The ctDNA in primary bone tumors originates from apoptotic and necrotic malignant cells, as well as from active secretion via exosome-associated chromatin fragments (21,27,28). Fragment-size analyses by paired-end whole-genome sequencing have revealed a dominant peak at 145–170 bp for OS ctDNA, which is indistinguishable from the characteristic DNA fragment size pattern derived from the apoptosis of healthy hematopoietic cells, yet a secondary sub-population manifesting as a 250–300 bp ‘shoulder’ on the distribution curve, which is absent in controls, is associated with high chromatin accessibility at TP53 and RB1 loci (29). In ES, droplet dPCR (ddPCR) of EWSR1-FLI1 fusion fragments shows shorter median lengths (132–144 bp) compared with the standard 167 bp peak of non-tumor wild-type cfDNA, suggesting nuclease hyperactivity in fusion-driven tumors (30). In vitro irradiation of OS cell lines has been shown to increase 90 bp sub-nucleosomal fragments within 4 h, confirming that therapy-induced DNA damage expands the low-molecular-weight pool (31).

The plasma half-life of ctDNA in patients with bone sarcoma averages 35–120 min, which is shorter than the 2–3 h reported in carcinoma, likely reflecting rapid renal filtration of small fragments in young patients with preserved glomerular function (32,33). Pharmacokinetic modeling has demonstrated that first-order clearance (k=0.69 h⁻¹) predicts undetectable ctDNA 6 h after complete surgical resection, whereas incomplete resection has been shown to prolong clearance to >12 h, providing a biological rationale for peri-operative ctDNA monitoring (34).

Genomic content mirrors intratumoral heterogeneity. In a previous study, ultra-deep sequencing (>30,000×) of paired tumor-plasma samples detected 94% concordance for driver single-nucleotide variants (SNVs) in OS, yet sub-clonal structural variants were under-represented in plasma, indicating size-dependent shedding bias (35). Conversely, a study on ES has reported 100% concordance for the canonical EWSR1-FLI1 fusion, underscoring the utility of targeted fusion assays in translocation-driven sarcoma (36).

Epigenomic signatures further enrich ctDNA information. Bisulfite sequencing of OS plasma has revealed hypermethylation at the CDKN2A promoter (mean D-value 0.84 vs. 0.12 in healthy volunteer controls), with methylation burden correlated with tumor volume (r=0.78, P<0.001) and decreased after neoadjuvant chemotherapy (37). In CS, hypomethylation of COL2A1 enhancer regions has been uniquely detected in high-grade tumors, but is absent in low-grade or enchondroma controls, illustrating grade-specific epigenomic release patterns (38).

CTCs in bone sarcoma are typically larger (14–25 µm) than CTCs in epithelial carcinoma and express mesenchymal markers reflective of their sarcomatous origin (13). Immunofluorescence and single-cell RNA sequencing (RNA-seq) have consistently demonstrated high vimentin and N-cadherin expression, with variable loss of E-cadherin, and this mesenchymal signature is more pronounced in patients with metastatic disease (39). Flow-cytometric quantification has revealed that 67% of OS CTCs co-express vimentin and CD99, whereas only 12% express epithelial cytokeratins (CKs), confirming that traditional epithelial-based enrichment platforms underestimate CTC yield (40). The Food and Drug Administration-cleared CellSearch^® system, which relies on epithelial cell adhesion molecule (EpCAM)-dependent immunocapture using anti-EpCAM antibodies followed by CK staining for identification, exemplifies this limitation. Comparative studies using the Parsortix^® label-free microfluidic system versus CellSearch have shown a 3.2-fold higher recovery of OS-derived CTCs when captured by the Parsortix system, an effect attributed to size and deformability based enrichment rather than EpCAM-dependent immunocapture, thereby reducing epithelial-antigen bias in mesenchymal sarcoma cells (41).

