ADER+

Exposure-Response Analysis Data

Introduction

Exposure-Response (ER) modeling is a critical tool in drug development for evaluating the relationships between drug exposure, safety, and efficacy. These analyses help characterize dose-response relationships, optimize dosing regimens, and support regulatory decision-making.

While CDISC released the Population Pharmacokinetic (PopPK) Implementation Guide in 2023 (see Basic Data Structure for ADaM Population PK), no equivalent standards currently exist specifically for ER data. However, ER datasets share many structural similarities with PopPK datasets, including numeric covariates, relative time variables, and pharmacokinetic exposure metrics. The examples below demonstrate how PopPK principles can be extended to create standardized ER analysis datasets.

The following four specialized datasets support different aspects of exposure-response analysis:

Dataset	Full Name	Primary Purpose	Key Features
ADER	Exposure-Response Analysis Dataset	Integrated exposure-response relationships across safety and efficacy endpoints	Combines PK metrics (AUC, Cmax) with time-to-event outcomes (OS, PFS), tumor response, and baseline covariates
ADEE	Exposure-Efficacy Analysis Dataset	Exposure relationships with efficacy and tumor response endpoints	Focuses on efficacy outcomes such as tumor measurements, response rates, and progression metrics linked to drug exposure
ADES	Exposure-Safety Analysis Dataset	Exposure relationships with adverse events and safety endpoints	Links exposure metrics to adverse event occurrence, severity, and time-to-onset for safety signal detection
ADTRR	Exposure-Tumor Response Rate Analysis Dataset	Exposure relationships with categorical tumor response metrics	Analyzes exposure impact on response categories (CR, PR, SD, PD) and best overall response

Each dataset builds upon standard ADaM datasets (ADSL, ADRS, ADTTE, ADAE, ADLB, ADVS) and incorporates pharmacokinetic parameters from ADPC/ADPP to create analysis-ready datasets for exposure-response modeling.

The top of this page includes common derivations. See the tabs below for specifics for each dataset.

First Load Packages

First we will load the packages required for our project. We will use {admiral} and {admiralonco} for the creation of analysis data. We will source these from {pharmaverseadam}. {admiral} requires {dplyr}, {lubridate} and {stringr}. Find other {admiral} functions and related variables by searching admiraldiscovery. We will use {metacore} and {metatools} to store and manipulate metadata from our specifications. We will use {xportr} to perform checks on the final data and export to a transport file.

Exposure Response data typically use ADaM data as source, so this example will depend on {pharmaverseadam} with data from existing {admiral} and {admiralonco} templates.

# Load Packages
library(admiral)
library(admiralonco)
# pharmaverseadam contains example datasets generated from the CDISC pilot
# project SDTM ran through admiral templates
library(pharmaverseadam)
library(dplyr)
library(lubridate)
library(stringr)
library(purrr)
library(rlang)
library(metacore)
library(metatools)
library(xportr)

Load Specifications for Metacore

We have saved our specifications in an Excel file and will load them into {metacore} with the metacore::spec_to_metacore() function. We will subset for ADER first and load the other datasets below.

# ---- Load Specs for Metacore ----
metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>%
  select_dataset("ADER")

Load Source Datasets

We will load ADaM data from {pharmaverseadam}. The main sources will be ADRS, ADTTE, ADAE, and ADTR. We will use ADSL for baseline characteristics and we will derive additional baselines from vital signs ADVS and laboratory data ADLB. We will use exposure metrics from ADPP.

# ---- Load source datasets ----
# Load ADRS, ADTTE, ADSL, ADLB, ADVS, ADEX, ADPP, ADAE and ADTR
adrs <- pharmaverseadam::adrs_onco
adtte <- pharmaverseadam::adtte_onco
adsl <- pharmaverseadam::adsl %>%
  filter(TRT01A != "Screen Failure")
adlb <- pharmaverseadam::adlb
advs <- pharmaverseadam::advs
adex <- pharmaverseadam::adex %>%
  filter(PARCAT1 == "INDIVIDUAL")
adpp <- pharmaverseadam::adpp
adae <- pharmaverseadam::adae
adtr <- pharmaverseadam::adtr_onco

Common Derivations

We will include common derivations for all exposure-response datasets first. These include covariates and exposure metrics.

Derive Covariates Using `{metatools}`

In this step we will create our numeric covariates using the metatools::create_var_from_codelist() function. Because there are 12 separate input/output variable pairs to map from controlled terminology codelists, we use purrr::reduce() to chain the calls programmatically rather than repeating them individually. Additional numeric identifiers (STUDYIDN, USUBJIDN, etc.) and study-level constants (ROUTE, FORM, REGION1) are derived with mutate().

Show codelist derivations

# ---- Derive Covariates ----
# Include numeric values for STUDYIDN, USUBJIDN, SEXN, RACEN etc.

covar <- purrr::reduce(
  list(
    c("STUDYID", "STUDYIDN"),
    c("SEX", "SEXN"),
    c("RACE", "RACEN"),
    c("ETHNIC", "ETHNICN"),
    c("ARMCD", "COHORT"),
    c("ARMCD", "COHORTC"),
    c("ARM", "ARMN"),
    c("ACTARM", "ACTARMN"),
    c("TRT01A", "TRT01AN"),
    c("TRT01P", "TRT01PN"),
    c("COUNTRY", "COUNTRYN"),
    c("COUNTRY", "COUNTRYL")
  ),
  ~ create_var_from_codelist(.x, metacore,
    input_var = !!rlang::sym(.y[1]),
    out_var   = !!rlang::sym(.y[2])
  ),
  .init = adsl
) %>%
  mutate(
    STUDYIDN = as.numeric(word(USUBJID, 1, sep = fixed("-"))),
    SITEIDN = as.numeric(word(USUBJID, 2, sep = fixed("-"))),
    USUBJIDN = as.numeric(word(USUBJID, 3, sep = fixed("-"))),
    SUBJIDN = as.numeric(SUBJID),
    ROUTE = unique(adex$EXROUTE)[1],
    FORM = unique(adex$EXDOSFRM)[1],
    REGION1 = COUNTRY,
    REGION1N = COUNTRYN
  ) %>%
  create_var_from_codelist(metacore, input_var = FORM, out_var = FORMN) %>%
  create_var_from_codelist(metacore, input_var = ROUTE, out_var = ROUTEN)

Sample of Data

Derive Additional Baselines

Next we add additional baselines from vital signs and laboratory data. We will use admiral::derive_vars_merged() and admiral::derive_vars_transposed() to merge these datasets. Note that we use get_admiral_option("subject_keys") throughout to identify unique subjects instead of listing STUDYID and USUBJID.

Show baseline derivations

# ---- Derive additional baselines from ADVS and ADLB ----

labsbl <- adlb %>%
  filter(ABLFL == "Y" & PARAMCD %in% c("CREAT", "ALT", "AST", "BILI")) %>%
  mutate(PARAMCDB = paste0(PARAMCD, "BL")) %>%
  select(!!!get_admiral_option("subject_keys"), PARAMCDB, AVAL)

covar_vslb <- covar %>%
  derive_vars_merged(
    dataset_add = advs,
    filter_add = PARAMCD == "HEIGHT" & ABLFL == "Y",
    by_vars = get_admiral_option("subject_keys"),
    new_vars = exprs(HTBL = AVAL)
  ) %>%
  derive_vars_merged(
    dataset_add = advs,
    filter_add = PARAMCD == "WEIGHT" & ABLFL == "Y",
    by_vars = get_admiral_option("subject_keys"),
    new_vars = exprs(WTBL = AVAL)
  ) %>%
  derive_vars_transposed(
    dataset_merge = labsbl,
    by_vars = get_admiral_option("subject_keys"),
    key_var = PARAMCDB,
    value_var = AVAL
  ) %>%
  mutate(
    BMIBL = compute_bmi(height = HTBL, weight = WTBL),
    BSABL = compute_bsa(
      height = HTBL,
      weight = WTBL,
      method = "Mosteller"
    ),
    CRCLBL = compute_egfr(
      creat = CREATBL, creatu = "SI", age = AGE, weight = WTBL, sex = SEX,
      method = "CRCL"
    ),
    EGFRBL = compute_egfr(
      creat = CREATBL, creatu = "SI", age = AGE, weight = WTBL, sex = SEX,
      method = "CKD-EPI"
    )
  ) %>%
  rename(TBILBL = BILIBL)

Sample of Data

Add Exposure Metrics from `ADPP`

An essential component of ER modeling is the exposure metrics calculated from the PK concentration data in ADPC, ADPPK, or ADPP. Here we will use AUCLST and CMAX from the ADPP dataset from {admiral} in {pharmaverseadam}. See the {aNCA} package for details about calculating AUC, CMAX, and other parameters. If ADPP contains parameters from multiple visits (e.g., Cycle 1 Day 1 and steady-state), add a filter condition to restrict to the steady-state visit — for example filter = PARAMCD %in% c("AUCLST", "CMAX") & AVISIT == "Cycle 1 Day 8" — to ensure one record per subject is transposed.

