Blog

As molecularly targeted and immune-based strategies continue to reshape the treatment paradigm for biliary tract cancers, recent advances in HER2-directed therapy and chemoimmunotherapy have expanded the range of effective options, according to Tanios S. Bekaii-Saab, MD.

“There’s a lot of focus on bringing those therapies that have shown activity in the more refractory setting into the first-line setting, either as single agents or in combination,” he said. “That’s how we’re going to transform the field, while continuing to dig deeper for other targets.”

In an interview with OncLive®, he outlined current treatment standards for advanced biliary tract cancers; highlighted the role of targeted therapies for this patient population; and discussed the need for biomarker-driven immunotherapy selection and other ongoing research.

Saab is a professor of medicine at Mayo Clinic College of Medicine and Science and a consultant in the Division of Hematology and Oncology at Mayo Clinic in Phoenix, Arizona.

OncLive: How would you define the current treatment landscape for biliary tract cancers in 2025?

Bekaii-Saab: Treatment in 2025 biliary tract cancers are an interesting group of cancers. It essentially refers to cancers that involve the bile ducts, and then that also includes gallbladder cancer. All these fall under this denomination.

This is a rare cancer, although if you look across the world, it’s approximately 100,000 patients [diagnosed each year]. In the United States, depending on where you look, it’s approximately 7000 to 12,000 patients [diagnosed each year] with this cancer. If you think about breast cancer, [with] 220,000-plus patients [per year], biliary tract cancer is, relatively speaking, a rare cancer.

What’s unique about this cancer is this is one of the cancers where immunotherapy has a role along with chemotherapy. This is also a very target-rich cancer. There are a number of targets that have been identified that we go after with targeted therapies, and many of them are approved; many are still under study.

Talking first about the landscape, chemotherapy—primarily gemcitabine and cisplatin—had been the backbone for the longest time in treating this cancer in the first line. Over the last few years, we’ve had PD-1 and PD-L1 inhibitors—durvalumab [Imfinzi] first, and then pembrolizumab [Keytruda] came in. When combined with gemcitabine and cisplatin, they appear to do better than gemcitabine/cisplatin alone, and so they became part of our standard of care. Chemoimmunotherapy is our standard in the first line.

[Chemoimmunotherapy] did move the needle. If you look at the 3-year overall survival [OS] for durvalumab, there’s certainly a doubling of the 3-year OS rate, but you can still see that it’s dismal, with only approximately 1 out of 5 patients surviving beyond 3 years.

How often can targeted therapy be utilized for patients with biliary tract cancer?

Approximately 30% to 40% of these cancers have a target that we can match with a proper therapy. For example, for intrahepatic cholangiocarcinoma, approximately 5% to 7% of patients have an FGFR2 fusion. Pemigatinib [Pemazyre] and futibatinib [Lytgobi] both give you a response rate close to 40% and really good long-term outcomes for some patients. Second- and third-generation FGFR inhibitors are being developed that seem to have higher response rates, but also more toxicity.

For gallbladder cancers and extrahepatic cholangiocarcinoma, HER2 amplifications are a major driver. Outside the United States, up to 30% of patients will have HER2 amplifications. We have treatments such as fam-trastuzumab deruxtecan-nxki [Enhertu] and trastuzumab [Herceptin] plus pertuzumab [Perjeta]; more recently, zanidatamab-hrii [Ziihera], which is approved specifically for [previously treated] patients with immunohistochemistry [IHC] 3+ [HER2-positive disease].

We also have rare KRAS G12C mutations, with data for adagrasib [Krazati], which is now included in the National Comprehensive Cancer Network [NCCN] guidelines. We have BRAF V600E mutations, present in approximately 5% to 10% of patients, with dabrafenib [Tafinlar] plus trametinib [Mekinist]. We have IDH1 mutations, present in approximately 20% of patients with intrahepatic cholangiocarcinoma, treatable with ivosidenib [Tibsovo]. Then there are other rare fusions, with more being discovered. Even for a rare cancer, in 30% to 40% of patients—[in addition to chemoimmunotherapy]—we have targeted agents that can induce meaningful responses.

How do you approach sequencing in patients eligible for HER2-targeted therapy?

Looking at the data globally, the key studies, at least over the last year, that have changed practice were mostly in the HER2-amplified space.

We had an approval by the FDA with zanidatamab based on the phase 2b HERIZON-BTC-01 trial [ NCT04466891]. This agent is bispecific, targeting both trastuzumab- and pertuzumab-like epitopes—and preclinically, at least, it seemed to perform better than trastuzumab [plus] pertuzumab.

We know that pertuzumab plus trastuzumab has a very modest response rate—around 20%—and can have some toxicities. With zanidatamab, we’ve seen is a response rate a little north of 40%, with many of these responses appearing to be quite durable, and even better outcomes in the IHC 3+ subgroup compared with others. This led to the approval of this agent specifically in IHC 3+ patients.

