Time-dependent recurrence patterns after curative resection for stage II/III colorectal cancer: An integrated analysis of four phase III randomized controlled trials (JCOG2310A-S4).

BACKGROUND

Postoperative surveillance for colorectal cancer (CRC) remains largely uniform despite potentially time-dependent recurrence hazards. Understanding temporal recurrence dynamics may help refine surveillance strategies. However, data on temporal recurrence patterns in CRC remain limited, particularly across stage, risk, and recurrence sites.

METHODS

This retrospective pooled individual participant data analysis integrated four RCTs conducted by the Japan Clinical Oncology Group. A total of 3655 patients with curatively resected pathological stage (pStage) II/III CRC were analysed (816 pStage II; 2839 pStage III). Time-dependent recurrence hazards were estimated using biweight kernel-smoothed hazard functions. Peak timing was defined as the maximum hazard within the first 5 years postoperatively. Analyses were stratified by pStage, prognostic factors identified using multivariable Cox regression, and recurrence site.

RESULTS

Five-year relapse-free survival (RFS) was 87.9% in pStage II and 77.0% in pStage III. Peak recurrence hazard was observed at 10.0 and 12.8 months in pStage II and III, respectively. In pStage II, age ≥ 70 years and pT4 were independent poor prognostic factors. In pStage III, age ≥ 70 years, male sex, pT4, pN2 and absence of adjuvant chemotherapy (ACT) were independently associated with poorer RFS. Patients aged < 70 years in pStage II showed earlier peak timing than those aged ≥ 70 years (9.0 versus 12.9 months). In pStage III, patients without ACT showed earlier peak timing than those receiving ACT (3.3 versus 12.8 months). Site-specific analysis showed that liver metastasis had the earliest peak recurrence hazard (8.0 months), particularly in pStage II (6.0 months).

CONCLUSIONS

Postoperative recurrence hazards showed distinct temporal patterns according to pStage, prognostic factors, and metastatic sites. These findings suggest that considering not only who is at risk, but also when the risk is highest, supports the development of risk- and time-adapted postoperative surveillance strategies.