A prospective analysis from the CIRCULATE-Japan GALAXY study demonstrates that postsurgical circulating tumor DNA (ctDNA)–defined molecular residual disease (MRD) status strongly predicts survival outcomes and identifies which patients with resected colorectal liver metastases (CRLM) are most likely to benefit from adjuvant chemotherapy (ACT). Published in JAMA Oncology, the updated analysis of 298 patients provides the first evidence that ctDNA-guided risk stratification can identify an overall survival (OS) benefit from ACT in MRD-positive patients undergoing upfront surgery — a finding with significant implications for personalizing postsurgical management in this setting.

Clinical Takeaway

✔ MRD positivity 2–10 weeks after CRLM resection is the strongest independent prognostic factor for inferior DFS and OS, outperforming established clinicopathologic variables in multivariable analysis.

✔ Among upfront surgery patients with MRD positivity, ACT was associated with significantly improved OS (adjusted HR, 0.27; 95% CI, 0.08–0.88; P = .03) — the first reported OS benefit tied to ctDNA-guided ACT selection in resected CRLM.

✔ MRD-negative patients experienced excellent long-term OS without clear benefit from ACT, suggesting observation may be appropriate for this low-risk subgroup.

✔ ACT was not associated with improved outcomes in the neoadjuvant chemotherapy (NAC) cohort regardless of MRD status, potentially reflecting chemoresistant biology and prior treatment exposure.

Background & Rationale

The liver is the most common site of distant metastasis in colorectal cancer (CRC), and recurrence occurs in up to two-thirds of patients following curative-intent hepatectomy for CRLM. Despite decades of investigation, the role of adjuvant chemotherapy after CRLM resection remains highly contested. Randomized trials have consistently demonstrated improvements in disease-free survival (DFS) but have failed to establish a conclusive overall survival benefit, and no validated biomarker has been established to guide patient selection for postsurgical therapy. The Signatera personalized, tumor-informed ctDNA assay (Natera) has emerged as a leading tool for detecting MRD after curative-intent resection, with prior data from CIRCULATE-Japan demonstrating its prognostic power across CRC stages. This updated analysis sought to extend prior DFS findings by evaluating whether ctDNA-defined MRD status predicts OS benefit from ACT in resected CRLM, and to examine this association separately in patients receiving NAC prior to surgery.

Assay Profile: Signatera (Tumor-Informed ctDNA Testing)

  • Assay type: Personalized, tumor-informed multiplex polymerase chain reaction next-generation sequencing (PCR-NGS) assay
  • Methodology: Whole-exome sequencing of formalin-fixed, paraffin-embedded tumor tissue and matched normal peripheral blood DNA to identify patient-specific somatic single-nucleotide variants (SNVs); up to 16 tumor-specific variants tracked in plasma
  • MRD positivity definition: Detection of at least 2 tumor-specific variants above a predefined threshold in cell-free DNA
  • Quantification: ctDNA levels expressed as mean tumor molecules (MTM) per milliliter of plasma
  • Testing window: 2–10 weeks postsurgery for the primary landmark analysis; serial surveillance samples collected at 4, 12, 24, 36, 48, 72, and 96 weeks after surgery

    Study Design

    • Study: Prospective subset analysis of the CIRCULATE-Japan GALAXY study — a nationwide, prospective registry of patients with clinical stage II–IV CRC undergoing curative resection
    • Enrollment period: May 2020 – July 2024
    • Analytic period: January – April 2026
    • Inclusion criteria: Surgically resected CRLM without extrahepatic metastases, with available MRD results 2–10 weeks after curative-intent hepatectomy
    • Exclusions: Patients with extrahepatic metastases; patients enrolled in interventional CIRCULATE-Japan trials (ALTAIR, VEGA, FANTASTIC)

    NAC cohort (n = 107): Neoadjuvant chemotherapy administered prior to hepatectomy

    • Statistical methods: Kaplan-Meier survival analysis; Cox proportional hazards models adjusted for age, sex, RAS status, size and number of liver metastases, prior oxaliplatin exposure, and synchronicity; landmark analyses at 70 days postsurgery to address immortal time bias; time-dependent Cox models for longitudinal MRD surveillance

