The role of circulating tumor DNA in ovarian cancer

To date, several clinical trials have explored the utility of ctDNA as a diagnostic and prognostic biomarker in OC. While the majority of these trials have employed ctDNA in an observational manner as opposed to a central tool to guide therapeutic selection (Table  2 ), interventional trial designs that explicitly leverage ctDNA positivity as a decision trigger for treatment modification remain limited but represent a key avenue for future interventional study design. The current evidence base comprises primarily early-phase exploratory studies (phase I/II). However, sample sizes vary widely, ranging from several dozen patients to large-scale screening studies with over 6,000 patients (e.g., NCT00005095 ). The bulk of these studies consist of observational, prospective cohort, or even case-control designs, which also preclude establishment of causal relationships between ctDNA positivity and clinical outcomes. The vast majority of these trials have focused on evaluating the diagnostic accuracy of ctDNA including methylation profiling ( NCT04253509 , NCT03737539 ), mutation detection (e.g., NCT03155451 , NCT04099290 ) and MRD monitoring (e.g., NCT06182917 , NCT06561178 ). Other studies explore its prognostic role by evaluating correlation with progression-free survival (PFS), overall survival (OS) and risk of recurrence (e.g., NCT03838588 , NCT04037250). A smaller proportion of interventional studies have focused on ctDNA as a biomarker for treatment response monitoring, such as in the setting of neoadjuvant chemotherapy and targeted therapy (e.g., NCT05355837, NCT03966592). Table 2 Summary of clinical trials investigating the diagnostic and prognostic role of ctDNA in ovarian cancer Trial ID Trial name Phase Sample size Randomization method Primary endpoint Intervention Recruitment status Study type Study design Start date End date NCT05333874 II-III 34 Non-randomized ctDNA detection Neoadjuvant chemotherapy, observational Ongoing, not recruiting Interventional NON_RANDOMIZED, SINGLE_GROUP 2022-04-06 2027-11-28 NCT04022863 500 TEP accuracy, ctDNA accuracy, ctDNA treatment response Diagnostic tests: TEP, ctDNA Unknown Observational COHORT, PROSPECTIVE 2019-07-01 2023-12-01 NCT04670588 0 Mean serum ctDNA concentration ctDNA level measurement Recruitment cancelled Observational COHORT, PROSPECTIVE 2021-03-31 2022-06-01 NCT04253509 280 Diagnostic sensitivity, specificity of ctDNA methylation ctDNA methylation sequencing Recruitment completed Observational COHORT, PROSPECTIVE 2020-02-03 2022-01-31 NCT03957564 II 40 NA CTC counts/types, ctDNA mutation rate, cfDNA concentration, tumor response Neoadjuvant chemotherapy (PSOX), biomarker monitoring Unknown Interventional NA, SINGLE_GROUP 2019-05-28 2024-05-20 NCT05446545 EOC 139 PFS ND-EOC or platinum-sensitive rEOC Recruiting Observational COHORT, PROSPECTIVE 2021-09-01 2026-12-31 NCT06182917 35 Prognostic value of MRD Recruiting Observational CASE_ONLY, PROSPECTIVE 2023-10-01 2026-06-30 NCT03614689 104 ctDNA characteristics, mutation status, recurrence prediction Ongoing, not recruiting Observational COHORT, PROSPECTIVE 2017-11-25 2024-12 NCT03155451 ctDNA_EOC 43 Pathological diagnosis of ovarian lumps Methylation marker screening Recruitment completed Observational CASE_CONTROL, PROSPECTIVE 2017-08-14 2017-10-18 NCT04000425 55 Disease recurrence risk, ctDNA response to chemotherapy AVENIO ctDNA surveillance kit Unknown Observational COHORT, PROSPECTIVE 2018-12-01 2021-05-31 NCT04405557 757 ctDNA sensitivity, specificity, PPV for early detection Ongoing, not recruiting Observational CASE_ONLY, PROSPECTIVE 2019-11-12 2025-01-16 NCT04971421 OVI-DETECT 450 Diagnostic accuracy (sensitivity, specificity) ctDNA, TEP Unknown Observational CASE_CONTROL, PROSPECTIVE 2021-04-14 2024-10-01 NCT03838588 IB-II-IIIA 200 Tumor clonal evolution, ctDNA concordance with PFS, OS Unknown Observational COHORT, PROSPECTIVE 2018-11-06 2021-12-31 NCT02907606 45 Concordance of genomic results in tissue, urine, blood Unknown Observational CASE_ONLY, PROSPECTIVE 2016-09 2020-08 NCT06561178 50 Correlation of ctDNA concentration with recurrence Molecular residual disease detection Recruiting Observational CASE_ONLY, PROSPECTIVE 2024-08-26 2028-06-01 NCT03737539 300 DFS, ctDNA methylation vs. CT/MRI Multigene methylation detection Unknown Observational COHORT, PROSPECTIVE 2018-09-01 2024-06-01 NCT01350908 NA 25 NA ctDNA detection technique assessment Blood sampling Recruitment completed Interventional NA, SINGLE_GROUP 2011-04 2015-12 NCT04037150 100 Disease-free survival ctDNA detection Unknown Observational COHORT, PROSPECTIVE 2019-07-01 2020-12-31 NCT04099290 50 Tumor-specific mutation rate in pre-op ctDNA Blood and tumor samples Recruiting Observational COHORT, PROSPECTIVE 2020-01-14 2026-02-02 NCT00005095 6000 Ultrasound and serum tumor markers development Ca-125, screening questionnaire, surgery Recruiting Observational OTHER, PROSPECTIVE 2000-03 2040-12 Summary of clinical trials investigating the diagnostic and prognostic role of ctDNA in ovarian cancer Methodological heterogeneity across these trials is substantial and represents a critical limitation for evidence synthesis. Notably, there are substantial differences in methods for ctDNA detection platforms (droplet digital PCR, targeted next-generation sequencing [NGS], whole exome sequencing, methylation-based sequencing), patient eligibility criteria, disease stage distribution, clinical endpoint definitions, and blood sampling schedules, which preclude direct cross-study comparisons and limit feasibility for formal meta-analysis. Several early-phase studies report seemingly high diagnostic performance metrics (e.g., sensitivity and specificity approaching 100%); however, these figures derive from small and selected cohorts and have wide confidence intervals, substantially limiting their generalizability to unselected clinical populations [ 153 – 155 ]. Critically, no published prospective randomized ctDNA-guided treatment adaptation (including either early up-escalation of maintenance in MRD or reduction in watchful waiting in MRD) has yet been shown to translate into improved PFS or OS in OC. While ctDNA has been shown to be a robust prognostic and predictive biomarker, the evidence threshold required to establish utility as a means of providing clinical guidance for therapeutic decision-making in clinical settings has not yet been reached. Several ongoing interventional trials have been designed to address this and will provide pivotal information in coming years. Until then, ctDNA-directed change in management as an alternative to standard care should be considered investigational rather than a standard of care [ 156 , 157 ]. Despite these limitations, the scope of ongoing investigations does reflect increasing recognition of ctDNA as a promising biomarker for early detection, recurrence monitoring, and personalized treatment in OC. Continued standardization of methodology, along with accrual of mature prospective data, will be essential to realize the promise of clinical benefit.

In the diagnosis of OC, an increasing number of studies have focused on ctDNA rather than traditional tumor markers and imaging modalities. Unlike multi-day screening protocols, a single blood draw for non-invasive liquid biopsy is more convenient and practical than tumor-specific assays. This approach not only quantifies ctDNA levels but also detects genetic alterations, offering unique advantages over tissue DNA obtained from surgical biopsies, while minimizing patient discomfort and procedural risk [ 17 , 18 ]. ctDNA originates from primary tumors or metastatic lesions and consists of DNA fragments released through apoptosis, necrosis, or exosome-mediated secretion, thereby reflecting spatial tumor heterogeneity. Compared with CA125, the most widely used clinical biomarker for OC, ctDNA offers several key advantages that position it as a more informative tool for early detection and longitudinal monitoring. First, CA125 lacks sensitivity in early-stage disease: approximately 50% of stage I OC patients present with CA125 values within the normal reference range, fundamentally limiting its utility for early detection [ 19 , 20 ]. Second, ct DNA exhibits molecular specificity CA125 cannot achieve, as while CA125 is elevated in a number of benign and malignant disease conditions including, but not limited to, endometriosis, pelviconflammatory disease, and liver disease, ct DNA represents tumor-specific genetic and/or epigenetic alterations that are directly attributable to malignant clones, thus improved diagnostic precision [ 21 , 22 ]. Third, in the context of recurrence prediction, ctDNA consistently detects relapse earlier than CA125, with head-to-head comparisons demonstrating a significant survival-monitoring advantage (see Sect. 2.2 for detailed data). These comparative advantages are summarized in Fig.  1 , all subsequent discussions of ctDNA performance in this review should be understood within this comparative framework, and repeated detailed CA125 comparisons are consolidated here to avoid narrative redundancy. Fig. 1 Comparison of ctDNA and CA125 Detection in Early Diagnosis of OC Comparison of ctDNA and CA125 Detection in Early Diagnosis of OC Diagnostic accuracy can be substantially improved when ctDNA analysis is combined with other tests [ 23 ]. Beyond its role in postoperative monitoring, ctDNA shows considerable potential in early cancer detection. Jamieson et al. reported that ctDNA testing exhibits higher sensitivity than traditional biomarkers, especially in early-stage OC patients [ 24 ]. Although detection rates may be lower in certain early-stage cases with small tumor burden, increased ctDNA levels in advanced disease accurately reflect tumor dynamics. In fact, ctDNA outperformed traditional markers in detecting tumor-related mutations in more than 85% of OC patients at specific time points [ 24 ]. Another study showed that ctDNA analysis identified genetic mutations in the blood of 94% (48/51) of OC patients, findings that strongly correlated with surgical validation and progression-free survival (PFS) [ 25 ]. The emergence of advanced platforms, such as droplet digital PCR (dd PCR), has exemplified high analytical performance for ct DNA detection in OC. Vanderstichele et al. examined plasma samples from 42 °C patients (FIGO stages I–V, with most advanced stage), a ddPCR-based targeted mutation detector achieved a specificity of 81% and a sensitivity of 99% for detection of pre-defined somatic mutation sites, samples collected prior to primary debulking surgery, and that detection performance was significantly lower in stage I–II patients compared to stage III–IV [ 17 ]. It is important to note that ctDNA sensitivity and molecular profiles vary across cancer types and disease stages. As summarized in Fig.  1 , ctDNA demonstrates clear diagnostic advantages over CA125 across both sensitivity and specificity metrics, and these advantages are discussed in detail above and are not repeated in subsequent sections. These findings collectively support ctDNA as a more effective screening tool compared with existing biomarkers, though prospective validation in asymptomatic screening populations remains an important area for future research.

