ctDNA in the Management of Resectable & Advanced Colorectal Cancer: Current Status and Future Directions Authors: Humaid O Al-Shamsi, Saeed Rafii, Syed Hammad Hassan Tirmazy, Deborah Mukherji, Ahmad Y Abuhelwa, Faryal Iqbal, Siddig Ibrahim Abdelwahab, Hampig Raphael Kourie, Maen Abdelrahim Published: 2026 ABSTRACT AbstractCirculating tumor DNA (ctDNA) is an increasingly important biomarker for molecular residual disease (MRD) after curative-intent treatment of colorectal cancer (CRC), but its prognostic value must be distinguished from proven clinical utility. This structured narrative review synthesizes evidence from PubMed/MEDLINE, Embase, ClinicalTrials.gov, and major oncology conference proceedings, covering stage I–III CRC and selected patients with resected or ablated oligometastatic stage IV disease. We review tumor-informed and tumor-agnostic assay strategies, postoperative sampling, serial versus landmark testing, low-shedding disease, clonal hematopoiesis, and major prospective and randomized studies. Postoperative ctDNA positivity consistently identifies patients at substantially increased risk of recurrence, while serial ctDNA dynamics provide additional prognostic information. In stage II colon cancer, the randomized DYNAMIC trial showed that a ctDNA-guided strategy reduced adjuvant chemotherapy use without compromising long-term recurrence outcomes. By contrast, DYNAMIC-III did not establish non-inferiority for ctDNA-guided de-escalation and showed no recurrence-free survival benefit from conventional chemotherapy escalation in ctDNA-positive stage III disease. The Phase III ALTAIR trial likewise did not meet its pre specified primary disease-free survival endpoint for treatment of molecular relapse with trifluridine/tipiracil. Current professional guidance therefore supports a cautious distinction between clinical validity and clinical utility: ctDNA can inform prognosis, counseling, and clinical-trial eligibility, but should not routinely replace standard surveillance or independently determine treatment escalation or de-escalation outside evidence-based indications or clinical trials. Future progress requires harmonized assays, randomized demonstration of outcome benefit, and prospective validation of integrated ctDNA, pathology, transcriptomic, and AI-derived risk models.Keywords: colorectal cancer, ctDNA, MRD, liquid biopsy, MSI-H/dMMR, adjuvant Introduction Colorectal cancer (CRC) remains a major and expanding global health challenge. The most recent comprehensive worldwide estimates come from GLOBOCAN 2022, reporting 1.93 million new CRC cases and 904,000 CRC deaths in 2022, placing CRC among the most commonly diagnosed cancers and a leading cause of cancer mortality globally,1,2 Rising by 1.1% per year since 2005, colorectal cancer is now the leading cause of cancer death among US adults under 50 as of 2023.3 Looking ahead, demographic growth and population ageing are expected to substantially increase the burden as IARC’s Cancer Tomorrow projections anticipate a marked rise in CRC incidence by 2040, with global projections around 3.2 million new cases annually by that time.4 These trends are unfolding alongside ongoing epidemiologic transition, including obesity and lifestyle risk factors, and widening inequities in access to early detection and effective treatment, factors that collectively shape both CRC incidence and outcomes worldwide.2 Paradoxically, improvements in screening and diagnostic pathways are increasing detection of earlier-stage CRC, yet the clinical problem of post-surgical recurrence risk persists.5 Curative-intent surgery is foundational for localized CRC and selected oligometastatic disease, but a substantial proportion of patients who appear cured by conventional assessments ultimately relapse due to clinically occult micrometastatic disease; for example, in patients with oligometastases treated with curative-intent radiotherapy, new metastases develop in about 73% of cases despite initial local control.6 This creates a persistent gap between what clinicians can measure and what patients need to know. Standard post-operative risk stratification still relies largely on TNM staging and clinicopathologic features.7 While these tools are essential for population-level prognostication and guideline-based decision-making, they are imperfect surrogates for residual tumor biology in the individual patient, especially when recurrence risk is driven by microscopic dissemination that is invisible to routine imaging and not reliably reflected by conventional markers.8,9 This limitation has tangible consequences in clinical care. In the adjuvant setting, clinicians must balance two competing harms; overtreatment, where patients already cured by surgery receive adjuvant chemotherapy (ACT) with limited or no benefit but meaningful toxicity, and undertreatment, where patients at high risk of relapse are not identified early enough to justify escalation or intensified surveillance.10 This dilemma is particularly prominent in resected stage II disease, where ACT benefit is modest and heterogeneous, and remains relevant in stage III disease, where ACT is standard but the magnitude of benefit varies across patients and risk groups.11,12 The overarching clinical objective is therefore not simply to add more therapy, but personalized therapy, reducing recurrence while minimizing unnecessary exposure to systemic treatment. Surveillance strategies after curative therapy are also constrained. Carcinoembryonic antigen (CEA) is widely used, yet multiple evaluations emphasize that CEA has insufficient sensitivity and specificity to serve as a standalone biomarker for early detection of recurrence, contributing to false-negative reassurance and delayed identification of molecular relapse.13,14 Imaging likewise detects recurrence only after tumor burden reaches radiographic thresholds, leaving a window during which residual disease may be biologically active but clinically silent.15 This explains why recurrence can occur despite apparently successful surgery and guideline-concordant follow-up. In this context, circulating tumor DNA (ctDNA) has emerged as a non-invasive biomarker for molecular residual disease (MRD) detection and personalized management in CRC.16 ctDNA directly assess tumor-derived DNA fragments in plasma, enabling detection of residual malignant clones after surgery and dynamic monitoring during systemic therapy.17 Clinically, postoperative ctDNA positivity is consistently associated with substantially higher recurrence risk, supporting ctDNA as a biologically grounded tool that complements, and may outperform, traditional risk factors in identifying patients who remain at risk after resection.18,19 Importantly, ctDNA is transitioning from prognostication to treatment decision-making. In the randomized DYNAMIC trial, a ctDNA-guided strategy for stage II colon cancer reduced ACT use without compromising recurrence-free survival, establishing proof-of-concept for de-escalation when ctDNA is undetectable.20 Current professional guidance highlights an important distinction between the clinical validity of ctDNA as a prognostic biomarker and its clinical utility as a treatment-directive biomarker. Postoperative ctDNA is consistently associated with recurrence risk, but therapeutic modification or replacement of established surveillance based on ctDNA alone remains insufficiently validated across most curative-intent CRC settings. The present review provides a 2026 perspective that explicitly separates prognostic, predictive, and treatment-strategy evidence and incorporates mature 5-year DYNAMIC data, the randomized DYNAMIC-III trial, the phase III ALTAIR trial, contemporary CIRCULATE-platform evidence, emerging assay-comparison data, and recent professional guidance. This evidence hierarchy is particularly important because an association between ctDNA status and outcome does not, by itself, demonstrate that changing treatment on the basis of ctDNA improves patient outcomes.1,2,4–8,11 Throughout this review, we distinguish prognostic from predictive evidence. A prognostic biomarker identifies groups with different expected outcomes irrespective of treatment, whereas a predictive biomarker identifies differential benefit from a specific intervention. Demonstrating that postoperative ctDNA positivity is associated with a high recurrence hazard therefore establishes clinical validity as a prognostic MRD marker, but does not establish that escalation, de-escalation, or earlier treatment triggered by ctDNA improves clinical outcomes. Evidence of such clinical utility requires prospective interventional validation, preferably in randomized trials.11 Objectives of This Review This review examines the biological rationale, evolving technologies, and evaluate the clinical evidence supporting ctDNA-guided management across the neoadjuvant and post-surgical/adjuvant settings, highlighting practice-relevant advances. We additionally address MSI-H/dMMR-specific