Day 9: VS (Vital Signs) & Repeated Measures

Visit-Level Data and Positional Readings

1 Learning Objectives

By the end of Day 9, you will be able to:

Understand the VS (Vital Signs) domain structure and its role as a Findings class domain
Handle repeated measures - multiple readings per subject per visit per timepoint
Work with positional variables like VSPOS (SITTING/STANDING) and VSLOC (ARM)
Derive summary statistics (mean, min, max) across multiple readings
Understand how VS data maps to the ADaM BDS (Basic Data Structure) for ADVS

2 Introduction to the VS Domain

2.1 What is the VS Domain?

The VS (Vital Signs) domain captures measurements of a subject’s vital functions during a clinical trial. Like the LB domain we covered on Day 8, VS belongs to the Findings class of SDTM domains.

Common vital signs measured include:

Blood Pressure (Systolic and Diastolic)
Heart Rate (Pulse)
Temperature
Respiratory Rate
Weight and Height
Oxygen Saturation

2.2 Why is VS Special?

The VS domain introduces an important concept: repeated measures at the same timepoint. Unlike demographics (one row per subject) or adverse events (one row per event), vital signs are often measured multiple times:

Multiple blood pressure readings at the same visit (e.g., 3 readings taken 2 minutes apart)
Position-dependent measurements (sitting vs. standing blood pressure)
Location-specific readings (left arm vs. right arm)

Repeated Measures Challenge

When analyzing vital signs, we need to decide:

Do we use the first reading?
Do we calculate the mean of all readings?
Do we report the minimum or maximum?

This is a key consideration when building ADaM datasets (ADVS) from SDTM VS data.

2.3 The VS Domain Structure

┌──────────────────────────────────────────────────────────────────────────────┐
│                        VS DOMAIN - KEY VARIABLES                             │
├──────────────────────────────────────────────────────────────────────────────┤
│  IDENTIFIER VARIABLES                                                        │
│    STUDYID    = Study identifier                                             │
│    USUBJID    = Unique subject identifier (links to DM)                      │
│    VSSEQ      = Sequence number within subject                               │
├──────────────────────────────────────────────────────────────────────────────┤
│  TOPIC VARIABLE                                                              │
│    VSTESTCD   = Vital sign test short name (e.g., "SYSBP", "DIABP", "HR")    │
│    VSTEST     = Vital sign test name (e.g., "Systolic Blood Pressure")       │
├──────────────────────────────────────────────────────────────────────────────┤
│  RESULT VARIABLES                                                            │
│    VSORRES    = Result as originally collected (character)                   │
│    VSORRESU   = Unit of original result                                      │
│    VSSTRESC   = Standardized result (character)                              │
│    VSSTRESN   = Standardized result (numeric)                                │
│    VSSTRESU   = Standardized unit                                            │
├──────────────────────────────────────────────────────────────────────────────┤
│  POSITIONAL VARIABLES (Unique to VS)                                         │
│    VSPOS      = Position of subject (SITTING, STANDING, SUPINE)              │
│    VSLOC      = Location on body (LEFT ARM, RIGHT ARM)                       │
├──────────────────────────────────────────────────────────────────────────────┤
│  TIMING VARIABLES                                                            │
│    VISITNUM   = Visit number                                                 │
│    VISIT      = Visit name (e.g., "SCREENING", "WEEK 2")                     │
│    VSDY       = Study day of measurement                                     │
│    VSTPT      = Planned time point name (PRE-DOSE, 1 HOUR POST-DOSE)         │
│    VSTPTNUM   = Planned time point number                                    │
│    VSDTC      = Date/time of measurement (ISO 8601)                          │
└──────────────────────────────────────────────────────────────────────────────┘

3 Package Installation & Loading

3.1 Required Packages

Package	Purpose
`dplyr`	Data manipulation (filter, mutate, summarise)
`tidyr`	Data reshaping (pivot operations)
`pharmaversesdtm`	Example SDTM datasets including VS

3.2 Install Packages (if needed)

# Install packages if not already installed
if (!requireNamespace("dplyr", quietly = TRUE)) suppressMessages(install.packages("dplyr"))
if (!requireNamespace("tidyr", quietly = TRUE)) suppressMessages(install.packages("tidyr"))
if (!requireNamespace("pharmaversesdtm", quietly = TRUE)) suppressMessages(install.packages("pharmaversesdtm"))

3.3 Load Packages

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

4 Exploring pharmaversesdtm VS Data

Let’s start by loading and exploring the VS domain from pharmaversesdtm.

