Preparing analysis results in a tidy ARD format is a prerequisite for
a {tfrmt} table. A useful tool for ARD creation is the {cards}
package. By design, the output of {cards} is display-agnostic and only
contains information about the summaries themselves. Thus, the output
requires a bit of manipulation to get the ARD in a display-ready format.
For example, a cards object will label all denominators as
“N”, but {tfrmt} expects header (i.e. population-level) big Ns to have a
unique label that sets them apart from other denominators, like
“bigN”.
{tfrmt} contains several helper functions to ease the transition from
the cards output to a display-ready ARD fit for {tfrmt}.
Below are two examples utilizing these helpers for a demographics and AE
table.
Demog table
Demographics tables commonly contain a mix of categorical and continuous summaries by treatment group. We can perform these summaries using the following {cards} code:
ard_demog <- ard_stack(
data = pharmaverseadam::adsl |> dplyr::filter(SAFFL == "Y"),
ard_categorical(
variables = "AGEGR1",
statistic = ~ c("p")
),
ard_continuous(
variables = "AGE",
statistic = ~ continuous_summary_fns(c("mean","sd"))),
.by = "ARM"
)
ard_demog
#> {cards} data frame: 21 x 11
#> group1 group1_level variable variable_level stat_name stat_label stat
#> 1 ARM Placebo AGEGR1 >64 p % 0.837
#> 2 ARM Placebo AGEGR1 18-64 p % 0.163
#> 3 ARM Placebo AGE mean Mean 75.209
#> 4 ARM Placebo AGE sd SD 8.59
#> 5 ARM Xanomeli… AGEGR1 >64 p % 0.869
#> 6 ARM Xanomeli… AGEGR1 18-64 p % 0.131
#> 7 ARM Xanomeli… AGE mean Mean 74.381
#> 8 ARM Xanomeli… AGE sd SD 7.886
#> 9 ARM Xanomeli… AGEGR1 >64 p % 0.905
#> 10 ARM Xanomeli… AGEGR1 18-64 p % 0.095
#> ℹ 11 more rows
#> ℹ Use `print(n = ...)` to see more rows
#> ℹ 4 more variables: context, fmt_fun, warning, errorThe first step in preparing the data for {tfrmt} is calling
shuffle_card() which performs general operations to get the
results in a tidy data frame without list-columns and extra
metadata.
ard_demog_shuffled <- ard_demog |>
shuffle_card()
ard_demog_shuffled
#> # A tibble: 21 × 8
#> ARM AGEGR1 AGE context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo >64 NA catego… AGEGR1 p % 0.837
#> 2 Placebo 18-64 NA catego… AGEGR1 p % 0.163
#> 3 Placebo NA Over… contin… AGE mean Mean 75.2
#> 4 Placebo NA Over… contin… AGE sd SD 8.59
#> 5 Xanomeline H… >64 NA catego… AGEGR1 p % 0.869
#> 6 Xanomeline H… 18-64 NA catego… AGEGR1 p % 0.131
#> 7 Xanomeline H… NA Over… contin… AGE mean Mean 74.4
#> 8 Xanomeline H… NA Over… contin… AGE sd SD 7.89
#> 9 Xanomeline L… >64 NA catego… AGEGR1 p % 0.905
#> 10 Xanomeline L… 18-64 NA catego… AGEGR1 p % 0.0952
#> # ℹ 11 more rowsNote that the variables have been transformed wide and renamed. NULL
values have also been filled with “Overall
For our {tfrmt} table, AGEGR1 and AGE need
to be combined into a single variable. We can use the
prep_combine_vars() helper for this:
ard_demog_shuffled <- ard_demog_shuffled |>
prep_combine_vars(vars = c("AGEGR1","AGE"))
ard_demog_shuffled
#> # A tibble: 21 × 7
#> ARM variable_level context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo >64 catego… AGEGR1 p % 0.837