considerations, where distinct tumor biology and immunotherapy strategies may require trial designs and ctDNA applications that differ from microsatellite-stable CRC. Methods Evidence was identified through a structured literature search to inform this narrative synthesis, but study selection and reporting were not conducted under a formal PRISMA framework. This structured narrative review was informed by a literature search of PubMed/MEDLINE, Embase, and major oncology conference proceedings. The search is complemented by trial registry screening in ClinicalTrials.gov to capture ongoing, completed, and updated protocols relevant to ctDNA and MRD in colorectal cancer. Because ctDNA evidence in this field frequently appears first as late-breaking or interim data, conference proceedings were also searched through the ASCO Meeting Library (including ASCO Annual Meeting and ASCO GI) and ESMO-related abstract publications, including Annals of Oncology Congress abstract supplements, to identify emerging datasets and trial updates not yet available as full-text manuscripts. Reference lists of pivotal trials and contemporary reviews were manually searched by 2 independent authors, and forward citation screening was used for landmark ctDNA-guided studies to capture follow-up publications and subgroup analyses. Search terms combined controlled vocabulary and free-text keywords, adapted to each database, and included “circulating tumor DNA” and “ctDNA”, “liquid biopsy”, “minimal residual disease” and “molecular residual disease”, “molecular recurrence/relapse”, together with disease terms (“colorectal cancer”, “colon cancer”, “rectal cancer”) and setting terms (“neoadjuvant”, “preoperative”, “adjuvant”, “postoperative”, “surveillance”, “metastasectomy”, “oligometastatic”). Conference searches used analogous term combinations within ASCO and ESMO search interfaces and abstract books. Eligibility criteria were pre specified to prioritize clinically applicable evidence for curative-intent disease. Peer-reviewed studies published from 2015 through 25th of Jan 2026 were eligible, trials before 2015 were excluded as they were mainly initial and preclinical ctDNA trials. Eligible populations included adults with resectable CRC treated with curative-intent surgery in stages I–III, and patients undergoing curative-intent resection or ablation of metastatic disease in a post-metastasectomy setting (resected stage IV/oligometastatic CRC), since ctDNA-guided MRD assessment is increasingly studied as a recurrence-risk and treatment-tailoring tool after metastasectomy. Studies were required to evaluate ctDNA in plasma as an index biomarker for MRD detection, risk stratification, treatment guidance, response monitoring, or surveillance, using either tumor-informed or tumor-agnostic approaches. Exclusion criteria included studies focused exclusively on population screening or early detection in average-risk asymptomatic cohorts, unresectable metastatic CRC without a curative-intent resection/metastasectomy context, review and non-original reports without primary data, and very small case series where estimates are highly unstable. Data extraction was performed at the study level and focused on trial design and intended clinical role of ctDNA, patient and tumor characteristics, treatment context, ctDNA assay approach, and reported outcomes including recurrence-related endpoints such as Disease Free Survival (DFS)/Recurrence Free Survival (RFS) and patterns of ctDNA clearance or conversion. When multiple reports referred to the same cohort or trial, the most mature peer-reviewed dataset was prioritized, while earlier interim reports were used only to clarify longitudinal consistency or evolution of endpoints. Evidence was interpreted according to study design and maturity, with randomized interventional trials given greater weight than prospective observational studies, retrospective cohorts, registries, conference reports, modeling studies, and ongoing trial protocols. Findings were further categorized as prognostic, predictive or treatment-directive, or hypothesis-generating, as appropriate. No original statistical pooling was performed; effect estimates, confidence intervals, and P values are reported from the source studies. Biological Rationale for ctDNA in Colorectal Cancer ctDNA represents short fragments of tumor-derived DNA shed into the bloodstream, primarily through tumor cell apoptosis and necrosis, with additional contribution from active release in some contexts.21 Because ctDNA has a relatively short circulating half-life reported on the order of minutes to a few hours, it functions as a dynamic, near–real-time readout of tumor burden and treatment impact, in contrast to static pathologic variables assessed at resection.22,23 This biology is particularly compelling in resectable CRC, where the clinical question after curative-intent surgery is not whether tumor existed, but whether microscopic residual disease persists below the resolution of imaging and routine biomarkers.24,25 ctDNA provides an analytically direct means to detect MRD by identifying tumor-specific genomic or epigenomic signals in plasma when radiographic disease is absent.26 Multiple prospective CRC studies have consistently shown that postoperative ctDNA positivity identifies a subgroup with markedly higher recurrence risk, supporting ctDNA as a reliable MRD surrogate and a biologically grounded predictor of relapse.24 Early longitudinal work demonstrated that serial ctDNA tracking after resection can anticipate recurrence and reflect changes in tumor burden across follow-up, aligning ctDNA kinetics with the natural history of microscopic progression.25 Subsequent prospective studies using targeted sequencing approaches in localized colon cancer reinforced that postoperative ctDNA detection is independently associated with inferior disease-free outcomes, consistent with the interpretation that detectable ctDNA after surgery reflects persistent micrometastatic clones capable of driving clinical relapse.26,27 Large contemporary datasets and meta-level syntheses further support the prognostic validity of ctDNA across resected CRC populations, linking ctDNA positivity not only to recurrence risk but also to broader survival endpoints, strengthening the argument that ctDNA captures clinically consequential residual disease biology rather than being a purely correlative marker.28,29 Conceptually, this creates a new framework for perioperative management as ctDNA converts an otherwise binary post-resection state into a measurable continuum of residual disease burden, enabling earlier risk-adapted intervention and potentially more efficient allocation of adjuvant therapy and surveillance intensity. Compared with established biomarkers, ctDNA offers important advantages but should be positioned with appropriate nuance. CEA remains widely used, yet its diagnostic performance for recurrence is variable across studies and is influenced by baseline CEA secretion biology, recurrence volume, and timing; false negatives are common, particularly in early molecular relapse.14 Several CRC studies and reviews suggest ctDNA can provide a clinically meaningful lead time over conventional surveillance, sometimes detecting molecular relapse months before radiographic confirmation, although real-world performance is heterogeneous and depends on assay type and sampling schedule.24,30 Importantly, not all datasets show ctDNA superiority for every surveillance endpoint; for example, a cohort analysis comparing ctDNA with imaging and CEA reported higher sensitivity when imaging and CEA were combined than with ctDNA alone at the time of clinically confirmed recurrence, underscoring that ctDNA is best viewed as complementary and context-dependent rather than universally replacing established modalities.31 Circulating tumor cells (CTCs) are also prognostically informative in CRC, but their rarity in early-stage disease and technical complexity of enrichment/characterization have constrained broad MRD implementation relative to plasma ctDNA workflows.32 Technical Advances in ctDNA Detection ctDNA detection has evolved from early single-variant PCR assays toward highly sensitive, error-suppressed next-generation sequencing (NGS) strategies capable of MRD assessment at extremely low variant allele fractions.33 Contemporary MRD testing is commonly divided into tumor-informed approaches, which personalize a patient-specific assay from tissue-identified variants, and tumor-agnostic (plasma-only/tissue-free) approaches, which infer ctDNA using fixed genomic panels and/or epigenomic signatures without requiring tumor tissue.34 In CRC, tumor-informed assays such as Signatera exemplify individualized multiplex designs that prioritize a small set of patient-specific variants and incorporate matched normal sequencing to reduce germline and clonal hematopoiesis interference, whereas tissue-free assays such as Guardant Reveal integrate genomic alterations with epigenomic/methylation