Fully Nested ATC2 → ATC4 → Drug (CMDECOD) Table by Treatment (wide)

Description

Builds a three-level nested summary table of concomitant medications (or similar data), grouped as ATC2 → ATC4 → Drug (CMDECOD), with counts and percentages by treatment arm. Outputs a wide data frame where each treatment column contains n (pct).

Two indent modes are supported for the display label column stat:

• RTF mode (default): If atc4_spaces and cmdecod_spaces are both NULL, and rtf_safe = TRUE, stat will include the provided RTF indent strings (atc4_rtf, cmdecod_rtf) before the label text.

• SAS blanks mode: If atc4_spaces or cmdecod_spaces is provided (non-NULL), stat will use only blank spaces (no RTF codes) as visual indents (SAS-style), regardless of rtf_safe.

Sorting can be controlled by sort_by:

• "count" (default): within each level, sort descending by counts for the column ⁠n__<trtan_coln>⁠ (e.g., n__21), then alphabetically.

• "alpha": alphabetical ascending order at each level.

Rows where all three levels are "UNCODED" (case-insensitive) are pushed to the very end of the table (after all other rows), preserving the nested order.

Usage

ATCbyDrug(
  indata,
  dmdata,
  group_vars,
  trtan_coln,
  rtf_safe = TRUE,
  sort_by = c("count", "alpha"),
  atc4_spaces = NULL,
  cmdecod_spaces = NULL,
  atc4_rtf = "(*ESC*)R/RTF\"\\li180 \"",
  cmdecod_rtf = "(*ESC*)R/RTF\"\\li360 \""
)

Arguments

Details

Denominator (N) is computed from dmdata as distinct USUBJID per main_group. For each level (ATC2, ATC4 within ATC2, Drug/CMDECOD within ATC4), the function computes distinct-subject counts by main_group, the percentage w.r.t. N, and forms "n (pct)". The wide result has:

• stat = display label with indent (RTF or blanks, depending on mode).

• ⁠trt<value>⁠ columns (e.g., trt21, trt22, …): "n (pct)" per treatment value.

• ⁠n__<value>⁠ columns mirroring raw counts (useful for custom sorting or QC).

• Ordering columns: sec_ord, psec_ord, sort_ord (help keep nested order).

Indent modes:

• RTF mode: Use when you want RTF control words in the output for direct RTF rendering. Do not set atc4_spaces/cmdecod_spaces; keep rtf_safe = TRUE.

• SAS blanks mode: Provide atc4_spaces and/or cmdecod_spaces to indent using blanks only (friendly for plain-text outputs or RTF pipelines that inject formatting later).

UNCODED handling: Rows are considered UNCODED only if all three of ATC2, ATC4, and Drug (CMDECOD) equal "UNCODED" (case-insensitive, leading/trailing space ignored). Such rows are assigned to the end of the table after sorting.

Value

A tibble with nested rows containing:

• stat (indented label),

• treatment columns ⁠trt*⁠ (string "n (pct)"),

• raw-count columns ⁠n__*⁠,

• helper ordering columns (sec_ord, psec_ord, sort_ord).

Examples

library(dplyr)

cm <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~ATC2,      ~ATC4,         ~CMDECOD,
  "01",       21,   "A - Alim.", "A01A",        "CHLORHEXIDINE",
  "01",       21,   "A - Alim.", "A01A",        "CHLORHEXIDINE",
  "02",       21,   "A - Alim.", "A01A",        "NYSTATIN",
  "03",       22,   "A - Alim.", "A01A",        "NYSTATIN",
  "04",       22,   "J - Anti.", "J01C",        "AMOXICILLIN",
  "05",       21,   "J - Anti.", "J01C",        "AMOXICILLIN",
  "06",       22,   "UNCODED",   "UNCODED",     "UNCODED"
)



dm <- tibble::tribble(
  ~USUBJID, ~TRTAN,
  "01",       21,
  "02",       21,
  "05",       21,
  "03",       22,
  "04",       22,
  "06",       22
)

