The development and implementation of an individualised prognostic modelling tool for nonmetastatic prostate cancer: PREDICT Prostate

2017 David Thurtle

David Thurtle
Clinical Research Associate and Honorary Urology Registrar
Academic Urology Group, University of Cambridge and Cambridge University Hospitals NHS Foundation Trust

Date of report:
December 2019

Project start date:
TUF funding commenced 1st January 2018
Funding completion date 30th June 2019

“Many thanks to The Urology Foundation and all of its supporters. This work would not have been possible without the generous and encouraging support of TUF. This was a project that would have struggled to gain support from ‘larger’ national funding bodies, yet its clinical impact and importance is considerable. We hope that patients and clinicians will continue to benefit from this work, made possible by TUF, for many years to come.”

Research hypothesis:
A new individualised prognostic model can be devised to predict the personal risk of cancer-specific mortality in men with newly-diagnosed non-metastatic prostate cancer. This model can be used to test the impact of different treatment choices on this outcome and to inform optimum treatment strategies.

Target study numbers:
PREDICT Prostate patient study – 150 participants

Problems encountered:
Delays between study introduction and study appointments due to patient pathways.

Progress to date:
8 sites included onto study. Further 3 pending.

Study recruitment to date:
70 completed

Key findings:

  • Development of a survival model which accurately predicts prostate cancer-specific and overall survival outcomes at diagnosis, with concordance indices in excess of 0.80.
  • The robustness and generalisability of the model has been demonstrated within an independent cohort of 69,000 Swedish men.
  • The model was developed into a free-to-use patient-facing website which has been hosted on an NHS server and has subsequently been endorsed by NICE as a decision aid.
  • Our work has shown health care professionals tend to overestimate prostate cancer mortality, when compared to this multivariable model. Predictions of prostate cancer mortality were 1.9-fold higher, and predictions of survival benefit from treatment were 4-fold higher than estimates from PREDICT prostate. Treatment recommendations of health care professionals may be affected by seeing PREDICT Prostate estimates. Our results suggest this might be most evident in intermediate-risk cases.
  • Interim analysis of the patient study is pending. Results will assess predominantly the impact of PREDICT Prostate on patients’ decisional confidence and anxiety.

Significance of findings:

Published output already available on:

  • Model development
  • External validation in large Swedish cohort of 69,000 men
  • Potential impact on clinician decision-making

Potential Clinical Impact:
The development of an accurate model and user-friendly web interface for PREDICT Prostate allows for widespread accessibility and clinical usability. The model enables standardised, individualised information to be provided to men newly diagnosed with prostate cancer – and their healthcare team. This should inform the predominant decision dilemma for a significant proportion of patients, i.e. whether to undergo immediate radical treatment or commence on active surveillance. As our clinician study suggests, this information may reduce over-treatment in more favourable-risk cases. The impact on patient decision-making is still being assessed in the ongoing multi-centre randomised controlled trial of the decision aid.

Key publications:

  • Understanding of prognosis in non-metastatic prostate cancer: a randomised comparative study of clinician estimates measured against the PREDICT prostate prognostic model. Thurtle DR, Jenkins V, Pharoah PD, Gnanapragasam VJ.
  • Br J Cancer. 2019 Sep 16. doi: 10.1038/s41416-019-0569-4. PMID: 31523057
  • Models predicting survival to guide treatment decision-making in newly diagnosed primary non-metastatic prostate cancer: a systematic review.
  • Thurtle D, Rossi SH, Berry B, Pharoah P, Gnanapragasam VJ.
  • BMJ Open. 2019 Jun 22;9(6):e029149. doi: 10.1136/bmjopen-2019-029149. PMID: 31230029
  • Individual prognosis at diagnosis in nonmetastatic prostate cancer: Development and external validation of the PREDICT Prostate multivariable model.
  • Thurtle DR, Greenberg DC, Lee LS, Huang HH, Pharoah PD, Gnanapragasam VJ.
  • PLoS Med. 2019 Mar 12;16(3):e1002758. doi: 10.1371/journal.pmed.1002758. eCollection 2019 Mar. PMID: 30860997


  • PREDICT External validation at EAU 2019, Barcelona, Spain and at BAUS 2019, Glasgow
  • PREDICT Clinician Study at EAU 2019, Barcelona, Spain and BAUS 2019, Glasgow UK
  • PREDICT Prostate presentation at BUG 2018, Birmingham, UK
  • PREDICT: Prostate – a novel prognostic model that estimates individual survival in newly diagnosed primary non-metastatic prostate cancer. AUA 2018, San Francisco USA, EAU 2018, Copenhagen, & BAUS 2018, Liverpool, UK.


  • Vice Chancellors Research Impact and Engagement Award, University of Cambridge, for PREDICT Prostate work
  • NICE Endorsement of PREDICT Prostate
  • Best poster: Prostate Cancer at BAUS 2019
  • Best poster: Individualising prostate cancer treatment decision-making at EAU 2019
  • Second Prize for best abstract by a resident at EAU Congress 2019
  • Best poster: Prostate Cancer Staging: Imaging versus statistics. EAU 2018
  • Best poster: Prostate Cancer Diagnostics and Active Surveillance. EAU 2018
  • Best ePoster winner: Prostate Cancer Treatment. BAUS 2017

Future Studies and Plans:

  • The ‘PREDICT Prostate Patient Study’ is ongoing ( . It is due to close in July 2020. The study seeks to assess what impact the model might have on patients’ decisional conflict, anxiety around decision-making, and final treatment decisions.
  • We are exploring the development of a ‘Prostate Plus’ model – focussing on men with advanced disease, with or without metastases, using similar methodology. This has the potential to enable modelling of benefits from novel agents such as Abiraterone and Docetaxel.
  • Analytical protocols and web-designs have been made in a reproducible way such that similar models can be developed for other tumour-types which may be exploited for kidney, bladder or testicular cancer.

Are you likely to apply for further funding from The Urology Foundation for this study?
No – unlikely to be necessary for this study

Have you secured further follow-on research funding?
Not required currently.



Find out more