# ---- Add Exposure Metrics  ----
# Add appropriate exposure metrics from ADPP. Here we use AUCLST and CMAX as examples
# this could be extended to include other parameters such as AUCINF, AUCALL, Tmax, Tlast etc.
# depending on the needs of the analysis.
# NOTE: If ADPP contains multiple visits (e.g., Day 1 and steady-state), add an AVISIT
# filter below to select only the steady-state visit, e.g. filter_add = AVISIT == "Cycle 1 Day 8",
# to ensure derive_vars_transposed() produces one record per subject.
covar_auc <- covar_vslb %>%
  derive_vars_transposed(
    dataset_merge = adpp,
    filter = PARAMCD %in% c("AUCLST", "CMAX"),
    by_vars = get_admiral_option("subject_keys"),
    key_var = PARAMCD,
    value_var = AVAL
  ) %>%
  rename(AUCSS = AUCLST, CMAXSS = CMAX)

Sample of Data

Conclusions

These Exposure-Response datasets demonstrate a practical approach to standardizing ER data using CDISC ADaM principles and the Pharmaverse ecosystem. The four datasets work together to support different modeling objectives: ADER for integrated exposure-response analysis, ADEE for efficacy endpoints, ADES for safety evaluation, and ADTRR for tumor response assessment. The framework demonstrated here can be adapted for study-specific ER endpoints. By standardizing the structure of exposure-response data and maintaining consistency with other ADaM formats, these datasets facilitate regulatory review, enhance traceability, and enable reproducible pharmacometric analyses. We encourage the community to adopt and refine these approaches as CDISC continues to develop formal ER data standards.

Code for Individual Datasets

Derivations specific to each dataset continue below. Select the appropriate tab to view.

Dataset-Specific Derivations

Derive Time to Event Variables from `ADTTE`

We will use parameters from ADTTE for Overall Survival, Progression Free Survival and Duration of Response. We will use admiral::derive_vars_transposed() to transpose the AVAL for each PARAMCD. We also transpose the censor variables CNSR here as indicator variables appending “IND” to the parameter name.

# ---- Create ADER base dataset

# For ADTTE censor variables add "IND" to PARAMCD
adttei <- adtte %>%
  mutate(PARAMCD = paste0(PARAMCD, "IND"))

ader_tte <- adsl %>%
  select(!!!get_admiral_option("subject_keys")) %>%
  # Create OS and PFS variables from ADTTE
  derive_vars_transposed(
    dataset_merge = adtte,
    by_vars = get_admiral_option("subject_keys"),
    key_var = PARAMCD,
    value_var = AVAL
  ) %>%
  # Create OS and PFS censor variables
  derive_vars_transposed(
    dataset_merge = adttei,
    by_vars = get_admiral_option("subject_keys"),
    key_var = PARAMCD,
    value_var = CNSR
  )

Sample of Data

We will use Best Overall Response by Investigator (BOR) from ADRS. We attach this to the data with admiral::derive_vars_merged(). Note that derive_vars_transposed() and derive_vars_merged() will be our primary ways of combining data from the different {admiral} and {admiralonco} datasets.

# ---- Add ADRS data ----
# Add response date to ADER for duration of response calculation
ader_bor <- ader_tte %>%
  derive_vars_merged(
    dataset_add = adrs,
    filter_add = PARAMCD == "BOR" & SAFFL == "Y" & ANL01FL == "Y",
    by_vars = get_admiral_option("subject_keys"),
    new_vars = exprs(BOR = AVAL, BORC = AVALC)
  )

Add Analysis Sequence Number using admiral::derive_var_obs_number().

# ---- Add Sequence Number ----
ader_aseq <- ader_bor %>%
  derive_var_obs_number(
    by_vars = get_admiral_option("subject_keys"),
    check_type = "error"
  )

Sample of Data

Combine with Covariates and Exposure

We combine our covariates with the rest of the data

# Combine covariates with ADER data

ader_prefinal <- ader_aseq %>%
  derive_vars_merged(
    dataset_add = covar_auc,
    by_vars = get_admiral_option("subject_keys")
  )

Check Data With metacore and metatools

We use {metacore} objects with {metatools} functions to perform a number of checks on the data. We will drop variables not in the specs and make sure all the variables from the specs are included.

ader <- ader_prefinal %>%
  drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs
  check_variables(metacore) %>% # Check all variables specified are present and no more
  check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT
  order_cols(metacore) %>% # Orders the columns according to the specs
  sort_by_key(metacore) # Sorts the rows by the sort keys

Apply Labels and Formats with xportr

Using {xportr} we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with xportr::xportr_write().

dir <- tempdir() # Change to whichever directory you want to save the dataset in

ader_xpt <- ader %>%
  xportr_type(metacore, domain = "ADER") %>% # Coerce variable type to match specs
  xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata
  xportr_label(metacore) %>% # Assigns variable label from metacore specifications
  xportr_format(metacore) %>% # Assigns variable format from metacore specifications
  xportr_df_label(metacore) %>% # Assigns dataset label from metacore specifications
  xportr_write(file.path(dir, "ader.xpt")) # Write xpt v5 transport file

Load Specifications

metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>%
  select_dataset("ADEE")

Create Base `ADEE` Dataset

ADEE is a BDS dataset structured one record per subject per time-to-event parameter. It links pharmacokinetic exposure metrics from ADPP with efficacy endpoints (OS, PFS, TTP, TTNT) sourced from ADTTE, enabling exposure-efficacy modeling and analysis of progression and survival outcomes.

# ---- Create ADEE base dataset

# Get variable names from both datasets
adsl_vars <- names(adsl)
adtte_vars <- names(adtte)

# Find common variables
common_vars <- intersect(adsl_vars, adtte_vars)

# Remove key variables to get variables to drop
vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID"))

# Ensure PARAMN exists in ADTTE
if (!"PARAMN" %in% names(adtte)) {
  adtte <- adtte %>%
    mutate(
      PARAMN = case_when(
        PARAMCD == "PFS" ~ 1,
        PARAMCD == "OS" ~ 2,
        PARAMCD == "TTP" ~ 3,
        PARAMCD == "TTNT" ~ 4,
        TRUE ~ 99
      )
    )
}

# ---- Create ADEE Base

adee_base <- adtte %>%
  # Filter to efficacy endpoints
  filter(PARAMCD %in% c("OS", "PFS", "TTP", "TTNT")) %>%
  # Add derived variables
  mutate(
    EVENT = 1 - CNSR,
    AVALU = if_else(!is.na(AVAL), "DAYS", NA_character_),
  ) %>%
  # Remove overlapping variables (use clean method)
  select(-any_of(vars_to_drop))

Sample of Data

Add Analysis Variables

# ---- Add Analysis variables

adee_aseq <- adee_base %>%
  # Analysis flags
  mutate(
    ANL01FL = if_else(PARAMCD == "PFS", "Y", ""),
    ANL02FL = if_else(PARAMCD == "OS", "Y", ""),
    ANL03FL = if_else(PARAMCD == "TTP", "Y", ""),
    ANL04FL = if_else(PARAMCD == "TTNT", "Y", "")
  ) %>%
  # Parameter categories
  mutate(
    PARCAT1 = "EFFICACY",
    PARCAT2 = "TIME TO EVENT"
  ) %>%
  # Sequence number
  derive_var_obs_number(
    by_vars = get_admiral_option("subject_keys"),
    order = exprs(PARAMCD),
    new_var = ASEQ,
    check_type = "error"
  )

Sample of Data

Combine with Covariates and Exposure

We combine our covariates with the rest of the data

# Combine covariates with ADEE data

adee_prefinal <- adee_aseq %>%
  derive_vars_merged(
    dataset_add = covar_auc,
    by_vars = get_admiral_option("subject_keys")
  )

Sample of Data

Check Data With metacore and metatools

adee <- adee_prefinal %>%
  drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs
  check_variables(metacore) %>% # Check all variables specified are present and no more
  check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT
  order_cols(metacore) %>% # Orders the columns according to the spec
  sort_by_key(metacore) # Sorts the rows by the sort keys

Apply Labels and Formats with xportr

Using {xportr} we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with xportr::xportr_write().

dir <- tempdir() # Change to whichever directory you want to save the dataset in

adee_xpt <- adee %>%
  xportr_type(metacore, domain = "ADEE") %>% # Coerce variable type to match specs
  xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata
  xportr_label(metacore) %>% # Assigns variable label from metacore specifications
  xportr_format(metacore) %>% # Assigns variable format from metacore specifications
  xportr_df_label(metacore) %>% # Assigns dataset label from metacore specifications
  xportr_write(file.path(dir, "adee.xpt")) # Write xpt v5 transport file

Load Specifications

metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>%
  select_dataset("ADES")

Create Base `ADES` Dataset

ADES contains two levels of records: subject-level flag parameters and event-level AE records. Here we create the five subject-level summary flags using admiral::derive_param_exist_flag(), which evaluates a condition against ADAE for each subject in ADSL and sets AVALC to "Y" or "N" accordingly. admiral::yn_to_numeric() is used within set_values_to to derive the numeric AVAL directly. Note that pharmaverseadam records severity in ASEV/ASEVN rather than AETOXGR/AETOXGRN.

