01 | PD-L1 Negative. Still Responds: What Are We Missing?

In CheckMate 067, a first-line trial in advanced melanoma, 41.3% of patients classified as PD-L1 negative who received nivolumab alone achieved an objective response. The trial used a 5% tumour-cell staining threshold to define PD-L1 positivity. That was the observation. What the result did not explain was why those patients responded.

I came to this question as a patient, not a scientist. Having experienced Stage IV cholangiocarcinoma and a complete response to immunotherapy, I wanted to understand what had made that response possible. More importantly, I wanted to understand what the science could reveal for patients who were still searching for a treatment opportunity.

The number that stopped me was 41.3%.

These were not patients whose tumours had tested positive for the protein PD-L1. They were patients classified as negative under the trial's testing threshold. Yet roughly four in ten responded to a medicine that blocks the PD-1 immune checkpoint.

How could that happen?

To understand the contradiction, we first need to enter the world in which it occurred.

Inside the Tumour

Visualise a tumour as a growing collection of cells. Each cell is its own tiny City, carrying instructions and operating systems that determine how it functions. Around these cancer cells are blood vessels, supporting cells and immune cells, all interacting within the same environment.

Among those immune cells are T cells, capable of recognising and attacking abnormal cells. But T cells also carry control systems that prevent an immune attack from damaging healthy tissue unnecessarily.

One of those controls is a receptor called PD1. Think of it as a brake receiver on the T cell.

All cells including cancer cells have a corresponding PD-L1 ligand, a protein that can connect to the T Cells receiver. When PD-L1 binds to PD1, it delivers an inhibitory signal that tells the T cell's to reduce its activity.

In our Cell City picture, imagine PD-L1 displaying a message that says, "Maintenance in progress. Stand down."

Healthy cells use immune-regulatory signals for protection. Cancer cells like any cell can still use the same system to weaken an attack against them.

Nivolumab (Nivo) and pembrolizumab (Keytruda) work differently from chemotherapy. They do not directly attack the cancer cell. They bind to PD-1, preventing inhibitory signals from PD-L1 and another protein, PD-L2, from engaging and applying that receptor brake.

If the T cell has already recognised the cancer, removing that restraint allows its attack to continue or recover.

The Contradiction

Now return to CheckMate 067.

A tumour was classified as PD-L1 positive when at least 5% of the assessed tumour cells met the staining threshold. Below that threshold, the tumour was classified as negative.

But negative did not necessarily mean that every cancer cell lacked PD-L1. Importantantly, nor did it measure PD-L2, or reveal every immune interaction occurring elsewhere in the tumour.

Most importantly, it did not tell investigators how those signals might change after a PD1 inhibitor treatment removed the brake.

And yet the responses were there.

This was melanoma, not cholangiocarcinoma. Its response rate cannot be transferred to patients with bile duct cancer.

But the question exposed by the result travels beyond melanoma.

Was the PD-L1 defence mechanism truly absent in those patients measure as PD-L1 negative, or had the original test captured only part of the biological picture?

That became the investigation.

The question a study sets out to answer may not be the most important question its results uncover.

02 | How Allison and Honjo Changed Cancer Treatment Forever

James Allison and Tasuku Honjo uncovered different controls that restrain T-cell activity. Their discoveries established that cancer treatment could work by releasing the immune system's brakes, rather than directly attacking cancer cells.

The CheckMate 067 result could not have occurred without a change in how scientists understood the immune system.

For decades, cancer treatment had largely been organised around attacking cancer cells directly. Surgery removed them. Radiation damaged them. Chemotherapy interfered with their ability to survive and divide.

But two scientists were investigating a different possibility. What if the immune system already possessed the ability to attack cancer, but something was restraining it?

Two Scientists. Two Different Brakes.

James Allison studied CTLA-4, a protein that helps regulate T-cell activation.

Tasuku Honjo's laboratory identified PD-1, another receptor involved in controlling immune-cell activity.

They were not studying two names for the same mechanism.

