Create Mockup Table Based on Metadata

In this document, we illustrate a way to create mockup table by using the metadata created by metalite. An easy-to-follow example is used for illustration purpose. If users want to get a comprehensive production ready work, then users need to refine this simple example to align with their cases.

Overview of the example we used

In this example, the available datasets are adae and adsl, which are available in the r2rtf package, i.e., r2rtf::r2rtf_adae and r2rtf::r2rtf_adsl. For these two datasets, our objective is to conduct two adverse event (AE) analysis: (1) AE summary analysis and (3) specific AE analysis. And the population for these two analysis is all participants as treated (APaT). Besides, there is one observations for these two analysis: weeks 0-12. In the analysis, we are interested in three types of AEs: (1) any AEs, (2) series AEs, and (3) drug-related AEs.

Please note the above example servers for illustration purpose, instead of a comprehensive production ready work. If users are interested in more AE analysis with more AE categories, or get more types of populations/observations, modification of code in this vignette is needed.

Prepare the metadata

In this section, we introduce the steps to prepare the metadata used for the aforementioned two AE analysis.

Step 1: Define the metadata

To define the metadata, users need to specify both population and observations datasets. In our example, the observation dataset comes from r2rtf::r2rtf_adae and the population dataset comes from r2rtf::r2rtf_adsl. So we define the metadata by meta_adam() as follows, where users can specify the observation dataset name and population dataset name.

adae <- r2rtf::r2rtf_adae
adsl <- r2rtf::r2rtf_adsl |> rename(TRTA = TRT01A)
meta <- meta_adam(
  observation = adae,
  population = adsl
)

Step 2: Define the analysis plan

To start mockup tables, we begin with defining the mockup analysis plan. To define an analysis plan, we use plan() and add_plan() to add details per one mockup analysis.

Recall in this example, we have 2 analysis plans:

AE summary;
Specific AE analysis.

For the AE summary analysis, we will define analysis = "ae_summary". Similarly we define proper population, observation and parameter key words for

population = "apat": All Participants as Treated (APaT).
observation = "wk12": Weeks 0-12 analyses.
parameter = "any;rel;ser": a table to cover “any”, “drug related” and “serious” adverse events.

For the AE specific analysis, users can follow the similar logic as in AE summary.

So, we can define an analysis plan as below for a total of 2 mockup to generate 4 TLFs.

plan <- plan(
  analysis = "ae_summary", population = "apat",
  observation = "wk12", parameter = "any;rel;ser"
) |>
  add_plan(
    analysis = "ae_specific", population = "apat",
    observation = "wk12",
    parameter = c("any", "rel", "ser")
  )
plan
#>   mock    analysis population observation   parameter
#> 1    1  ae_summary       apat        wk12 any;rel;ser
#> 2    2 ae_specific       apat        wk12         any
#> 3    2 ae_specific       apat        wk12         rel
#> 4    2 ae_specific       apat        wk12         ser

Then, we use incorporate the above defined plan into the metadata by using define_plan(), i.e.,

meta <- meta |> define_plan(plan)

Step 3: Define the keywords

After defining the overview of the analysis plan, we begin to define the meaning of key words.

Define keywords in population:

The following code define the keywords "apat", which is a population keywords for APaT population. And here we also cover the key information such as group variable, the group order, etc…

meta <- meta |>
  define_population(
    name = "apat",
    group = "TRTA",
    group_order = c(
      "High Dose" = "Xanomeline High Dose",
      "Placebo" = "Placebo"
    ),
    subset = SAFFL == "Y"
  )

Define keywords in observation:

The following code define the keywords "wk12", which is an observation keywords for Weeks 0-12 observations.

meta <- meta |>
  define_observation(
    name = "wk12",
    group = "TRTA",
    ae_var = "AEDECOD",
    var = c("AEREL", "AESER"),
    subset = SAFFL == "Y",
    label = "Weeks 0 to 12"
  )

Define keywords in analysis plans:

Note that in the analysis plan, there are 5 keywords:

"any": a parameter keywords for any adverse events.
"rel": a parameter keywords for drug-related adverse events.
"ser": a parameter keywords for serious adverse events.
"ae_summary": an analysis keywords for AE summary analysis.
"ae_specific": an analysis keywords for specific AE analysis.