features to improve sensitivity when tissue is unavailable and to accelerate turnaround time.33,35 Assay performance should also be interpreted in the context of sampling strategy and tumor biology. Landmark postoperative testing may miss low-level residual disease, particularly in low-shedding tumors, whereas serial testing can improve sensitivity by capturing molecular recurrence over time. Tumor-agnostic approaches may also be affected by clonal hematopoiesis unless appropriate filtering is incorporated. Importantly, discordant results between assays should not be interpreted as definitive evidence for treatment escalation or de-escalation, as differences may reflect assay design, sampling timing, biological shedding, or analytical sensitivity rather than true absence or presence of residual disease.36,37 Key technical gains enabling MRD applications include unique molecular identifiers and other error-correction strategies, ultra-deep sequencing, and improved bioinformatics that collectively push analytic sensitivity toward the range required for postoperative MRD, with MRD-focused discussions often referencing 0.01% variant allele frequency as a practical/theoretical benchmark for early-stage detection in plasma.38–40 Parallel advances in methylation and broader epigenomic profiling have expanded mutation-agnostic detection, as demonstrated by validation of tissue-free epigenomic MRD assays in resected CRC cohorts showing clinically meaningful stratification of recurrence risk and lead time for molecular relapse.35 Standardization remains an active frontier, with increasing emphasis on harmonizing pre-analytic variables, sampling timepoints, and reporting conventions, while regulatory momentum is reflected by FDA Breakthrough Device Designations granted to selected MRD assays intended to accelerate clinical evaluation, designations that recognize potential impact but do not, by themselves, establish clinical utility.33,41 Ultra-sensitive whole-genome approaches are now pushing MRD detection into the parts-per-million range: a 2024 open-access cfDNA WGS method using single-read error correction (AccuScan) reported an error rate of 4.2×10−7 and analytical sensitivity down to 10−6 variant allele frequency (99% sample-level specificity), achieving 90% landmark sensitivity and 100% specificity for relapse prediction in post-surgical colorectal cancer.42 Assay-related limitations will be discussed later in this review. Clinical Evidence from Key Trials The clinical evidence supporting ctDNA-guided management in resectable colorectal cancer (CRC) has matured rapidly from prognostic observational studies to randomized trials testing de-escalation and escalation strategies. In this review, prognostic trials refers to the association between ctDNA status, typically measured in an early postoperative MRD window, and subsequent recurrence or survival, regardless of treatment, whereas predictive trials design addresses whether ctDNA identifies patients who derive differential benefit from a specific therapeutic strategy or whether ctDNA-guided decision algorithms can safely improve outcomes and/or reduce overtreatment. Across multiple trials, postoperative ctDNA has shown to be a predictor for survival and with significant hazard ratios for recurrence-free or disease-free survival underscoring its biological relevance as a marker of residual micrometastatic disease.28,43,44 Importantly, the strength of the postoperative ctDNA signal is consistent across different analytical platforms and appears to refine recurrence risk beyond conventional clinicopathologic staging. In a stage III colon cancer cohort using a tumor-informed assay, postsurgical ctDNA was the strongest independent correlate of recurrence, and persistent detectability after completion of adjuvant therapy identified patients at particularly high risk, supporting ctDNA as a dynamic on-treatment biomarker.26 In another larger phase III dataset analyzed with a tissue-free epigenomic assay, ctDNA positivity (20%) remained independently associated with shorter time-to-recurrence, disease-free survival, and overall survival (hazard ratios 4–6), and quantitative ctDNA burden further stratified outcomes among ctDNA-positive patients.45 Complementing trial cohorts, prospective real-world implementation with routine result return demonstrated marked early DFS separation and a high negative predictive value with few false-negative cases, informing ongoing efforts to embed ctDNA into practical surveillance and ctDNA-guided treatment algorithms.46 Postoperative MRD as a High-Amplitude Risk Classifier The CIRCULATE-Japan observational registry (GALAXY) provides the most influential prognostic study because of its scale, standardized sampling, and integration into downstream interventional trials. In a recent update (n=2,240 across stage II–III resectable colon cancer and resected stage IV settings), ctDNA positivity during the prespecified MRD window was associated with profoundly inferior outcomes (DFS HR ~11.99; OS HR ~9.68), and ctDNA dynamics further stratified risk beyond a single binary measurement.28 Importantly, this analysis also reported that “ctDNA clearance” is not a theoretical concept, rather sustained clearance after adjuvant chemotherapy (ACT) was associated with markedly better outcomes than transient clearance, suggesting that longitudinal ctDNA trajectories may serve as an intermediate phenotype for treatment efficacy and relapse biology. These findings align with earlier prospective work in stage III CRC demonstrating the strong prognostic value of serial ctDNA monitoring and its potential to quantify recurrence kinetics.47 Randomized trials reinforce the same prognostic message. In the DYNAMIC-III randomized Phase 2/3 trial in stage III colon cancer (n=968), ctDNA status measured 5–6 weeks after surgery clearly separated recurrence risk, with ctDNA-negative patients experiencing substantially fewer recurrences than ctDNA-positive patients (3-year RFS 87% vs 49%).43 The magnitude and consistency of this prognostic effect across diverse cohorts and various other progostic trials explains why ctDNA is now widely viewed as the most direct clinically available biomarker of MRD in CRC and why trial programs increasingly use ctDNA to enrich for high-risk populations and to further operationalize treatment adaptation in newer predictive trial designs.48,49 De-Escalation and Escalation Strategies The most practice-influential signal to date comes from stage II colon cancer. In the randomized DYNAMIC trial (n=455), a ctDNA-guided approach reduced ACT use without compromising outcomes at initial readout, and the mature 5-year analysis confirmed durable safety.20 At a median follow-up of ~60 months, 5-year recurrence-free survival was 88% with ctDNA-guided management versus 87% with standard management, with similar 5-year overall survival (93.8% vs 93.3%). The updated analysis also highlighted two clinically meaningful refinements for the ctDNA-positive subgroup; higher postoperative ctDNA burden predicted worse recurrence risk, and end-of-treatment ctDNA persistence identified an ultra–high-risk phenotype. These results support ctDNA-guided de-escalation in stage II disease as a mature example of how MRD assessment can reduce overtreatment while maintaining oncologic safety.50 Yet, The DYNAMIC trial had several limitations. The relatively small sample size limited robust subgroup analyses. In addition, treatment decisions were guided by ctDNA status rather than randomization within subgroups, which may have introduced selection bias. Furthermore, the study evaluated ctDNA-guided management only at the initial adjuvant decision point, without assessing the long-term role of serial ctDNA monitoring. Larger prospective studies are needed to validate these findings. In Contrast, in DYNAMIC-III, for patients with ctDNA-negative stage III colon cancer following surgery, recurrence risk was low, with a 3-year recurrence-free survival rate of 87%.41 ctDNA-informed treatment de-escalation was feasible and achieved high adherence, leading to a marked reduction in oxaliplatin exposure and an improved safety profile, including fewer treatment-related hospitalizations and grade ≥3 adverse events. Although non-inferiority compared with standard of care was not formally confirmed, clinical outcomes were comparable, particularly in low-risk (T1–3N1) disease. These findings support the potential role of ctDNA-guided strategies in facilitating individualized risk–benefit discussions and highlight the need for further prospective studies to validate de-escalation approaches in this setting.51 Among patients with ctDNA-positive stage III colon cancer following surgery, recurrence risk remained high despite adjuvant chemotherapy and increased with rising ctDNA burden. In the escalation arm, chemotherapy intensification including intensification from oxaliplatin-based doublets to FOLFOXIRI did not improve outcomes compared with standard management (2-year