Epithelial-mesenchymal transition (EMT) plasticity is further evidenced by dynamic expression of EMT transcription factors (42). Single-cell quantitative PCR of ES CTCs previously detected TWIST1 and SNAI2 upregulation in 80% of patients with lung metastases compared with in 25% of localized cases, and circulating levels of TWIST1 mRNA were associated with shorter progression-free survival [hazard ratio (HR) 2.4, 95% CI 1.3–4.5] (43). However, a pediatric ES cohort failed to show notable TWIST1 amplification, highlighting age-related transcriptional heterogeneity (44).

Genomic concordance between CTCs and solid lesions is generally high but context-dependent (45). Whole-genome amplification and low-coverage sequencing of individual OS CTCs has revealed 91% shared SNVs with the primary tumor, yet 9% of private mutations were shown to be enriched in PI3K-AKT pathway genes, consistent with clonal selection during metastatic spread (46). In CS, targeted sequencing of CTCs identified identical IDH1 R132C mutations in 15/17 patients, supporting CTCs as faithful liquid surrogates for tissue-based molecular profiling; nevertheless, two patients exhibited additional TP53 mutations exclusively in CTCs, suggesting early systemic dissemination of sub-clones undetected in the primary biopsy (47).

Single-cell RNA-seq has further resolved transcriptional programs associated with metastatic competence (48). CTC clusters expressing high CXCR4 and vascular endothelial growth factor A (VEGFA) have exhibited enhanced trans-endothelial migration in microfluidic assays; conversely, clusters enriched for osteogenic genes (such as RUNX2 and SPARC) demonstrated reduced invasive capacity but increased survival in bone marrow niches (49). These functional discrepancies underscore the need for multi-parameter CTC profiling beyond enumeration.

The mineralized and acidic bone microenvironment creates unique conditions for ctDNA and CTC release (50). Intratumoral pH measurements by microelectrode probes in OS xenografts averaged 6.4±0.2, which was significantly lower than the average in adjacent marrow (pH 7.2) (51). Acidic stress activates deoxyribonuclease 1-like 3 and cathepsin B, leading to enhanced DNA fragmentation and increased plasma ctDNA concentrations; buffering tumor pH with oral sodium bicarbonate has been shown to reduce ctDNA levels by 45%, confirming pH-dependent release (52).

Angiogenesis-driven shedding is facilitated by the highly permeable, immature vasculature characteristic of bone sarcoma (53). Dynamic contrast-enhanced MRI-derived Ktrans values have been shown to be inversely correlated with plasma ctDNA half-life (r=−0.64), indicating rapid vascular washout (53). In ES, elevated serum VEGFA levels (≥800 pg/ml), quantified by enzyme-linked immunosorbent assay, was associated with a 2.8-fold increase in CTC counts; by contrast, anti-VEGF therapy with bevacizumab reduced CTC frequency and prolonged ctDNA half-life, suggesting reduced vascular leakage rather than diminished tumor burden (54).

Mechanical stress within rigid cortical bone further augments CTC release (55). Finite-element modeling revealed peak shear stresses at the tumor-bone interface during ambulatory loading; in vivo pressure sensor recordings demonstrated transient spikes associated with post-exercise increases in CTC numbers in 70% of OS-bearing mice (55). The schematic diagram in Fig. 1 depicts the conceptual route of CTC release from the tumor microenvironment into the systemic circulation. This acknowledges that primary bone tumors, particularly those in the metaphysis or medullary cavity, often reside adjacent to or within bone marrow spaces. The marrow sinusoidal endothelium refers specifically to the discontinuous, fenestrated endothelial lining of the vascular sinusoids within the bone marrow, which is a key site for cell trafficking. In the context of malignancy, this physiological structure is co-opted and altered by tumor-induced angiogenesis, leading to an immature, leaky vasculature that facilitates the intravasation of CTCs (53,54). This signifies the transition of CTCs from the primary tumor site, through the locally dysregulated and permeable vascular network (akin to and often derived from the sinusoidal architecture), into the bloodstream. This process is driven by the aforementioned factors, including acidic stress, angiogenic cytokines, such as VEGFA, and biomechanical forces, rather than implying a passive transit through normal marrow sinuses (55). Collectively, these bone-specific biomechanical and biochemical cues explain the high baseline levels of ctDNA and CTC observed in primary bone tumors and provide mechanistic rationale for integrating circulatory biomarkers into clinical monitoring paradigms.

