Modelling dynamic covariates effect on survival via Functional Data Analysis: application to the MRC BO06 trial in osteosarcoma

Keywords

Statistical learning
Health Analytics
Code:
27/2020
Title:
Modelling dynamic covariates effect on survival via Functional Data Analysis: application to the MRC BO06 trial in osteosarcoma
Date:
Friday 8th May 2020
Author(s):
Spreafico, M.; Ieva, F.; Fiocco, M.
Download link:
Abstract:
Time-varying covariates are of great interest in clinical research since they represent dynamic patterns which reflect disease progression. In cancer studies biomarkers values change as functions of time and chemotherapy treatment is modified by delaying a course or reducing the dose intensity, according to patient’s toxicity levels. Models for time-to-event to deal with the dynamic nature of time-varying covariates during follow-up are necessary, still not well developed. In this work, innovative methods to represent time-dependent covariates by means of Functional Data Analysis (FDA) and how to include them into survival models are discussed. This new approach was applied to osteosarcoma data from the MRC BO06/EORTC 80931 randomized clinical trial, new insights into the clinical research. Time-varying covariates related to alkaline phosphatase (ALP) and chemotherapy dose during treatment were considered. Processes dynamics over time were investigated and additional information that may be related to the survival were included into the time-to-event models. High ALP levels reflected poor overall survival. Although dose-intense profiles were not associated with a better survival, the strength of our method is the ability to detect differences between patients with different biomarker evolution and treatment response, even when randomised to the same regimen. This aspect is seldom addressed in the literature.
This report, or a modified version of it, has been also submitted to, or published on
Statistical Methods & Applications, 2022. https://doi.org/10.1007/s10260-022-00647-0