes the building blocks of each arm (e.g., washout, treatment)
TV	Trial Visits	Lists all planned visits and their timing windows
TI	Trial Inclusion/Exclusion	Documents the inclusion/exclusion criteria
TS	Trial Summary	Key parameters about the study (phase, indication, design)

When Do You Need These?

Trial Design domains are required for FDA submissions. They provide the context for understanding the subject-level data. An FDA reviewer needs to know:

What arms existed? (TA)
What was the study schedule? (TV)
What were the eligibility criteria? (TI)
What type of study was this? (TS)

9.2 TA - Trial Arms

# Create a Trial Arms (TA) domain example
ta <- tibble::tribble(
  ~STUDYID,  ~DOMAIN, ~ARMCD,  ~ARM,           ~TAESSION, ~TATRANS,       ~EPOCH,       ~ELEMENT,
  "CDISC01", "TA",    "PBO",   "Placebo",      1L,        "No criteria",  "SCREENING",  "Screen",
  "CDISC01", "TA",    "PBO",   "Placebo",      2L,        "Randomized",   "TREATMENT",  "Placebo",
  "CDISC01", "TA",    "PBO",   "Placebo",      3L,        "End treatment","FOLLOW-UP",  "Follow-up",
  "CDISC01", "TA",    "ACT10", "Active 10mg",  1L,        "No criteria",  "SCREENING",  "Screen",
  "CDISC01", "TA",    "ACT10", "Active 10mg",  2L,        "Randomized",   "TREATMENT",  "Active 10mg",
  "CDISC01", "TA",    "ACT10", "Active 10mg",  3L,        "End treatment","FOLLOW-UP",  "Follow-up",
  "CDISC01", "TA",    "ACT20", "Active 20mg",  1L,        "No criteria",  "SCREENING",  "Screen",
  "CDISC01", "TA",    "ACT20", "Active 20mg",  2L,        "Randomized",   "TREATMENT",  "Active 20mg",
  "CDISC01", "TA",    "ACT20", "Active 20mg",  3L,        "End treatment","FOLLOW-UP",  "Follow-up"
)

cat("Trial Arms (TA):\n")

Trial Arms (TA):

print(ta)

# A tibble: 9 × 8
  STUDYID DOMAIN ARMCD ARM         TAESSION TATRANS       EPOCH     ELEMENT    
  <chr>   <chr>  <chr> <chr>          <int> <chr>         <chr>     <chr>      
1 CDISC01 TA     PBO   Placebo            1 No criteria   SCREENING Screen     
2 CDISC01 TA     PBO   Placebo            2 Randomized    TREATMENT Placebo    
3 CDISC01 TA     PBO   Placebo            3 End treatment FOLLOW-UP Follow-up  
4 CDISC01 TA     ACT10 Active 10mg        1 No criteria   SCREENING Screen     
5 CDISC01 TA     ACT10 Active 10mg        2 Randomized    TREATMENT Active 10mg
6 CDISC01 TA     ACT10 Active 10mg        3 End treatment FOLLOW-UP Follow-up  
7 CDISC01 TA     ACT20 Active 20mg        1 No criteria   SCREENING Screen     
8 CDISC01 TA     ACT20 Active 20mg        2 Randomized    TREATMENT Active 20mg
9 CDISC01 TA     ACT20 Active 20mg        3 End treatment FOLLOW-UP Follow-up

9.3 TV - Trial Visits

# Create a Trial Visits (TV) domain example
tv <- tibble::tribble(
  ~STUDYID,  ~DOMAIN, ~VISITNUM, ~VISIT,       ~TVSTRL,     ~TVENRL,
  "CDISC01", "TV",    -1,        "SCREENING",  "Day -21",   "Day -1",
  "CDISC01", "TV",     1,        "BASELINE",   "Day 1",     "Day 1",
  "CDISC01", "TV",     2,        "WEEK 2",     "Day 8",     "Day 18",
  "CDISC01", "TV",     3,        "WEEK 4",     "Day 22",    "Day 32",
  "CDISC01", "TV",     4,        "WEEK 8",     "Day 50",    "Day 64",
  "CDISC01", "TV",     5,        "WEEK 12",    "Day 78",    "Day 92",
  "CDISC01", "TV",     6,        "END OF STUDY","Day 92",   "Day 106"
)

cat("Trial Visits (TV) - Study Schedule:\n")

Trial Visits (TV) - Study Schedule:

print(tv)

# A tibble: 7 × 6
  STUDYID DOMAIN VISITNUM VISIT        TVSTRL  TVENRL 
  <chr>   <chr>     <dbl> <chr>        <chr>   <chr>  
1 CDISC01 TV           -1 SCREENING    Day -21 Day -1 
2 CDISC01 TV            1 BASELINE     Day 1   Day 1  
3 CDISC01 TV            2 WEEK 2       Day 8   Day 18 
4 CDISC01 TV            3 WEEK 4       Day 22  Day 32 
5 CDISC01 TV            4 WEEK 8       Day 50  Day 64 
6 CDISC01 TV            5 WEEK 12      Day 78  Day 92 
7 CDISC01 TV            6 END OF STUDY Day 92  Day 106