# ---- Create ADES base dataset

# Derive subject-level summary parameters from ADAE using derive_param_exist_flag()
# NOTE: pharmaverseadam uses ASEV/ASEVN (severity) not AETOXGR/AETOXGRN

# Get all subjects from ADSL
adsl_sub <- adsl %>%
  select(!!!get_admiral_option("subject_keys"))

# Derive binary subject-level AE flag parameters
subject_params <- derive_param_exist_flag(
  dataset_ref = adsl_sub,
  dataset_add = adae,
  condition = TRTEMFL == "Y",
  false_value = "N",
  missing_value = "N",
  set_values_to = exprs(
    AVAL = yn_to_numeric(AVALC),
    PARAMCD = "TEAE",
    PARAM = "Any Treatment-Emergent Adverse Event",
    PARAMN = 1
  )
) %>%
  derive_param_exist_flag(
    dataset_ref = adsl_sub,
    dataset_add = adae,
    condition = TRTEMFL == "Y" & ASEVN == 3,
    false_value = "N",
    missing_value = "N",
    set_values_to = exprs(
      AVAL = yn_to_numeric(AVALC),
      PARAMCD = "TEAESEV",
      PARAM = "Any Severe Treatment-Emergent Adverse Event",
      PARAMN = 2
    )
  ) %>%
  derive_param_exist_flag(
    dataset_ref = adsl_sub,
    dataset_add = adae,
    condition = AESER == "Y",
    false_value = "N",
    missing_value = "N",
    set_values_to = exprs(
      AVAL = yn_to_numeric(AVALC),
      PARAMCD = "SAE",
      PARAM = "Any Serious Adverse Event",
      PARAMN = 3
    )
  ) %>%
  derive_param_exist_flag(
    dataset_ref = adsl_sub,
    dataset_add = adae,
    condition = AREL %in% c("POSSIBLE", "PROBABLE", "RELATED"),
    false_value = "N",
    missing_value = "N",
    set_values_to = exprs(
      AVAL = yn_to_numeric(AVALC),
      PARAMCD = "TRAE",
      PARAM = "Any Treatment-Related Adverse Event",
      PARAMN = 4
    )
  ) %>%
  derive_param_exist_flag(
    dataset_ref = adsl_sub,
    dataset_add = adae,
    condition = AEACN == "DRUG WITHDRAWN",
    false_value = "N",
    missing_value = "N",
    set_values_to = exprs(
      AVAL = yn_to_numeric(AVALC),
      PARAMCD = "AEDCN",
      PARAM = "AE Leading to Treatment Discontinuation",
      PARAMN = 5
    )
  ) %>%
  arrange(USUBJID, PARAMCD)

Sample of Data

Create Event Level Parameters

While subject-level parameters capture whether an event occurred at all, event-level records provide one row per individual adverse event and preserve granular information such as preferred term (AEDECOD), system organ class (AEBODSYS), severity (AESEV/AESEVN), seriousness (AESER), and relatedness (AEREL/AERELN). These records are restricted to treatment-emergent AEs (TRTEMFL == "Y") and combined with the subject-level parameters into a single ADES dataset.

# ---- Create event level parameters

# Get variable names for clean dropping
adsl_vars <- names(covar_auc)
adae_vars <- names(adae)
common_vars <- intersect(adsl_vars, adae_vars)
vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID"))

# Create event-level records from ADAE
# NOTE: Using actual pharmaverseadam variables (ASEV/ASEVN, AREL)
event_params <- adae %>%
  filter(TRTEMFL == "Y") %>% # Treatment-emergent only
  mutate(
    PARAMCD = "AETERM",
    PARAM = "Adverse Event Term",
    PARAMN = 10,
    AVAL = 1, # Event occurred
    AVALC = "Y",

    # Keep AE-specific variables (8-char names)
    # Using actual pharmaverseadam ADAE variables
    AEDECOD = AEDECOD, # Preferred term
    AEBODSYS = AEBODSYS, # System organ class
    AESEV = ASEV, # Severity (char): MILD, MODERATE, SEVERE
    AESEVN = ASEVN, # Severity (num): 1, 2, 3
    AESER = AESER, # Serious flag: Y/N
    AEREL = AREL, # Relationship (char): NOT RELATED, POSSIBLE, etc.

    # Create numeric relationship for analysis
    AERELN = case_when(
      AREL == "NOT RELATED" ~ 0,
      AREL == "UNLIKELY RELATED" ~ 1,
      AREL == "POSSIBLE" ~ 2,
      AREL == "PROBABLE" ~ 3,
      AREL == "RELATED" ~ 4,
      TRUE ~ NA_real_
    ),
    AESTDT = ASTDT, # AE start date (8 chars)
    AEENDT = AENDT # AE end date (8 chars)
  ) %>%
  select(-any_of(vars_to_drop))

# ---- Combine subject and event levels

# Ensure all AE-specific variables exist in subject_params (as NA)
# This prevents issues when binding with event_params
subject_params_complete <- subject_params %>%
  derive_vars_merged(
    dataset_add = adsl_sub,
    by_vars = get_admiral_option("subject_keys"),
  ) %>%
  mutate(
    # Add event-level variables as NA for subject-level records
    AESTDT = as.Date(NA),
    AEENDT = as.Date(NA),
    AEDECOD = NA_character_,
    AEBODSYS = NA_character_,
    AESEV = NA_character_,
    AESEVN = NA_integer_,
    AESER = NA_character_,
    AEREL = NA_character_,
    AERELN = NA_real_
  )

event_params_complete <- event_params %>%
  derive_vars_merged(
    dataset_add = adsl_sub,
    by_vars = get_admiral_option("subject_keys")
  )

# Combine both levels
ades_base <- bind_rows(
  subject_params_complete,
  event_params_complete
) %>%
  arrange(USUBJID, PARAMCD)

Sample of Data

Add Analysis Variables

# ---- Add analysis variables

ades_flags <- ades_base %>%
  # Analysis flags
  mutate(
    ANL01FL = if_else(PARAMCD == "TEAE", "Y", ""),
    ANL02FL = if_else(PARAMCD == "TEAESEV", "Y", ""),
    ANL03FL = if_else(PARAMCD == "SAE", "Y", ""),
    ANL04FL = if_else(PARAMCD == "TRAE", "Y", ""),
    ANL05FL = if_else(PARAMCD == "AETERM", "Y", "")
  ) %>%
  # Parameter categories
  mutate(
    PARCAT1 = "SAFETY",
    PARCAT2 = case_when(
      PARAMN <= 5 ~ "SUBJECT-LEVEL",
      PARAMN >= 10 ~ "EVENT-LEVEL",
      TRUE ~ NA_character_
    )
  ) %>%
  # Analysis timepoint
  mutate(
    AVISIT = if_else(PARAMN <= 5, "OVERALL", "AT EVENT"),
    AVISITN = if_else(PARAMN <= 5, 99, 0)
  ) %>%
  # Sort and create sequence number
  # Use coalesce to handle NA dates (puts them first in sort)
  arrange(USUBJID, PARAMN, coalesce(AESTDT, as.Date("1900-01-01"))) %>%
  group_by(!!!get_admiral_option("subject_keys")) %>%
  mutate(ASEQ = row_number()) %>%
  ungroup()

Combine with Covariates and Exposure

We combine our covariates with the rest of the data

# Combine covariates with ADES data

ades_prefinal <- ades_flags %>%
  derive_vars_merged(
    dataset_add = covar_auc,
    by_vars = get_admiral_option("subject_keys")
  )

Check Data With metacore and metatools

## Check Data With metacore and metatools

ades <- ades_prefinal %>%
  drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs
  check_variables(metacore) %>% # Check all variables specified are present and no more
  check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT
  order_cols(metacore) %>% # Orders the columns according to the spec
  sort_by_key(metacore) # Sorts the rows by the sort keys

The following variable(s) were dropped:
  AESTDT
  AEENDT
  AESEV
  AESEVN
  DOMAIN
  AETERM
  AEBDSYCD
  AELLT
  AELLTCD
  AEPTCD
  AEHLT
  AEHLTCD
  AEHLGT
  AEHLGTCD
  AESOC
  AESOCCD
  AESTDTC
  ASTDTM
  ASTDTF
  ASTTMF
  AEENDTC
  AENDTM
  AENDTF
  AENTMF
  AESTDY
  AEENDY
  AOCCIFL
  AESDTH
  AESLIFE
  AESHOSP
  AESDISAB
  AESCONG
  AREL
  AESCAN
  AESOD
  AEDTC
  LDOSEDTM
  DOSEON
  DOSEU
  ANL03FL
  ANL04FL
  ANL05FL
  PARCAT1
  PARCAT2
  AVISIT
  AVISITN
  COUNTRY
  RFSTDTC
  RFENDTC
  RFXSTDTC
  RFXENDTC
  RFPENDTC
  SCRFDT
  FRVDT
  DTHDTC
  DTHADY
  DTHFL
  LDDTHELD
  LDDTHGR1
  DTH30FL
  DTHA30FL
  DTHDOM
  DTHB30FL
  REGION1
  DMDTC
  DMDY
  AGEU
  AGEGR1
  RACEGR1
  ACTARM
  ACTARMCD
  TRTSDTM
  TRTSTMF
  TRTEDTM
  TRTETMF
  EOSSTT
  EOSDT
  RFICDTC
  RANDDT
  LSTALVDT
  DTHDT
  DTHDTF
  DTHCAUS
  DTHCGR1
  BRTHDTC
  COHORT
  COHORTC
  ACTARMN
  COUNTRYN
  COUNTRYL
  ROUTE
  FORM
  REGION1N
  FORMN
  ROUTEN

No missing or extra variables

Apply Labels and Formats with xportr

Using {xportr} we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with xportr::xportr_write().

dir <- tempdir() # Change to whichever directory you want to save the dataset in

ades_xpt <- ades %>%
  xportr_type(metacore, domain = "ADES") %>% # Coerce variable type to match specs
  xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata
  xportr_label(metacore) %>% # Assigns variable label from metacore specifications
  xportr_format(metacore) %>% # Assigns variable format from metacore specifications
  xportr_df_label(metacore, domain = "ADES") %>% # Assigns dataset label from metacore specifications
  xportr_write(file.path(dir, "ades.xpt")) # Write xpt v5 transport file

Load Specifications

metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>%
  select_dataset("ADTRR")

Create Base `ADTRR` Dataset

We will derive the tumor size parameter PARAMCD = "TSIZE" from ADTR. We will add nominal time and actual time using admiral::derive_var_nfrlt() and admiral::derive_vars_duration(). In traditional Exposure-Response datasets these might be TIME and NTIM variables.

