AUTHORS
Jinhua Liang, PhD, Yi-Fan Wu, PhD2, Yi-Lun Ji3, Yue Li4, Liuqing Zhu5, Hua Yin6, Li Wang7, Jianyong Li, MD8, Xu Wei, MD, PhD9
AFFILIATIONS
1 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
2 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
3 Nanjing Geneseeq Technology Inc., Nanjing, China
4 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
5 Nanjing Geneseeq Technology Inc., Nanjing, China
6 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
7 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
8 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
9 The First Affiliated Hospital of Nanjing Medical University, Jiangsu Province Hospital, NANJING, China
ABSTRACT
Background: The role of central nervous system (CNS) prophylaxis in patients with diffuse large B-cell lymphoma (DLBCL) remains controversial, largely due to the lack of prospective randomized controlled trials. Recent large retrospective studies, including the study by Wilson MR et al. (JCO, 2026), have shown that high-dose methotrexate (HD-MTX) prophylaxis does not significantly reduce CNS relapse rates, even among patients classified as ultra-high risk. These findings, together with accumulating evidence, have raised concerns about the routine use of HD-MTX for CNS prophylaxis and underscored the need for more accurate risk stratification to identify patients at the highest risk of CNS progression. In this context, cerebrospinal fluid circulating tumor DNA (CSF-ctDNA) has emerged as a promising biomarker for identifying patients who are likely to develop CNS relapse.
Aims: This study aimed to evaluate, in a large-cohort setting, the ability of CSF-ctDNA to predict future CNS relapse in DLBCL compared with the established CNS-IPI risk score.
Methods: Between December 2021 and January 2026, 410 newly-diagnosed DLBCL (ND-DLBCL) patients without overt CNS involvement at baseline were enrolled in the cohort. In this cohort, biospecimens (tumor tissue, plasma and CSF samples) were collected at baseline and analyzed using a 475-gene lymphoma-specific panel (NGS was performed via ATG-Seq with duplex UMIs and Monte Carlo denoising) after patients completed the induction therapy. These results did not affect any clinical decisions regarding induction therapy or CNS prophylaxis. All patients received frontline therapy with R-CHOP or Pola-R-CHP regimens. CNS prophylaxis (intrathecal chemotherapy or high-dose methotrexate) was administered for patients classified as NCCN-defined CNS high-risk group (≥1 of the following criteria: CNS international prognostic index (CNS-IPI) score 4-6; testicular, renal/adrenal; or breast involvement). Patients were identi?ed as CSF-ctDNA positive based on either of the following criteria: (1) at least one mutant variant detected in CSF-ctDNA was shared with tissue; (2) at least one mutant variant detected in CSF-ctDNA was canonical driver mutation as previously reported in DLBCL. This criteria were validated in previous studies. The median follow-up was 34.1 months (interquartile range, 19.9-42.7 months).
Result: Of the 410 ND-DLBCL patients, 98 (23.9%) were identified as CSF-ctDNA positive. Among the 144 patients (144/410, 35.1%) who were classified as NCCN-defined CNS high-risk, only 57 patients (57/144, 39.6%) were defined as CSF-ctDNA positive. Among the 98 CSF-ctDNA positive patients, the top 5 frequently mutated genes detected in CSF were MYD88 (26.5%), CD79B (20.4%), KMT2D (18.4%), PIM1(18.4%) and BCL6 (15.3%). Among all genetic subtypes, MCD subtype showed highest CSF-ctDNA(+) ratio (37/108, 34.3%).
By the last follow-up date, CNS progression had occurred in 18 patients (18/410, 4.4%). All 18 patients (100.0%) were positive for CSF-ctDNA at baseline, whereas only 10 (55.6%) of them met the NCCN guideline criteria for high-risk status. There was no statistically significant difference in the CNS relapse rate between NCCN guideline criteria for high-risk status and low-risk status (3-year rate, 6.8% vs 3.3%, P=0.066) while statistically significant differences were observed between CSF-ctDNA positivity and negativity at baseline (3-year rate, 19.6% vs 0.0%, P<0.001). This proportion may be confounded by the use (or non-use) of CNS prophylaxis.
Interestingly, similar conclusions were reached in the subgroup after accounting for confounding factors. CSF-ctDNA positive patients had a significantly higher 3-year CNS relapse rate than CSF-ctDNA negative patients in both the NCCN high-risk group receiving CNS prophylaxis (n=144, 17.3% vs 0.0%, P<0.001) and the non-high-risk cohort without CNS prophylaxis (n=266, 23.4% vs 0.0%, P<0.001).
Multivariable analysis confirmed that CSF-ctDNA positivity rather than NCCN-defined CNS high-risk status remained an independent predictor of CNS relapse after adjustment for non-GCB subtype, involvement of high-risk sites, CNS-IPI score 4-6 and receiving CNS prophylaxis, highlighting the superior predictive value of CSF-ctDNA for CNS relpase.
Conclusion: In summary, findings from our study strongly suggest that CSF-ctDNA could serve as a sensitive and specific biomarker for CNS relapse prediction, with the potential to guide risk-adapted CNS prophylaxis strategies.
DISCLOSURES
Jinhua Liang, PhD: No relevant conflicts of interest to declare.