The fortnightly BRAG meeting will be held this Thursday the 31st of October at 1 pm via Zoom/GP-Y801. This week we will have a presentation by Mildred Mmbone Lumwamu along with a discussion.
Zoom link: https://qut.zoom.us/j/89563332139?pwd=ILeHs644gDYNFOZphKab5e8ALldb9Q.1
Meeting ID: 895 6333 2139
Passcode: 292233
Mildred’s Talk
Title: New Statistical and Machine Learning Insights into Cancer
Abstract: Accurate prediction of survival time in cancer patients, particularly those with comorbidities, is crucial for improving outcomes and personalising treatment strategies. Patients with comorbidities are often excluded from traditional clinical trials, resulting in limited information about their survival time and the specific impact of comorbidities on survival outcomes. This study aims to develop and compare advanced survival models to enhance predictive accuracy of survival time particularly for cancer patients with comorbidities. The first approach employs a one stage and novel two stage flexible parametric survival model (FPSM) that addresses the unique challenges of comorbidity data and quantify its effect on survival. A one stage and novel two stage random survival forest (RSF) will also be employed. This non-parametric approach is well- suited for capturing complex interactions between variables such as comorbidities which are challenging to model through traditional approaches. To mitigate model and parameter uncertainty, predictions from the FPSM and RSF models will be integrated to enhance robustness and predictive performance. The research focuses on two case studies: (1)The Breast Cancer Outcomes Study (BCOS) cohort consisting of 3,323 women diagnosed with invasive breast cancer in Queensland aged between 20 and 79 years, with follow-up until December 2018 and (2) a longitudinal cohort of Queensland residents diagnosed with colorectal cancer, followed until 2019. The case studies will facilitate the development and implementation of statistical methodology for the design of a target trial framework, leveraging observational data to emulate a randomized clinical trial. The aim will be to inform treatment guidelines for patients with comorbidities, who are often excluded from traditional clinical trials. The ultimate goal of this study is to enhance cancer survival predictions, particularly for individuals with comorbidities, by advancing statistical methodology. This research will provide new insights into the application and extension of state- of-the-art statistical models, offering methodological innovations that contribute to the field of survival analysis.
Thanks,
Abi & Jamintha