We’ve also seen data with tucatinib [Tukysa] plus trastuzumab [from a phase 2 study (NCT04579380)], again amounting to a decent response rate and ultimately placement on the NCCN guidelines.

Then there is trastuzumab deruxtecan [which has] a tumor-agnostic approval that includes biliary tract cancer. In the biliary tract cancer subgroup, the response rate was around 40%.

Overall, HER2 amplifications have become a very important target, added to the others we already consider. The question is: how do you sequence them in the clinic?

In my practice, after gemcitabine/cisplatin plus durvalumab or gemcitabine/cisplatin plus pembrolizumab, the first two options would be tucatinib plus trastuzumab or zanidatamab. Zanidatamab may be preferred in the IHC 3+ group, but for IHC 2+ and in situ hybridization–positive disease, I would go with tucatinib/trastuzumab, largely based on our experience in colorectal cancer.

That leaves trastuzumab deruxtecan, which I like to keep as a backup option [after patients progress on the] other 2 [options]. Trastuzumab deruxtecan is an antibody-drug conjugate, and it doesn’t rely on HER2 activity, per se. It binds to HER2, is internalized, and then releases deruxtecan into the tumor environment, leading to cancer cell killing.

Importantly, across studies, the response rate with trastuzumab deruxtecan does not appear to be affected by prior HER2-directed therapy, which means we can still see benefit even in pre-exposed patients. This effectively gives us two lines of HER2-directed therapy. Of course, one of the main concerns with trastuzumab deruxtecan remains interstitial lung disease, which continues to require close monitoring.

What does the future hold for biliary tract cancer research?

The key is to try to bring a lot of these targeted therapies into earlier lines of treatment. That’s being done with zanidatamab. We tried to do that with pemigatinib, but had some difficulties because of the rarity of the target. [We] also will see others, like ivosidenib, ultimately making their way in—primarily in combination with chemotherapy, with or without immunotherapy.

The goal here is to try to bring the advantages we’re seeing with [targeted] those therapies into the first-line setting. That is where I think they’ll have a much bigger benefit. We see that across a lot of different malignancies—if you hit the biology sooner, you end up with significantly better outcomes.

Now, the challenge is that these are rare cancers, and those are targets [within these rare cancers]. The question becomes: how do you design your study in a way that will allow you to test that question while also allowing patients to start treatment while you’re waiting on results? Those are designs that are being looked at in many places.

The other element is immunotherapy. With PD-1/PD-L1 inhibitors, only approximately 10% to 20% of patients seem to benefit. We don’t have a biomarker today, but a lot of biomarker discovery is focusing on how we can better select those patients who may benefit—rather than exposing 100 patients to find the 20 or 30 who will respond. Those are the patients we would want to target, and for the rest, we need to identify different pathways.

I think we’re pretty much maxed out on chemotherapy options beyond [gemcitabine/cisplatin]. All other chemotherapy agents have a very modest role, and I don’t think our goal would be to continue developing along the cytotoxic pathway, except with antibody-drug conjugates, where there may be other targets of interest, such as MET, that could be developed in this pathway.

The job market has begun looking shakier. How much is artificial intelligence to blame?

Not a whole lot. At least not yet.

A review of employment surveys, interviews with labor market analysts and recent company earnings reports shows little evidence, so far, that would support assertions of a widespread economic impact from AI’s growing usage.

“It’s such an emotional thing for people, many of whom are determined to see it in the data,” said Martha Gimbel, executive director and co-founder of the Budget Lab at Yale University and a former President Joe Biden economic adviser. “And it’s just not there yet.”

Much is riding on the payoff from AI. The stock market has been hitting record highs largely thanks to gains from tech giants like Nvidia, Google parent Alphabet, Facebook parent Meta and Microsoft, which have made enormous investments in pumping out AI-related products.

For precisely that reason, analysts say, some businesses may be incentivized to hype AI’s potential as a disruptive force. Through the end of July, the term “AI” has been cited on about two-thirds of second-quarter earnings calls conducted by S&P 500 companies, according to the data provider Factset. That’s up from less than half in the first quarter.

Amid a downshifting economy, cost pressures are mounting, prompting corporate leaders to hype AI’s potential as a savings source — even if it’s not quite there yet.

“In 2023, you’d have a high-profile public company do a job cut and cite rising interest rates or uncertain macro conditions,” Roger Lee, a tech entrepreneur who also runs a website that tracks tech industry layoffs, said. ”Today, it’s AI.”

The most extreme warning about AI’s short-term impact has come from Dario Amodei, co-founder and CEO of AI firm Anthropic. In May, he told Axios that he foresees half of all entry-level white-collar jobs being wiped out in the next one to five years, spiking unemployment to between 10% and 20%.

So far, evidence for this scenario is mixed. All job openings, entry-level or otherwise, have been declining since 2023, according to labor market analytics company Revelio Labs, though the trend has not been linear. Revelio said entry-level jobs exposed to AI have been declining fastest — but senior roles exposed to AI have actually begun to recover.

