Background
Prospective, multicentre, randomised phase 3 (FIND; NCT05904665). N=584 in the modified intention-to-treat population (289 ctDNA-guided, 295 control). Non-metastatic colorectal cancer after curative resection. Hypothesis: a dynamic surveillance strategy driven by circulating tumour DNA (ctDNA) methylation increases the proportion of recurrences that can still be treated with curative intent, by detecting relapse earlier and at lower metastatic burden.
Results
Interventions and follow up: Arm A: ctDNA methylation-guided dynamic surveillance — quarterly ctDNA testing; a positive result triggered immediate CT; if negative, bimonthly CT continued alongside quarterly ctDNA; after two consecutive negative ctDNA results imaging reverted to standard frequency (n=289)
Arm B: Standard CT-based surveillance (n=295)
Primary endpoint: Proportion of patients with recurrence who received curative-intent metastasis-directed therapy
mFollow up: 23.3 months.
Results: Curative-intent therapy for recurrence (primary): 48.1% vs 23.6%, relative risk 2.03, P=.008.
Recurrence rate: 18.0% vs 18.6%, P=.919.
Time to clinical recurrence: 9.5 vs 13.4 months, P<.001 (lead time 3.9 months).
Curative resection, liver and/or lung-confined recurrence: 42.3% vs 18.2%, P=.002.
Hepatic metastases ≤3 lesions: 75.0% vs 28.6%, P=.005.
Hepatic lesion ≤3 cm: 90.0% vs 57.1%, P=.033.
Unilobar hepatic disease: 80.0% vs 28.6%, P=.002.
DFS / OS: NR — pending mature data.
Arm B: Standard CT-based surveillance (n=295)
Primary endpoint: Proportion of patients with recurrence who received curative-intent metastasis-directed therapy
mFollow up: 23.3 months.
Results: Curative-intent therapy for recurrence (primary): 48.1% vs 23.6%, relative risk 2.03, P=.008.
Recurrence rate: 18.0% vs 18.6%, P=.919.
Time to clinical recurrence: 9.5 vs 13.4 months, P<.001 (lead time 3.9 months).
Curative resection, liver and/or lung-confined recurrence: 42.3% vs 18.2%, P=.002.
Hepatic metastases ≤3 lesions: 75.0% vs 28.6%, P=.005.
Hepatic lesion ≤3 cm: 90.0% vs 57.1%, P=.033.
Unilobar hepatic disease: 80.0% vs 28.6%, P=.002.
DFS / OS: NR — pending mature data.
Adverse events
Investigational drug exposure: none — this is a diagnostic and surveillance-strategy trial, not a therapeutic one.
Treatment-emergent adverse events: NR; no protocol therapy was administered.
Strategy-related harms: NR — incremental CT radiation exposure, false-positive-triggered imaging, and testing-related anxiety were not quantified.
Procedural morbidity of additional metastasectomy: NR.
Treatment-emergent adverse events: NR; no protocol therapy was administered.
Strategy-related harms: NR — incremental CT radiation exposure, false-positive-triggered imaging, and testing-related anxiety were not quantified.
Procedural morbidity of additional metastasectomy: NR.
Conclusions
ctDNA methylation-guided dynamic surveillance did not change how often patients recurred (18.0% vs 18.6%), but doubled the proportion of recurrences amenable to curative-intent treatment (48.1% vs 23.6%) by detecting relapse a median 3.9 months earlier and at lower metastatic burden. This is the first randomised phase 3 demonstration that a ctDNA-driven surveillance algorithm changes the actionability of recurrence after curative resection of colorectal cancer — though whether that translates into longer survival is unresolved.
Key Limitations
The primary endpoint is a process and actionability measure, not survival — earlier detection with more resections is exactly the pattern lead-time and length bias produce, and DFS and OS remain immature at 23.3 months. The shorter median time to clinical recurrence in the intervention arm is a direct manifestation of lead time and does not by itself establish benefit. Open-label, with unavoidable performance bias: knowing a patient is ctDNA-positive plausibly lowers the threshold for offering metastasectomy, inflating the primary endpoint independent of tumour biology. Several subgroup comparisons rest on very small denominators (hepatic-feature analyses involve roughly 10–14 patients per arm). The assay is a specific methylation platform with industry co-author affiliation and is not interchangeable with tumour-informed ctDNA assays. Predominantly Chinese multicentre population. Non-metastatic disease was pooled across stages and across colon and rectal primaries without reported stage-stratified results.
Clinical Context
Standard post-resection surveillance in non-metastatic colorectal cancer is CEA plus periodic CT imaging per NCCN, ASCO and ESMO. ctDNA is currently recognised as prognostic but is not endorsed for routine treatment or surveillance decision-making outside a clinical trial: ASCO's ctDNA guideline advises against using ctDNA measures as surrogate markers of disease burden for routine decisions outside trials, and NCCN's colon and rectal guidelines state the evidence is insufficient to recommend routine ctDNA assay use outside a trial. FIND is the surveillance-strategy counterpart to the therapeutic ctDNA trials (the DYNAMIC series and adjuvant escalation/de-escalation designs) and is the first to show a randomised effect on curative-intent salvage. It strengthens the case for ctDNA-guided surveillance but does not yet clear the survival bar that guidelines have set, so it should not by itself displace standard CEA-and-CT follow-up.