For the first three keywords, we can define them by define_parameter() as follows. Here we can add (1) the endnotes, (2) labels, (3) the criterion to filter, etc.

meta <- meta |>
  define_parameter(
    name = "any",
    end_notes = c()
  ) |>
  define_parameter(
    name = "rel",
    subset = AEREL %in% c("POSSIBLE", "PROBABLE"),
    label = "drug-realted adverse events"
  ) |>
  define_parameter(
    name = "ser",
    subset = AESER == "Y",
    label = "serious adverse events",
    end_notes = c("Serious adverse events up to 90 days of last dose are included.")
  )

For the last two keywords, we can define them by define_analysis(). In this vignette, we only present the mockup table of AE summary. If users are interested in specific AE analysis, please define the keywords for "ae_specific" as define_analysis(name = "ae_specific", title = ...).

meta <- meta |>
  define_analysis(
    name = "ae_summary",
    title = "Summary of Adverse Events"
  )

Step 4: Build and review the metadata

Once all keywords, datasets, analysis plan are defined, users can build the metadata to compose all the information.

meta <- meta |> meta_build()

After the entire metadata is built, users can review the analysis plan by

meta$plan
#>   mock    analysis population observation   parameter
#> 1    1  ae_summary       apat        wk12 any;rel;ser
#> 2    2 ae_specific       apat        wk12         any
#> 3    2 ae_specific       apat        wk12         rel
#> 4    2 ae_specific       apat        wk12         ser

And the entire meta is

meta
#> ADaM metadata: 
#>    .$data_population     Population data with 254 subjects 
#>    .$data_observation    Observation data with 1191 records 
#>    .$plan    Analysis plan with 4 plans 
#> 
#> 
#>   Analysis population type:
#>     name        id  group var       subset                         label
#> 1 'apat' 'USUBJID' 'TRTA'     SAFFL == 'Y' 'All Participants as Treated'
#>                     group_order
#> 1 Xanomeline High Dose, Placebo
#> 
#> 
#>   Analysis observation type:
#>     name        id  group          var       subset           label    ae_var
#> 1 'wk12' 'USUBJID' 'TRTA' AEREL, AESER SAFFL == 'Y' 'Weeks 0 to 12' 'AEDECOD'
#> 
#> 
#>   Analysis parameter type:
#>    name                         label                               subset
#> 1 'any'          'any adverse events'                                     
#> 2 'rel' 'drug-realted adverse events' AEREL %in% c('POSSIBLE', 'PROBABLE')
#> 3 'ser'      'serious adverse events'                         AESER == 'Y'
#> 
#> 
#>   Analysis function:
#>            name                           label
#> 1  'ae_summary'  'Table: adverse event summary'
#> 2 'ae_specific' 'Table: specific adverse event'

Add components of mockup table

Based on the mockup table, we specify additional parameters that are required in the function we will define in the next section.

mockup = TRUE as we will create mockup for both analysis.
output_report to define the output path based on analysis keywords.

meta$plan <- meta$plan |>
  mutate(
    mockup = TRUE,
    output_report = paste0("./tlf/mock-", analysis, ".rtf")
  )
meta$plan
#>   mock    analysis population observation   parameter mockup
#> 1    1  ae_summary       apat        wk12 any;rel;ser   TRUE
#> 2    2 ae_specific       apat        wk12         any   TRUE
#> 3    2 ae_specific       apat        wk12         rel   TRUE
#> 4    2 ae_specific       apat        wk12         ser   TRUE
#>                output_report
#> 1  ./tlf/mock-ae_summary.rtf
#> 2 ./tlf/mock-ae_specific.rtf
#> 3 ./tlf/mock-ae_specific.rtf
#> 4 ./tlf/mock-ae_specific.rtf

After the plan is updated, we can construct call program. Here we also provide a global argument for the data source.

