Skip to contents

Forest Plot of Hazard Ratios with Confidence Intervals

Source: R/plot_cox.R
plot_cox_forest.Rd

Generates a publication/deck-ready horizontal forest plot of hazard ratios.

Usage

plot_cox_forest(
  data,
  label_col = "index_label",
  hr_col = "estimate",
  low_col = "conf.low",
  high_col = "conf.high",
  p_col = NULL,
  color = "#9D2235",
  scale = c("hr", "log_hr"),
  color_by = c("none", "significance"),
  order_by = c("none", "magnitude", "pvalue"),
  x_limits = NULL,
  x_breaks = NULL,
  title = NULL,
  caption = NULL,
  base_size = 13,
  label_map = NULL,
  style = NULL,
  exponentiate = NULL,
  log_scale = FALSE,
  sig_level = 0.05
)

Arguments

data

Data frame containing tidy Cox regression results (e.g. from broom::tidy or multi-model runs).

label_col

Column name for labels / terms (character string or unquoted).

hr_col

Column name for hazard ratios, always on the HR (not log) scale (default: "estimate").

low_col

Column name for lower CI bounds, on the HR scale (default: "conf.low").

high_col

Column name for upper CI bounds, on the HR scale (default: "conf.high").

p_col

Optional column name for p-values. Required for color_by = "significance" and order_by = "pvalue".

color

Primary color for points and error bars (default: Temple Cherry "#9D2235").

scale

One of "hr" (default; plots hazard ratios with a reference line at 1) or "log_hr" (log-transforms hr_col/low_col/high_col for plotting, with a reference line at 0). Equivalent to exponentiate = TRUE and FALSE; kept for existing code. Give scale or exponentiate, not both with different meanings.

color_by

One of "none" (default) or "significance", which colors points/error bars by whether p_col is below sig_level (a term is significant when p < sig_level; a missing p-value counts as not significant) and adds a legend.

order_by

One of "none" (default; original row order), "magnitude" (sorts by abs(log(estimate)) descending), or "pvalue" (sorts by p_col ascending).

x_limits

Optional 2-element numeric vector for x-axis limits.

x_breaks

Optional numeric vector of axis break points.

title

Optional plot title.

caption

Optional plot caption.

base_size

Base font size (default: 13).

label_map

Optional label mapping (see [cbe_labels()]).

style

Optional label style (`"short"` or `"full"`).

exponentiate

`NULL` (default: follow `scale`), `TRUE` to show hazard ratios (reference line at 1) or `FALSE` to show the raw coefficients, `log(HR)` (reference line at 0, linear axis). The columns you pass are always on the HR scale; this only chooses what is displayed, and the confidence limits are transformed with the estimate.

log_scale

Logical. With hazard ratios, `TRUE` draws the axis on a log scale (reference line still at 1, so equal ratios in either direction are the same distance from it). This is a separate choice from `exponentiate`; `log_scale = TRUE` with `exponentiate = FALSE` is an error, because the coefficients are already logs. All HR and confidence limits must be positive.

sig_level

Significance threshold for `color_by = "significance"` (default 0.05).

Value

A ggplot2 object.

See also

[plot_cox_forest_multi()]