Comparing Against Baselines or Control

Gabriel Becker and Adrian Waddell

2023-12-07

Introduction

Often the data from one column is considered the reference/baseline/comparison group and is compared to the data from the other columns.

For example, lets calculate the average age:

library(rtables)

lyt <- basic_table() %>%
  split_cols_by("ARM") %>%
  analyze("AGE")

tbl <- build_table(lyt, DM)
tbl
#        A: Drug X   B: Placebo   C: Combination
# ——————————————————————————————————————————————
# Mean     34.91       33.02          34.57

and then the difference of the average AGE between the placebo arm and the other arms:

lyt2 <- basic_table() %>%
  split_cols_by("ARM", ref_group = "B: Placebo") %>%
  analyze("AGE", afun = function(x, .ref_group) {
    in_rows(
      "Difference of Averages" = rcell(mean(x) - mean(.ref_group), format = "xx.xx")
    )
  })

tbl2 <- build_table(lyt2, DM)
tbl2
#                          A: Drug X   B: Placebo   C: Combination
# ————————————————————————————————————————————————————————————————
# Difference of Averages     1.89         0.00           1.55

Note that the column order has changed and the reference group is displayed in the first column.

In cases where we want cells to be blank in the reference column, (e.g., “B: Placebo”) we use non_ref_rcell() instead of rcell(), and pass .in_ref_col as the second argument:

lyt3 <- basic_table() %>%
  split_cols_by("ARM", ref_group = "B: Placebo") %>%
  analyze(
    "AGE",
    afun = function(x, .ref_group, .in_ref_col) {
      in_rows(
        "Difference of Averages" = non_ref_rcell(mean(x) - mean(.ref_group), is_ref = .in_ref_col, format = "xx.xx")
      )
    }
  )

tbl3 <- build_table(lyt3, DM)
tbl3
#                          A: Drug X   B: Placebo   C: Combination
# ————————————————————————————————————————————————————————————————
# Difference of Averages     1.89                        1.55
lyt4 <- basic_table() %>%
  split_cols_by("ARM", ref_group = "B: Placebo") %>%
  analyze(
    "AGE",
    afun = function(x, .ref_group, .in_ref_col) {
      in_rows(
        "Difference of Averages" = non_ref_rcell(mean(x) - mean(.ref_group), is_ref = .in_ref_col, format = "xx.xx"),
        "another row" = non_ref_rcell("aaa", .in_ref_col)
      )
    }
  )

tbl4 <- build_table(lyt4, DM)
tbl4
#                          A: Drug X   B: Placebo   C: Combination
# ————————————————————————————————————————————————————————————————
# Difference of Averages     1.89                        1.55     
# another row                 aaa                        aaa

You can see which arguments are available for afun in the manual for analyze().

Row Splitting

When adding row-splitting the reference data may be represented by the column with or without row splitting. For example:

lyt5 <- basic_table(show_colcounts = TRUE) %>%
  split_cols_by("ARM", ref_group = "B: Placebo") %>%
  split_rows_by("SEX", split_fun = drop_split_levels) %>%
  analyze("AGE", afun = function(x, .ref_group, .ref_full, .in_ref_col) {
    in_rows(
      "is reference (.in_ref_col)" = rcell(.in_ref_col),
      "ref cell N (.ref_group)" = rcell(length(.ref_group)),
      "ref column N (.ref_full)" = rcell(length(.ref_full))
    )
  })

tbl5 <- build_table(lyt5, subset(DM, SEX %in% c("M", "F")))
tbl5
#                                A: Drug X   B: Placebo   C: Combination
#                                 (N=121)     (N=106)        (N=129)    
# ——————————————————————————————————————————————————————————————————————
# F                                                                     
#   is reference (.in_ref_col)     FALSE        TRUE          FALSE     
#   ref cell N (.ref_group)         56           56             56      
#   ref column N (.ref_full)        106         106            106      
# M                                                                     
#   is reference (.in_ref_col)     FALSE        TRUE          FALSE     
#   ref cell N (.ref_group)         50           50             50      
#   ref column N (.ref_full)        106         106            106

The data assigned to .ref_full is the full data of the reference column whereas the data assigned to .ref_group respects the subsetting defined by row-splitting and hence is from the same subset as the argument x or df to afun.