MRD in Colorectal Cancer: Is ctDNA Ready for Routine Clinical Use?

For physicians treating colorectal cancer, the most honest answer is that circulating tumor DNA (ctDNA) is ready for selective routine clinical use, but not for universal routine use across every postoperative decision. That distinction matters. 

In stage II colon cancer, randomized data now support ctDNA-guided adjuvant decision-making that reduces chemotherapy use without compromising long-term outcomes. In stage III colon cancer, ctDNA is a powerful prognostic classifier, but the most important randomized trial to date did not show that ctDNA-guided de-escalation could meet noninferiority, and exploratory escalation did not improve recurrence-free survival. 

So, if the question is whether ctDNA is ready to be a standard decision layer in every resected colorectal cancer case, the answer is still no. If the question is whether ctDNA is already useful enough to influence selected physician discussions in clinic, especially in stage II colon cancer, the answer is increasingly yes. 

That is why the current debate around molecular residual disease in colorectal cancer is not really about whether ctDNA is biologically meaningful. That part is largely settled. The debate is about clinical utility. A test can be highly prognostic and still fall short of being ready for broad routine use if the assay is not sensitive enough for de-escalation, turnaround is too slow for adjuvant decisions, a positive result does not have an agreed action plan, or randomized evidence is still incomplete in the setting where the test is being used. 

That is the lens physicians should apply to ctDNA colorectal cancer testing today. 

What MRD means in colorectal cancer

In colorectal cancer, minimal residual disease (MRD) refers to clinically occult residual disease that remains after curative-intent therapy but is below the threshold of conventional imaging. ctDNA is the main blood-based method being used to try to detect that residual disease. In metastatic practice, ctDNA is often used as a liquid biopsy tool for molecular profiling. 

In the postoperative setting, the question is different. The clinician is not primarily asking what targetable mutation is present, but whether tumor-derived DNA is still detectable after surgery or after completion of adjuvant therapy, which would imply persistent disease biology and a higher likelihood of recurrence. The NCI Colon and Rectal-Anal Task Forces white paper identified MRD detection as one of the core colorectal settings most likely to change care. 

That difference between general liquid biopsy and MRD testing is easy to blur, but it is critical in practice. A plasma assay that is sufficient for detecting RAS or BRAF alterations in metastatic disease is not automatically sufficient for detecting microscopic residual disease after resection (when tumor burden is far lower and the consequences of a false-negative result are much greater). This is why the conversation about routine use in colorectal cancer has focused so heavily on assay sensitivity, sampling strategy, and whether ctDNA should be interpreted as a binary yes-or-no signal or as one component of a broader clinical model. 

Why physicians are asking this question now

This question is pressing now because ctDNA colorectal cancer testing has moved from early observational studies into randomized trials, commercial availability, and real-world adoption. NCI noted in 2024 that commercial ctDNA recurrence-monitoring assays were already being used in some clinics while also acknowledging the central practical problem: Clinicians still do not always know what to do when ctDNA is positive but imaging is negative. That tension is the clearest sign that the field has matured. The biomarker is no longer hypothetical. The unresolved issue is how broadly it should drive management. 

The literature reflects the same transition. A 2024 JNCCN review argued that, because large, randomized data with adequate follow-up were still lacking for high-risk stage II and stage III colon cancer, routine ctDNA-guided therapy outside a clinical trial could not yet be recommended in those populations. Then, in 2025, the authors’ reply to a letter on that review emphasized that convincing data now support the prognostic value of ctDNA and its predictive relevance for adjuvant chemotherapy benefit in stage II colon cancer while still cautioning against overinterpreting nonrandomized data, such as GALAXY. That exchange captures the real status of the field better than either extreme position does. 

Where ctDNA is closest to routine clinical use

Stage II colon cancer

If physicians want the most defensible example of ctDNA being ready for routine clinical use in colorectal cancer, stage II colon cancer is it. The strongest evidence comes from the randomized DYNAMIC trial. In mature five-year follow-up, recurrence-free survival was 88% in the ctDNA-guided group and 87% with standard management, and overall survival was 93.8% versus 93.3%, respectively. At the same time, the ctDNA-guided strategy reduced the proportion of patients receiving adjuvant chemotherapy, which had already dropped from 28% with standard management to 15% in the original analysis. The investigators concluded that the mature data support a ctDNA-guided approach to adjuvant chemotherapy in stage II colon cancer. 

