The word and the way: strategies for domain-specific BERT pre-training in German medical NLP.

Glioblastoma is an aggressive primary brain tumor with an unfavorable prognosis. In unselected contemporary cohorts, median overall survival is generally between 12 and 18 months, whereas selected molecular subgroups may experience substantially longer survival [–]. It remains among the solid tumors with the shortest survival despite multimodal therapy. Management of newly diagnosed glioblastoma typically includes maximal safe tumor resection, followed by combined radiotherapy and temozolomide chemotherapy. In selected patients, treatment may be supplemented with tumor-treating fields. Since the introduction of radiotherapy plus temozolomide in 2005, only a small number of randomized phase III trials have demonstrated an overall survival benefit in newly diagnosed glioblastoma, largely driven by alkylating regimens rather than targeted or immune therapies []. Nearly all patients ultimately develop disease progression despite initial therapy. At recurrence, management is individualized and may include repeat surgery, reirradiation, and systemic agents, but evidence is heterogeneous and no universally accepted overall-survival–prolonging standard exists []. Given the poor prognosis and high neurological morbidity, a glioblastoma diagnosis can profoundly disrupt emotional well-being, physical function, and social roles [, ]. Compared with many other cancers, brain tumors uniquely threaten cognition, communication, behavior, and autonomy, complicating decision-making and social participation [, ]. Psycho-oncological consultations frequently reveal high psychological burden, with many patients expressing concerns about death and a strong need for medical information []. Hence, patients with glioma frequently experience psychiatric comorbidities such as depression and anxiety, which are associated with diminished quality of life and may negatively affect treatment adherence and survival. By comparison, suicidality remains understudied and is rarely assessed in a standardized manner [–]. Depressive symptoms may reflect direct neurobiological and treatment-related effects (e.g., lesion location, inflammation, corticosteroid exposure) and/or a psychological response to diagnosis and functional loss []. Longitudinal trajectories of depression and anxiety after surgery are heterogeneous, with some patients improving transiently and others worsening as treatment burden and neurological disability accumulate [, ]. Suicide mortality is elevated in cancer patients compared with the general population, with the magnitude of risk varying by cancer type and being particularly high in poor-prognosis tumors []. Data on suicidality in glioblastoma specifically are limited. Large population-based analyses indicate higher suicide mortality after cancer diagnosis, with risk varying by tumor type and time since diagnosis; poor-prognosis central nervous system (CNS) malignancies consistently rank among higher-risk groups []. Existing literature suggests that suicidality in brain tumor populations likely reflects interacting neuroanatomical, treatment-related, inflammatory, and psychosocial factors, although evidence is heterogeneous and often observational. While inflammatory biomarkers such as tumor necrosis factor alpha (TNF-α) and interleukin-6 (IL-6) have shown associations with suicidal ideation in psychiatric samples, evidence regarding their significance in glioblastoma is limited, and tumor-specific studies are needed to clarify their role [–].