Creating a BDS Time-to-Event ADaM

In admiral: ADaM in R Asset Library

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

library(admiraldev)

Introduction

This article describes creating a BDS time-to-event ADaM.

The main part in programming a time-to-event dataset is the definition of the events and censoring times. {admiral} supports single events like death or composite events like disease progression or death. More than one source dataset can be used for the definition of the event and censoring times.

Note: All examples assume CDISC SDTM and/or ADaM format as input unless otherwise specified.

Required Packages

The examples of this vignette require the following packages.

library(admiral)
library(dplyr, warn.conflicts = FALSE)
library(pharmaversesdtm)

library(lubridate)

Programming Workflow

• Read in Data
• Derive Parameters (CNSR, ADT, STARTDT)
• Derive Analysis Value (AVAL)
• Derive Analysis Sequence Number (ASEQ)
• Add ADSL Variables
• Add Labels and Attributes

Read in Data {#readdata}

To start, all datasets needed for the creation of the time-to-event dataset should be read into the environment. This will be a company specific process.

For example purpose, the ADSL dataset---which is included in {admiral}---and the SDTM datasets from {pharmaversesdtm} are used.

data("ae")
data("admiral_adsl")

ae <- convert_blanks_to_na(ae)
adsl <- admiral_adsl

ae <- filter(ae, USUBJID %in% c("01-701-1015", "01-701-1023", "01-703-1086", "01-703-1096", "01-707-1037", "01-716-1024"))

The following code creates a minimally viable ADAE dataset to be used throughout the following examples.

adae <- ae %>%
  left_join(adsl, by = c("STUDYID", "USUBJID")) %>%
  derive_vars_dt(
    new_vars_prefix = "AST",
    dtc = AESTDTC,
    highest_imputation = "M"
  ) %>%
  derive_vars_dt(
    new_vars_prefix = "AEN",
    dtc = AEENDTC,
    highest_imputation = "M",
    date_imputation = "last"
  ) %>%
  mutate(TRTEMFL = if_else(ASTDT >= TRTSDT &
    AENDT <= TRTEDT + days(30), "Y", NA_character_))

Derive Parameters (CNSR, ADT, STARTDT) {#parameters}

To derive the parameter dependent variables like CNSR, ADT, STARTDT, EVNTDESC, SRCDOM, PARAMCD, ... the derive_param_tte() function can be used. It adds one parameter to the input dataset with one observation per subject. Usually it is called several times.

For each subject it is determined if an event occurred. In the affirmative the analysis date ADT is set to the earliest event date. If no event occurred, the analysis date is set to the latest censoring date.

The events and censorings are defined by the event_source() and the censor_source() class respectively. It defines

• which observations (filter parameter) of a source dataset (dataset_name parameter) are potential events or censorings,
• the value of the CNSR variable (censor parameter), and
• which variable provides the date (date parameter).

The date can be provided as date (--DT variable) or datetime (--DTM variable).

CDISC strongly recommends CNSR = 0 for events and positive integers for censorings. {admiral} enforces this recommendation. Therefore the censor parameter is available for censor_source() only. It is defaulted to 1.

The dataset_name parameter expects a character value which is used as an identifier. The actual data which is used for the derivation of the parameter is provided via the source_datasets parameter of derive_param_tte(). It expects a named list of datasets. The names correspond to the identifiers specified for the dataset_name parameter. This allows to define events and censoring independent of the data.

Pre-Defined Time-to-Event Source Objects

The table below shows all pre-defined tte_source objects which should cover the most common use cases.

knitr::kable(admiral:::list_tte_source_objects())

These pre-defined objects can be passed directly to derive_param_tte() to create a new time-to-event parameter.

adtte <- derive_param_tte(
  dataset_adsl = adsl,
  start_date = TRTSDT,
  event_conditions = list(ae_ser_event),
  censor_conditions = list(lastalive_censor),
  source_datasets = list(adsl = adsl, adae = adae),
  set_values_to = exprs(PARAMCD = "TTAESER", PARAM = "Time to First Serious AE")
)

dataset_vignette(
  adtte,
  display_vars = exprs(USUBJID, PARAMCD, PARAM, STARTDT, ADT, CNSR)
)