out_rtf <- ATCbyDrug(
  indata      = cm,
  dmdata      = dm,
  group_vars  = c("TRTAN", "ATC2", "ATC4", "CMDECOD"),
  trtan_coln  = "21",
  rtf_safe    = TRUE,
  sort_by     = "count"
)

out_rtf



out_spaces <- ATCbyDrug(
  indata          = cm,
  dmdata          = dm,
  group_vars      = c("TRTAN", "ATC2", "ATC4", "CMDECOD"),
  trtan_coln      = "21",
  sort_by         = "count",
  atc4_spaces     = 2,
  cmdecod_spaces  = 4
)

out_spaces



out_alpha <- ATCbyDrug(
  indata      = cm,
  dmdata      = dm,
  group_vars  = c("TRTAN", "ATC2", "ATC4", "CMDECOD"),
  trtan_coln  = "21",
  sort_by     = "alpha",
  rtf_safe    = FALSE
)

out_alpha

SOC → PT summary by treatment (wide), with optional BY-grouping, SOC totals, UNCODED positioning, BY-specific Big-N, and optional Big-N printing

Description

Build a System Organ Class (SOC) → Preferred Term (PT) summary by treatment in a wide layout suitable for clinical TLFs. Optionally stratify the display by a BY variable from the AE dataset, order BY groups by a separate key, add TOTAL rows, control UNCODED placement, and optionally calculate percentages using BY-specific denominators.

Usage

SOCbyPT(
  indata,
  dmdata,
  pop_data = NULL,
  group_vars,
  trtan_coln,
  by_var = NULL,
  by_sort_var = NULL,
  by_sort_numeric = TRUE,
  id_var = "USUBJID",
  rtf_safe = TRUE,
  indent_str = "(*ESC*)R/RTF\"\\li360 \"",
  use_sas_round = FALSE,
  header_blank = FALSE,
  soc_totals = FALSE,
  total_label = "TOTAL SUBJECTS WITH AN EVENT",
  uncoded_position = c("count", "last"),
  bigN_by = NULL,
  print_bigN = FALSE
)

Arguments

Value

A tibble with columns:

• stat

• ⁠trt*⁠ treatment columns

• sort_ord, sec_ord

• by_var, by_sort_var (when BY used)

Examples

library(dplyr)


adae <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~AEBODSYS,          ~AEDECOD,
  "01",       11,   "GASTROINTESTINAL", "NAUSEA",
  "01",       11,   "GASTROINTESTINAL", "VOMITING",
  "02",       11,   "NERVOUS SYSTEM",   "HEADACHE",
  "03",       12,   "GASTROINTESTINAL", "NAUSEA",
  "04",       12,   "NERVOUS SYSTEM",   "DIZZINESS",
  "05",       12,   "UNCODED",          "UNCODED"
)

adsl <- tibble::tribble(
  ~USUBJID, ~TRTAN,
  "01",       11,
  "02",       11,
  "03",       12,
  "04",       12,
  "05",       12
)

out1 <- SOCbyPT(
  indata     = adae,
  dmdata     = adsl,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12"   # reference arm for sorting
)

out1




out2 <- SOCbyPT(
  indata       = adae,
  dmdata       = adsl,
  group_vars   = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln   = "12",
  rtf_safe     = FALSE,
  header_blank = TRUE
)

out2



adae_sex <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~SEX, ~AEBODSYS,          ~AEDECOD,
  "01",       11,   "M",  "GASTROINTESTINAL", "NAUSEA",
  "02",       11,   "F",  "GASTROINTESTINAL", "VOMITING",
  "03",       12,   "M",  "NERVOUS SYSTEM",   "HEADACHE",
  "04",       12,   "F",  "NERVOUS SYSTEM",   "DIZZINESS",
  "05",       12,   "F",  "UNCODED",          "UNCODED"
)

