Circulating cell-free DNA (cfDNA) refers to fragmented DNA released into the bloodstream through apoptosis, necrosis and active secretion. In patients with cancer, a fraction of cfDNA derives from tumour cells and is termed circulating tumour DNA (ctDNA). ctDNA carries the genetic and epigenetic signatures of the tumour, providing a non-invasive “liquid biopsy” that can reflect tumour dynamics, heterogeneity and treatment response without repeated tissue sampling [1].

cfDNA circulates as free fragments, as DNA bound to proteins or nucleosomes, and within extracellular vesicles. ctDNA analysis typically targets tumour-specific alterations such as single-nucleotide variants, copy-number alterations and epigenetic modifications, including aberrant DNA methylation, which is often tissue-specific and therefore valuable for diagnosis and prognosis [1]. Across the care continuum, cfDNA has been explored for early detection, primary-tumour characterisation, minimal-residual-disease (MRD) assessment, treatment-response monitoring and early recurrence detection; ctDNA levels can track tumour burden more closely than protein biomarkers such as CA 15-3 [2].

Breast cancer is the most common cancer among women in Southeast Asia, but incidence, risk factors and health-system capacity vary widely across countries [3]. Reported patterns include comprehensive registry data and rapid urbanisation in Singapore, heterogeneous incidence and access in Thailand, likely under-representation of rural burden in the Philippines and Indonesia, and emerging real-world ctDNA experience in Vietnam [3][4]. Region-specific challenges include a high prevalence of radiologically dense breast tissue, an earlier peak incidence (often in the 40s), frequent late-stage presentation, and resource limitations affecting screening infrastructure and specimen quality [3][5][6].

Mammographic sensitivity is markedly reduced in dense breasts. Reported sensitivity was 47.4% in asymptomatic Japanese women in their 40s in the J-START study and 54% - 67% for Korean women aged 45 - 49 [5]. Supplemental ultrasound and MRI can improve detection but are constrained by cost, availability and interpretive expertise, particularly in rural areas [3][5]. These limitations motivate complementary, blood-based approaches; importantly, the mammography evidence cited here derives from East Asian (Japanese and Korean) cohorts rather than from Southeast Asia specifically, a distinction we maintain throughout (Section 5).

The convergence of dense breast tissue, earlier onset and imaging limitations creates a clear rationale for cfDNA-based approaches, which are independent of breast density [7]. This review synthesises evidence on cfDNA in early detection, treatment monitoring and precision management of breast cancer, with emphasis on Southeast Asia. Throughout, we (i) separate analytical validity, diagnostic accuracy and clinical utility; (ii) distinguish breast-specific from multi-cancer evidence; (iii) qualify outcome claims; and (iv) distinguish Southeast Asian from other Asian and global data. We frame cfDNA assays as complements to, not replacements for, established breast-imaging pathways.

Analytical validity concerns whether an assay accurately and reproducibly measures the analyte it targets. Multi-omics platforms such as SPOT-MAS (Screening for the Presence of Tumour by Methylation and Size) integrate next-generation sequencing (NGS) and machine learning to interrogate genetic, fragmentomic and epigenetic features of ctDNA [8][9]. An analytical-validation study reported a cancer-signal sensitivity of 73.9% and specificity of 95.9% across 285 patients with five cancer types versus 222 healthy individuals [10]. This is a multi-cancer analytical-validation estimate in a case-control design, not a breast-specific screening result, and case-control designs are known to overstate accuracy relative to intended-use populations.

Diagnostic accuracy concerns whether assay results correspond to disease status in the intended population. Here it is essential to separate breast-specific from multi-cancer evidence, and to note stage, comparator and cohort type. The large prospective K-DETEK cohort (9024 asymptomatic participants in Vietnam) reported an overall sensitivity of 70.8% for cancer and a specificity of 99.7% with a negative predictive value of 99.92%, but these are aggregate multi-cancer figures across all screened cancers, not breast-specific, and should not be read as breast-cancer screening performance [8]. Breast-specific classifiers report an area under the curve (AUC) of about 0.90 for distinguishing breast cancer from non-cancer, with stage-dependent sensitivity (62.3% Stage I, 73.9% Stage II, 88.3% Stage IIIA), reflecting the difficulty of detecting early, low-shedding disease [11]. Table 1 disaggregates the principal estimates by test type, population and stage, comparator and cohort type.