CTLA-4 acts principally during the processes that activate T cells. PD1 can restrain T cells after activation, including when they encounter cancer within a tumour.

Think of an immune response as a journey. CTLA-4 can restrict the T cells from being activated. PD1, the brake, can slow its activity after it has been activated to seek and destroy.

Both controls are necessary. An immune system without restraints could damage the very person it exists to protect. As is the case in auto immune diseases.

But cancer can like any cell can also use those restraints.

Allison's work helped establish that blocking CTLA-4 could strengthen T Cell activation and anti-tumour immunity. This discovery contributed to the development of ipilimumab (Ipi).

Research following Honjo's PD1 discovery established the therapeutic potential of blocking PD1. That work contributed to medicines including nivolumab and pembrolizumab.

CTLA4, PD1 and PD-L1 along with PD-L2 another cell inhibitory signaling protien became known as the "Checkpoint Pathway"

In 2018, Allison and Honjo shared the Nobel Prize in Physiology or Medicine for discoveries that transformed cancer immunotherapy.

What They Had Actually Changed

Their discoveries introduced a different treatment strategy.

Instead of directing a medicine at the cancer cell itself, treatment could remove a restraint on the patient's own immune system.

But that distinction contains a limitation.

Releasing a brake does not create the ability to recognise cancer. The T cell must still have the biological capacity to identify and attack its target.

Think of a car with its brakes applied. Releasing the brakes permits movement. It does not supply the engine, fuel or direction.

The same principle applies here.

A PD1 inhibitor may release an immune restraint, but it cannot guarantee an effective anti-cancer response.

That brought me back to the patients classified as PD-L1 negative.

If their tumours appeared to lack the expected signal, what restraint was nivolumab releasing? Would there be any benefit.

But the results showed there was benefit measured, and it was not minor.

Perhaps the answer was not that the treatment behaved unexpectedly. Perhaps our understanding of what the test revealed was incomplete.

Releasing the brake creates an opportunity. Understanding when and why the immune system can use that opportunity is the next question.

03 | Was CheckMate 067 an Isolated Result?

CheckMate 067 was not the only melanoma trial reporting responses among patients classified as PD-L1 negative. Other studies found responses with nivolumab, pembrolizumab and checkpoint combinations. The pattern confirmed that a negative baseline classification did not rule out response. It did not establish why those responses occurred.

One unexpected result can be a statistical accident, an unusual patient population or a finding that disappears under closer examination.

So I went looking for the same observation in other trials.

CheckMate 066 provided another piece. Among nivolumab-treated patients whose PD-L1 status was negative or indeterminate, the reported objective response rate was 33.1%.

The classification was not identical to CheckMate 067, but the observation was important. Patients without a positive baseline PD-L1 classification were responding.

Then came KEYNOTE-006, investigating pembrolizumab against ipilimumab. In the reported PD-L1-negative subgroup, approximately 25% responded to pembrolizumab, compared with 13% receiving ipilimumab.

Now the finding extended beyond one drug. Two different medicines blocking PD1 had produced responses among patients classified as PD-L1 negative.

Then the Combination Trials

In CheckMate 067, nivolumab combined with ipilimumab produced an objective response in 54.8% of the PD-L1-negative subgroup.

CheckMate 069 also reported substantial responses among PD-L1-negative patients receiving the combination.

These results did not isolate the contribution of PD-1 blockade from CTLA-4 blockade. The combination changes more than one immune control.

But the evidence was accumulating. Negative baseline PD-L1 classification did not make a response impossible.

Response Is Not the Same as Cure

The language of clinical trials matters.

An objective response includes both partial and complete responses.

A partial response generally requires at least a 30% reduction in the measured target lesions under RECIST 1.1.

A complete response means that the disease meets the trial's criteria for disappearance of detectable cancer on assessment. It does not prove that every malignant cell has been eliminated.

Duration is a separate question. A response recorded at one assessment does not establish how long it lasted.

KEYNOTE-029 added another observation. In its reported PD-L1-negative subgroup, 50.0% achieved an objective response and 33.3% achieved a complete response.