adsl_sex <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~SEX,
  "01",       11,   "M",
  "02",       11,   "F",
  "03",       12,   "M",
  "04",       12,   "F",
  "05",       12,   "F"
)

out3 <- SOCbyPT(
  indata           = adae_sex,
  dmdata           = adsl_sex,
  group_vars       = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln       = "12",
  by_var           = "SEX",
  by_sort_var      = "SEX",
  by_sort_numeric  = FALSE,
  uncoded_position = "last"
)

out3



out4 <- SOCbyPT(
  indata      = adae_sex,
  dmdata      = adsl_sex,
  group_vars  = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln  = "12",
  by_var      = "SEX",
  bigN_by     = "YES",
  print_bigN  = TRUE
)

out4


out4_trtN <- SOCbyPT(
  indata     = adae_sex,
  dmdata     = adsl_sex,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12",
  by_var     = "SEX",
  bigN_by    = "NO",
  print_bigN = TRUE
)

out4_trtN



pop_adsl <- tibble::tribble(
  ~USUBJID, ~TRTAN,
  "01",       11,
  "02",       11,
  "03",       12,
  "04",       12,
  "05",       13
)

out5 <- SOCbyPT(
  indata     = adae,
  dmdata     = adsl,
  pop_data   = pop_adsl,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12"
)

SOC → PT summary by treatment with Grade split (wide)

Description

Summarises AEs by System Organ Class (SOC) → Preferred Term (PT) per treatment arm and splits each arm into Grade buckets (1–5 + NOT REPORTED). The table includes a first TOTAL SUBJECTS WITH AN EVENT row, optional SOC subtotal rows, and RTF-safe indenting for PT lines. The SOC/PT block order can be driven by a reference arm (e.g., TRTAN = 12) and a specific grade via sort_grade (default 5).

Usage

SOCbyPT_Grade(
  indata,
  dmdata,
  pop_data = NULL,
  group_vars,
  trtan_coln,
  grade_num = "AETOXGRN",
  grade_char = NULL,
  by_var = NULL,
  by_sort_var = NULL,
  by_sort_numeric = TRUE,
  bigN_by = NULL,
  print_bigN = FALSE,
  id_var = "USUBJID",
  rtf_safe = TRUE,
  indent_str = "(*ESC*)R/RTF\"\\li360 \"",
  use_sas_round = FALSE,
  header_blank = TRUE,
  soc_totals = FALSE,
  total_label = "TOTAL SUBJECTS WITH AN EVENT",
  nr_char_values = c("NOT REPORTED", "NOT_REPORTED", "NOTREPORTED", "NOT REPRTED", "NR",
    "N", "NA"),
  sort_grade = 5,
  debug = FALSE,
  uncoded_position = c("count", "last")
)

Arguments

Value

A tibble with columns:

• stat

• For each treatment and each grade bucket: TRT<trt>_GRADE1, …, TRT<trt>_GRADE5, TRT<trt>_NOT_REPORTED

• sort_ord, sec_ord

Key features

• Grades from numeric and/or character sources: Uses grade_num (1–5). If a character grade column exists (e.g., "AETOCGR"/"AETOXGR"), it is cleaned and mapped, with values in nr_char_values treated as Not Reported.

• NR logic: (a) For PT rows, a subject contributes the max numeric grade among 1–5 (NR ignored). (b) For the top TOTAL row, if any PT for the subject is NR-only (no numeric grade), the subject contributes to NOT REPORTED; otherwise to their max numeric grade.

• Ordering: Within SOC/PT, order is determined using counts from the reference arm trtan_coln filtered to sort_grade (fallback = all grades).

• BY support: Optional by_var (from AE) adds strata with optional by_sort_var to control strata ordering (numeric or character).

• SOC totals: soc_totals = TRUE adds a SOC subtotal row (max-grade logic).

• Denominators: Ns are computed from dmdata (or pop_data, if provided).

• Big N behavior with BY: controlled by bigN_by (TRT-only vs BY×TRT).