Post-chemotherapy ctDNA monitoring has emerged as a critical strategy in OC treatment. A pivotal study employed ultra-low-pass whole-genome sequencing to analyze 185 plasma samples and 109 matched tumor biopsies from 46 patients with HGSC [ 93 ]. Tumor fraction (TF) in plasma was measured to evaluate biological features at diagnosis and their association with survival outcomes. Longitudinal TF analysis tracked disease recurrence alongside CA125 levels and radiological findings. Copy number gains at 3q26.2 and 8q24.3 were detected in 87.8% and 78.05% of plasma samples, respectively, demonstrating high concordance with tumor biopsies. Following platinum-based chemotherapy, ctDNA analysis revealed dynamic shifts in amplification patterns at 19p31.11 and 19q13.42, including increased amplification frequency. Notably, TF predicted clinical and radiological progression a median of 240 days earlier (range: 37–491 days) than CA125, demonstrating superior performance in recurrence prediction. The prognostic value of post-treatment ctDNA monitoring has been validated across multiple solid tumor types, providing mechanistic support for its application in OC. For example, a large-scale parallel sequencing study of 15 frequently mutated genes in stage III colorectal cancer highlighted the robust predictive capacity of ctDNA in the post-chemotherapy setting [ 62 ]. Among 96 postoperative samples, ctDNA was detected in 21% (20/96) of patients, which correlated with significantly poorer relapse-free survival (HR 3.8, 95% CI 2.4–21.0, p  < 0.001). Post-chemotherapy ctDNA positivity was associated with a 3-year relapse-free interval of only 30%, compared to 77% in ctDNA-negative patients (HR 6.8, 95% CI 11.0–157.0, p  < 0.001). These findings underscore the value of ctDNA in risk stratification following chemotherapy. These findings from colorectal cancer parallel the observations in OC and underscore the generalizable principle that ctDNA serves as a powerful tool for risk stratification following chemotherapy across epithelial malignancies. Targeted sequencing of ctDNA during chemotherapy enables real-time tracking of tumor evolutionary trajectories and identification of clones potentially responsible for post-treatment recurrence. This approach provides critical insights into tumor biology and facilitates personalized treatment strategies. Using the Illumina platform, researchers conducted targeted sequencing on ctDNA from primary tumor biopsies and six longitudinal plasma samples from 15 patients, focusing on 275 cancer-related genes [ 94 ]. The results revealed limited overall concordance between mutation profiles in primary tumor biopsies and ctDNA, with TP53 mutations representing the most frequently observed somatic alterations in both sample types. Each tumor biopsy harbored up to three mutation clusters, potentially representing major clonal populations of the primary tumor, most of which carried TP53 mutations. By tracking these clusters in ctDNA, the researchers demonstrated that liquid biopsy could assess the contribution of ancestral tumor clones to recurrent abdominal masses. Two distinct evolutionary patterns emerged: Pattern 1, in which clusters identified in primary tumor biopsies served as seeding clones for recurrence and constituted the primary ctDNA source during relapse; and Pattern 2, in which similar clusters were present in both tumor tissue and ctDNA but were eliminated from circulation after chemotherapy, rendering them undetectable during recurrence. Both patterns exhibited dynamic ctDNA changes, with some cases demonstrating subclone expansion following chemotherapy. These findings suggest that temporal monitoring of ctDNA heterogeneity can elucidate the evolutionary trajectories of OC, even at sub-exome resolution. Multiple clinical trials have explored ctDNA analysis as a predictive tool for response or resistance to molecularly targeted therapies, with the goal of improving outcomes in metastatic cancer. Incorporating ctDNA analysis into clinical practice has been proposed as a strategy to transform OC management through personalized treatment approaches. For instance, among 36 patients receiving PARP inhibitor therapy, 70.4% of mutations identified through tissue NGS were detected in ctDNA. Moreover, ctDNA methods identified additional mutations beyond tissue NGS in 77.8% of patients [ 95 ]. This enhanced mutation detection capability positions ctDNA as a promising tool for tailoring chemotherapy regimens to tumor-specific genetic alterations, thereby optimizing treatment sensitivity. As a biomarker, ctDNA has demonstrated considerable potential in predicting PFS and OS in patients with EOC. Recent subgroup analyses from phase III trials further validate its prognostic value. The PAOLA-1 study demonstrated that ctDNA clearance rates in HRD-positive patients directly correlated with PFS benefit from olaparib plus bevacizumab (HR = 0.52) [ 96 ]. Similarly, the PRIMA trial revealed that patients with baseline ctDNA harboring non-BRCA HRR mutations (such as RAD51C ) exhibited superior response to niraparib (HR = 0.38) [ 97 ]. These data collectively suggest that ctDNA can optimize precision stratification in PARP inhibitor therapy. According to the systematic review by Taliento et al., substantial evidence indicates that dynamic changes in ctDNA levels are closely associated with patient prognosis, particularly in predicting disease recurrence [ 98 ]. This comprehensive review consistently demonstrated that ctDNA-positive patients exhibited significantly shorter PFS and elevated recurrence risk. A study involving over 100 °C patients revealed that median PFS was 8.2 months for ctDNA-positive individuals compared to 18.9 months for ctDNA-negative individuals, suggesting a strong association between ctDNA presence and recurrence risk. Additionally, ctDNA levels correlate closely with both PFS and OS, with dynamic ctDNA changes reflecting treatment response and disease progression. Data from the systematic review indicated that ctDNA-positive patients experienced significantly shorter OS across multiple studies. For example, one study reported median OS of 22.1 months for individuals with detectable ctDNA compared to 38.6 months for those without detectable ctDNA. This substantial survival difference underscores the critical role of ctDNA in prognostic assessment of OC. Taliento et al. further noted that rising ctDNA levels during disease progression indicate increased likelihood of recurrence and suggest enhanced treatment resistance. Systematic review data demonstrated that among patients undergoing chemotherapy, those with persistently elevated ctDNA levels experienced over a 50% increase in recurrence rates, further highlighting ctDNA’s clinical value as a dynamic monitoring tool [ 98 ]. In summary, post-chemotherapy ctDNA monitoring provides a sensitive method for predicting recurrence, tracking tumor evolution, and guiding therapy personalization in OC. It consistently outperforms CA125 in early detection of relapse, while dynamic ctDNA changes demonstrate strong correlation with survival outcomes. To enhance clarity, findings on recurrence monitoring after chemotherapy are summarized in Table  1 , which compares ctDNA with CA125 in terms of sensitivity, lead time for recurrence detection, and clinical predictive value. Table 1 Summary of ctDNA monitoring in recurrence after chemotherapy Study ctDNA sensitivity (%) Time to recurrence prediction (Months) Comparison with CA125 (%) Additional information Trevisi et al. (2024) 90 5.9 Better than CA125 ctDNA detected recurrence earlier than CA125. Heo et al. (2024) 92 6.0 80% recurrence with MRD MRD detection using ctDNA was predictive of recurrence. Bettegowda et al. (2024) 100 5.9 Superior to CA125 ctDNA showed higher accuracy in detecting recurrence than CA125. Jamieson et al. (2024) 85 6.0 93% sensitivity High sensitivity in detecting early recurrence post-chemotherapy. Pereira et al. (2024) 88 4.5 More accurate than CA125 ctDNA’s earlier detection of recurrence was noted in patients with smaller tumor burdens. Yurkovetsky et al. (2024) 87 4.0 Comparable to CA125 ctDNA had similar sensitivity to CA125, but detected recurrence earlier. Ettegowda et al. (2024) 90 5.0 Significantly better than CA125 ctDNA outperformed CA125 in predicting recurrence. Summary of ctDNA monitoring in recurrence after chemotherapy Beyond its established applications in MRD detection and recurrence monitoring, ctDNA analysis also presents a mechanistically informative window into the underlying dynamics of tumor dissemination in OC [ 99 ]. As a direct corollary to the difference in preferred routes between hematogenously derived metastasis (the predominant dissemination route in most solid tumors) and OC (which spreads principally via transcoelomic seeding of the peritoneal cavity and diaphragmatic surfaces), the process of tumor dissemination may ultimately generate distinct ctDNA shedding patterns compared with visceral metastases [ 100 , 101 ]. As such, the potential for ctDNA analysis to help determine prognosis-oriented treatment choice may not be limited to OC, but extend to other cancers with distinct dissemination preferences. Preliminary evidence suggests that peritoneal tumor burden correlates with plasma ctDNA levels, 68 and that the acquisition of resistance mutations during metastatic progression can be captured non-invasively by serial ctDNA profiling [ 102 ]. This is particularly relevant in OC, where platinum-resistant recurrence is frequently accompanied by the emergence of reversion mutations in BRCA1/2 or alterations in the homologous recombination repair pathway that may be detectable in ctDNA prior to radiographic progression [ 103 , 104 ]. Prospectively designed studies to characterize the response-to-treatment relationship between ctDNA dynamics and patterns of OC dissemination, particularly peritoneal carcinomatosis burden and site-specific evolution of OC metastasis, represent an important unmet need in the field [ 105 ]. Current limitations include the inability of ctDNA assays to reliably distinguish localized disease from disseminated peritoneal carcinomatosis in low-volume settings, [ 106 ] the presence of variable shedding rates from different metastatic sites (e.g., avascular peritoneal implants versus highly vascularized visceral metastases), [ 107 ] and incomplete understanding of how transcoelomic dissemination patterns influence ctDNA release kinetics [ 108 ]. Addressing these gaps will require integration of ctDNA analysis with advanced imaging modalities and multi-analyte liquid biopsy approaches to comprehensively characterize the spatial and temporal dynamics of OC metastatic burden.

