Real-world survival benefit of adding durvalumab to chemotherapy in cholangiocarcinoma: a retrospective cohort analysis from a German Comprehensive Cancer Centre

Background Advanced biliary tract cancers (BTCs) carry a poor prognosis. Gemcitabine/platinum chemotherapy was the first-line standard until TOPAZ-1 and KEYNOTE-966 established immune checkpoint inhibitor-based combinations as a new standard. Real-world data with a chemotherapy-only comparator remain limited. Patients and methods We retrospectively analyzed 172 patients with advanced BTC treated at the West German Cancer Centre (2018-2024): a durvalumab cohort ( n = 64, treated 2022-2024) and a historical comparator cohort ( n = 108, chemotherapy alone). Because cohort assignment was based on treatment era rather than randomization, propensity score matching (incorporating age, sex, metastatic burden, CA19-9, tumor location, and chemotherapy regimen) was carried out; analyses are reported for both the unmatched ( n = 172) and matched ( n = 77) cohorts. The primary endpoint was median overall survival (mOS); secondary endpoints included objective response rate (ORR), disease control rate (DCR), time to treatment failure (TTF), and toxicity. Results In the matched cohort ( n = 77), durvalumab was associated with significantly longer mOS (14.6 versus 11.1 months; hazard ratio 0.54, 95% confidence interval 0.307-0.956, P = 0.032) and a higher ORR (45% versus 20%; P = 0.015); DCR was numerically higher (65% versus 47.5%; P = 0.083). In the unmatched cohort ( n = 172), durvalumab was associated with a larger mOS difference (18.7 versus 12.8 months; P = 0.003), higher ORR (42.2% versus 24.1%; P = 0.018) and DCR (65.6% versus 45.4%; P = 0.011); TTF trended in favor of durvalumab (4.8 versus 3.4 months; P = 0.066). No new safety signals were observed. Conclusions These real-world data support the clinical benefit of adding durvalumab to gemcitabine/platinum chemotherapy in patients with advanced BTC, consistent with TOPAZ-1 and KEYNOTE-966. Given the retrospective design, findings should be regarded as supportive of, rather than exceeding, pivotal trial data.