4.1 Load the VS Domain

# Load VS domain from pharmaversesdtm
data("vs", package = "pharmaversesdtm")

# Quick overview
cat("VS domain dimensions:", nrow(vs), "rows x", ncol(vs), "columns\n")

VS domain dimensions: 29643 rows x 24 columns

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

Number of unique subjects: 254

cat("Number of unique tests:", n_distinct(vs$VSTESTCD), "\n")

Number of unique tests: 6

4.2 Explore VS Structure

# View the structure of the VS domain
dplyr::glimpse(vs)

Rows: 29,643
Columns: 24
$ STUDYID  <chr> "CDISCPILOT01", "CDISCPILOT01", "CDISCPILOT01", "CDISCPILOT01…
$ DOMAIN   <chr> "VS", "VS", "VS", "VS", "VS", "VS", "VS", "VS", "VS", "VS", "…
$ USUBJID  <chr> "01-701-1015", "01-701-1015", "01-701-1015", "01-701-1015", "…
$ VSSEQ    <dbl> 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18…
$ VSTESTCD <chr> "DIABP", "DIABP", "DIABP", "DIABP", "DIABP", "DIABP", "DIABP"…
$ VSTEST   <chr> "Diastolic Blood Pressure", "Diastolic Blood Pressure", "Dias…
$ VSPOS    <chr> "SUPINE", "STANDING", "STANDING", "SUPINE", "STANDING", "STAN…
$ VSORRES  <chr> "64", "83", "57", "68", "59", "71", "56", "51", "61", "67", "…
$ VSORRESU <chr> "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg…
$ VSSTRESC <chr> "64", "83", "57", "68", "59", "71", "56", "51", "61", "67", "…
$ VSSTRESN <dbl> 64, 83, 57, 68, 59, 71, 56, 51, 61, 67, 61, 65, 56, 50, 54, 6…
$ VSSTRESU <chr> "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg", "mmHg…
$ VSSTAT   <chr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N…
$ VSLOC    <chr> NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, NA, N…
$ VSBLFL   <chr> NA, NA, NA, NA, NA, NA, "Y", "Y", "Y", NA, NA, NA, NA, NA, NA…
$ VISITNUM <dbl> 1.0, 1.0, 1.0, 2.0, 2.0, 2.0, 3.0, 3.0, 3.0, 3.5, 3.5, 3.5, 4…
$ VISIT    <chr> "SCREENING 1", "SCREENING 1", "SCREENING 1", "SCREENING 2", "…
$ VISITDY  <dbl> -7, -7, -7, -1, -1, -1, 1, 1, 1, 13, 13, 13, 14, 14, 14, 28, …
$ VSDTC    <chr> "2013-12-26", "2013-12-26", "2013-12-26", "2013-12-31", "2013…
$ VSDY     <dbl> -7, -7, -7, -2, -2, -2, 1, 1, 1, 13, 13, 13, 15, 15, 15, 29, …
$ VSTPT    <chr> "AFTER LYING DOWN FOR 5 MINUTES", "AFTER STANDING FOR 1 MINUT…
$ VSTPTNUM <dbl> 815, 816, 817, 815, 816, 817, 815, 816, 817, 815, 816, 817, 8…
$ VSELTM   <chr> "PT5M", "PT1M", "PT3M", "PT5M", "PT1M", "PT3M", "PT5M", "PT1M…
$ VSTPTREF <chr> "PATIENT SUPINE", "PATIENT STANDING", "PATIENT STANDING", "PA…

4.3 Key VS Variables

Let’s examine the vital sign tests available in our data:

# View unique vital sign tests
vs %>%
  distinct(VSTESTCD, VSTEST, VSSTRESU) %>%
  arrange(VSTESTCD)

# A tibble: 9 × 3
  VSTESTCD VSTEST                   VSSTRESU 
  <chr>    <chr>                    <chr>    
1 DIABP    Diastolic Blood Pressure mmHg     
2 DIABP    Diastolic Blood Pressure <NA>     
3 HEIGHT   Height                   cm       
4 PULSE    Pulse Rate               BEATS/MIN
5 PULSE    Pulse Rate               <NA>     
6 SYSBP    Systolic Blood Pressure  mmHg     
7 SYSBP    Systolic Blood Pressure  <NA>     
8 TEMP     Temperature              C        
9 WEIGHT   Weight                   kg

Understanding VS Test Codes

Common CDISC vital sign test codes:

SYSBP = Systolic Blood Pressure
DIABP = Diastolic Blood Pressure
PULSE = Pulse Rate / Heart Rate
TEMP = Temperature
RESP = Respiratory Rate
WEIGHT = Weight
HEIGHT = Height
BMI = Body Mass Index (often derived)

5 Understanding Visit Structure

5.1 Visits and Timepoints

In clinical trials, vital signs are collected at specific visits (e.g., Screening, Week 2, Week 8) and sometimes at specific timepoints within a visit (e.g., pre-dose, 1 hour post-dose).

# View visit structure
vs %>%
  distinct(VISITNUM, VISIT) %>%
  arrange(VISITNUM)

# A tibble: 16 × 2
   VISITNUM VISIT              
      <dbl> <chr>              
 1      1   SCREENING 1        
 2      2   SCREENING 2        
 3      3   BASELINE           
 4      3.1 UNSCHEDULED 3.1    
 5      3.5 AMBUL ECG PLACEMENT
 6      4   WEEK 2             
 7      5   WEEK 4             
 8      6   AMBUL ECG REMOVAL  
 9      7   WEEK 6             
10      8   WEEK 8             
11      9   WEEK 12            
12     10   WEEK 16            
13     11   WEEK 20            
14     12   WEEK 24            
15     13   WEEK 26            
16    201   RETRIEVAL

