Circulating tumor DNA (ctDNA) has become one of the most closely watched biomarkers in gastrointestinal oncology. Its ability to detect molecular residual disease, estimate recurrence risk, and follow changes in tumor biology has created considerable interest across colorectal, gastro-esophageal, pancreatic, and biliary tract cancers.
However, an important distinction remains. A test can reliably identify patients at high risk without necessarily showing which treatment will improve their outcomes.
In a Viewpoint published in eClinicalMedicine, Sarbajit Mukherjee, Azza Sarfraz, Kyaw Lin Aung, and Manmeet Singh Ahluwalia examine the evolving evidence for circulating tumor DNA, or ctDNA, across gastrointestinal cancers. Their analysis presents a field moving beyond prognostic validation while still confronting major questions about treatment selection, assay interpretation, and clinical implementation.
From Molecular Signal to Clinical Decision
ctDNA consists of short fragments of tumor-derived genetic material released into the bloodstream. Because these fragments have a short half-life, they can provide a dynamic picture of tumor burden and genomic evolution during treatment.
The likelihood of detecting ctDNA is influenced by several factors, including disease stage, metastatic location, tumor biology, timing of blood collection, and the assay being used.
Tumor-informed assays are designed around alterations identified in an individual patient’s tumor and generally provide greater analytical sensitivity for detecting minimal residual disease. Tumor-agnostic assays do not require personalized tumor sequencing and can capture a broader range of emerging alterations, but their sensitivity can differ.
Prospective observational studies have consistently shown that postoperative ctDNA positivity is associated with a markedly increased risk of recurrence. Serial changes, including persistent positivity or clearance, can further refine this risk.
The more difficult question is whether acting on that information improves survival or other outcomes that matter to patients.
Colorectal Cancer Leads the Evidence Base
Among gastrointestinal malignancies, colorectal cancer has the most mature evidence supporting ctDNA-based assessment of molecular residual disease.
The DYNAMIC trial provided an important demonstration of how ctDNA could reduce unnecessary treatment in stage II colon cancer. A ctDNA-guided strategy lowered adjuvant chemotherapy use from 28% to 15% without compromising 2-year recurrence-free survival.
The findings supported selective chemotherapy de-escalation. Nevertheless, the trial did not establish that ctDNA-positive patients benefit from chemotherapy because these patients were not randomized between chemotherapy and observation.
The study was also designed as a phase II noninferiority trial with an 8.5-percentage-point margin. It resulted in greater use of oxaliplatin in the ctDNA-guided group—9.5% compared with 2.7%—without evidence of superior outcomes.
DYNAMIC therefore showed that ctDNA can help spare chemotherapy in selected patients. It did not establish postoperative ctDNA positivity as an automatic indication for intensified treatment.
CIRCULATE Delivers an Important but Cautious Signal
The 2026 CIRCULATE AIO-KRK-0217/ABCSG trial offered the first prospective randomized indication that postoperative ctDNA positivity might help identify patients with stage II proficient mismatch repair and microsatellite-stable colon cancer who benefit from adjuvant chemotherapy.
In the per-protocol analysis, 3-year disease-free survival reached 77% with chemotherapy, compared with 38% with observation. The 3-year recurrence rate was 19% versus 62%, respectively.
However, the intention-to-treat primary endpoint did not reach statistical significance. Three-year disease-free survival was 61% with chemotherapy and 38% with observation.
Interpretation is further limited by the trial’s premature closure, small sample size, treatment nonadherence, and reliance on the per-protocol analysis.
The results move ctDNA closer to treatment-selection utility in colorectal cancer, but they do not support a universal or assay-independent treatment mandate.
DYNAMIC-III Separates Prognosis From Treatment Benefit
The DYNAMIC-III trial reinforced the strong prognostic value of postoperative ctDNA in stage III colon cancer.
Three-year recurrence-free survival was 87% among ctDNA-negative patients, compared with 49% among ctDNA-positive patients.
Yet this substantial separation in risk did not translate into a clear benefit from ctDNA-guided treatment modification.
Among ctDNA-negative patients, de-escalation reduced oxaliplatin use from 88.6% to 34.8% and lowered treatment-related hospitalizations from 13.2% to 8.5%. However, 3-year recurrence-free survival was 85.3% with de-escalation and 88.1% with standard management, and the strategy did not meet the prespecified noninferiority margin.
Among ctDNA-positive patients, treatment escalation also failed to improve outcomes. Two-year recurrence-free survival was 51% in the escalation group and 61% in the standard-management group.