recurrence-free survival: 51% vs 61%), indicating limited benefit from escalation with conventional cytotoxic regimens. In contrast, post-treatment ctDNA clearance was strongly associated with favorable prognosis, whereas persistent ctDNA identified a subgroup with extremely poor outcomes (3-year recurrence-free survival: 14% vs 79%). These findings emphasize that simple intensification of standard chemotherapy may be insufficient once molecular residual disease is detectable and highlight the need for novel therapeutic strategies and biomarker-driven endpoints in future trials.52,53 Large platform programs are designed to move beyond prognostic certainty toward true predictive utility. CIRCULATE-US/NRG-GI008 (NCT05174169) is a phase II/III MRD-directed platform that randomizes ctDNA-negative patients to immediate ACT versus serial ctDNA surveillance with delayed ACT and randomizes ctDNA-positive patients to standard ACT versus intensified therapy (including FOLFIRINOX), with DFS as a primary endpoint in Phase III.54 This design directly tests whether ctDNA can safely support de-escalation (non-inferiority trial design framework) and whether ctDNA-positive patients benefit from escalation (superiority framework), addressing questions that observational datasets cannot resolve.55,56 Complementary evidence from CIRCULATE-Japan links prognostic MRD detection to treatment responsiveness at the population level. The CIRCULATE-Japan molecular MRD program includes GALAXY as the observational backbone and interventional trials such as VEGA (de-escalation among MRD-negative high-risk stage II/low-risk stage III) and ALTAIR (post-adjuvant or surveillance treatment in molecular relapse).57,58 As of January 2026, VEGA’s mature non-inferiority outcomes have not been published in a full peer-reviewed report; thus, VEGA is best discussed as a landmark design that operationalizes ctDNA-negative de-escalation within a national registry infrastructure.59 For treatment of molecular relapse, the phase III ALTAIR trial evaluated trifluridine/tipiracil (FTD/TPI) versus placebo in ctDNA-positive patients without radiological evidence of disease. In the prespecified primary analysis, median DFS was 9.30 months with FTD/TPI versus 5.55 months with placebo (HR 0.79, 95% CI 0.60–1.05; P=0.107); therefore, the primary DFS endpoint was not met. These findings further illustrate that identification of molecular recurrence by ctDNA does not necessarily establish that earlier systemic treatment with an available therapy will improve clinical outcomes.60,61 In ASCO GI 2026 In the GALAXY study at ALTAIR-participating sites reported that postoperative ctDNA dynamics are critical for identifying patients eligible for molecularly guided trials were approximately 15% of patients with stage II–III disease and 30% with stage IV disease met ctDNA-based eligibility criteria, and 77% of patients with persistent ctDNA positivity after adjuvant therapy were eligible for enrollment, although only 6.5% were ultimately recruited.61 Higher minimal residual disease positivity and trial eligibility rates were observed with advancing disease stage, and elevated ctDNA levels were associated with early clinical relapse. These findings underscore the importance of serial ctDNA monitoring during adjuvant treatment and surveillance to optimize precision trial enrollment. Notably, not all interventional experiences have been positive. The NRG-GI005 COBRA trial in low-risk stage IIA colon cancer did not meet its prespecified Phase II endpoint and halted further enrolment based on stopping rules, illustrating how assay choice, population risk, endpoint selection, and the biologic meaning of low-level ctDNA signals can materially influence trial interpretability.62 When considered together, the results of DYNAMIC (positive), DYNAMIC-III (mixed), and COBRA (negative) define the current evidence landscape, in which ctDNA is consistently validated as a prognostic biomarker, yet its translation into effective, risk-adapted therapeutic strategies remains highly dependent on clinical context, analytical methodology, and treatment approach. Table 1 summarizes randomized and controlled ctDNA-guided trials in resectable colorectal cancer, outlining study populations, assay strategies, interventions or comparisons, primary endpoints, and key signals from mature or available outcome data that collectively illustrate the clinical utility and limitations of prognostic and predictive ctDNA approaches in adjuvant management. Table 1Randomized and Controlled Trials of ctDNA-Guided Adjuvant Management in CRCStudyPopulation (N)Assay ParadigmIntervention/ComparisonPrimary Endpoint(s)Key Signal (Maturity)DYNAMIC.20Stage II colon cancer, total N=455 (ctDNA-guided n=302; standard n=153)Tumor-informed ctDNA-guided strategyctDNA-guided ACT vs standard clinicopathologic managementRFS (non-inferiority); OS (updated)5-year RFS 88% vs 87%; 5-year OS 93.8% vs 93.3%; reduced ACT with maintained outcomes; ctDNA burden and end-of-treatment persistence further stratified ctDNA-positive riskDYNAMIC-III.43Stage III colon cancer, evaluable N=968; ctDNA-negative n=702Tumor-informed ctDNA-guided strategyctDNA-guided de-escalation (ctDNA−)/escalation (ctDNA+) vs standard managementctDNA−: 3-year RFS (non-inferiority); ctDNA+: 2-year RFSPrognostic separation: 3-year RFS 87% (ctDNA−) vs 49% (ctDNA+). De-escalation reduced oxaliplatin (34.8% vs 88.6%) but did not meet non-inferiority (85.3% vs 88.1%). Escalation did not improve 2-year RFS (51% vs 61%). Persistent ctDNA after treatment predicted very poor outcomes (3-year RFS 14% vs 79%)COBRA/NRG-GI005.62Low-risk stage IIA colon cancer (phase II/III design)Tissue-free (plasma-only)ctDNA-directed chemotherapy vs observationPhase II endpoint focused on ctDNA clearance/early efficacy signalPhase II endpoint not met; enrollment halted per prespecified stopping rules (conference-level evidence; full publication awaited)TRACC Part C.63High-risk stage II and stage III CRC; target N≈1621Tissue-free (plasma-only; opened with Guardant Reveal)ctDNA-guided ACT vs standard-of-care ACT3-year DFS (non-inferiority)Pragmatic non-inferiority trial testing de-escalation among MRD-negative patients; ongoingCIRCULATE-US/NRG-GI008.54Resected colon cancer; ctDNA-stratified cohorts (phase II/III platform)Tumor-informed (Signatera)ctDNA−: immediate ACT vs serial ctDNA surveillance with delayed therapy; ctDNA+: standard ACT vs intensified therapy (eg, FOLFIRINOX)Cohort A: time to ctDNA+ (phase II) and DFS (phase III non-inferiority); Cohort B: DFS (superiority)Definitive platform testing both de-escalation safety and escalation benefit using ctDNA-defined MRD; ongoingVEGA (CIRCULATE-Japan).57High-risk stage II/low-risk stage III colon cancer; ctDNA− at week 4 (within GALAXY)Tumor-informedObservation vs 3-month CAPOXDFS (non-inferiority)Major de-escalation trial embedded in GALAXY; mature outcomes not yet available in peer-reviewed full text (as of Jan 2026)ALTAIR (CIRCULATE-Japan).57,61Resected stage II–IV CRC with molecular relapse/MRD+ during surveillanceTumor-informedTrifluridine/tipiracil vs placeboDFSFirst large placebo-controlled molecular-relapse interventionPhase III randomized trial of molecular relapse; the prespecified primary DFS endpoint was not metAbbreviations: ACT, adjuvant chemotherapy; CAPOX, capecitabine plus oxaliplatin; CT, computed tomography; CTC, circulating tumor cell; DFS, disease-free survival; MRD, minimal residual disease; OS, overall survival; RFS, recurrence-free survival; tdDNA, tumor-directed circulating tumor DNA assay. Randomized and Controlled Trials of ctDNA-Guided Adjuvant Management in CRC Abbreviations: ACT, adjuvant chemotherapy; CAPOX, capecitabine plus oxaliplatin; CT, computed tomography; CTC, circulating tumor cell; DFS, disease-free survival; MRD, minimal residual disease; OS, overall survival; RFS, recurrence-free survival; tdDNA, tumor-directed circulating tumor DNA assay. Registries and Real-World Programs: Clinical Validity, Decision Impact, and Assay-Dependent Performance Beyond randomized trials, prospective registries provide critical insights about implementation and generalizability. GALAXY is the cornerstone, demonstrating reproducible prognostic separation at scale and informing downstream trial eligibility and timing rules.28 BESPOKE CRC, the first large prospective US-based trial (133 sites; 1,792 patients) evaluating tumor-informed ctDNA after surgery, supports postoperative ctDNA MRD detection as a powerful prognostic and predictive tool in stage II–III CRC; serial monitoring also enabled oligometastasis-directed therapy in 40% of patients who recurred, and ctDNA status during surveillance remained prognostic regardless of adjuvant chemotherapy (median DFS 9 months), with patient-reported perceived utility and well-being measures suggesting broad acceptability of ctDNA testing.64 A key translational bridge between “prognostic” and “predictive” evidence is ctDNA clearance after ACT.65 Across large cohorts