5.2 SDTM Timing Variables Explained

Variable	Description	Example
`VISITNUM`	Numeric visit identifier	1, 2, 3
`VISIT`	Visit name	“BASELINE”, “WEEK 2”
`VSDY`	Study day of measurement	1, 15, 29
`VSTPT`	Planned time point name	“PRE-DOSE”, “1HR POST-DOSE”
`VSTPTNUM`	Planned time point number	1, 2, 3
`VSDTC`	Date/time (ISO 8601)	“2023-03-15T09:30:00”

# Example of timing variables
vs %>%
  filter(USUBJID == first(vs$USUBJID), VSTESTCD == "SYSBP") %>%
  select(USUBJID, VSTESTCD, VISITNUM, VISIT, VSDY, VSTPT, VSTPTNUM) %>%
  arrange(VISITNUM, VSTPTNUM) %>%
  head(10)

# A tibble: 10 × 7
   USUBJID     VSTESTCD VISITNUM VISIT                VSDY VSTPT        VSTPTNUM
   <chr>       <chr>       <dbl> <chr>               <dbl> <chr>           <dbl>
 1 01-701-1015 SYSBP         1   SCREENING 1            -7 AFTER LYING…      815
 2 01-701-1015 SYSBP         1   SCREENING 1            -7 AFTER STAND…      816
 3 01-701-1015 SYSBP         1   SCREENING 1            -7 AFTER STAND…      817
 4 01-701-1015 SYSBP         2   SCREENING 2            -2 AFTER LYING…      815
 5 01-701-1015 SYSBP         2   SCREENING 2            -2 AFTER STAND…      816
 6 01-701-1015 SYSBP         2   SCREENING 2            -2 AFTER STAND…      817
 7 01-701-1015 SYSBP         3   BASELINE                1 AFTER LYING…      815
 8 01-701-1015 SYSBP         3   BASELINE                1 AFTER STAND…      816
 9 01-701-1015 SYSBP         3   BASELINE                1 AFTER STAND…      817
10 01-701-1015 SYSBP         3.5 AMBUL ECG PLACEMENT    13 AFTER LYING…      815

6 Positional Variables: VSPOS and VSLOC

6.1 Why Position and Location Matter

Blood pressure measurements can vary significantly based on:

Position: Sitting, standing, or lying down (supine)
Location: Which arm the reading was taken from

Clinical protocols often specify:

“Blood pressure should be measured after the patient has been seated for at least 5 minutes, using the right arm.”

6.2 VSPOS - Position of Subject

The VSPOS variable captures the subject’s position during measurement:

VSPOS Value	Description
`SITTING`	Subject is seated
`STANDING`	Subject is standing
`SUPINE`	Subject is lying flat
`RECLINING`	Subject is partially reclined

# Check position values in our data
vs %>%
  filter(VSTESTCD %in% c("SYSBP", "DIABP")) %>%
  count(VSPOS, name = "Count") %>%
  arrange(desc(Count))

# A tibble: 2 × 2
  VSPOS    Count
  <chr>    <int>
1 STANDING 10942
2 SUPINE    5473

6.3 VSLOC - Location on Body

The VSLOC variable specifies where on the body the measurement was taken:

VSLOC Value	Description
`LEFT ARM`	Blood pressure cuff on left arm
`RIGHT ARM`	Blood pressure cuff on right arm
`ARM`	Arm (unspecified)
`ORAL`	Oral temperature
`AXILLARY`	Under the armpit (temperature)

# Check location values
vs %>%
  filter(VSTESTCD %in% c("SYSBP", "DIABP", "TEMP")) %>%
  count(VSTESTCD, VSLOC, name = "Count") %>%
  arrange(VSTESTCD, desc(Count))

# A tibble: 4 × 3
  VSTESTCD VSLOC       Count
  <chr>    <chr>       <int>
1 DIABP    <NA>         8207
2 SYSBP    <NA>         8208
3 TEMP     ORAL CAVITY  1765
4 TEMP     EAR           955

7 Handling Multiple Readings

7.1 The Problem: Multiple Measurements

In many studies, vital signs are measured multiple times at the same visit to ensure accuracy. For blood pressure, it’s common to take 3 readings and use the average.

7.2 Creating Sample Data with Multiple Readings

Let’s create a dataset that demonstrates this scenario:

# Create sample data with 3 BP readings per visit
set.seed(42)

subjects <- c("CDISC01-001-001", "CDISC01-001-002", "CDISC01-001-003")
visits <- c("SCREENING", "BASELINE", "WEEK 2", "WEEK 4")
visit_nums <- c(-1, 1, 2, 3)