Persistent ctDNA after treatment identified a particularly poor-risk population, with a 3-year recurrence-free survival of only 14%.
DYNAMIC-III confirmed that ctDNA can distinguish patients with profoundly different prognoses. It did not demonstrate that currently available escalation strategies can overcome the adverse biology associated with persistent molecular disease.
COBRA Exposes the Risks of Assay-Dependent Decisions
The COBRA trial tested ctDNA-guided treatment escalation in low-risk stage II colon cancer, a setting in which the baseline risk of recurrence is relatively low and assay specificity becomes especially important.
In its preplanned phase II analysis, ctDNA clearance at 6 months occurred in three of seven patients assigned to surveillance, or 43%, and in one of nine patients assigned to FOLFOX or CAPOX, or 11%.
The trial was closed early according to prespecified futility rules.
These findings differed substantially from previous reports, including results from GALAXY, where spontaneous clearance occurred in 12.2% of patients and chemotherapy-associated clearance in 68.5%.
The unexpected pattern in COBRA raised questions about biological fluctuations close to the assay’s detection threshold, false-positive classifications, and differences between testing platforms.
The study did not invalidate ctDNA as a prognostic biomarker. Instead, it demonstrated that ctDNA positivity in a low-risk population is not yet a sufficiently standardized therapeutic trigger for routine treatment escalation.
Earlier Detection Does Not Always Mean Better Salvage
ctDNA can identify molecular recurrence before disease becomes visible on imaging. Nevertheless, earlier detection is clinically valuable only when it leads to an intervention capable of improving outcomes.
In a real-world cohort of 184 patients with resected stage II to IV colorectal cancer, 20 patients had ctDNA-positive but imaging-negative recurrence at initial detection.
Only three patients—1.6% of the entire cohort—ultimately achieved durable disease-free status attributable to an intervention initiated because of ctDNA findings.
At the same time, 14 recurrences were detected through imaging despite negative ctDNA results. Three patients with positive ctDNA later cleared the molecular signal without developing radiographic recurrence.
These observations show that molecular detection can provide valuable prognostic information, but it may add limited curative benefit beyond high-quality, guideline-concordant imaging surveillance.
Beyond Colorectal Cancer, Prognosis Outpaces Utility
The evidence for ctDNA-directed treatment is less developed in other gastrointestinal cancers.
In gastro-esophageal adenocarcinoma, ctDNA detected after neoadjuvant therapy or surgery is associated with shorter progression-free survival and worse disease-specific survival. However, localized gastro-esophageal tumors can release relatively little ctDNA into the bloodstream.
In one cohort of patients with resected esophageal adenocarcinoma, 66% of those without detectable postoperative ctDNA still died from the disease within 40 months. This illustrates the risk of relying on a negative result in a low-shedding malignancy.
Several approaches are under investigation to improve detection, including tumor-informed mutation tracking, plasma-only next-generation sequencing, digital polymerase chain reaction, methylation signatures, and fragmentomic analysis.
A pan-gastrointestinal cell-free DNA methylation platform achieved an area under the curve of 0.90 for gastric cancer detection and also provided information about the likely tissue of origin.
Although these methods could strengthen recurrence-risk assessment, they require prospective validation before being used to direct treatment.
Pancreatic Cancer: Powerful Risk Stratification, Unproven Intervention
Pancreatic ductal adenocarcinoma provides another example of ctDNA’s strong prognostic value.
Among patients with resected disease, perioperative tumor-informed ctDNA positivity during the minimal residual disease window was associated with a median disease-free survival of 6.4 months, compared with 33.3 months among ctDNA-negative patients.
During surveillance, ctDNA positivity was associated with a 12.4-fold higher risk of recurrence. In multivariable analysis, it was the strongest prognostic factor, with a hazard ratio of 24.3.
These findings identify a population at exceptionally high risk of early relapse. They do not yet demonstrate that changing treatment according to ctDNA results can improve that outcome.
For now, ctDNA in pancreatic cancer remains a promising tool for perioperative risk stratification rather than a validated basis for treatment adaptation.
Cholangiocarcinoma Presents a Biological Challenge
ctDNA analysis in cholangiocarcinoma is complicated by the characteristics of the tumor itself.
Dense stromal fibrosis and limited vascular distribution can reduce the amount of tumor-derived DNA entering the circulation. Tumor-naive assays are therefore more likely to produce negative results when disease burden is low.
Testing remains exploratory and has not been standardized across studies or clinical settings.