These percentages came from a particular subgroup and treatment combination. The subgroup size, assessment method and follow-up must be considered. They are not general predictions for all PD-L1-negative patients.

What the Pattern Established

Stand back from the individual trials.

Patients classified as PD-L1 negative responded to PD1 inhibitors. Some responded to combinations involving PD1 and CTLA-4 blockade. Some achieved complete responses.

These were different studies with different populations, treatment regimens and testing methods. Their percentages cannot be treated as a head-to-head comparison.

But the central finding survived the comparison.

A negative baseline PD-L1 classification did not establish that the immune system was incapable of responding to checkpoint treatment. In fact it had provided strong an consistent results to the opposite.

What the trials could not tell us was what the original tests had missed, or which biological mechanisms had made the responses possible.

To investigate that, we needed to examine the test itself.

The trials confirmed the responses. They did not establish which biological mechanisms made those responses possible.

04 | What Does a PD-L1-Negative Biopsy Actually Prove?

A PD-L1-negative biopsy reports the staining observed in the sampled tissue under a particular testing method and threshold. It does not measure every region of the tumour, every relevant cell or the biological changes that may occur before and after the sample was taken.

The original melanoma trials had produced an uncomfortable question.

Patients were responding despite a test classification that might have suggested less opportunity for benefit.

But perhaps the problem was not the result. Perhaps it was what we were asking the result to prove.

A Biopsy Is a Sample, Not the Whole Tumour

Imagine a large city with many neighbourhoods. You take a photograph of one street and use it to describe what is happening across the entire city.

The photograph may be perfectly accurate. But it cannot reveal every street.

A tumour biopsy works under a similar limitation. It samples a particular location at a particular time. The laboratory then examines that sample using a defined staining and scoring method.

Some tests score PD-L1 expression on tumour cells. Others include relevant immune cells. Different studies can use different assays, thresholds and scoring systems.

Therefore, the word negative cannot be interpreted independently of the test that produced it.

In CheckMate 067, a tumour expressing below the 5% tumour-cell staining threshold was classified as PD-L1 negative.

Below 5% does not mean zero. And zero expression in the assessed sample would still not prove that PD-L1 was absent throughout the entire tumour.

What Else Was Outside the Photograph?

PD-L1 may be expressed by cancer cells and by cells within the surrounding immune environment.

Its distribution can vary between different regions of a tumour.

Another protein, PD-L2, can also bind to PD-1 and deliver an inhibitory signal.

A PD-L1 staining result does not measure that second ligand.

So a negative result may be accurate without describing the complete checkpoint environment.

And there is a further limitation. The sample is taken at one moment. It cannot continuously observe what happens as immune cells interact with the tumour.

The biopsy is a photograph. The immune response is a movie.

What Should the Patient Ask?

The immediate question is not whether the laboratory was wrong.

It is whether the result is being interpreted within the limits of what was actually measured.

A patient can ask which threshold was used, what TPS score was recorded for the PD-L1 status.

Those answers may determine whether the result has any practical bearing on treatment eligibility in that setting.

But one question remained.

What if the biology itself changes after the photograph is taken?

A biopsy can accurately report what it sees without revealing everything the tumour can do.

05 | What Happens When PD-L1 Changes During an Immune Response?

PD-L1 expression is not necessarily fixed. Activated immune T cells can release interferon-gamma, which can increase PD-L1 expression in surrounding cells. A tumour classified as PD-L1 negative at baseline may therefore behave differently when in the presence of a T cell attack.

The normal biopsy investigation reveals a limitation in what we can see.

The next discovery concerned something more fundamental: the tumour environment itself can change.

Scientists investigating immune responses observed a relationship that initially seems contradictory.

Where immune cells were attacking tumours, PD-L1 expression could increase.

In other words, evidence of an immune attack could be accompanied by the appearance of a signal that restrained that same attack.

How the Defence Is Activated

When an activated T cell recognises its target, it can release a signalling molecule called interferon-gamma (IFN-gamma).