Visit Windows

Notice TVSTRL and TVENRL - these define the visit window. In practice:

A “WEEK 2” visit might be acceptable if it occurs between Day 8 and Day 18
Programming logic must map actual visit dates to planned visits based on these windows
Visits outside windows may be flagged as unscheduled visits

9.4 TS - Trial Summary

# Create a Trial Summary (TS) domain example
ts <- tibble::tribble(
  ~STUDYID,  ~DOMAIN, ~TSSEQ, ~TSPARMCD,  ~TSPARM,                    ~TSVAL,
  "CDISC01", "TS",    1L,     "ADDON",     "Added on to Existing Treatments", "N",
  "CDISC01", "TS",    2L,     "SSTDTC",    "Study Start Date",         "2024-01-01",
  "CDISC01", "TS",    3L,     "SENDTC",    "Study End Date",           "2024-12-31",
  "CDISC01", "TS",    4L,     "INDIC",     "Trial Indication",        "Type 2 Diabetes Mellitus",
  "CDISC01", "TS",    5L,     "TRT",       "Investigational Therapy", "Study Drug XYZ",
  "CDISC01", "TS",    6L,     "TPHASE",    "Trial Phase",             "PHASE III",
  "CDISC01", "TS",    7L,     "STYPE",     "Study Type",              "INTERVENTIONAL",
  "CDISC01", "TS",    8L,     "RANDOM",    "Trial is Randomized",     "Y",
  "CDISC01", "TS",    9L,     "STOTEFFC",  "Study to Evaluate",       "EFFICACY AND SAFETY",
  "CDISC01", "TS",    10L,    "PCLAS",     "Pharmacological Class",   "SGLT2 INHIBITOR"
)

cat("Trial Summary (TS) - Key Study Parameters:\n")

Trial Summary (TS) - Key Study Parameters:

print(ts)

# A tibble: 10 × 6
   STUDYID DOMAIN TSSEQ TSPARMCD TSPARM                          TSVAL          
   <chr>   <chr>  <int> <chr>    <chr>                           <chr>          
 1 CDISC01 TS         1 ADDON    Added on to Existing Treatments N              
 2 CDISC01 TS         2 SSTDTC   Study Start Date                2024-01-01     
 3 CDISC01 TS         3 SENDTC   Study End Date                  2024-12-31     
 4 CDISC01 TS         4 INDIC    Trial Indication                Type 2 Diabete…
 5 CDISC01 TS         5 TRT      Investigational Therapy         Study Drug XYZ 
 6 CDISC01 TS         6 TPHASE   Trial Phase                     PHASE III      
 7 CDISC01 TS         7 STYPE    Study Type                      INTERVENTIONAL 
 8 CDISC01 TS         8 RANDOM   Trial is Randomized             Y              
 9 CDISC01 TS         9 STOTEFFC Study to Evaluate               EFFICACY AND S…
10 CDISC01 TS        10 PCLAS    Pharmacological Class           SGLT2 INHIBITOR

TS Parameters

The TS domain uses a parameter-value structure where each row is one study-level parameter. Common parameters include:

TPHASE - Phase I, II, III, IV
INDIC - What disease is being studied
RANDOM - Is this a randomized trial
STYPE - Interventional vs. Observational
REGID - Regulatory registration number (e.g., IND number)

The full list of valid TS parameters is in the CDISC CT Package.

10 How DS Feeds Into ADSL

The DS domain is a primary source for the ADaM ADSL (Subject-Level Analysis Dataset). Here’s the connection:

# Preview: Deriving ADSL disposition variables from DS
adsl_disp <- raw_disp %>%
  mutate(
    # Core ADSL variables derived from DS logic
    RANDFL = if_else(SCREEN_RESULT == "PASSED", "Y", "N"),
    RANDDT = RAND_DATE,
    
    # Treatment completion
    EOSSTT = case_when(
      SCREEN_RESULT == "FAILED" ~ "SCREEN FAILURE",
      TRT_RESULT == "COMPLETED" ~ "COMPLETED",
      TRUE ~ "DISCONTINUED"
    ),
    
    # Reason for discontinuation
    DCSREAS = case_when(
      EOSSTT == "COMPLETED" ~ NA_character_,
      EOSSTT == "SCREEN FAILURE" ~ SCREEN_REASON,
      TRUE ~ TRT_RESULT
    ),
    