#> 2 Placebo 18-64 catego… AGEGR1 p % 0.163
#> 3 Placebo Overall AGE contin… AGE mean Mean 75.2
#> 4 Placebo Overall AGE contin… AGE sd SD 8.59
#> 5 Xanomeline… >64 catego… AGEGR1 p % 0.869
#> 6 Xanomeline… 18-64 catego… AGEGR1 p % 0.131
#> 7 Xanomeline… Overall AGE contin… AGE mean Mean 74.4
#> 8 Xanomeline… Overall AGE contin… AGE sd SD 7.89
#> 9 Xanomeline… >64 catego… AGEGR1 p % 0.905
#> 10 Xanomeline… 18-64 catego… AGEGR1 p % 0.0952
#> # ℹ 11 more rowsNext, we want to make a row label column that is the variable level
for categorical AGEGR1 and the stat label for continuous
AGE. We can use the prep_label() helper for
this:
ard_demog_shuffled <- ard_demog_shuffled |>
prep_label()
ard_demog_shuffled
#> # A tibble: 21 × 8
#> ARM variable_level context stat_variable stat_name stat_label stat label
#> <chr> <chr> <chr> <chr> <chr> <chr> <dbl> <chr>
#> 1 Plac… >64 catego… AGEGR1 p % 0.837 >64
#> 2 Plac… 18-64 catego… AGEGR1 p % 0.163 18-64
#> 3 Plac… Overall AGE contin… AGE mean Mean 75.2 Mean
#> 4 Plac… Overall AGE contin… AGE sd SD 8.59 SD
#> 5 Xano… >64 catego… AGEGR1 p % 0.869 >64
#> 6 Xano… 18-64 catego… AGEGR1 p % 0.131 18-64
#> 7 Xano… Overall AGE contin… AGE mean Mean 74.4 Mean
#> 8 Xano… Overall AGE contin… AGE sd SD 7.89 SD
#> 9 Xano… >64 catego… AGEGR1 p % 0.905 >64
#> 10 Xano… 18-64 catego… AGEGR1 p % 0.0952 18-64
#> # ℹ 11 more rowsNote that if we take a peek at our stats for the treatment groups, we
have n, N, p from the
ard_stack() call:
tail(ard_demog_shuffled)
#> # A tibble: 6 × 8
#> ARM variable_level context stat_variable stat_name stat_label stat label
#> <chr> <chr> <chr> <chr> <chr> <chr> <dbl> <chr>
#> 1 Xanom… NA tabula… ARM n n 84 NA
#> 2 Xanom… NA tabula… ARM N N 254 NA
#> 3 Xanom… NA tabula… ARM p % 0.331 NA
#> 4 Xanom… NA tabula… ARM n n 84 NA
#> 5 Xanom… NA tabula… ARM N N 254 NA
#> 6 Xanom… NA tabula… ARM p % 0.331 NACommonly we only want to display the n in the column
headers for treatment group. To distinguish these as big N’s, we need to
relabel the stat, and drop the unnecessary N and
p. The prep_big_n() helper takes care of this
for us:
ard_demog_shuffled <- ard_demog_shuffled |>
prep_big_n(vars = "ARM") # the grouping variable we want to prepare big N's for
ard_demog_shuffled
#> # A tibble: 15 × 8
#> ARM variable_level context stat_variable stat_name stat_label stat label
#> <chr> <chr> <chr> <chr> <chr> <chr> <dbl> <chr>
#> 1 Plac… >64 catego… AGEGR1 p % 0.837 >64
#> 2 Plac… 18-64 catego… AGEGR1 p % 0.163 18-64
#> 3 Plac… Overall AGE contin… AGE mean Mean 75.2 Mean
#> 4 Plac… Overall AGE contin… AGE sd SD 8.59 SD
#> 5 Xano… >64 catego… AGEGR1 p % 0.869 >64
#> 6 Xano… 18-64 catego… AGEGR1 p % 0.131 18-64
#> 7 Xano… Overall AGE contin… AGE mean Mean 74.4 Mean
#> 8 Xano… Overall AGE contin… AGE sd SD 7.89 SD
#> 9 Xano… >64 catego… AGEGR1 p % 0.905 >64
#> 10 Xano… 18-64 catego… AGEGR1 p % 0.0952 18-64
#> 11 Xano… Overall AGE contin… AGE mean Mean 75.7 Mean
#> 12 Xano… Overall AGE contin… AGE sd SD 8.29 SD
#> 13 Plac… NA tabula… ARM bigN n 86 NA
#> 14 Xano… NA tabula… ARM bigN n 84 NA
#> 15 Xano… NA tabula… ARM bigN n 84 NAFinally, we can do any additional table-specific manipulations and we’re ready for our tfrmt!
