Background: Early prediction of therapeutic response is an important issue in glioblastoma patients. Cell-free DNA (cfDNA) concentration in plasma shown potential prognostic value. cfDNA release is influenced by non-tumor parameters. Short-length cfDNA (slDNA) may better reflect tumor evolution. The aim was to investigate plasmatic slDNA as prognostic marker in unresected glioblastoma patients. Methods: An ancillary study of a prospective trial was conducted (GLIOPLAK trial). Patients had newly diagnosed and histomolecular confirmed glioblastoma IDH wild-type (WHO 2021 classification). Patients with resected tumor were excluded. All patients underwent radiotherapy/temozolomide (RT/TMZ) schedule after biopsy. Plasmas were collected at three times during first-line treatment: before (pre-), after (post-) RT and at the time of disease progression/relapse. Cell-free DNA (cfDNA) was extracted from plasma using the QIAamp circulating nucleic acid kit (Qiagen). Short-length DNA (slDNA) fragments (equal or lower than 250 bp length) concentration was calculated based on AUC after automated electrophoresis using 4200 TapeStation System. Primary objective was to investigate the impact on survival of slDNA during RT/TMZ phase. Secondary objectives were to explore the association between tumor volume, corticosteroid exposition and slDNA. Results: Thirty-eight patients were included: median age was 63 [interquartile range 55-66], 84.2% (n=32) had Karnofsky index higher than 80% at baseline and median overall survival (OS) was 13.3 months [CI 95% 11.5; 15.5]. slDNA was detected in 69.4% at the pre-RT time (mean 722.3 pg/ml ±sd 1257.7) and in 83.3% at the post-RT-time (686.1 pg/ml ±1217.1). One third of the patients experienced a slDNA decrease between pre- and post-RT time. Presence of slDNA at the pre-RT time was associated with improved OS: median of 11.7 months [9.8-19.6] in the slDNA(+) group (n=25) versus 8.8months [5.5-NA] in the slDNA(-) group (n=11), HR 0.309 [0.133-0.716], log rank p=0.004. Decrease of slDNA concentration was also associated with better outcome compared to stability or increase of slDNA between pre- and post-RT: respectively median OS 14.1 vs 10.1 months, 0.424 [0.197-0.909], p=0.02. Corticosteroid exposition and dose were not correlated to slDNA concentration at pre-RT or slDNA variation (respectively correlation coefficient 0.081 and 0.047). At pre-RT, presence of slDNA was independent of tumor volume: mean 265.0 cm3 ±84.7 in slDNA(+) group vs 237.7 cm3 ±115.7 in slDNA(-) group, p=0.492. Total circulating cfDNA concentration did not influence OS in our cohort. Conclusions: slDNA during the RT/TMZ phase is an independent prognostic marker in unresected glioblastoma patients. The role of slDNA as a biomarker for adaptative treatment after RT as well as non-invasive molecular characterization tool can be explored in dedicated trial. Clinical trial information: NCT02617745.