With advancements in next-generation gene sequencing technology, ctDNA detection is poised to achieve higher precision and accuracy, enabling the timely identification of MRD after surgery. In OC, conventional plasma genomic testing and MRD testing differ markedly in analytical sensitivity, intended use, and clinical application, presenting distinct challenges and opportunities. From a technical perspective, MRD testing requires exceptionally high analytical sensitivity to detect ultra-low concentrations of tumor signals. For example, ctDNA MRD assays that integrate epigenomic and genomic features can increase sensitivity by 25%–36% compared with relying on genomic variants alone [ 26 ]. In OC, microfluidic platforms can detect methylated DNA fragments at levels as low as 0.00005%. By contrast, conventional plasma genomic testing is typically used for screening or diagnosis and has relatively lower sensitivity requirements; for instance, CA125—an established OC marker—lacks sufficient sensitivity for early detection [ 27 ]. Although novel multi-marker panels have demonstrated improved sensitivity [ 27 ], they still fall short of the stringent detection thresholds required for MRD testing applications [ 28 ]. The choice of detection platform significantly influences both analytical performance and clinical applicability [ 29 ]. Tumor-informed approaches, which utilize patient-specific mutation profiles derived from whole exome sequencing of tumor tissue, generally achieve higher baseline detection rates (97.9%−100%) compared to tumor-type informed methods based on DNA methylation patterns [ 15 ]. However, tumor-informed strategies utilizing personalized structural variants and single nucleotide variants have demonstrated extraordinarily high pre- and postoperative detection rates (96% and 81%, respectively) with sensitivity down to 0.00009% VAF [ 30 ]. These personalized approaches employ whole exome sequencing on archived tumor tissue to identify patient-specific somatic variants, which are then tracked longitudinally in plasma samples using targeted panels. While tumor-informed methods offer superior sensitivity, they require accessible tumor tissue and entail additional sequencing costs, whereas tumor-agnostic approaches provide broader applicability but may sacrifice detection sensitivity in low-burden disease states [ 15 , 29 ]. In terms of intended use, MRD testing is primarily employed to monitor residual disease after treatment and to predict recurrence. Persistent postoperative ctDNA is significantly associated with recurrence-free survival (RFS) and can identify patients at high risk of relapse [ 26 , 31 ]. In contrast, conventional plasma genomic assays are mainly used for early diagnosis or risk stratification, such as detecting OC–related markers in urine or self-collected cervical samples [ 32 , 33 ]. From a clinical perspective, MRD testing must demonstrate a strong link with outcomes. For instance, ctDNA MRD has shown clinical relevance in resectable non–small-cell lung cancer (NSCLC), while in OC, MRD status may potentially guide adjuvant immunotherapy or targeted therapies based on iTMB or actionable mutations [ 34 ]. By contrast, conventional plasma assays are mostly applied to initial screening, such as cfDNA extraction and mutation analysis using microfluidic chips [ 35 ]. An emerging analytical dimension that complements mutation-based and methylation-based ctDNA detection is fragmentomics, the systematic analysis of the size distribution and genomic positioning of cfDNA fragments [ 36 , 37 ]. Cancer-derived cfDNA exhibits characteristic alterations in fragment length, nucleosomal positioning patterns, and end-motif composition compared with cfDNA released from normal cells, reflecting differences in chromatin accessibility and nuclease activity at sites of active transcription in tumor cells [ 38 ]. Cristiano et al. demonstrated that genome-wide cfDNA fragmentation profiles could distinguish cancer patients from healthy individuals with high accuracy, and that fragment size distributions carry tissue-of-origin information that complements methylation-based classification [ 36 , 39 ]. In the context of OC surveillance, fragmentomics holds particular promise for low tumor burden states, such as post-treatment MRD monitoring or early-stage screening, where the absolute quantity of ctDNA carrying detectable somatic mutations may fall below the analytical limits of conventional mutation-based assays [ 40 , 41 ]. Integration of fragmentomic features with mutation and methylation signals in multimodal analytical pipelines represents a logical next step for enhancing ctDNA sensitivity across OC disease stages [ 42 , 43 ]. Chapman et al. presented findings at the 2021 American Association of Cancer Research Annual Meeting demonstrating the utility of ctDNA for dynamic monitoring of MRD status, thereby enabling prediction of postoperative recurrence and prognosis in epithelial OC patients [ 44 ]. Among 21 patients followed for a median of 15 months, all ctDNA-positive patients experienced relapse (8/8), while ctDNA-negative patients remained disease-free within 6 months (13/13). MRD monitoring detected recurrence 2–36 months earlier than imaging (median lead time: 9 months). ctDNA positivity was strongly associated with inferior RFS (HR 12.75, 95% CI: 1.7–94, p  < 0.0001), whereas CA125 showed no significant correlation (HR 1.3, p  = 0.09). Similarly, Hou et al. analyzed serial blood samples from 25 epithelial OC patients and found that 23% (5/22) were ctDNA-positive at the first post-treatment sampling. All ctDNA-positive patients ultimately progressed, with ctDNA predicting recurrence on average 10 months earlier than imaging ( p  < 0.0001). Consistent with Chapman’s findings, CA125 was not significantly predictive ( p  = 0.056) [ 45 ]. Other studies further confirm the predictive value of ctDNA. In a retrospective analysis, ctDNA detection of recurrence demonstrated both 100% sensitivity and specificity, identifying relapse 10 months earlier than computed tomography (CT) imaging, compared to only 0.9 months for CA125 [ 45 ]. In another investigation of stage I–IV epithelial OC patients (post-surgery/under surveillance cohorts, N  = 25, median follow-up 2 years), ctDNA was exclusively detected in relapsed patients (100% sensitivity/specificity), exhibiting strong predictive power for relapse (HR: 17.6, p  = 0.001; p  < 0.0001), while CA125 positivity showed no significant correlation with relapse ( p  = 0.113; p  = 0.056), with ctDNA demonstrating clear superiority over CA125 in recurrence prediction [ 46 ]. Moreover, CTCs also provide valuable complementary information: CTC-positive patients have up to a 60% risk of developing chemotherapy resistance. The combination of ctDNA and CTC analyses offers complementary and synergistic insights into genetic mutations and phenotypic tumor changes, with emerging evidence suggesting that multi-analyte liquid biopsy strategies may improve overall recurrence detection sensitivity compared to single-analyte approaches [ 47 , 48 ]. To avoid conceptual redundancy with the subsequent section, this section will focus explicitly on prognostic and predictions of ctDNA-derived genomic alterations, how mutation status, variant allele fraction (VAF) dynamics, and clonal evolution patterns are linked to treatment response, resistance emergence and PFS and OS in OC patients. Prevalence, histotype-specific enrichment and co-occurence frequency of these mutations across OC histological subtypes (HGSC) endomentrioid, clear cell and mucinous are addressed separately in Sect.  2.4 to which cross-references are provided where relevant. Further studies reinforce the clinical importance of ctDNA. Pereira et al. demonstrated that ctDNA predicted recurrence 7 months earlier than CT imaging, even when scans were negative [ 49 ]. Wimberger et al. reported that cfDNA levels correlated with residual tumor burden > 1 cm and were linked to higher recurrence risk ( p  = 0.0001) and poorer OS ( p  = 0.021) [ 50 ]. Harris et al. identified chromosomal rearrangements in tumor tissue and tracked them in cfDNA of 10 advanced OC patients, detecting persistent disease in 3 patients with postoperative ctDNA positivity [ 51 ]. Other analyses confirmed that ctDNA mutations often appear 2–5 months before clinical or radiological progression [ 24 ]. A meta-analysis of eight studies further showed ctDNA positivity as an independent risk factor for worse OS (HR 2.36, 95% CI 1.76–3.17) and PFS (HR 2.51, 95% CI 1.83–3.45) [ 52 ]. Among ctDNA-detectable genetic alterations, TP53 mutant allele fraction (VAF) dynamics form a sensitive readout of treatment efficacy and disease trajectory, between patients receiving platinum-based chemotherapy a over 60% reduction in TP53 VAF after one cycle of treatment correlates with prolonged PFS, whereas continued TP53-mutant ct DNA after six weeks of treatment predicts radiographic progression and the increasing tumor size [ 43 ]. Crucially, the prognostic value of TP53 VAF kinetics is independent from its mutational prevalence across OC subtypes as it is not the existence of TP53-mutant ctDNA, but rather the dynamic change in VAF which yields prognostic value for serial ctDNA monitoring [ 43 ]. In addition to the TP53 VAF dynamics, KRAS-mutant ctDNA (most frequently detectable within low-grade serous and mucinous histological types) forms a real-time pharmacokinetic readout of MEK inhibitor response and serves as an early predictor of acquired resistance, as VAF rebound is observed during treatment [ 47 ]. Beyond TP53 and KRAS , emerging evidence highlights the prognostic value of other genetic alterations, including BRCA1/2 mutations, homologous recombination deficiency (HRD) signatures, and copy number variations. ctDNA-based detection of BRCA1/2 mutations has been shown to predict treatment response to PARP inhibitors and guide maintenance therapy selection, with longitudinal profiling enabling real-time tracking of clonal evolution and the emergence of resistance mechanisms, including BRCA reversion mutations and drug efflux transporter upregulation, prior to radiographic progression [ 53 , 54 ]. The implementation of tumor-informed versus tumor-agnostic ctDNA detection approaches represents a critical methodological consideration. Tumor-informed assays, which utilize personalized panels based on tumor sequencing, demonstrate superior sensitivity (up to 95%, tumor-informed personalized panel, high-shedding HGSOC, LOD < 0.01% VAF) for MRD detection compared to tumor-agnostic approaches (60–75%, fixed gene panel or methylation-based; across mixed OC stages) [ 29 ]. However, tumor-agnostic methods offer practical advantages including faster turnaround time and applicability when tissue samples are unavailable [ 15 ]. Recent hybrid approaches combining both strategies show promise in optimizing detection accuracy while maintaining clinical feasibility [ 55 ]. In conclusion, ctDNA dynamics can be used to monitor tumor progression and prognosis, offering a more precise reflection of tumor changes than traditional CA125 detection. The integration of multi-omic ctDNA analysis, encompassing somatic mutations, epigenetic modifications, and fragmentomic patterns, further enhances prognostic stratification and enables more comprehensive disease monitoring [ 56 ]. However, ctDNA may not capture all genetic variations present in tissue, and some researchers argue that ctDNA testing lacks sufficient validation for widespread clinical implementation, With accurate treatment matching feasible in only a limited number of patients, technical confounders, including leukocyte-derived background DNA release, which increases noise and impairs variant allele frequency detection at low tumor fractions – represent key analytic challenges [ 48 ]. A further source of false-positive ctDNA signals arises from clonal hematopoiesis of indeterminate potential (CHIP), whereby somatic mutations in hematopoietic stem cells – unrelated to the primary tumor – are detectable in cfDNA; the latter phenomenon and the strategies to mitigate it are discussed in more detail in the Limitations section. Single-cell analysis of CTCs and ctDNA provides insights into tumor heterogeneity and dormant MRD populations that may seed future recurrence [ 56 ]. These dormant cells exhibit distinct metabolic and epigenetic signatures detectable through liquid biopsy, potentially enabling targeted therapeutic interventions before clinical relapse [ 57 ]. This underscores the irreplaceable role of tissue biopsy in providing comprehensive tumor information and definitive diagnosis [ 58 ]. Thus, the question of whether ctDNA can fully replace tissue sequencing remains to be explored in future studies. Current consensus advocates for a complementary approach wherein tissue biopsy establishes initial diagnosis and molecular profiling, while serial ctDNA monitoring provides real-time disease surveillance and treatment response assessment [ 59 ]. Monitoring ctDNA levels and mutations allows researchers to detect postoperative MRD early, providing valuable guidance for clinicians in postoperative and post-chemotherapy treatment decisions. While ctDNA demonstrates notable advantages in sensitivity and specificity, its broader adoption in clinical practice requires further research and validation. The following section will focus on the role of ctDNA in predicting OC recurrence and its applications in early identification and management of recurrence. As a novel biomarker, ctDNA demonstrates significant potential in post-surgical monitoring of OC. Research indicates that ctDNA detection enables early post-surgical identification of MRD and can predict recurrence risk. For instance, ctDNA mutation detection has been used to monitor disease status in patients with endometrial cancer (EC) and OC. Data from 44 patients (24 EC, 17 °C, 2 with concurrent EC and OC, and 1 endocervical adenocarcinoma [EA]) indicated that at least one somatic mutation was detected in the tumor tissue of 91% of patients (20/24 EC, all OC/SEOC/EA), while 12 