• RTF-safe indent: PT stat values can be indented using indent_str.

• SAS-style rounding: Percentages can follow SAS “round half away from zero” via use_sas_round = TRUE.

• UNCODED placement: uncoded_position = c("count","last"). With "last", the block where SOC == "UNCODED" is forced to the very end (per BY stratum), and within that SOC the PT == "UNCODED" line is forced last. Detection is case-insensitive and robust to extra spaces/non-breaking spaces.

Examples

library(dplyr)

adae <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~AEBODSYS,           ~AEDECOD,          ~AETOXGRN,
  "01",       11,   "GASTROINTESTINAL",  "NAUSEA",          2,
  "01",       11,   "GASTROINTESTINAL",  "VOMITING",        3,
  "02",       11,   "GASTROINTESTINAL",  "NAUSEA",          5,
  "03",       12,   "NERVOUS SYSTEM",    "HEADACHE",        1,
  "03",       12,   "NERVOUS SYSTEM",    "DIZZINESS",       2,
  "04",       12,   "GASTROINTESTINAL",  "NAUSEA",          4
)

adsl <- tibble::tribble(
  ~USUBJID, ~TRTAN,
  "01",       11,
  "02",       11,
  "03",       12,
  "04",       12
)

out1 <- SOCbyPT_Grade(
  indata     = adae,
  dmdata     = adsl,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12"   # reference arm for ordering
)

out1



out2 <- SOCbyPT_Grade(
  indata       = adae,
  dmdata       = adsl,
  group_vars   = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln   = "12",
  soc_totals   = TRUE,
  header_blank = TRUE
)

out2


adae2 <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~AEBODSYS,          ~AEDECOD,     ~AETOXGRN, ~AETOXGR,
  "01",       11,   "GASTROINTESTINAL", "NAUSEA",     2,         "",
  "02",       11,   "GASTROINTESTINAL", "NAUSEA",     NA,        "NR",
  "03",       12,   "NERVOUS SYSTEM",   "HEADACHE",   3,         NA,
  "04",       12,   "UNCODED",          "UNCODED",    NA,        "NOT REPORTED"
)

out3 <- SOCbyPT_Grade(
  indata       = adae2,
  dmdata       = adsl,
  group_vars   = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln   = "12",
  grade_num    = "AETOXGRN",
  grade_char   = "AETOXGR",
  sort_grade   = "NOT REPORTED",
  rtf_safe     = FALSE,
  uncoded_position = "last"
)

out3


adae_sex <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~SEX, ~AEBODSYS,          ~AEDECOD,    ~AETOXGRN,
  "01",       11,   "M",  "GASTROINTESTINAL", "NAUSEA",    2,
  "02",       11,   "F",  "GASTROINTESTINAL", "NAUSEA",    5,
  "03",       12,   "M",  "NERVOUS SYSTEM",   "HEADACHE",  3,
  "04",       12,   "F",  "NERVOUS SYSTEM",   "DIZZINESS", 1
)

adsl_sex <- tibble::tribble(
  ~USUBJID, ~TRTAN, ~SEX,
  "01",       11,   "M",
  "02",       11,   "F",
  "03",       12,   "M",
  "04",       12,   "F"
)

out4_trtN <- SOCbyPT_Grade(
  indata     = adae_sex,
  dmdata     = adsl_sex,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12",
  by_var     = "SEX",
  bigN_by    = "NO",
  print_bigN = TRUE
)

out4_byN <- SOCbyPT_Grade(
  indata     = adae_sex,
  dmdata     = adsl_sex,
  group_vars = c("TRTAN", "AEBODSYS", "AEDECOD"),
  trtan_coln = "12",
  by_var     = "SEX",
  bigN_by    = "YES",
  print_bigN = TRUE
)

out4_trtN
out4_byN

Frequency Table by Group (wide): n (%) with flexible ordering and formats

Description

freq_by() produces a one-level frequency table by treatment (wide layout) where each row is a category of last_group (e.g., a bucketed lab value), and each treatment column shows n (%) using distinct subject counts.