and analytic syntheses, ctDNA positivity after completion of adjuvant therapy is consistently associated with very high recurrence risk, whereas clearance is associated with improved outcomes, supporting ctDNA dynamics as a biologically evident intermediate marker of treatment effectiveness.18,66,67 Although these associations are compelling, they do not automatically imply that escalating therapy based on ctDNA dynamics improves outcomes; this is precisely the clinical-utility question that ongoing adaptive platform trials aim to answer. Table 2 presents major prospective cohorts and platform programs evaluating postoperative ctDNA in resectable colorectal cancer. Table 2Major Prospective Cohorts/Platforms Evaluating Postoperative ctDNA in Colorectal CancerProgramPopulation (N)Design and TimingCore Findings/ContributionGALAXY (CIRCULATE-Japan).57Expanded cohort n=2,240 (stage II–III resectable colon cancer plus resected stage IV settings in updated analysis)Large prospective observational registry; MRD window sampling with serial drawsPostoperative ctDNA positivity is strongly associated with inferior DFS and OS (DFS HR ~11.99; OS HR ~9.68). ctDNA dynamics (sustained vs transient clearance) further stratify outcomes and underpin interventional trials (eg, VEGA/ALTAIR)DYNAMIC/DYNAMIC-III programs.20,43Stage II (n=455) and stage III (n=968) colon cancerProspective randomized programs with defined post-op sampling timepointsStage II: ctDNA-guided strategy reduces ACT without compromising 5-year outcomes. Stage III: ctDNA is strongly prognostic; de-escalation reduced toxicity but did not meet non-inferiority; escalation with existing chemotherapy did not improve outcomes in ctDNA+ diseaseBESPOKE CRC.64Stage II–III enriched cohort within a broader prospective registry (trial NCT04264702)Prospective observational program delivering ctDNA results to clinicians to assess decision impactEvaluates real-world adoption and decision impact; interim evidence disseminated via journal supplements/meeting proceedings; complements randomized trials by informing implementation barriers and practice patternsAbbreviations: CRC, colorectal cancer; DFS, disease-free survival; GALAXY, large observational CIRCULATE-Japan registry; MRD, minimal residual disease; Prospective obs, prospective observational; RFS, recurrence-free survival; Signatera, tumor-informed ctDNA assay; US, United States ctDNA platform cohort. Major Prospective Cohorts/Platforms Evaluating Postoperative ctDNA in Colorectal Cancer Abbreviations: CRC, colorectal cancer; DFS, disease-free survival; GALAXY, large observational CIRCULATE-Japan registry; MRD, minimal residual disease; Prospective obs, prospective observational; RFS, recurrence-free survival; Signatera, tumor-informed ctDNA assay; US, United States ctDNA platform cohort. Meta-Analyses: Pooled Estimates Supporting Recurrence Risk Stratification and Post-ACT Dynamics Systematic reviews and meta-analyses broadly corroborate the prognostic value of postoperative ctDNA. A recent systematic review and meta-analysis focused on stage II CRC found that postoperative ctDNA positivity was associated with increased recurrence risk (pooled RR 3.66), and it also emphasized that ctDNA positivity after adjuvant chemotherapy is a particularly adverse sign while serial monitoring can identify recurrence earlier than conventional approaches.19 Complementary meta-analytic work across broader stage ranges similarly supports strong associations between postoperative ctDNA and recurrence, while highlighting heterogeneity introduced by assay design (tumor-informed vs tissue-free), sampling timepoints, and definitions of positivity, reinforcing the need to interpret pooled metrics within the context of platform-specific performan.68 Prognostic and Predictive Value of ctDNA A clinically useful way to frame ctDNA is that it provides three layers of information that conventional postoperative assessment cannot easily supply; whether residual disease is molecularly detectable at all, how that signal evolves over time, and how reliably that signal maps onto outcomes in different treatment contexts e,g. monitoring clonal evolution. The first layer, postoperative detectability, has repeatedly behaved like a rule-in marker for residual malignant clones in plasma, but it is not a perfect rule-out test.69 A key limitation that shapes both prognostic interpretation and trial design is imperfect sensitivity in early-stage disease, as observed in a dual-cohort model-development study, where postoperative ctDNA was absent in a substantial fraction of patients who later recurred, motivating hybrid strategies that do not treat ctDNA negativity as synonymous with cure.70 The second layer, longitudinal behaviour, is where ctDNA begins to resemble a quantitative disease activity biomarker rather than a single yes/no test. Real-world and trial-linked datasets indicate that repeated ctDNA positivity during surveillance is tightly aligned with clinical relapse and, in some cohorts, can anticipate recurrence while standard markers remain uninformative.18,71,72 In an analysis explicitly comparing ctDNA with CEA, ctDNA monitoring showed higher discriminatory performance for recurrence than CEA and appeared to behave similarly in both a prospective trial cohort and a commercial real-world cohort, supporting external validity in an underrepresented population.18 Importantly, serial ctDNA is not only an early detection tool; it can be read as a trajectory. In a prospective stage III cohort with scheduled postoperative sampling through adjuvant therapy and follow-up, ctDNA growth patterns were linked to the clinical behaviour of recurrence and survival, suggesting that ctDNA kinetics may capture aggressiveness and not merely the presence of relapse.47 This kinetic framing matters because it shifts ctDNA from being a passive prognostic label to an on-treatment signal that could support adaptive strategies, such as intensifying therapy when molecular burden rises rapidly or shortening surveillance intervals when molecular risk accelerates. The third layer, predictive utility, requires a more careful, context-dependent definition. A prognostic biomarker separates risk groups; a predictive biomarker identifies differential benefit from a specific intervention.73 Some of the most instructive predictive-adjacent signals come from settings where systemic therapy benefits are debated, such as after resection of colorectal liver metastases.74 In a prospective cohort of resectable colorectal liver metastases, postoperative ctDNA positivity strongly associated with inferior recurrence-free and overall survival, whereas ctDNA clearance during neoadjuvant chemotherapy was not clearly linked to improved outcomes, underscoring that ctDNA response is not interchangeable across time windows or treatment phases.74 Even more clinically actionable was the observation that persistent ctDNA after completion of surgery and adjuvant therapy identified a near-certain relapse phenotype in that cohort, suggesting a rational target population for post-adjuvant escalation trials that test new drug classes rather than simply prolonging conventional therapy. Predictive signals may also be therapy-class specific, which is particularly relevant to MSI-H/dMMR disease where immunotherapy introduces new decision problems such as early identification of resistance and determining appropriate treatment duration.75 In a prespecified secondary analysis of the SAMCO-PRODIGE 54 randomized trial in dMMR/MSI-H metastatic CRC, early change in ctDNA, using tumor-specific methylation markers, was associated with long-term outcomes and, critically, the strength of this association was substantially greater in the immune checkpoint inhibitor arm than in the chemotherapy arm.76 This pattern supports a clinically important idea for ctDNA-guided management as ctDNA can function not only as a baseline risk marker but also as an early pharmacodynamic readout that helps distinguish biologic benefit from apparent radiographic stability, and it may be most informative when aligned to mechanisms where conventional response assessment is imperfect. Because postoperative ctDNA is an imperfect rule-out test, integration with clinicopathologic factors is not a fallback but a design principle for real-world decision-making. Hybrid models that combine ctDNA with stage and adverse histopathologic features and baseline biomarkers have demonstrated improved recurrence prediction compared with ctDNA alone, explicitly addressing the false-negative problem and offering a more clinically credible framework for tailoring adjuvant intensity and surveillance schedules.70 In practice, this integrated approach enables risk messaging as ctDNA positivity can justify escalation or trial enrollment with high confidence, whereas ctDNA negativity can support de-escalation only when contextualized by clinicopathologic and molecular risk, sampling timing, and assay strategy.70 Clinical Implications and Current