Joint Modelling

This is the third research project during my PhD studies. Most of the methodological work was done at KAUST under supervision of Denis Rustand,phd, and Prof. Haavard Rue.

This work was accepted for presentation at Baltic-24 EAU meeting (May 2024, Tallinn, Estonia), the Abstract is below, and full paper will be added later!

Abstract

Introduction and Objectives

Conflicting evidence exists for the effect of prostate cancer (PCa) treatment delay on outcomes after curative treatment. PCa is generally slow-growing, and treatment delays up to 6 months have not been associated with impaired outcomes. However, the effect of treatment delay on outcomes in intermediate PCa is more controversial, especially during the MRI era. In this study, we assess the effect of delaying or expedited curative treatment (surgery or radiation therapy) on the time to treatment relapse.

Materials and Methods:

We used the Helsinki University Hospital registry for data mining and to categorise PCa patients by Gleason grade group (1-5), treatment (RP/RT), and follow-up outcome (BCR) for a final sample size of 11719 patients (1993–2019). A broader definition of biochemical recurrence was established considering secondary treatments as an additional indicator of relapse alongside traditional PSA cut-offs (PSA of 0.2 ug/l for surgery and PSA nadir +2 ug/l for radiation therapy). A joint survival model (utilising INLAJoint R package) was employed to model the relationship between the time from diagnosis to curative treatment (treatment risk) and the time from diagnosis to treatment recurrence (relapse risk).

Results:

Conditional on covariates, including age at diagnosis and Gleason grade group, our joint survival model revealed a negative association γ = -0.99 [-0.99, -0.81] between the risk of treatment and risk of treatment relapse. Regardless of the grade group, individuals within the top 15% with the lowest risk of receiving treatment (i.e., the longest time to treatment) exhibited a 1.96 [1.05, 7.17] increased risk of recurrence compared to the average individual. Conversely, individuals within the top 15% with the highest risk of receiving treatment (i.e., shortest time to treatment) showed a 0.51 [0.14, 0.95] decreased risk of recurrence. When fitting separate models for each grade group, the association decreased in higher grade groups. Additionally, associations of delaying treatment were higher than associations of earlier treatment.

If you have any feedback or comments, please feel free to leave them in the section below. Thank you!