New: If fmt is not provided (NULL), labels are derived from the unique values present in data[[last_group]] (post na_to_code mapping, if used).

It supports:

• SAS-style rounding (use_sas_round = TRUE) for the percent.

• Format mapping via either a named vector or a tibble/data.frame with columns value (codes) and raw (labels).

• Ordering by the numeric value of last_group found in the data, or optionally the union of format + data codes (include_all_fmt_levels).

• Counting NA under a chosen code/label using na_to_code (e.g., code "4" = "MISSING").

• Auto-detecting the subject ID column when id_var is not provided.

Usage

freq_by(
  data,
  denom_data = NULL,
  main_group,
  last_group,
  label,
  sec_ord,
  fmt = NULL,
  use_sas_round = FALSE,
  indent = 2,
  id_var = "USUBJID",
  include_all_fmt_levels = TRUE,
  na_to_code = NULL
)

Arguments

Details

• Counting uses n_distinct(id_var) within each ⁠(main_group, last_group)⁠ cell.

• Percent is 100 * n / N where N = distinct subjects in denom_data by main_group.

• When fmt = NULL, both codes and labels are taken from the observed values of last_group (after applying na_to_code mapping), ordered numerically where possible.

• Output treatment columns are normalized to trtXX if original names start with digits.

• Missing treatment arms are added as "0".

Value

A tibble with:

• stat (character), sort_ord (integer), sec_ord (integer),

• One column per treatment arm (e.g., trt1, trt2, …), with "n (pct)" or "0".

Examples

set.seed(1)

toy_adsl <- tibble::tibble(
  USUBJID = sprintf("ID%03d", 1:60),
  TRTAN   = sample(c(1, 2), size = 60, replace = TRUE),
  AGE     = sample(18:85, size = 60, replace = TRUE),
  SEX     = sample(c("Male", "Female"), size = 60, replace = TRUE),
  ETHNIC  = sample(
    c("Hispanic or Latino",
      "Not Hispanic or Latino",
      "Unknown",
      NA_character_),
    size = 60, replace = TRUE
  )
) |>
  dplyr::mutate(
    AGEGR1 = dplyr::case_when(
      AGE < 65            ~ "<65 years",
      AGE >= 65 & AGE < 75 ~ "65–<75 years",
      AGE >= 75           ~ ">=75 years"
    )
  )

toy_dm <- toy_adsl |>
  dplyr::select(USUBJID, TRTAN)

freq_by(
  data       = toy_adsl,
  denom_data = toy_dm,
  main_group = "TRTAN",
  last_group = "AGEGR1",
  label      = "Age group, n (%)",
  sec_ord    = 1,
  fmt        = NULL,
  na_to_code = NULL
)

freq_by(
  data       = toy_adsl,
  denom_data = toy_dm,
  main_group = "TRTAN",
  last_group = "SEX",
  label      = "Sex, n (%)",
  sec_ord    = 2,
  fmt        = NULL,
  na_to_code = "99"
)

fmt_ethnic <- c(
  "Hispanic or Latino"         = "Hispanic or Latino",
  "Not Hispanic or Latino"     = "Not Hispanic or Latino",
  "Unknown"                    = "Unknown",
  "99"                         = "Missing"
)

freq_by(
  data       = toy_adsl,
  denom_data = toy_dm,
  main_group = "TRTAN",
  last_group = "ETHNIC",
  label      = "Ethnic group, n (%)",
  sec_ord    = 3,
  fmt        = fmt_ethnic,
  include_all_fmt_levels = TRUE,
  na_to_code = "99"
)

One-Line Frequency Summary by Treatment Group

Description

Generates a single-row frequency summary table across treatment groups, reporting counts and percentages of subjects meeting a filter condition.