# Generate VS data with multiple readings per visit
vs_multiple <- expand_grid(
  USUBJID = subjects,
  VISIT_INFO = tibble(VISIT = visits, VISITNUM = visit_nums)
) %>%
  unnest(VISIT_INFO) %>%
  # Create 3 readings per vital sign per visit
  crossing(
    READING = 1:3,
    tibble::tribble(
      ~VSTESTCD, ~VSTEST,                     ~VSSTRESU, ~BASE_VALUE, ~SD_VALUE,
      "SYSBP",   "Systolic Blood Pressure",   "mmHg",    125,         10,
      "DIABP",   "Diastolic Blood Pressure",  "mmHg",    80,          8,
      "PULSE",   "Pulse Rate",                "beats/min", 72,        8
    )
  ) %>%
  # Generate values with some variation between readings
  rowwise() %>%
  mutate(
    VSORRES = as.character(round(rnorm(1, BASE_VALUE, SD_VALUE), 0)),
    VSSTRESN = as.numeric(VSORRES),
    VSSTRESC = VSORRES,
    VSORRESU = VSSTRESU,
    VSPOS = "SITTING",
    VSLOC = ifelse(VSTESTCD %in% c("SYSBP", "DIABP"), "RIGHT ARM", NA_character_)
  ) %>%
  ungroup() %>%
  # Add identifiers
  mutate(
    STUDYID = "CDISC01",
    VSTPT = paste("READING", READING),
    VSTPTNUM = READING
  ) %>%
  group_by(USUBJID) %>%
  mutate(VSSEQ = row_number()) %>%
  ungroup() %>%
  select(STUDYID, USUBJID, VSSEQ, VSTESTCD, VSTEST, 
         VSORRES, VSORRESU, VSSTRESC, VSSTRESN, VSSTRESU,
         VSPOS, VSLOC, VISITNUM, VISIT, VSTPT, VSTPTNUM)

cat("Sample VS data with multiple readings:\n")

Sample VS data with multiple readings:

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

Total records: 108

cat("Subjects:", n_distinct(vs_multiple$USUBJID), "\n")

Subjects: 3

cat("Visits:", n_distinct(vs_multiple$VISIT), "\n\n")

Visits: 4

# Show example for one subject at one visit
vs_multiple %>%
  filter(USUBJID == first(vs_multiple$USUBJID), VISIT == "BASELINE") %>%
  print()

# A tibble: 9 × 16
  STUDYID USUBJID       VSSEQ VSTESTCD VSTEST VSORRES VSORRESU VSSTRESC VSSTRESN
  <chr>   <chr>         <int> <chr>    <chr>  <chr>   <chr>    <chr>       <dbl>
1 CDISC01 CDISC01-001-…     1 DIABP    Diast… 91      mmHg     91             91
2 CDISC01 CDISC01-001-…     2 PULSE    Pulse… 67      beats/m… 67             67
3 CDISC01 CDISC01-001-…     3 SYSBP    Systo… 129     mmHg     129           129
4 CDISC01 CDISC01-001-…     4 DIABP    Diast… 85      mmHg     85             85
5 CDISC01 CDISC01-001-…     5 PULSE    Pulse… 75      beats/m… 75             75
6 CDISC01 CDISC01-001-…     6 SYSBP    Systo… 124     mmHg     124           124
7 CDISC01 CDISC01-001-…     7 DIABP    Diast… 92      mmHg     92             92
8 CDISC01 CDISC01-001-…     8 PULSE    Pulse… 71      beats/m… 71             71
9 CDISC01 CDISC01-001-…     9 SYSBP    Systo… 145     mmHg     145           145
# ℹ 7 more variables: VSSTRESU <chr>, VSPOS <chr>, VSLOC <chr>, VISITNUM <dbl>,
#   VISIT <chr>, VSTPT <chr>, VSTPTNUM <int>

Understanding the Data Structure

Notice that for each subject at each visit, we have:

3 readings × 3 tests = 9 records per subject per visit
Each reading is identified by VSTPT and VSTPTNUM
This granular data allows for various analysis approaches

8 Deriving Statistics Across Readings

Now let’s derive mean, minimum, and maximum values across the multiple readings.

8.1 Calculate Summary Statistics

# Calculate statistics across readings for each subject/visit/test
vs_summary <- vs_multiple %>%
  group_by(STUDYID, USUBJID, VSTESTCD, VSTEST, VSSTRESU, 
           VSPOS, VSLOC, VISITNUM, VISIT) %>%
  summarise(
    N_READINGS = n(),
    VSSTRESN_MEAN = round(mean(VSSTRESN, na.rm = TRUE), 1),
    VSSTRESN_MIN = min(VSSTRESN, na.rm = TRUE),
    VSSTRESN_MAX = max(VSSTRESN, na.rm = TRUE),
    VSSTRESN_SD = round(sd(VSSTRESN, na.rm = TRUE), 2),
    .groups = "drop"
  )

cat("Summary statistics by subject/visit/test:\n")

Summary statistics by subject/visit/test:

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

# A tibble: 12 × 14
   STUDYID USUBJID         VSTESTCD VSTEST   VSSTRESU VSPOS VSLOC VISITNUM VISIT
   <chr>   <chr>           <chr>    <chr>    <chr>    <chr> <chr>    <dbl> <chr>
 1 CDISC01 CDISC01-001-001 DIABP    Diastol… mmHg     SITT… RIGH…       -1 SCRE…
 2 CDISC01 CDISC01-001-001 DIABP    Diastol… mmHg     SITT… RIGH…        1 BASE…
 3 CDISC01 CDISC01-001-001 DIABP    Diastol… mmHg     SITT… RIGH…        2 WEEK…
 4 CDISC01 CDISC01-001-001 DIABP    Diastol… mmHg     SITT… RIGH…        3 WEEK…
 5 CDISC01 CDISC01-001-001 PULSE    Pulse R… beats/m… SITT… <NA>        -1 SCRE…
 6 CDISC01 CDISC01-001-001 PULSE    Pulse R… beats/m… SITT… <NA>         1 BASE…
 7 CDISC01 CDISC01-001-001 PULSE    Pulse R… beats/m… SITT… <NA>         2 WEEK…
 8 CDISC01 CDISC01-001-001 PULSE    Pulse R… beats/m… SITT… <NA>         3 WEEK…
 9 CDISC01 CDISC01-001-001 SYSBP    Systoli… mmHg     SITT… RIGH…       -1 SCRE…
10 CDISC01 CDISC01-001-001 SYSBP    Systoli… mmHg     SITT… RIGH…        1 BASE…
11 CDISC01 CDISC01-001-001 SYSBP    Systoli… mmHg     SITT… RIGH…        2 WEEK…
12 CDISC01 CDISC01-001-001 SYSBP    Systoli… mmHg     SITT… RIGH…        3 WEEK…
# ℹ 5 more variables: N_READINGS <int>, VSSTRESN_MEAN <dbl>,
#   VSSTRESN_MIN <dbl>, VSSTRESN_MAX <dbl>, VSSTRESN_SD <dbl>