Think of interferon-gamma as an alert moving through the tumour's Cell Cities.

Cells receiving that alert can respond by changing which proteins they produce. In responsive cells, one of those changes can be increased PD-L1.

The cancer may therefore strengthen an inhibitory defence in response to immune activity by producing and expressing more PD-L1 on its surface.

This is known as adaptive immune resistance.

The sequence is important.

The T cell recognises cancer. The T cell releases interferon-gamma. Cells increase PD-L1. PD-L1 can then engage PD1 applying the brake to restrain the T cell.

The attack can help trigger the very mechanism that slows the attack.

Why the Timing of the Biopsy Matters

Now consider a tumour sampled before a strong immune response has developed or is in a period of exhaustion.

Its PD-L1 expression may be low. Under the assay's threshold, the sample is classified as negative.

But that classification does not establish whether the same cells could increase PD-L1 expression after receiving and responding to interferon-gamma.

The original test measured expression. It did not directly test inducibility.

This offers one possible explanation for some responses among patients classified as PD-L1 negative.

If a tumour retains the capacity to increase PD-L1 during an immune attack, blocking PD1 could prevent that induced PD-L1 from connecting and applying the brakes to restrain the T cell through PD-1.

But possibility is not proof.

We cannot conclude that dynamic PD-L1 induction explains the CheckMate 067 responses without directly investigating the patients and tumour samples.

What Has Actually Been Observed?

Research by Taube and colleagues linked inflammatory immune activity with PD-L1 expression in melanoma.

Spranger and colleagues further investigated how CD8 T cells could drive adaptive changes within the tumour environment.

Vilain and colleagues examined paired tumour biopsies before and during PD1 treatment, documenting changes in immune infiltration and PD-L1 expression.

These studies support the biological mechanism.

They do not establish that every PD-L1-negative responder followed the same pathway.

For today's patient, this creates a practical distinction.

A negative baseline test may describe the sampled tumour accurately while leaving its later biological behaviour unresolved.

But the evidence supporting treatment must still come from the patient's cancer type, biomarkers, clinical setting and available trials.

And that brings us to cholangiocarcinoma.

What a tumour expresses today in one static bioply sample does not establish what it can express tomorrow.

06 | The Immune-Cold Cholangiocarcinoma Question

Cholangiocarcinoma frequently has an immune-suppressive tumour environment with limited effective T-cell activity. But immune-cold describes the conditions observed in the tumour. It does not prove that those conditions are permanent or that every patient has identical immune biology.

The melanoma trials had revealed the mystery. The biopsy studies had exposed the limits of a baseline measurement. The interferon-gamma research had shown that PD-L1 could change.

But my investigation was never about melanoma alone.

I wanted to know what this meant for cholangiocarcinoma.

And here the biological environment presents a different challenge.

What Does Immune-Cold Mean?

Think again of the tumour as a collection of Cell Cities.

An effective immune attack requires T cells that can recognise cancer, enter the tumour and remain capable of attacking their targets.

In many cholangiocarcinomas, those conditions are difficult to achieve.

The tumour may become surrounded by supporting tissue called stroma. Think of it like scar tissue forming around a wound to help protect it. But in this case, the tissue forms around the cancer cells, making it harder for attacking T cells to get through. The tumour may also contain immune cells that display PD-L1 and PD-L2, proteins that tell T cells to slow down or stop, weakening their ability to attack the cancer.

Researchers often describe such environments as immune-cold.

It is a useful description, but not a complete biological verdict.

The tumour may contain thick supporting tissue called stroma. Think of it like dense scar tissue forming a barrier around and between cancer cells, making it harder for attacking T cells to get through. Even those that reach the cancer may face difficult conditions, including too little oxygen and nourishment, and signals from other cells telling them to slow down or stop. These conditions can weaken their ability to attack and contribute to T-cell exhaustion, where repeated attempts to fight gradually leave them less effective.

These are different problems, and blocking PD1 may not overcome all of them.

What Does PD-L1 Testing Show in CCA?

