LIQUID BIOPSY FOR EARLY CANCER DETECTION: A SYSTEMATIC REVIEW OF CFDNA, CTDNA, METHYLATION, AND FRAGMENTOMIC BIOMARKERS

Abstract

Background: Cancer diagnosed before dissemination is substantially more amenable to curative treatment; however, population screening is currently available for only a limited number of malignancies. Blood-based liquid biopsy has consequently emerged as a potential strategy for detecting molecular evidence of cancer before conventional clinical presentation. Circulating cell-free DNA (cfDNA) provides several distinct layers of information, including tumor-derived mutations, DNA methylation changes, copy-number alterations, fragment-size distributions, nucleosome-associated patterns, and fragment-end signatures. Objective: This systematic review critically evaluates cfDNA quantity and integrity, circulating tumor DNA (ctDNA) mutations, methylation signatures, and fragmentomic biomarkers for early cancer detection, with particular emphasis on recent independent-validation and multi-cancer studies. Methods: A structured literature search was performed for studies available through January 2026 using combinations of terms relating to cell-free DNA, circulating tumor DNA, DNA methylation, fragmentomics, early cancer detection, and multi-cancer detection. A total of 1,066 records were identified through database and supplementary searching. After removal of 289 duplicate or otherwise ineligible records, 777 records underwent title and abstract screening. Of these, 635 records were excluded, and 142 reports were sought for retrieval. Five reports could not be retrieved, leaving 137 full-text reports for eligibility assessment. Following exclusion of 111 full-text reports, 26 primary studies were included in the qualitative synthesis. Human studies evaluating blood-based cfDNA biomarkers for detection of treatment-naive, localized, non-metastatic, or preclinical malignancy were prioritized. Diagnostic performance, stage-specific sensitivity, specificity, area under the receiver operating characteristic curve (AUC), cancer-signal-origin performance, study design, and validation strategy were extracted. Methodological limitations were assessed using principles of QUADAS-3. Owing to substantial clinical and analytical heterogeneity, a quantitative meta-analysis was not performed. Results: A total of 26 primary studies were included in the final qualitative synthesis. Mutation-based ctDNA assays demonstrated high biological specificity but were limited by low circulating tumor fractions in small cancers and confounding by clonal hematopoiesis. Methylation assays provided a larger number of tumor-associated loci and generally stronger tissue-of-origin information. Recent enzyme-assisted methylation testing in the INSPECTOR study achieved 81.9% sensitivity and 99.0% specificity in independent validation, with stage I sensitivity of 65.5%. The 2024 TOTEM methylation assay showed 55.9% sensitivity and 98.6% specificity in independent validation. Fragmentomic technologies demonstrated strong discrimination using shallow whole-genome sequencing: a 2025 multidimensional fragmentomics study reported 87.4% sensitivity, 97.8% specificity, and 82.4% tissue-of-origin accuracy in an independent cohort. In a subsequent prospective cohort of asymptomatic individuals, sensitivity was lower at 53.5%, emphasizing the difference between case-control and screening performance. A pancreatic cancer fragmentomic classifier achieved an AUC of 0.987 in validation, while multimodal assays combining methylation, mutations, proteins, or fragmentomic signals improved detection compared with single-feature approaches. Conclusion: The field has shifted from measuring total circulating DNA toward extracting multidimensional genomic and epigenomic information from cfDNA molecules. Methylation and fragmentomics currently offer the broadest signals for early-stage detection, whereas mutation analysis contributes high molecular specificity. Multimodal approaches are particularly promising. Nevertheless, prospective population-based validation, standardization of pre-analytical procedures, robust control of confounding, and demonstration of clinical benefit remain essential before widespread adoption.