Usage

freq_by_line(data, id_var, trt_var, filter_expr, label, denom_data = NULL)

Arguments

Details

This function calculates the number and percentage of unique subjects per treatment group (trt_var) satisfying a given filter condition (filter_expr). The result is formatted as "n (pct)" and returned in a single-row tibble, labeled by the provided label. An optional denominator dataset (denom_data) can be specified to override the default denominator population (used to calculate percentages).

Useful for producing compact summary rows (e.g., "SAF Population", "Subjects >= 65") in clinical tables.

Value

A one-row tibble containing "n (pct)" summaries per treatment group.

Examples

set.seed(123)
adsl <- data.frame(
  USUBJID = paste0("SUBJ", 1:100),
  TRT01P = sample(c("0", "54", "100"), 100, replace = TRUE),
  SAFFL = sample(c("Y", "N"), 100, replace = TRUE),
  AGE = sample(18:80, 100, replace = TRUE)
)


freq_by_line(adsl, USUBJID, TRT01P, SAFFL == "Y", label = "SAF population")


saf <- adsl[adsl$SAFFL == "Y", ]

freq_by_line(
  adsl, USUBJID, TRT01P,
  AGE >= 65,
  label = "Age >=65 in SAF",
  denom_data = saf
)

Compare DEV vs VAL datasets (PROC COMPARE-style) with robust file detection

Description

generate_compare_report() compares a developer (DEV) dataset and a validation (VAL) dataset for a given domain and produces outputs similar to SAS ⁠PROC COMPARE⁠.

This function is intended for ADaM/SDTM/TFL validation workflows and supports:

• Directory-driven inputs: DEV and VAL locations are provided via dev_dir and val_dir.

• Case-insensitive domain matching: domain = "ADAE" will match files like ⁠adae.*⁠.

• VAL prefix flexibility: resolves prefix_val variants such as v_, ⁠v-⁠, and v (no separator).

• Automatic extension detection for DEV and VAL files: .sas7bdat, .xpt, .csv, .rds.

• Optional filtering using filter_expr prior to comparison.

• Optional PROC COMPARE-style CSV output with BASE, COMPARE, and DIF triplets.

• Optional LST-like report using arsenal::comparedf() for summarized differences.

Usage

generate_compare_report(
  domain,
  dev_dir,
  val_dir,
  by_vars = c("STUDYID", "USUBJID"),
  vars_to_check = NULL,
  report_dir = NULL,
  prefix_val = "v_",
  max_print = 50,
  write_csv = FALSE,
  run_comparedf = TRUE,
  filter_expr = NULL,
  study_id = NULL,
  author = NULL
)

Arguments

Details

The function looks for exactly one matching domain file per directory:

• DEV: ⁠<domain>.<ext>⁠

• VAL: ⁠<prefix><domain>.<ext>⁠ where ⁠<prefix>⁠ is prefix_val plus common variants supporting underscore/hyphen/no-separator forms (e.g., v_, ⁠v-⁠, v).

Supported extensions (priority order) are: sas7bdat, xpt, csv, rds.

If multiple matches exist for the same domain in a directory (e.g., adae.csv and adae.xpt), the function stops with an ambiguous match error to prevent accidental comparisons.

PROC COMPARE-style CSV behavior When write_csv = TRUE, the output includes:

• ⁠_TYPE_⁠ with values BASE, COMPARE, DIF

• ⁠_OBS_⁠ sequence within each BY key

• For numeric variables, DIF = DEV - VAL

• For Date variables, DIF is integer day difference (as.integer(DEV - VAL))

• For POSIXct variables, DIF is seconds difference (as.numeric(DEV - VAL))

• For other types, DIF is a character mask (X indicates difference)

Value

Invisibly returns a list with:

• only_in_dev: rows present only in DEV (set-difference result)

• only_in_val: rows present only in VAL (set-difference result)

• comparedf: arsenal::comparedf object (or NULL if run_comparedf = FALSE)

