Investigating the potential of real-world data to improve outcomes for patients with lung cancer

Abstract

There is increasing interest in using real-world data, routinely collected data relating to the health or healthcare delivery of patients, to generate evidence that has the potential to alter clinical decision making. Such real-world evidence could help to fill gaps in clinical knowledge, particularly for patients under-represented in clinical trials and for changes in radiotherapy workflows which occur as technology and techniques advance, often without clinical evidence to support the potential benefits. This thesis investigates the potential of real-world data to improve outcomes for patients with lung cancer through the analysis of routinely collected clinical and imaging data. First, the radiomics literature was reviewed to assess whether radiomics had the potential to personalise lung cancer treatment. The reviewed literature suffered from significant limitations, and no single radiomics biomarker or methodological approach was used widely, suggesting substantial barriers to clinical translation remain. Next, the reliability of radiomic features was assessed across four feature extraction platforms. It was found that choice of feature extraction platform, Imaging Biomarker Standardisation Initiative (IBSI) compliance, parameter settings and platform version affected feature reliability. This highlights the difficulty in trusting radiomics biomarkers, and the importance of using the latest version of an IBSI compliant software to ensure reproducibility of radiomics, a key requirement for clinical translation. The potential of real-world clinical data was then evaluated in the context of various retrospective changes to practice. First, the introduction of Intensity-modulated radiotherapy (IMRT) at The Christie NHS Foundation Trust was investigated, finding that the proportion of patients treated with curative-intent radiotherapy had increased and patient survival had improved following the introduction of IMRT. Second, the impact of the COVID-19 pandemic on outcomes for patients with lung cancer was evaluated, finding that patients who had a change to their radiotherapy or chemotherapy treatment did not have significantly worse survival or relapse rates compared to patients whose treatments were not changed; however, patients who had a change to their radiotherapy did have increased odds of grade 3+ acute toxicity. Finally, the potential of Bayesian methodology for assessing changes to clinical practice was investigated. A Bayesian analysis of a change to image-guided radiotherapy protocol found a reduced hazard of death for patients who had residual set-up errors towards the heart post-protocol change. This suggests the potential for Bayesian methodology to evaluate prospective incremental changes to practice. Together, these results demonstrate the potential real-world datasets have to monitor and improve outcomes for patients with lung cancer.

Date of Award	31 Dec 2023
Original language	English
Awarding Institution	The University of Manchester
Supervisor	James O'Connor (Supervisor), Corinne Faivre-Finn (Supervisor) & Gareth Price (Supervisor)