By: Yayehirad A Melsew

March 2025

Objective

Understanding chemotherapy sequencing is essential in oncology research. By visualising treatment transitions, we can gain insights into patient care pathways, identify trends, and improve decision-making in cancer treatment. Sankey diagrams provide an intuitive way to map these transitions and highlight end transitions such as ongoing treatment (if there is no treatment failure) or death.

Introduction

Tracking and analysing treatment pathways is crucial in oncology. Sankey diagrams allow researchers and clinicians to visualise sequential treatments and better understand how patients progress through different lines of therapy. Unlike traditional flowcharts or timelines, Sankey diagrams emphasize the volume of transitions between stages, making them ideal for complex treatment data. To facilitate this, I have developed an R function, create_sankey_diagram, that enables easy creation of such diagrams. This blog post will walk you through its usage and applications.

Installation and Setup

First, install the necessary R packages:

install.packages(c("ggalluvial", "dplyr", "ggplot2", 
                   "readxl", "RCurl", "RColorBrewer"))
install.packages("devtools")
devtools::install_github("davidsjoberg/ggsankey")

Then, source the function directly from GitHub:

source("https://raw.githubusercontent.com/Yayehirad/Sankey/master/create_sankey_diagram.R")

Note: If you encounter issues installing ggsankey (e.g., GitHub access errors), ensure devtools is installed correctly and check your internet connection. Alternatively, consult the ggsankey GitHub page for manual installation instructions.

Id	FirstLine	SecondLine	ThirdLine	FourthLine
P1	ChemoG	ChemoD	ChemoA	ChemoG
P2	ChemoG	Ongoing	NA	NA
P3	ChemoC	ChemoD	ChemoD	Ongoing
P4	ChemoF	ChemoF	ChemoF	Ongoing
P5	ChemoC	ChemoG	ChemoC	ChemoF
P6	ChemoB	ChemoB	ChemoB	ChemoB
P7	ChemoB	ChemoB	Ongoing	NA
P8	ChemoF	ChemoF	ChemoF	ChemoF
P9	ChemoC	ChemoD	ChemoB	ChemoE

Note: In this table, terms like “ChemoG” or “ChemoD” are placeholders representing different chemotherapy regimens, while “Ongoing” indicates the patient remains on that treatment, and “NA” means no further treatment was recorded.

Loading Simulated Treatment Data

We will use a simulated dataset that represents treatment pathways for 100 patients across multiple lines of therapy.

simulated_data <- read.csv("https://raw.githubusercontent.com/Yayehirad/Sankey/master/simulated_data.csv")

Generating Sankey Diagrams

Now, let’s use create_sankey_diagram to visualise treatment transitions.

Example 1: Two Lines of Treatment with Legend

plot12_with_legend <- create_sankey_diagram(
  data = simulated_data,              # Dataset containing treatment data
  id_col = "Id",                      # Column identifying unique patients
  lot_cols = c("FirstLine", "SecondLine"), # Columns for treatment lines to plot
  show_legend = TRUE,                 # Include a legend for treatment types
  show_numbers = TRUE                 # Display patient counts on flows
)
print(plot12_with_legend)

Example 2: Three Lines of Treatment Without Legend

plot123_no_numbers <- create_sankey_diagram(
  data = simulated_data,              # Dataset containing treatment data
  id_col = "Id",                      # Column identifying unique patients
  lot_cols = c("FirstLine", "SecondLine", "ThirdLine"), # Columns for treatment lines
  show_legend = FALSE,                # Exclude the legend
  show_numbers = TRUE                 # Display patient counts on flows
)
print(plot123_no_numbers)

Example 3: Four Lines of Treatment

plot1234 <- create_sankey_diagram(
  data = simulated_data,              # Dataset containing treatment data
  id_col = "Id",                      # Column identifying unique patients
  lot_cols = c("FirstLine", "SecondLine", "ThirdLine", "FourthLine"), # Columns for treatment lines
  show_legend = FALSE,                # Exclude the legend
  show_numbers = FALSE                # Hide patient counts on flows
)
print(plot1234)

Conclusion

The create_sankey_diagram function is a powerful tool for visualising complex treatment pathways in cancer research. By incorporating these diagrams into your workflow, you can better understand patient journeys, treatment choices, and outcomes. This can support data-driven decision-making and enhance oncology research.