    Patient Population

    • Total patients: 298
    • Median age: 67 years (range, 33–85)
    • Sex: 64.4% male, 35.6% female
    • Median follow-up: 43.2 months (range, 1–68)
    • R0 resection rates: 96.3% (upfront surgery cohort) vs 86.0% (NAC cohort)
    • ACT administration: 25.6% of upfront surgery patients; 33.6% of NAC patients
    • NAC cohort characteristics: Higher rates of multiple liver metastases, larger tumors, synchronous disease, prior oxaliplatin exposure, and major hepatectomy compared with the upfront surgery cohort

    Primary Endpoints

    • Disease-free survival (DFS)
    • Overall survival (OS)

    Efficacy Outcomes: Prognostic Impact of MRD Status

    • MRD positivity rate: 32.4% (62/191) at the 2–10 week postsurgical window
    • MRD positivity vs negativity — DFS: HR, 4.14 (95% CI, 2.69–6.36; P < .001); 48-month DFS 15.6% vs 59.6%
    • MRD positivity vs negativity — OS: HR, 9.13 (95% CI, 4.44–18.78; P < .001); 48-month OS 41.4% vs 91.1%
    • Multivariable analysis: MRD positivity was the strongest independent prognostic factor for inferior DFS (HR, 7.89; 95% CI, 4.33–14.38; P < .001) and OS (HR, 12.03; 95% CI, 5.08–28.46; P < .001)
    • Additional adverse prognostic factors: RAS variants (DFS and OS) and liver metastasis size >5 cm (DFS only)
    • Longitudinal surveillance: Serial MRD negativity associated with substantially improved DFS (HR, 11.27; 95% CI, 6.61–19.21; P < .001) and OS (HR, 12.37; 95% CI, 4.22–36.29; P < .001) vs MRD positivity at any surveillance time point

    NAC Cohort

    • MRD positivity rate: 47.7% (51/107) postsurgically — notably higher than the upfront surgery cohort
    • Presurgical ctDNA positivity rate (post-NAC): 82.2% (83/101)
    • Presurgical ctDNA positivity — OS: HR, 11.42 (95% CI, 1.58–1452.50; P = .009); no significant association with DFS
    • Postsurgical MRD positivity vs negativity — DFS: HR, 4.82 (95% CI, 2.92–7.97; P < .001); median DFS 3.5 months (MRD positive) vs 20.7 months (MRD negative)
    • Postsurgical MRD positivity vs negativity — OS: HR, 9.43 (95% CI, 2.78–31.96; P < .001); 48-month OS 52.9% vs 92.9%
    • ctDNA negativity post-NAC: 94% of patients who were ctDNA-negative presurgically remained MRD negative at the postsurgical landmark

    Efficacy Outcomes: ACT Benefit by MRD Status

    Upfront Surgery Cohort — MRD-Positive Patients

    • DFS (ACT vs observation): Adjusted HR, 0.07 (95% CI, 0.02–0.24; P < .001); 48-month DFS 37.5% vs not reached in the observation arm (note: the not-reached arm reflects the observation group’s DFS plateau at a lower level due to near-universal early events)
    • OS (ACT vs observation): Adjusted HR, 0.27 (95% CI, 0.08–0.88; P = .03); 48-month OS 65.3% vs 32.9%

    Upfront Surgery Cohort — MRD-Negative Patients

    • DFS (ACT vs observation): Adjusted HR, 0.69 (95% CI, 0.32–1.50; P = .36); 48-month DFS 72.4% vs 55.0% — no statistically significant difference
    • OS (ACT vs observation): Adjusted HR, 0.54 (95% CI, 0.10–2.88; P = .47); 48-month OS 93.2% vs 90.3% — no statistically significant difference

    NAC Cohort — MRD-Positive Patients

    • DFS (ACT vs observation): HR, 0.69 (95% CI, 0.32–1.49; P = .35) — no benefit
    • OS (ACT vs observation): HR, 0.81 (95% CI, 0.27–2.41; P = .70) — no benefit

