Real-world outcomes of adjuvant therapy after curative-intent resection for biliary tract cancer.

BACKGROUND

The role of adjuvant therapy after biliary tract cancer (BTC) resection remains debated, particularly in real-world practice. This study examined determinants of adjuvant therapy receipt and its association with survival.

METHODS

Patients undergoing curative-intent resection for intrahepatic (iCCA), perihilar (pCCA), and distal cholangiocarcinoma (dCCA), and gallbladder cancer (GBC) at Mayo Clinic Rochester (2000-2024) were retrospectively analyzed. OS and RFS were assessed by Kaplan-Meier analysis. Adjusted analyses excluded 90-day mortality and included multivariable Cox, time-dependent Cox, and propensity-score-matched sensitivity analyses.

RESULTS

Among 770 patients (iCCA, n = 343; pCCA, n = 205; GBC, n = 135; dCCA, n = 87), 341 (44.3%) received adjuvant therapy. Non-receipt was independently associated with older age, longer hospitalization, and surgery-related major complications. After excluding 90-day mortality, adjuvant therapy was not associated with improved OS in the overall cohort (p = 0.978), while subtype-specific effects were heterogeneous. Among high-risk patients (N + , R1, or TNM III-IV), adjuvant therapy was associated with longer OS (37.3 vs 30.1 months; p = 0.015). After harmonization of adjusted analyses to 90-day survivors, capecitabine remained independently associated with improved OS, whereas the gemcitabine-based estimate was attenuated to a non-significant trend. High-risk features remained strongly prognostic, and time-dependent Cox models yielded concordant estimates.

CONCLUSION

Major postoperative morbidity was a key determinant of adjuvant therapy non-receipt. After excluding 90-day mortality and harmonizing adjusted analyses, adjuvant therapy was not associated with improved survival in the overall cohort, including in propensity-score-matched analysis. Survival associations were most consistent in risk-enriched subgroups but remain exploratory and require prospective randomized validation.