In a study by Mody and colleagues examining 652 biliary tract cancers, PD-L1 positivity was reported in 8.6% overall under the study's particular assay and threshold.

The reported proportions were 7.3% in intrahepatic and 5.2% in extrahepatic cholangiocarcinoma.

These figures cannot be treated as universal PD-L1 prevalence. Different assays and scoring thresholds can produce different classifications.

But they reinforce why a negative result is a common clinical issue in this disease.

Can Cholangiocarcinoma Increase PD-L1?

This was the next question I wanted answered.

If PD-L1 can be induced by interferon-gamma in melanoma, can the same biological response occur in cholangiocarcinoma?

Research involving human intrahepatic cholangiocarcinoma cell lines has demonstrated increased PD-L1 expression after exposure to interferon-gamma.

That establishes that inducible PD-L1 biology is possible in cholangiocarcinoma cells.

It does not establish that this mechanism explains responses to PD-1 treatment in patients.

Nor does it establish that increasing PD-L1 will make an immune-cold tumour respond.

The T cells must still be present, recognise the cancer and retain the capacity to attack it.

The Question of Timing

There is another issue patients with advanced cancer understand particularly well.

Time changes the options.

A surgical opportunity may disappear as disease progresses. A patient's condition may deteriorate until a treatment can no longer be tolerated.

Could the opportunity for an effective immune response also change with time?

Biologically, that is a reasonable question. T-cell access, activation, exhaustion and tumour resistance can change.

But we do not yet have evidence establishing a specific PD1 treatment window in cholangiocarcinoma or showing that earlier PD1 monotherapy improves outcomes outside established treatment indications.

This uncertainty should drive research, not premature prescribing.

The patient's Earlier Right Action is to challenge the exclusion of PD1 inhibitors such as nivolumab (Nivo) or pembrolizumab (Keytruda) based solely on low or negative PD-L1 expression. Clinical trials have demonstrated that patients classified as PD-L1 negative have achieved significant response rates. Ask the treating oncologist to review these trial results and support access to PD1 treatment where clinically appropriate, based on the full clinical evidence, not the PD-L1 result alone.

And that brings us to a question patients are already asking.

What happens when they progress on Durva?

Immune-cold describes the conditions we find. It does not prove those conditions can never change.

07 | Has Durva Really Closed the Door on Keytruda and Nivo?

Durvalumab blocks PD-L1. Pembrolizumab and nivolumab block PD-1, preventing inhibitory signalling from both PD-L1 and PD-L2 through that receptor. Their molecular targets are different. Whether that difference improves outcomes after durvalumab progression remains unproven and unanswered in cholangiocarcinoma.

This is where the investigation moves from the laboratory to a conversation I regularly encounter with patients.

A patient with advanced cholangiocarcinoma has received gemcitabine, cisplatin and durvalumab (Durva).

The cancer progresses while on Durva.

The patient asks whether Keytruda or Nivo might offer another opportunity.

Most patients encounter a roadblock. They are told, "You have already tried immunotherapy. They are the same."

But are they the same? Durvalumab (Durva) blocks PD-L1, while pembrolizumab (Keytruda) and nivolumab (Nivo) block PD1. They act on different parts of the immune checkpoint pathway.

That answer may reflect sound clinical reasoning about the evidence and likely benefit.

But the treatment label alone does not explain the biology.

Have they actually tried the same molecular treatment?

Return to the T Cell

We established that PD1 is an inhibitory receptor carried by T cells.

We also know there are two ligands that can signal, PD-L1 and PD-L2, can both connect to PD1, applying its brake to reduce the T-cell attack.

Durvalumab (Durva) binds to PD-L1, preventing it from connecting with the PD1 receiver on a T cell.

Pembrolizumab (Keytruda) and nivolumab (Nivo) work differently. They bind directly to the PD1 receiver on the T cell, blocking the inhibitory signals from both PD-L1 and PD-L2.

But Durva cannot prevent PD-L2 from binding to PD1.

But Pembrolizumab and nivolumab cover PD1.