    NAC Cohort — MRD-Negative Patients

    • DFS (ACT vs observation): HR, 0.63 (95% CI, 0.29–1.37; P = .24) — no benefit
    • OS (ACT vs observation): HR, 0.73 (95% CI, 0.07–8.20; P = .80) — no benefit

    ctDNA Level Analysis and Recurrence Patterns

    • In the upfront surgery cohort, ACT was associated with improved DFS compared with observation in both low ctDNA (≤1 MTM/mL) and high ctDNA (>1 MTM/mL) subgroups, suggesting that the binary MRD classification captures the primary predictive signal without added benefit from quantitative stratification
    • No significant association between ACT and DFS was observed by ctDNA level subgroup in the NAC cohort
    • Recurrence patterns differed by MRD status:
        • MRD-positive patients: Higher rates of liver metastases at recurrence (upfront surgery: 70% vs 36%; NAC: 68% vs 48%)
        • MRD-negative patients: Higher rates of lung metastases at recurrence (upfront surgery: 50% vs 13%; NAC: 38% vs 14%), potentially reflecting lower ctDNA shedding from pulmonary sites
        • Among a small subset of patients (n = 6) with available post-metastasis-directed therapy (MDT) ctDNA data, 2 ctDNA-negative patients remained recurrence-free through 28–40 months of follow-up, while all 4 ctDNA-positive patients experienced recurrence within 1–13 months post-MDT

      Why ACT May Not Benefit NAC-Treated Patients

      The absence of ACT benefit in the NAC cohort — even among MRD-positive patients — may reflect several interrelated factors that clinicians should consider:

    • Higher baseline disease burden: The NAC cohort had greater tumor burden, higher rates of synchronous metastases, and more frequent major hepatectomy, suggesting more aggressive tumor biology less amenable to postsurgical systemic therapy
    • Chemotherapy resistance: Prior exposure to oxaliplatin-based regimens may select for chemoresistant clones, attenuating the effectiveness of subsequent ACT — a hypothesis supported by emerging multi-omic data on CRC evolution after treatment
    • Higher MRD positivity rate: The 47.7% MRD positivity rate in the NAC cohort (vs 32.4% upfront) may reflect enrichment for biologically aggressive, therapy-resistant residual disease that cannot be eradicated by conventional cytotoxic regimens
    • Insufficient intensification: Consistent with findings from the DYN,AMIC-III trial (escalation to FOLFOXIRI in MRD-positive locally advanced colon cancer) and the FANTASTIC study (mFOLFOXIRI after metastatic CRC resection), intensification of cytotoxic chemotherapy alone appears insufficient to clear MRD in high-risk patients, highlighting the need for novel biologically targeted strategies

    Key Clinical Implications

    ✔ ctDNA-defined MRD status at 2–10 weeks post-surgery should be considered a powerful tool for risk stratification after CRLM resection, providing prognostic information beyond standard clinicopathologic variables including RAS status, tumor size, and synchronicity.

    ✔ For patients undergoing upfront surgery, MRD-positive status identifies those most likely to derive meaningful OS benefit from ACT, with a 48-month OS of 65.3% with ACT versus 32.9% with observation — a clinically compelling difference that supports biomarker-driven treatment selection.

    ✔ MRD-negative patients after upfront surgery experience excellent long-term survival (48-month OS >90%) without ACT benefit, and may reasonably be managed with observation, avoiding unnecessary chemotherapy toxicity.

    ✔ These findings may help explain the historical failure of randomized ACT trials in CRLM to show OS benefit, as those trials enrolled biomarker-unselected, biologically heterogeneous populations in which the benefit in MRD-positive patients was diluted by the lack of benefit in MRD-negative patients.

    ✔ For NAC-treated patients, alternative strategies beyond conventional ACT are needed for those with persistent postsurgical MRD positivity, including clinical trial enrollment evaluating targeted therapies or novel combinations.

    ✔ Serial ctDNA monitoring beyond the initial postsurgical window provides additional prognostic value, with continued MRD negativity during surveillance strongly associated with long-term DFS and OS in both cohorts.

    ✔ Prospective, randomized trials evaluating ctDNA-guided ACT decisions in resected CRLM are needed to validate these findings and establish ctDNA-directed management as a standard-of-care approach in this setting.

    Study Limitations

      • This was a nonrandomized, retrospective subset analysis; residual confounding cannot be excluded despite multivariable

     

    Source:

      • Kataoka K, Ito K, Nakamura Y, et al. Circulating Tumor DNA Status and Adjuvant Chemotherapy in Resected Colorectal Liver Metastases. JAMA Oncol. Published online July 03, 2026. doi:10.1001/jamaoncol.2026.2191