Day 18: ADAE - Adverse Events Analysis Dataset

Week 3, Day 18: OCCDS structure with treatment emergent flags

1 Welcome to ADAE

ADAE is the backbone of safety reporting. Every safety table flows from this dataset.

Today’s focus: Build ADAE with treatment emergent flags and occurrence indicators.

2 Setup

library(admiral)
library(pharmaversesdtm)
library(pharmaverseadam)
library(dplyr)
library(lubridate)
library(stringr)
library(xportr)

# Load SDTM domains
ae <- pharmaversesdtm::ae
ex <- pharmaversesdtm::ex

# Load pre-built ADSL
adsl <- pharmaverseadam::adsl

cat("Loaded AE:", nrow(ae), "records\n")

Loaded AE: 1191 records

cat("Loaded ADSL:", nrow(adsl), "subjects\n")

Loaded ADSL: 306 subjects

Why pharmaverseadam::adsl?

Pre-built reference ADSL
Has all required variables
Saves rebuild time

3 Step 1: Merge ADSL to AE

# Merge treatment dates to AE
adae <- ae %>%
  admiral::derive_vars_merged(
    dataset_add = adsl,
    new_vars = exprs(TRTSDT, TRTEDT, TRT01A),
    by_vars = exprs(STUDYID, USUBJID)
  )

cat("\nADAE initialized:", nrow(adae), "records\n")


ADAE initialized: 1191 records

# Check
adae %>%
  select(USUBJID, AETERM, TRTSDT, TRTEDT) %>%
  head(5)

# A tibble: 5 × 4
  USUBJID     AETERM                               TRTSDT     TRTEDT    
  <chr>       <chr>                                <date>     <date>    
1 01-701-1015 APPLICATION SITE ERYTHEMA            2014-01-02 2014-07-02
2 01-701-1015 APPLICATION SITE PRURITUS            2014-01-02 2014-07-02
3 01-701-1015 DIARRHOEA                            2014-01-02 2014-07-02
4 01-701-1023 ATRIOVENTRICULAR BLOCK SECOND DEGREE 2012-08-05 2012-09-01
5 01-701-1023 ERYTHEMA                             2012-08-05 2012-09-01

4 Step 2: Derive AE Start Date

# Derive AE start date
adae <- adae %>%
  admiral::derive_vars_dt(
    new_vars_prefix = "AST",
    dtc = AESTDTC,
    highest_imputation = "M",
    flag_imputation = "auto"
  )

cat("\nAE start date derived\n")


AE start date derived

# Check
adae %>%
  select(USUBJID, AESEQ, AESTDTC, ASTDT, ASTDTF) %>%
  head(5)

# A tibble: 5 × 5
  USUBJID     AESEQ AESTDTC    ASTDT      ASTDTF
  <chr>       <dbl> <chr>      <date>     <chr> 
1 01-701-1015     1 2014-01-03 2014-01-03 <NA>  
2 01-701-1015     2 2014-01-03 2014-01-03 <NA>  
3 01-701-1015     3 2014-01-09 2014-01-09 <NA>  
4 01-701-1023     3 2012-08-26 2012-08-26 <NA>  
5 01-701-1023     1 2012-08-07 2012-08-07 <NA>

Imputation:

Partial dates like “2024-03” → “2024-03-01”
ASTDTF flag shows if imputed

5 Step 3: Derive AE End Date

# Derive AE end date
adae <- adae %>%
  admiral::derive_vars_dt(
    new_vars_prefix = "AEN",
    dtc = AEENDTC,
    highest_imputation = "M",
    date_imputation = "last",
    flag_imputation = "auto"
  )

cat("\nAE end date derived\n")


AE end date derived

# Check
adae %>%
  select(USUBJID, AESEQ, AEENDTC, AENDT, AENDTF) %>%
  head(5)

# A tibble: 5 × 5
  USUBJID     AESEQ AEENDTC    AENDT      AENDTF
  <chr>       <dbl> <chr>      <date>     <chr> 
1 01-701-1015     1 <NA>       NA         <NA>  
2 01-701-1015     2 <NA>       NA         <NA>  
3 01-701-1015     3 2014-01-11 2014-01-11 <NA>  
4 01-701-1023     3 <NA>       NA         <NA>  
5 01-701-1023     1 2012-08-30 2012-08-30 <NA>

Difference from start:

date_imputation = "last" → “2024-03” becomes “2024-03-31”

6 Step 4: Derive Study Days

# Calculate relative days
adae <- adae %>%
  admiral::derive_vars_dy(
    reference_date = TRTSDT,
    source_vars = exprs(ASTDT, AENDT)
  )

cat("\nStudy days derived\n")


