10.1 Introduction

As part of a systematic review and meta-analysis, you may also want to examine the internal validity (risk of bias) of included studies using the relevant domain-based risk of bias assessment tool, and present the results of this assessment in a graphical format. The Cochrane Handbook recommends two types of figure: a summary barplot figure showing the proportion of studies with a given risk of bias judgement within each domain, and a traffic light plot which presents the domain level judgments for each study.

However, the options available to researchers when creating these figures are limited. While RevMan has the functionality to create the plots, many researchers do not use it to conduct thei systematic review and so copying the relevant data into the system is an inefficient solution. Similarly, producing the graphs by hand, using software such as MS PowerPoint, is time consuming and means the figures have to manually updated if changes are needed. Additionally, journals usually require figures to be of publication quality (above ~300-400dpi), which can be hard to achieve when exporting the risk of bias figures from RevMan or creating them by hand.

Example RevMan output.

Example RevMan output.

To avoid all of this, you can now easily plot the risk of bias figures yourself within RStudio, using the robvis package which provides functions to convert a risk of bias assessment summary table into a summary plot or a traffic-light plot.

10.1.1 Load robvis

Assuming that you have already installed the dmetar package (see Section 2.2.1), load the robvis package using:

library(robvis)

10.1.2 Importing your risk of bias summary table data

To produce our plots, we first have to import the results of our risk of bias assessment from Excel into R. Please note that robvis expects certain facts about the data you provide it, so be sure to follow the guidance below when setting up your table in Excel:

  1. The first column is labelled “Study” and contains the study identifier (e.g. Anthony et al, 2019)
  2. The second-to-last column is labelled “Overall” and contains the overall risk-of-bias judgments
  3. The last column is labelled “Weight” and contains some measure of study porecision e.g. the weight assigned to each study in the meta-analysis, or if no meta-analysis was performed, the sample size of each study). See Section 10.2.2.3 for more details.
  4. All other columns contain the results of the risk-of bias assessment for a specific domain.

To elaborate on the above guidance, consider as an example the ROB2 tool which has 5 domains. The resulting dataset that robvis would expect for this tool would have 8 columns:

  • Column 1: Study identifier
    • Column 2-6: One RoB2 domain per column
    • Column 7: Overall risk-of-bias judgments
    • Column 8: Weight

In Excel, this risk of bias summary table would look like this:

Note: for three of the four tool templates (ROB2, ROBINS-I, QUADAS-2), what you name the columns containing the domain-level judgments is not important, as the templates within robvis will relabel each domain with the correct tool-specific heading.

Once you have saved the table you created in Excel to the working directory as a comma-separated-file (e.g. “robdata.csv”; see Section 3.2), you can either read the file into R programatically using the command below or via the “import assistant” method as described in Section 3.2.3.

my_rob_data <- read.csv("robdata.csv", header = TRUE)

10.1.3 Templates

robvis produces the risk of bias figures by using the data you provide to populate a template figure specific to the risk of bias assessment tool you used. At present, robvis contains templates for the following three tools:

  • ROB2, the new Cochrane risk of bias tool for randomized controlled trials;
  • ROBINS-I, the Risk of Bias In Non-randomized Studies - of Interventions tool;
  • QUADAS-2, the Quality and Applicability of Diagnostic Accuracy Studies, Version 2;

robvis also contains a special generic template, labelled as ROB1. Designed for use with the original Cochrane risk of bias tool for randomized controlled trials, it can also be used to visualize the results of assessments performed with other domain-based tools not included in the list above. See Section 10.4 for more information on the additional steps required when using this template.

10.1.4 Example datasets

The robvis package contains an example dataset for each template outlined above. These are stored in the following objects:

  • data_rob2 : Example data for the ROB2 tool
  • data_robins : Example data for the ROBINS-I tool
  • data_quadas : Example data for the QUADAS-2 tool
  • data_rob1 : Example data for the RoB-1 tool

You can explore these datasets using the head() function. For example, once you have loaded the package using library(robvis), viewing the ROBINS-I example dataset can be achieved by running the following command:

head(data_robins)

These example datasets are used to create the plots presented through the remainder of this guide.



banner