Table 1. Principal cfDNA early-detection performance estimates, disaggregated by assay type (breast-specific vs. multi-cancer), population and stage, comparator, and cohort type (screening vs. clinically ascertained).

Clinical utility concerns whether using the test improves patient-relevant outcomes. This is the least mature domain: no breast-cancer cfDNA screening assay has yet demonstrated, in prospective controlled studies, that its use reduces late-stage diagnosis or mortality relative to standard pathways. Accordingly, early-detection cfDNA assays should currently be regarded as complementary to, rather than replacements for, established breast-imaging pathways. In practice, this means a cfDNA signal should trigger, not bypass, confirmatory imaging (mammography with ultrasound or MRI as indicated) and tissue biopsy, and a negative cfDNA result should not be used to defer guideline-recommended screening.

Aberrant DNA methylation is a hallmark of cancer, and cfDNA methylation signatures can be detected before clinical diagnosis. Using cfDNA methylation immunoprecipitation sequencing (cfMeDIP-seq) on pre-diagnosis samples, one study identified signatures predicting breast cancer up to six years before clinical detection, with an AUC of 0.930 in an independent test set [12]; this is a nested pre-diagnostic case-control result requiring prospective validation before clinical use. Fragmentomics fragment size, distribution and end motifs adds orthogonal information; cytosine-starting 4-mer end motifs (e.g., CGCC, CCCC) are enriched and certain guanine-starting motifs depleted in breast-cancer cfDNA, and these features improve machine-learning classifiers [11].

Interpreting the estimates above requires attention to well-recognised error sources that are especially consequential in screening, where disease prevalence is low.

Clonalhaematopoiesis(CH/CHIP). Somatic mutations arising in expanded haematopoietic clones are shed into plasma and are a major cause of false-positive ctDNA calls, particularly for larger panels covering genes such as TP53; in one high-intensity sequencing study a majority of plasma variants in cancer patients had features consistent with clonal haematopoiesis [13][14]. Sequencing paired white-blood-cell DNA is the standard mitigation and should be built into screening assays [14][15].Low ctDNA shedding in early-stage disease. ctDNA constitutes only a small fraction of total cfDNA, and this fraction is lowest in early-stage and minimal-residual settings; limited plasma genome-equivalents cap the detectable variant-allele frequency and drive false negatives, and some tumours shed little ctDNA even when advanced [15]. This biological floor explains the stage-dependent sensitivity in Table 1 and cautions against over-interpreting a negative result.Pre-analytical variation. Blood-collection tube type, draw-to-processing interval, temperature, centrifugation and extraction method all affect cfDNA yield, integrity and background from leukocyte lysis, producing both false negatives and spurious signals; without harmonised pre-analytical protocols, cross-study performance is not directly comparable [16][17].

Plasma cfDNA methylation/promoter profiles can predict pathological complete response (pCR) to neoadjuvant chemotherapy; a Random Forest model achieved an AUC of 0.980 with 95.3% accuracy for predicting pCR [18]. Serial ctDNA can track tumour burden during therapy and, in metastatic disease, correlates with response more dynamically than CA 15-3, including earlier detection of progression and emerging resistance during CDK4/6-inhibitor therapy [2]. These are predictive/prognostic associations; whether acting on them improves outcomes is addressed in Section 4.4.

In Southeast Asian cohorts, tumour-informed ctDNA assays show high prognostic accuracy. For 110 early-stage patients, the K-TRACK assay predicted recurrence with 80.0% sensitivity and 98.3% specificity, with lead times up to 11.0 months before clinical or radiological recurrence [19]. A higher tumour fraction (e.g., >10%) is associated with worse survival and a low fraction (<1%) with better real-world overall survival, including in bone-only metastatic disease [2]. Lead time and prognostic separation are consistently demonstrated; a demonstrated survival benefit from earlier, ctDNA-triggered intervention is not (Section 4.4).

Liquid biopsy can detect resistance-conferring alterations before clinical progression. Rising ESR1 mutations can be tracked in ER-positive metastatic disease on aromatase inhibitors; the PADA-1 trial, one of the few interventional data sets in this area, showed that switching therapy upon ESR1 detection improved progression-free survival [2]. In a Southeast Asian real-world study, ESR1 mutations were found in 12.5% of metastatic ER-positive patients (variants including Y537N, Y538G) [19]. ERBB2 (HER2) amplification in ctDNA has been associated with high response rates to anti-HER2 therapy when tissue is unavailable [20]. Beyond PADA-1, most therapy-guidance evidence remains associative.