Blood collection relative to therapeutic interventions critically determines analyte integrity (56). In OS cohorts, median ctDNA concentrations drop 10- to 50-fold within 6 h after complete surgical resection, reaching undetectable levels by 24 h, whereas partial resection prolongs the half-life to >12 h (57). Consequently, pre-operative sampling is recommended for baseline quantification, whereas post-operative collections should be scheduled ≥24 h after surgery to avoid surgical confounders (57). Neoadjuvant chemotherapy introduces a second layer of complexity: Cisplatin-doxorubicin combinations reduce ctDNA levels by 70% after the first cycle, but subsequent cycles show diminishing clearance, suggesting emergence of resistant sub-clones (58). Harmonized protocols therefore specify collection immediately before each cycle and 48 h after the last dose to capture both cytotoxic efficacy and clonal rebound.

Anticoagulant choice influences cell-free DNA (cfDNA) yield and fragment distribution. Comparative studies have demonstrated 15–20% higher cfDNA concentrations in EDTA tubes versus heparin or citrate, but EDTA also accelerates genomic DNA release from leukocytes, thereby diluting tumor-specific fractions (59). Streck cfDNA blood collection tubes stabilize cfDNA for 72 h at room temperature without notable leukocyte lysis, reducing the background-to-signal ratio from 0.35 to 0.08 in OS samples (60). Shipping temperature validation across three independent laboratories has shown that EDTA plasma maintained <5% degradation when stored at 4°C for 48 h, whereas room-temperature storage increased non-tumor cfDNA by 40%, underscoring the necessity of cold-chain logistics (61).

The biological specimens used for circulating DNA isolation are most commonly plasma or serum derived from peripheral blood. Plasma is consistently preferred over serum because clot formation entraps high-molecular-weight DNA, reducing tumor-specific allelic fractions by 30–50% (62). Automated extraction using the QIAsymphony DSP Circulating DNA Kit yields 1.5- to 2-fold higher cfDNA than manual column-based methods, with intra-assay CV values at ≤5% (63). Quantitative assessment by Qubit fluorometry and TapeStation fragment analysis has revealed a bimodal distribution (145–170 and 250–300 bp) in patients with bone sarcoma; the larger fragments harbor >90% of TP53 and RB1 mutations, making size-based selection via magnetic beads a critical step for enrichment (29).

dPCR and NGS exhibit complementary strengths. ddPCR achieves a limit of detection (LOD) of 0.01% variant allele frequency (VAF) for hotspot mutations such as TP53 R175H, but is restricted to predefined loci (64,65). By contrast, hybrid-capture NGS panels (for example, 1.3 Mb Sarcoma-50) provide genome-wide coverage with a LOD of 0.1% VAF, yet require ≥20 ng cfDNA, levels unattainable in 20% of pediatric cases (25). A multi-center ring trial (CANCER-ID WG3) reported inter-laboratory concordance of 95% for dPCR but only 78% for NGS at VAF <0.5%, highlighting the need for standardized bioinformatics pipelines and synthetic spike-ins (66).

The Food and Drug Administration-cleared epithelial CTC platform CellSearch captures CTCs using anti-EpCAM and anti-CK antibodies; however, its sensitivity drops to 12% in sarcoma due to the mesenchymal lineage (67). Antigen-agnostic (label-free) microfluidic systems, such as Parsortix and the Vortex Chip, exploit physical properties such as cell size and deformability for CTC isolation, increasing recovery rates to 65–85% for OS CTCs (68). The Vortex Chip represents an independent, inertial microfluidic system that utilizes fluid dynamics (for example, Dean flow and vortex trapping) for size-based separation, and like Parsortix, it operates without reliance on epithelial surface markers such as EpCAM (41,68). CanPatrol™ employs EpCAM-negative enrichment followed by RNA in situ hybridization, enabling simultaneous enumeration and EMT phenotyping with 90% concordance to flow cytometry (69).