That matters because stage II colon cancer has always been the most uncomfortable postoperative gray zone. Many patients are cured with surgery alone. Some are overtreated because clinicians and patients alike are uneasy about “doing nothing.” Others recur despite apparently favorable pathology. ctDNA changes this discussion because it provides a biologic estimate of residual disease rather than relying only on stage and conventional high-risk features. In practical terms, a negative postoperative ctDNA result now gives oncologists stronger footing when recommending observation, and a positive result makes the argument for adjuvant therapy substantially more compelling. 

That does not mean ctDNA should replace the rest of the case. Stage, T4 status, mismatch repair biology, nodal yield, and the patient’s tolerance for toxicity and recurrence risk still matter. What DYNAMIC changed is the level of confidence behind the recommendation. In stage II colon cancer, ctDNA is no longer just a research signal. It is a clinically meaningful risk stratifier that can alter what physicians recommend and how they explain that recommendation to patients. 

Postoperative positivity and clearance

Part of what makes stage II colon cancer so important is that ctDNA may already be doing more than sorting patients into positive and negative groups. In the five-year DYNAMIC report, ctDNA clearance at the end of adjuvant therapy was observed in 87.5% of treated ctDNA-positive patients. That does not yet justify a universal mid-treatment algorithm, but it does suggest that dynamic ctDNA behavior may become increasingly relevant as the field moves from static postoperative risk assessment to response-adapted management. 

The broader observational literature points in the same direction. In GALAXY, adjuvant chemotherapy was associated with substantially higher cumulative ctDNA clearance among patients who were ctDNA-positive after surgery, and patients who failed to clear ctDNA had much worse disease-free survival than those who did clear it. This is one reason many clinicians now see ctDNA as more than a prognostic biomarker. It is beginning to function as a real-time biologic readout of whether adjuvant therapy is controlling the disease that pathology and imaging cannot yet see. 

Where routine use is still premature

Stage III colon cancer

Stage III colon cancer is where physicians most need to separate prognostic value from readiness for routine decision-making. Postsurgical and postchemotherapy ctDNA are unquestionably informative. In the 2019 JAMA Oncology cohort study, three-year recurrence-free interval was 47% vs. 76% for detectable versus undetectable ctDNA after surgery and 30% vs. 77%, respectively, after completion of adjuvant chemotherapy. The same study concluded that serial postsurgical and postchemotherapy ctDNA analyses may identify patients at high risk of recurrence despite standard treatment. 

GALAXY reinforced this at scale. In that prospective observational cohort, postsurgical ctDNA positivity four weeks after surgery was associated with a hazard ratio of 10.0 for recurrence overall and 10.82 in stage II or III disease. ctDNA-positive patients with stage II or III colorectal cancer appeared more likely to derive benefit from adjuvant chemotherapy. Those are very strong data for prognosis and biologic enrichment, but they are still observational. They help explain why the biomarker is compelling, but they do not settle how physicians should alter standard therapy. 

That is why DYNAMIC-III matters so much. In this randomized Phase 2/3 trial, ctDNA-negative patients in the ctDNA-guided arm received de-escalated therapy, and ctDNA-positive patients received escalated therapy. With a median follow-up of 47 months, ctDNA-negative patients had markedly better outcomes than ctDNA-positive patients overall, confirming the biomarker’s prognostic strength. 

But de-escalation did not meet the noninferiority margin compared with standard management, despite reducing oxaliplatin use and hospitalizations. And in ctDNA-positive patients, escalation did not improve outcomes over standard management. The investigators concluded that ctDNA is validated as a strong prognostic classifier, but those results are not the same as proving it is ready to replace current adjuvant standards in routine stage III care. 

