Adjuvant Imatinib in High-Risk Gastrointestinal Stromal Tumors: Time-Dependent Treatment Exposure, Genotype, and Recurrence-Free Survival in a Real-World Cohort
Hao Wang, Wei Zhao, Qi Zhang, Lei Shi, Aoqing Wang, Linsen ZhouBackground
The relationship between adjuvant imatinib exposure, treatment duration, tumor genotype, and recurrence-free survival (RFS) in high-risk gastrointestinal stromal tumors (GISTs) remains incompletely defined, particularly in patients who are still receiving treatment or who relapse during treatment.
Methods
We retrospectively analyzed 259 consecutive patients with high-risk primary GIST who underwent R0 resection at two tertiary centers between 2008 and 2025. Follow-up was updated through June 30, 2026. Adjuvant imatinib exposure was modeled as a time-dependent cumulative-duration variable in a start-stop Cox model. The prespecified duration categories were no exposure or not-yet-started, <36 months, 36 to <60 months, and ≥60 months. A 36-month landmark analysis and a treated-only post-discontinuation RFS (PD-RFS) analysis were performed as sensitivity and conditional analyses, respectively.
Results
The cohort included 69 untreated and 190 treated patients; 91 patients were still receiving adjuvant imatinib at the cutoff. There were 101 RFS events and 158 censored observations. In the adjusted time-dependent Cox model, compared with no exposure or not-yet-started person-time, the hazard ratios were 0.399 (95% CI 0.252–0.633) for <36 months, 0.080 (95% CI 0.027–0.233) for 36 to <60 months, and 0.075 (95% CI 0.014–0.392) for ≥60 months. The global treatment-by-genotype interaction was not significant (P=0.581). Among 161 patients eligible for the 36-month landmark analysis, continued treatment at the landmark was associated with lower subsequent RFS hazard than no prior treatment (HR 0.376, 95% CI 0.182–0.775; P=0.008). In the 91-patient PD-RFS analysis, genotype remained associated with post-discontinuation outcomes.
Conclusion
In this retrospective cohort, cumulative adjuvant imatinib exposure modeled over time was associated with RFS. The findings support time-dependent and landmark analyses for treatment-duration questions, while genotype-specific effects should be interpreted as associations and exploratory subgroup findings rather than causal or definitive treatment-selection rules.