ESTRO 2025 - Abstract Book

S3862

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2025

Results: The ATE estimate in our population is 0.92 [0.73, 1.15] in favour of concurrent chemo-radiotherapy. Standard multi variable analysis overestimates the effect of concurrent treatment compared to PROTECT and RCT data (figure 2, bottom). The RCT meta-analysis reports a larger effect size than causal analysis but has a younger (19% > 70 vs 43%) and fitter (50% ECOG 0 vs 27%) population than our study. ITE estimates (figure 2, top) suggest concurrent treatment may have improved survival in some patients treated sequentially.

Conclusion: Our results demonstrate causal analysis can provide more credible ATEs than standard multi-variable statistical approaches. The difference between our ATE estimate and the RCT meta-analysis can be explained by RCT inclusion criteria. Our ITE estimates suggest increased use of concurrent treatment might improve survival, but the model does not consider treatment toxicity. Further research into the use of causal models for counterfactual prediction on an individual patient basis is needed.