Single Event

For example, the overall survival time could be defined from treatment start to death. Patients alive or lost to follow-up would be censored to the last alive date. The following call defines a death event based on ADSL variables.

death <- event_source(
  dataset_name = "adsl",
  filter = DTHFL == "Y",
  date = DTHDT
)

A corresponding censoring based on the last known alive date can be defined by the following call.

lstalv <- censor_source(
  dataset_name = "adsl",
  date = LSTALVDT
)

The definitions can be passed to derive_param_tte() to create a new time-to-event parameter.

adtte <- derive_param_tte(
  dataset_adsl = adsl,
  source_datasets = list(adsl = adsl),
  start_date = TRTSDT,
  event_conditions = list(death),
  censor_conditions = list(lstalv),
  set_values_to = exprs(PARAMCD = "OS", PARAM = "Overall Survival")
)

dataset_vignette(
  adtte,
  display_vars = exprs(USUBJID, PARAMCD, PARAM, STARTDT, ADT, CNSR)
)

Note that in practice for efficacy parameters you might use randomization date as the time to event origin date.

Add Additional Information for Events and Censoring (EVNTDESC, SRCVAR, ...)

To add additional information like event or censoring description (EVNTDESC) or source variable (SRCVAR) the set_values_to parameter can be specified in the event/censoring definition.

# define death event #
death <- event_source(
  dataset_name = "adsl",
  filter = DTHFL == "Y",
  date = DTHDT,
  set_values_to = exprs(
    EVNTDESC = "DEATH",
    SRCDOM = "ADSL",
    SRCVAR = "DTHDT"
  )
)

# define censoring at last known alive date #
lstalv <- censor_source(
  dataset_name = "adsl",
  date = LSTALVDT,
  set_values_to = exprs(
    EVNTDESC = "LAST KNOWN ALIVE DATE",
    SRCDOM = "ADSL",
    SRCVAR = "LSTALVDT"
  )
)

# derive time-to-event parameter #
adtte <- derive_param_tte(
  dataset_adsl = adsl,
  source_datasets = list(adsl = adsl),
  event_conditions = list(death),
  censor_conditions = list(lstalv),
  set_values_to = exprs(PARAMCD = "OS", PARAM = "Overall Survival")
)

dataset_vignette(
  adtte,
  display_vars = exprs(USUBJID, EVNTDESC, SRCDOM, SRCVAR, CNSR, ADT)
)
# save adtte and adsl for next section
adtte_bak <- adtte
adsl_bak <- adsl

Handling Subjects Without Assessment

If a subject has no event and has no record meeting the censoring rule, it will not be included in the output dataset. In order to have a record for this subject in the output dataset, another censoring_source() object should be created to specify how those patients will be censored. Therefore the start censoring is defined below to achieve that subjects without data in adrs are censored at the start date.

The ADaM IG requires that a computed date must be accompanied by imputation flags. Thus, if the function detects a --DTF and/or --TMF variable corresponding to start_date then STARTDTF and STARTTMF are set automatically to the values of these variables. If a date variable from one of the event or censoring source datasets is imputed, the imputation flag can be specified for the set_values_to parameter in event_source() or censor_source() (see definition of the start censoring below).

As the CDISC pilot does not contain a RS dataset, the following example for progression free survival uses manually created datasets.