# ---- Create ADTRR base dataset

# Get variable names for clean dropping
adsl_vars <- names(covar_auc)
adtr_vars <- names(adtr)
common_vars <- intersect(adsl_vars, adtr_vars)
vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID"))

tsize_final <- adtr %>%
  filter(PARAMCD == "SDIAM") %>%
  mutate(
    PARAMCD = "TSIZE",
    PARAM = "Target Lesions Sum of Diameters",
    PARAMN = 1
  ) %>%
  # Derive Nominal Relative Time from First Dose (NFRLT)
  derive_var_nfrlt(
    new_var = NFRLT,
    new_var_unit = FRLTU,
    out_unit = "DAYS",
    visit_day = ADY
  ) %>%
  # Derive Actual Relative Time from First Dose (AFRLT)
  derive_vars_duration(
    new_var = AFRLT,
    start_date = TRTSDT,
    end_date = ADT,
    out_unit = "DAYS",
    floor_in = FALSE,
    add_one = FALSE
  ) %>%
  select(-any_of(vars_to_drop))

Sample of Data

Add BOR from `ADRS`

Best Overall Response (BOR) is sourced from ADRS and included as a second parameter in ADTRR. We filter to safety-evaluable subjects (SAFFL == "Y") and derive a numeric response code BORN ordered from most to least favourable: CR (4), PR (3), SD (2), PD (1), NE (0). Overlapping ADSL variables are dropped before combining with the tumour size records.

# ---- Add BOR from ADRS

adrs_vars <- names(adrs)
common_vars_adrs <- intersect(adsl_vars, adrs_vars)
vars_to_drop_adrs <- setdiff(common_vars_adrs, c("STUDYID", "USUBJID"))

bor <- adrs %>%
  filter(PARAMCD == "BOR" & SAFFL == "Y" & ANL01FL == "Y") %>%
  mutate(
    PARAMN = 2,
    # Create BORN from AVALC if AVAL doesn't exist
    BORN = if ("AVAL" %in% names(.)) {
      AVAL
    } else {
      case_when(
        AVALC == "CR" ~ 4,
        AVALC == "PR" ~ 3,
        AVALC == "SD" ~ 2,
        AVALC == "PD" ~ 1,
        AVALC == "NE" ~ 0,
        TRUE ~ NA_real_
      )
    }
  ) %>%
  select(-any_of(vars_to_drop_adrs))

Sample of Data

Derive Nadir

The nadir is the minimum post-baseline tumour size observed for each subject. It is derived from TSIZE records and included as a third ADTRR parameter (PARAMCD = "NADIR"). The percentage change from baseline at the nadir timepoint (NADPCHG) and the visit label at which nadir was reached (NADVST) are preserved to support waterfall plots and other tumour response visualisations.

# ---- Derive Nadir

# Calculate nadir from TSIZE records
# Keep BASE, CHG, PCHG from the nadir timepoint
nadir <- tsize_final %>%
  filter(AVISITN > 0 & !is.na(AVAL)) %>%
  group_by(!!!get_admiral_option("subject_keys")) %>%
  filter(AVAL == min(AVAL, na.rm = TRUE)) %>%
  slice(1) %>%
  ungroup() %>%
  mutate(
    PARAMCD = "NADIR",
    PARAM = "Nadir Tumor Size",
    PARAMN = 3,
    NADIR = AVAL,
    NADPCHG = PCHG, # Keep PCHG at nadir
    NADVST = AVISIT # Keep visit of nadir
  )

Sample of Data

Combine Parameters

The three parameter datasets — tumour size (TSIZE), best overall response (BOR), and nadir (NADIR) — are stacked with bind_rows() and sorted by subject and parameter order. Analysis flags, parameter categories, units, and the analysis sequence number are then added. We use derive_var_obs_number() to compute ASEQ consistently, ordered by PARAMN and AVISITN.

# ---- Combine parameters

adtrr_base <- bind_rows(
  tsize_final,
  bor,
  nadir
) %>%
  arrange(USUBJID, PARAMN, AVISITN)

# ---- Add analysis variables

# Ensure AVALU exists before mutating
if (!"AVALU" %in% names(adtrr_base)) {
  adtrr_base <- adtrr_base %>%
    mutate(AVALU = NA_character_)
}

adtrr_seq <- adtrr_base %>%
  # Analysis flags
  mutate(
    # Baseline flag
    ABLFL = case_when(
      !is.na(ABLFL) ~ ABLFL,
      !is.na(AVISITN) & AVISITN == 0 ~ "Y",
      TRUE ~ ""
    ),

    # Post-baseline flag
    ANL01FL = if_else(!is.na(AVISITN) & AVISITN > 0, "Y", ""),

    # Responders (CR or PR)
    ANL02FL = if_else(!is.na(AVALC) & AVALC %in% c("CR", "PR"), "Y", ""),

    # Has change from baseline
    ANL03FL = if_else(!is.na(PCHG), "Y", "")
  ) %>%
  # Parameter categories
  mutate(
    PARCAT1 = "TUMOR RESPONSE",
    PARCAT2 = case_when(
      PARAMCD == "TSIZE" ~ "MEASUREMENT",
      PARAMCD == "BOR" ~ "OVERALL RESPONSE",
      PARAMCD == "NADIR" ~ "BEST RESPONSE",
      TRUE ~ NA_character_
    )
  ) %>%
  # Set AVALU (now guaranteed to exist)
  mutate(
    AVALU = case_when(
      !is.na(AVALU) & AVALU != "" ~ AVALU, # Keep existing non-empty
      PARAMCD == "TSIZE" ~ "mm",
      PARAMCD == "NADIR" ~ "mm",
      TRUE ~ NA_character_
    )
  ) %>%
  # Sequence number
  derive_var_obs_number(
    by_vars = get_admiral_option("subject_keys"),
    order = exprs(PARAMN, AVISITN),
    new_var = ASEQ,
    check_type = "error"
  ) %>%
  arrange(USUBJID, PARAMN, AVISITN)

Sample of Data

Combine with Covariates and Exposure

# ---- Combine with covariates

adtrr_prefinal <- adtrr_seq %>%
  derive_vars_merged(
    dataset_add = covar_auc,
    by_vars = get_admiral_option("subject_keys")
  )

Check Data With metacore and metatools

## Check Data With metacore and metatools

adtrr <- adtrr_prefinal %>%
  drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs
  check_variables(metacore) %>% # Check all variables specified are present and no more
  check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT
  order_cols(metacore) %>% # Orders the columns according to the spec
  sort_by_key(metacore) # Sorts the rows by the sort keys

The following variable(s) were dropped:
  DOMAIN
  PDFL
  ADTF
  PARCAT1
  PARCAT2
  PARCAT3
  CHGNAD
  PCHGNAD
  ANL04FL
  TRSEQ
  TRGRPID
  TRLNKID
  TRTESTCD
  TRTEST
  TRORRES
  TRORRESU
  TRSTRESC
  TRSTRESN
  TRSTRESU
  TREVAL
  TREVALID
  TRACPTFL
  VISITNUM
  VISIT
  TRDTC
  TULOC
  TULOCGR1
  LSEXP
  LSASS
  RSTESTCD
  RSTEST
  RSORRES
  RSSTRESC
  RSEVAL
  RSEVALID
  RSACPTFL
  RSDTC
  RSSEQ
  NADVST
  COUNTRY
  RFSTDTC
  RFENDTC
  RFXSTDTC
  RFXENDTC
  RFPENDTC
  SCRFDT
  FRVDT
  DTHDTC
  DTHADY
  DTHFL
  LDDTHELD
  LDDTHGR1
  DTH30FL
  DTHA30FL
  DTHDOM
  DTHB30FL
  REGION1
  DMDTC
  DMDY
  AGEU
  AGEGR1
  RACEGR1
  ACTARM
  ACTARMCD
  TRTSDTM
  TRTSTMF
  TRTEDTM
  TRTETMF
  EOSSTT
  EOSDT
  RFICDTC
  RANDDT
  LSTALVDT
  DTHDT
  DTHDTF
  DTHCAUS
  DTHCGR1
  BRTHDTC
  COHORT
  COHORTC
  ACTARMN
  COUNTRYN
  COUNTRYL
  ROUTE
  FORM
  REGION1N
  FORMN
  ROUTEN

No missing or extra variables

Apply Labels and Formats with xportr

Using {xportr} we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with xportr::xportr_write().

dir <- tempdir() # Change to whichever directory you want to save the dataset in

adtrr_xpt <- adtrr %>%
  xportr_type(metacore, domain = "ADTRR") %>% # Coerce variable type to match specs
  xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata
  xportr_label(metacore) %>% # Assigns variable label from metacore specifications
  xportr_format(metacore) %>% # Assigns variable format from metacore specifications
  xportr_df_label(metacore, domain = "ADTRR") %>% # Assigns dataset label from metacore specifications
  xportr_write(file.path(dir, "adtrr.xpt")) # Write xpt v5 transport file