Guidelines In routine practice, ctDNA most directly changes the postoperative conversation by adding a molecular assessment of risk that is not captured by pathologic staging alone, but major guidelines remain cautious about using ctDNA to direct adjuvant decisions outside trials. As of version 5.2025, NCCN has explicitly acknowledged ctDNA as prognostic in colorectal cancer, while emphasizing that evidence is still insufficient for routine, decision-directive use; an NCCN update summarized this position as “ctDNA is a prognostic marker; however, there is currently insufficient evidence to recommend routine use of ctDNA assays outside of a clinical trial,” and that “care de-escalation and treatment decision-making are not recommended based on ctDNA findings,” with participation in clinical trials encouraged.77,78 In parallel, the ESMO Precision Medicine Working Group recommendations note that using ctDNA to detect molecular residual disease is not recommended at present because clinical utility has not been established, reinforcing a trial-first approach for MRD-directed intervention.79 Patients’ Perspective on ctDNA Testing Patient-reported outcomes have been formally reported only from BESPOKE CRC, making it the sole large prospective CRC ctDNA program to directly assess perceived utility and well-being domains alongside serial MRD testing.64 In their survey-based interim dataset (1,130 responses from 413 patients), participants generally viewed tumor-informed ctDNA results as reassuring and actionable; 73% reported reduced anxiety about recurrence, 87% felt more confident they were receiving the right treatment after seeing their results, 92% would continue ctDNA monitoring, and 96% valued the additional information provided by ctDNA testing. Consistent with these findings, an NCI summary of BESPOKE notes that anxiety/depression scores were not meaningfully worse in ctDNA-positive vs ctDNA-negative patients, and that most patients preferred ongoing ctDNA surveillance, supporting ctDNA as not only a biologically informative biomarker but also a generally acceptable component of survivorship care when implemented with clear counseling and pathways.80 Health Economics in ctDNA and Adjuvant CRC Management Health-economic signals for ctDNA-guided adjuvant decision-making remain mixed and highly context-dependent, largely because value is driven by how much chemotherapy use is avoided in ctDNA-negative patients, whether ctDNA-positive patients experience material outcome gains from escalation, and local test pricing, chemotherapy costs, toxicity management costs, and willingness-to-pay thresholds. In a US payer-focused budget impact analysis (commercial and Medicare Advantage) modeling adoption of tumor-informed ctDNA to guide adjuvant therapy in stage II colon cancer, ctDNA testing was projected to be cost-saving at acceptable uptake levels, with savings driven primarily by reduced adjuvant chemotherapy utilization and associated administration and adverse-event costs; the model also estimated a “budget-neutral” test price that differed by payer mix.81 Importantly, such budget impact results are sensitive to baseline practice patterns and may not generalize to settings where ACT use is already low or drug pricing/administration costs differ substantially. By contrast, cost-effectiveness evaluations performed in other health systems often conclude that ctDNA-guided strategies are not automatically cost-effective under current assumptions, because quality-adjusted life years (QALY) gains are modest unless ctDNA meaningfully improves selection for effective therapy. A Dutch model-based economic evaluation comparing ctDNA-guided approaches against guideline-based selection in stage II colon cancer found that combining ctDNA with clinicopathologic high-risk criteria could improve health outcomes but was not cost-effective at base-case inputs; ctDNA-guided selection became cost-effective only under threshold conditions such as lower test cost, improved test performance, or a scenario where ctDNA positivity predicts a larger chemotherapy benefit.82 These findings reinforce that, while de-escalation supported by randomized evidence can produce payer savings in some contexts, the largest incremental value proposition, and thus the most consequential driver of cost-effectiveness, likely depends on whether ctDNA-guided escalation in MRD-positive patients yields sufficiently large absolute reductions in recurrence and downstream metastatic-treatment costs. Finally, implementation economics are inseparable from coverage, access, and workflow costs. Even when modeled results appear favorable, real-world value depends on assay reimbursement, equitable availability, standardized timing, and care pathways that translate results into consistent actions without excessive downstream imaging or clinic utilization. Broader liquid-biopsy policy work highlights that payer coverage and access barriers can limit real-world uptake and concentrate benefits in better-resourced systems, which should be accounted for in future prospective economic endpoints embedded within randomized ctDNA-guided trials.83 Suggested Clinical Integration of ctDNA in CRC Management in 2026 Across disease stages, ctDNA should currently be interpreted primarily as a prognostic marker and as a tool for risk stratification, counseling, and clinical-trial enrollment. Outside settings supported by prospective randomized evidence, ctDNA results should not independently determine treatment escalation or de-escalation, replace guideline-recommended surveillance, or dictate systemic therapy after curative-intent metastasectomy. Neoadjuvant Chemotherapy In colon cancer treated with preoperative chemotherapy, ctDNA is best positioned in 2026 as an investigational pharmacodynamic marker rather than a stand-alone decision tool, because the evidence base is smaller and more heterogeneous than in postoperative MRD. Early-stage colon neoadjuvant ctDNA work supports feasibility and hypothesis-generating associations between ctDNA dynamics and response but does not yet justify routine regimen switching based on ctDNA alone outside trials.79,80 In rectal cancer receiving chemoradiotherapy or total neoadjuvant therapy (TNT), ctDNA is more consistently prognostic, and persistent positivity across treatment windows has been associated with inferior metastasis-free and survival outcomes in prospective cohorts; however, integration should remain adjunctive to established response assessment (MRI/endoscopy) until outcome-driven randomized strategies mature.80–82 Theoretically, Serial ctDNA dynamics across neoadjuvant chemotherapy and the perioperative period can inform surgical planning and tailor adjuvant therapy intensity based on post-surgical ctDNA status, yet this still need to be evaluated in the setting of a clinical trial. Neoadjuvant Immunotherapy For dMMR/MSI-H disease, neoadjuvant checkpoint blockade has produced striking pathologic response rates, and emerging datasets support ctDNA kinetics as a biologic response readout that may help distinguish deep responders from potential non-responders in the preoperative window.84,85 Emerging NICHE-2 kinetics suggest ctDNA may function as a rapid biologic response readout: ctDNA was detectable in 92% at baseline, with 45% demonstrating ctDNA clearance by day 15; preoperatively, ctDNA clearance was reported in 94% of patients achieving pCR (vs 70% with MPR), and all patients were ctDNA-negative at the 3-week post-surgery MRD timepoint.86 In NICHE-3 (nivolumab/relatlimab), pathological response remained high (97% overall; 68% pCR), with ctDNA listed among secondary endpoints, but mature ctDNA–outcome correlations have not yet been fully reported; analogous response-adapted rectal organ-preservation programs are now evaluating ultrasensitive ctDNA monitoring to support risk-stratified surveillance and early molecular relapse detection.87,88 More broadly, neoadjuvant immunotherapy studies have shown that pre-surgical ctDNA clearance can correlate with pathological response, reinforcing its role as a mechanistically aligned monitoring biomarker, although clinical thresholds for changing therapy remain investigational.89 For dMMR locally advanced rectal cancer managed with nonoperative intent after immunotherapy, highly durable complete responses have been reported, but ctDNA-guided stopping rules are not yet guideline-standard; ctDNA can be considered a complementary tool within protocolized surveillance rather than a determinant of omission of surgery off trial.90 ctDNA can serve as a unifying biomarker to support patient selection, response monitoring during neoadjuvant immunotherapy, perioperative decision-making, and post-treatment management, yet this needs to be further assessed in future clinical trials. Stage II Colon Cancer A pragmatic 2026 approach is to obtain ctDNA in an MRD window around 4 weeks after surgery, with a confirmatory draw in the subsequent weeks when feasible, recognizing that timing and assay