Session Item

Diffuse intrinsic pontine glioma (DIPG) is a lethal pediatric brain tumor. Radiation therapy (RT) is the only effective treatment and standard of care. Reirradiation (rRT) remains an option at the time of progression. However, available data on prognostic factors for rRT are limited so far. The objective of our study is to review our experience of DIPG rRT and identify prognostic clinical and MRI-based radiomic features.

We collected patients with histopathologically proven DIPG who received rRT in our Department of Radiotherapy. We analyzed clinical, demographic, and radiological characteristics and compared them to the cohort of patients who underwent radiotherapy only once (non-reirradiated group). We delineated pre-rRT tumors using 3D Slicer software and extracted radiomic features from FLAIR MRI sequences using the python package PyRadiomics. We used the least absolute shrinkage and selection operator (Lasso) for Cox’s proportional hazard model to obtain the best features to predict the hazard ratio after rRT.

We analyzed 16 patients who received rRT between January 2016 and September 2022. They received a total dose ranging from 20 Gy to 24 Gy (in 2 Gy fractions). rRT was well tolerated, we did not record grade >2 toxicity. 75% of patients with neurological impairments improved after rRT. Median survival time from the end of rRT was 24 weeks (range 0 – 52 weeks). We compared the reirradiated group with our cohort of 22 DIPG non-reirradiated patients who were potentially eligible to receive reirradiation because of survival time longer than 3 months after the first irradiation. The median survival time from the end of the first radiotherapy was 66.5 weeks (range 25 – 156 weeks) in the reirradiated group versus 24.5 weeks (range 14 – 109 weeks) in the non-reirradiated group (p = 0.01). Figure 1 presents Kaplan-Meier curves for these groups.

We estimated Cox model regularized by Lasso penalty using clinical and pre-rRT MRI-based radiomic features. Our optimal model chose only radiomic features. The model demonstrated a concordance of 0.94 (95% CI: 0.89–0.99). The radiomic features that strongly predict the hazard ratio in the Cox model are linked to the tumor sphericity and homogeneity of grey levels. The hazard ratio is negatively associated with tumor sphericity and positively associated with structural homogeneity. Figure 2 presents the shapes of tumors from our dataset with good and poor prognosis according to the Cox model.

Reirradiation is a safe and effective treatment method in patients with progressive DIPG. The multiparametric MRI-based radiomic models could be useful for predicting survival in DIPG after rRT. Our results revealed that tumors with low sphericity and high homogeneity of grey levels are associated with a worse prognosis after rRT.