how to summarise multiple logistic regression models in a table?

Question

I have a dataset with age as continuous and as a factor, sex as a factor and 4 groups.

structure(list(Age = c(9, 12, 16, 57), Age_1 = structure(c(2L, 
3L, 3L, 7L), .Label = c("8", "1", "2", "3", "4", "5", "6", "7"
), class = "factor"), Sex = structure(c(2L, 1L, 2L, 1L), .Label = c("M", 
"F", "U"), class = "factor"), N = structure(c(2L, 2L, 2L, 
2L), .Label = c("0", "1"), class = "factor"), G = structure(c(1L, 
1L, 1L, 1L), .Label = c("0", "1"), class = "factor"), L_1 = 
structure(c(1L, 
1L, 1L, 1L), .Label = c("0", "1"), class = "factor"), C_1 = 
structure(c(1L, 
1L, 1L, 1L), .Label = c("0", "1"), class = "factor"), G_1 = 
structure(c(1L, 
1L, 1L, 1L), .Label = c("0", "1"), class = "factor"), m = structure(c(1L, 
1L, 1L, 1L), .Label = c("0", "1"), class = "factor"), A = c(1, 
1, 1, 1)), row.names = c(NA, 4L), class = "data.frame")

I want to do logistic regression for each variable (Age, Age_1 and sex) for each of the groups (N, G, L_1, C_1, G_1, m). eg.

mylogit <- glm(N  ~ Sex, data = logistic_s, family = "binomial")
mylogit <- glm(N  ~ Age, data = logistic_s, family = "binomial")

I am using gtsummary for combining the variables in a table.

library(gtsummary) 

tbl_n <-
      tbl_uvregression(
        logistic_s[c("N", "Age", "sex", "Age_1")],
        method = glm,
        y = N,
        method.args = list(family = binomial),
        exponentiate = TRUE
      )

tbl_n  

This produces the output for one group (eg. N) with the variables Age, Age_1, Sex.

I want to repeat this with each of the groups (eg N, G, L_1 etc) and then combine the tables to make one combined table.

I am open to using different packages if there are other options that suit this better. I want to make a table that can be exported in word.

Answer 1

I agree that some reproducible code would be helpful. I am not 100% certain what kind of output you're looking to get. You want to build some univariate logistic regression models, separately for 2 or more groups?

If that is correct, here is one way to go about it: I will use the trial data set in the gtsummary package as an example. I'll make the grouping variable treatment (trt).

library(gtsummary)
library(tidyverse)

trial_subset <-
  trial %>%
  select(trt, response, age, marker, grade) 

We'll begin by constructing univariate regression tables stratified by trt using the tbl_uvregression() function from gtsummary package. They will be stored in a new column in the data frame called tbl_uv.

df_uv <-
  trial_subset %>%
  # group by trt, and nest data within group
  group_by(trt) %>%
  nest() %>%
  # build univariate logistic regression models separately within grouping variable
  mutate(
    tbl_uv = map(
      data,
      ~tbl_uvregression(
        data = .x, 
        y = response,
        method = glm, 
        method.args = list(family = binomial),
        exponentiate = TRUE
      )
    )
  )
#> # A tibble: 2 x 3
#> # Groups:   trt [2]
#>   trt    data               tbl_uv    
#>   <chr>  <list>             <list>    
#> 1 Drug A <tibble [98 x 4]>  <tbl_vrgr>
#> 2 Drug B <tibble [102 x 4]> <tbl_vrgr>

We can now use the tables saved in df_uv to merge them into a single table using the tbl_merge() function.

tbl_merge(
  tbls = df_uv$tbl_uv, # list of the univariate logistic regression tables
  tab_spanner = paste0("**", df_uv$trt, "**") # adding stars to bold the header
)

This produces the table below. I hope this is helpful!