--- title: "ADER+" subtitle: "Exposure-Response Analysis Data" order: 12 --- ```{r setup script, include=FALSE, purl=FALSE} invisible_hook_purl <- function(before, options, ...) { knitr::hook_purl(before, options, ...) NULL } knitr::knit_hooks$set(purl = invisible_hook_purl) source("functions/print_df.R") ``` ## Introduction Exposure-Response (ER) modeling is a critical tool in drug development for evaluating the relationships between drug exposure, safety, and efficacy. These analyses help characterize dose-response relationships, optimize dosing regimens, and support regulatory decision-making. While CDISC released the Population Pharmacokinetic (PopPK) Implementation Guide in 2023 (see [Basic Data Structure for ADaM Population PK](https://www.cdisc.org/standards/foundational/adam/basic-data-structure-adam-poppk-implementation-guide-v1-0)), no equivalent standards currently exist specifically for ER data. However, ER datasets share many structural similarities with PopPK datasets, including numeric covariates, relative time variables, and pharmacokinetic exposure metrics. The examples below demonstrate how PopPK principles can be extended to create standardized ER analysis datasets. The following four specialized datasets support different aspects of exposure-response analysis: | Dataset | Full Name | Primary Purpose | Key Features | |------------------|------------------|-------------------|------------------| | [**ADER**](#ader) | Exposure-Response Analysis Dataset | Integrated exposure-response relationships across safety and efficacy endpoints | Combines PK metrics (AUC, Cmax) with time-to-event outcomes (OS, PFS), tumor response, and baseline covariates | | [**ADEE**](#adee) | Exposure-Efficacy Analysis Dataset | Exposure relationships with efficacy and tumor response endpoints | Focuses on efficacy outcomes such as tumor measurements, response rates, and progression metrics linked to drug exposure | | [**ADES**](#ades) | Exposure-Safety Analysis Dataset | Exposure relationships with adverse events and safety endpoints | Links exposure metrics to adverse event occurrence, severity, and time-to-onset for safety signal detection | | [**ADTRR**](#adtrr) | Exposure-Tumor Response Rate Analysis Dataset | Exposure relationships with categorical tumor response metrics | Analyzes exposure impact on response categories (CR, PR, SD, PD) and best overall response | Each dataset builds upon standard ADaM datasets (ADSL, ADRS, ADTTE, ADAE, ADLB, ADVS) and incorporates pharmacokinetic parameters from ADPC/ADPP to create analysis-ready datasets for exposure-response modeling. The top of this page includes common derivations. See the tabs below for specifics for each dataset. ## First Load Packages First we will load the packages required for our project. We will use `{admiral}` and `{admiralonco}` for the creation of analysis data. We will source these from `{pharmaverseadam}`. `{admiral}` requires `{dplyr}`, `{lubridate}` and `{stringr}`. Find other `{admiral}` functions and related variables by searching [admiraldiscovery](https://pharmaverse.github.io/admiraldiscovery/articles/reactable.html). We will use `{metacore}` and `{metatools}` to store and manipulate metadata from our specifications. We will use `{xportr}` to perform checks on the final data and export to a transport file. Exposure Response data typically use ADaM data as source, so this example will depend on `{pharmaverseadam}` with data from existing `{admiral}` and `{admiralonco}` templates. ```{r echo=TRUE, message=FALSE} #| label: Load Packages # Load Packages library(admiral) library(admiralonco) # pharmaverseadam contains example datasets generated from the CDISC pilot # project SDTM ran through admiral templates library(pharmaverseadam) library(dplyr) library(lubridate) library(stringr) library(purrr) library(rlang) library(metacore) library(metatools) library(xportr) ``` ## Load Specifications for Metacore We have saved [our specifications](../metadata/pk_spec.xlsx) in an Excel file and will load them into `{metacore}` with the `metacore::spec_to_metacore()` function. We will subset for `ADER` first and load the other datasets below. ```{r echo=TRUE, message=FALSE} #| label: Load Specs #| warning: false # ---- Load Specs for Metacore ---- metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>% select_dataset("ADER") ``` ## Load Source Datasets We will load ADaM data from `{pharmaverseadam}`. The main sources will be `ADRS`, `ADTTE`, `ADAE`, and `ADTR`. We will use `ADSL` for baseline characteristics and we will derive additional baselines from vital signs `ADVS` and laboratory data `ADLB`. We will use exposure metrics from `ADPP`. ```{r} #| label: Load Source # ---- Load source datasets ---- # Load ADRS, ADTTE, ADSL, ADLB, ADVS, ADEX, ADPP, ADAE and ADTR adrs <- pharmaverseadam::adrs_onco adtte <- pharmaverseadam::adtte_onco adsl <- pharmaverseadam::adsl %>% filter(TRT01A != "Screen Failure") adlb <- pharmaverseadam::adlb advs <- pharmaverseadam::advs adex <- pharmaverseadam::adex %>% filter(PARCAT1 == "INDIVIDUAL") adpp <- pharmaverseadam::adpp adae <- pharmaverseadam::adae adtr <- pharmaverseadam::adtr_onco ``` ## Common Derivations We will include common derivations for all exposure-response datasets first. These include covariates and exposure metrics. ### Derive Covariates Using `{metatools}` In this step we will create our numeric covariates using the `metatools::create_var_from_codelist()` function. Because there are 12 separate input/output variable pairs to map from controlled terminology codelists, we use `purrr::reduce()` to chain the calls programmatically rather than repeating them individually. Additional numeric identifiers (`STUDYIDN`, `USUBJIDN`, etc.) and study-level constants (`ROUTE`, `FORM`, `REGION1`) are derived with `mutate()`. ```{r} #| label: Common Covariates #| code-fold: true #| code-summary: "Show codelist derivations" # ---- Derive Covariates ---- # Include numeric values for STUDYIDN, USUBJIDN, SEXN, RACEN etc. covar <- purrr::reduce( list( c("STUDYID", "STUDYIDN"), c("SEX", "SEXN"), c("RACE", "RACEN"), c("ETHNIC", "ETHNICN"), c("ARMCD", "COHORT"), c("ARMCD", "COHORTC"), c("ARM", "ARMN"), c("ACTARM", "ACTARMN"), c("TRT01A", "TRT01AN"), c("TRT01P", "TRT01PN"), c("COUNTRY", "COUNTRYN"), c("COUNTRY", "COUNTRYL") ), ~ create_var_from_codelist(.x, metacore, input_var = !!rlang::sym(.y[1]), out_var = !!rlang::sym(.y[2]) ), .init = adsl ) %>% mutate( STUDYIDN = as.numeric(word(USUBJID, 1, sep = fixed("-"))), SITEIDN = as.numeric(word(USUBJID, 2, sep = fixed("-"))), USUBJIDN = as.numeric(word(USUBJID, 3, sep = fixed("-"))), SUBJIDN = as.numeric(SUBJID), ROUTE = unique(adex$EXROUTE)[1], FORM = unique(adex$EXDOSFRM)[1], REGION1 = COUNTRY, REGION1N = COUNTRYN ) %>% create_var_from_codelist(metacore, input_var = FORM, out_var = FORMN) %>% create_var_from_codelist(metacore, input_var = ROUTE, out_var = ROUTEN) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print covar print_df(covar %>% select(USUBJID, TRT01A, TRT01AN, SEX, SEXN, RACE, RACEN, FORM, FORMN)) ``` ### Derive Additional Baselines Next we add additional baselines from vital signs and laboratory data. We will use `admiral::derive_vars_merged()` and `admiral::derive_vars_transposed()` to merge these datasets. Note that we use `get_admiral_option("subject_keys")` throughout to identify unique subjects instead of listing `STUDYID` and `USUBJID`. ```{r} #| label: Baselines #| code-fold: true #| code-summary: "Show baseline derivations" # ---- Derive additional baselines from ADVS and ADLB ---- labsbl <- adlb %>% filter(ABLFL == "Y" & PARAMCD %in% c("CREAT", "ALT", "AST", "BILI")) %>% mutate(PARAMCDB = paste0(PARAMCD, "BL")) %>% select(!!!get_admiral_option("subject_keys"), PARAMCDB, AVAL) covar_vslb <- covar %>% derive_vars_merged( dataset_add = advs, filter_add = PARAMCD == "HEIGHT" & ABLFL == "Y", by_vars = get_admiral_option("subject_keys"), new_vars = exprs(HTBL = AVAL) ) %>% derive_vars_merged( dataset_add = advs, filter_add = PARAMCD == "WEIGHT" & ABLFL == "Y", by_vars = get_admiral_option("subject_keys"), new_vars = exprs(WTBL = AVAL) ) %>% derive_vars_transposed( dataset_merge = labsbl, by_vars = get_admiral_option("subject_keys"), key_var = PARAMCDB, value_var = AVAL ) %>% mutate( BMIBL = compute_bmi(height = HTBL, weight = WTBL), BSABL = compute_bsa( height = HTBL, weight = WTBL, method = "Mosteller" ), CRCLBL = compute_egfr( creat = CREATBL, creatu = "SI", age = AGE, weight = WTBL, sex = SEX, method = "CRCL" ), EGFRBL = compute_egfr( creat = CREATBL, creatu = "SI", age = AGE, weight = WTBL, sex = SEX, method = "CKD-EPI" ) ) %>% rename(TBILBL = BILIBL) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print covar_vslb print_df(covar_vslb %>% select(USUBJID, WTBL, HTBL, BMIBL, CRCLBL)) ``` ### Add Exposure Metrics from `ADPP` An essential component of ER modeling is the exposure metrics calculated from the PK concentration data in `ADPC`, `ADPPK`, or `ADPP`. Here we will use `AUCLST` and `CMAX` from the `ADPP` dataset from `{admiral}` in `{pharmaverseadam}`. See the `{aNCA}` package for details about calculating `AUC`, `CMAX`, and other parameters. If `ADPP` contains parameters from multiple visits (e.g., Cycle 1 Day 1 and steady-state), add a `filter` condition to restrict to the steady-state visit — for example `filter = PARAMCD %in% c("AUCLST", "CMAX") & AVISIT == "Cycle 1 Day 8"` — to ensure one record per subject is transposed. ```{r} #| label: AUCSS # ---- Add Exposure Metrics ---- # Add appropriate exposure metrics from ADPP. Here we use AUCLST and CMAX as examples # this could be extended to include other parameters such as AUCINF, AUCALL, Tmax, Tlast etc. # depending on the needs of the analysis. # NOTE: If ADPP contains multiple visits (e.g., Day 1 and steady-state), add an AVISIT # filter below to select only the steady-state visit, e.g. filter_add = AVISIT == "Cycle 1 Day 8", # to ensure derive_vars_transposed() produces one record per subject. covar_auc <- covar_vslb %>% derive_vars_transposed( dataset_merge = adpp, filter = PARAMCD %in% c("AUCLST", "CMAX"), by_vars = get_admiral_option("subject_keys"), key_var = PARAMCD, value_var = AVAL ) %>% rename(AUCSS = AUCLST, CMAXSS = CMAX) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print covar_auc print_df(covar_auc %>% select( USUBJID, TRT01A, SEX, WTBL, HTBL, AUCSS, CMAXSS )) ``` ## Conclusions These Exposure-Response datasets demonstrate a practical approach to standardizing ER data using CDISC ADaM principles and the Pharmaverse ecosystem. The four datasets work together to support different modeling objectives: `ADER` for integrated exposure-response analysis, `ADEE` for efficacy endpoints, `ADES` for safety evaluation, and `ADTRR` for tumor response assessment. The framework demonstrated here can be adapted for study-specific ER endpoints. By standardizing the structure of exposure-response data and maintaining consistency with other ADaM formats, these datasets facilitate regulatory review, enhance traceability, and enable reproducible pharmacometric analyses. We encourage the community to adopt and refine these approaches as CDISC continues to develop formal ER data standards. ## Code for Individual Datasets Derivations specific to each dataset continue below. Select the appropriate tab to view. ## Dataset-Specific Derivations ::: panel-tabset ## ADER {#ader} ### Derive Time to Event Variables from `ADTTE` We will use parameters from `ADTTE` for Overall Survival, Progression