Diagnostic accuracy of circulating tumor DNA for post-remission surveillance in diffuse large B-cell lymphoma: A systematic review and meta-analysis.

Abstract

7060 Background: Relapse after complete remission remains a major cause of treatment failure in Diffuse Large B-Cell Lymphoma (DLBCL). Standard surveillance using clinical assessment and imaging often detects recurrence after substantial tumor regrowth. Circulating tumor DNA (ctDNA) has emerged as a promising biomarker for minimal residual disease detection; however, its diagnostic accuracy for relapse surveillance in DLBCL has not been systematically quantified. Methods: A systematic review and meta-analysis was performed per a prespecified PROSPERO protocol (CRD420251179858). MEDLINE, Embase, CENTRAL, and major conference proceedings were searched through October 2025. Following systematic screening of B-cell non-Hodgkin lymphoma studies, those eligible for quantitative synthesis enrolled adults with DLBCL in complete remission undergoing post-treatment ctDNA surveillance, with clinically or radiologically confirmed relapse as the reference standard. Studies were included if sufficient data were available to construct 2×2 contingency tables. Pooled sensitivity and specificity were estimated using a bivariate random-effects logistic regression model with a logit link, accounting for between-study heterogeneity. A continuity correction of 0.5 was applied to all cells in studies with zero counts prior to model fitting. Results: Three studies (N=157) comprising patients in complete remission after first-line therapy were included in the quantitative synthesis. Using a bivariate random-effects model, the pooled sensitivity of ctDNA for detection of clinical relapse was 58.6% (95% CI 32.2–80.8%), while pooled specificity was 93.9% (95% CI 83.6–97.9%), indicating high specificity but moderate sensitivity during surveillance. Specificity was consistently high, supporting strong rule-in performance. Sensitivity varied across studies (range: 31%–88%), likely reflecting differences primarily in sampling density, as well as assay platforms (tumor-informed Ig-HTS vs. targeted NGS). Across included cohorts, ctDNA positivity generally preceded or coincided with clinically detected relapse, while ctDNA negativity was associated with durable remission during follow-up. Conclusions: In patients with DLBCL in complete remission after first-line therapy, ctDNA surveillance demonstrates very high specificity but only moderate sensitivity for detection of clinical relapse. These findings support ctDNA as a highly specific rule-in biomarker for relapse during surveillance, potentially resolving ambiguity in equivocal imaging findings. However, its moderate sensitivity limits its standalone ability to detect relapse. Prospective validation of standardized testing intervals and assay harmonization are needed to optimize sensitivity before integration with standard monitoring strategies.