So, in stage III colon cancer, ctDNA is highly useful, but its most defensible routine role today is in sharpening prognosis, refining counseling, and identifying patients who should be prioritized for clinical trials or closer follow-up. It is not yet a universal stand-alone switch for de-escalating oxaliplatin in ctDNA-negative patients or intensifying chemotherapy in ctDNA-positive patients outside carefully considered settings. This is exactly the kind of clinical nuance that makes the answer to “is ctDNA ready?” dependent on which colorectal scenario the physician is facing. 

Rectal cancer

Rectal cancer is even less ready for routine MRD-based decision-making. The NCI white paper explicitly states that conclusive data are lacking and that the role of ctDNA status after surgery in tailoring adjuvant therapy for rectal cancer still needs to be assessed. It also emphasizes that nonmetastatic rectal cancer typically has lower tumor burden, often further reduced by neoadjuvant therapy, making high assay sensitivity and rapid turnaround especially important. 

The 2025 review in The Oncologist reached a similar conclusion, stating that the utility of ctDNA in locally advanced rectal cancer remains under investigation and that ctDNA should currently be used as an adjunct to standard testing strategies. The same review highlighted why routine use is difficult. In one cited real-world analysis, imaging and CEA were more sensitive than ctDNA for identifying recurrence, and in post-total-neoadjuvant-therapy settings, ctDNA showed poor sensitivity and negative predictive value for predicting residual disease. That is not the profile of a test ready to independently decide who can safely skip surgery or who no longer needs standard surveillance. 

For physicians, this means colon and rectal cancer should not be discussed as though they were interchangeable MRD settings. The strongest data supporting routine clinical use remain in colon cancer, especially stage II colon cancer. In rectal cancer, ctDNA remains promising, but the current evidence supports adjunctive use rather than routine stand-alone decision-making. 

Surveillance after curative-intent treatment

Surveillance is another area where ctDNA is both promising and not yet fully routine. The signal is certainly strong. In a prospective multicenter study of stage II and III colorectal cancer, serial ctDNA analyses identified recurrence with 92.0% overall accuracy and detected recurrence ahead of radiologic imaging by a mean lead time of 5.01 months. In BESPOKE CRC, NCI reported that recurrent cancer was being detected by ctDNA six to nine months before scans in some patients. From a biologic standpoint, that is impressive. 

But lead time is not the same thing as clinical utility. A surveillance biomarker is only truly ready for routine use if a positive result leads to an action that improves patient outcomes or meaningfully changes management. NCI highlighted the unresolved problem directly: When ctDNA is positive but imaging shows no recurrence, clinicians still face uncertainty about whether to treat, intensify imaging, or simply watch more closely. That uncertainty is one reason surveillance ctDNA is not yet a universal routine standard even though it is clearly biologically informative. 

Assay and workflow issues that still limit routine use

Tumor-informed and tumor-agnostic assays are not interchangeable

Another reason ctDNA is not uniformly “ready” is that the assays themselves differ in important ways. The current MRD landscape in colorectal cancer is dominated by two broad strategies: tumor-informed assays, which sequence the patient’s tumor and then track patient-specific variants in plasma, and tumor-agnostic assays, which look for predefined genomic or epigenomic signals in plasma without needing prior tumor sequencing. These approaches have different tradeoffs in sensitivity, turnaround time, and operational complexity. 

That is not just a technical detail. In a 2025 diagnostic accuracy meta-analysis, tumor-informed assays had markedly higher sensitivity than tumor-agnostic assays in serial monitoring after colorectal cancer resection (0.88 vs. 0.59, respectively) without a significant difference in false-positive rates. For a physician deciding whether to lean on a negative result when discussing de-escalation, that difference matters. 

The article’s core clinical message is simple: ctDNA readiness depends not only on the concept of MRD, but also on which assay is being used and how it performs in the exact workflow being proposed. 

Sensitivity is still the central problem

The hardest technical problem in postoperative MRD testing is sensitivity, not specificity. The NCI white paper states that assays used in de-escalation trials should ideally have sensitivity of at least 95% to reduce the risk of false-negative results, and it notes that serial sampling and/or larger plasma volumes currently provide the best level of sensitivity for MRD detection. That is a very high bar, but it is appropriate. In the adjuvant setting, a falsely reassuring result can mean undertreating a potentially curable patient. 

