ctDNA Liquid Biopsy Predicts Colorectal Cancer Recurrence, But Acting on It Remains Contested

A blood test that detects molecular traces of colorectal cancer after surgery has become one of the most powerful prognostic tools in oncology, yet a new Viewpoint published in The Lancet Regional Health – Western Pacific argues that the field has reached a critical juncture. Circulating tumour DNA, or ctDNA, can identify residual disease with extraordinary precision, but whether acting on the result actually improves patient survival remains unresolved, with recent interventional trials yielding mixed and in some cases cautionary results. The authors, Alicia Yioli Lefas and Ian Chau, propose that discordant findings should be interpreted through four linked questions: what biological state ctDNA positivity represents, when the result is obtained and acted upon, which assay and sampling strategy is used, and whether the trial randomises a management strategy or a treatment within a ctDNA-defined group.

The prognostic evidence is now overwhelming. In the GALAXY study within CIRCULATE-Japan, the largest prospective observational dataset evaluating postoperative ctDNA in resected colorectal cancer, 15.9 per cent of 2109 patients were ctDNA-positive during the molecular residual disease window of two to ten weeks after surgery. Among these, 78.3 per cent relapsed, compared with just 13.1 per cent of ctDNA-negative patients. The hazard ratios were staggering: ctDNA positivity conferred a disease-free survival hazard ratio of 11.99 and an overall survival hazard ratio of 9.68 in the overall cohort, rising to a disease-free survival hazard ratio of 15.97 in resected stage I–III colon cancer. Three-year overall survival in stage I–III colon cancer was 69.3 per cent for ctDNA-positive patients versus 96.8 per cent for ctDNA-negative patients. A recent meta-analysis confirmed a more than threefold increased recurrence risk in postoperative ctDNA-positive stage II colorectal cancer, with a pooled risk ratio of 3.66.

Comparable prognostic separation has been demonstrated across continents and platforms. In the US-based BESPOKE study, ctDNA positivity was observed in 8.1 per cent of stage II and 24.9 per cent of stage III patients, and was strongly associated with inferior disease-free survival in both stages, with hazard ratios exceeding 10. At 24 months, disease-free survival for stages II–III combined was 41.4 per cent in ctDNA-positive patients and 91.7 per cent in ctDNA-negative patients. In a post hoc combined analysis of the randomised phase III PRODIGE-GERCOR IDEA-France and HORG-IDEA-Greece trials, five-year overall survival was 89.4 per cent in ctDNA-negative patients versus 52.7 per cent in ctDNA-positive patients, with a median follow-up of 6.7 years. ctDNA status remained an independent prognostic factor after adjustment for established clinicopathological variables.

Longitudinal kinetics add a further layer of information. In GALAXY, sustained ctDNA clearance after adjuvant chemotherapy identified a favourable subgroup, with disease-free survival events occurring in only 10.3 per cent of patients with sustained clearance, compared with 86.2 per cent with transient clearance and 100 per cent with no clearance. Twenty-four-month disease-free survival was 89.0 per cent versus 3.3 per cent for sustained and transient clearance, respectively. True spontaneous clearance without clinical recurrence was uncommon at 1.9 per cent. During surveillance, ctDNA positivity typically preceded radiological recurrence by a median of 5.91 months, with lead times across prospective studies ranging from six to ten months, underscoring the potential of serial monitoring to identify molecular recurrence well before imaging.

Yet the predictive value of ctDNA remains contested. The Australian DYNAMIC trial randomised patients with resected stage II colon cancer to a ctDNA-informed strategy or standard management. Chemotherapy use was substantially lower in the ctDNA-guided arm at 15 per cent versus 28 per cent, and the trial met its prespecified non-inferiority endpoint for two-year recurrence-free survival. However, among patients who received adjuvant chemotherapy, oxaliplatin-based doublets were used substantially more often in the ctDNA-guided arm at 62 per cent versus 10 per cent, suggesting that the strategy redistributed treatment intensity rather than uniformly reducing burden. With five-year follow-up, recurrence-free and overall survival curves showed little separation between arms. Exploratory subgroup analyses suggested a signal in T4 disease, but because treatment selection was not randomised and oxaliplatin use differed substantially between strategies, the analysis cannot separate an effect of ctDNA stratification from the effect of preferential oxaliplatin allocation.

More cautionary results have emerged from stage III disease. DYNAMIC-III, a multicentre phase II/III randomised trial in resected stage III colon cancer, found that ctDNA-guided de-escalation substantially reduced oxaliplatin use from 88.6 per cent to 34.8 per cent, but non-inferiority was not demonstrated, with three-year recurrence-free survival numerically lower in the ctDNA-guided arm at 85.3 per cent versus 88.1 per cent, narrowly crossing the prespecified non-inferiority margin of −7.5 percentage points. In the ctDNA-positive cohort, escalation including FOLFOXIRI did not improve recurrence-free survival or meaningfully enhance ctDNA clearance compared with standard adjuvant chemotherapy. The COBRA/NRG-GI005 trial was stopped for futility in its phase II component because chemotherapy did not improve six-month ctDNA clearance among the first 16 ctDNA-positive patients. These findings question both the benefit of ctDNA-guided escalation in low-risk disease and the validity of ctDNA clearance as a surrogate endpoint.

