Highlight
– Postoperative circulating tumor DNA (ctDNA) positivity strongly predicts worse disease-free and overall survival in stage III colon cancer (adjusted HRs ≈ 6 for both endpoints).
– In a post hoc analysis of CALGB (Alliance)/SWOG 80702 (n=940), addition of adjuvant celecoxib was associated with improved DFS (aHR 0.61; 95% CI, 0.42–0.89) and OS (aHR 0.62; 95% CI, 0.40–0.96) among ctDNA-positive patients.
– Celecoxib did not show benefit in ctDNA-negative patients; however, interaction tests were not statistically significant, so predictive utility remains hypothesis-generating.
Background
Colon cancer remains a leading cause of cancer morbidity and mortality worldwide. For stage III disease, adjuvant combination chemotherapy (fluoropyrimidine plus oxaliplatin) reduces recurrence risk but does not eliminate it, and optimal personalization of adjuvant therapy remains an unmet need. Circulating tumor DNA (ctDNA) measured after surgery has emerged as a robust prognostic biomarker for minimal residual disease (MRD) and recurrence risk; trials such as DYNAMIC have demonstrated that ctDNA can guide adjuvant treatment decisions in stage II disease.
Separately, observational and mechanistic evidence implicates cyclooxygenase (COX) pathway modulation in colorectal carcinogenesis: aspirin use has been associated with reduced colorectal cancer incidence and improved survival in some observational cohorts, and benefit may be enriched in tumors with PIK3CA mutations. Selective COX-2 inhibition with celecoxib has biological plausibility to influence micrometastatic tumor growth, but randomized adjuvant trials have not shown an overall survival benefit across unselected populations. The CALGB (Alliance)/SWOG 80702 randomized clinical trial tested addition of celecoxib to standard adjuvant chemotherapy in stage III colon cancer without overall efficacy in the parent trial. The post hoc companion study examined whether postoperative ctDNA status might identify a subgroup that derives benefit from celecoxib.
Study design
This analysis is a predefined post hoc companion study of CALGB (Alliance)/SWOG 80702 (NCT01150045), a phase 3 randomized trial enrolling patients with resected stage III colon cancer between 2010 and 2015. The parent trial randomized participants to receive adjuvant celecoxib versus placebo in addition to standard chemotherapy; the trial also included randomization of 3 versus 6 months of adjuvant oxaliplatin-containing therapy as part of a factorial design.
For the companion biomarker analysis, 940 patients consented to biospecimen collection and had postoperative ctDNA measured using a tumor-informed, personalized 16-plex PCR–NGS assay (Signatera, Natera Inc.) on plasma obtained after surgery and before adjuvant therapy. The primary clinical endpoints evaluated were disease-free survival (DFS) and overall survival (OS). The pre-specified statistical plan assessed prognostic associations of ctDNA and whether ctDNA status predicted differential benefit from celecoxib.
Key findings
Population and follow-up: 940 patients were included (mean age 60.9 years; 45.3% female). Median follow-up was 6.0 years. Postoperative ctDNA was positive in 173 patients (18.4%) and negative in 767 (81.6%). Approximately 23.6% of patients reported prior low-dose aspirin use.
Prognostic performance of postoperative ctDNA
ctDNA positivity after surgery was strongly prognostic. Compared with ctDNA-negative patients, ctDNA-positive patients experienced substantially worse outcomes: adjusted hazard ratio (aHR) for DFS 6.12 (95% CI, 4.66–8.03) and for OS aHR 5.86 (95% CI, 4.19–8.19). These effect sizes confirm prior observations that ctDNA identifies patients at very high risk of recurrence after curative-intent resection.
Association of celecoxib with survival by ctDNA status
Among ctDNA-positive patients (n=173), the addition of celecoxib to standard chemotherapy was associated with improved outcomes versus placebo: DFS aHR 0.61 (95% CI, 0.42–0.89) and OS aHR 0.62 (95% CI, 0.40–0.96). In contrast, among ctDNA-negative patients (n=767), celecoxib did not confer a statistically significant survival benefit (DFS aHR 0.76; 95% CI, 0.53–1.09; OS aHR 0.85; 95% CI, 0.54–1.36).
However, formal tests for interaction between ctDNA status and treatment assignment were not significant (P for interaction = .41 for DFS and .33 for OS), indicating that while the point estimates suggest greater benefit in ctDNA-positive patients, the study could not definitively prove a predictive (treatment-by-biomarker) effect.
Subgroup analyses
The pattern of apparent benefit in ctDNA-positive patients persisted across strata defined by microsatellite instability (MSI) status and tumor PIK3CA mutation status. This is notable because observational data previously linked aspirin-related benefits to PIK3CA-mutant tumors; in this analysis, celecoxib effects did not appear limited to PIK3CA-mutant cases. The companion study did not report a statistically significant interaction by PIK3CA or MSI that would confirm selective benefit.