strategy influence sensitivity.91 This can support shared decision-making about adjuvant chemotherapy in a setting where absolute benefit is often modest. The randomized DYNAMIC provides evidence (despite the limitations) to support that ctDNA-guided management can safely reduce adjuvant chemotherapy use in stage II colon cancer while maintaining long-term recurrence outcomes.20 ctDNA negativity can be used to frame a discussion about treatment omission for selected patients, whereas ctDNA positivity identifies a higher-risk group in whom adjuvant treatment and clinical-trial options should be considered alongside standard clinicopathologic factors, with serial ctDNA used primarily for prognostic counselling and to identify persistent molecular disease after therapy.91,92 Stage III Colon Cancer Standard, risk-adapted oxaliplatin-based adjuvant chemotherapy remains the default in 2026, and ctDNA should not be used to intensify cytotoxic regimens off protocol.93,94 DYNAMIC-III reinforced that postoperative ctDNA is strongly prognostic in stage III disease, but escalation with more intensive chemotherapy did not improve outcomes, supporting a no off-trial escalation principle while trials test novel strategies.43 The most actionable integration is routing ctDNA-positive patients toward biomarker-directed trials designed to test true predictive utility, such as NRG-GI008/CIRCULATE-North America, which explicitly evaluates escalation and de-escalation strategies in ctDNA-stratified cohorts.95 Serial ctDNA may be used to refine prognosis and counselling, but ACT modification should remain trial-based until definitive platform results are available. Stage IV After Curative Resection After resection of oligometastatic disease, ctDNA is potential tool as a prognostic risk marker to identify patients at exceptionally high risk of recurrence and to prioritize clinical trial enrollment, rather than as a trigger for empiric chemotherapy intensification.96 Prospective data in resected colorectal liver metastases demonstrate that postoperative ctDNA-based MRD stratifies outcomes and can be used to study who benefits from adjuvant therapy in a setting where standard benefit remains debated.97 NCCN also note that ctDNA is not recommended for surveillance in stage IV colon/rectal cancer, reinforcing that post-metastasectomy ctDNA use should be selective and ideally protocolized.77,78 Rectal Cancer After Neoadjuvant Therapy Post-neoadjuvant ctDNA is consistently prognostic and may complement MRI, endoscopy, and pathology when planning adjuvant therapy intensity or considering nonoperative management strategies, but it should be integrated as an adjunct rather than a replacement for established modalities.98 Prospective studies in locally advanced rectal cancer suggest ctDNA detection and persistence across neoadjuvant windows are associated with worse outcomes, and ctDNA negativity after therapy is more common in favorable response states, supporting its use for risk-adapted multidisciplinary discussion and trial enrollment rather than unilateral decision-making.99 A practical, 2026-oriented algorithm for integrating ctDNA across neoadjuvant, adjuvant, surveillance, and post-surgical settings is summarized in Table 3. Table 3Evidence-Based Interpretation and Current Boundaries of ctDNA Use Across Curative-Intent Colorectal Cancer SettingsClinical SettingSuggested ctDNA Role in 2026Practical ApproachHow to Act on ResultsKey CaveatsNeoadjuvant chemotherapy (colon)Investigational pharmacodynamic marker (feasibility/response association), not a stand-alone decision toolConsider serial ctDNA across neoadjuvant course and perioperative window (protocolized when possible)Do not switch regimens solely on ctDNA outside trials; use for risk discussion and trial enrollmentEvidence smaller/heterogeneous vs postoperative MRD; low-shedding disease can yield false negatives; interpret alongside imaging/clinical courseNeoadjuvant CRT/TNT (rectal)Adjunct prognostic marker; persistent positivity associated with worse outcomesOptional serial draws across treatment windows + perioperative samplingUse to support MDT risk stratification and trial referral; maintain MRI/endoscopy as primary response assessmentNot yet outcome-driven randomized strategies; do not replace MRI/endoscopy/pathologyNeoadjuvant immunotherapy (dMMR/MSI-H)Mechanistically aligned response biomarker (kinetics/clearance correlate with response)If using, prioritize kinetics/clearance pre-op and early post-op (protocolized)No ctDNA-based stopping/omission rules off-trial; use for monitoring and surveillance support within protocolsThresholds for therapy change remain investigational; nonoperative pathways still require rigorous surveillanceStage II colon cancer (post-op)Most practice-ready use-case: supports de-escalation discussionsDraw ctDNA in an MRD window ~4 weeks post-surgery, with confirmatory repeat in subsequent weeks when feasiblectDNA-negative: support omission discussion when clinico-path risk is low/modest; ctDNA-positive: strengthen rationale for adjuvant therapy and/or trial; consider serial ctDNA mainly for prognosis and persistenceTiming and assay strategy materially affect sensitivity; beware clonal hematopoiesis; avoid over-interpreting a single low-level positiveStage III colon cancer (post-op)Prognostic, not yet predictive for cytotoxic escalationctDNA may be obtained post-op primarily to identify high-risk patientsDo not intensify cytotoxic regimens off-protocol; route ctDNA-positive patients to ctDNA-stratified escalation/de-escalation trials; serial ctDNA for prognosis/counseling“No off-trial escalation” principle; platform trial readouts needed before routine predictive claimsStage IV after curative resection (eg, oligometastatic/metastasectomy)Selective prognostic risk marker to prioritize trial enrollmentConsider post-op ctDNA to stratify recurrence risk in a protocolized pathwayUse primarily for trial triage and prognosis; avoid empiric intensification solely on ctDNAGuidelines do not recommend ctDNA for surveillance in stage IV colon/rectal cancer; evidence is evolvingRectal cancer after neoadjuvant therapyAdjunct prognostic marker complementing MRI/endoscopy/pathologyPost-neoadjuvant ctDNA can be considered, especially when decisions about adjuvant intensity/nonoperative pathways are being discussedUse to support MDT discussion and trial enrollment, not unilateral decision-makingShould not replace established modalities; persistence across windows signals higher risk but does not yet mandate a specific off-trial change Evidence-Based Interpretation and Current Boundaries of ctDNA Use Across Curative-Intent Colorectal Cancer Settings Limitations and Knowledge Gaps Assay-related limitations remain the most immediate constraint on ctDNA-driven care. Diagnostic performance varies widely by assay paradigm and sampling strategy, with landmar” (single timepoint) testing generally less sensitive than serial monitoring, and tumor-informed approaches typically outperforming tumor-agnostic approaches for recurrence detection in resected CRC.36 A recent diagnostic-accuracy meta-analysis reported substantially higher sensitivity with serial tumor-informed testing than with serial tumor-agnostic testing (approximately 0.88 vs 0.59), while false-positive rates were broadly similar, highlighting that a large fraction of missed events is driven by biology and sampling frequency rather than by any single platform claim.36 Reported performance metrics for representative MRD assay paradigms are summarized in Supplementary Table 1. False positives can occur when non-tumor variants enter the cfDNA pool, most importantly through clonal hematopoiesis (CH), a well-established source of spurious plasma genotyping signals; mitigation requires thoughtful assay design and, when feasible, paired leukocyte (WBC) sequencing or matched-normal strategies to filter CH-derived variants.100,101 Tumor heterogeneity adds a second layer of risk as recurrence may be driven by subclones not represented in the tissue region used for assay design, particularly when evolutionary divergence occurs between primary tumor and metastatic relapse, which can weaken tumor-informed tracking if the monitored variants are not retained.102 Clinical limitations are dominated by variability in when and how ctDNA is measured and interpreted. Postoperative sampling is often performed weeks after surgery, yet adjuvant decisions are time-sensitive, and the optimal timing window is still not fully standardized across trials and practice; a systematic review focused on the immediate postoperative setting highlighted the operational challenge of aligning ctDNA biology with real-world decision timepoints.103 Evidence gaps are also stage-specific as stage I and low-risk populations have low event rates and lower ctDNA shedding, making it difficult to define stable performance estimates and to prove utility beyond standard surveillance.16,104 Finally, while ctDNA has strong and consistent