The broader picture for white-collar professions most at risk of disruption actually indicates fairly stable employment trends. Last week’s official jobs report showed office and administrative roles have actually returned to their pandemic-era highs, while employment in other professional sectors, like accounting and legal services, has held relatively steady.

It’s a gloomier story in tech — but also a more nuanced one when it comes to AI’s impact. The leaders of Amazon and Microsoft have both signaled the ability to run their businesses with reduced headcount thanks to AI. Tech layoffs tracked by Lee’s website hit a three-month high in July, with three companies — Intel, Microsoft and Recruit Holdings, the parent of Indeed and Glassdoor — largely responsible.

All three of those companies cited artificial intelligence as playing a role in the job reductions, Lee said. But he noted that in the case of Recruit Holdings, there were no specifics about how AI had impacted the lost positions. The company simply said the technology was “changing the world.”

“It does seem like many of the roles being cut are in line with ones being used by AI,” Lee said. “But it’s still being used as a cover in other cases.”

A representative for Recruit did not respond to a request for comment.

The simple calculus behind AI is that businesses will be able to do more with less, increasing overall productivity while reducing hiring needs. Yet economists say it is difficult to calculate accurate changes in productivity over the short term — though so far, the broadest national measure has shown a deceleration in recent quarters.

Most of the benefits of AI are instead accruing to consumers, not businesses, according to a forthcoming paper from researchers at Carnegie Mellon and Stanford University. If it feels like much of the value from the current generation of AI seems mostly to allow ordinary people to generate emails and papers faster, or do quicker research, you’re not imagining things.

“Free goods are invisible in the GDP numbers, even if they make consumers better off,” the authors, Avinash Collis and Erik Brynjolfsson, wrote in a recent Wall Street Journal op-ed. They calculate consumers derived the equivalent of $97 billion in surplus welfare from generative AI in 2024, compared with $7 billion in revenues logged by the tech firms actually creating AI products.

Economies typically see a “J-curve” effect when transformative technologies are introduced, Collis told NBC News. At first there is a bottleneck that can cause some disruptions, though these initial effects are often not captured in official figures. For example, the iPhone increased the total global volume of photos from billions to trillions, something that directly impacted workers at camera giant Kodak, but created incalculable opportunities elsewhere, Collis said.

“There will likely be a lot of impact, perhaps on some sectors negatively,” Collis said. “But at the same time lots of new jobs could be created as well.”

Other indicators do suggest the stirrings of a more pronounced AI effect on jobs. The July employment survey from consultancy Challenger, Gray and Christmas found companies have blamed “automation and AI implementation” for 20,000 job cuts in 2025, with another 10,000 or so directly attributable to artificial intelligence. Challenger said this shows “a significant acceleration in AI-related restructuring.”

Those figures are dwarfed by cuts related to government spending declines and general economic and market conditions, which account for nearly 500,000 lost roles this year, Challenger said.

Some companies appear to be keeping payroll counts steady in response to the broad uncertainty in the economy, and using any additional resources to explore AI’s potential to boost their bottom lines. Stacy Spikes, CEO of MoviePass, told NBC News that internal workflows at his company become vastly more efficient thanks to AI. That’s made him more gun-shy about bringing on new workers into certain departments, like software. As of Tuesday, MoviePass’ careers page showed no open positions.

“We haven’t seen headcount need to increase,” Spikes said.

Businesses like MoviePass still appear to be the exception, however. Analysts at Goldman Sachs say only about 9% of all companies are regularly using new AI tools to produce goods or services. As a result, they see only limited effects at the moment.

“When I look at the impact that AI has had on the overall labor market data so far, it looks pretty small to me,” Joseph Briggs, head of the global economics team at Goldman Sachs Research, said on a recent company podcast. Even for recent college grads, who have seen unemployment rates tick higher, “the anecdotes and the relationship that the anecdotes have to AI is often a little bit overstated,” Briggs said.

JP Morgan analysts came to a similar conclusion, finding that, for now, its research “failed to find a significant impact on job growth.”

But they cautioned that this could change at the next economic downturn.

For white-collar workers, “we think that during the course of the next recession the speed and the breadth of the adoption of the AI tools and applications in the workplace might induce large scale displacement for occupations,” they said in a recent note to clients.

Others remain more optimistic about the potential for new opportunities to overcome any negative effects. That’s how Nvidia co-founder and CEO Jensen Huang sees it. As the head of an AI giant, he may also have reason to hype its potential — but his outlook is notably rosier than Anthropic’s Amodei’s. Huang told Axios last month that the technology would ultimately lead to more jobs, even if there are some redundancies elsewhere.

“Everyone’s jobs will change,” he said. “Some jobs will be unnecessary. Some people will lose jobs. But many new jobs will be created. … The world will be more productive.”