The field’s newer assay-development work shows why this issue remains active. A 2025 Nature Cancer study noted that existing assays help identify high-risk patients after surgery, but their limited sensitivity immediately after surgery still creates challenges for adjuvant decisions. The same study reported that a plasma whole-exome, sequencing-based, tumor-agnostic assay showed higher MRD sensitivity than current assays. That is encouraging, but it also underscores the point that current commercial-generation assays are not the endpoint. If better sensitivity is still being actively engineered, then routine use must remain conditional on exactly which performance problem the test is being asked to solve. 

Biology creates a ceiling that technology cannot fully erase

Even perfect sequencing chemistry cannot overcome all biologic constraints. The NCI white paper notes that blood-based ctDNA assays likely face a ceiling in some settings because certain tumors are low shedders, and some metastatic sites (including peritoneal or nervous system disease) may be less reliably detected in plasma. That has an important practical consequence. A negative ctDNA result can be reassuring, but it should not be interpreted as a synonym for absence of disease in every patient and every disease pattern. 

This is especially important when physicians try to translate ctDNA results across disease settings. A negative postoperative result in a standard-risk stage II colon cancer patient is a very different clinical object from a negative result in a rectal cancer patient after total neoadjuvant therapy or in a patient whose risk of recurrence may center on disease sites that shed little ctDNA. In other words, a ctDNA result always carries the biology of the disease setting inside it. That is one reason that routine use has to be setting-specific. 

False positives, CHIP, and turnaround time still matter

False positives are less common than false negatives in these discussions, but they are clinically important. The NCI white paper identifies clonal hematopoiesis as a major source of false-positive results and notes that alterations, such as TP53 and KRAS, can arise from clonal hematopoiesis of indeterminate potential (CHIP), rather than residual colorectal cancer. It also outlines mitigation strategies, including tumor-informed assay design, barcoding, and matching ctDNA sequencing with leukocyte and/or tumor sequencing. Those steps are not academic niceties. They directly affect whether a positive result should trigger anxiety, imaging, or treatment discussion.  

Turnaround time is equally practical. A postoperative MRD assay is only useful if the result comes back soon enough to influence adjuvant chemotherapy decisions. The NCI white paper explicitly states that rapid turnaround is crucial in the curative-intent setting. This means routine clinical use is not just about analytical validity. It is also about whether the workflow is fast enough for real oncology practice. A technically excellent assay that arrives too late for treatment planning is not routine-ready in the way physicians actually need. 

So, is ctDNA ready for routine clinical use?

If “routine clinical use” means a test that should be used across resected stage II colon cancer, stage III colon cancer, rectal cancer, and surveillance with similar confidence, then no, ctDNA is not ready for that kind of broad standardization. The randomized evidence is not strong enough across all settings, assay sensitivity still limits de-escalation confidence, and action pathways remain unsettled when ctDNA is positive but imaging is negative. 

If “routine clinical use” means selective, disciplined use by physicians who understand the evidence and the limitations, then the answer is more favorable. In stage II colon cancer, ctDNA is now clinically useful enough to influence adjuvant recommendations in a routine way. In stage III disease, it is ready to inform prognosis and trial referral, even if not yet to replace standard adjuvant algorithms. In rectal cancer and surveillance, it is best viewed as an adjunctive tool rather than a routine stand-alone decision maker. That is probably the most accurate physician-level answer in 2026. 

How physicians can use ctDNA now

The most practical current use of ctDNA colorectal cancer testing is in postoperative stage II colon cancer discussions where the adjuvant decision is genuinely uncertain. That is where randomized data best support using ctDNA to reduce unnecessary chemotherapy without compromising long-term outcomes. 

In stage III colon cancer, physicians can use ctDNA to sharpen estimates of recurrence risk and to identify patients who may benefit from clinical trial enrollment or closer follow-up, but they should be cautious about using ctDNA alone to justify routine de-escalation or escalation of chemotherapy. 

In rectal cancer, ctDNA should currently be interpreted alongside magnetic resonance imaging, endoscopy, pathology, and standard surveillance tools, rather than as a replacement for them. A negative ctDNA result after total neoadjuvant therapy should not be treated as sufficient evidence that surgery or standard follow-up can safely be omitted.