See Also

comparedf, fsetdiff, fintersect

Examples

td <- tempdir()
dev_dir <- file.path(td, "dev")
val_dir <- file.path(td, "val")
rpt_dir <- file.path(td, "rpt")
dir.create(dev_dir, showWarnings = FALSE)
dir.create(val_dir, showWarnings = FALSE)
dir.create(rpt_dir, showWarnings = FALSE)


dev <- data.frame(
  STUDYID = "STDY1",
  USUBJID = c("01", "02"),
  AESEQ   = c(1, 1),
  AETERM  = c("HEADACHE", "NAUSEA"),
  stringsAsFactors = FALSE
)
val <- dev
val$AETERM[2] <- "VOMITING"

utils::write.csv(dev, file.path(dev_dir, "adae.csv"), row.names = FALSE)
utils::write.csv(val, file.path(val_dir, "v-adae.csv"), row.names = FALSE)


generate_compare_report(
  domain        = "adae",
  dev_dir       = dev_dir,
  val_dir       = val_dir,
  by_vars       = c("STUDYID","USUBJID","AESEQ"),
  report_dir    = rpt_dir,
  write_csv     = TRUE,
  run_comparedf = FALSE
)


generate_compare_report(
  domain        = "ADAE",
  dev_dir       = dev_dir,
  val_dir       = val_dir,
  by_vars       = c("STUDYID","USUBJID","AESEQ"),
  report_dir    = rpt_dir,
  write_csv     = FALSE,
  run_comparedf = FALSE
)


generate_compare_report(
  domain        = "adae",
  dev_dir       = dev_dir,
  val_dir       = val_dir,
  by_vars       = c("STUDYID","USUBJID","AESEQ"),
  report_dir    = rpt_dir,
  filter_expr   = "USUBJID == '02'",
  write_csv     = TRUE,
  run_comparedf = FALSE
)

Extract Column Metadata from a Data Frame

Description

Inspects a data frame and returns a summary of metadata for each column, including column name, label, format, class/type, missingness, uniqueness, and (optionally) SAS-style display for Date variables (e.g., DATE9 -> 09JUL2012).

Usage

get_column_info(
  df,
  include_attributes = TRUE,
  exclude_attributes = c("class", "row.names"),
  label_attr = c("label", "var.label", "labelled", "Label"),
  format_attr = c("format", "format.sas", "Format", "displayWidth"),
  compute_ranges = TRUE,
  sas_date_display = TRUE
)

Arguments

Value

A tibble with one row per column and metadata fields.

• column: Column name

• label: Label attribute (if present)

• format: Format attribute (if present; e.g., DATE9.)

• class: Class(es)

• typeof: Underlying storage type

• n: Total length

• n_missing: Number of NAs

• n_unique: Number of unique values

• min_raw/max_raw: Min/max as raw values (Date/numeric)

• min_disp/max_disp: Min/max as display strings (SAS-like for dates when enabled)

• sample_disp: First non-missing value as display string (SAS-like for dates when enabled)

• attribute_names: Comma-separated attribute names (after exclusions)

• attributes: List column of attributes (optional)

Examples

df <- data.frame(
  USUBJID = c("01", "02", "03"),
  AGE     = c(45, 50, NA),
  TRTAN   = c(1L, 2L, 1L),
  ASTDT   = as.Date(c("2024-01-01", "2024-01-02", "2024-01-03")),
  stringsAsFactors = FALSE
)

get_column_info(df)

Load Data Files of Various Formats

Description

Loads one or more data files from a given directory. Supports multiple file types commonly used in clinical trials: .sas7bdat, .xpt, .csv, .xls, and .xlsx.

Usage

get_data(dir, file_names = NULL)

Arguments

Details

Automatically detects file extensions and returns each dataset using its base file name (e.g., "adsl.xpt" becomes adsl).

If multiple files with the same base name but different extensions exist (e.g., adsl.csv and adsl.sas7bdat), the function stops and reports the duplicates to avoid ambiguity.