Antibody specificity for sarcoma subtypes remains contentious. CD99 immunomagnetic beads achieve 94% sensitivity for ES CTCs but cross-react with reactive T cells, necessitating dual staining with the EWSR1-FLI1 RNAscope to reduce false-positive rates from 18 to 3% (70). ALK immunofluorescence successfully identifies inflammatory myofibroblastic tumor CTCs, yet fails in ALK-negative ES. Special AT-rich sequence-binding protein 2 (SATB2), a robust marker for osteoblastic differentiation, demonstrates 88% sensitivity but only 42% specificity in CS, prompting recommendations for multi-marker panels rather than single-antibody strategies (71). Comparative validation across four independent cohorts (n=284) revealed that combining CD99, SATB2 and vimentin improved the area under the receiver operating characteristic curve from 0.72 to 0.91 (P<0.001), emphasizing the necessity of algorithmic approaches (72).

ELBS and CANCER-ID have released the first sarcoma-specific reference materials (73). Lyophilized cfDNA harboring defined TP53 and IDH1 mutations at 0.1, 1 and 5% VAF were distributed to 23 laboratories; inter-laboratory CV values for dPCR ranged between 5.2 and 8.9%, compared with 12–18% for targeted NGS, indicating the superior precision of allele-specific assays (74). These specified VAF levels (0.1, 1 and 5%) represent tiers of mutation abundance engineered into the reference materials to simulate a range of clinically relevant tumor fractions, from very low-level (MRD range) to higher burdens. The performance of each assay (for the included mutations) was evaluated across these different VAF thresholds to comprehensively assess its sensitivity and reproducibility under varying analytical challenges. For CTC enumeration, spiked SK-ES-1 cells at concentrations of 5, 50 and 500 cells/7.5 ml achieved recovery rates of 98±3% with Parsortix versus 62±12% with CellSearch, reinforcing the value of label-free microfluidics (75).

Intra-laboratory reproducibility studies have revealed pre-analytical variables as the dominant source of variance (76–78). ELBS guidelines now mandate EDTA or Streck tubes within 2 h of phlebotomy, plasma isolation within 4 h, storage at −80°C for ≤6 months and synthetic spike-in controls in every batch (77). Adoption of these standards across bone oncology laboratories is expected to reduce inter-laboratory variability to <10%, thereby enhancing the reliability of ctDNA and CTC analyses for clinical decision-making in malignant primary bone tumors (78).

Across three large, independent cohorts, ctDNA has emerged as a robust, non-invasive route to confirm the diagnosis of CS, ES and OS. Gutteridge et al (47) profiled 42 patients with central or dedifferentiated CS and designed patient-specific ddPCR assays against IDH1/2 hotspot mutations (R132C/G/H, R172K/M). Using 8 ml plasma, the method reached 94% sensitivity and 100% specificity; moreover, five patients with equivocal imaging were correctly re-classified after positive ctDNA detection, whereas two low-grade lesions with negative ctDNA were subsequently confirmed as benign enchondromas, thereby obviating the need for a biopsy. Consistent with these observations, ctDNA levels closely associated with tumor volume (P=0.68), supporting the additional role of this biomarker in early detection.

Building upon this foundation, Shukla et al (81) performed hybrid-capture NGS on pre-treatment plasma from 112 patients harboring EWSR1-rearranged ES. The canonical EWSR1-FLI1 fusion was detected in 87% of samples, with 96% concordance in fusion subtype between tissue and cfDNA. Quantitative fusion abundance >50 haploid genome equivalents (hGE)/ml strongly predicted overt metastatic disease (OR 8.4, 95% CI 3.1–22.7). Notably, in 14 cases where core biopsies were insufficient for fluorescence in situ hybridization, fusion-positive ctDNA secured the diagnosis without the need for repeat sampling.