8.2 Create Analysis-Ready Dataset

When creating ADaM ADVS, we typically select one representative value from the multiple readings. Common approaches:

# Create analysis-ready dataset using MEAN of readings
vs_analysis <- vs_summary %>%
  mutate(
    AVAL = VSSTRESN_MEAN,
    AVALC = as.character(AVAL),
    DTYPE = "MEAN"  # Derivation type
  ) %>%
  select(STUDYID, USUBJID, VSTESTCD, VSTEST, 
         VISITNUM, VISIT, VSPOS, VSLOC,
         AVAL, AVALC, VSSTRESU, N_READINGS, DTYPE)

cat("Analysis-ready dataset (using mean values):\n")

Analysis-ready dataset (using mean values):

head(vs_analysis, 12)

# A tibble: 12 × 13
   STUDYID USUBJID        VSTESTCD VSTEST VISITNUM VISIT VSPOS VSLOC  AVAL AVALC
   <chr>   <chr>          <chr>    <chr>     <dbl> <chr> <chr> <chr> <dbl> <chr>
 1 CDISC01 CDISC01-001-0… DIABP    Diast…       -1 SCRE… SITT… RIGH…  77.7 77.7 
 2 CDISC01 CDISC01-001-0… DIABP    Diast…        1 BASE… SITT… RIGH…  89.3 89.3 
 3 CDISC01 CDISC01-001-0… DIABP    Diast…        2 WEEK… SITT… RIGH…  73.7 73.7 
 4 CDISC01 CDISC01-001-0… DIABP    Diast…        3 WEEK… SITT… RIGH…  75   75   
 5 CDISC01 CDISC01-001-0… PULSE    Pulse…       -1 SCRE… SITT… <NA>   74   74   
 6 CDISC01 CDISC01-001-0… PULSE    Pulse…        1 BASE… SITT… <NA>   71   71   
 7 CDISC01 CDISC01-001-0… PULSE    Pulse…        2 WEEK… SITT… <NA>   74.3 74.3 
 8 CDISC01 CDISC01-001-0… PULSE    Pulse…        3 WEEK… SITT… <NA>   76.7 76.7 
 9 CDISC01 CDISC01-001-0… SYSBP    Systo…       -1 SCRE… SITT… RIGH… 123.  123.3
10 CDISC01 CDISC01-001-0… SYSBP    Systo…        1 BASE… SITT… RIGH… 133.  132.7
11 CDISC01 CDISC01-001-0… SYSBP    Systo…        2 WEEK… SITT… RIGH… 127   127  
12 CDISC01 CDISC01-001-0… SYSBP    Systo…        3 WEEK… SITT… RIGH… 121.  120.7
# ℹ 3 more variables: VSSTRESU <chr>, N_READINGS <int>, DTYPE <chr>

ADaM Considerations

When building ADVS (Analysis Dataset for Vital Signs):

AVAL = Analysis Value (numeric)
AVALC = Analysis Value (character)
DTYPE = Derivation Type (documents how AVAL was calculated)

Common DTYPE values: “MEAN”, “MEDIAN”, “LAST”, “FIRST”, “WORST”

9 Creating a Wide Format View

For some analyses, having one row per subject per visit (with BP columns side by side) is useful.

9.1 Pivot to Wide Format

# Create wide format with one row per subject/visit
vs_wide <- vs_summary %>%
  select(USUBJID, VISITNUM, VISIT, VSTESTCD, VSSTRESN_MEAN) %>%
  pivot_wider(
    names_from = VSTESTCD,
    values_from = VSSTRESN_MEAN,
    names_prefix = "MEAN_"
  )

cat("Wide format - One row per subject per visit:\n")

Wide format - One row per subject per visit:

print(vs_wide)

# A tibble: 12 × 6
   USUBJID         VISITNUM VISIT     MEAN_DIABP MEAN_PULSE MEAN_SYSBP
   <chr>              <dbl> <chr>          <dbl>      <dbl>      <dbl>
 1 CDISC01-001-001       -1 SCREENING       77.7       74         123.
 2 CDISC01-001-001        1 BASELINE        89.3       71         133.
 3 CDISC01-001-001        2 WEEK 2          73.7       74.3       127 
 4 CDISC01-001-001        3 WEEK 4          75         76.7       121.
 5 CDISC01-001-002       -1 SCREENING       78         76         133.
 6 CDISC01-001-002        1 BASELINE        80         68.3       111 
 7 CDISC01-001-002        2 WEEK 2          79.7       65         130.
 8 CDISC01-001-002        3 WEEK 4          86.7       70         132 
 9 CDISC01-001-003       -1 SCREENING       77         76.3       127.
10 CDISC01-001-003        1 BASELINE        81         68.3       130 
11 CDISC01-001-003        2 WEEK 2          84.3       63.7       125.
12 CDISC01-001-003        3 WEEK 4          79.3       79.3       126.

