S830
Clinical - Lung
ESTRO 2026
Digital Poster Highlight 4054
Patient selection improves prediction of tumour volume regression in NSCLC, in a single parameter, Proliferation Saturation Index mathematical model Sarah Barrett 1,2 , Vincent Bourbonne 1,3 , Conor K McGarry 4,5 , Mohammad U Zahid 6 , Heiko Enderling 6,7 , Gerard M Walls 4,5 , Laure Marignol 1,2 1 Discipline of Radiation Therapy, Trinity College Dublin, Dublin, Ireland. 2 Trinity St. James’s Cancer Institute, Trinity College Dublin, Dublin, Ireland. 3 Radiation Oncology Department, , University Hospital of Brest, Belfast, Ireland. 4 Patrick G. Johnston Centre for Cancer Research, Queen’s University Belfast, Belfast, United Kingdom. 5 Northern Ireland Cancer Centre, Belfast Health & Social Care Trust, Belfast, United Kingdom. 6 Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, USA. 7 Institute for Data Science in Oncology, The University of Texas MD Anderson Cancer Center, Houston, USA Purpose/Objective: The Proliferation Saturation Index (PSI) model predicts Tumour Volume Regression (TVR) in NSCLC during conventionally fractionated RT using early treatment response data [1–4]. It incorporates a radiation sensitivity parameter ( α ), a tumour growth rate ( λ ), and a patient-specific PSI value. An underlying assumption of the model is that a tumour will shrink in response to RT. This study tests TVR prediction using tumour volumes extracted from the first 4 CBCTs acquired during RT, in an external cohort of patients with NSCLC, and seeks to examine the impact of patient selection on predictive performance. Material/Methods: A cohort of n=110 patients treated with 2Gy per fraction, to a radical total dose of 60–66Gy were included in this study. n=38 had concurrent chemoradiotherapy (CCRT), n=24 had RT alone, n=32 had induction chemo and n=16 had induction and CCRT.Tumours were contoured on all available CBCTs by a Radiation Oncologist, and resultant volumes extracted from the first 4 scans acquired during treatment delivery were input into the PSI model alongside previously derived parameter values of α = 0.3, and λ =0.012 [1]. Model predicted volume at final CBCT (CBCTFINAL) was compared to measured volume at the same timepoint, using the Coefficient of determination (R ² ) and Pearson correlation coefficient (PCC).As the model assumes that tumours will shrink in response to RT, a subgroup of patients, without a volume increase between CBCT-1 and CBCT-2 was evaluated for prediction of CBCTFINAL volume. Results: In this dataset, CBCT-4 was acquired on day 14 (median; range 4-32) and CBCTFINAL was acquired on
Conclusion: The α and λ parameters derived from the RT alone cohort did not generalise well to the new population, likely due to clinical and imaging differences. Notably, dataset 2 used 3DCT for planning, potentially introducing systematic uncertainty in pre-treatment growth rate estimation. This was somewhat mitigated in Dataset 1 by using AVIP reconstructions, which resemble CBCT more closely. Furthermore there was substantial variation in clinical management between the groups. Re-optimisation of parameters is recommended for patients undergoing CCRT, and further validation should be pursued across both RT- alone and CCRT cohorts. References: [1] Barrett S., Z.M.U., Enderling H., Marignol L., Predicting Individual Tumour Response Dynamics in Locally Advanced Non-Small Cell Lung Cancer Radiation Therapy; a Mathematical Modelling Study. IJROB 2024.[2] Sunassee, E.D., et al., Proliferation saturation index in an adaptive Bayesian approach to predict patient-specific radiotherapy responses. IJROPB 2019[3] Prokopiou, S., et al., A proliferation saturation index to predict radiation response and personalize radiotherapy fractionation. R&O 2015. [4] Barrett, S. et al. Monitoring lung tumour volume on daily cone beam CT; is it achievable in a real-world setting? TIPSRO 2025, V 36, 100352 Keywords: Tumour regression, mathematical modelling
Made with FlippingBook - Share PDF online