Safety and broader trial context
The parent trial did not demonstrate overall survival benefit of celecoxib when applied to the entire study population. COX-2 inhibitors, including celecoxib, have known cardiovascular risk considerations, and clinical decisions about adding a COX-2 inhibitor must weigh potential antineoplastic benefit against known adverse event profiles. The current companion analysis did not emphasize new safety signals but underscores the importance of individualized risk–benefit assessment.
Expert commentary and interpretation
These results illustrate several key points for clinicians and trialists:
- ctDNA is a robust prognostic marker in stage III colon cancer. A roughly sixfold increase in recurrence and mortality risk for ctDNA-positive patients reinforces ctDNA’s role in risk stratification.
- Signal for relative benefit of celecoxib among ctDNA-positive patients is clinically intriguing. The observed aHRs (~0.6) for DFS and OS suggest a potentially meaningful relative risk reduction in the MRD-positive subgroup.
- Because the interaction tests were not significant, the analysis is hypothesis-generating rather than definitive proof of a predictive biomarker. Post hoc analyses, even with pre-specified plans, are susceptible to confounding and require prospective validation.
- Biological plausibility exists: COX-2 inhibitors may alter tumor-promoting inflammation and prostaglandin-mediated pathways that support micrometastatic survival. Integration with prior observational and molecular data (e.g., aspirin and PIK3CA) suggests overlapping but distinct mechanisms.
- Cardiovascular safety considerations remain. Prior cardiovascular safety trials (for example, the PRECISION trial) provide context for celecoxib’s risk profile but do not obviate the need for individualized assessment in the adjuvant oncology setting.
Limitations
Key limitations include the post hoc nature of the companion study, potential imbalance in unmeasured confounders, and limited event counts within subgroups. The non-significant interaction p-values mean that the observed subgroup benefit may be due to chance, and definitive predictive biomarker claims require prospective ctDNA-guided randomization. Moreover, the study used a specific tumor-informed ctDNA assay (Signatera); generalizability to other assays depends on comparable sensitivity/specificity and assay timing.
Clinical implications and next steps
For practicing clinicians, the analysis reinforces ctDNA’s utility as a prognostic tool for stage III colon cancer and introduces an important hypothesis: patients with detectable postoperative ctDNA (MRD) may derive incremental benefit from adding celecoxib to standard adjuvant therapy. However, given the post hoc design and non-significant treatment-by-biomarker interaction, clinicians should not change standard practice solely on these data.
Strategic next steps include prospective randomized trials that stratify or randomize patients by postoperative ctDNA status to evaluate celecoxib (or other interventions) in MRD-positive patients. Such trials should prespecify endpoints, safety monitoring—particularly cardiovascular events—and standardized ctDNA assays and timing. Parallel translational studies probing COX-2 pathway activation in MRD and potential biomarkers of response (beyond ctDNA) would strengthen biological rationale and patient selection.
Conclusion
This prespecified post hoc companion analysis of CALGB (Alliance)/SWOG 80702 confirms the powerful prognostic value of postoperative ctDNA in stage III colon cancer and generates the hypothesis that adjuvant celecoxib may improve outcomes specifically among ctDNA-positive patients. These findings are promising but not definitive; prospective, ctDNA-stratified randomized trials are needed before routine use of celecoxib as MRD-directed adjuvant therapy can be recommended.
Funding and trial registration
The parent trial, CALGB (Alliance)/SWOG 80702, was conducted under cooperative group sponsorship; details and funding sources are reported in the trial publications. Trial registration: ClinicalTrials.gov Identifier: NCT01150045.
References
- Zhang GQ, Meyerhardt JA, Shi Q, et al. Predictive Role of Circulating Tumor DNA in Stage III Colon Cancer Treated With Celecoxib: A Post Hoc Analysis of the CALGB (Alliance)/SWOG 80702 Phase 3 Randomized Clinical Trial. JAMA Oncol. 2025 Dec 4:e255144. doi:10.1001/jamaoncol.2025.5144. PMID: 41343176.
- Tie J, et al. Circulating Tumor DNA–Guided Management of Stage II Colon Cancer. N Engl J Med. 2022;386(3):226–237. (DYNAMIC trial)
- Liao X, et al. Aspirin use, PIK3CA mutation, and colorectal-cancer survival. N Engl J Med. 2012;367(17):1596–1606.
- Rothwell PM, et al. Long-term effect of aspirin on colorectal cancer incidence and mortality: 20-year follow-up of five randomised trials. Lancet. 2010;376(9754):1741–1750.
- Steinbach G, et al. The effect of celecoxib, a cyclooxygenase-2 inhibitor, in familial adenomatous polyposis. N Engl J Med. 2000;342(26):1946–1952.
- FitzGerald GA, et al. Cardiovascular safety of celecoxib, naproxen, or ibuprofen for arthritis. N Engl J Med. 2016;375(26):2519–2529. (PRECISION trial)