The ALTAIR trial within CIRCULATE-Japan provides perhaps the most instructive negative result. Patients with ctDNA positivity after curative-intent resection and no radiological evidence of disease were randomised to trifluridine/tipiracil or placebo. The primary endpoint was not met, with median disease-free survival of 9.30 versus 5.55 months and a hazard ratio of 0.79, alongside substantial haematological toxicity. Critically, ctDNA clearance was observed in 12.4 per cent of placebo-treated patients, a finding that challenges the assumption that a single positive result invariably represents a stable and inexorably progressive disease state. Analytical variation near the limit of detection, intermittent tumour shedding, delayed effects of prior treatment, and genuine biological clearance may each contribute, supporting confirmatory serial testing and predefined persistence criteria in future molecular residual disease-directed trials.

The authors argue that a positive postoperative test can represent several biological states: small-volume treatment-sensitive disease, residual clones intrinsically resistant to standard chemotherapy, intermittent shedding near the analytical threshold, or already established micrometastatic disease below radiological detection. These states are unlikely to respond uniformly to the same intervention. Timing further complicates interpretation, as the molecular residual disease window varies across trials from two to twelve weeks, and a result obtained before adjuvant therapy should not be assumed to have the same biological meaning as one emerging months after treatment. Assay platforms also differ substantially: tumour-informed assays such as Signatera and NeXT Personal track patient-specific variants and generally achieve higher sensitivity in serial sampling at 0.88 versus 0.59 for tumour-agnostic approaches, while tumour-agnostic assays such as Guardant Reveal and Tempus xM avoid tissue dependency and can be deployed rapidly. A recent meta-analysis found no statistically significant difference in specificity between the two categories. Trial results should therefore not be generalised as class effects of ctDNA.

Surveillance trials add a further dimension. The Chinese FIND study randomised patients to standard or ctDNA-guided surveillance following curative resection, and found that ctDNA-guided monitoring shortened the time to clinical recurrence and increased curative-intent treatment among patients who recurred, from 23.6 per cent to 48.1 per cent. Among recurrences confined to the liver and/or lungs, curative resection was also more frequent at 42.3 per cent versus 18.2 per cent. However, the authors caution that several biases can inflate the apparent value of earlier detection, including lead-time bias, length-time bias, and site-dependent shedding that favours detection of liver-dominant recurrence, which is also more amenable to metastasis-directed therapy. A higher metastasectomy rate is an intermediate endpoint, not a substitute for clinical benefit, and a survival advantage has not yet been established.

The Western Pacific region has been central to evidence generation, with GALAXY and ALTAIR recruiting across Japan and Taiwan, Australia leading DYNAMIC and DYNAMIC-III, China developing the FIND surveillance strategy and CINTS-R neoadjuvant programme, and Korea evaluating intensified therapy in the CLAUDIA platform. However, prospective clinical-utility evidence remains sparse across much of Southeast Asia and lower-resource settings, where implementation requires pathology access, standardised plasma processing, reliable cold-chain transport, high-complexity sequencing, and rapid communication of results. The authors conclude that outside clinical trials, ctDNA testing may be considered to provide additional prognostic information when interpreted with the patient in the context of stage, pathological risk, and comorbidity, but should not currently be used alone to omit, shorten, intensify, or switch adjuvant chemotherapy, initiate systemic therapy for molecular recurrence without radiological disease, or replace standard surveillance. Until the field proves when, how, and for whom acting on ctDNA improves survival, quality of life, or treatment burden, ctDNA should refine prognosis and support research, not replace clinicopathological judgement or dictate postoperative treatment.

Subject of Research: Circulating tumour DNA for molecular residual disease detection in colorectal cancer

Article Title: Consensus and controversy of liquid biopsy using circulating tumour DNA (ctDNA) for molecular residual disease detection in colorectal cancer

Article References: Yioli Lefas, A., & Chau, I. (2026). Consensus and controversy of liquid biopsy using circulating tumour DNA (ctDNA) for molecular residual disease detection in colorectal cancer. The Lancet Regional Health – Western Pacific, 75, Article 102001. https://doi.org/10.1016/j.lanwpc.2026.102001

Image Credits: AI Generated

DOI: 10.1016/j.lanwpc.2026.102001

Keywords: ctDNA, liquid biopsy, colorectal cancer, molecular residual disease, adjuvant chemotherapy, GALAXY, DYNAMIC, DYNAMIC-III, ALTAIR, prognostic biomarker, surveillance, Western Pacific