View(adsl)

adsl <- tibble::tribble(
  ~USUBJID, ~DTHFL, ~DTHDT,            ~TRTSDT,           ~TRTSDTF,
  "01",     "Y",    ymd("2021-06-12"), ymd("2021-01-01"), "M",
  "02",     "N",    NA,                ymd("2021-02-03"), NA,
  "03",     "Y",    ymd("2021-08-21"), ymd("2021-08-10"), NA,
  "04",     "N",    NA,                ymd("2021-02-03"), NA,
  "05",     "N",    NA,                ymd("2021-04-01"), "D"
) %>%
  mutate(STUDYID = "AB42")

dataset_vignette(
  adsl,
  display_vars = exprs(USUBJID, DTHFL, DTHDT, TRTSDT, TRTSDTF)
)

View(adrs)

adrs <- tibble::tribble(
  ~USUBJID, ~AVALC, ~ADT,              ~ASEQ,
  "01",     "SD",   ymd("2021-01-03"), 1,
  "01",     "PR",   ymd("2021-03-04"), 2,
  "01",     "PD",   ymd("2021-05-05"), 3,
  "02",     "PD",   ymd("2021-02-03"), 1,
  "04",     "SD",   ymd("2021-02-13"), 1,
  "04",     "PR",   ymd("2021-04-14"), 2,
  "04",     "CR",   ymd("2021-05-15"), 3
) %>%
  mutate(
    STUDYID = "AB42",
    PARAMCD = "OVR",
    PARAM = "Overall Response"
  ) %>%
  select(STUDYID, USUBJID, PARAMCD, PARAM, ADT, ASEQ, AVALC)

dataset_vignette(
  adrs,
  display_vars = exprs(USUBJID, AVALC, ADT, ASEQ, PARAMCD, PARAM)
)

An event for progression free survival occurs if

• progression of disease is observed or
• the subject dies.

Therefore two event_source() objects are defined:

• pd for progression of disease and
• death for death.

Some subjects may experience both events. In this case the first one is selected by derive_param_tte().

# progressive disease event #
pd <- event_source(
  dataset_name = "adrs",
  filter = AVALC == "PD",
  date = ADT,
  set_values_to = exprs(
    EVNTDESC = "PD",
    SRCDOM = "ADRS",
    SRCVAR = "ADT",
    SRCSEQ = ASEQ
  )
)

# death event #
death <- event_source(
  dataset_name = "adsl",
  filter = DTHFL == "Y",
  date = DTHDT,
  set_values_to = exprs(
    EVNTDESC = "DEATH",
    SRCDOM = "ADSL",
    SRCVAR = "DTHDT"
  )
)

Subjects without event must be censored at the last tumor assessment. For the censoring the lastvisit object is defined as all tumor assessments. Please note that it is not necessary to select the last one or exclude assessments which resulted in progression of disease. This is handled within derive_param_tte().

# last tumor assessment censoring (CNSR = 1 by default) #
lastvisit <- censor_source(
  dataset_name = "adrs",
  date = ADT,
  set_values_to = exprs(
    EVNTDESC = "LAST TUMOR ASSESSMENT",
    SRCDOM = "ADRS",
    SRCVAR = "ADT"
  )
)

Patients without tumor assessment should be censored at the start date. Therefore the start object is defined with the treatment start date as censoring date. It is not necessary to exclude patient with tumor assessment in the definition of start because derive_param_tte() selects the last date across all censor_source() objects as censoring date.

# start date censoring (for patients without tumor assessment) (CNSR = 2) #
start <- censor_source(
  dataset_name = "adsl",
  date = TRTSDT,
  censor = 2,
  set_values_to = exprs(
    EVNTDESC = "TREATMENT START",
    SRCDOM = "ADSL",
    SRCVAR = "TRTSDT",
    ADTF = TRTSDTF
  )
)

# derive time-to-event parameter #
adtte <- derive_param_tte(
  dataset_adsl = adsl,
  source_datasets = list(adsl = adsl, adrs = adrs),
  start_date = TRTSDT,
  event_conditions = list(pd, death),
  censor_conditions = list(lastvisit, start),
  set_values_to = exprs(PARAMCD = "PFS", PARAM = "Progression Free Survival")
)

dataset_vignette(
  adtte %>%
    select(
      STUDYID, USUBJID, PARAMCD, PARAM, STARTDT, ADT, ADTF, CNSR,
      EVNTDESC, SRCDOM, SRCVAR
    ),
  display_vars = exprs(USUBJID, PARAMCD, STARTDT, ADT, ADTF, CNSR)
)

Deriving a Series of Time-to-Event Parameters

If several similar time-to-event parameters need to be derived the call_derivation() function is useful.