At present, ctDNA appears more established as a tool for molecular profiling and prognostic stratification in cholangiocarcinoma than as a validated method for directing minimal residual disease treatment.
One Biomarker, Many Assays
Assay heterogeneity remains one of the greatest barriers to routine ctDNA implementation.
Tumor-informed and tumor-agnostic platforms differ in their tissue requirements, turnaround times, analytical thresholds, serial sensitivity, and susceptibility to false-positive or false-negative results.
Some tumor-informed assays can detect variant allele fractions close to 0.01%. Serial sensitivity has been reported at approximately 88% for tumor-informed approaches, compared with 59% for some tumor-agnostic methods.
Reported ctDNA clearance rates across major colorectal cancer datasets range from 11% to 68.5%. Such variation indicates that clearance cannot yet be treated as a standardized or stand-alone treatment endpoint.
ctDNA burden also appears to function as a biological continuum. Recurrence risk rises as the mutant allele fraction increases, rather than falling into a simple positive-or-negative classification.
Timing Can Change the Result
The postoperative timing of ctDNA collection is another unresolved issue.
Across pooled colorectal cancer cohorts, ctDNA was detected in 11.0% of patients at 4–6 weeks, 12.5% at 6–8 weeks, and 13.8% at 8–10 weeks after surgery.
Postoperative studies have reported ctDNA half-lives ranging from 35 to 139 minutes across tumor types.
Because ctDNA levels can change rapidly and may be influenced by surgery, treatment, and biological variation, no single postoperative sampling window has been fully standardized for treatment-directing assessment.
A positive or negative result must therefore be interpreted in the context of when the sample was collected.
When Molecular Risk Exceeds Clinical Actionability
A positive ctDNA result in a patient with no radiographic disease can reveal substantial recurrence risk. It does not establish that beginning systemic therapy before visible relapse will improve disease-free or overall survival.
This creates a difficult clinical situation. Patients and clinicians receive information about molecular recurrence without a validated strategy for responding to it.
Positive results can also carry psychological consequences, particularly when no disease is visible on imaging and the benefit of early intervention is uncertain.
Negative results present a different challenge. Current assays can miss patients who later experience recurrence, particularly in cancers with limited ctDNA shedding. In some settings, up to half of patients who eventually relapse can initially have negative molecular residual disease testing.
For this reason, ctDNA cannot replace imaging-based surveillance.
Building a More Reliable ctDNA Framework
The next phase of ctDNA development will likely depend less on isolated postoperative measurements and more on serial molecular trajectories.
Future approaches could interpret ctDNA alongside imaging, clinicopathological risk factors, metastatic patterns, treatment timing, and assay characteristics. Genomic alterations could also be combined with methylation and fragmentomic features to improve analytical sensitivity and tissue-of-origin assessment.
Other liquid-biopsy components—including circulating tumor cells, extracellular vesicles, circulating RNA, and protein biomarkers—could provide complementary information about viable disease, tumor phenotype, treatment resistance, and the tumor microenvironment.
Before these strategies enter routine practice, they will require standardized sample collection, prospectively validated decision thresholds, calibrated risk models, and trials demonstrating that biomarker-guided interventions improve patient outcomes.
A Promising Biomarker, Not Yet a Universal Mandate
The National Cancer Institute GI Oncology ctDNA Working Group has emphasized that ctDNA-guided treatment strategies require prospective validation. ctDNA has not yet been established as a surrogate endpoint for progression-free survival or overall survival in late-phase gastrointestinal cancer trials.
Colorectal cancer is advancing most rapidly toward clinical utility. Emerging randomized data suggest that postoperative ctDNA positivity could identify some patients with stage II proficient mismatch repair and microsatellite-stable colon cancer who benefit from adjuvant chemotherapy.
However, the nonsignificant intention-to-treat result and limitations of the CIRCULATE trial prevent that finding from being treated as definitive or universally applicable.
Across other gastrointestinal malignancies, ctDNA remains primarily a powerful prognostic biomarker.
Its clinical role must be defined according to cancer type, disease stage, testing platform, sampling schedule, and the maturity of prospective evidence. Until trials demonstrate that acting on ctDNA improves patient-important outcomes, it cannot serve as a universal stand-alone treatment mandate.
The promise of ctDNA is substantial. The path forward now depends on proving that increasingly precise molecular risk assessment can be converted into better care and better outcomes.
Written by Nare Hovhannisyan, MD
Explore more oncology news and features on OncoDaily