ard_final <- ard_demog_shuffled |>
# drop unneeded variables
dplyr::select(-c(variable_level, context, stat_label)) |>
# add sorting
dplyr::mutate(ord1 = dplyr::case_when(
label == "18-64" ~ 1,
label == ">64" ~ 2,
TRUE ~ 3
))
ard_final
#> # A tibble: 15 × 6
#> ARM stat_variable stat_name stat label ord1
#> <chr> <chr> <chr> <dbl> <chr> <dbl>
#> 1 Placebo AGEGR1 p 0.837 >64 2
#> 2 Placebo AGEGR1 p 0.163 18-64 1
#> 3 Placebo AGE mean 75.2 Mean 3
#> 4 Placebo AGE sd 8.59 SD 3
#> 5 Xanomeline High Dose AGEGR1 p 0.869 >64 2
#> 6 Xanomeline High Dose AGEGR1 p 0.131 18-64 1
#> 7 Xanomeline High Dose AGE mean 74.4 Mean 3
#> 8 Xanomeline High Dose AGE sd 7.89 SD 3
#> 9 Xanomeline Low Dose AGEGR1 p 0.905 >64 2
#> 10 Xanomeline Low Dose AGEGR1 p 0.0952 18-64 1
#> 11 Xanomeline Low Dose AGE mean 75.7 Mean 3
#> 12 Xanomeline Low Dose AGE sd 8.29 SD 3
#> 13 Placebo ARM bigN 86 NA 3
#> 14 Xanomeline High Dose ARM bigN 84 NA 3
#> 15 Xanomeline Low Dose ARM bigN 84 NA 3Create the {tfrmt}
tfrmt(
group = stat_variable,
label = label ,
param = stat_name,
value = stat,
column = ARM,
sorting_cols = ord1,
body_plan = body_plan(
frmt_structure(
group_val = ".default",
label_val = ".default",
frmt("xx.x")
),
frmt_structure(
group_val = ".default",
label_val = ".default",
p = frmt("xx%", transform = ~.*100)
)
),
big_n = big_n_structure(param_val = "bigN"),
col_plan = col_plan(-ord1)
) |>
print_to_gt(ard_final)| Placebo N = 86 | Xanomeline High Dose N = 84 | Xanomeline Low Dose N = 84 | |
|---|---|---|---|
| AGEGR1 | |||
| 18-64 | 16% | 13% | 10% |
| >64 | 84% | 87% | 90% |
| AGE | |||
| Mean | 75.2 | 74.4 | 75.7 |
| SD | 8.6 | 7.9 | 8.3 |
AE table
AE tables contain hierarchical summaries. We can perform these summaries using the following {cards} code:
adae <- pharmaverseadam::adae |>
dplyr::filter(SAFFL=="Y") |>
dplyr::filter(AESOC %in% unique(AESOC)[1:2])|>
dplyr::group_by(AESOC) |>
dplyr::filter(AEDECOD %in% unique(AEDECOD)[1:3]) |>
dplyr::ungroup()
ard_ae <- ard_stack_hierarchical(
data = adae,
variables = c(AESOC, AEDECOD),
by = ARM,
denominator = pharmaverseadam::adsl |> dplyr::filter(SAFFL=="Y"),
id = USUBJID,
statistic = ~ c("n","p")
)
ard_ae
#> {cards} data frame: 57 x 13
#> group1 group1_level group2 group2_level variable variable_level stat_name
#> 1 <NA> <NA> ARM Placebo n
#> 2 <NA> <NA> ARM Placebo N
#> 3 <NA> <NA> ARM Placebo p
#> 4 <NA> <NA> ARM Xanomeli… n
#> 5 <NA> <NA> ARM Xanomeli… N
#> 6 <NA> <NA> ARM Xanomeli… p
#> 7 <NA> <NA> ARM Xanomeli… n
#> 8 <NA> <NA> ARM Xanomeli… N
#> 9 <NA> <NA> ARM Xanomeli… p
#> 10 ARM Placebo <NA> AESOC GASTROIN… n
#> stat_label stat
#> 1 n 86
#> 2 N 254
#> 3 % 0.339
#> 4 n 84
#> 5 N 254
#> 6 % 0.331
#> 7 n 84
#> 8 N 254
#> 9 % 0.331
#> 10 n 12
#> ℹ 47 more rows
#> ℹ Use `print(n = ...)` to see more rows
#> ℹ 4 more variables: context, fmt_fun, warning, errorLike the demog table example, the first step in preparing the data
for {tfrmt} is calling shuffle_card() which performs
general operations to get the results in a tidy data frame without
list-columns and extra metadata.