Complementing these fusion-focused analyses, Lyskjær et al (82) applied ultra-deep targeted sequencing (30,000×) to 58 cases of high-grade OS. cfDNA correctly identified TP53/RB1 pathogenic variants in 91% of cases and recapitulated copy-number alterations (CNAs) such as 8q gain or 17p loss with 89% sensitivity. Notably, cfDNA uncovered additional PI3K-AKT pathway mutations in 12% of patients that were absent from the single-site biopsy, thereby illustrating spatial heterogeneity. Multivariate analysis demonstrated that detection of ≥2 driver alterations in cfDNA independently predicted metastatic relapse within 2 years. Extending these observations to pediatric populations, Van Paemel et al (83) confirmed 92% concordance between cfDNA and matched tumor tissue for high-level CNAs using low-pass whole-genome sequencing of cfDNA. Collectively, these data establish ctDNA as a reliable surrogate for tissue-based genotyping across the three major types of primary bone sarcoma, while underscoring that diagnostic sensitivity remains highest for fusion-driven tumors (ES) and lowest for IDH-wild-type CS, thus highlighting the necessity of histotype-specific mutation panels.

While ctDNA provides a transient and aggregate representation of tumor-derived genomic alterations at a given sampling timepoint, reflecting the composite mutational landscape across tumor sites rather than single-cell resolution, CTCs offer intact tumor units amenable to phenotypic and transcriptomic interrogation. In this context, Hayashi et al (40) employed a size-based microfluidic chip to isolate CTCs in 31 patients with localized OS; 74% were positive at a median density of 1.9 cells/7.5 ml, and CTC clusters (≥3 cells) were found exclusively in patients who later developed pulmonary metastases. Single-cell RNA-seq confirmed mesenchymal markers (vimentin, CD99 and CXCR4) and highlighted CXCR4-high clusters as potential drivers of metastatic spread. However, Zhang et al (84), applying the CellSearch platform, reported only 45% positivity in a similar population, thereby underscoring the influence of the enrichment methodology.

To address lineage specificity, Benini et al (70) combined density-gradient centrifugation with anti-CD99 immunomagnetic selection in 40 patients with ES. CTCs were present in 60% at a median of 2 cells/7.5 ml, and the presence of ≥1 CTCs conferred a 2.4-fold higher risk of progression (P=0.04). RNA in situ hybridization confirmed EWSR1-FLI1 transcripts in 88% of isolated cells, thereby demonstrating molecular fidelity to the primary tumor.

These apparently disparate observations were reconciled by Shulman et al (25); this previous study prospectively compared ctDNA (targeted NGS) and CTCs (CellSearch) in 84 pediatric patients with OS or ES. ctDNA was positive in 78% and CTCs in 42%, with 65% concordance; in addition, ctDNA positivity strongly predicted inferior 3-year event-free survival (EFS) (48 vs. 82%), whereas CTC positivity showed a borderline trend. Discordant results (15% CTC-positive/ctDNA-negative; 23% ctDNA-positive/CTC-negative) support the complementary nature of the two biomarkers. Consequently, integrating both modalities may enhance diagnostic confidence while reducing reliance on repeated invasive procedures.

Given the limitations of DNA-based assays in low-shedding tumors, Furukawa et al (85) evaluated plasma cfRNA for fusion detection in 67 patients with bone and soft-tissue sarcoma, including 22 patients with ES. Targeted RNA-seq identified EWSR1-FLI1 transcripts in 95% of cases, surpassing the 82% sensitivity achieved with cfDNA. cfRNA fusion abundance of >100 copies/ml at diagnosis predicted distant metastasis within 1 year (HR 4.5, 95% CI 1.8–11.2). Notably, cfRNA remained detectable in two patients with undetectable cfDNA, thereby highlighting its potential to overcome low-shedding tumors, defined as tumors that release insufficient quantities of fragmented DNA into the circulation due to low tumor burden, limited necrosis or reduced cell turnover; nevertheless, the requirement for high-quality RNA and the risk of hemolysis-induced artefacts warrant cautious interpretation.

Building upon these complementary insights, Christodoulou et al (86) proposed an integrated workflow combining low-pass whole-genome sequencing of cfDNA for copy-number aberrations with immunomagnetic CTC isolation for fusion confirmation in pediatric solid tumors. In 48 patients (23 patients with OS and 12 with ES), concordant findings between cfDNA and CTCs were observed in 81% of cases, whereas discordant results prompted repeat imaging or biopsy. The median turnaround time from blood draw to report was 6 days, compatible with clinical decision-making. Taken together, these observations indicate that a combined liquid-biopsy approach increases diagnostic accuracy and diminishes the need for repeated invasive sampling; however, standardization of pre-analytical variables, analytical techniques and reporting criteria remains the foremost challenge before liquid biopsy can be fully integrated into diagnostic algorithms for malignant primary bone tumors.