patients (27%) exhibited detectable mutations in preoperative plasma ctDNA (6/24 [25%] EC and 6/17 [35%] OC). Preoperative ctDNA mutation detection correlated with advanced disease stages, elevated preoperative CA125 levels, and recurrence. Among the 12 patients with detectable preoperative ctDNA mutations, 5 (42%) were classified as clinical stage I by preoperative imaging, but final pathology revealed stage II/III disease. Moreover, in 11 patients who underwent serial ctDNA and CA125 assessments during treatment, ctDNA clearance was observed before CA125 normalization. Among the 13 patients who experienced disease recurrence (4 EC, 8 °C, 1 EA), 8 had ctDNA mutations detected post-surgery, while 4 were monitored through ctDNA mutations upon recurrence, with these mutations detected 2–5 months before clinical, radiological, or biomarker progression [ 24 ]. These findings indicate that ctDNA testing holds clinical potential for early MRD detection post-surgery and for predicting recurrence risk. Monitoring ctDNA levels and mutations enables earlier identification of disease recurrence, allowing for more timely treatment interventions for patients [ 24 ]. Importantly, the kinetics of ctDNA clearance during treatment provide dynamic prognostic information. Rapid ctDNA clearance (> 90% reduction within first treatment cycle) correlates with improved outcomes, while slow clearance or rebound indicates primary or acquired resistance [ 60 ]. Multi-analyte liquid biopsy approaches incorporating ctDNA, CTCs, tumor-educated platelets, and extracellular vesicles demonstrate enhanced sensitivity and specificity compared to single-analyte strategies [ 53 ]. For OC patients, ctDNA detection provides a more sensitive early indication of recurrence compared to conventional imaging and tumor marker surveillance. For instance, a study involving 63 EOC patients analyzed 264 plasma samples, detecting ctDNA in 51 of 55 diagnostic samples (93%) and in all 18 samples (100%) during disease progression. Throughout treatment, the presence of ctDNA in the final treatment sample was associated with accelerated disease progression (median time 1.02 years vs. 3.38 years, HR = 5.63, P  < 0.001) and decreased OS (median time 2.31 years vs. not reached, HR = 8.22, P  < 0.001), especially among patients with HGSC. In 12 patients, ctDNA analysis identified disease progression earlier than conventional monitoring, with a median lead time of 5.9 months. To approach the detection limits of ctDNA, researchers employed an ultra-sensitive method on 5 patients, detecting ctDNA fractions as low as 0.0004% and encompassing 479 to 1,856 tumor mutations. Research by Trevisi et al. indicates that ctDNA has significant clinical value in monitoring MRD and predicting recurrence in OC patients. Routine ctDNA analysis after surgery and chemotherapy achieves an accuracy rate of over 90% (post-surgical and post-chemotherapy monitoring; HGSOC predominant cohort) for predicting recurrence. Especially when imaging does not show clear signs of recurrence, ctDNA can already detect MRD [ 61 ]; Heo et al. [ 62 ] used a ctDNA analysis method without tumor sequencing to successfully detect MRD and predict recurrence in patients. The study demonstrated a sensitivity of 92% ( n  = 328 °C patients, tumor-agnostic cfDNA, post-platinum chemotherapy setting) for ctDNA detection among 328 °C patients, with particularly significant clinical value following platinum chemotherapy. For patients who tested ctDNA-positive after treatment, the recurrence risk was significantly higher than in ctDNA-negative patients. The study further found that within 6 months post-surgery, the recurrence rate was 80% (MRD-positive, within 6 months post-surgery, n  = 328) and 35% (MRD-negative, same cohort) among patients with detectable MRD, compared to 35% among those without detectable MRD. These findings suggest that ctDNA testing offers high sensitivity and specificity for early detection of MRD post-surgery and for predicting recurrence risk in EOC patients.The outcomes of ctDNA detection are significantly associated with patient survival. In one study, cfDNA, tumor DNA, and germline DNA were extracted from plasma, formalin-fixed paraffin-embedded tumor tissue, and white blood cells of 48 women diagnosed with serous OC (47 high-grade, 1 low-grade). Sequencing targeted 59 driver genes linked to breast and OC using plasma samples collected post-surgery and pre-chemotherapy. ctDNA testing identified tumor-specific variants in all 47 patients, and these variants were further analyzed in postoperative plasma samples [ 63 ]. Among the patients, fifteen (31.9%) exhibited visible MRD, with ctDNA detected in all cases. In contrast, 31 patients (65.9%) showed no visible MRD, yet ctDNA was detected in 24 (77.4%) of them. Among those without visible residual disease, patients with detectable ctDNA had a higher mortality rate (20 of 27 deaths) compared to those without detectable ctDNA (3 of 7 deaths) (HR 2.32; 95% CI: 0.67–8.05), though this difference did not reach statistical significance ( p  = 0.18). Emerging evidence supports the integration of ctDNA monitoring into clinical decision-making algorithms. MRD-directed therapy, wherein treatment intensification or maintenance strategies are guided by ctDNA status, shows promise in early-phase trials. Patients with persistent MRD after frontline therapy may benefit from extended PARP inhibitor maintenance or novel immunotherapy combination; Conversely, ctDNA-negative patients with low recurrence risk may safely de-escalate surveillance intensity, reducing healthcare costs and patient burden; prospective randomized trials evaluating ctDNA-guided treatment adaptation are currently underway and will determine whether this precision medicine approach translates to survival benefits [ 64 ]. It should be noted that these represent hypothesis-generating observations, prospective randomized evidence demonstrating that ctDNA-guided treatment modification improves survival in OC remains to be established, as addressed in detail in Sect. 2.4 . The study suggests that ctDNA in postoperative serum may signal the presence of MRD and potentially predict recurrence in OC patients. These findings imply that ctDNA testing could be a valuable tool for evaluating the risk of disease recurrence following surgery. As shown in Fig.  2 , ctDNA shows high sensitivity in detecting MRD after surgery, with sensitivity values reported in various studies ranging from 80% to 90%. This chart presents the data from studies by Ettegowda et al., Yurkovetsky Z et al., and others, highlighting the consistency and effectiveness of ctDNA in detecting MRD. Future directions include the development of cost-effective, point-of-care ctDNA assays enabling real-time monitoring, as well as the integration of artificial intelligence algorithms to interpret complex ctDNA patterns and predict clinical trajectories with greater precision [ 29 ]. The prognostic and predictive implications of ctDNA-detectable mutations discussed above are contextualized by their differential prevalence across OC histological subtypes, which is addressed separately in Sect. 2.4 . Fig. 2 Sensitivity of ctDNA in Post-Surgical MRD Detection Sensitivity of ctDNA in Post-Surgical MRD Detection A fundamental distinction must be drawn between two levels of clinical evidence regarding ctDNA in OC: prognostic association and interventional benefit. The current literature robustly supports the former: ctDNA positivity in the postoperative or post-treatment setting is an independent predictor of inferior RFS and OS, consistently outperforming CA125 in lead-time to recurrence detection across multiple cohorts [ 49 ]. However, evidence for the latter, that ctDNA-guided treatment modification actually improves patient survival outcomes, is currently absent in OC. To date, no prospective randomized controlled trial has demonstrated that the adaptation of therapeutic strategy based on ctDNA MRD status (for example, intensifying maintenance PARP inhibitor therapy in the MRD-positive setting, or safely de-escalating surveillance intensity in the MRD-negative setting) translates into statistically significant improvements in PFS or OS in OC [ 65 ]. This gap in demonstrating clinical utility from predictive biomarker performance is not confined to OC, analogous challenges in validation have been documented in other solid tumors in which ctDNA-guided interventional trials are similarly ongoing [ 66 , 67 ]. Several prospective interventional trials specifically designed to determine the value of ctDNA-directed treatment decision-making in OC are currently underway, including the evaluation of ctDNA-guided escalation of maintenance therapy following primary treatment ( NCT06182917 , NCT06561178 ) [ 68 ]. These pivotal trials will provide the critical data needed to determine if MRD status as detected by ctDNA is a robust trigger for therapeutic intervention with measurable survival benefit [ 69 ]. Until such results are mature, the translation of ctDNA-directed treatment modification should be considered investigational, and the integration of these tests into clinical practice outside of the trial setting must proceed with caution. This evidence gap also has implications for health technology assessment and reimbursement: regulatory bodies and payers increasingly require demonstration of clinical utility (not just prognostic association) before approving coverage of novel biomarker tests [ 70 ]. Prospective validation of ctDNA-guided therapeutic algorithms in OC, therefore, represents not only a scientific priority, but also a prerequisite for equitable clinical implementation [ 66 ]. While ctDNA monitoring can reflect tumor burden and treatment response, its clinical efficacy depends on several critical factors, including detection sensitivity, inter-patient variability, and optimization of clinical treatment strategies. A fundamental gap exists between the demonstrated analytical performance of ctDNA assays and their clinical utility—the ability to meaningfully impact patient outcomes through actionable interventions [ 70 ]. In postoperative monitoring of OC patients, ctDNA detection is closely related to recurrence risk, but no adequately powered randomized controlled trials have demonstrated a direct correlation between ctDNA-guided interventions and improvements in OS or PFS [ 71 ]. For instance, patients with ctDNA detected postoperatively experienced recurrence approximately 9 months earlier than those without detectable ctDNA [ 72 ]. However, the critical question is not whether ctDNA can detect recurrence early, but whether this early detection translates into therapeutic interventions that improve survival outcomes [ 73 ]. Several ongoing phase III trials (TRACER, DYNAMIC-Recur) are evaluating whether treatment intensification based on ctDNA positivity improves outcomes compared to standard surveillance [ 74 ]. Standardization of pre-analytical workflows remains a major barrier, as evidenced by significant inter-laboratory variability even when analyzing identical samples [ 70 ]. Detection sensitivity varies considerably depending on tumor burden and biology. While tumor-informed approaches can achieve sensitivities exceeding 95% for high-shedding tumors [ 75 ], sensitivity drops to 40–60% for low-volume disease [ 76 ]. Certain histological subtypes, particularly mucinous and clear cell carcinomas, demonstrate lower ctDNA shedding rates compared to HGSC [ 77 ]. Additionally, CHIP can cause false-positive MRD calls, requiring white blood cell sequencing for filtering [ 78 , 79 ]. Tumor heterogeneity poses another challenge, as ctDNA may not capture spatial heterogeneity comprehensively, particularly for metastases in sanctuary sites or minor subclones harboring resistance mutations [ 80 , 81 ]. Current commercial ctDNA assays cost $2,000-$5,000 per test, with longitudinal monitoring requiring cumulative costs of $10,000-$30,000 per patient annually [ 82 ]. Cost-effectiveness analyses suggest favorable outcomes only if early intervention improves survival and reduces downstream costs [ 83 ]. Reimbursement policies lag behind, with Medicare coverage for OC monitoring not yet established [ 84 ]. Integration into clinical workflows requires addressing sampling frequency, validated action thresholds, rapid turnaround times, and clinician training [ 85 ]. Technological innovations promise to overcome current limitations. Next-generation approaches including duplex sequencing, multi-analyte strategies (combining ctDNA with CTCs, ctRNA), fragmentomic analysis, and epigenetic profiling demonstrate improved sensitivity and specificity [ 86 – 90 ]. Machine learning algorithms trained on longitudinal ctDNA trajectories may predict treatment response with greater accuracy [ 91 ], while point-of-care devices under development aim to reduce costs and turnaround times [ 92 ]. These trials will determine whether ctDNA-guided management improves patient outcomes beyond prognostic information. Patient-centered considerations warrant attention, including the psychological impact of persistent MRD positivity and health equity concerns regarding access disparities [ 70 ]. Shared decision-making frameworks and patient education materials should clearly communicate the probabilistic nature of ctDNA results. In conclusion, while ctDNA demonstrates remarkable potential for MRD detection in OC, substantial barriers remain before routine clinical implementation. The field must transition from proof-of-concept studies to rigorous clinical utility trials proving that ctDNA-guided management improves patient outcomes. This requires standardization of workflows, large-scale prospective trials, cost-effectiveness validation, clear regulatory frameworks, technological innovations, and patient-centered research. The next decade will determine whether ctDNA fulfills its promise as a transformative tool in precision oncology for OC. Collaborative efforts among researchers, clinicians, industry, regulators, and patient advocates are essential to improve outcomes for women with OC.

