Circulating tumor DNA (ctDNA) is being investigated as a tool to identify molecular residual disease after surgery and potentially guide the intensity of adjuvant treatment in colon cancer. A new phase 2 study has evaluated whether serial ctDNA testing can be used to adapt postoperative chemotherapy in patients with resected high-risk stage II or stage III disease.
The results of the PEGASUS trial were published on September 14, 2026, in Nature Cancer under the original title, “Circulating tumor DNA-guided de-escalation or escalation of adjuvant therapy in high-risk stage II and stage III colon cancer: the phase 2 PEGASUS trial.”
Authors: Silvia Marsoni, Clara Montagut, Filippo Pietrantonio, Andrea Sartore-Bianchi, Luca Lazzari, Francesca Bergamo, Maria Giulia Zampino, Noelia Tarazona, Mario Mandalà, Stefano Tamberi, Elena Elez, Cristina Santos Vivas, Paolo Luraghi, Michele Prisciandaro, Federica Tosi, Davide Ciardiello, Joana Vidal, Victor Seguì, Michele Palazzo, Giulia Maddalena, Katia Bencardino, Ida Taglialatela, Marta Guix, Eliana Rulli, Valentina Vettore, Federica Morano, Claudio Isella, Giovanni Crisafulli, Giorgio Patelli, Enzo Medico, Alberto Bardelli, Andrés Cervantes, Valter Torri, Josep Tabernero, Salvatore Siena, and Sara Lonardi.
The PEGASUS Trial
PEGASUS (NCT04259944) was an investigator-initiated, multicenter, prospective, single-arm, nonrandomized phase 2 trial conducted across 11 centers in Italy and Spain. The study included patients with completely resected, microsatellite-stable (MSS) high-risk stage II pT4N0 or stage III pT1–4N1–2 colon cancer. The trial evaluated the feasibility of dynamically adapting adjuvant chemotherapy according to serial ctDNA results.
Between July 2020 and July 2022, 156 patients were screened and 135 were included in the final cohort. At the first postoperative liquid biopsy, 30 patients were ctDNA positive and 105 were ctDNA negative. Five initially ctDNA-negative patients tested positive at the planned confirmatory assessment, resulting in 35 patients (26%) being classified as ctDNA positive and 100 (74%) as ctDNA negative at the postoperative landmark.
Treatment was determined by ctDNA status. Patients with confirmed ctDNA negativity received 6 months of capecitabine monotherapy, while patients with postoperative ctDNA positivity received 3 months of CAPOX. Subsequent ctDNA assessments were used to further adapt treatment. Patients who remained ctDNA positive after CAPOX were escalated to 6 months of FOLFIRI, while patients who cleared ctDNA after CAPOX received 3 additional months of capecitabine with continued molecular monitoring. The primary endpoint was the 2-year relapse-free rate among patients with two consecutive negative postoperative ctDNA assessments.
Primary Endpoint and Clinical Outcomes
At a median follow-up of 41.0 months, 28 recurrences were observed. Recurrence occurred in 15 of 100 ctDNA-negative patients (15%) and 13 of 35 ctDNA-positive patients (37%). The trial had originally planned to evaluate 134 ctDNA-negative patients, but only 100 were accrued. Twelve relapses occurred within 2 years in this group, corresponding to a 2-year relapse-free rate of 88% (90% CI, 81–93%). This was below the prespecified threshold of at least 92%, meaning that the trial did not meet its primary endpoint.
Postoperative ctDNA status was nevertheless associated with subsequent outcomes. The estimated 2-year disease-free survival (DFS) was 87.7% among ctDNA-negative patients compared with 61.8% among ctDNA-positive patients, with an HR of 2.71 (95% CI, 1.35–5.45; P=0.0036). The median time to recurrence among patients who relapsed was 14.8 months in the ctDNA-negative group and 10.9 months in the ctDNA-positive group. At 3 years, DFS was 82.8% among ctDNA-negative patients and 58.4% among ctDNA-positive patients. Three-year overall survival was 93.8% and 83.9%, respectively.
Serial ctDNA Dynamics and Treatment Adaptation
Serial ctDNA monitoring provided additional information beyond the initial postoperative assessment. Among the 35 patients classified as ctDNA positive after surgery, 34 underwent the planned post-CAPOX ctDNA assessment, as one patient relapsed immediately after CAPOX. At this time point, 23 of 34 patients (68%) remained ctDNA positive, while 11 (32%) converted to ctDNA negative.
However, clearance was transient in 8 of these 11 patients. Only three patients achieved sustained ctDNA negativity after 3 months of CAPOX and remained relapse free beyond 38 months.