associations with recurrence, long-term overall survival (OS) impact of ctDNA-guided strategies remains incompletely established in prospective randomized data, and large-scale analyses have noted that OS-level validation has lagged behind recurrence endpoints.28 Equity and access are standing as implementation-limiting barriers. Even where clinical enthusiasm is high, cost and reimbursement constraints restrict broad adoption, and availability is uneven across health systems and geographies.105 For example, in a US referral-center real-world cohort of tumor-informed ctDNA testing in CRC, 94% of tested patients were White, underscoring the limited representativeness of available validation/implementation evidence and the need for broader inclusion and access pathways across diverse populations and health systems.106 Economic models suggest ctDNA-guided decision-making can be cost-neutral or cost-saving under certain assumptions, for example, by reducing unnecessary chemotherapy use in stage II colon cancer, but these analyses are payer- and price-sensitive and do not resolve access inequities in low-resource settings.81 Finally, external validation in diverse and underrepresented populations is still limited; real-world studies have explicitly highlighted the scarcity of performance data in groups such as Southeast Asian patients, raising the risk that assay performance and downstream decision frameworks may not generalize uniformly.18,107 Future Research Priorities The next phase of ctDNA-guided care in resectable colorectal cancer should shift from proving that MRD is prognostic to proving what to do with MRD, especially when postoperative ctDNA remains detectable despite standard therapy. The most urgent priority is escalation trials that test novel, mechanism-aligned agents in MRD-positive disease rather than simply intensifying cytotoxic chemotherapy. Platform designs such as CIRCULATE–North America (NRG-GI008) are central to this agenda because they operationalize both de-intensification in MRD-negative disease and escalation strategies in MRD-positive cohorts within one trial infrastructure, enabling definitive testing of predictive utility at scale.95 Parallel efforts in other regions, including pragmatic non-inferiority designs like TRACC Part C, are also essential because they test whether ctDNA can safely reduce unnecessary adjuvant exposure in real-world eligible populations and health systems.63 TRACC Part C is a pragmatic, multicenter non-inferiority randomized trial in resected high-risk stage II and stage III CRC comparing a ctDNA-guided adjuvant strategy, primarily de-escalating chemotherapy in postoperative ctDNA-negative patients using a plasma-only assay, versus standard-of-care adjuvant therapy, with 3-year disease-free survival as the primary endpoint; the study is currently recruiting and has not yet reported results.63 Completing these large, randomized programs is the single most important step toward guideline-grade clinical utility and should be treated as a field-wide priority rather than an incremental research goal. A second priority is improving de-escalation precision by reducing false-negative MRD classification. The additional MRD-positive patients detected only by higher-sensitivity approaches are the very subgroup most likely to benefit from treatment intensification or trial enrollment. Whole-genome strategies with error suppression are rapidly improving analytic limits of detection and are being evaluated for MRD monitoring in solid tumors, including colorectal cancer, suggesting a pathway to expand the reliably actionable MRD-positive population without sacrificing specificity.42 Critically, however, ultra-sensitive WGS-based MRD assays require prospective, outcome-linked validation rather than purely analytical benchmarking. The practical research question is not merely analytical sensitivity but whether WGS-defined MRD improves patient-relevant endpoints by better selecting patients for effective escalation and avoiding unsafe omission of therapy. Third, risk stratification should evolve from single-modality biomarker testing into integrated models that combine ctDNA with orthogonal measures of recurrence biology. Digital pathology and AI pathomics have already shown promise in identifying which patients are most likely to benefit from chemotherapy, offering a complementary, tissue-based lens on risk and treatment responsiveness that could be integrated with ctDNA MRD status into composite decision tools.108 In parallel, transcriptomic recurrence signatures and other tumor-intrinsic expression programs provide a biologic context for why some MRD-positive patients progress quickly and why some MRD-negative patients still relapse, and recent work continues to refine predictive gene signatures that could support treatment selection frameworks extending beyond chemotherapy alone.17,109 A key 2026 priority is prospective, clinically testable integration linking ctDNA plus AI-derived pathomics plus transcriptomics to prespecified decision rules and endpoints. The most impactful multi-omics agenda for 2026 is therefore not additive complexity, but clinically testable integration; ctDNA (dynamic MRD) plus pathomics (spatial microenvironment and architecture) plus transcriptomics (intrinsic programs), linked to prospective endpoints. Finally, the field needs consensus-building that is practical enough to standardize workflows across trials and clinics. Harmonization priorities include prespecified MRD timepoints, definitions of ctDNA positivity across assay classes, and reporting conventions for ctDNA kinetics to ensure that results from platform trials can be translated into implementable care pathways. A particularly high-yield translational niche is the post-adjuvant and molecular relapse setting, where ctDNA can be used as a rapid screening tool to identify patients for early-intervention studies before radiographic recurrence, enabling faster assessment of novel agents and potentially shortening the timeline from biologic signal to patient-relevant outcomes. Future risk models may further integrate ctDNA with orthogonal measures of tumor biology, including AI-derived and transcriptomic signatures. Emerging studies in gastrointestinal cancers have demonstrated the potential of AI-based prognostic indices and transcriptomic profiling to capture biologic heterogeneity and improve risk stratification. However, integration of these approaches with ctDNA remains investigational and requires prospective validation before it can inform routine treatment or surveillance decisions.110,111 Conclusion Postoperative ctDNA is among the strongest available prognostic markers of molecular residual disease after curative-intent treatment of colorectal cancer, and serial testing can provide additional information about molecular disease dynamics. Randomized evidence establishes proof-of-concept for a ctDNA-guided treatment strategy in stage II colon cancer, but actionability is not uniform across settings: DYNAMIC-III did not establish non-inferiority for stage III de-escalation or benefit from conventional chemotherapy escalation, and ALTAIR did not meet its primary DFS endpoint for treatment of molecular recurrence. Accordingly, ctDNA should currently be used primarily to refine prognosis, support informed counseling, and facilitate clinical-trial enrollment; it should not routinely replace standard surveillance or independently dictate escalation, de-escalation, or post-metastasectomy therapy outside validated evidence-based settings. The next phase of the field requires harmonized assays, appropriately timed serial sampling, randomized demonstration of clinical utility, and prospective validation of integrated clinicopathologic and multi-omic risk models. Ultimately, the goal is to translate molecular residual-disease information into more precise care that reduces unnecessary treatment without compromising oncologic outcomes. Acknowledgments During the preparation of this manuscript/study, the author(s) used Grok 4.1, for editing and improving the English language only; it was not used for generating text or any part of the manuscript. The authors have reviewed and edited the output and take full responsibility for the content of this publication. Moreover, we acknowledge the support of Muhammad Farooq Latif, Amin Abyad, Nadia Abdelwahed, Dina Mohamed Hamza, Mawada Hussein, Ashok Sebastian komaranchath, Mudhasir Ahmad, Dima Abdul Jabbar, Hala Abdellatif, Mehdi Afrit, Tariq Alhamid, Hassan Jaafar for their expert opinion and final review of the manuscript. Funding Statement The APC was funded by Burjeel Cancer Institute, Burjeel Medical City, Abu Dhabi, United Arab Emirates. Author Contributions All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure Dr Deborah Mukherji reports Honoraria from Novartis, MSD, Pfizer, Amgen, Astellas, Bayer, Janssen, Roche, Merck Serono and BMS, outside the submitted work. The authors declare no other conflicts of interest. References