They prevent both PD-L1 and PD-L2 from engaging that receptor and delivering their inhibitory signals through PD1.

In our Cell City picture, Durva blocks a particular inhibitory message.

Keytruda and Nivo cover the PD1 receiver so that neither PD-L1 nor PD-L2 can deliver its inhibitory signal through that receiver.

That distinction is established molecular biology.

But Does the Difference Matter Clinically?

Here the investigation encounters its most important boundary.

The fact that PD1 blockade prevents signalling from both ligands does not prove that PD-L2 alone continued to apply the brake, causing Durva to fail.

Nor does it prove that switching to Keytruda or Nivo after Durva progression will work.

The cancer may have progressed because T cells could not recognise the tumour, could not enter it, or could not overcome other suppressive mechanisms.

Blocking PD1 may not solve any of those problems.

But the molecular difference creates a legitimate research question.

Has the difference been investigated directly in patients whose cholangiocarcinoma progressed on durvalumab?

And if so, what happened?

The Patient-Led Conversation

The patient should not have to argue that another checkpoint inhibitor must work.

The patient should be able to ask what evidence supports or argues against considering it.

A useful question is:

"Durva blocked PD-L1, while Keytruda and Nivo block PD1. Has that difference been investigated after Durva progression in cholangiocarcinoma? What evidence supports or argues against another checkpoint inhibitor in my circumstances?"

The answer may be that the evidence is too weak, the risks too high, or another treatment offers a better prospect.

Those conclusions may be entirely appropriate.

But the patient deserves to understand the reasoning.

The question is no longer simply whether these medicines belong to the same family.

It is whether their differences have been tested in the clinical situation the patient now faces.

Nor should exposing this argument resonably restrain following the consistent evidence produced in trhe negative PD-L1 trial groups.

Sharing a treatment family does not mean sharing the same target. But a different target is not proof of a different outcome.

08 | What Does the Clinical Evidence Actually Show?

The molecular difference between PD-L1 and PD-1 blockade is established. But the question that matters to a patient is whether that difference produces a meaningful clinical response, particularly after progression on durvalumab. Biology provides a reason to investigate. Clinical evidence must establish whether the treatment works.

We began with a clinical observation in melanoma. Patients classified as PD-L1 negative responded to nivolumab, even though their baseline biopsy did not meet the trial's PD-L1 positivity threshold.

We then examined why that could happen.

PD-L1 expression can change. A biopsy cannot observe every part of a tumour or continuously follow its biology. PD-L2 can also deliver inhibitory signals through PD-1.

Those mechanisms establish biological possibilities. They do not establish the effectiveness of a treatment in cholangiocarcinoma.

We must now examine what patients with biliary tract cancer have actually experienced in clinical studies.

What Has Pembrolizumab Demonstrated?

Pembrolizumab (Keytruda) blocks PD-1, preventing inhibitory signalling through that receptor from both PD-L1 and PD-L2.

In KEYNOTE-158, a study that included previously treated patients with advanced biliary tract cancer, pembrolizumab monotherapy produced objective responses in approximately 5.8% of the biliary tract cancer cohort.

The majority of patients did not achieve an objective response. However, some responses were durable.

KEYNOTE-028, which enrolled patients with PD-L1-positive biliary tract cancer, also reported responses to pembrolizumab.

These studies establish that PD-1 blockade can produce clinical responses in a minority of patients with advanced biliary tract cancer.

But they do not establish that pembrolizumab will work for an individual patient whose tumour is PD-L1 negative.

Nor do they establish its effectiveness after progression on durvalumab.

A low overall response rate does not mean nobody responds. But the existence of responders does not establish which patient will be one of them.

What Has Nivolumab Demonstrated?

Nivolumab (Nivo) also blocks PD-1.

A phase 2 study led by Kim and colleagues examined nivolumab monotherapy in previously treated patients with advanced refractory biliary tract cancer.

The study reported an investigator-assessed objective response rate of 22% among evaluable patients. Independent central review reported a lower response rate of 11%.