In the following example parameters for time to first AE, time to first serious AE, and time to first related AE are derived. The censoring is the same for all three. Only the definition of the event differs.

adtte <- adtte_bak
adsl <- adsl_bak

# define censoring #
observation_end <- censor_source(
  dataset_name = "adsl",
  date = pmin(TRTEDT + days(30), EOSDT),
  censor = 1,
  set_values_to = exprs(
    EVNTDESC = "END OF TREATMENT",
    SRCDOM = "ADSL",
    SRCVAR = "TRTEDT"
  )
)

# define time to first AE #
tt_ae <- event_source(
  dataset_name = "ae",
  date = ASTDT,
  set_values_to = exprs(
    EVNTDESC = "ADVERSE EVENT",
    SRCDOM = "AE",
    SRCVAR = "AESTDTC"
  )
)

# define time to first serious AE #
tt_ser_ae <- event_source(
  dataset_name = "ae",
  filter = AESER == "Y",
  date = ASTDT,
  set_values_to = exprs(
    EVNTDESC = "SERIOUS ADVERSE EVENT",
    SRCDOM = "AE",
    SRCVAR = "AESTDTC"
  )
)

# define time to first related AE #
tt_rel_ae <- event_source(
  dataset_name = "ae",
  filter = AEREL %in% c("PROBABLE", "POSSIBLE", "REMOTE"),
  date = ASTDT,
  set_values_to = exprs(
    EVNTDESC = "RELATED ADVERSE EVENT",
    SRCDOM = "AE",
    SRCVAR = "AESTDTC"
  )
)

# derive all three time to event parameters #
adaette <- call_derivation(
  derivation = derive_param_tte,
  variable_params = list(
    params(
      event_conditions = list(tt_ae),
      set_values_to = exprs(PARAMCD = "TTAE")
    ),
    params(
      event_conditions = list(tt_ser_ae),
      set_values_to = exprs(PARAMCD = "TTSERAE")
    ),
    params(
      event_conditions = list(tt_rel_ae),
      set_values_to = exprs(PARAMCD = "TTRELAE")
    )
  ),
  dataset_adsl = adsl,
  source_datasets = list(
    adsl = adsl,
    ae = filter(adae, TRTEMFL == "Y")
  ),
  censor_conditions = list(observation_end)
)

adaette %>%
  select(STUDYID, USUBJID, PARAMCD, STARTDT, ADT, CNSR, EVNTDESC, SRCDOM, SRCVAR) %>%
  arrange(USUBJID, PARAMCD) %>%
  dataset_vignette(display_vars = exprs(USUBJID, PARAMCD, STARTDT, ADT, CNSR, EVNTDESC, SRCDOM, SRCVAR))

Deriving Time-to-Event Parameters Using By Groups

If time-to-event parameters need to be derived for each by group of a source dataset, the by_vars parameter can be specified. Then a time-to-event parameter is derived for each by group.

Please note that CDISC requires separate parameters (PARAMCD, PARAM) for the by groups. Therefore the variables specified for the by_vars parameter are not included in the output dataset. The PARAMCD variable should be specified for the set_value_to parameter using an expression on the right hand side which results in a unique value for each by group. If the values of the by variables should be included in the output dataset, they can be stored in PARCATn variables.

In the following example a time-to-event parameter for each preferred term in the AE dataset is derived.