Free Survival and Duration of Response. We will use `admiral::derive_vars_transposed()` to transpose the `AVAL` for each `PARAMCD`. We also transpose the censor variables `CNSR` here as indicator variables appending "IND" to the parameter name. ```{r} #| label: Create ADER base dataset # ---- Create ADER base dataset # For ADTTE censor variables add "IND" to PARAMCD adttei <- adtte %>% mutate(PARAMCD = paste0(PARAMCD, "IND")) ader_tte <- adsl %>% select(!!!get_admiral_option("subject_keys")) %>% # Create OS and PFS variables from ADTTE derive_vars_transposed( dataset_merge = adtte, by_vars = get_admiral_option("subject_keys"), key_var = PARAMCD, value_var = AVAL ) %>% # Create OS and PFS censor variables derive_vars_transposed( dataset_merge = adttei, by_vars = get_admiral_option("subject_keys"), key_var = PARAMCD, value_var = CNSR ) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print ader_tte print_df(ader_tte %>% select(USUBJID, OS, OSIND, PFS, PFSIND, RSD, RSDIND)) ``` We will use Best Overall Response by Investigator (BOR) from `ADRS`. We attach this to the data with `admiral::derive_vars_merged()`. Note that `derive_vars_transposed()` and `derive_vars_merged()` will be our primary ways of combining data from the different `{admiral}` and `{admiralonco}` datasets. ```{r} #| label: ADER Add ADRS # ---- Add ADRS data ---- # Add response date to ADER for duration of response calculation ader_bor <- ader_tte %>% derive_vars_merged( dataset_add = adrs, filter_add = PARAMCD == "BOR" & SAFFL == "Y" & ANL01FL == "Y", by_vars = get_admiral_option("subject_keys"), new_vars = exprs(BOR = AVAL, BORC = AVALC) ) ``` Add Analysis Sequence Number using `admiral::derive_var_obs_number()`. ```{r} #| label: ADER ASEQ # ---- Add Sequence Number ---- ader_aseq <- ader_bor %>% derive_var_obs_number( by_vars = get_admiral_option("subject_keys"), check_type = "error" ) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print ader_aseq print_df(ader_aseq %>% select( USUBJID, ASEQ, OS, PFS, BOR, BORC )) ``` ### Combine with Covariates and Exposure We combine our covariates with the rest of the data ```{r} #| label: ADER Combine with Covariates # Combine covariates with ADER data ader_prefinal <- ader_aseq %>% derive_vars_merged( dataset_add = covar_auc, by_vars = get_admiral_option("subject_keys") ) ``` ### Check Data With metacore and metatools We use `{metacore}` objects with `{metatools}` functions to perform a number of checks on the data. We will drop variables not in the specs and make sure all the variables from the specs are included. ```{r} #| label: ADER Metacore #| warning: false ader <- ader_prefinal %>% drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs check_variables(metacore) %>% # Check all variables specified are present and no more check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT order_cols(metacore) %>% # Orders the columns according to the specs sort_by_key(metacore) # Sorts the rows by the sort keys ``` ### Apply Labels and Formats with xportr Using `{xportr}` we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with `xportr::xportr_write()`. ```{r} #| label: ADER xportr dir <- tempdir() # Change to whichever directory you want to save the dataset in ader_xpt <- ader %>% xportr_type(metacore, domain = "ADER") %>% # Coerce variable type to match specs xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata xportr_label(metacore) %>% # Assigns variable label from metacore specifications xportr_format(metacore) %>% # Assigns variable format from metacore specifications xportr_df_label(metacore) %>% # Assigns dataset label from metacore specifications xportr_write(file.path(dir, "ader.xpt")) # Write xpt v5 transport file ``` ## ADEE {#adee} ### Load Specifications ```{r echo=TRUE, message=FALSE} #| label: ADEE Load Specs #| warning: false metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>% select_dataset("ADEE") ``` ### Create Base `ADEE` Dataset `ADEE` is a BDS dataset structured one record per subject per time-to-event parameter. It links pharmacokinetic exposure metrics from `ADPP` with efficacy endpoints (OS, PFS, TTP, TTNT) sourced from `ADTTE`, enabling exposure-efficacy modeling and analysis of progression and survival outcomes. ```{r} #| label: ADEE Base dataset # ---- Create ADEE base dataset # Get variable names from both datasets adsl_vars <- names(adsl) adtte_vars <- names(adtte) # Find common variables common_vars <- intersect(adsl_vars, adtte_vars) # Remove key variables to get variables to drop vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID")) # Ensure PARAMN exists in ADTTE if (!"PARAMN" %in% names(adtte)) { adtte <- adtte %>% mutate( PARAMN = case_when( PARAMCD == "PFS" ~ 1, PARAMCD == "OS" ~ 2, PARAMCD == "TTP" ~ 3, PARAMCD == "TTNT" ~ 4, TRUE ~ 99 ) ) } # ---- Create ADEE Base adee_base <- adtte %>% # Filter to efficacy endpoints filter(PARAMCD %in% c("OS", "PFS", "TTP", "TTNT")) %>% # Add derived variables mutate( EVENT = 1 - CNSR, AVALU = if_else(!is.na(AVAL), "DAYS", NA_character_), ) %>% # Remove overlapping variables (use clean method) select(-any_of(vars_to_drop)) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print adee_base print_df(adee_base %>% select(USUBJID, PARAMCD, PARAM, AVAL, AVALU)) ``` ### Add Analysis Variables ```{r} #| label: ADEE Analysis variables # ---- Add Analysis variables adee_aseq <- adee_base %>% # Analysis flags mutate( ANL01FL = if_else(PARAMCD == "PFS", "Y", ""), ANL02FL = if_else(PARAMCD == "OS", "Y", ""), ANL03FL = if_else(PARAMCD == "TTP", "Y", ""), ANL04FL = if_else(PARAMCD == "TTNT", "Y", "") ) %>% # Parameter categories mutate( PARCAT1 = "EFFICACY", PARCAT2 = "TIME TO EVENT" ) %>% # Sequence number derive_var_obs_number( by_vars = get_admiral_option("subject_keys"), order = exprs(PARAMCD), new_var = ASEQ, check_type = "error" ) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print adee_seq print_df(adee_aseq %>% select(USUBJID, PARAMCD, PARAM, AVAL, AVALU, ANL01FL, ANL02FL, ASEQ)) ``` ### Combine with Covariates and Exposure We combine our covariates with the rest of the data ```{r} #| label: ADEE Combine with Covariates # Combine covariates with ADEE data adee_prefinal <- adee_aseq %>% derive_vars_merged( dataset_add = covar_auc, by_vars = get_admiral_option("subject_keys") ) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print adee_prefinal print_df(adee_prefinal %>% select(USUBJID, TRT01PN, AUCSS, CMAXSS, PARAMCD, AVAL, AVALU), n = 20) ``` ### Check Data With metacore and metatools We use `{metacore}` objects with `{metatools}` functions to perform a number of checks on the data. We will drop variables not in the specs and make sure all the variables from the specs are included. ```{r} #| label: ADEE Metacore #| warning: false adee <- adee_prefinal %>% drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs check_variables(metacore) %>% # Check all variables specified are present and no more check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT order_cols(metacore) %>% # Orders the columns according to the spec sort_by_key(metacore) # Sorts the rows by the sort keys ``` ### Apply Labels and Formats with xportr Using `{xportr}` we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with `xportr::xportr_write()`. ```{r} #| label: ADEE xportr dir <- tempdir() # Change to whichever directory you want to save the dataset in adee_xpt <- adee %>% xportr_type(metacore, domain = "ADEE") %>% # Coerce variable type to match specs xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata xportr_label(metacore) %>% # Assigns variable label from metacore specifications xportr_format(metacore) %>% # Assigns variable format from metacore specifications xportr_df_label(metacore) %>% # Assigns dataset label from metacore specifications xportr_write(file.path(dir, "adee.xpt")) # Write xpt v5 transport file ``` ## ADES {#ades} ### Load Specifications ```{r echo=TRUE, message=FALSE} #| label: ADES Load Specs #| warning: false metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>% select_dataset("ADES") ``` ### Create Base `ADES` Dataset `ADES` contains two levels of records: subject-level flag parameters and event-level AE records. Here we create the five subject-level summary flags using `admiral::derive_param_exist_flag()`, which evaluates a condition against `ADAE` for each subject in `ADSL` and sets `AVALC` to `"Y"` or `"N"` accordingly. `admiral::yn_to_numeric()` is used within `set_values_to` to derive the numeric `AVAL` directly. Note that `pharmaverseadam` records severity in `ASEV`/`ASEVN` rather than `AETOXGR`/`AETOXGRN`. ```{r} #| label: ADES Base data # ---- Create ADES base dataset # Derive subject-level summary parameters from ADAE using derive_param_exist_flag() # NOTE: pharmaverseadam uses ASEV/ASEVN (severity) not AETOXGR/AETOXGRN # Get all subjects from ADSL adsl_sub <- adsl %>% select(!!!get_admiral_option("subject_keys")) # Derive binary subject-level AE flag parameters subject_params <- derive_param_exist_flag( dataset_ref = adsl_sub, dataset_add = adae, condition = TRTEMFL == "Y", false_value = "N", missing_value = "N", set_values_to = exprs( AVAL = yn_to_numeric(AVALC), PARAMCD = "TEAE", PARAM = "Any Treatment-Emergent Adverse Event", PARAMN = 1 ) ) %>% derive_param_exist_flag( dataset_ref = adsl_sub, dataset_add = adae, condition = TRTEMFL == "Y" & ASEVN == 3, false_value = "N", missing_value = "N", set_values_to = exprs( AVAL = yn_to_numeric(AVALC), PARAMCD = "TEAESEV", PARAM = "Any Severe Treatment-Emergent Adverse Event", PARAMN = 2 ) ) %>% derive_param_exist_flag( dataset_ref = adsl_sub, dataset_add = adae, condition = AESER == "Y", false_value = "N", missing_value = "N", set_values_to = exprs( AVAL = yn_to_numeric(AVALC), PARAMCD = "SAE", PARAM = "Any Serious Adverse Event", PARAMN = 3 ) ) %>% derive_param_exist_flag( dataset_ref = adsl_sub, dataset_add = adae, condition = AREL %in% c("POSSIBLE", "PROBABLE", "RELATED"), false_value = "N", missing_value = "N", set_values_to = exprs( AVAL = yn_to_numeric(AVALC), PARAMCD = "TRAE", PARAM = "Any Treatment-Related Adverse Event", PARAMN = 4 ) ) %>% derive_param_exist_flag( dataset_ref = adsl_sub, dataset_add = adae, condition = AEACN == "DRUG WITHDRAWN", false_value = "N", missing_value = "N", set_values_to = exprs( AVAL = yn_to_numeric(AVALC), PARAMCD = "AEDCN", PARAM = "AE Leading to Treatment Discontinuation", PARAMN = 5 ) ) %>% arrange(USUBJID, PARAMCD) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print subject_params print_df(subject_params %>% select(USUBJID, PARAMCD, PARAM, AVAL, AVALC), n = 20) ``` ### Create Event Level Parameters While subject-level parameters capture whether an event occurred at all, event-level records provide one row per individual adverse event and preserve granular information such as preferred term (`AEDECOD`), system organ class (`AEBODSYS`), severity (`AESEV`/`AESEVN`), seriousness (`AESER`), and relatedness (`AEREL`/`AERELN`). These records are restricted to treatment-emergent AEs (`TRTEMFL == "Y"`) and combined with the subject-level parameters into a single `ADES` dataset. ```{r} #| label: ADES Event level parameters # ---- Create event level parameters # Get variable names for clean dropping adsl_vars <- names(covar_auc) adae_vars <- names(adae) common_vars <- intersect(adsl_vars, adae_vars) vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID")) # Create event-level records from ADAE # NOTE: Using actual pharmaverseadam variables (ASEV/ASEVN, AREL) event_params <- adae %>% filter(TRTEMFL == "Y") %>% # Treatment-emergent only mutate( PARAMCD = "AETERM", PARAM = "Adverse Event Term", PARAMN = 10, AVAL = 1, # Event occurred AVALC = "Y", # Keep AE-specific variables (8-char names) # Using actual pharmaverseadam ADAE variables AEDECOD = AEDECOD, # Preferred term AEBODSYS = AEBODSYS, # System organ class AESEV = ASEV, # Severity (char): MILD, MODERATE, SEVERE AESEVN = ASEVN, # Severity (num): 1, 2, 3 AESER = AESER, # Serious flag: Y/N AEREL = AREL, # Relationship (char): NOT RELATED, POSSIBLE, etc. # Create numeric relationship for analysis AERELN = case_when( AREL == "NOT RELATED" ~ 0, AREL == "UNLIKELY RELATED" ~ 1, AREL == "POSSIBLE" ~ 2, AREL == "PROBABLE" ~ 3, AREL == "RELATED" ~ 4, TRUE ~ NA_real_ ), AESTDT = ASTDT, # AE start date (8 chars) AEENDT = AENDT # AE end date (8 chars) ) %>% select(-any_of(vars_to_drop)) # ---- Combine subject and event levels # Ensure all AE-specific variables exist in subject_params (as NA) # This prevents issues when binding with event_params subject_params_complete <- subject_params %>% derive_vars_merged( dataset_add = adsl_sub, by_vars = get_admiral_option("subject_keys"), ) %>% mutate( # Add event-level variables as NA for subject-level records AESTDT = as.Date(NA), AEENDT = as.Date(NA), AEDECOD = NA_character_, AEBODSYS = NA_character_, AESEV = NA_character_, AESEVN = NA_integer_, AESER = NA_character_, AEREL = NA_character_, AERELN = NA_real_ ) event_params_complete <- event_params %>% derive_vars_merged( dataset_add = adsl_sub, by_vars = get_admiral_option("subject_keys") ) # Combine both levels ades_base <- bind_rows( subject_params_complete, event_params_complete ) %>% arrange(USUBJID, PARAMCD) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print ades_base print_df(ades_base %>% select(USUBJID, PARAMCD, PARAM, AEDECOD, AVAL, AVALC, AESTDT), n = 20) ``` ### Add Analysis Variables ```{r} #| label: ADES analysis variables # ---- Add analysis variables ades_flags <- ades_base %>% # Analysis flags mutate( ANL01FL = if_else(PARAMCD == "TEAE", "Y", ""), ANL02FL = if_else(PARAMCD == "TEAESEV", "Y", ""), ANL03FL = if_else(PARAMCD == "SAE", "Y", ""), ANL04FL = if_else(PARAMCD == "TRAE", "Y", ""), ANL05FL = if_else(PARAMCD == "AETERM", "Y", "") ) %>% # Parameter categories mutate( PARCAT1 = "SAFETY", PARCAT2 = case_when( PARAMN <= 5 ~ "SUBJECT-LEVEL", PARAMN >= 10 ~ "EVENT-LEVEL", TRUE ~ NA_character_ ) ) %>% # Analysis timepoint mutate( AVISIT = if_else(PARAMN <= 5, "OVERALL", "AT EVENT"), AVISITN = if_else(PARAMN <= 5, 99, 0) ) %>% # Sort and create sequence number # Use coalesce to handle NA dates (puts them first in sort) arrange(USUBJID, PARAMN, coalesce(AESTDT, as.Date("1900-01-01"))) %>% group_by(!!!get_admiral_option("subject_keys")) %>% mutate(ASEQ = row_number()) %>% ungroup() ``` ### Combine with Covariates and Exposure We combine our covariates with the rest of the data ```{r} #| label: ADES Combine with Covariates # Combine covariates with ADES data ades_prefinal <- ades_flags %>% derive_vars_merged( dataset_add = covar_auc, by_vars = get_admiral_option("subject_keys") ) ``` ### Check Data With metacore and metatools ```{r} #| label: ADES metatools ## Check Data With metacore and metatools ades <- ades_prefinal %>% drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs check_variables(metacore) %>% # Check all variables specified are present and no more check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT order_cols(metacore) %>% # Orders the columns according to the spec sort_by_key(metacore) # Sorts the rows by the sort keys ``` ### Apply Labels and Formats with xportr Using `{xportr}` we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with `xportr::xportr_write()`. ```{r} #| label: ADES xportr dir <- tempdir() # Change to whichever directory you want to save the dataset in ades_xpt <- ades %>% xportr_type(metacore, domain = "ADES") %>% # Coerce variable type to match specs xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata xportr_label(metacore) %>% # Assigns variable label from metacore specifications xportr_format(metacore) %>% # Assigns variable format from metacore specifications xportr_df_label(metacore, domain = "ADES") %>% # Assigns dataset label from metacore specifications xportr_write(file.path(dir, "ades.xpt")) # Write xpt v5 transport file ``` ## ADTRR {#adtrr} ### Load Specifications ```{r echo=TRUE, message=FALSE} #| label: ADTRR Load Specs #| warning: false metacore <- spec_to_metacore("./metadata/pk_spec.xlsx", quiet = TRUE) %>% select_dataset("ADTRR") ``` ### Create Base `ADTRR` Dataset We will derive the tumor size parameter `PARAMCD = "TSIZE"` from `ADTR`. We will add nominal time and actual time using `admiral::derive_var_nfrlt()` and `admiral::derive_vars_duration()`. In traditional Exposure-Response datasets these might be `TIME` and `NTIM` variables. ```{r} #| label: ADTRR Base Data # ---- Create ADTRR base dataset # Get variable names for clean dropping adsl_vars <- names(covar_auc) adtr_vars <- names(adtr) common_vars <- intersect(adsl_vars, adtr_vars) vars_to_drop <- setdiff(common_vars, c("STUDYID", "USUBJID")) tsize_final <- adtr %>% filter(PARAMCD == "SDIAM") %>% mutate( PARAMCD = "TSIZE", PARAM = "Target Lesions Sum of Diameters", PARAMN = 1 ) %>% # Derive Nominal Relative Time from First Dose (NFRLT) derive_var_nfrlt( new_var = NFRLT, new_var_unit = FRLTU, out_unit = "DAYS", visit_day = ADY ) %>% # Derive Actual Relative Time from First Dose (AFRLT) derive_vars_duration( new_var = AFRLT, start_date = TRTSDT, end_date = ADT, out_unit = "DAYS", floor_in = FALSE, add_one = FALSE ) %>% select(-any_of(vars_to_drop)) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print tsize_final print_df(tsize_final %>% select(USUBJID, PARAMCD, PARAM, AVISIT, NFRLT, AFRLT, FRLTU, BASE, AVAL, CHG), n = 20) ``` ### Add BOR from `ADRS` Best Overall Response (BOR) is sourced from `ADRS` and included as a second parameter in `ADTRR`. We filter to safety-evaluable subjects (`SAFFL == "Y"`) and derive a numeric response code `BORN` ordered from most to least favourable: CR (4), PR (3), SD (2), PD (1), NE (0). Overlapping `ADSL` variables are dropped before combining with the tumour size records. ```{r} #| label: ADTRR add BOR # ---- Add BOR from ADRS adrs_vars <- names(adrs) common_vars_adrs <- intersect(adsl_vars, adrs_vars) vars_to_drop_adrs <- setdiff(common_vars_adrs, c("STUDYID", "USUBJID")) bor <- adrs %>% filter(PARAMCD == "BOR" & SAFFL == "Y" & ANL01FL == "Y") %>% mutate( PARAMN = 2, # Create BORN from AVALC if AVAL doesn't exist BORN = if ("AVAL" %in% names(.)) { AVAL } else { case_when( AVALC == "CR" ~ 4, AVALC == "PR" ~ 3, AVALC == "SD" ~ 2, AVALC == "PD" ~ 1, AVALC == "NE" ~ 0, TRUE ~ NA_real_ ) } ) %>% select(-any_of(vars_to_drop_adrs)) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print bor print_df(bor %>% select(USUBJID, PARAMCD, PARAM, AVAL, AVALC), n = 20) ``` ### Derive Nadir The nadir is the minimum post-baseline tumour size observed for each subject. It is derived from `TSIZE` records and included as a third `ADTRR` parameter (`PARAMCD = "NADIR"`). The percentage change from baseline at the nadir timepoint (`NADPCHG`) and the visit label at which nadir was reached (`NADVST`) are preserved to support waterfall plots and other tumour response visualisations. ```{r} #| label: ADTRR add NADIR # ---- Derive Nadir # Calculate nadir from TSIZE records # Keep BASE, CHG, PCHG from the nadir timepoint nadir <- tsize_final %>% filter(AVISITN > 0 & !is.na(AVAL)) %>% group_by(!!!get_admiral_option("subject_keys")) %>% filter(AVAL == min(AVAL, na.rm = TRUE)) %>% slice(1) %>% ungroup() %>% mutate( PARAMCD = "NADIR", PARAM = "Nadir Tumor Size", PARAMN = 3, NADIR = AVAL, NADPCHG = PCHG, # Keep PCHG at nadir NADVST = AVISIT # Keep visit of nadir ) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print nadir print_df(nadir %>% select(USUBJID, PARAMCD, PARAM, NADIR, NADPCHG), n = 20) ``` ### Combine Parameters The three parameter datasets — tumour size (`TSIZE`), best overall response (`BOR`), and nadir (`NADIR`) — are stacked with `bind_rows()` and sorted by subject and parameter order. Analysis flags, parameter categories, units, and the analysis sequence number are then added. We use `derive_var_obs_number()` to compute `ASEQ` consistently, ordered by `PARAMN` and `AVISITN`. ```{r} #| label: ADTRR combine parameters # ---- Combine parameters adtrr_base <- bind_rows( tsize_final, bor, nadir ) %>% arrange(USUBJID, PARAMN, AVISITN) # ---- Add analysis variables # Ensure AVALU exists before mutating if (!"AVALU" %in% names(adtrr_base)) { adtrr_base <- adtrr_base %>% mutate(AVALU = NA_character_) } adtrr_seq <- adtrr_base %>% # Analysis flags mutate( # Baseline flag ABLFL = case_when( !is.na(ABLFL) ~ ABLFL, !is.na(AVISITN) & AVISITN == 0 ~ "Y", TRUE ~ "" ), # Post-baseline flag ANL01FL = if_else(!is.na(AVISITN) & AVISITN > 0, "Y", ""), # Responders (CR or PR) ANL02FL = if_else(!is.na(AVALC) & AVALC %in% c("CR", "PR"), "Y", ""), # Has change from baseline ANL03FL = if_else(!is.na(PCHG), "Y", "") ) %>% # Parameter categories mutate( PARCAT1 = "TUMOR RESPONSE", PARCAT2 = case_when( PARAMCD == "TSIZE" ~ "MEASUREMENT", PARAMCD == "BOR" ~ "OVERALL RESPONSE", PARAMCD == "NADIR" ~ "BEST RESPONSE", TRUE ~ NA_character_ ) ) %>% # Set AVALU (now guaranteed to exist) mutate( AVALU = case_when( !is.na(AVALU) & AVALU != "" ~ AVALU, # Keep existing non-empty PARAMCD == "TSIZE" ~ "mm", PARAMCD == "NADIR" ~ "mm", TRUE ~ NA_character_ ) ) %>% # Sequence number derive_var_obs_number( by_vars = get_admiral_option("subject_keys"), order = exprs(PARAMN, AVISITN), new_var = ASEQ, check_type = "error" ) %>% arrange(USUBJID, PARAMN, AVISITN) ``` ```{r eval=TRUE, echo=FALSE, purl=FALSE} #| label: print adtrr_seq print_df(adtrr_seq %>% select(USUBJID, PARAMCD, PARAM, AVISIT, NFRLT, AFRLT, BASE, AVAL, CHG), n = 20) ``` ### Combine with Covariates and Exposure ```{r} #| label: ADTRR prefinal # ---- Combine with covariates adtrr_prefinal <- adtrr_seq %>% derive_vars_merged( dataset_add = covar_auc, by_vars = get_admiral_option("subject_keys") ) ``` ### Check Data With metacore and metatools ```{r} #| label: ADTRR metatools ## Check Data With metacore and metatools adtrr <- adtrr_prefinal %>% drop_unspec_vars(metacore) %>% # Drop unspecified variables from specs check_variables(metacore) %>% # Check all variables specified are present and no more check_ct_data(metacore) %>% # Checks all variables with CT only contain values within the CT order_cols(metacore) %>% # Orders the columns according to the spec sort_by_key(metacore) # Sorts the rows by the sort keys ``` ### Apply Labels and Formats with xportr Using `{xportr}` we check variable type, assign variable length, add variable labels, add variable formats, and save a transport file with `xportr::xportr_write()`. ```{r} #| label: ADTRR xportr dir <- tempdir() # Change to whichever directory you want to save the dataset in adtrr_xpt <- adtrr %>% xportr_type(metacore, domain = "ADTRR") %>% # Coerce variable type to match specs xportr_length(metacore) %>% # Assigns SAS length from a variable level metadata xportr_label(metacore) %>% # Assigns variable label from metacore specifications xportr_format(metacore) %>% # Assigns variable format from metacore specifications xportr_df_label(metacore, domain = "ADTRR") %>% # Assigns dataset label from metacore specifications xportr_write(file.path(dir, "adtrr.xpt")) # Write xpt v5 transport file ``` :::

