ESTRO 2025 - Abstract Book

S3861

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Conclusion: KM curves and Cox model showed that the Inferior Vena Cava might be a radiosensitive structure, with an impact visible despite the weight of the tumour volume, suggesting its consideration in future studies to determine a tolerable dose constraint.

Keywords: Cardio-toxicity, Propensity score matching

References: The study was supported by AIRC MFAG 27480.

4299

Proffered Paper External reproduction of a causal estimation of the individual and cohort benefit of sequential vs concurrent chemoradiotherapy in SIII NSCLC patients Charlie Cunniffe 1,2 , Wouter van Amsterdam 3 , Rajesh Ranganath 4 , Fiona Blackhall 1,5 , Matthew Sperrin 6 , Gareth Price 1,2 1 Division of Cancer Sciences, University of Manchester, Manchester, United Kingdom. 2 Radiotherapy Related Research, The Christie NHS Foundation Trust, Manchester, United Kingdom. 3 Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, Utrecht, Netherlands. 4 Courant Institute of Mathematical Sciences, New York University, New York City, USA. 5 Medical Oncology, The Christie NHS Foundation Trust, Manchester, United Kingdom. 6 Division of Informatics, Imaging and Data Science, University of Manchester, Manchester, United Kingdom Purpose/Objective: Randomised controlled trials (RCTs) provide the best evidence to guide treatment decision-making. However, Non Small Cell Lung Cancer (NSCLC) patients are often elderly, frail and from disadvantaged backgrounds, populations often under-represented in clinical trials. This means there are patient groups where there is uncertainty about which treatment strategy, e.g. sequential vs concurrent chemo-radiotherapy, will lead to the best outcomes. Using causal inference we can supplement RCT evidence with evidence from observational datasets. A challenge with observational causal inference is that important confounders, e.g. underlying patient fitness, are unobserved. Information in the patient record, such as performance score, provide proxy measurements of patient fitness, and can be used to infer unobserved confounding and allow estimation of treatment effects. In this study we use a recently introduced causal inference method to estimate individualised treatment effect (ITE) in a cohort of routinely treated NSCLC patients. Material/Methods: We employ the proxy-based individual treatment effect modelling in cancer (PROTECT) methodology to estimate the ITEs of concurrent vs sequential chemo-radiotherapy on overall survival1. A local model was developed adapting the PROTECT directed acyclic graph (DAG), figure 1, to include our selected proxies - performance score, comorbidity score and frailty score. The model was trained on 1117 SIII NSCLC patients treated between 2013 and 2023. ITEs for each patient are averaged to get the population's average treatment effect (ATE) estimate. We compared the ATE to standard multi-variable Cox regression, the ATE from the original PROTECT model 1 , and the results of a meta-analysis of RCTs 2 .