Could Circulating Tumor DNA Offer a New Noninvasive Tool for CNSL Prediction?

According to new research presented during the 2021 ASH Annual Meeting, circulating tumor DNA (ctDNA) accurately reflects tumor burden and may serve as a useful clinical biomarker for risk, outcome, and surgery-free lymphoma classification in patients with central nervous system lymphoma (CNSL).
In this study, Florian Scherer, MD, of the University Medical Center Freiburg in Germany, and colleagues applied Cancer Personalized Profiling by Deep Sequencing (CAPP-Seq) and Phased Variant Enrichment and Detection Sequencing (PhasED-Seq) to tumor biopsies (n = 85), plasma samples (n = 131), and cerebrospinal fluid (CSF) specimens (n = 62) from a total of 92 patients with CNSL and 44 patients with other brain cancers or inflammatory cerebral diseases. The researchers targeted 794 distinct genetic regions.
The investigators correlated concentrations of ctDNA with radiologic tumor burden measures and evaluated associations with clinical outcomes at certain clinical time points. Using a supervised machine learning approach from tumor whole-genome sequencing data and genotyping analyses, the investigators also developed a novel classifier to distinguish CNSL from other CNS tumors noninvasively. The classifier was also based on mutational landscapes in both plasma and CSF. An independent validation was subsequently performed on the dataset.
All of the CNSL tumor biopsies featured genetic aberrations. The investigators reported a median of 262 mutations per patient. There was pretreatment plasma ctDNA detectable in 78% of plasma samples as well as 100% of CSF specimens. Concentrations of ctDNA ranged from 0.0004% to 5.94% allele frequency in plasma and 0.0049% to 50.47% allele frequency in CSF.
Compared with the ctDNA concentrations in a previously published cohort of patients with systemic diffuse large B-cell lymphoma, levels of plasma ctDNA were more than 200-fold lower in patients with CNSL. According to the researchers, there was a significant association between concentrations of ctDNA and total radiographic tumor volumes (TRTV) as measured by magnetic resonance imaging. No association was found between ctDNA concentrations and clinical risk scores or concurrent steroid therapy.
The assessment of ctDNA concentrations at pretreatment time points was predictive of progression-free survival (PFS) and overall survival (OS). Using ctDNA and TRTV as pretreatment biomarkers, the investigators were able to stratify patients into favorable or poor prognostic risk groups. During curative-intent induction therapy, ctDNA positivity was significantly associated with both PFS and OS.
The novel machine learning classifier was finally applied to 207 specimens obtained from an independent validation cohort of patients with and without CNSL. For the noninvasive diagnosis of CNSL, the system demonstrated high specificity (100%) and positive predictive value (100%). Additionally, there was a sensitivity CSF (57%) and plasma (21%). The researchers noted that these findings suggest that a significant proportion of patients with CNSL “might be able to forego invasive surgical biopsies” for the purposes of diagnosis.