BACKGROUND: GBM genomic profiling relies on sequencing a limited tumor tissue sample, which is invasive and may underrepresent molecular heterogeneity. CSF, which can be obtained less invasively, is in intimate contact with tumor lesions and may better capture the full genomic profile of the tumor. However, datasets with contemporaneously collected CSF and tissue to support this claim have been lacking. To evaluate the performance of CSF NGS, we conducted a pilot study in patients with GBM undergoing a resection for suspected recurrence following first-line chemoradiotherapy.

METHODS: Paired CSF and tissue samples were sequenced using a hybrid capture-based NGS assay. Clinically meaningful variants were defined as those that are potentially targetable using off-label or clinical trial options, exhibit prognostic value, or may predict response or resistance to specific treatments.

RESULTS: Eighteen patients were enrolled, and 13 of 18 CSF samples (72.2%) were sequenced successfully. At least one variant was detected in all CSF samples analyzed. A median of 7 variants (range 1—67) was detected per sample across 54 genes. The median variant allele fraction was 0.6% (range 0.2—72.4%). Among 38 clinically meaningful genes, 102 variants were detected; 25 (24.5%) were detected in both tissue and CSF, while 60 (58.8%) were detectable solely in CSF. Hypermutation was detected by CSF in one patient. Of the 82 variants detected in this patient’s tumor, 15 (18.3%) were identified in both tissue and CSF, 15 (18.3%) were identified only in the tissue, and 52 (63.2%) were identified only in the CSF.

CONCLUSIONS: CSF NGS detects clinically meaningful variants at a substantial rate and frequently identifies mutations not detected by matched tissue NGS. These results suggest that CSF may be a suitable source material for tumor profiling, overcoming the limitations of tissue, and may also provide a more comprehensive tumor profile.

VIEW THE POSTER