ctDNA has demonstrated clinical utility in diagnosis, recurrence prediction, and postoperative residual disease monitoring in OC. It is important to contextualize the current level of evidence linking early ctDNA positivity to survival outcomes in OC. While multiple studies have demonstrated that postoperative ctDNA detection is associated with significantly shorter progression-free survival (PFS) and overall survival (OS), the critical question of whether ctDNA-guided therapeutic intervention—triggered by MRD positivity prior to clinical or radiological relapse, translates into improved survival outcomes remains unanswered by prospective randomized evidence [ 132 ]. The existing data are largely prognostic and observational in nature: ctDNA positivity identifies patients at high relapse risk, but no completed randomized controlled trial has yet demonstrated that initiating or intensifying therapy based on ctDNA MRD status improves OS in OC 65 . This distinction between prognostic utility (ctDNA predicts who will relapse) and predictive/therapeutic utility (acting on ctDNA results improves outcomes) is clinically fundamental. Several ongoing trials are designed to prospectively test whether MRD-guided treatment decisions improve survival, but mature outcome data are not yet available [ 69 ]. Until such evidence is established, ctDNA MRD positivity should be interpreted as a high-risk stratification tool warranting intensified surveillance, rather than a validated trigger for treatment escalation outside of clinical trial settings. Nevertheless, despite its promising prospects in numerous studies, ctDNA faces a range of technical and clinical challenges that must be addressed before routine clinical implementation can be achieved. These include limitations in analytical sensitivity at low tumor fractions, the absence of standardized pre-analytical protocols, the biological heterogeneity of ctDNA shedding across OC histotypes, and the potential for false-positive signals arising from non-tumor sources. Among the latter, clonal hematopoiesis of indeterminate potential (CHIP)—introduced briefly in Sect. 2.3 , represents one of the most clinically significant confounders affecting ctDNA interpretation in OC, particularly in the context of MRD detection and recurrence monitoring, and is therefore addressed in dedicated subsections below. CHIP is defined as the presence of somatic mutations in hematopoietic stem cells with a VAF of over 2%, occurring in individuals without diagnosed hematologic malignancies [ 133 , 134 ]. CHIP is characterized by somatic mutations occurring most commonly in genes involved in epigenetic regulation and DNA damage response, including DNMT3A , TET2 , ASXL1 , TP53 , and PPM1D . These mutations arise as a consequence of age-related replicative stress and can confer a selective growth advantage to mutant clones. The prevalence of CHIP increases exponentially with age. In the general population, CHIP is detected in approximately 10–15% of individuals aged 65–70 years and rises to over 30% in those older than 80 years. In cancer patients, particularly those receiving genotoxic therapies, CHIP frequency may be even higher due to therapy-induced selective pressure on resistant clones [ 135 ]. Studies have shown that platinum-based chemotherapy and PARP inhibitors, commonly used in OC treatment, can promote expansion of pre-existing CHIP clones, with VAF increases of 2–5 fold observed in 20–40% of treated patients [ 78 ]. In OC cohorts specifically, limited but emerging data suggest CHIP prevalence rates of approximately 15–20%, consistent with age-matched general population controls [ 136 ]. DNMT3A mutations represent the most prevalent CHIP alteration in OC patients (approximately 40–42% of CHIP cases), followed by TET2 (∼26%), ASXL1 (∼12%), and TP53 (∼8%), the latter being of particular concern given its near-universal presence as a tumor-derived mutation in HGSC [ 137 ]. Importantly, the median VAF of CHIP mutations typically ranges from 2 to 10%, which critically overlaps with the VAF range of MRD (0.01–5%), creating substantial challenges for ctDNA interpretation in the MRD monitoring context. Given that the median age at OC diagnosis is 63 years, with approximately 40% of patients diagnosed after age 65, CHIP represents a critical confounder that must be systematically addressed in ctDNA-based MRD strategies. While the theoretical mechanisms by which CHIP may confound ctDNA interpretation have been well characterized, it is critical to distinguish between the theoretical risk and the frequency with which CHIP has actually been empirically documented to cause analytical errors in published ctDNA studies specific to OC. 136,138 Data specifically quantifying the proportion of false-positive ctDNA results arising from CHIP in OC-dedicated cohorts remain limited [ 133 , 137 ]. However, an increasing body of literature from pan-cancer studies, gynecologic oncology studies, and cfDNA methodology studies provide important quantitative estimates [ 133 , 139 , 140 ]. One of the largest and most methodologically rigorous analyses to date performed tumor-normal matched genomic profiling in a total of 8,812 patients across multiple solid tumor types and found CHIP mutations in 25.2% of cases [ 140 ]. Critically, however – when tumor-only sequencing was performed without matched normal tissue (as is commonly performed in clinical ctDNA assays), a substantial proportion of CHIP-derived variants were misclassified as somatic tumor mutations, yielding false-positive calls that could not be distinguished from true tumor-derived signal without germline comparators. Although the OC-specific subgroup statistics were not reported in this analysis, the overall misclassification rate of this work highlighted the systemic risk of CHIP confounding in unfiltered plasma sequencing workflows. The challenge of CHIP-derived false positives in the context of plasma sequencing was directly quantified by Hu et al., in which the authors evaluated paired plasma and WBC DNA samples from cancer patients undergoing cell-free DNA profiling [ 138 ]. In this study, 32% of plasma variants initially detected in cfDNA were found to originate from clonal hematopoiesis rather than tumor tissue when matched WBC sequencing was applied as a filter. The majority of these CHIP-derived false positives involved mutations in DNMT3A, TET2, ASXL1, and TP53, genes that are not only the most prevalent CHIP drivers but are also recurrently mutated in OC, particularly in high-grade serous carcinoma (HGSOC), where TP53 mutation is present in over 95% of cases. These findings directly illustrate why CHIP confounding is most problematic in the OC context: the near-universal overlap between tumor-defining TP53 mutations in HGSOC and CHIP-derived TP53 variants results in the emergence of analytically indistinguishable signals without orthogonal WBC-based filtering. Evidence specifically examining CHIP confounding in the gynecologic malignancy and ovarian cancer context was provided by Razavi et al., who applied ultra-deep sequencing to cfDNA samples of solid tumor patients and systematically characterized the cellular origin profiles of plasma variants [ 140 ]. This study revealed that the fraction of detectable cfDNA mutations derived from CHIP was substantial and previously underestimated, particularly in patients over age 70, and that the variant allele frequency (VAF) of this CHIP-derived signal (typically 2–10%) critically overlapped the range predicted for detection of minimal residual disease (MRD) at 0.01–5%, making discrimination between these two cellular sources analytically challenging in the absence of paired leukocyte sequencing [ 140 ]. The clinical implications of unfiltered CHIP signal within serial ctDNA monitoring were further demonstrated by Bolton et al., who examined the dynamics of CHIP clone expansion in cancer patients undergoing cytotoxic therapy [ 135 ]. In patients receiving platinum-based chemotherapy, the standard backbone of treatment for OC in particular, CHIP clones harboring TP53, PPM1D, and DNMT3A mutations exhibited selective expansion, with VAF increases of 2- to 5-fold documented in 20–40% of treated patients, resulting in therapy-induced clonal expansion with rising VAF trajectories in serial cfDNA measurements which are phenotypically indistinguishable from emergent MRD or early relapse signals and represent a direct risk for misinterpretation in the post-treatment monitoring context [ 135 ]. Regarding OC-specific cohort data, Adalsteinsson et al. performed a scalable whole-exome sequencing of cfDNA from 1,439 tumor cases across multiple cancer types, including a subset of ovarian cancer patients, and showed high concordance between cfDNA and matched tumor profiles when appropriately controlled with germline filtering [ 141 ]. However, their analysis also showed that without matched normal tissue comparator samples, a meaningful fraction of variant calls, particularly those in TP53, epigenetic regulator genes, and others could not be reliably attributed to tumor origin versus hematopoietic origin, reinforcing the need for WBC-matched sequencing for any ctDNA-based workflow intended for MRD applications in OC [ 141 ]. Collectively, the available evidence suggests that CHIP-attributable false-positive rates in unfiltered ctDNA assays range from approximately 14–25% across studies of cancer populations of comparable age distribution to OC cohorts, and that these rates are strongly reduced to the range of 3–5%, upon implementation of paired WBC sequencing and CHIP-aware bioinformatic filtering [ 140 ]. However, it should be acknowledged that fully comprehensive prospectively designed studies specifically quantifying CHIP-attributable false-positive rates in OC ctDNA MRD assays represent an unmet need. Existing estimates are partially extrapolated from pan-cancer or methodology-focused datasets and may not fully capture the OC-specific context, where the near-universal prevalence (over 95%) of TP53 mutations in HGSOC creates an especially complex overlap between tumor-derived and CHIP-derived variants. The median age at OC diagnosis of 63 years, with approximately 40% of patients diagnosed after age 65, further amplifies this risk given the exponential age dependence of CHIP prevalence [ 142 ]. Dedicated prospectively designed studies in OC cohorts with paired WBC sequencing as a mandatory standard component of ctDNA MRD assay analysis are urgently needed to generate robust, OC-specific estimates of CHIP confounding frequency, and to validate the performance of computational deconvolution strategies in this clinical setting. The presence of CHIP poses significant challenges to the accuracy and interpretation of mutation-based ctDNA assays, particularly in the context of MRD detection following curative-intent surgery. CHIP can compromise ctDNA testing through multiple mechanisms. First, when the same genes are mutated in both hematopoietic and tumor tissues (e.g., TP53 , which is mutated in > 90% of high-grade serous OCs and also represents 5–10% of CHIP mutations), cell-free DNA derived from CHIP clones can be erroneously attributed to tumor origin [ 79 , 143 ]. Without paired white blood cell sequencing to identify germline and CHIP variants, false-positive rates in ctDNA assays have been reported to reach 15–30% in elderly patient cohorts [ 67 , 106 ], with the highest rates observed in assays employing tumor-agnostic panels that cannot distinguish tissue of origin for TP53 variants. Second, CHIP-derived cell-free DNA can constitute 30–50% of total cfDNA in elderly patients, effectively diluting tumor-derived ctDNA and reducing analytical sensitivity for true MRD detection. This is especially problematic in the post-surgical setting, where ctDNA levels are inherently low and maximal sensitivity is required. Third, serial ctDNA measurements are increasingly used to monitor treatment response and detect early recurrence. However, CHIP clones can expand during chemotherapy, mimicking disease progression [ 78 ]. Studies have documented that therapy-related clonal expansion can result in rising VAFs that are misinterpreted as emerging MRD, leading to unnecessary interventions or premature escalation of therapy. Fourth, the confounding effect of CHIP on ctDNA interpretation is highly age-dependent, necessitating risk-stratified approaches. In patients younger than 50 years, CHIP prevalence is less than 5% with minimal impact on ctDNA interpretation. In patients aged 50–65 years, CHIP prevalence reaches 8–12%, making baseline white blood cell sequencing recommended. In patients older than 65 years, CHIP prevalence exceeds 15%, mandating paired WBC sequencing to filter hematopoietic variants [ 133 ]. These age-dependent effects underscore the necessity for tailored approaches to ctDNA analysis in OC, where a substantial proportion of patients fall into high-risk age categories. Addressing these CHIP-related confounders through systematic incorporation of matched WBC sequencing, prospective validation of computational filtering pipelines, and dedicated OC-specific cohort studies represents a critical priority for elevating the clinical reliability of ctDNA-based MRD assessment. The following sections address additional technical and biological challenges alongside emerging directions for methodological advancement in OC liquid biopsy research. To address the challenge of CHIP in ctDNA analysis, several complementary approaches have been developed and are being increasingly integrated into clinical practice. Tumor-informed sequencing approaches, which first sequence the primary tumor tissue and then design patient-specific assays targeting tumor-confirmed mutations, can achieve ultra-high sensitivity (down to 0.001% VAF) while avoiding CHIP-related false positives [ 144 ]. This strategy