MPN Personalised Risk Calculator

MPN Personalised Risk Calculator

Input New Patient Data

Enter data for new patient. Unknown data will be imputed from available variables.

Input New Patient Data From File

Download data template (.csv file) Download Template
Complete template using 0/1 for absence/presence. Age in years, Haemoglobin in g/l, white cell and platelet counts x10^9/l, 1 for female, 2 for male For unknown variables: leave as NA and this will be imputed

Patient Description
Patient Outcomes
Generate report

This application is based on data and prognostic models from Grinfeld and Nangalia et al. 2018

The Genomics tab allows the user to view the frequency of mutations(s) across MPN subtypes.

Alternatively, to generate individual patient predictions, first select the diagnosis of interest: ET, PV, MF or other (MPNu, MDS/MPN overlap etc)

Then choose between:

  • Selecting a patient already used in the analysis to view their clinical and genomic parameters, predicted and actual outcomes,
  • Inputing variables for a new or hypothetical patient by manually inputting variables, or
  • Inputing variables for a new or hypothetical patient by downloading, completing and uploading a csv template file

The output is viewed on the Patient Prediction tab.

This calculator is intended as an adjunct to the paper and for research purposes only.

It has not been prospectively validated and predictions derived from it should be used with caution.

Data regarding the accuracy of the model are provided in the paper. In general, predictions are accurate in approximately 80% of cases

Outcome predictions are from diagnosis and uses the risk associated with variables from time of diagnosis. If time of genomic sampling is post diagnosis then we suggest adjusting patient age to time of genomic sampling, and to use this as the starting time for predictions.

Requests related to the MPN prediction model should be addressed to Dr. Jacob Grinfeld (, queries related to the web site and the app functionality should be addressed to Dr. Eugene Nadezhdin (

Shiny implementation - Jacob Grinfeld, with additional work by Eugene Nadezhdin.
CoxHD package and multistate models - Moritz Gerstung, with additional work by Rob Cantrill and Jacob Grinfeld.
Last update: July 2020

Frequency of genomic variables across MPN phenotypes