    # Safety population flag (received at least one dose)
    SAFFL = if_else(!is.na(FIRST_DOSE), "Y", "N"),
    
    # ITT population flag (randomized)
    ITTFL = RANDFL,
    
    # Treatment duration (days)
    TRTDURD = if_else(
      !is.na(FIRST_DOSE) & !is.na(TRT_END_DATE),
      as.numeric(TRT_END_DATE - FIRST_DOSE) + 1,
      NA_real_
    )
  )

cat("ADSL Disposition Variables Preview:\n")

ADSL Disposition Variables Preview:

adsl_disp %>%
  select(USUBJID, RANDFL, EOSSTT, DCSREAS, SAFFL, ITTFL, TRTDURD) %>%
  head(10)

# A tibble: 10 × 7
   USUBJID         RANDFL EOSSTT         DCSREAS             SAFFL ITTFL TRTDURD
   <chr>           <chr>  <chr>          <chr>               <chr> <chr>   <dbl>
 1 CDISC01-001-001 Y      DISCONTINUED   WITHDRAWAL BY SUBJ… Y     Y          56
 2 CDISC01-001-002 Y      COMPLETED      <NA>                Y     Y          85
 3 CDISC01-001-003 Y      COMPLETED      <NA>                Y     Y          85
 4 CDISC01-001-004 Y      COMPLETED      <NA>                Y     Y          85
 5 CDISC01-001-005 Y      COMPLETED      <NA>                Y     Y          85
 6 CDISC01-001-006 Y      COMPLETED      <NA>                Y     Y          85
 7 CDISC01-001-007 Y      COMPLETED      <NA>                Y     Y          85
 8 CDISC01-001-008 N      SCREEN FAILURE EXCLUSION CRITERIA… N     N          NA
 9 CDISC01-001-009 Y      DISCONTINUED   ADVERSE EVENT       Y     Y          42
10 CDISC01-001-010 Y      DISCONTINUED   LOST TO FOLLOW-UP   Y     Y          61

# Population flags summary
cat("\n=== Analysis Population Summary ===\n")


=== Analysis Population Summary ===

cat("Screened:", nrow(adsl_disp), "\n")

Screened: 30

cat("Randomized (ITT):", sum(adsl_disp$ITTFL == "Y"), "\n")

Randomized (ITT): 25

cat("Safety Population:", sum(adsl_disp$SAFFL == "Y"), "\n\n")

Safety Population: 25

cat("End of Study Status:\n")

End of Study Status:

adsl_disp %>%
  count(EOSSTT, name = "N") %>%
  mutate(Percent = round(100 * N / sum(N), 1)) %>%
  arrange(desc(N))

# A tibble: 3 × 3
  EOSSTT             N Percent
  <chr>          <int>   <dbl>
1 COMPLETED         19    63.3
2 DISCONTINUED       6    20  
3 SCREEN FAILURE     5    16.7

Key ADSL Variables from DS

ADSL Variable	Source	Description
RANDFL	DS (RANDOMIZED record)	Randomized flag (Y/N)
RANDDT	DS (DSSTDTC where DSDECOD = RANDOMIZED)	Randomization date
EOSSTT	DS (treatment disposition)	End of study status
DCSREAS	DS (DSTERM for discontinuation)	Reason for discontinuation
SAFFL	EX (received at least one dose)	Safety population flag
ITTFL	DS (randomized)	ITT population flag
TRTDURD	Derived from EX dates	Treatment duration

Understanding these mappings is essential for ADaM work in Week 3.

11 Deliverable Summary

Today you completed the following:

Task	Status
Understood DS domain structure and purpose	✓ Done
Built a DS domain with consent, randomization, and treatment disposition	✓ Done
Explored EPOCH and study period concepts	✓ Done
Surveyed Trial Design domains (TA, TV, TS)	✓ Done
Created subject disposition flow (CONSORT-style)	✓ Done
Previewed how DS feeds into ADSL	✓ Done

12 Key Takeaways

DS is an Events class domain - Each record is a milestone or disposition event
Multiple records per subject - Consent, randomization, treatment disposition, study disposition
DSCAT/DSSCAT distinguish event types - Milestones vs. disposition decisions
EPOCH defines study periods - Screening, Treatment, Follow-up
Trial Design domains are required - TA, TE, TV, TI, TS describe the study blueprint
DS drives ADSL - Randomization, completion, discontinuation all come from DS

13 Resources

CDISC SDTM Implementation Guide - DS Domain - Official DS specification
Trial Design Domains Guide - TA, TE, TV, TI, TS specifications
CONSORT Statement - Reporting standards for clinical trials
Admiral ADSL Vignette - Building ADSL from DS
Pharmaverse.org - R packages for clinical data

14 What’s Next?

In Day 12, we will focus on Data Cuts with datacutr:

What clinical data cuts are and why they matter for interim analyses
Using the datacutr package to apply cutoff dates across domains
Handling ongoing AEs, partial dates, and records at the cutoff boundary
Patient-level vs. record-level cut logic