ESTRO 2024 - Abstract Book

S5050

Physics - Radiomics, functional and biological imaging and outcome prediction

ESTRO 2024

Figure 2. Plot of toxicity risk for the cumulative V20 univariable model. Plot presentational details are as for Figure 1.

Conclusion:

We developed two univariable dose-response models for ≥grade 3 lung toxicity following reirradiation, based on cMLD and cV20Gy. Predictions of both models were well correlated with the toxicity rates the models were built from (Pearson coefficients >0.7). Multivariable modelling insignificantly improved toxicity prediction, due to data collinearity. cMLD and cV20Gy levels associated with 20% toxicity rates were both lower than suggested in the expert consensus constraints, but were similar to the constraints for initial radiotherapy. This suggests there is little lung recovery. The model predictions have wide confidence intervals and future models may include some factors for which we lacked data (e.g. lung function, smoking status). Thus, further data is required to validate and refine our estimates, but limiting cMLD and cV20Gy following reirradiation to the dose limits used in initial radiotherapy may be prudent.

Keywords: Re-irradiation, lung cancer, lung toxicity

References: