Session Item

The aims of this study were: i) to analyze the relationship between radiation induced temporal lobe necrosis (TLN) and dosimetric features in patients with skull base tumors treated with pencil beam scanning proton therapy (PT); ii) to test available normal tissue complication probability (NTCP) models.

Clinical and dosimetric data of 110 adult consecutive skull base chordoma and chondrosarcoma patients treated at CNAO between September 2011 and July 2020 were retrospectively analyzed. A median dose of 70 Gy (RBE) (range 70-74 Gy (RBE)) was planned and delivered in 35-37 daily fractions. Current treatment planning goal for TL was limiting the maximum dose to the hottests 2 cm³ (D_2cc ) to 71 Gy (RBE). Magnetic Resonance Imaging (MRI) studies showing TLN at the first occurrence were co-registered with the original planning images. TLN was graded using Common Terminology Criteria for Adverse Events (CTCAE) v 5.0. The dosimetric parameters of TLs (Figure 1) were investigated for finding a correlation with the TLN’s occurrence and severity and then included in NTCP univariate models. Their classification and predictive performances were evaluated by means of goodness  of fit (R²); Area Under ROC Curve (AUC); accuracy, sensitivity and F-score.

Figure 1 - Parameters extracted from DVHs for NTCP model evaluation.

Median follow-up was 36 months (range 9-98 months). Grade G1 TLN was reported in 26 patients (24%), while G2 in 14 cases (13%). No TLN > G2 was recorded. 95% of patients had TLN within 50 months from the end of treatment. Among the dosimetric parameters analyzed, D_0.5cc, D_2cc and D_Max better correlate with the risk of TLN. The corresponding NTCP curves are reported in Figure 2. For the G2 TLN models based on D_0.5cc, D_2cc and D_Max, R² was 0.88, 0.95 and 0.92 whereas AUC was 0.80, 0.82 and 0.79, respectively. Moreover, when considering the TD reproducing experimental G2 TLN occurrence, the  D_2cc -based TLN model reached the best predictive performance (accuracy = 0.71, sensitivity = 0.82, F-score = 0.33). Tolerance dose at 2 cm³ of TL for 5% and 20% probability of developing G2 TLN in 5 years were 62.9 Gy (RBE) and 72 Gy (RBE), respectively, confirming the validity of the dose constraint used for plan optimization.  

Figure 2 - Figure 1 - D2cc, D0.5cc, DMax NTCP models from univariate analysis. Shading limits the region within the 95% confidence interval. Solid line-open circles (yellow) TLN G > 0; dashed line-filled circles (red) TLN ≥ G2.

Patterns of TLN occurrence in patients with skull-base chordoma and chondrosarcoma underwent PT were comparable with literature data. High doses to very small TL volumes were the major predictors of TLN. This analysis showed that the TL dose constraint, currently in use in clinical practice, correctly predicted TLN G2 occurrence.