Tumor-informed versus tumor-agnostic ctDNA assays for molecular residual disease detection in colorectal cancer: A systematic review and meta-analysis.

3595 Background: Circulating tumor DNA (ctDNA) is increasingly used for detection of molecular residual disease (MRD) following curative-intent treatment for colorectal cancer (CRC). Two principal assay strategies exist: tumor-informed (TI) assays requiring prior tumor sequencing, and tumor-agnostic (TA) assays using fixed plasma-based panels. The relative prognostic and diagnostic performance of these approaches remain uncertain. Despite growing clinical adoption, no systematic comparison of their relative performance exists to guide assay selection. Methods: We performed a systematic review and meta-analysis in accordance with PRISMA 2020 guidelines. PubMed, Embase, Cochrane Library, and Web of Science were searched from inception to January 2025 for original studies evaluating ctDNA-based MRD detection after curative-intent treatment for CRC. Primary outcomes were hazard ratios (HRs) for recurrence-free survival (RFS) and diagnostic accuracy metrics (sensitivity and specificity). Random-effects models (DerSimonian-Laird estimator) were used to pool effect estimates; heterogeneity was assessed using I² statistics. Subgroup analyses were conducted by assay type (tumor-informed vs tumor-agnostic) and sampling strategy (landmark vs serial). Quality assessment was performed using QUADAS-2 for diagnostic studies and Newcastle-Ottawa Scale for prognostic studies. Results: Twenty-three studies comprising 9,847 patients met inclusion criteria. Eighteen studies evaluated tumor-informed assays (n=7,892) and eight evaluated tumor-agnostic assays (n=2,156), with three directly comparing both approaches. For tumor-informed assays, the pooled HR for recurrence in ctDNA-positive versus ctDNA-negative patients was 7.58 (95% CI, 6.21–9.25; I²=24%). For tumor-agnostic assays, the pooled HR was 4.12 (95% CI, 3.18–5.34; I²=18%). Subgroup analysis demonstrated significantly stronger prognostic discrimination for tumor-informed assays (ratio of HRs 1.84; 95% CI, 1.38–2.45; p<0.001). In serial sampling settings, tumor-informed assays achieved higher sensitivity for MRD detection than tumor-agnostic assays (88% vs 62%; p=0.002). At landmark timepoints, specificity was similar between approaches (97% vs 94%; p=0.08). Between study heterogeneity was low to moderate across analyses. Conclusions: Tumor-informed ctDNA assays demonstrate superior prognostic performance and higher sensitivity with serial sampling compared with tumor-agnostic approaches for MRD detection in colorectal cancer. TI assays may be preferred when longitudinal monitoring is feasible, while TA assays remain useful when tumor tissue is unavailable, or rapid results are needed. These findings support selection of ctDNA assay strategies based on clinical context.