tlf-ae-spec.Rmd

# Specific AE

Following [ICH E3 guidance](https://database.ich.org/sites/default/files/E3_Guideline.pdf),
we need to summarize which participants were included in each efficacy
analysis in Section 12.2, Adverse Events (AEs).

```{r}
library(haven) # Read SAS data
library(dplyr) # Manipulate data
library(tidyr) # Manipulate data
library(r2rtf) # Reporting in RTF format
```

In this chapter, we illustrate how to summarize simplified specific AE
information in a study.

```{r, out.width = "100%", out.height = "400px", echo = FALSE, fig.align = "center"}
knitr::include_graphics("tlf/tlf_spec_ae.pdf")
```

The data used to summarize AE information is in `adsl` and `adae` datasets.

```{r}
adsl <- read_sas("adam_data/adsl.sas7bdat")
adae <- read_sas("adam_data/adae.sas7bdat")
```

For illustration purposes, we only provide count in the simplified
table. The percentage of participants for each AE criteria can be
calculated as in Chapter \@ref(aesummary).

In this way, let's focus on the analysis script for two advanced
features for a table layout.

- group content: AE can be summarized in multiple nested layers.
  (e.g., by system organ class (SOC, `AESOC`) and specific AE term
  (`AEDECOD`))
- pagenization: there are many AE terms that can not be covered in one
  page. Column headers and SOC information need to be repeated on
  every page.

In the code below, we count the number of subjects in each AE term by
SOC and treatment group, and we create a new variable `order` and set it
as `0` for these counted numbers. The variable `order` will help with
the data manipulation later.

```{r}
ana <- adae %>%
  mutate(
    AESOC = toTitleCase(tolower(AESOC)),
    AEDECOD = toTitleCase(tolower(AEDECOD))
  )

t1 <- ana %>%
  group_by(TRTAN, AESOC) %>%
  summarise(n = fmt_num(n_distinct(USUBJID), digits = 0)) %>%
  mutate(AEDECOD = AESOC, order = 0)

t1 %>% head(4)
```

In the code below, we count the number of subjects in each AE term by
SOC, AE term, and treatment group. Here we also create a new variable
`order` and set it as `1` for these counted numbers.

```{r}
t2 <- ana %>%
  group_by(TRTAN, AESOC, AEDECOD) %>%
  summarise(n = fmt_num(n_distinct(USUBJID), digits = 0)) %>%
  mutate(order = 1)

t2 %>% head(4)
```

We prepare reporting data for AE information.

```{r}
t_ae <- bind_rows(t1, t2) %>%
  pivot_wider(
    id_cols = c(AESOC, order, AEDECOD),
    names_from = TRTAN,
    names_prefix = "n_",
    values_from = n,
    values_fill = fmt_num(0, digits = 0)
  ) %>%
  arrange(AESOC, order, AEDECOD) %>%
  select(AESOC, AEDECOD, starts_with("n"))

t_ae %>% head(4)
```

We prepare reporting data for analysis population.

```{r}
t_pop <- adsl %>%
  filter(SAFFL == "Y") %>%
  count_by("TRT01AN", "SAFFL",
    var_label = "Participants in population"
  ) %>%
  mutate(
    AESOC = "pop",
    AEDECOD = var_label
  ) %>%
  select(AESOC, AEDECOD, starts_with("n_"))

t_pop
```

The final report data is saved in `tbl_ae_spec`.
We also add a blank row between population and AE information in the reporting table.

```{r}
tbl_ae_spec <- bind_rows(
  t_pop,
  data.frame(AESOC = "pop"),
  t_ae
) %>%
  mutate(AEDECOD = ifelse(AEDECOD == AESOC,
    AEDECOD, paste0("  ", AEDECOD)
  ))

tbl_ae_spec %>% head(4)
```

We start to define the format of the output.

To obtain the nested layout, we use the `page_by` argument in the
`rtf_body` function. By defining `page_by="AESOC"`, `r2rtf` recognize
the variable as a group indicator.

After setting `pageby_row = "first_row"`, the first row is displayed as
group header. If a group of information is broken into multiple pages,
the group header row is repeated on each page by default.

We can also customize the text format by providing a matrix that has the
same dimension as the input dataset (i.e., `tbl_ae_spec`). In the code
below, we illustrate how to display **bold** text for group headers to
highlight the nested structure of the table layout.

```{r}
n_row <- nrow(tbl_ae_spec)
n_col <- ncol(tbl_ae_spec)
id <- tbl_ae_spec$AESOC == tbl_ae_spec$AEDECOD
id <- ifelse(is.na(id), FALSE, id)

text_format <- ifelse(id, "b", "")
```

More discussion on `page_by`, `group_by` and `subline_by`
features can be found in the
[`r2rtf` package website](https://merck.github.io/r2rtf/articles/example-sublineby-pageby-groupby.html.

```{r}
tbl_ae_spec %>%
  rtf_title(
    "Analysis of Participants With Specific Adverse Events",
    "(Safety Analysis Population)"
  ) %>%
  rtf_colheader(" | Placebo | Xanomeline Low Dose| Xanomeline High Dose",
    col_rel_width = c(3, rep(1, 3))
  ) %>%
  rtf_colheader(" | n |  n | n ",
    border_top = "",
    border_bottom = "single",
    col_rel_width = c(3, rep(1, 3))
  ) %>%
  rtf_body(
    col_rel_width = c(1, 3, rep(1, 3)),
    text_justification = c("l", "l", rep("c", 3)),
    text_format = matrix(text_format, nrow = n_row, ncol = n_col),
    page_by = "AESOC",
    pageby_row = "first_row"
  ) %>%
  rtf_footnote("Every subject is counted a single time for each applicable row and column.") %>%
  rtf_encode() %>%
  write_rtf("tlf/tlf_spec_ae.rtf")
```

```{r, out.width = "100%", out.height = "400px", echo = FALSE, fig.align = "center"}
knitr::include_graphics("tlf/tlf_spec_ae.pdf")
```

The procedure to generate a baseline characteristics table can be
summarized as follow:

- Step 1: Read data into R, i.e., `adae` and `adsl`.
- Step 2: Count the number of subjects by SOC and treatment group
  (rows with bold text) and save into `t1`.
- Step 3: Count the number of subjects in each AE term by SOC, AE
  term, and treatment group (rows without bold text) and save into `t2`.
- Step 4: Bind `t1` and `t2` by row into `t_ae`.
- Step 5: Count the number of subjects in each arm as `t_pop`.
- Step 6: Bind `t_pop` and `t_ae` by row into `tbl_ae_spec`.
- Step 7: Format the output by `r2rtf`.