Study days derived

# Check
adae %>%
  select(USUBJID, TRTSDT, ASTDT, ASTDY, AENDT, AENDY) %>%
  filter(!is.na(ASTDY)) %>%
  head(5)

# A tibble: 5 × 6
  USUBJID     TRTSDT     ASTDT      ASTDY AENDT      AENDY
  <chr>       <date>     <date>     <dbl> <date>     <dbl>
1 01-701-1015 2014-01-02 2014-01-03     2 NA            NA
2 01-701-1015 2014-01-02 2014-01-03     2 NA            NA
3 01-701-1015 2014-01-02 2014-01-09     8 2014-01-11    10
4 01-701-1023 2012-08-05 2012-08-26    22 NA            NA
5 01-701-1023 2012-08-05 2012-08-07     3 2012-08-30    26

Study day convention:

Day 1 = first treatment day
No Day 0

7 Step 5: Derive Duration

# Calculate duration
adae <- adae %>%
  mutate(
    ADURN = case_when(
      !is.na(ASTDT) & !is.na(AENDT) ~ as.numeric(AENDT - ASTDT) + 1,
      TRUE ~ NA_real_
    )
  )

cat("\nDuration calculated\n")


Duration calculated

# Summary
adae %>%
  filter(!is.na(ADURN)) %>%
  summarise(
    N = n(),
    Mean = round(mean(ADURN), 1),
    Median = median(ADURN),
    Min = min(ADURN),
    Max = max(ADURN)
  )

# A tibble: 1 × 5
      N  Mean Median   Min   Max
  <int> <dbl>  <dbl> <dbl> <dbl>
1   718  24.2     11     1   444

Formula: AENDT - ASTDT + 1

8 Step 6: Derive Treatment Emergent Flag

# Flag AEs that occurred during treatment
adae <- adae %>%
  mutate(
    TRTEMFL = case_when(
      !is.na(ASTDT) & !is.na(TRTSDT) & ASTDT >= TRTSDT ~ "Y",
      TRUE ~ NA_character_
    )
  )

cat("\nTreatment emergent flag derived\n")


Treatment emergent flag derived

# Distribution
adae %>%
  count(TRTEMFL)

# A tibble: 2 × 2
  TRTEMFL     n
  <chr>   <int>
1 Y        1126
2 <NA>       65

Logic:

AE started on or after treatment start
Used in all safety tables

9 Step 7: Derive Severity Numeric

# Create numeric severity
adae <- adae %>%
  mutate(
    ASEV = AESEV,
    ASEVN = case_when(
      AESEV == "MILD" ~ 1,
      AESEV == "MODERATE" ~ 2,
      AESEV == "SEVERE" ~ 3,
      TRUE ~ NA_real_
    )
  )

cat("\nSeverity numeric created\n")


Severity numeric created

# Distribution
adae %>%
  count(ASEV, ASEVN)

# A tibble: 3 × 3
  ASEV     ASEVN     n
  <chr>    <dbl> <int>
1 MILD         1   770
2 MODERATE     2   378
3 SEVERE       3    43

10 Step 8: Derive First Occurrence Flag

# Flag first AE per subject
adae <- adae %>%
  arrange(USUBJID, ASTDT, AESEQ) %>%
  group_by(USUBJID) %>%
  mutate(
    AOCCFL = if_else(row_number() == 1, "Y", NA_character_)
  ) %>%
  ungroup()

cat("\nFirst occurrence flag derived\n")


First occurrence flag derived

# Check
adae %>%
  filter(AOCCFL == "Y") %>%
  select(USUBJID, AETERM, ASTDT, AOCCFL) %>%
  head(5)

# A tibble: 5 × 4
  USUBJID     AETERM                    ASTDT      AOCCFL
  <chr>       <chr>                     <date>     <chr> 
1 01-701-1015 APPLICATION SITE ERYTHEMA 2014-01-03 Y     
2 01-701-1023 ERYTHEMA                  2012-08-07 Y     
3 01-701-1028 APPLICATION SITE ERYTHEMA 2013-07-21 Y     
4 01-701-1034 APPLICATION SITE PRURITUS 2014-08-27 Y     
5 01-701-1047 HIATUS HERNIA             2013-02-12 Y

Logic:

First AE record per subject
Used for “subjects with at least one AE”

11 Step 9: Derive Serious AE Flag

# Copy serious flag
adae <- adae %>%
  mutate(
    ASER = AESER,
    ASERN = if_else(AESER == "Y", 1, 0)
  )

cat("\nSerious flag created\n")