Introduction

First Load Packages

Load Specifications for Metacore

Load Source Datasets

Common Derivations

Derive Covariates Using {metatools}

Sample of Data

Derive Additional Baselines

Sample of Data

Add Exposure Metrics from ADPP

Sample of Data

Conclusions

Code for Individual Datasets

Dataset-Specific Derivations

Derive Time to Event Variables from ADTTE

Sample of Data

Sample of Data

Combine with Covariates and Exposure

Check Data With metacore and metatools

Apply Labels and Formats with xportr

Load Specifications

Create Base ADEE Dataset

Sample of Data

Add Analysis Variables

Sample of Data

Combine with Covariates and Exposure

Sample of Data

Check Data With metacore and metatools

Apply Labels and Formats with xportr

Load Specifications

Create Base ADES Dataset

Sample of Data

Create Event Level Parameters

Sample of Data

Add Analysis Variables

Combine with Covariates and Exposure

Check Data With metacore and metatools

Apply Labels and Formats with xportr

Load Specifications

Create Base ADTRR Dataset

Sample of Data

Add BOR from ADRS

Sample of Data

Derive Nadir

Sample of Data

Combine Parameters

Sample of Data

Combine with Covariates and Exposure

Check Data With metacore and metatools

Apply Labels and Formats with xportr

Derive Covariates Using `{metatools}`

Add Exposure Metrics from `ADPP`

Derive Time to Event Variables from `ADTTE`

Create Base `ADEE` Dataset

Create Base `ADES` Dataset

Create Base `ADTRR` Dataset

Add BOR from `ADRS`