Optimising outcomes for potentially resectable pancreatic cancer through personalised predictive medicine : the application of complexity theory to probabilistic statistical modeling

  • Alison Bradley

Student thesis: Doctoral Thesis

Abstract

Survival outcomes for pancreatic cancer remain poor. Surgical resection with adjuvant therapy is the only potentially curative treatment, but for many people surgery is of limited benefit. Neoadjuvant therapy has emerged as an alternative treatment pathway however the evidence base surrounding the treatment of potentially resectable pancreatic cancer is highly heterogeneous and fraught with uncertainty and controversy. This research seeks to engage with conjunctive theorising by avoiding simplification and abstraction to draw on different kinds of data from multiple sources to move research towards a theory that can build a rich picture of pancreatic cancer management pathways as a complex system. The overall aim is to move research towards personalised realistic medicine by using personalised predictive modeling to facilitate better decision making to achieve the optimisation of outcomes. This research is theory driven and empirically focused from a complexity perspective. Combining operational and healthcare research methodology, and drawing on influences from complementary paradigms of critical realism and systems theory, then enhancing their impact by using Cilliers' complexity theory 'lean ontology', an open-world ontology is held and both epistemic reality and judgmental relativity are accepted. The use of imperfect data within statistical simulation models is explored to attempt to expand our capabilities for handling the emergent and uncertainty and to find other ways of relating to complexity within the field of pancreatic cancer research. Markov and discrete-event simulation modelling uncovered new insights and added a further dimension to the current debate by demonstrating that superior treatment pathway selection depended on individual patient and tumour factors. A Bayesian Belief Network was developed that modelled the dynamic nature of this complex system to make personalised prognostic predictions across competing treatments pathways throughout the patient journey to facilitate better shared clinical decision making with an accuracy exceeding existing predictive models.
Date of Award7 Jun 2021
Original languageEnglish
Awarding Institution
  • University Of Strathclyde
SupervisorRobert van der Meer (Supervisor) & Alec Morton (Supervisor)

Cite this

'