Biomarker-Negative Cholangiocarcinoma: Durvalumab or Pembrolizumab Alongside Chemotherapy
A newly diagnosed unresectable cholangiocarcinoma with no actionable mutation, and two real, FDA-approved immunotherapy combinations with no head-to-head trial to choose between them.
Grace T., 66, spent thirty years behind the counter at the same post office branch before retiring, and now runs a food pantry out of her church basement two mornings a week — the kind of work, she's told the team more than once, she doesn't intend to interrupt for very long. She came in three weeks ago with painless jaundice, worked up to a perihilar mass encasing a segment of the hepatic artery: unresectable extrahepatic cholangiocarcinoma. A percutaneous biliary drain placed a week ago has already brought her bilirubin down from 6.4 to 1.8 mg/dL, and comprehensive genomic profiling came back clean — no FGFR2 fusion, no IDH1 mutation, no HER2 amplification, nothing to target directly. Her hypothyroidism, well-controlled for over a decade on a stable levothyroxine dose, is the only other item in her chart, and cross-sectional imaging found no evidence of nodal or distant spread beyond the local arterial encasement that makes surgical resection unsafe. She's ECOG performance status 1, functional enough to be asking the team when she can get back to the pantry, not just whether she can.
With no actionable alteration, the real decision is which biomarker-agnostic immunotherapy combination to start alongside gemcitabine-cisplatin, and the field currently offers two genuine options rather than one clear standard. TOPAZ-1 was the first phase 3 trial to show a real overall-survival benefit from adding an immune checkpoint inhibitor — durvalumab — to this chemotherapy backbone in unresectable biliary tract cancer, and it's carried the longest track record of real-world use since. KEYNOTE-966 tested the same basic addition with pembrolizumab in the same population and was also positive on overall survival. No trial has ever compared the two checkpoint inhibitors directly against each other in this disease, which leaves the team choosing between two real, FDA-approved, biomarker-agnostic options with no head-to-head data to settle it. Both trials, notably, enrolled a similar mix of intrahepatic, perihilar, and gallbladder primaries, so Grace's perihilar site doesn't itself argue for one regimen over the other — whatever tips the decision has to come from somewhere else.
Choosing the immunotherapy partner before cycle one
TOPAZ-1 was the trial that actually moved the needle for biliary tract cancer — the first phase 3 to show a real overall-survival benefit from adding immunotherapy to gemcitabine-cisplatin, and it's now the regimen with the longest track record of real-world use behind it. Start durvalumab.
Doesn't dispute that TOPAZ-1 was first or that it works. KEYNOTE-966 tested the same basic idea — pembrolizumab added to the same chemotherapy backbone — in the same population, and it was also positive on overall survival. Nothing about being second makes pembrolizumab's own trial less real, and this practice's infusion nurses have run pembrolizumab schedules across half a dozen other tumor types; that familiarity is a genuine safety margin, not a tiebreaker of convenience.
"Longest track record" is really just "came out first" restated — it isn't evidence of a larger effect.
Then look at what each trial actually reported, since neither of the first two arguments settles it. TOPAZ-1's hazard ratio for death was 0.80, confidence interval 0.66 to 0.97; KEYNOTE-966's was 0.83, 0.72 to 0.95. Those intervals sit almost on top of one another, and each is measured against its own separately-run control arm — so the honest reading isn't that durvalumab wins by 0.03. It's that the numbers refuse to discriminate, and that refusal is the answer. Which means the tiebreaker has to be a difference that actually exists, and the only one in this room is the one he was too quick to wave off: this practice has run pembrolizumab schedules for years and has never run durvalumab. Administration and scheduling error has a real rate. A hazard ratio borrowed from someone else's control arm does not.
Agreed to start gemcitabine-cisplatin-pembrolizumab — not because its trial is better, but because the pharmacologist's own look at the numbers established that neither trial's result distinguishes the two, which left administration experience as the only real difference on the table. The first oncologist accepted the reasoning while noting he'd have been equally comfortable with durvalumab.
Not agreed: whether "our nurses know this one" should be allowed to settle future drug choices as a standing rule — the first oncologist is uneasy that an operational argument won today and worries it will be reached for next time before the evidence has been looked at properly, rather than after it came back indeterminate.