Among 24 patients who switched to FOLFIRI, 13 remained ctDNA positive, six of whom relapsed, while 11 converted to ctDNA negative. None of the 11 patients who converted to ctDNA negativity after FOLFIRI had relapsed at a median follow-up of 46.5 months. These observations were exploratory and do not establish FOLFIRI as an effective adjuvant escalation strategy, but they demonstrated differing molecular trajectories among patients with persistent postoperative MRD.
Across the study, 100 patients were ctDNA negative at the postoperative landmark. Of these, 97 remained persistently negative during subsequent testing, while three later converted to ctDNA positive. Among the 35 patients classified as ctDNA positive at the postoperative landmark, 15 remained persistently positive, eight achieved transient clearance, and 12 achieved sustained clearance. No relapses were recorded among patients who achieved sustained ctDNA clearance during the reported follow-up.
Molecular Detection Before Radiological Relapse
The study also evaluated whether ctDNA could identify recurrence before conventional imaging. Among patients with a positive ctDNA result during follow-up who subsequently developed radiologically confirmed recurrence, molecular detection preceded radiological relapse by a median of 9.1 months.
However, ctDNA detection differed according to the site of recurrence. Of 13 liver-only recurrences, nine were detected by ctDNA. In contrast, none of the five lung-only or three peritoneum-only recurrences were ctDNA positive at any tested time point. These findings highlight an important limitation of plasma-based MRD testing, as ctDNA negativity did not exclude subsequent recurrence, particularly for disease recurring in the lung or peritoneum.
Comparison With the TOSCA Cohort
The investigators compared the PEGASUS population with a propensity score-matched historical cohort from the TOSCA trial. Two-year DFS was 79.7% in PEGASUS and 82.2% in the matched TOSCA cohort, with no significant difference between the groups (HR, 0.89; 95% CI, 0.51–1.55; P=0.669). The ctDNA-guided strategy was associated with substantially less treatment-related neurotoxicity. Neurological toxicity was reduced by an absolute 43.2% compared with the matched TOSCA cohort.
Grade 1–2 neurological toxicity occurred in 13.8% of PEGASUS patients compared with 41.5% in TOSCA, while grade 3–4 neurological toxicity occurred in 3.8% and 19.3%, respectively. The calculated number needed to treat with the PEGASUS strategy to prevent one case of neurotoxicity was 2.32 (95% CI, 1.86–3.06).
Because TOSCA served as a matched historical control rather than a randomized comparator, these efficacy and safety comparisons remain exploratory.
Tumor Biology and ctDNA
The investigators also performed transcriptomic profiling of resected primary tumors. RNA-sequencing data were available for 132 of the 135 patients. CMS4 was the most frequent consensus molecular subtype, identified in 37 patients (28%), followed by CMS2 in 35 (26%), CMS1 in 21 (16%), and CMS3 in 18 (14%); 21 tumors (16%) were unclassified.
CMS4 was more frequent among patients who relapsed, including among some patients who were ctDNA negative after surgery. In multivariable analysis, postoperative ctDNA positivity was independently associated with inferior DFS (HR, 2.75; 95% CI, 1.31–5.76; P=0.0074), as was CMS4 status (HR, 3.02; 95% CI, 1.46–6.24; P=0.0029). The findings suggest that ctDNA status and tumor molecular characteristics may provide complementary information about recurrence risk.
Takeaway
PEGASUS showed that serial ctDNA-guided treatment adaptation can be operationally implemented in resected MSS high-risk stage II and stage III colon cancer and that postoperative and longitudinal ctDNA status provide important prognostic information.
However, the trial did not meet its predefined primary endpoint. The study was also a relatively small, nonrandomized phase 2 trial, and the comparison with TOSCA relied on a matched historical cohort. The plasma-only Reveal v1.2 assay used in PEGASUS has since been superseded by a more sensitive version.
The ctDNA-guided strategy was associated with substantially lower neurotoxicity than the historical TOSCA comparator, while sustained ctDNA clearance was associated with favorable outcomes. At the same time, recurrences among ctDNA-negative patients, particularly in lung- and peritoneum-limited disease, highlight the current limitations of plasma-based MRD detection.
The investigators concluded that randomized phase 3 validation is required. The findings support further evaluation of oxaliplatin de-escalation in selected ctDNA-negative patients and highlight persistent or recurrent ctDNA positivity as a molecular state requiring more effective therapeutic strategies beyond conventional cytotoxic chemotherapy.
The full article is available in Nature Cancer.