Value

If exactly one file is loaded, returns the dataset. If multiple files are loaded, returns a named list of datasets.

Examples

## Not run: 

adsl <- get_data("path/to/adam", "adsl")

ds <- get_data("path/to/adam")

adsl <- ds$adsl


## End(Not run)

Summary Table: Mean and Related Statistics by Group

Description

This function calculates common summary statistics (N, Mean, SD, Median, Q1, Q3, Min, Max) for a numeric variable, grouped by a treatment or category variable. It supports optional SAS-style rounding (round half away from zero) and formats the results for table-ready display. Missing treatment groups are automatically added with zero values.

Usage

mean_by(
  data,
  group_var,
  uniq_var,
  label,
  sec_ord,
  precision_override = NULL,
  indent = 3,
  use_sas_round = FALSE,
  id_var = "USUBJID"
)

Arguments

Details

The function:

1. Auto-detects precision if precision_override is NULL.

2. Calculates N, mean, SD, quartiles, min, max.

3. Applies SAS-style rounding if use_sas_round = TRUE.

4. Converts statistics into a display format suitable for RTF or text output.

5. Ensures all treatment columns appear in output, filling missing ones with "0".

SAS-style rounding logic: Values exactly halfway between two increments are rounded away from zero (e.g., 1.25 → 1.3, -1.25 → -1.3 with 1 decimal place).

Value

A tibble with the following columns:

• stats : internal statistic code (n1, mn, sd, etc.)

• stat : display label (" N", " MEAN", etc.)

• sort_ord : row ordering number

• sec_ord : section ordering number (from input)

• Treatment columns (trt1, trt2, ...): formatted values per treatment group

Examples

library(dplyr)

df <- tibble::tibble(
  USUBJID = rep(1:6, each = 1),
  TRTAN   = c(1, 1, 2, 2, 3, 3),
  BMIBL   = c(25.1, 26.3, 24.8, NA, 23.4, 27.6)
)
mean_by(
  data          = df,
  group_var     = TRTAN,
  uniq_var      = BMIBL,
  label         = "BMI (kg/m^2)",
  sec_ord       = 1
)


mean_by(
  data          = df,
  group_var     = TRTAN,
  uniq_var      = BMIBL,
  label         = "BMI (kg/m^2)",
  sec_ord       = 1,
  precision_override = 2
)


mean_by(
  data          = df,
  group_var     = TRTAN,
  uniq_var      = BMIBL,
  label         = "BMI (kg/m^2)",
  sec_ord       = 1,
  use_sas_round = TRUE
)


df2 <- tibble::tibble(
  USUBJID = c(1, 2, 3, 4),
  TRTAN   = c(1, 1, 3, 3),
  BMIBL   = c(25.1, 26.3, 23.4, 27.6)
)

mean_by(
  data      = df2,
  group_var = TRTAN,
  uniq_var  = BMIBL,
  label     = "BMI (kg/m^2)",
  sec_ord   = 1
)

SAS-Compatible Rounding

Description

Performs rounding in the same manner as SAS, where values exactly halfway between two integers are always rounded away from zero. This differs from R's default rounding (IEC 60559), which rounds to the nearest even number ("bankers' rounding").

Usage

sas_round(x, digits = 0)

Arguments

Details

In SAS, values like 1.5 or -2.5 are rounded to 2 and -3 respectively. This function emulates that behavior by manually adjusting and checking the fractional component of the value before applying rounding.

Value

A numeric vector with values rounded using SAS-compatible logic.

Examples

sas_round(c(1.5, 2.5, 3.5, -1.5, -2.5, -3.5))

sas_round(c(1.25, 1.35, -1.25, -1.35), digits = 1)

sas_round(c(1.235, 1.245, -1.235, -1.245), digits = 2)

sas_round(c(1.2345, 1.2355), digits = 3)

sas_round(c(1.23445, 1.23455), digits = 4)

sas_round(c(1.234445, 1.234455), digits = 5)