# load libraries
library(cards)AET02 - Adverse Events
Adverse Events (AET02) QC Workflow
1. Generate a table using {chevron}
Show the code
tlg_aet02 <- chevron::aet02_main(chevron::syn_data, arm_var = "ARM", lbl_overall = "Overall ARM")
head(tlg_aet02, n = 15) A: Drug X B: Placebo C: Combination Overall ARM
(N=15) (N=15) (N=15) (N=45)
———————————————————————————————————————————————————————————————————————————————————————————————————————————————————
Total number of patients with at least one adverse event 13 (86.7%) 14 (93.3%) 15 (100%) 42 (93.3%)
Overall total number of events 58 59 99 216
cl A.1
Total number of patients with at least one adverse event 7 (46.7%) 6 (40.0%) 10 (66.7%) 23 (51.1%)
Total number of events 8 11 16 35
dcd A.1.1.1.1 3 (20.0%) 1 (6.7%) 6 (40.0%) 10 (22.2%)
dcd A.1.1.1.2 5 (33.3%) 6 (40.0%) 6 (40.0%) 17 (37.8%)
cl B.1
Total number of patients with at least one adverse event 5 (33.3%) 6 (40.0%) 8 (53.3%) 19 (42.2%)
Total number of events 6 6 12 24
dcd B.1.1.1.1 5 (33.3%) 6 (40.0%) 8 (53.3%) 19 (42.2%)
cl B.2
Total number of patients with at least one adverse event 11 (73.3%) 8 (53.3%) 10 (66.7%) 29 (64.4%)
Total number of events 18 15 20 53
dcd B.2.1.2.1 5 (33.3%) 6 (40.0%) 5 (33.3%) 16 (35.6%) 2. Flatten the table into a data.frame
A {rtables} based output can be flattened into a data.frame using the as_results_df() function from the {rtables} package. The make_ard argument set to TRUE, will format the data similar to the output generated by the {cards} package. Setting the add_tbl_str_decimals to FALSE will not return a column with the statistic as a formatted string. We also include a step to remove the “label” attribute for the statistics. The diffdf() function is sensitive to attribute mismatch, so we will remove them (set to NULL) to match the ARD results exactly.
rtables_result <- rtables::as_result_df(tlg_aet02, make_ard = TRUE, add_tbl_str_decimals = FALSE)
attr(rtables_result$stat, "label") <- NULL
rtables_result[1:6, c("group2", "group2_level", "variable", "variable_level", "stat")]| group2 | group2_level | variable | variable_level | stat |
|---|---|---|---|---|
| ARM | A: Drug X | USUBJID | unique | 13.0000000 |
| ARM | A: Drug X | USUBJID | unique | 0.8666667 |
| ARM | A: Drug X | USUBJID | nonunique | 58.0000000 |
| ARM | A: Drug X | AEBODSYS | unique | 7.0000000 |
| ARM | A: Drug X | AEBODSYS | unique | 0.4666667 |
| ARM | A: Drug X | AEBODSYS | nonunique | 8.0000000 |
3. Create a comparable ARD
In the code below, we perform different data pre-processing for the different ARDs we want to create. Then, we generate separate ARDs using the {cards} package for the different sections of the AET02 table and compare them as subsets. Note: If your table does not have an “overall” column, you can use the ARD creation steps at the bottom of the page using primarily ard_categorical(). ard_hierarchical() is useful here to calculate the statistics for the overall column and allows for better filtering using the “context” column.
# data pre-processing. filter observations with ANL01FL = "Y"
adae <- chevron::syn_data$adae |>
dplyr::filter(ANL01FL == "Y")
adsl <- chevron::syn_data$adsl |>
dplyr::filter(ANL01FL == "Y")
# ----- ARDS -----
ard_result <- ard_stack_hierarchical(
data = adae,
variables = c(ANL01FL, AEBODSYS, AEDECOD),
by = c(ARM),
denominator = adsl,
overall = TRUE,
id = USUBJID,
statistic = ~ c("n", "p")
) |>
apply_fmt_fn() |>
unlist_ard_columns() |>
dplyr::filter(context == "hierarchical")
ard_count_result <-
ard_stack_hierarchical_count(
data = adae,
variables = c(ANL01FL, AEBODSYS),
by = ARM,
overall = TRUE,
denominator = ADSL
) |>
apply_fmt_fn() |>
unlist_ard_columns() |>
dplyr::filter(context == "hierarchical_count")4. Statistics Comparison
We’ll modify the rtables result to closely match the ARD result:
- Rename “group2_level” to “ARM” to match the ARD.