Serious flag created

# Distribution
adae %>%
  count(ASER, ASERN)

# A tibble: 2 × 3
  ASER  ASERN     n
  <chr> <dbl> <int>
1 N         0  1188
2 Y         1     3

12 Step 10: Derive Analysis Flag

# Primary analysis flag
adae <- adae %>%
  mutate(
    ANL01FL = if_else(TRTEMFL == "Y", "Y", NA_character_)
  )

cat("\nAnalysis flag created\n")


Analysis flag created

# Distribution
adae %>%
  count(ANL01FL, TRTEMFL)

# A tibble: 2 × 3
  ANL01FL TRTEMFL     n
  <chr>   <chr>   <int>
1 Y       Y        1126
2 <NA>    <NA>       65

ANL01FL:

Primary analysis records
Typically = TRTEMFL

13 Validation

cat("\n=== ADAE Validation ===\n\n")


=== ADAE Validation ===

# Check 1: Start <= End
check1 <- adae %>%
  filter(!is.na(ASTDT) & !is.na(AENDT) & ASTDT > AENDT)
cat("Check 1 - ASTDT > AENDT:", nrow(check1), "\n")

Check 1 - ASTDT > AENDT: 0

# Check 2: Duration
check2 <- adae %>%
  filter(!is.na(ADURN)) %>%
  mutate(
    Expected = as.numeric(AENDT - ASTDT) + 1,
    Match = ADURN == Expected
  ) %>%
  filter(!Match)
cat("Check 2 - ADURN mismatch:", nrow(check2), "\n")

Check 2 - ADURN mismatch: 0

# Check 3: One AOCCFL per subject
check3 <- adae %>%
  filter(AOCCFL == "Y") %>%
  count(USUBJID) %>%
  filter(n > 1)
cat("Check 3 - Multiple AOCCFL:", nrow(check3), "\n")

Check 3 - Multiple AOCCFL: 0

cat("\n✓ Validation complete\n")


✓ Validation complete

14 Summary

cat("\n=== ADAE Summary ===\n\n")


=== ADAE Summary ===

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

Total records: 1191

cat("Variables:", ncol(adae), "\n\n")

Variables: 52

# By treatment
adae %>%
  filter(TRTEMFL == "Y") %>%
  group_by(TRT01A) %>%
  summarise(
    N_AEs = n(),
    N_Subjects = n_distinct(USUBJID),
    N_Serious = sum(ASER == "Y", na.rm = TRUE),
    .groups = "drop"
  )

# A tibble: 3 × 4
  TRT01A               N_AEs N_Subjects N_Serious
  <chr>                <int>      <int>     <int>
1 Placebo                281         65         0
2 Xanomeline High Dose   418         69         1
3 Xanomeline Low Dose    427         84         2

# By severity
adae %>%
  filter(TRTEMFL == "Y") %>%
  count(ASEV)

# A tibble: 3 × 2
  ASEV         n
  <chr>    <int>
1 MILD       731
2 MODERATE   354
3 SEVERE      41

15 Export

# Add labels
attr(adae$ASTDT, "label") <- "Analysis Start Date"
attr(adae$ASTDY, "label") <- "Analysis Start Relative Day"
attr(adae$AENDT, "label") <- "Analysis End Date"
attr(adae$AENDY, "label") <- "Analysis End Relative Day"
attr(adae$ADURN, "label") <- "Analysis Duration (Days)"
attr(adae$TRTEMFL, "label") <- "Treatment Emergent Flag"
attr(adae$ASEV, "label") <- "Analysis Severity"
attr(adae$ASEVN, "label") <- "Analysis Severity (N)"
attr(adae$AOCCFL, "label") <- "First Occurrence Flag"
attr(adae$ANL01FL, "label") <- "Analysis Record Flag 01"

# Export - xportr is already loaded via library()
xportr_write(adae, path = "adae.xpt", domain = "ADAE")

cat("\n✓ ADAE exported to: adae.xpt\n")


✓ ADAE exported to: adae.xpt

16 Key Takeaways

Use pharmaverseadam::adsl for reference ADSL
Impute start date “first”, end date “last”
TRTEMFL = AE on or after treatment start
AOCCFL = first AE per subject (arrange + group_by + row_number)
Always validate dates and flags

17 Next Steps

Day 19: ADLB - Lab Analysis Dataset
Day 20: ADVS - Vitals Analysis Dataset
Day 21: ADTTE - Time-to-Event

18 Resources

OCCDS Vignette: https://pharmaverse.github.io/admiral/articles/occds.html
ADAE Example: https://pharmaverse.github.io/examples/adam/adae.html
pharmaverseadam: https://pharmaverse.github.io/pharmaverseadam/
xportr package: https://atorus-research.github.io/xportr/

End of Day 18