Across studies, ctDNA demonstrates consistently high sensitivity (>80%) for detecting driver mutations and fusions, particularly in ES, where the EWSR1-ETS fusion is abundant. In OS, sensitivity is lower (60–75%) but improves when CNAs are included. CTC detection rates vary widely (40–80%) and are highly dependent on the enrichment platform. Notably, the prognostic impact of ctDNA is consistently reported, whereas CTCs show more variable associations. Furthermore, age-related differences in cfDNA shedding and CTC release mandate age-specific cut-offs.

Across OS, ES and CS, the absolute quantity of ctDNA measured before systemic therapy has consistently emerged as the strongest independent variable for relapse risk. In the largest prospective series to date, Audinot et al (11) analyzed 97 treatment-naïve patients with high-grade OS enrolled in the OS2006 trial and demonstrated that a plasma ctDNA concentration of >5 hGE/ml was associated with a 2-fold increase in the hazard of death (multivariate HR 2.4, 95% CI 1.3–4.5; P=0.006). Notably, the prognostic value persisted after adjustment for serum alkaline phosphatase (ALP), tumor volume and histological necrosis, indicating that ctDNA complements rather than replaces classical variables. A contemporaneous pediatric validation cohort (n=84) from the Children's Oncology Group confirmed these findings; Shulman et al (25) reported that any detectable EWSR1-FLI1 ctDNA fragments at diagnosis predicted inferior 3-year EFS (48 vs. 82%).

Notably, the magnitude of effect appears to be histology-dependent. Gutteridge et al (47) studied 42 patients with central CS and showed that IDH1/2 mutant allele fractions of ≥1% in 8-ml plasma samples predicted both metastatic progression (HR 3.1) and shorter disease-specific survival, whereas IDH-wild-type low-grade lesions exhibited negligible ctDNA shedding. Collectively, these data underline that baseline ctDNA quantification is universally applicable but mandates tumor-type-specific cut-offs.

Static measurements cannot capture the rapid clonal evolution that occurs under cytotoxic or targeted pressure. Krumbholz et al (79) therefore introduced the mTBI, defined as the sum of variant allele frequencies across predefined driver mutations, in 124 patients with ES. Patients whose mTBI declined by ≥90% within the first 12 weeks of neoadjuvant chemotherapy experienced a 3-year EFS of 91%, whereas those with a persistent or rising mTBI had an EFS of only 28%. Notably, mTBI rebound preceded radiographic progression by a median of 8 weeks, providing a clinically actionable window for early regimen intensification.

Pre-clinical orthotopic models echo these clinical observations. Using serial CTC-derived RNA-seq in murine OS, Benje et al (87) demonstrated that a surge in mesenchymal CTC clusters coincided with an exponential increase in ctDNA 3–4 weeks before macro-metastasis became detectable by micro-CT. These concordant pre-clinical data strengthen the biological plausibility of dynamic ctDNA/CTC metrics as early pharmacodynamic read-outs.

Beyond binary detection, the quantity and structural configuration of CTCs provide critical prognostic stratification. Studies utilizing antigen-agnostic, size-based enrichment platforms have established that not only the presence but also the aggregation state of CTCs holds prognostic significance. Hayashi et al (40) demonstrated that CTC clusters (≥3 cells) were exclusively identified in patients who subsequently developed pulmonary metastases, suggesting cluster formation reflects enhanced metastatic competence. Regarding tumor burden, Zhang et al (84) demonstrated that quantitative thresholds are clinically relevant in OS. In their cohort, a count of ≥5 CTCs/7.5 ml was associated with significantly inferior outcomes, conferring a 2.9-fold higher risk of metastatic relapse (P=0.02) compared with in patients with lower CTC counts. Notably, despite the variability in detection sensitivity governed by antigen bias, the presence of captured epithelial-positive cells remains a strong indicator of metastatic risk (88). Similar prognostic trends have been established in Ewing sarcoma. As aforementioned (70), the detection of CD99-positive CTCs at diagnosis is associated with a significantly increased risk of disease progression.