is particularly effective for MRD monitoring in the post-surgical setting. Paired white blood cell sequencing, involving concurrent sequencing of peripheral blood mononuclear cells or buffy coat DNA, allows for direct identification of CHIP variants, which can then be filtered from ctDNA analysis [ 79 ]. This approach has been shown to reduce false-positive rates by 60–80% in mutation-based ctDNA assays. Methylation-based detection offers another powerful solution, as CHIP mutations do not typically alter genome-wide methylation patterns in a cancer-specific manner, meaning methylation-based ctDNA assays are largely unaffected by CHIP interference [ 145 ]. Platforms utilizing cell-free methylated DNA immunoprecipitation sequencing (cfMeDIP-seq) or bisulfite sequencing can detect cancer-specific methylation signatures independent of somatic mutations. Multi-marker integration, combining ctDNA analysis with orthogonal biomarkers such as CA125, HE4, or CTCs, can provide additional specificity, as concordance between rising ctDNA levels and elevated protein biomarkers or CTCs significantly increases positive predictive value for true disease recurrence. Bioinformatic filtering algorithms, particularly machine learning approaches trained on databases of known CHIP mutations (e.g., COSMIC, ClinVar), can computationally flag likely CHIP variants based on gene identity, VAF range, and age-associated mutational signatures [ 146 ]. While imperfect, these tools provide an additional layer of filtering when paired WBC sequencing is unavailable. Finally, longitudinal trajectory analysis, rather than relying on single time-point measurements, employs serial sampling with kinetic modeling to help distinguish CHIP expansion (typically gradual, linear increases) from true MRD emergence (often exponential growth) [ 75 ]. Correlation with radiographic imaging and clinical assessment further enhances specificity, creating a multidimensional framework for accurate disease monitoring. The confounding effects of CHIP on ctDNA testing have important implications for clinical implementation in OC across different clinical scenarios. For screening and early detection in asymptomatic populations, where cancer prevalence is low (typically less than 1%), even modest false-positive rates from CHIP can result in poor positive predictive values. Age-adjusted thresholds and methylation-based approaches may be preferable for population-level screening [ 145 ]. For MRD monitoring, the post-surgical MRD setting, where ctDNA levels are extremely low (often below 0.1% VAF), is most vulnerable to CHIP interference. Tumor-informed approaches with paired WBC sequencing should be considered the standard of care for accurate MRD assessment in OC patients, particularly those over age 60 [ 79 , 144 ]. For treatment response monitoring during active therapy, where ctDNA levels are higher and kinetic changes more dramatic, CHIP interference is less problematic. However, clonal expansion during chemotherapy necessitates baseline WBC sequencing for accurate interpretation of longitudinal trends [ 78 ]. Importantly, current guidelines from the European Society for Medical Oncology and the American Society of Clinical Oncology emphasize that ctDNA assays lacking CHIP-filtering strategies should not be used for clinical decision-making outside of research protocols. As ctDNA testing becomes increasingly integrated into OC management, systematic approaches to identify and mitigate CHIP interference will be essential to ensure clinical validity and utility, particularly as regulatory approval and reimbursement frameworks evolve. Beyond CHIP interference, ctDNA detection faces significant technical limitations that must be addressed for widespread clinical adoption. The concentration of ctDNA in peripheral blood is typically very low, particularly in early-stage disease or during remission, and ctDNA release varies by tumor type, location, and vascular perfusion, with certain OC exhibiting minimal shedding. This variability reduces the sensitivity of ctDNA-based tests, particularly for detecting small-volume disease. Sample quality also significantly impacts ctDNA detection accuracy, as variations in blood collection methods, storage conditions, and processing protocols can lead to inconsistent results. For instance, improper handling may cause contamination with genomic DNA from lysed white blood cells, increasing background noise and potentially obscuring tumor-derived signals. Tumor heterogeneity poses another major challenge, as OC is known for its significant intra- and inter-tumoral heterogeneity, meaning that different regions of the same tumor—or even different metastatic sites—may harbor distinct mutational profiles [ 75 ]. Because ctDNA represents a mixture of DNA from various tumor sites, it may not fully capture all genetic alterations present, leading to incomplete genomic characterization. Furthermore, the dynamic nature of ctDNA is influenced by various physiological and therapeutic factors. For example, surgical resection, chemotherapy, or radiation can temporarily alter ctDNA levels, complicating longitudinal monitoring. The kinetics of ctDNA clearance after treatment are not fully understood, and residual ctDNA may not always correlate with viable tumor burden, raising questions about optimal timing for post-treatment surveillance. These multifaceted challenges underscore the need for continued refinement of ctDNA technologies and standardization of clinical protocols. Currently, ctDNA detection strategies fall into two major categories with distinct advantages and limitations: tumor-informed and tumor-agnostic approaches. Tumor-informed approaches rely on prior sequencing of primary tumor tissue to identify patient-specific mutations, after which personalized ctDNA panels are designed to track these specific alterations in blood [ 144 ]. This method offers high sensitivity and specificity for MRD detection, as it focuses exclusively on known tumor-derived variants. However, it requires access to tumor tissue, which may not always be available, especially in cases where surgery is not performed or biopsy material is insufficient. Additionally, tumor-informed approaches are resource-intensive and time-consuming, limiting their scalability for widespread clinical use. Tumor-agnostic approaches, on the other hand, do not require prior knowledge of tumor mutations and instead employ broad genomic or epigenomic profiling to detect cancer-associated signals in plasma [ 145 ]. Methylation-based assays, for instance, analyze cancer-specific DNA methylation patterns that distinguish malignant from normal cells. These methods are particularly attractive for screening and early detection, as they can be applied universally without tissue requirements. However, tumor-agnostic approaches generally have lower sensitivity for MRD detection compared to tumor-informed methods and may be more susceptible to false positives from non-tumor sources. Recent advances have sought to combine the strengths of both approaches through hybrid strategies that integrate initial tumor-agnostic screening with subsequent tumor-informed tracking for confirmed cancer patients, optimizing both detection sensitivity and clinical applicability [ 82 ]. This integrated paradigm represents a promising direction for comprehensive liquid biopsy implementation across the cancer care continuum. The choice between tumor-informed and tumor-agnostic ctDNA strategies should be guided by the specific clinical objective and practical constraints of each setting. In the postoperative MRD monitoring context, where ctDNA concentrations are inherently low (often < 0.1% variant allele frequency [VAF]) and maximum analytical sensitivity is required, tumor-informed approaches utilizing patient-specific mutation panels designed from primary tumor sequencing are strongly preferred. Surgical resection routinely provides sufficient tissue for tumor sequencing, and the clinical stakes of false-negative MRD results (missed early recurrence) justify the additional cost and turnaround time [ 70 , 147 ]. During active treatment response monitoring, where ctDNA levels are generally higher and kinetic changes more pronounced, both tumor-informed and tumor-agnostic approaches can provide informative data. The higher signal-to-noise ratio in this setting reduces the sensitivity advantage of tumor-informed methods, making tumor-agnostic platforms a pragmatic alternative, particularly when tissue is unavailable or turnaround time is critical [ 29 ]. For population-level early detection screening in asymptomatic women, tumor-informed approaches are inherently inapplicable due to the absence of prior tumor characterization. Tumor-agnostic, methylation-based platforms such as Galleri (GRAIL) represent the only currently feasible strategy for this indication, though sensitivity limitations in stage I–II OC remain a significant challenge [ 145 ]. For patients who have completed primary treatment, tumor-informed monitoring offers the advantage of tracking known patient-specific variants, enabling earlier detection than imaging, at the cost of requiring accessible archival tumor tissue. When tissue is unavailable or inadequate for sequencing, a clinically common scenario in heavily pretreated patients, tumor-agnostic approaches or liquid biopsy-derived molecular profiling may serve as an acceptable alternative, with the understanding that sensitivity is reduced [ 148 ]. These scenario-specific considerations underscore that no single ctDNA strategy is universally optimal, rather, a clinically rational approach requires matching the detection paradigm to the specific monitoring objective, patient characteristics, and available resources. Liquid biopsies, particularly ctDNA-based assays, hold promising cost-effectiveness and accessibility benefits in real-world clinical applications [ 149 ]. In comparison to conventional screening methods which often rely on expensive imaging procedures or invasive tissue biopsies, ctDNA testing offers the ease of simple and minimally invasive blood draws, thereby reducing procedural risks, patient burden, and healthcare resource consumption. Recent real-world experience from large-scale MCED tests has yielded high assay success rates, fast turnaround times, as well as high patient acceptance, highlighting the potential for widespread implementation [ 150 ]. More importantly, ctDNA assays have the potential to expand the scope of screening for cancers that currently do not have dedicated screening protocols, thereby increasing overall early detection rates [ 151 ]. Furthermore, the portability and relatively low infrastructure requirements of blood-based testing increase accessibility in underserved and resource-limited populations, thus addressing existing disparities in cancer detection efforts [ 151 ]. Overall, these characteristics position ctDNA assays as potential tools to reduce individual patient as well as systemic healthcare costs, in line with aims for more sustainable and equitable cancer care. Looking forward, several key developments are expected to enhance the clinical utility of ctDNA in OC management and potentially transform precision oncology paradigms. Future ctDNA assays will likely incorporate multiple layers of molecular information through integration with multi-omics platforms, including genomic mutations, epigenetic modifications, fragmentomics (analysis of cfDNA fragment size and patterns), and protein biomarkers [ 152 ]. Multi-omics integration promises to improve both sensitivity and specificity, particularly for early detection and MRD monitoring, by capturing complementary dimensions of tumor biology. Artificial intelligence and machine learning represent another frontier, with advanced computational tools being developed to improve signal detection, distinguish tumor-derived from non-tumor DNA, and predict clinical outcomes based on ctDNA dynamics [ 146 ]. AI-driven models can integrate complex genomic data with clinical variables to provide personalized risk stratification and treatment recommendations, potentially enabling real-time clinical decision support. Standardization and clinical validation remain critical priorities, as robust standardization of pre-analytical variables (sample collection, processing, storage) and analytical protocols (sequencing platforms, bioinformatic pipelines) is essential for widespread clinical adoption. Large-scale prospective trials are needed to validate ctDNA-based strategies for guiding treatment decisions, such as escalation or de-escalation of therapy, and to establish evidence-based clinical utility that satisfies regulatory and payer requirements. Beyond ctDNA, the broader field of liquid biopsy is exploring other circulating biomarkers, including CTCs, extracellular vesicles (exosomes), and tumor-educated platelets (TEPs). Combining ctDNA with these complementary analytes may yield more comprehensive tumor profiling and enhance diagnostic accuracy across multiple cancer types. Finally, continuous or frequent ctDNA monitoring could enable real-time assessment of treatment response, allowing for adaptive therapy adjustments [ 82 ]. This dynamic approach may help identify resistance mechanisms earlier and facilitate timely therapeutic switches, moving toward a vision of truly personalized, data-driven cancer care. In conclusion, while ctDNA holds tremendous promise for revolutionizing OC management, addressing current technical, biological, and clinical challenges—particularly CHIP interference—is essential for realizing its full potential. Continued innovation in detection technologies, coupled with rigorous clinical validation and thoughtful integration into clinical workflows, will determine the extent to which ctDNA can improve patient outcomes in this challenging disease.