spec_call_program(meta,
  data_source = "[adam-adsl; adae]"
)
#> [1] "ae_summary(meta = meta, population = 'apat', observation = 'wk12', parameter = 'any;rel;ser', mockup = TRUE, output_report = './tlf/mock-ae_summary.rtf', data_source = '[adam-adsl; adae]')"
#> [2] "ae_specific(meta = meta, population = 'apat', observation = 'wk12', parameter = 'any', mockup = TRUE, output_report = './tlf/mock-ae_specific.rtf', data_source = '[adam-adsl; adae]')"      
#> [3] "ae_specific(meta = meta, population = 'apat', observation = 'wk12', parameter = 'rel', mockup = TRUE, output_report = './tlf/mock-ae_specific.rtf', data_source = '[adam-adsl; adae]')"      
#> [4] "ae_specific(meta = meta, population = 'apat', observation = 'wk12', parameter = 'ser', mockup = TRUE, output_report = './tlf/mock-ae_specific.rtf', data_source = '[adam-adsl; adae]')"

The pending effort is to define an ae_summary() and an ae_specific() function that can generate the mock up tables based on the inputs.

Create functions

Let’s create a simplified version of ae_summary() to illustrate the idea.

Click to view the code for function ae_summary()

ae_summary <- function(meta,
                       population,
                       observation,
                       parameter,
                       mockup,
                       output_report,
                       data_source,
                       ...) {
  # Identify parameter keywords
  para_list <- unlist(strsplit(parameter, ";"))

  # Row label
  ae_label <- vapply(para_list,
    FUN = function(x) {
      collect_adam_mapping(meta, x)$label
    },
    FUN.VALUE = character(1)
  )

  # Get treatment grouping order
  trt_order <- eval(collect_adam_mapping(meta, population)$group_order)

  # Get the title/endnotes of TLF
  title_text <- collect_title(
    meta = meta,
    population = population,
    observation = observation,
    parameter = parameter, analysis = "ae_summary"
  )
  title_text[2] <- "{Week}"

  # Logic to create mockup table
  if (mockup) {
    # Generate RTF
    x <- tibble::tibble(
      ae_label = ae_label,
      n_1 = rep("x", length(ae_label)),
      n_2 = rep("x", length(ae_label)), # no display_total/CI
      pct_1 = rep("x.xx", length(ae_label)),
      pct_2 = rep("x.xx", length(ae_label))
    ) |>
      dplyr::select(ae_label, n_1, pct_1, n_2, pct_2, everything()) |>
      # r2rtf::rtf_title(rtf_mock_color(title_text)) |>
      r2rtf::rtf_title(title_text) |>
      r2rtf::rtf_colheader(paste0(" | ", paste(names(trt_order), collapse = " | "), " "),
        col_rel_width = c(3, rep(2, length(trt_order)))
      ) |>
      r2rtf::rtf_colheader(" | n | (%) | n | (%) ",
        border_top = c("", rep("single", 2 * length(trt_order))),
        border_bottom = "single",
        border_left = c("single", rep(c("single", ""), length(trt_order))),
        col_rel_width = c(3, rep(1, 2 * length(trt_order)))
      ) |>
      r2rtf::rtf_body(
        col_rel_width = c(3, rep(1, 2 * length(trt_order))),
        border_left = c("single", rep(c("single", ""), length(trt_order))),
        text_justification = c("l", rep("c", 2 * length(trt_order)))
      ) |>
      r2rtf::rtf_source(data_source)

    # Require r2rtf to use color
    attr(x, "page")$use_color <- TRUE

    # Save RTF to a path
    if (!is.null(output_report)) {
      x |>
        r2rtf::rtf_encode() |>
        r2rtf::write_rtf(output_report)
    }
  }

  if (!mockup) {
    # Logic to perform actual AE summary table.
    # Omit here
  }

  output_report
}

For illustration purposes, we only create mock up table for the first row.

meta_run(meta,
  i = 1,
  data_source = "[adam-adsl; adae]"
)
#> $`ae_summary(meta = meta, population = 'apat', observation = 'wk12', parameter = 'any;rel;ser', mockup = TRUE, output_report = './tlf/mock-ae_summary.rtf', data_source = '[adam-adsl; adae]')`
#> [1] "./tlf/mock-ae_summary.rtf"