When to Use Wide Format

Wide format is useful for:

Correlation analysis (e.g., systolic vs. diastolic BP)
Patient profiles (view all vitals at once)
Exploratory data analysis

SDTM VS is always in long format (one row per measurement).

10 Complete Example: Building a VS Domain

Let’s build a complete VS domain from simulated raw data, similar to what you might receive from an EDC system.

10.1 Step 1: Create Raw Vital Signs Data

# Simulate raw vital signs data as it might come from EDC
set.seed(123)

subjects <- paste0("CDISC01-001-00", 1:5)
visits <- c("SCREENING", "BASELINE", "WEEK 2", "WEEK 4", "WEEK 8")
visit_nums <- c(-1, 1, 2, 3, 4)

# Generate comprehensive VS data
raw_vs <- expand_grid(
  USUBJID = subjects,
  VISIT_INFO = tibble(VISIT = visits, VISITNUM = visit_nums)
) %>%
  unnest(VISIT_INFO) %>%
  # Add vital sign tests with 3 readings each for BP, 1 for others
  crossing(
    tibble::tribble(
      ~TEST, ~TESTCD, ~UNIT,       ~MEAN_VAL, ~SD_VAL, ~N_READ,
      "Systolic Blood Pressure",  "SYSBP",  "mmHg",      125,     12,    3,
      "Diastolic Blood Pressure", "DIABP",  "mmHg",      78,       8,    3,
      "Pulse Rate",               "PULSE",  "beats/min", 72,       8,    1,
      "Temperature",              "TEMP",   "C",         36.8,   0.3,    1,
      "Respiratory Rate",         "RESP",   "breaths/min", 16,    2,    1,
      "Weight",                   "WEIGHT", "kg",        75,      10,    1,
      "Height",                   "HEIGHT", "cm",        170,     10,    1
    )
  ) %>%
  # Expand for number of readings
  rowwise() %>%
  mutate(data = list(tibble(READING = 1:N_READ))) %>%
  unnest(data) %>%
  ungroup() %>%
  # Generate values
  rowwise() %>%
  mutate(
    VALUE = round(rnorm(1, MEAN_VAL, SD_VAL), 1),
    # Height stays constant within subject
    VALUE = ifelse(TESTCD == "HEIGHT", round(rnorm(1, 170, 10), 0), VALUE),
    # Weight stays relatively constant
    VALUE = ifelse(TESTCD == "WEIGHT", round(rnorm(1, 75 + (VISITNUM * 0.5), 2), 1), VALUE)
  ) %>%
  ungroup() %>%
  # Clean up - HEIGHT only at SCREENING
  filter(!(TESTCD == "HEIGHT" & VISIT != "SCREENING")) %>%
  mutate(
    STUDYID = "CDISC01",
    VSPOS = case_when(
      TESTCD %in% c("SYSBP", "DIABP", "PULSE") ~ "SITTING",
      TRUE ~ NA_character_
    ),
    VSLOC = case_when(
      TESTCD %in% c("SYSBP", "DIABP") ~ "RIGHT ARM",
      TESTCD == "TEMP" ~ "ORAL",
      TRUE ~ NA_character_
    ),
    VSTPT = ifelse(N_READ > 1, paste("READING", READING), NA_character_),
    VSTPTNUM = ifelse(N_READ > 1, READING, NA_integer_)
  )

cat("Raw VS data generated:\n")

Raw VS data generated:

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

Total records: 255

10.2 Step 2: Transform to SDTM VS Format

# Transform to proper SDTM VS format
sdtm_vs <- raw_vs %>%
  # Add required SDTM variables
  mutate(
    DOMAIN = "VS",
    VSTESTCD = TESTCD,
    VSTEST = TEST,
    VSORRES = as.character(VALUE),
    VSORRESU = UNIT,
    VSSTRESC = VSORRES,
    VSSTRESN = VALUE,
    VSSTRESU = UNIT,
    VSDTC = "2023-03-15",  # Simplified for this example
    EPOCH = case_when(
      VISITNUM < 1 ~ "SCREENING",
      VISITNUM == 1 ~ "TREATMENT",
      TRUE ~ "TREATMENT"
    )
  ) %>%
  # Add sequence number
  group_by(USUBJID) %>%
  mutate(VSSEQ = row_number()) %>%
  ungroup() %>%
  # Select and order SDTM variables
  select(
    STUDYID, DOMAIN, USUBJID, VSSEQ,
    VSTESTCD, VSTEST,
    VSORRES, VSORRESU,
    VSSTRESC, VSSTRESN, VSSTRESU,
    VSPOS, VSLOC,
    VISITNUM, VISIT, VSTPT, VSTPTNUM,
    EPOCH, VSDTC
  ) %>%
  arrange(STUDYID, USUBJID, VISITNUM, VSTESTCD, VSTPTNUM)