Epithelial ovarian cancer (EOC) represents a leading cause of gynecologic cancer-related mortality worldwide. In 2022, an estimated 324,603 new OC cases and 206,956 ovarian cancer-related deaths were reported globally among women, ranking ovarian cancer eighth in cancer incidence among women worldwide and accounting for [ 1 , 2 ] approximately 3.4% of all new cancer cases and approximately 4.8% of cancer deaths. OC is characterized by marked heterogeneity, with epithelial subtypes comprising over 90% of cases being EOC. The absence of effective screening modalities and the insidious nature of disease progression result in the majority of patients being diagnosed at advanced stages (International Federation of Gynecology and Obstetrics stages III/IV) according to the International Federation of Gynecology and Obstetrics (FIGO). Compounded by widespread chemotherapy resistance, the 5-year overall survival (OS) rate remains disappointingly low, ranging from 30% to 50% 2 . Despite decades of research, no universally accepted biomarkers exist for early detection, accurate diagnosis, or treatment response prediction beyond traditional tumor markers such as CA125 and imaging assessments [ 3 , 4 ]. Circulating cell-free DNA (cfDNA) comprises DNA fragments released into the bloodstream by both normal and malignant cells through mechanisms including apoptosis, necrosis, and active secretion. ctDNA, a tumor-derived subset of cfDNA, typically constitutes approximately 0.01% to 5% of the total cfDNA in cancer patients [ 5 ]. With a short half-life of approximately one hour, ctDNA offers a dynamic, real-time molecular snapshot of tumor biology. These DNA fragments carry tumor-specific genetic and epigenetic alterations linked to cancer progression, enabling non-invasive detection and monitoring of malignancy [ 6 , 7 ]. Recent large-scale translational studies have substantially advanced the clinical application of ctDNA across diverse oncologic contexts. The GRAIL consortium has developed the Galleri ® multi-cancer early detection (MCED) test, which employs targeted cfDNA methylation profiling to identify cancer-associated epigenetic signatures. Real-world validation in over 100,000 participants demonstrated specificity exceeding 99% for multi-cancer detection, with accurate tissue-of-origin localization in the majority of positive cases [ 8 , 9 ]. In the third and final substudy of the Circulating Cell-free Genome Atlas (CCGA; NCT02889978 ) study—a pre-specified, large-scale clinical validation in 4,077 participants (cancer: n  = 2,823; non-cancer: n  = 1,254 with confirmed status at one-year follow-up)—specificity for cancer signal detection was 99.5% (95% CI: 99.0%–99.8%), with accurate tissue-of-origin (CSO) prediction in 88.7% of true positive cases [ 11 ]. Galleri achieved a stage I–III sensitivity of 67.6% (95% CI: 64.4%–70.6%) in the 12 pre-specified cancer types that account for approximately two-thirds of annual US cancer deaths, a group that includes ovarian cancer [ 10 , 11 ]. Nevertheless, sensitivity in stage I disease remained modest 16.8% (95% CI: 14.5%–19.5%), reflecting the inherent biological challenge of detecting ctDNA at extremely low tumor burdens [ 9 , 12 ]. Recent prospective data from the PATHFINDER study ( N  = 107,654 asymptomatic participants aged over 50 years) provided the first large-scale real-world validation of the Galleri assay in an intended-use screening population [ 13 ]. The overall test positivity rate was 1.4%, with a specificity of 99.1% (95% CI: 99.0%–99.2%) and a cancer PPV of 38.0% among participants with complete diagnostic follow-up. Cancer signal origin (CSO) prediction accuracy was 79.2% (95% CI: 72.1%–85.1%), somewhat lower than the 88.7% reported in the CCGA case-control validation substudy, likely reflecting the greater biological heterogeneity of an unselected screening population. Importantly, PATHFINDER demonstrated feasibility of integrating MCED testing into routine clinical workflows, with a median time to diagnostic resolution of 79 days among test-positive participants. These data confirm the high specificity of methylation-based MCED in real-world screening settings while reinforcing that early-stage sensitivity limitations, particularly relevant for ovarian cancer, where stage I–II disease represents the optimal intervention window, remain a central challenge for population-level implementation [ 13 ]. In parallel, tumor-informed ctDNA assays have emerged as highly sensitive tools for disease monitoring in defined patient cohorts. SAGA diagnostics has pioneered personalized ctDNA detection platforms utilizing ultra-sensitive digital polymerase chain reaction (dPCR) technology tailored to patient-specific tumor mutations. In OC, preoperative ctDNA levels measured using the SAGA safe platform demonstrated significant prognostic value, with elevated ctDNA strongly associated with poor OS [ 14 ]. Furthermore, prospective feasibility studies have established that postoperative ctDNA detection can objectively identify MRD following primary debulking surgery, correlating with surgical completeness and subsequent recurrence risk [ 15 ]. These findings underscore the potential of tumor-informed approaches to refine risk stratification and guide personalized treatment decisions. Together, these representative studies illustrate two complementary paradigms in ctDNA application: tumor-agnostic, methylation-based multi-cancer screening exemplified by platforms such as Galleri (GRAIL), and tumor-informed, mutation-tracking strategies optimized for longitudinal monitoring in well-characterized patient populations as demonstrated by SAGA diagnostics. Both approaches highlight the transformative promise of ctDNA while underscoring current challenges, including limited sensitivity in low-burden disease states, the need for standardized analytical workflows, and ongoing efforts to define clinical utility endpoints that justify integration into routine OC management. To ensure comprehensive coverage of the evolving landscape of ctDNA research in OC, a systematic literature search was conducted across multiple academic databases, including PubMed, Google Scholar, and Web of Science. The search strategy employed combinations of key terms including “circulating tumor DNA,” “ovarian cancer,” “early diagnosis,” “minimal residual disease (MRD) monitoring,” and “recurrence prediction,” spanning publications from foundational studies to the most recent advances. All studies investigating the clinical or translational application of ctDNA in OC were considered for inclusion. Inclusion criteria encompassed: ① Studies involving OC patients with a primary focus on ctDNA detection, characterization, or clinical utility. ② Research designs including experimental investigations, prospective or retrospective clinical trials, and systematic reviews, and meta-analyses. ③ Original research articles or comprehensive reviews published in peer-reviewed journals with full-text availability. ① Studies involving OC patients with a primary focus on ctDNA detection, characterization, or clinical utility. ② Research designs including experimental investigations, prospective or retrospective clinical trials, and systematic reviews, and meta-analyses. ③ Original research articles or comprehensive reviews published in peer-reviewed journals with full-text availability. Exclusion criteria comprised: ① Studies unrelated to OC. ② Studies focusing exclusively on non-ctDNA biomarkers such as circulating tumor cells (CTCs) or exosomes without ctDNA analysis. ③ Articles lacking sufficient methodological detail or incomplete reporting of key experimental or clinical outcomes, such as those missing critical results. ① Studies unrelated to OC. ② Studies focusing exclusively on non-ctDNA biomarkers such as circulating tumor cells (CTCs) or exosomes without ctDNA analysis. ③ Articles lacking sufficient methodological detail or incomplete reporting of key experimental or clinical outcomes, such as those missing critical results. Literature Screening Process: ① Literature screening was conducted in two sequential stages: in the first stage, two independent reviewers screened all retrieved records by title and abstract to exclude clearly irrelevant studies, in the second stage, the same two reviewers independently assessed the full texts of potentially eligible articles against predefined inclusion and exclusion criteria. ② Any discrepancies between reviewers were resolved through structured discussion; where consensus could not be reached, a third reviewer served as arbitrator. ③ Duplicate records identified across databases were removed prior to screening using reference management software (EndNote X9), with residual duplicates identified and eliminated through manual cross-checking. ④ Given that this review was not prospectively registered in a systematic review registry (e.g., PROSPERO), and given the broad scope of the search strategy, some risk of selection bias cannot be entirely excluded. This represents an acknowledged methodological limitation. Future reviews in this area are encouraged to adopt formal PRISMA-compliant protocols and prospective registration to further minimize selection bias [ 13 , 16 ]. ① Literature screening was conducted in two sequential stages: in the first stage, two independent reviewers screened all retrieved records by title and abstract to exclude clearly irrelevant studies, in the second stage, the same two reviewers independently assessed the full texts of potentially eligible articles against predefined inclusion and exclusion criteria. ② Any discrepancies between reviewers were resolved through structured discussion; where consensus could not be reached, a third reviewer served as arbitrator. ③ Duplicate records identified across databases were removed prior to screening using reference management software (EndNote X9), with residual duplicates identified and eliminated through manual cross-checking. ④ Given that this review was not prospectively registered in a systematic review registry (e.g., PROSPERO), and given the broad scope of the search strategy, some risk of selection bias cannot be entirely excluded. This represents an acknowledged methodological limitation. Future reviews in this area are encouraged to adopt formal PRISMA-compliant protocols and prospective registration to further minimize selection bias [ 13 , 16 ]. Due to the rapid evolution of ctDNA technologies and the inherent limitations of database indexing, earlier foundational studies or emerging preprints may not be fully captured, representing an acknowledged limitation in comprehensiveness. Nonetheless, this search strategy ensures robust representation of the current state of knowledge and identifies critical gaps warranting further investigation.

