Bile duct cancer: real-world data suggest the durvalumab combo helps patients the TOPAZ-1 trial excluded

About a third of patients with advanced biliary tract cancer, the family of cancers that includes bile duct and gallbladder tumors, would not have qualified for the TOPAZ-1 trial that won durvalumab its FDA approval in this disease. A new 1,358-patient international cohort published in 2026 finds their survival on the same chemotherapy-plus-immunotherapy regimen was statistically indistinguishable from the patients the original phase III trial actually treated. That result reframes a set of eligibility rules many oncologists already saw as too narrow.

The study, published in the International Journal of Cancer, pulled together 1,358 patients with advanced biliary tract cancer who got first-line cisplatin, gemcitabine, and durvalumab (an immunotherapy drug sold as Imfinzi) across 55 centers. Researchers then scored each patient against the seven exclusion criteria TOPAZ-1 had used in 2022: 912 (67.1%) cleared every bar, 446 (32.9%) failed at least one. Median follow-up was 14.5 months.

Patients who fit TOPAZ-1 had a median overall survival of 16.1 months; those who would have been excluded lived 12.5 months (hazard ratio 0.69, p=0.0001). Progression-free survival, the time before the disease visibly worsened or the patient died, followed the same pattern: 8.2 months for the in-group, 6.5 months for the out-group (HR 0.73, p<0.0001). The 3.6-month overall survival gap is real, and the report does not pretend otherwise.

What the new data do argue is that the gap is not the drug's fault. Cross-trial comparison is the only honest move here, and the authors run it: the TOPAZ-1-out group's 12.5-month median overall survival sits within a hair of the 12.9 months the original phase III experimental arm reported (HR 1.15, p=0.13). Progression-free survival in the out-group lined up similarly against the trial arm (HR 1.12, 95% CI 0.93–1.42, p=0.09). Patients the trial would have kept out did about as well as patients the trial actually enrolled.

Inside the out-group, the seven TOPAZ-1 exclusion criteria split into two camps. Three were tied to worse overall survival: active infection, elevated bilirubin (a liver-function marker often raised in bile-duct obstruction), and ECOG performance status above 1 (a standard oncology yardstick for how sick a patient is in daily life). Five other criteria, including abnormal liver enzymes, corticosteroid use, abnormal kidney or blood counts, and prior surgery within six months, did not predict a worse outcome on the regimen. Adverse events were not significantly higher in the out-group than in the in-group.

That is the mechanism readers should leave with. The TOPAZ-1 exclusion list, written in 2018 and designed to keep the original trial clean, treats most of these seven rules as interchangeable risk markers. Real-world practice does not. Active infection, organ failure, and a patient too frail to get out of bed most of the day are not the same as an abnormal blood test or a recent surgery.

The cohort is retrospective, meaning oncologists chose to treat each patient with the regimen and the researchers looked back at who did and did not qualify. Cross-trial comparison is non-randomized, and the survival curves for the out-group are reconstructed rather than observed in a controlled arm. One number in the source string, the 95% CI "14.6–16.5" attached to the overall-survival hazard ratio in the abstract, looks like it has been mis-pasted from the median-survival range, and the version that lands in print should not carry it. The author conclusion is "appears effective and tolerable in TOPAZ-1-excluded patients, supporting expanded clinical use," not "should replace eligibility judgment."

For payers, tumor boards, and oncologists, the practical question is whether eligibility lists written for clean phase III data should still be used to gate access to the same regimen now that real-world numbers exist. The 2026 cohort does not answer that question. By design, it cannot. But it makes the answer harder to defer.