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

SDTM VS domain created:

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

Total records: 255

# Preview
head(sdtm_vs, 15)

# A tibble: 15 × 19
   STUDYID DOMAIN USUBJID        VSSEQ VSTESTCD VSTEST VSORRES VSORRESU VSSTRESC
   <chr>   <chr>  <chr>          <int> <chr>    <chr>  <chr>   <chr>    <chr>   
 1 CDISC01 VS     CDISC01-001-0…    11 DIABP    Diast… 80.9    mmHg     80.9    
 2 CDISC01 VS     CDISC01-001-0…    12 DIABP    Diast… 81.2    mmHg     81.2    
 3 CDISC01 VS     CDISC01-001-0…    13 DIABP    Diast… 78.9    mmHg     78.9    
 4 CDISC01 VS     CDISC01-001-0…    14 HEIGHT   Height 172     cm       172     
 5 CDISC01 VS     CDISC01-001-0…    15 PULSE    Pulse… 86.3    beats/m… 86.3    
 6 CDISC01 VS     CDISC01-001-0…    16 RESP     Respi… 17      breaths… 17      
 7 CDISC01 VS     CDISC01-001-0…    17 SYSBP    Systo… 101.4   mmHg     101.4   
 8 CDISC01 VS     CDISC01-001-0…    18 SYSBP    Systo… 133.4   mmHg     133.4   
 9 CDISC01 VS     CDISC01-001-0…    19 SYSBP    Systo… 119.3   mmHg     119.3   
10 CDISC01 VS     CDISC01-001-0…    20 TEMP     Tempe… 36.5    C        36.5    
11 CDISC01 VS     CDISC01-001-0…    21 WEIGHT   Weight 73      kg       73      
12 CDISC01 VS     CDISC01-001-0…     1 DIABP    Diast… 73.5    mmHg     73.5    
13 CDISC01 VS     CDISC01-001-0…     2 DIABP    Diast… 76.2    mmHg     76.2    
14 CDISC01 VS     CDISC01-001-0…     3 DIABP    Diast… 90.5    mmHg     90.5    
15 CDISC01 VS     CDISC01-001-0…     4 PULSE    Pulse… 73      beats/m… 73      
# ℹ 10 more variables: VSSTRESN <dbl>, VSSTRESU <chr>, VSPOS <chr>,
#   VSLOC <chr>, VISITNUM <dbl>, VISIT <chr>, VSTPT <chr>, VSTPTNUM <int>,
#   EPOCH <chr>, VSDTC <chr>

10.3 Step 3: Validate the VS Domain

# Validation checks
cat("=== VS Domain Validation ===\n\n")

=== VS Domain Validation ===

# Check 1: Required variables
required_vars <- c("STUDYID", "DOMAIN", "USUBJID", "VSSEQ", 
                   "VSTESTCD", "VSTEST", "VSORRES", "VSORRESU")
present <- required_vars %in% names(sdtm_vs)
cat("Required variables check:\n")

Required variables check:

for (i in seq_along(required_vars)) {
  cat("  ", required_vars[i], ":", ifelse(present[i], "✓", "✗"), "\n")
}

   STUDYID : ✓ 
   DOMAIN : ✓ 
   USUBJID : ✓ 
   VSSEQ : ✓ 
   VSTESTCD : ✓ 
   VSTEST : ✓ 
   VSORRES : ✓ 
   VSORRESU : ✓

# Check 2: Data completeness
cat("\nData completeness:\n")


Data completeness:

cat("  Missing VSSTRESN:", sum(is.na(sdtm_vs$VSSTRESN)), "\n")

  Missing VSSTRESN: 0

cat("  Missing VSPOS (for BP/pulse):", 
    sum(is.na(sdtm_vs$VSPOS) & sdtm_vs$VSTESTCD %in% c("SYSBP", "DIABP", "PULSE")), "\n")

  Missing VSPOS (for BP/pulse): 0

# Check 3: Value ranges
cat("\nValue ranges by test:\n")


Value ranges by test:

sdtm_vs %>%
  group_by(VSTESTCD, VSTEST, VSSTRESU) %>%
  summarise(
    Min = min(VSSTRESN, na.rm = TRUE),
    Max = max(VSSTRESN, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  print()

# A tibble: 7 × 5
  VSTESTCD VSTEST                   VSSTRESU      Min   Max
  <chr>    <chr>                    <chr>       <dbl> <dbl>
1 DIABP    Diastolic Blood Pressure mmHg         61.6  96.3
2 HEIGHT   Height                   cm          172   189  
3 PULSE    Pulse Rate               beats/min    60.3  86.3
4 RESP     Respiratory Rate         breaths/min  11.4  20.2
5 SYSBP    Systolic Blood Pressure  mmHg        101.  151. 
6 TEMP     Temperature              C            36.4  37.8
7 WEIGHT   Weight                   kg           70.5  81.1

# Check 4: Coverage
cat("\nData coverage:\n")


Data coverage:

cat("  Subjects:", n_distinct(sdtm_vs$USUBJID), "\n")