OC is a highly heterogeneous malignancy, with pathological subtypes profoundly influencing ctDNA characteristics and clinical applications. HGSC accounts for 70–80% of EOC, with most cases being diagnosed at FIGO stage III–IV [ 109 , 110 ]. The OS rate for HGSC patients (34.1%) is significantly lower than that of non-HGSC patients (57.0%) [ 111 ], underscoring its aggressive nature. Consequently, early pathological classification is critical for guiding treatment decisions. Beyond their prognostic significance discussed above, TP53 and BRCA1/2 alterations, along with a constellation of subtype-defining mutations, exhibit distinct prevalence patterns across OC histologic subtypes that have direct implications for ctDNA assay design and interpretation. Accumulating evidence demonstrates that various pathological types of OC exhibit significant differences in ctDNA release, mutation profiles, and associations with clinical characteristics [ 112 ]. Substantial advancements have been made in utilizing ctDNA across various OC subtypes, particularly in HGSC, clear cell carcinoma (CCC), and low-grade serous OC (LGSC), highlighting its considerable clinical potential. In HGSC, the most common and aggressive OC histotype, TP53 mutations are present in > 95% of cases, representing a highly reliable tumor-specific marker for ctDNA-based disease tracking [ 43 ]. BRCA1/2 mutations, present in approximately 20–25% of HGSC patients (germline and somatic combined), are detectable in ctDNA with a sensitivity exceeding 85% [ 113 , 114 ]. Given the high prevalence and near-universal presence of TP53 alterations in HGSC, this class of mutations is particularly suitable for tumor-informed design of ctDNA panels, while BRCA1/2 status additionally inform subtype-specific therapeutic vulumility, as discussed in Sect. 2.3 . Additionally, ctDNA applications in CCC suggest that for chemotherapy-resistant patients, ctDNA can guide the selection of appropriate targeted therapies prior to treatment initiation and enable dynamic monitoring of therapeutic efficacy during treatment. In LGSC, KRAS and BRAF mutations are present in approximately 70% of patients, providing a high-frequency ctDNA tracking target that is molecularly distinct from HGSC [ 91 ]. The subtype-specific mutation architecture across HGSC, CCC, and LGSC thus necessitates tailored ctDNA assay configurations, a consideration with direct implications for panel design, analytical sensitivity thresholds, and clinical interpretation frameworks. Collectively, the subtype-specific mutation landscapes of OC, TP53 and BRCA1/2 in HGSC, ARID1A and PIK3CA in CCC, and KRAS/BRAF in LGSC, define distinct molecular frameworks for ctDNA assay optimization, enabling histotype-matched surveillance strategies that extend the clinical utility of liquid biopsy beyond what uniform panel designs can achieve [ 113 , 114 ]. OC presents distinct characteristics across clinical stages. In early-stage OC, ctDNA detection may be limited by tumor burden and patient-specific factors, resulting in a lower detection rate. In contrast, ctDNA detection rates increase significantly in advanced stages (III and IV). For instance, Bettegowda et al. assessed ctDNA detection in 640 patients across multiple cancer types and stages, revealing that median ctDNA concentration in stage IV patients was 100 times higher than in stage I patients. Furthermore, ctDNA analysis is particularly valuable for guiding molecular classification in advanced disease. The European Society for Medical Oncology recommends routine use of comprehensive next-generation sequencing for tumor analysis in advanced NSCLC, prostate cancer, OC, and cholangiocarcinoma [ 115 ]. Research has demonstrated that uterine cavity lavage for ctDNA detection shows promise for preliminary OC identification, particularly in HGSC patients. Studies show that uterine lavage techniques can identify tumor-specific TP53 mutations, representing DNA fragments released into bodily fluids by tumor cells. NGS analysis of lavage fluid and biopsy samples revealed TP53 mutations in up to 27% of HGSC cases; furthermore, BRCA1 and BRCA2 mutations are relatively common among this patient group. Uterine lavage combined with ctDNA detection could be an effective tool for early HGSC diagnosis [ 116 ]. Pathological differences are also reflected in ctDNA methylation patterns. CCC and HGSC may differ in ctDNA mutation burden and the frequency of specific gene mutations. OCCC often shows elevated expression of genes associated with oxidative stress response and glycolysis, potentially leading to an accumulation of related mutations in ctDNA [ 112 ]. The relationship between tumor proliferative capacity and ctDNA shedding has been most systematically characterized in colorectal cancer and lung cancer. Although these findings derive from non-OC contexts, the underlying biological principle applies directly to understanding inter-subtype differences in ctDNA release in OC. Evidence from large-scale multi-cancer studies has established that tumor proliferation capacity is key determinants of ctDNA release in colorectal cancer. Studies indicate that secretory and CMS3 subtypes of colorectal cancer demonstrate reduced ctDNA release, while tumors with microsatellite instability show elevated ctDNA levels. A comparative analysis of transcriptome profiles across cancer types shows that colorectal cancer and lung squamous cell carcinoma exhibit high-proliferation ctDNA-shedding phenotypes, while lung adenocarcinoma includes several low-proliferation subgroups. Furthermore, proliferation rates are associated with ctDNA detection sensitivity across diverse cancer types. Different pathological types of OC can exhibit substantial variation in ctDNA release and mutation profiles [ 117 ]. These subtype-dependent shedding differences have direct implications for assay design in OC: detection thresholds calibrated for HGSC may be insufficiently sensitive for LGSC or mucinous OC, underscoring the need for subtype-stratified analytical frameworks. For instance, in well-differentiated OC, ctDNA mutation frequency may be lower, whereas in HGSC, both ctDNA mutation frequency and tumor burden may be substantially higher [ 62 ]. By analyzing ctDNA, researchers can identify mutations and methylation patterns associated with specific pathological types, facilitating the development of more precise treatment strategies [ 118 , 119 ]. The additional clinical relevance of detecting subtype-defining mutations by ctDNA can extend beyond subtype characterization. In HGSC, longitudinal tracking by ctDNA of BRCA1/2 somatic mutations can aid PARP inhibitor maintenance therapy selection [ 120 ] and enable early identification of BRCA reversion mutations which predict the development of acquired tissue resistance [ 102 , 103 ]. In LGSC, detection by ctDNA of KRAS and BRAF variant allele frequency offers a quantitative biomarker of MEK inhibitor treatment response, with elevated allele frequency predicted to precede radiological progression [ 121 , 122 ]. In CCC, detection by ctDNA of ARID1A mutations may aid histological classification when tissue is limited, [ 123 ] and has recently been proposed as a potential molecular surrogate for potential sensitivity to immune checkpoint inhibitors; however, the prospective evidence supporting this application in OC remains limited [ 124 , 125 ]. These subtype-specific clinical applications not only convert ctDNA from a generic tumor marker into a subtype-tailored specialized decision-support tool, but also demonstrate the ability of ctDNA in OC to impact treatment with multiple downstream clinical outcomes. However, prospective validation of each of these ctDNA-based decision-making applications in dedicated OC subtype cohorts is still required before routine clinical implementation can be recommended. Analyzing ctDNA methylation patterns allows researchers to identify specific methylation regions associated with HGSC. Regarding ESR1 methylation, a study by Liggett et al. examining plasma cfDNA from OC patients reported detection of methylation alterations in both primary tumor tissue and corresponding plasma cfDNA samples [ 98 ]. It should be noted that the sample size in this study was relatively small ( N  = 28 patients), and ESR1 was among a panel of genes assessed rather than the primary endpoint; this finding was derived from a discovery-phase cohort and requires independent prospective validation before clinical application [ 126 ]. Thus, cfDNA methylation characteristics may be utilized for early cancer diagnosis and provide essential insights for treatment evaluation and prognosis [ 127 ]. A study by Dobrzycka et al. in a cohort of 86 patients with ovarian tumours ( n  = 43 ovarian tumours of malignant histology, and n  = 43 benign histology) revealed a significant association of serum cfDNA with hTERT methylation in patients with large ovarian tumours ( p  = 0.001), as well as with the corresponding tumour tissue samples ( p  = 0.012) [ 100 ]. However, the study was lacking an independent validation cohort, and moreover that the inclusion of both malignant and benign cases may have influenced the reported associations. Hence, the results should be considered with caution until prospective replication. The methylation status of the hTERT gene may play a crucial role in tumorigenesis and development, similar to findings in breast cancer patients, where hTERT methylation is prevalent in both tumors and surrounding tissues and is detectable in ctDNA [ 128 ]. In addition to assessing the methylation status of the RASSF1A gene, promoter hypermethylation of RASSF1A has been evaluated as an epigenetic biomarker in the setting of OC [ 129 ]. Ibañez de Cáceres et al. found tumor cell-specific RASSF1A hypermethylation in serum, plasma, and peritoneal fluid collected from OC patients ( N  = 30), in the first evidence for the utility of circulating epigenetic biomarkers in OC [ 129 ]. It is important to note that the statistical associations with FIGO stage and OS outlined elsewhere in the literature require further independent confirmation in larger prospective cohorts, as the current evidence base for RASSF1A methylation in OC has been limited to small retrospective single-institution studies [ 130 ]. The methylation status of hTERT may also be associated with tumor and biomarkers that emulate those of the RASSF1A and CDKN2AIP genes, where promoter methylation has been evaluated as a potential diagnostic and prognostic marker across multiple cancer types, including OC [ 131 ]. Collectively, the ctDNA characteristics observed across different pathological types of OC provide valuable insights into tumor biology and support advancements in precision medicine. However, a critical limitation of the current methylation literature is the absence of standardized profiling platforms, differences in bisulfite conversion protocols, sequencing depth, and bioinformatic pipelines across studies substantially impair cross-study comparability. Prospective multicenter validation using harmonized analytical workflows will be essential before methylation-based ctDNA markers can be considered for routine clinical implementation in OC [ 99 ].