Systematic review and meta-analysis of ML with cfDNA for cancer detection across stages

Photo by Ousa Chea / Unsplash

European journal of cancer (Oxford, England : 1990) Published May 15, 2026 Medically reviewed May 15, 2026 Study authors: Filis Panagiotis, Markozannes Georgios, Salgkamis Dimitrios, Tsiknakis Nikos, Zerdes Ioannis, Pagkal… PubMed ↗ DOI ↗ By Dr. Lars van Dijk, PhD · Surgical, Procedural & Diagnostic

Key Takeaway

Consider the variable sensitivity of ML-cfDNA models across cancer stages when evaluating diagnostic potential.

This is a systematic review and meta-analysis of 109 articles that synthesized the diagnostic performance of integrating machine learning with circulating cell-free DNA (cfDNA) analysis for detecting multiple solid tumors. The analysis included cancer patients across stages I to III and non-cancer controls.

The authors found that specificity was consistently high across all tumor types and stages, ranging from 94% to 99%. Sensitivity for stages I to III ranged from 72% to 92%. When stratified by stage, sensitivity ranged from 44% to 91% for stage I, 71% to 98% for stage II, and 83% to 99% for stage III. Among machine learning methods, neural networks showed the highest sensitivity at 90% (95% CI: 81%-95%), followed by random forest at 86% (95% CI: 77%-92%), heterogeneous ensemble learning at 85% (95% CI: 79%-89%), and fragmentation analysis at 86% (95% CI: 80%-90%). Methylation analysis yielded a sensitivity of 81% (95% CI: 76%-85%). Specificity for fragmentation and methylation analyses was 92%-96%.

The authors acknowledge limitations, including heterogeneity across studies and the absence of reported adverse events or follow-up data. The review does not report a specific study population size beyond the 109 articles, nor does it detail specific interventions or comparators beyond the general approach.

Practice relevance is not reported. The findings suggest potential diagnostic utility but require cautious interpretation pending further validation in prospective studies.

Study Details

Study typeMeta analysis

EvidenceLevel 1

PublishedMay 2026

PMID41930854

View Original Abstract ↓

INTRODUCTION: The latest generation of liquid biopsies incorporates multi-omic features, including genomics, methylomics, and fragmentomics. Machine learning (ML) approaches have been proposed to synthesize these complex biological data for the development of diagnostic classifiers. This study aims to evaluate the integration of ML with circulating cell-free DNA (cfDNA) analysis for early cancer detection. METHODS: Medline, Embase, Cochrane, and Web of Science were searched in July 2025. Eligible studies combined ML and cfDNA features to distinguish cancer patients (stages I-III) from non-cancer controls. Summary diagnostic performance metrics and their 95% confidence intervals (CI) were calculated. RESULTS: The study included 109 articles permitting analyses for lung (n = 34), liver (n = 29), colorectal (n = 28), pancreatic (n = 16), breast (n = 17), esophageal (n = 12), ovarian (n = 13), gastric (n = 9), head and neck (n = 4), and mixed (n = 27) cancer types. Specificity was consistently high across all tumor types and stages (94%-99%). Sensitivity ranged from 72% to 92% for stage I-III, 44-91% for stage I, 71-98% for stage II and 83-99% for stage III. In the pooled study population, neural networks (90%, 95% CI: 81%-95%), random forest (86%, 95% CI: 77%-92%) and heterogeneous ensemble learning (85%, 95% CI: 79%-89%) demonstrated the highest sensitivity. The stratified analysis by classifier feature revealed 86% (95% CI: 80%-90%) sensitivity for fragmentation and 81% (95% CI: 76%-85%) for methylation, with 92%-96% specificity. CONCLUSION: ML and cfDNA profiling show potential for early cancer detection, with ensemble methods, neural networks and random forests achieving the best overall performance. Fragmentomic features provide the highest sensitivity.