- Remove any columns that aren’t in the ARD result (stat_name is being removed as it is a blank column)
rtables_result <- rtables_result |>
dplyr::rename(
group1_level = group2_level,
group1 = group2,
group2 = group1,
group2_level = group1_level
)
head(rtables_result, n = 10)| group2 | group2_level | group1 | group1_level | variable | variable_level | variable_label | stat_name | stat |
|---|---|---|---|---|---|---|---|---|
| NA | NA | ARM | A: Drug X | USUBJID | unique | Total number of patients with at least one adverse event | NA | 13.0000000 |
| NA | NA | ARM | A: Drug X | USUBJID | unique | Total number of patients with at least one adverse event | NA | 0.8666667 |
| NA | NA | ARM | A: Drug X | USUBJID | nonunique | Overall total number of events | NA | 58.0000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEBODSYS | unique | Total number of patients with at least one adverse event | NA | 7.0000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEBODSYS | unique | Total number of patients with at least one adverse event | NA | 0.4666667 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEBODSYS | nonunique | Total number of events | NA | 8.0000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEDECOD | dcd A.1.1.1.1 | dcd A.1.1.1.1 | NA | 3.0000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEDECOD | dcd A.1.1.1.1 | dcd A.1.1.1.1 | NA | 0.2000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEDECOD | dcd A.1.1.1.2 | dcd A.1.1.1.2 | NA | 5.0000000 |
| AEBODSYS | cl A.1 | ARM | A: Drug X | AEDECOD | dcd A.1.1.1.2 | dcd A.1.1.1.2 | NA | 0.3333333 |
We can compare the data in subsets to minimize reformatting steps
# subgroup analysis
rtables_result_sub1 <- rtables_result |>
dplyr::filter(variable == "AEDECOD") |>
dplyr::select(c("group1", "group1_level", "variable", "variable_level", "stat"))
ard_sub1 <- ard_result |>
dplyr::filter(variable == "AEDECOD") |>
dplyr::mutate(
group1 = dplyr::case_when(
group1 == "ANL01FL" ~ "ARM",
TRUE ~ group1
),
group1_level = dplyr::case_when(
group1_level == "Y" ~ "Overall ARM",
TRUE ~ group1_level
)
) |>
dplyr::select(c("group1", "group1_level", "variable", "variable_level", "stat"))
diffdf::diffdf(rtables_result_sub1, ard_sub1, keys = c("group1", "group1_level", "variable", "variable_level", "stat"))No issues were found!Compare the second subset
rtables_result_sub2 <- rtables_result |>
dplyr::filter(variable %in% c("USUBJID", "AEBODSYS") & variable_level != "nonunique") |>
dplyr::select(c("group1", "group1_level", "group2", "group2_level", "stat")) |>
dplyr::rename(
variable = group2,
variable_level = group2_level
) |>
dplyr::mutate(
variable = dplyr::coalesce(variable, "ANL01FL"),
variable_level = dplyr::coalesce(variable_level, "Y")
)
ard_sub2 <- ard_result |>
dplyr::filter(variable %in% c("ANL01FL", "AEBODSYS")) |>
dplyr::mutate(
group1 = dplyr::recode(group1, "ANL01FL" = "ARM"),
group1_level = dplyr::recode(group1_level, "Y" = "Overall ARM"),
group1 = dplyr::coalesce(group1, "ARM"),
group1_level = dplyr::coalesce(group1_level, "Overall ARM")
) |>
dplyr::select(c("group1", "group1_level", "variable", "variable_level", "stat"))
diffdf::diffdf(rtables_result_sub2, ard_sub2, keys = c("group1", "group1_level", "variable", "variable_level", "stat"))No issues were found!Total Number of Events (ARD counts data.frame)
# reformat rtables
rtables_count <- rtables_result |>
dplyr::filter(variable_label %in% c("Total number of events", "Overall total number of events")) |>
dplyr::mutate(
group2 = dplyr::case_when(
variable_label == "Overall total number of events" ~ "ANL01FL",
TRUE ~ group2
),
group2_level = dplyr::case_when(
variable_label == "Overall total number of events" ~ "Y",
TRUE ~ group2_level
)
) |>
dplyr::select("group1", "group1_level", "group2", "group2_level", "stat")
# reformat ARD result
ard_count_result <- ard_count_result |>
dplyr::mutate(
group1 = dplyr::coalesce(group1, "ARM"),
group1_level = dplyr::coalesce(group1_level, "Overall ARM"),
group1 = dplyr::case_when(
group1 == "ANL01FL" ~ "ARM",
TRUE ~ group1
),
group1_level = dplyr::case_when(
group1_level == "Y" ~ "Overall ARM",
TRUE ~ group1_level
)
) |>
dplyr::select("group1", "group1_level", "variable", "variable_level", "stat") |>
dplyr::rename(
group2 = variable,
group2_level = variable_level
)
diffdf::diffdf(rtables_count, ard_count_result, keys = c("group1", "group1_level", "group2", "group2_level", "stat"))No issues were found!