Value of Machine Learning Models for Cell-Free DNA-Based Multi-Cancer Early Detection: A Systematic Review and...

Area of Science:

• Biomarkers
• Machine Learning
• Oncology

Background:

• Machine learning (ML)-based analysis of cell-free DNA (cfDNA) is a promising strategy for multi-cancer early detection (MCED).
• Existing performance estimates often stem from training or enriched cohorts, limiting real-world applicability.
• Independent validation is crucial to assess the true diagnostic accuracy of these assays.

Purpose of the Study:

• To conduct a systematic review and meta-analysis of ML-based cfDNA assays for MCED.
• To evaluate diagnostic accuracy using only independent validation or testing datasets.
• To identify factors influencing assay performance and heterogeneity.

Main Methods:

• Systematic review and diagnostic accuracy meta-analysis of 13 studies.
• Inclusion of 23 independent datasets from 14,892 participants, excluding all training data.
• Bivariate random-effects model used to estimate pooled sensitivity, specificity, and diagnostic odds ratio (DOR), with subgroup analyses to explore heterogeneity.

Main Results:

• Pooled sensitivity of 0.78 (95% CI: 0.66-0.87) and pooled specificity of 0.96 (95% CI: 0.90-0.98).
• Summary area under the curve (AUC) of 0.94 and DOR of 76.6.
• Significant heterogeneity (I² > 90%) observed, influenced by geographic region, sample size, and cfDNA biomarker type.

Conclusions:

• ML-based cfDNA assays exhibit high specificity and moderate-to-high sensitivity in independent validation settings for MCED.
• Diagnostic performance is context-dependent, influenced by study design, population characteristics, and analytical choices.
• Large-scale, prospective, population-based validation is necessary prior to widespread clinical implementation.