ard_ae_shuffled <- ard_ae |>
shuffle_card()
ard_ae_shuffled
#> # A tibble: 57 × 8
#> ARM AESOC AEDECOD context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo NA NA tabula… ARM n n 86
#> 2 Placebo NA NA tabula… ARM N N 254
#> 3 Placebo NA NA tabula… ARM p % 0.339
#> 4 Xanomeline … NA NA tabula… ARM n n 84
#> 5 Xanomeline … NA NA tabula… ARM N N 254
#> 6 Xanomeline … NA NA tabula… ARM p % 0.331
#> 7 Xanomeline … NA NA tabula… ARM n n 84
#> 8 Xanomeline … NA NA tabula… ARM N N 254
#> 9 Xanomeline … NA NA tabula… ARM p % 0.331
#> 10 Placebo GAST… NA hierar… AESOC n n 12
#> # ℹ 47 more rowsNotice that our calculations of any AE by system organ class (AESOC) have a missing value for preferred term (AEDECOD):
ard_ae_shuffled |>
dplyr::filter(!is.na(AESOC) & is.na(AEDECOD))
#> # A tibble: 12 × 8
#> ARM AESOC AEDECOD context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo GAST… NA hierar… AESOC n n 12
#> 2 Placebo GAST… NA hierar… AESOC p % 0.140
#> 3 Placebo GENE… NA hierar… AESOC n n 9
#> 4 Placebo GENE… NA hierar… AESOC p % 0.105
#> 5 Xanomeline … GAST… NA hierar… AESOC n n 10
#> 6 Xanomeline … GAST… NA hierar… AESOC p % 0.119
#> 7 Xanomeline … GENE… NA hierar… AESOC n n 28
#> 8 Xanomeline … GENE… NA hierar… AESOC p % 0.333
#> 9 Xanomeline … GAST… NA hierar… AESOC n n 8
#> 10 Xanomeline … GAST… NA hierar… AESOC p % 0.0952
#> 11 Xanomeline … GENE… NA hierar… AESOC n n 27
#> 12 Xanomeline … GENE… NA hierar… AESOC p % 0.321We need to populate these missing values for our table. In order to
display these as group summaries, the value for AEDECOD must match the
value for AESOC. We can use the prep_hierarchical_fill()
helper to fill the value from AESOC into AEDECOD:
ard_ae_shuffled <- ard_ae_shuffled |>
prep_hierarchical_fill(
vars = c("AESOC","AEDECOD"),
fill_from_left = TRUE
)
ard_ae_shuffled |>
dplyr::filter(AESOC == AEDECOD)
#> # A tibble: 12 × 8
#> ARM AESOC AEDECOD context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo GAST… GASTRO… hierar… AESOC n n 12
#> 2 Placebo GAST… GASTRO… hierar… AESOC p % 0.140
#> 3 Placebo GENE… GENERA… hierar… AESOC n n 9
#> 4 Placebo GENE… GENERA… hierar… AESOC p % 0.105
#> 5 Xanomeline … GAST… GASTRO… hierar… AESOC n n 10
#> 6 Xanomeline … GAST… GASTRO… hierar… AESOC p % 0.119
#> 7 Xanomeline … GENE… GENERA… hierar… AESOC n n 28
#> 8 Xanomeline … GENE… GENERA… hierar… AESOC p % 0.333
#> 9 Xanomeline … GAST… GASTRO… hierar… AESOC n n 8
#> 10 Xanomeline … GAST… GASTRO… hierar… AESOC p % 0.0952
#> 11 Xanomeline … GENE… GENERA… hierar… AESOC n n 27
#> 12 Xanomeline … GENE… GENERA… hierar… AESOC p % 0.321Note that if we take a peek at our stats for the treatment groups, we
have n, N, p from the
ard_stack() call:
ard_ae_shuffled |>
dplyr::filter(is.na(AESOC))
#> # A tibble: 9 × 8
#> ARM AESOC AEDECOD context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo NA NA tabula… ARM n n 86
#> 2 Placebo NA NA tabula… ARM N N 254
#> 3 Placebo NA NA tabula… ARM p % 0.339
#> 4 Xanomeline H… NA NA tabula… ARM n n 84
