Skip to main content
v2026.11,986 entries · CC-BY 4.0

gtsummary in R: Table 1 and Regression Tables

How to use the gtsummary R package: install, build a Table 1, add p-values, make regression tables, read default output and fix common issues.

Written and maintained by CASRAI Editorial Board

Published

On this page

gtsummary is an R package that turns a data frame or a fitted model into a formatted table: a “Table 1” of participant characteristics, a regression results table, or a merged side-by-side table. This guide is for researchers and analysts who work in R and need to go from a data frame to a table they can drop into a manuscript. It covers what the package does, how to install it, a worked Table 1 and regression table, how to read the default output, and the documented causes of the most common surprises.

Facts about the package below were read from the package website and CRAN on 2026-11-07. At that date the current release on CRAN was gtsummary 2.6.1, published 2026-09-04, under an MIT licence.

What gtsummary is used for in research

The package documentation describes it as providing “an elegant and flexible way to create publication-ready analytical and summary tables.” CRAN describes it as creating “presentation-ready tables summarizing data sets, regression models, and more.” In research practice that covers three jobs:

  • Descriptive tables (usually Table 1) with tbl_summary(), optionally split by a grouping variable such as treatment arm.
  • Group comparisons by adding p-values to that table with add_p().
  • Model tables with tbl_regression(), for example odds ratios with 95% confidence intervals from a logistic model.

The package was described in the R Journal (2021, volume 13, pages 570-580, doi:10.32614/RJ-2021-053); cite that paper if gtsummary produced a table in your work. CRAN lists Daniel D. Sjoberg as maintainer.

Install or access

Install the released version from CRAN:

install.packages("gtsummary")
library(gtsummary)

The package website gives this route for the development version from GitHub:

pak::pkg_install("ddsjoberg/gtsummary", dependencies = TRUE)

CRAN lists, among its imports, the cards, cardx, cli, dplyr and gt packages (gt version 0.11.1 or later), which install automatically. gtsummary is R-only; the documentation and CRAN page describe an R package and no Python port is documented there.

Step-by-step: from a data frame to Table 1 and a regression table

The package website demonstrates the workflow with its built-in trial data set. Substitute your own data frame; the cleaning rules described under “Output and how to read it” apply to it.

  1. Load the package and your data. library(gtsummary). Know your variable types first; see data frames in R for the structure gtsummary expects.
  2. Build the overall summary. table1 <- trial |> tbl_summary(include = c(age, grade, response)). The include argument chooses which columns appear.
  3. Stratify by group. Add by = trt to split the columns by treatment arm, for example tbl_summary(by = trt, include = c(age, grade, response)).
  4. Add p-values and an overall column. Pipe the result to add_p() and, if wanted, add_overall().
  5. Relabel variables. Pass label = list(age = "Age, years") to tbl_summary().
  6. Build a regression table. Fit a model, then pass it to tbl_regression(). For a logistic model, set exponentiate = TRUE so the estimate column shows odds ratios; the reference page’s own example footer reads “OR = Odds Ratio, CI = Confidence Interval”.
  7. Combine tables. tbl_merge() places several gtsummary tables side by side.
  8. Export. as_gt() converts a table to a gt object, which can be saved with gt::gtsave(). as_flex_table() converts it to a flextable object; the reference page notes that the gt package does not support Word, which is why a flextable route exists for Word output.

Output and how to read it

The default statistics come from the tbl_summary() reference page (read 2026-11-07):

Variable type Default statistic What the cell shows
Continuous {median} ({p25}, {p75}) Median with the 25th and 75th percentiles (the interquartile range bounds)
Categorical {n} ({p}%) Count with percentage
  • Percent basis. The percent argument defaults to "column"; "row" and "cell" are the alternatives.
  • Missing data. missing defaults to "ifany", which adds a missing row only for variables that have missing values. The row is labelled “Unknown” by default (missing_text) and shows the count of missing observations (missing_stat = "{N_miss}"). "no" and "always" are the other options.
  • Which p-value test add_p() used. With a by variable and no group or adj.vars, the documented defaults are the Wilcoxon rank-sum test for continuous variables with two groups, the Kruskal-Wallis test for three or more groups, and a chi-squared test without continuity correction for categorical variables when all expected counts are at least 5, switching to Fisher’s exact test when any expected count is below 5. Override with test = everything() ~ "t.test", for example. The default p-value format rounds to one digit (label_style_pvalue(digits = 1)); pass pvalue_fun = label_style_pvalue(digits = 2) for two.

