2043 Background: Unique radiobiological and physical properties of carbon ion radiotherapy (CIRT) may be favorably utilized to improve outcome in recurrent High-Grade Glioma (rHGG). There are currently no standardized criteria for stratification of rHGG patients for re-irradiation (re-RT). This study evaluated the impact of morphological data (radiomics) and physical information (dosiomics) in stratifying rHGG patients for CIRT. Methods: Quantitative radiomics and dosiomics features were extracted from CIRT planning CTs with dose distribution (DD) and multiparametric MRIs (mpMRI, pre re-RT) of 141 patients (recurrent grade III: n=56 40%, grade IV: n=85 60%) treated with a median dose of 42 Gy (RBE) and a median fraction of 13. The MR sequences considered are T1 weighted pre-and post-contrast agent, fluid-attenuated inversion recovery (FLAIR) and apparent diffusion coefficient (ADC). Benefit of a re-RT risk score (RRRS), comprising the initial tumour grade, age and the Karnofsky Performance Score was shown to correlate with superior outcome in CIRT and conventional re-RT and was also studied here in parallel. Feature sets - a) RRRS, b) radiomics, c) dosiomics features - were evaluated both separately and combined. Multiple feature selection methods were used independently on the CT, DD and the MR sequences, followed by a stepwise Cox's Proportional Hazard model selection per modality or combination thereof. Multivariable models were ranked by 10-fold cross-validated concordance index (C-I). Results: Compared to the RRRS model (OS/PFS, C-I: 0.68/0.61), the multimodality model considering radiomics and dosiomics features (RD) allowed improved prognostic separation (OS/PFS, C-I: 0.77/0.70). The RD signature consisted of 12 and 10 textural features for the OS and PFS models. Combining the RD model with RRRS yielded the best performance (OS/PFS, C-I: 0.78/0.73). No significant correlation between the textural features and the prescribed dose, tumor grade and volume was found, with the Spearman's correlation coefficient ranging between -0.06 to 0.17. Conclusions: Integrating multimodal information outperforms unimodal prognostic separation of rHGG following CIRT, highlighting the importance to consider biological, physical and morphological data for patient stratification. Prospective validation studies of this multimodal stratifier is warranted.[Table: see text]