A central caveat applies to Sections 4.1 - 4.3: earlier molecular detection and prognostic association do not, by themselves, establish that earlier treatment changes improve outcomes. Lead-time and length biases can make earlier detection appear beneficial even when it is not, and a biomarker that stratifies prognosis need not identify patients who benefit from acting sooner. Establishing clinical utility requires interventional evidence randomised or well-controlled studies showing that ctDNA-guided decisions improve survival, quality of life, or other patient-relevant endpoints without net harm. PADA-1 provides such a signal for ESR1-guided switching; for MRD-guided escalation and for most monitoring applications, confirmatory interventional trials are still awaited. Claims of outcome benefit in this review should be read with this limitation in mind.

High costs are a primary obstacle. NGS-based multi-omics assays impose financial burdens often prohibitive for public systems and individual patients, and reimbursement varies widely, leaving many patients with substantial out-of-pocket expense [23][25]. Affordable, high-sensitivity assays aligned with price-sensitive markets are needed [23].

Laboratory capacity is uneven. While Singapore, Thailand and Malaysia have more advanced molecular infrastructure, others are still building capacity. Real-world Vietnamese experience highlights low DNA quality in FFPE tissue and white-blood-cell lysis in plasma affecting reliability [4]; Genomics Thailand notes that only about 10% of tumour tests reveal immediately actionable targets [6]. Sustained investment in laboratory networks, training and quality assurance is required.

The regulatory landscape is fragmented and evolving, with a lack of standardised pathways, inconsistent coverage and country-specific approval processes creating uncertainty and delaying access [24]. Progressive frameworks and expanding reimbursement in some countries are reducing barriers unevenly [23].

Geographic and socioeconomic disparities limit equitable access [22]. In the Philippines, access to cancer care is constrained by few specialised providers and geographic and socioeconomic barriers despite insurance coverage [21]. Vietnamese utilisation data show ctDNA tests were often ordered only once for breast cancer, likely for cost reasons, unlike lung or liver cancer [4]. Community-based interventions can improve screening participation in low-income settings [26].

Multi-omics analysis combined with AI can improve sensitivity for early-stage disease and may reduce long-term costs through automation, provided algorithms are trained on region-specific data such as that from the Genomics Thailand Initiative [6][8]. MCED platforms that screen several cancers from one draw may add value where multiple malignancies are prevalent, but breast-cancer performance must be reported and validated separately [10].

Public-private partnerships and regional knowledge-sharing can subsidise costs, build infrastructure, harmonise pre-analytical and analytical protocols, and share validation data. Academic-industry collaboration, as in the development of SPOT-MAS in Vietnam, illustrates a model for technology transfer suited to regional needs [10].

Given the region’s economic diversity, a tiered model is more realistic than a uniform one. Crucially, each tier should enter routine care only against a defined clinical indication, a specified confirmatory diagnostic pathway, and a stated minimum evidence threshold (Table 2). Across all tiers, cfDNA is positioned as an adjunct: a positive result triggers standard imaging and tissue confirmation, and no cfDNA result alone diagnoses or excludes breast cancer.

Table 2. Proposed tiered cfDNA testing pathway, specifying for each tier the clinical indication and target population, the confirmatory diagnostic pathway, and the minimum evidence required before routine implementation.

Regional harmonisation potentially through ASEAN frameworks could support mutual recognition of approvals, standardised evidence requirements that account for regional data, and expedited pathways for well-validated technologies [23]. A phased reimbursement approach is prudent: begin with established uses (e.g., response monitoring in metastatic disease), then adjuvant-therapy guidance, and finally early detection in defined high-risk groups as evidence ideally interventional solidifies. Workforce development in liquid-biopsy interpretation, bioinformatics and genetic counselling is essential.

Strategies should reflect the region’s dense-breast prevalence and distinct molecular profile, including differing triple-negative frequency and BRCA patterns, necessitating locally tailored panels and interpretation [5][6]. Higher cfDNA levels reported in triple-negative disease may have subtype-specific prognostic value [27]. Equity measures mobile units, simplified collection, telemedicine and trained community health workers should be built in from the outset, alongside local validation, cost-effectiveness analyses and implementation-science research.