  Subjects: 5

cat("  Visits:", n_distinct(sdtm_vs$VISIT), "\n")

  Visits: 5

cat("  Tests:", n_distinct(sdtm_vs$VSTESTCD), "\n")

  Tests: 7

11 Preview: ADaM BDS Structure

The VS domain maps directly to the ADaM ADVS dataset using the BDS (Basic Data Structure) format. Here’s a preview of how this transition works:

11.1 Key ADaM Concepts for ADVS

SDTM Variable	ADaM Variable	Description
VSSTRESN	AVAL	Analysis value (numeric)
VSSTRESC	AVALC	Analysis value (character)
VISIT	AVISIT	Analysis visit
VISITNUM	AVISITN	Analysis visit number
-	ABLFL	Baseline flag (“Y”)
-	BASE	Baseline value
-	CHG	Change from baseline
-	PCHG	Percent change from baseline

11.2 Mock ADVS Creation

# Create a simple ADVS preview
# First, aggregate to one value per subject/visit/test (using mean for BP)
advs_prep <- sdtm_vs %>%
  group_by(STUDYID, USUBJID, VSTESTCD, VSTEST, VSSTRESU, VISITNUM, VISIT) %>%
  summarise(
    AVAL = round(mean(VSSTRESN, na.rm = TRUE), 1),
    .groups = "drop"
  ) %>%
  rename(PARAMCD = VSTESTCD, PARAM = VSTEST, AVISITN = VISITNUM, AVISIT = VISIT)

# Add baseline flag and value
advs_preview <- advs_prep %>%
  group_by(USUBJID, PARAMCD) %>%
  mutate(
    # Flag baseline (VISITNUM = 1)
    ABLFL = ifelse(AVISITN == 1, "Y", NA_character_),
    # Get baseline value
    BASE = AVAL[AVISITN == 1][1],
    # Calculate change from baseline
    CHG = ifelse(AVISITN >= 1, round(AVAL - BASE, 1), NA_real_),
    # Calculate percent change
    PCHG = ifelse(AVISITN >= 1 & BASE != 0, 
                  round(100 * (AVAL - BASE) / BASE, 1), NA_real_)
  ) %>%
  ungroup()

cat("ADVS Preview (BDS Structure):\n")

ADVS Preview (BDS Structure):

advs_preview %>%
  filter(USUBJID == first(advs_preview$USUBJID), PARAMCD == "SYSBP") %>%
  select(USUBJID, PARAMCD, AVISITN, AVISIT, AVAL, ABLFL, BASE, CHG, PCHG) %>%
  print()

# A tibble: 5 × 9
  USUBJID         PARAMCD AVISITN AVISIT     AVAL ABLFL  BASE   CHG  PCHG
  <chr>           <chr>     <dbl> <chr>     <dbl> <chr> <dbl> <dbl> <dbl>
1 CDISC01-001-001 SYSBP        -1 SCREENING  118  <NA>    119  NA    NA  
2 CDISC01-001-001 SYSBP         1 BASELINE   119  Y       119   0     0  
3 CDISC01-001-001 SYSBP         2 WEEK 2     127. <NA>    119   8.1   6.8
4 CDISC01-001-001 SYSBP         3 WEEK 4     120. <NA>    119   0.9   0.8
5 CDISC01-001-001 SYSBP         4 WEEK 8     126. <NA>    119   6.7   5.6

Looking Ahead to Week 3

In Week 3 (Days 15-21), we’ll use the admiral package to properly build ADaM datasets like ADVS. The key admiral functions include:

derive_var_extreme_flag() - Create baseline flags
derive_var_base() - Derive baseline values
derive_vars_chg() - Calculate change from baseline

Learn more at: Admiral ADVS Vignette

12 Deliverable Summary

Today you completed the following:

Task	Status
Understood VS domain structure and repeated measures	✓ Done
Explored positioning variables (VSPOS, VSLOC)	✓ Done
Handled multiple readings per visit	✓ Done
Derived summary statistics (mean, min, max)	✓ Done
Created wide format view for analysis	✓ Done
Built a complete VS domain with 5+ vital signs	✓ Done
Previewed ADaM BDS structure (ADVS)	✓ Done

13 Key Takeaways

VS is a Findings class domain - It shares structure with LB (–TESTCD, –ORRES, –STRESN)
Repeated measures are common - Blood pressure often has 3 readings per visit
Position matters - VSPOS (SITTING/STANDING) affects measurements significantly
Location matters - VSLOC (LEFT ARM/RIGHT ARM) should be consistent per protocol
Summary statistics are essential - Choose mean, min, max based on analysis needs
VS maps to ADVS - BDS structure includes baseline, change, and percent change

14 Resources

CDISC SDTM Implementation Guide - VS Domain - Official VS specification
Admiral ADVS Vignette - Building vital signs analysis datasets
Pharmaverse.org - R packages for clinical data
tidyr pivot documentation - Wide/long format transformations
CDISC ADaM BDS Structure - Basic Data Structure specification

15 What’s Next?

In Day 10, we will focus on AE Domain Mastery & SAE Logic:

Deep dive into severity, causality, and outcomes
Understanding SAE criteria (AESER, AESDTH, AESHOSP)
Deriving AE duration and treatment-emergent flags
Preparing for ADaM ADAE creation