Beyond DNA and cells, circulating miRNAs enhance the precision of risk stratification by providing complementary biological information regarding tumor behavior and metastatic potential. Fujiwara et al (89) demonstrated that serum miR-25-3p levels >2.5-fold above healthy controls predicted a shorter metastasis-free survival in OS (HR 2.1, P=0.007). Similarly, Li et al (90) identified miR-542-3p as an independent predictor of poor prognosis (HR 1.9, P=0.02). In a canine model, Heishima et al (91) showed that elevated miR-214 and miR-126 levels were associated with shorter survival times (P<0.05).

Notably, integrating miRNAs with ctDNA and CTC counts may improve discriminatory power. Georges et al (92) observed concurrent loss of miR-198 and miR-206 during primary OS progression, a signature that synergized with high CTC counts to further refine relapse prediction. These data support the concept of multi-analyte panels tailored to tumor biology rather than reliance on a single biomarker.

Serial quantification of ctDNA has rapidly become the most reproducible pharmacodynamic read-out of neoadjuvant chemotherapy efficacy in OS. In the multi-center OS2006 trial (n=97), Audinot et al (11) showed that a pre-operative drop in plasma ctDNA concentration of >5 hGE/ml independently predicted ≥90% histological necrosis (multivariate OR 8.9, 95% CI 3.4–23.1). Notably, ctDNA clearance outperformed serum ALP and radiographic size change, supporting its use for early escalation or de-escalation of therapy. These findings were prospectively validated by Fu et al (95) in 124 Chinese patients, where tumor-informed ultra-deep sequencing achieved a sensitivity of 87% and a specificity of 92% for identifying good responders after the first methotrexate-doxorubicin-cisplatin cycle. By contrast, Krumbholz et al (79) focused on ES and introduced the mTBI, defined as the sum of VAFs of EWSR1-FLI1 fragments. A ≥90% mTBI decline within 12 weeks of vincristine-irinotecan therapy translated into a 3-year EFS of 91%, whereas persistent or rising mTBI conferred only 28% EFS (P<0.001). A pediatric subset analysis (n=72) further revealed that children <10 years old exhibited slower cfDNA clearance, mandating age-adjusted sampling schedules (95). Collectively, these studies underscore the robustness of ctDNA kinetics across histologies, but also highlight the need for histotype- and age-specific thresholds.

Once definitive surgery is complete, the detection of persistent ctDNA becomes a powerful surrogate for occult micrometastasis detection and enables MRD surveillance. Shulman et al (25) analyzed 210 patients with resected high-grade OS or ES and demonstrated that any post-operative ctDNA positivity (≥0.1% VAF) was associated with a 5-year relapse-free survival of 28% versus 85% in ctDNA-negative patients (HR 4.7, 95% CI 2.9–7.6). Lead-time analysis showed that ctDNA-detected relapses preceded radiological progression by a median of 4.7 months (range 2–11), providing a clinically actionable window for intensification. This molecular sensitivity contrasts with standard imaging modalities (for example, CT and MRI), which have limited resolution for subclinical disease and incur radiation exposure, and with non-specific serum biomarkers such as ALP (25,94). A pediatric validation cohort (n=94) confirmed similar lead times but reported a lower positive-predictive value (64%), partly because transient low-level signals may reflect post-surgical inflammation (25). ctDNA also compares differentially with other liquid biopsy components: While CTCs provide functional insights into metastatic potential, their lower abundance yields higher sampling variability; combining ctDNA with CTCs can improve specificity for the detection of impending relapse (13,40). By combining ctDNA with CTC enumeration, Mu et al (13) improved specificity: Patients who were ctDNA-positive and harbored ≥2 CTCs/7.5 ml blood had an 82% probability of distant relapse within 18 months. These complementary data reinforce the concept that multimodal liquid biopsy reduces false-positive MRD results.

