Rethinking Early Recurrence in Gallbladder Cancer: Why One Cutoff Does Not Fit All Stages

Early recurrence (ER) after curative-intent resection for gallbladder carcinoma (GBC) is clinically important but lacks a consistent definition. Fixed 6- or 12-month cutoffs are commonly used, however, recurrence timing is strongly stage dependent, which can systematically misclassify ER and compromise cross-study comparability. We conducted a retrospective cohort study of patients undergoing curative-intent resection for GBC with complete recurrence and pathology data. Among recurred patients, ER was defined within each AJCC stage using the prespecified stage-specific median time to recurrence (mTTR) (ER versus LR: TTR≤mTTR vs TTR>mTTR). The primary outcome was survival after recurrence (SAR), evaluated within stage. Sensitivity analyses included bootstrap assessment of stage-specific cut-point stability, RMST contrasts at prespecified horizons, and time-varying effects to address non-proportional hazards. Of 537 screened patients, 507 met eligibility criteria. AJCC stage distribution was I (n=42), IIA (n=55), IIB (n=108), IIIA (n=131), IIIB (n=87), and IVA (n=84). Recurrence occurred in 320/507 (63.1%) patients. In recurred patients, mTTR decreased monotonically with stage (30.4 months in stage I; 14.2, 11.3, 10.4, 7.9, and 6.5 months in stages IIA, IIB, IIIA, IIIB, and IVA), demonstrating fundamentally distinct stage-specific recurrence kinetics. Fixed calendar cutoffs yielded marked stage heterogeneity in the fraction labeled “early”: for TTR≤12 months, 0.0%, 17.4%, 61.8%, 95.1%, 98.6%, and 100.0% across stages I–IVA; for TTR≤6 months, 0.0%, 0.0%, 3.6%, 4.9%, 12.5%, and 34.7%. Under the stage-adapted mTTR definition, ER (vs LR) was associated with significantly shorter SAR in stages IIA (P=0.00519), IIB (P=0.0149), IIIA (P=0.00243), and IIIB (P<0.001), but no in stage I (P=0.208) or IVA (P=0.101). ER (vs LR) was also enriched for adverse pathologic features within stage, most prominently in stage III-IVA disease, including higher perineural invasion (stage IIIA: 52% vs 15%; P<0.001; IIIB: 50% vs 25%; P=0.028; IVA: 90% vs 33%; P<0.001), vascular invasion (IIIA: 52% vs 13%; P<0.001; IIIB: 67% vs 19%; P<0.001; IVA: 95% vs 31%; P<0.001), and tumor necrosis (IIIA: 62% vs 38%; P=0.035; IIIB: 97% vs 36%; P<0.001; IVA: 85% vs 47%; P<0.001). ER in GBC is not a fixed calendar time but a stage-dependent, time-varying phenotype. A stage-adapted ER framework based on within-stage TTR improves interpretability and identifies a high-risk subgroup with shorter SAR and enrichment of adverse pathology. External validation is warranted to confirm transportability across treatment eras and surveillance patterns.