Response Forest Graphs Within Treatment Arms by Continuous Biomarker Cutoff
For response endpoints it is good to show how to obtain within-treatment-arms comparisons of biomarker subgroups. This is similar to SFG04.
Similarly like in RFG1.
library(tern)
library(ggplot2.utils)
library(dplyr)
adrs <- random.cdisc.data::cadrs %>%
df_explicit_na() %>%
filter(PARAMCD == "BESRSPI", BMEASIFL == "Y") %>%
mutate(
is_rsp = AVALC %in% c("CR", "PR"),
ARM_BIN = fct_collapse_only(
ARM,
CTRL = c("B: Placebo"),
TRT = c("A: Drug X", "C: Combination")
),
BMRKR2 = fct_explicit_na_if(BMRKR2, BEP01FL == "N")
) %>%
var_relabel(
BEP01FL = "BEP",
BMRKR2 = "Biomarker (Categorical)"
)We define a vector of all cutpoints to use for a numeric biomarker (here BMRKR1).
We lapply() over this vector, each time generating a binary factor variable BMRKR1_cut and then tabulating the resulting statistics similar to SFG4, this time including the treatment arms in the subgroups argument. Then we rbind() all tables in the list together.
# across arm ----
cutpoints <- c(4, 5, 8)
adrs2 <- adrs %>%
mutate(
BMRKR1 = ifelse(BEP01FL == "N", NA, BMRKR1),
BMRKR1_cut = explicit_na(cut(BMRKR1, c(-Inf, cutpoints, Inf), right = FALSE))
) %>%
var_relabel(BMRKR1_cut = "Biomarker (Binned Continuous)")
tables_all_cutpoints <- lapply(cutpoints, function(cutpoint) {
adrs3 <- adrs2 %>%
filter(!is.na(BMRKR1)) %>%
mutate(
BMRKR1_thresh = explicit_na(factor(
ifelse(BMRKR1 > cutpoint, "Greater", "Less")
))
)
tbl <- extract_rsp_subgroups(
variables = list(
rsp = "is_rsp",
arm = "BMRKR1_thresh",
subgroups = c("ARM_BIN")
),
label_all = paste0("BMRKR1 (", cutpoint, ")"),
data = adrs3
)
basic_table() %>%
tabulate_rsp_subgroups(
df = tbl,
vars = c("n_tot", "n", "n_rsp", "prop", "or", "ci")
)
})
result <- do.call(rbind, tables_all_cutpoints)We can look at the result in the console already.
Greater Less
Baseline Risk Factors Total n n Responders Response (%) n Responders Response (%) Odds Ratio 95% CI
—————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————————
BMRKR1 (4) 97 62 61 98.4% 35 34 97.1% 0.56 (0.03, 9.20)
Description of Planned Arm
CTRL 36 21 21 100.0% 15 15 100.0% 1.00 (0.00, >999.99)
TRT 61 41 40 97.6% 20 19 95.0% 0.48 (0.03, 8.01)
BMRKR1 (5) 97 50 49 98.0% 47 46 97.9% 0.94 (0.06, 15.45)
Description of Planned Arm
CTRL 36 16 16 100.0% 20 20 100.0% 1.00 (0.00, >999.99)
TRT 61 34 33 97.1% 27 26 96.3% 0.79 (0.05, 13.21)
BMRKR1 (8) 97 16 16 100.0% 81 79 97.5% <0.01 (0.00, >999.99)
Description of Planned Arm
CTRL 36 3 3 100.0% 33 33 100.0% 1.00 (0.00, >999.99)
TRT 61 13 13 100.0% 48 46 95.8% <0.01 (0.00, >999.99)We can now produce the forest plot using the g_forest() function. Similarly as in SFG4 we need to specify the col_x, col_y and forest_header arguments for g_forest() by recovering them from one of the original tables.
