Adverse Events

This is a data visualisation tool for adverse events. The web based application takes a json data file and provides tools to explore the data at a patient and aggregated level in the form of tables and graphs. There are a number of filters to allow further investigation of the data.

There is also some support to look at measure data (such as bloods) and treatment received data

This is still a work in progress and may be updated considerably.

Usage

Go here to start using this application. There is an example dataset with dummy data in the Data folder of this repository if you want to test it out. Note: you will need to download the data and select it from your local machine.

For best results use the Google Chrome browser.

File format requirements

The file should be of type json and contain all the information listed below.

treatment: An array of objects containing the treatment information.

• value: the numeric or string value which is found in patientData
• label: the label to display for this treatment group
• color: a color associated with this treatment group

keyDates: An array of objects containing date information for key time periods.

• column: the column name of the column in patientData
• label: A display name for this date

causality: An array of objects containing causality information.

• column: the column name of the column in patientData
• label: A display name for this causality

patientData: A file containing one row per patient.

• patid: A unique patient identifier such as trial number
• treatment: the treatment assined to this patient. Should be one of the treatment values
• keyDates: As defined in the keyDates object. Should include an end date such as end of treatment

toxData: A file containing adverse event data over time.

• patid
• aestartdate: the adverse event start date or the date the adverse event grade changed to this grade
• aestopdate: the adverse event end date or the date the adverse event grade changed from this grade
• aecategory: the adverse event category
• aeterm: the adverse event name
• aegrade: the adverse event grade
• sae: Was this an SAE?
• causalities: As defined in the causality object. The data store here should be numeric 1 for definitely not related through to 5 for definitely related

keyGroups: (optional) A column within toxData containing a logical value. A value of 1 will be included in this group. This can be used to group adverse events such as haematological toxicities, or to group patients such as those with metatstatic disease.

• column: the column name of the column in toxData
• label: A display name for this group

keyEvents: (optional) A column label and color for an event to be placed on the adverse event patient level plots.

• column: the column name of the column in patientData
• label: A display name for this event
• color: a color associated with this event type

measureData: (optional) A dataset of measure data at various time points such as blood test results.

• patid: A unique patient identifier such as trial number
• dateOfMeasure: The date the measurement was taken
• measureColum: The column named in measureColumns

measureColumns: (optional) A list of measure types to plot from measure data.

• column: the column name of the column in measureData label: A display name for this measure
• min: (optional) A minimum value for the range of this measure. A red line will be drawn at this level in the measure plot. Counld represent normal range or a CTCAE grade.
• max: (optional) A maximum value for the range of this measure. A red line will be drawn at this level in the measure plot. Counld represent normal range or a CTCAE grade.

treatmentSpecification: (optional, experimental) A list of treatment data metadata for treatment compliance data, for plotting patient treatment data.

• index: A reference to the treatment type. In the case that there is multiple data for a single treatment type these should reference the same index and will be plotted overlaid in order.
• type: One of "Single" or "Double". Define if the data is a single day or a time period between two dates.
• label: A display name for this measure
• datasetName: The name of the dataset containing the information for this treatment. This dataset should be provided as part of the data object column: A column containing a numerical representation of dose
• startDate: The date treatment was given or the start of a time period that treatment was given on
• endDate: the end of a treatment period for type = "Double"
• doseColors: An array of dose colors. Each object should contain a value and a color. The treatment dose is assigned the color of the largest value in this array which is less than or equal to the dose. labels may be added and are used for keys in place of values. {"value":"0", "color": "#000000", "label": "0-25mg"}

Creating a .json file in R

R is able to convert a list object into a json file. One option is to use the package jsonlite (library(jsonlite)). Below is an example.

library(jsonlite)
# patientData and toxData are already defined
# create the datafile
aeMetaData = list()

# Add metadata
# Treatment
aeMetaData$treatment = data.frame(
  value = c("Treatment", "Placebo"),
  color = c("#00FF00","#FF0000"), # rgb colors
  stringsAsFactors = FALSE
)

# Key Dates
aeMetaData$keyDates = data.frame(
  column = c("randomisationdate", "date_of_treatment_PartA", "date_of_treatment_PartB", "date_stopped_treatment"),
  label = c("Date of Randomisation", "Part A", "Part B", "Date stopped treatment"),
  stringsAsFactors = FALSE
)

# Causality
aeMetaData$causality = data.frame(
  column = c("imp_causality"),
  label = c("IMP Causality"),
  stringsAsFactors = FALSE
)

# add main data
aeMetaData$patientData = patientData
aeMetaData$toxData = toxData

# create json string
string = toJSON(aeMetaData, pretty = TRUE, Date = "ISO8601")
# save to file
writeLines(string, con="trialName_AEdata.json")

Development version

To set-up a local development server:

1. Download this repository.
2. Download and install Node.js. Node comes with a package manager which we will use to download all the dependencies and also to run the app.
3. Open the command line or terminal.
4. Navigate to the folder containing this repository (e.g. in windows cd C:/folderpath
5. Install dependencies using the command npm install -s
6. Start the app using the command npm start
7. Once built the app will be running on http://localhost:3000. Go to this page in a web browser. Note that testing was performed in chrome.

Editing

There is a structure summary in the file /src/index.js.

Technical details

This web application is written in javascript (jsx) relying heavily on React and d3. React is a data binding package which means when the data is changed the interface updates automatically. d3 is a graphing and data visualisation library. Bootstrap is used for styling and CSS grid for layout.

Author

Peter Dutton (Centre for Statistics in Medicine, University of Oxford)

License

GPL-3