#> 5 Xanomeline H… NA NA tabula… ARM N N 254
#> 6 Xanomeline H… NA NA tabula… ARM p % 0.331
#> 7 Xanomeline L… NA NA tabula… ARM n n 84
#> 8 Xanomeline L… NA NA tabula… ARM N N 254
#> 9 Xanomeline L… NA NA tabula… ARM p % 0.331Commonly we only want to display the n in the column
headers for treatment group. To distinguish these as big N’s, we need to
relabel the stat, and drop the unnecessary N and
p. The prep_big_n() helper takes care of this
for us:
ard_ae_shuffled <- ard_ae_shuffled |>
prep_big_n(vars = "ARM") # the grouping variable we want to prepare big N's for
ard_ae_shuffled
#> # A tibble: 51 × 8
#> ARM AESOC AEDECOD context stat_variable stat_name stat_label stat
#> <chr> <chr> <chr> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo NA NA tabula… ARM bigN n 86
#> 2 Xanomeline H… NA NA tabula… ARM bigN n 84
#> 3 Xanomeline L… NA NA tabula… ARM bigN n 84
#> 4 Placebo GAST… GASTRO… hierar… AESOC n n 12
#> 5 Placebo GAST… GASTRO… hierar… AESOC p % 0.140
#> 6 Placebo GENE… GENERA… hierar… AESOC n n 9
#> 7 Placebo GENE… GENERA… hierar… AESOC p % 0.105
#> 8 Placebo GAST… DIARRH… hierar… AEDECOD n n 9
#> 9 Placebo GAST… DIARRH… hierar… AEDECOD p % 0.105
#> 10 Placebo GAST… HIATUS… hierar… AEDECOD n n 1
#> # ℹ 41 more rowsFinally, we can do any additional table-specific manipulations and we’re ready for our tfrmt!
ard_final <- ard_ae_shuffled |>
# drop uneeded variables
dplyr::select(-c(context, stat_label, stat_variable))
ard_final
#> # A tibble: 51 × 5
#> ARM AESOC AEDECOD stat_name stat
#> <chr> <chr> <chr> <chr> <dbl>
#> 1 Placebo NA NA bigN 86
#> 2 Xanomeline High Dose NA NA bigN 84
#> 3 Xanomeline Low Dose NA NA bigN 84
#> 4 Placebo GASTROINTESTINAL DISORDERS GASTRO… n 12
#> 5 Placebo GASTROINTESTINAL DISORDERS GASTRO… p 0.140
#> 6 Placebo GENERAL DISORDERS AND ADMINIST… GENERA… n 9
#> 7 Placebo GENERAL DISORDERS AND ADMINIST… GENERA… p 0.105
#> 8 Placebo GASTROINTESTINAL DISORDERS DIARRH… n 9
#> 9 Placebo GASTROINTESTINAL DISORDERS DIARRH… p 0.105
#> 10 Placebo GASTROINTESTINAL DISORDERS HIATUS… n 1
#> # ℹ 41 more rowsCreate the {tfrmt}
tfrmt(
group = AESOC,
label = AEDECOD,
param = stat_name,
value = stat,
column = ARM,
body_plan = body_plan(
frmt_structure(
group_val = ".default",
label_val = ".default",
frmt_combine(
"{n} ({p}%)",
n = frmt("xx"),
p = frmt("xx", transform = ~ . *100)
)
)
),
big_n = big_n_structure(
param_val = "bigN"
)
) |>
print_to_gt(ard_final)| Placebo N = 86 | Xanomeline High Dose N = 84 | Xanomeline Low Dose N = 84 | |
|---|---|---|---|
| GASTROINTESTINAL DISORDERS | 12 (14%) | 10 (12%) | 8 (10%) |
| DIARRHOEA | 9 (10%) | 4 ( 5%) | 5 ( 6%) |
| HIATUS HERNIA | 1 ( 1%) | 0 ( 0%) | 0 ( 0%) |
| VOMITING | 3 ( 3%) | 7 ( 8%) | 3 ( 4%) |
| GENERAL DISORDERS AND ADMINISTRATION SITE CONDITIONS | 9 (10%) | 28 (33%) | 27 (32%) |
| APPLICATION SITE ERYTHEMA | 3 ( 3%) | 15 (18%) | 12 (14%) |
| APPLICATION SITE PRURITUS | 6 ( 7%) | 22 (26%) | 22 (26%) |
| FATIGUE | 1 ( 1%) | 5 ( 6%) | 5 ( 6%) |