View(adsl)

adsl <- tibble::tribble(
  ~USUBJID, ~TRTSDT,           ~EOSDT,
  "01",     ymd("2020-12-06"), ymd("2021-03-06"),
  "02",     ymd("2021-01-16"), ymd("2021-02-03")
) %>%
  mutate(STUDYID = "AB42")

dataset_vignette(adsl)

View(ae)

ae <- tibble::tribble(
  ~USUBJID, ~AESTDTC,           ~AESEQ, ~AEDECOD,
  "01",     "2021-01-03T10:56", 1,      "Flu",
  "01",     "2021-03-04",       2,      "Cough",
  "01",     "2021",             3,      "Flu"
) %>%
  mutate(
    STUDYID = "AB42",
    AESTDT = convert_dtc_to_dt(dtc = AESTDTC, highest_imputation = "M")
  )

dataset_vignette(ae)

# define time to first adverse event event #
ttae <- event_source(
  dataset_name = "ae",
  date = AESTDT,
  set_values_to = exprs(
    EVNTDESC = "AE",
    SRCDOM = "AE",
    SRCVAR = "AESTDTC",
    SRCSEQ = AESEQ
  )
)

# define censoring at end of study #
eos <- censor_source(
  dataset_name = "adsl",
  date = EOSDT,
  set_values_to = exprs(
    EVNTDESC = "END OF STUDY",
    SRCDOM = "ADSL",
    SRCVAR = "EOSDT"
  )
)

# derive time-to-event parameter #
adtte <- derive_param_tte(
  dataset_adsl = adsl,
  by_vars = exprs(AEDECOD),
  start_date = TRTSDT,
  event_conditions = list(ttae),
  censor_conditions = list(eos),
  source_datasets = list(adsl = adsl, ae = ae),
  set_values_to = exprs(
    PARAMCD = paste0("TTAE", as.numeric(as.factor(AEDECOD))),
    PARAM = paste("Time to First", AEDECOD, "Adverse Event"),
    PARCAT1 = "TTAE",
    PARCAT2 = AEDECOD
  )
)

dataset_vignette(
  adtte %>%
    select(
      USUBJID, STARTDT, PARAMCD, PARAM, PARCAT1, PARCAT2, ADT, CNSR,
      EVNTDESC, SRCDOM, SRCVAR, SRCSEQ
    ),
  display_vars = exprs(USUBJID, STARTDT, PARAMCD, PARAM, ADT, CNSR, SRCSEQ)
)

Derive Analysis Value (AVAL) {#aval}

The analysis value (AVAL) can be derived by calling derive_vars_duration().

This example derives the time to event in days. Other units can be requested by the specifying the out_unit parameter.

adtte <- adtte_bak
adsl <- adsl_bak

adtte <- derive_vars_duration(
  adtte,
  new_var = AVAL,
  start_date = STARTDT,
  end_date = ADT
)

dataset_vignette(
  adtte
)

Derive Analysis Sequence Number (ASEQ) {#aseq}

The {admiral} function derive_var_obs_number() can be used to derive ASEQ:

adtte <- derive_var_obs_number(
  adtte,
  by_vars = exprs(STUDYID, USUBJID),
  order = exprs(PARAMCD),
  check_type = "error"
)

dataset_vignette(adtte)

Add ADSL Variables {#adslvars}

Variables from ADSL which are required for time-to-event analyses, e.g., treatment variables or covariates can be added using derive_vars_merged().

adtte <- derive_vars_merged(
  adtte,
  dataset_add = adsl,
  new_vars = exprs(ARMCD, ARM, ACTARMCD, ACTARM, AGE, SEX),
  by_vars = exprs(STUDYID, USUBJID)
)

dataset_vignette(
  adtte,
  display_vars = exprs(USUBJID, PARAMCD, CNSR, AVAL, ARMCD, AGE, SEX)
)

Add Labels and Attributes {#attributes}

Adding labels and attributes for SAS transport files is supported by the following packages:

• metacore: establish a common foundation for the use of metadata within an R session.

• metatools: enable the use of metacore objects. Metatools can be used to build datasets or enhance columns in existing datasets as well as checking datasets against the metadata.

• xportr: functionality to associate all metadata information to a local R data frame, perform data set level validation checks and convert into a transport v5 file(xpt).

NOTE: All these packages are in the experimental phase, but the vision is to have them associated with an End to End pipeline under the umbrella of the pharmaverse. An example of applying metadata and perform associated checks can be found at the pharmaverse E2E example.

admiral documentation built on Oct. 19, 2023, 1:08 a.m.