Because the defaults are non-parametric and report medians, state in your methods text which statistic and test each row used. For guidance on interpreting the model table, see how to read a regression table; for the descriptive statistics themselves, see summary statistics in R. Reporting conventions for statistics in journals are covered in APA statistics reporting.

Troubleshooting

Symptom Documented cause and fix
A numeric variable (for example a 1-5 score) is summarised as counts and percentages instead of a median Per the reference page, numeric variables with fewer than 10 unique levels default to categorical. Set type = list(score ~ "continuous") to force it. Valid types are "continuous", "continuous2", "categorical" and "dichotomous".
A yes/no variable shows one row instead of one row per level Variables coded TRUE/FALSE, 0/1 or yes/no are assumed dichotomous and display a single row. Use type = list(var ~ "categorical") to show every level.
You want a mean (SD) instead of a median (IQR) Set statistic = all_continuous() ~ "{mean} ({sd})".
No p-value for a grouped (clustered) comparison with three or more groups The add_p.tbl_summary reference states there is no default for grouped data when the by variable has more than two levels; choose a test explicitly with test.
Regression estimates look like log-odds exponentiate defaults to FALSE; set exponentiate = TRUE for odds ratios from a logistic model.

Free alternatives

Other CRAN packages build Table 1-style summaries. Descriptions below are quoted or paraphrased from each package’s CRAN page (read 2026-11-07).

Package What CRAN says it does Version and date on CRAN Licence
gtsummary Presentation-ready tables summarising data sets and regression models 2.6.1, 2026-09-04 MIT
tableone Creates “Table 1”, a description of baseline patient characteristics; supports continuous and categorical variables, p-values and standardized mean differences; weighted data via the survey package 0.13.2, 2022-04-15 GPL-2
arsenal tableby() for Table 1-style summaries by group, paired(), modelsum(), freqlist(), comparedf() and write2() for document output 3.7.4, 2026-10-17 GPL-2 | GPL-3

Pick by the output you need: only gtsummary’s documentation covers regression tables with tbl_regression() and conversion to gt and flextable objects, while tableone’s CRAN description highlights standardized mean differences and survey-weighted data.

FAQ

What is gtsummary in R?

An R package for creating publication-ready summary and regression tables from data frames and fitted models. As of 2026-11-07 the current CRAN version is 2.6.1.

How do I install gtsummary?

Run install.packages("gtsummary") in R, then library(gtsummary).

Is there a Python version of gtsummary?

The package documentation and CRAN page describe an R package only. This guide does not cover a Python equivalent.

How do I make a Table 1 grouped by treatment?

Use tbl_summary(by = trt) and pipe to add_p() to add between-group p-values.

Can gtsummary export to Word?

Convert the table with as_flex_table(), which returns a flextable object; the gtsummary reference notes the gt package does not support Word.

Sources

  • gtsummary package website, danieldsjoberg.com/gtsummary, including the reference pages for tbl_summary, add_p.tbl_summary, tbl_regression and as_flex_table (read 2026-11-07).
  • CRAN: gtsummary, tableone, arsenal (read 2026-11-07).
  • The R Journal, 2021, volume 13, pages 570-580, doi:10.32614/RJ-2021-053 (the gtsummary package paper).

Follow CASRAI

Research-administration guidance, standards updates and independent tool reviews.

Ask CASRAI · free to try

Ask about gtsummary in R: Table 1 and Regression Tables

2 free questions a day, no account.

An AI assistant for research: funder rules, compliance, methods and publishing. It links the sources behind its answers and labels answers that come from a web search.

Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.

Works on this site and inside Claude, Cursor and the AI tools you already use.

Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays open to read, with no account and no card.

Referenced across the research world

University of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logoUniversity of Cambridge logoColumbia University logoCrossref logoUniversity of Edinburgh logoHarvard University logoUniversity of Oxford logoPrinceton University logoStanford School of Medicine logoUniversity College London logoORCID logo
  • University of Cambridge logo
  • Columbia University logo
  • Crossref logo
  • University of Edinburgh logo
  • Harvard University logo
  • University of Oxford logo
  • Princeton University logo
  • Stanford School of Medicine logo
  • University College London logo
  • ORCID logo

View CASRAI adoption →