Automated Support for Personalised Oncology Treatment Planning: A Case Study in Prostate Cancer
We are delighted to announce that the ONCO-TP project, a collaboration between the School of Medicine at the University of St. Gallen (HSG) and the department of oncology at the Kantonsspital St. Gallen (KSSG) has been awarded seed funding by the HSG Health Forward programme.
Summary
Oncology treatment planning should ideally be based on the best available evidence. The gold standard of evidence to inform treatment selection is randomised controlled trials. However, such trials are slow, expensive, and not always able to enroll patients to match the characteristics of all patients presenting in the clinic. Clinicians therefore often have to make decisions in the absence of sufficiently exact evidence. Therefore, there is a pressing need for automated systems that are able to extrapolate from the best available evidence to support clinicians making personalised treatment choices for individual patients. Treatment selection is informed by multiple relevant factors including patient characteristics, tumor molecular information, previous treatments, and available evidence at the time. The overarching objective of this proposal is to build the foundations for a comprehensive decision support tool for personalised oncology treatment planning, through a detailed case study of evidence synthesis, exploration of language-model-based automation approaches for evidence detection and ranking, and a feasibility evaluation of approaches to molecular data integration for treatment recommendation in metastatic castration resistant prostate cancer (mCRPC).
The proposed Health Forward research project will develop a detailed case study in mCRPC, beginning with (a) a systematic review and network meta-analysis of the available clinical trials data to inform the overall treatment possibilities and their relative effectiveness as well as to identify the trials from which to request individual patient data. We will then evaluate the feasibility of approaches such as individualised decision curve analysis for informing the personalised treatment recommendations for patients based on the individual patient data from trials. In addition, we will lay a comprehensive foundation for future treatment planning by (b) evaluating the feasibility of including molecular information in the treatment recommendation system based on assembling and integrating data from various molecular databases, and developing algorithms based on large language models for identifying and triangulating different types of evidence and ranking the association of molecular evidence to individualised treatments for a given patient.