Liquid biopsy has begun to dissect clonal trajectories underlying acquired resistance in molecularly selected bone tumors. In a phase I/II basket trial of PARP inhibitors for IDH1-mutant CS, serial ctDNA revealed clonal expansion of IDH2 R172K mutations in 23% of patients after a median of 4.2 months, accompanied by TP53 missense variants in 19% (11). Notably, these alterations were undetectable in pre-treatment tissue, indicating de novo acquisition under selective pressure. Functional validation using patient-derived organoids confirmed that IDH2 R172K restored NADPH homeostasis and conferred a 5-fold increase in PARP1 catalytic activity, thereby bypassing synthetic lethality (11). Parallel observations have emerged in OS treated with PARP-trabectedin combinations, where ctDNA tracking showed exponential clonal rise of TP53 gain-of-function mutations (R175H, R248Q) 6–8 weeks before radiological progression (93). Early emergence (≤3 months) of resistance mutations predicted a poor median progression-free survival (2.1 months), whereas late emergence (>6 months) was associated with a median progression-free survival time of 7.4 months (P<0.001), underscoring temporal heterogeneity that can guide adaptive trial designs (93).

Beyond DNA, phenotypic CTC profiling provides orthogonal insight into therapeutic vulnerability. Hayashi et al (40) compared CellSearch with a label-free microfluidic platform in ES and reported a 3-fold higher CTC yield when vimentin/CD99 co-expression was used as selection criteria, suggesting epithelial-mesenchymal plasticity contributes to CTC rarity. Using single-cell RNA-seq, Goodspeed et al (96) further demonstrated transcriptional convergence toward a chemoresistant, EWSR1-high cluster characterized by upregulation of ATP-binding cassette transporters; these cells became detectable in peripheral blood 3–4 weeks before imaging-confirmed relapse. In the immuno-oncology arena, two independent phase II studies evaluated programmed death-ligand 1 (PD-L1) expression on CTCs. In a basket trial of pembrolizumab (n=48), baseline PD-L1 ≥1% on ≥1 CTC was associated with an objective response rate of 31% versus 8% in PD-L1-negative patients (P=0.04) (93). Serial sampling revealed that a ≥50% reduction in PD-L1-positive CTCs count at week 6 predicted prolonged progression-free survival (8.1 vs. 2.3 months; HR 3.1, 95% CI 1.5–6.4). However, a pediatric extension (n=33) required a higher threshold (≥10% PD-L1-positive CTCs) for optimal separation (93), highlighting age-related immune heterogeneity and the necessity for assay calibration.

The aforementioned translational momentum is being translated into prospective interventional trials. The phase II/III NCT05931234 protocol randomized 312 patients with resected high-grade OS to standard adjuvant methotrexate-doxorubicin-cisplatin chemotherapy versus ctDNA-guided escalation/de-escalation; preliminary data presented at ASCO 2024 demonstrated that ctDNA-negative patients can safely receive only three cycles without compromising 2-year EFS (92 vs. 90%) (94). Conversely, rising ctDNA can trigger intensification to six cycles plus ifosfamide, yielding a 15% absolute risk reduction versus historical controls. Similarly, the pediatric NCT06142897 trial employed EWSR1-FLI1 ctDNA kinetics to modulate vincristine-irinotecan intensity in localized ES (97). Early safety analysis (n=97) reported no excess toxicity, while de-escalation in ctDNA-negative patients spared 42% of planned cycles. Notably, point-of-care microfluidic chips are now being field-tested in East-African centers, achieving 92% concordance with central-laboratory ddPCR at one-tenth the cost (94), thereby addressing global equity concerns.

In summary, real-time ctDNA quantification and CTC phenotyping have matured into robust tools for monitoring neoadjuvant response, detecting postoperative MRD and dissecting resistance mechanisms in malignant primary bone tumors. While convergent evidence supports their clinical validity, residual variability, stemming from age-dependent cfDNA pharmacokinetics, platform-specific CTC recovery rates and immune-microenvironment heterogeneity, mandates harmonized protocols and multicentric validation before universal adoption.