BACKGROUND: Glioma recurrence can be divided into in situ recurrence and non-in situ recurrence, and the mutation evolution of gliomas with different recurrence patterns is still unknown. We used sequential sequencing of circulating tumor DNA (ctDNA) to compare the somatic mutation profile and clonal evolution of gliomas with different recurrence patterns. To investigate the value of ctDNA in predicting early postoperative tumor recurrence and guiding prognosis stratification in patients with glioma. METHODS: We prospectively recruited 92 patients with near-total resection of gliomas from our center. Two hundred and thirty-four postoperative tissue and Tumor In Situ Fluid (TISF) samples from 69 eligible patients were included in ctDNA analysis. RESULTS: Among the 69 patients, 37 glioblastoma (GBM) patients experienced recurrence, and the median progression-free survival (mPFS) was not significantly different between the situ recurrence group and the non-in situ recurrence group (8.6 vs. 6.1 months). The ctDNA of recurrent tissue and TISF were significantly consistent. Before and after initial treatment, TISF-ctDNA mutant allele fraction (MAF), subclonal mutation, and alterations in related pathways (lysine degradation and PI3K pathway) were negatively correlated with treatment response and PFS. Among recurrent GBM patients, EGFR mutations were the most common. Mutations related to the RTK-RAS pathway (NF1) were most common in patients with situ recurrent GBM, while mutations in the MUC family and TP53 pathway (MUC16, CHEK2) were prevalent and continuously increased in patients with non-in situ recurrent GBM. CONCLUSIONS: In glioma patients undergoing primary surgery, dynamic monitoring of ctDNA and genotyping can be used for early risk stratification, efficacy monitoring, and early recurrence detection, and provide a basis for clinical research to evaluate early therapeutic intervention.