With gridExtra::grid.arrange() the plot can be combined e.g. with a between-treatment-arms comparison, like RFG1A
R version 4.4.2 (2024-10-31)
Platform: x86_64-pc-linux-gnu
Running under: Ubuntu 24.04.1 LTS
Matrix products: default
BLAS: /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.26.so; LAPACK version 3.12.0
locale:
[1] LC_CTYPE=en_US.UTF-8 LC_NUMERIC=C
[3] LC_TIME=en_US.UTF-8 LC_COLLATE=en_US.UTF-8
[5] LC_MONETARY=en_US.UTF-8 LC_MESSAGES=en_US.UTF-8
[7] LC_PAPER=en_US.UTF-8 LC_NAME=C
[9] LC_ADDRESS=C LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
time zone: Etc/UTC
tzcode source: system (glibc)
attached base packages:
[1] stats graphics grDevices utils datasets methods base
other attached packages:
[1] dplyr_1.1.4 ggplot2.utils_0.3.2 ggplot2_3.5.1
[4] tern_0.9.7 rtables_0.6.11 magrittr_2.0.3
[7] formatters_0.5.10
loaded via a namespace (and not attached):
[1] generics_0.1.3 tidyr_1.3.1 EnvStats_3.0.0
[4] stringi_1.8.4 lattice_0.22-6 digest_0.6.37
[7] evaluate_1.0.3 grid_4.4.2 fastmap_1.2.0
[10] jsonlite_1.9.0 Matrix_1.7-2 backports_1.5.0
[13] survival_3.8-3 purrr_1.0.4 scales_1.3.0
[16] codetools_0.2-20 Rdpack_2.6.2 cli_3.6.4
[19] ggpp_0.5.8-1 nestcolor_0.1.3 rlang_1.1.5
[22] rbibutils_2.3 cowplot_1.1.3 munsell_0.5.1
[25] splines_4.4.2 withr_3.0.2 yaml_2.3.10
[28] tools_4.4.2 polynom_1.4-1 checkmate_2.3.2
[31] colorspace_2.1-1 forcats_1.0.0 ggstats_0.8.0
[34] broom_1.0.7 vctrs_0.6.5 R6_2.6.1
[37] lifecycle_1.0.4 stringr_1.5.1 htmlwidgets_1.6.4
[40] MASS_7.3-64 pkgconfig_2.0.3 pillar_1.10.1
[43] gtable_0.3.6 glue_1.8.0 xfun_0.51
[46] tibble_3.2.1 tidyselect_1.2.1 knitr_1.49
[49] farver_2.1.2 htmltools_0.5.8.1 labeling_0.4.3
[52] rmarkdown_2.29 random.cdisc.data_0.3.16 compiler_4.4.2 ---
title: RFG3
subtitle: Response Forest Graphs Within Treatment Arms by Continuous Biomarker Cutoff
categories: [RFG]
---
------------------------------------------------------------------------
::: panel-tabset
## Setup
For response endpoints it is good to show how to obtain within-treatment-arms comparisons of biomarker subgroups.
This is similar to [SFG04](../graphs/sfg04.qmd).
Similarly like in [RFG1](../graphs/RFG1/rfg01.qmd).
```{r, message = FALSE}
library(tern)
library(ggplot2.utils)
library(dplyr)
adrs <- random.cdisc.data::cadrs %>%
df_explicit_na() %>%
filter(PARAMCD == "BESRSPI", BMEASIFL == "Y") %>%
mutate(
is_rsp = AVALC %in% c("CR", "PR"),
ARM_BIN = fct_collapse_only(
ARM,
CTRL = c("B: Placebo"),
TRT = c("A: Drug X", "C: Combination")
),
BMRKR2 = fct_explicit_na_if(BMRKR2, BEP01FL == "N")
) %>%
var_relabel(
BEP01FL = "BEP",
BMRKR2 = "Biomarker (Categorical)"
)
```
## Plot
We define a vector of all cutpoints to use for a numeric biomarker (here `BMRKR1`).
We `lapply()` over this vector, each time generating a binary factor variable `BMRKR1_cut` and then tabulating the resulting statistics similar to [SFG4](../graphs/sfg04.qmd), this time including the treatment arms in the `subgroups` argument.
Then we `rbind()` all tables in the list together.
```{r}
# across arm ----
cutpoints <- c(4, 5, 8)
adrs2 <- adrs %>%
mutate(
BMRKR1 = ifelse(BEP01FL == "N", NA, BMRKR1),
BMRKR1_cut = explicit_na(cut(BMRKR1, c(-Inf, cutpoints, Inf), right = FALSE))
) %>%
var_relabel(BMRKR1_cut = "Biomarker (Binned Continuous)")
tables_all_cutpoints <- lapply(cutpoints, function(cutpoint) {
adrs3 <- adrs2 %>%
filter(!is.na(BMRKR1)) %>%
mutate(
BMRKR1_thresh = explicit_na(factor(
ifelse(BMRKR1 > cutpoint, "Greater", "Less")
))
)
tbl <- extract_rsp_subgroups(
variables = list(
rsp = "is_rsp",
arm = "BMRKR1_thresh",
subgroups = c("ARM_BIN")
),
label_all = paste0("BMRKR1 (", cutpoint, ")"),
data = adrs3
)
basic_table() %>%
tabulate_rsp_subgroups(
df = tbl,
vars = c("n_tot", "n", "n_rsp", "prop", "or", "ci")
)
})
result <- do.call(rbind, tables_all_cutpoints)
```
We can look at the result in the console already.
```{r}
result
```
We can now produce the forest plot using the `g_forest()` function.
Similarly as in [SFG4](../graphs/sfg04.qmd) we need to specify the `col_x`, `col_y` and `forest_header` arguments for `g_forest()` by recovering them from one of the original tables.
```{r, fig.width = 15}
one_table <- tables_all_cutpoints[[1]]
g2 <- g_forest(
result,
col_x = attr(one_table, "col_x"),
col_ci = attr(one_table, "col_ci"),
forest_header = attr(one_table, "forest_header"),
col_symbol_size = attr(one_table, "col_symbol_size")
)
```
With `gridExtra::grid.arrange()` the plot can be combined e.g. with a between-treatment-arms comparison, like [RFG1A](../graphs/RFG1/rfg01a.qmd)
{{< include ../misc/session_info.qmd >}}
:::