The distinction matters. Different assessment methods can produce different response estimates, even within the same study.

The trial also reported responses among patients whose tumours were microsatellite stable.

This is relevant because it demonstrates that responses were not confined to patients with MSI-high disease.

But the study was not a randomised comparison against an alternative treatment, and its findings do not establish nivolumab as an effective treatment for every biomarker-negative patient.

Most importantly for our investigation, it did not establish that nivolumab is effective after progression on durvalumab.

What About Combining PD-1 and CTLA-4 Blockade?

There is another biological approach.

Nivolumab blocks PD-1. Ipilimumab blocks CTLA-4. The combination attempts to release two different restraints on T-cell activity.

Studies of nivolumab plus ipilimumab in advanced biliary tract cancer have reported objective responses in some patients.

These findings show that combining checkpoint inhibitors can produce activity in this disease.

But combination treatment can also increase the risk of serious immune-related adverse effects.

Releasing more than one immune brake does not guarantee that the immune system can recognise and destroy cancer.

The cancer may still lack effective tumour-reactive T cells, exclude them from the tumour, or resist their attack through other mechanisms.

Blocking more immune checkpoints is not the same as proving that more patients will benefit.

What Happens After Durvalumab Progression?

This is the question Section 07 brought us to.

Durvalumab is commonly used with gemcitabine and cisplatin as first-line treatment for advanced biliary tract cancer.

When the cancer progresses, a patient may ask whether a PD-1 inhibitor could offer another opportunity.

The molecular targets are different.

But we need clinical evidence from patients who have actually received PD-1 treatment after progressing on durvalumab.

Earlier pembrolizumab and nivolumab monotherapy studies cannot answer that question directly because they largely preceded the routine first-line use of durvalumab.

More recent studies, including investigations of nivolumab plus ipilimumab, have enrolled some patients previously exposed to durvalumab.

However, the presence of previously treated patients within a study does not, by itself, establish a response rate for the post-durvalumab subgroup.

To answer the question, we need to know how many patients had progressed on durvalumab, what treatment they received next, how many responded, and how long those responses lasted.

We also need to distinguish patients who received nivolumab alone from those who received nivolumab combined with ipilimumab.

Evidence that PD-1 blockade can work in biliary tract cancer is not automatically evidence that switching from durvalumab to PD-1 blockade works after progression.

What Does This Mean for the Patient?

The evidence supports two conclusions.

First, PD-1 inhibitors have produced objective responses in some patients with biliary tract cancer.

Second, their clinical benefit after progression on durvalumab remains insufficiently established.

These conclusions must remain separate.

For a patient who has exhausted established treatment options, the Earlier Right Action is to ask whether there is direct evidence for treatment in their circumstances, whether relevant biomarkers have been assessed, and whether a suitable clinical trial is available.

A treatment should not be recommended simply because its mechanism offers a plausible explanation.

Equally, the absence of definitive evidence should not be misrepresented as proof that a biological response is impossible.

The mechanism tells us what might work. The clinical evidence tells us what has worked. The patient's next decision requires knowing the difference.

09 | What Must We Investigate to Find the Answer?

We know that some patients respond to PD-1 blockade despite negative baseline PD-L1 testing. We know PD-L1 expression can change during an immune response. What we have not established is which biological conditions make those responses possible, or whether those conditions can identify patients with cholangiocarcinoma who may benefit.

An investigation should not end with an explanation that sounds biologically reasonable.

It must produce a question that can be tested.

We have identified several possible explanations for responses among patients classified as PD-L1 negative.

PD-L1 may have been present in tumour regions the original biopsy did not sample.

PD-L1 may have been expressed by immune or supporting cells not included in the tumour-cell scoring method.

PD-L1 expression may have increased after interferon-gamma signalling developed.

PD-L2 or other inhibitory pathways may also have contributed to the original restraint on tumour-reactive T cells.

These explanations may overlap.

The task is to establish which mechanisms were actually present in patients who responded.

First, Examine the Tumour Before Treatment

The starting point is the original tumour sample.

But measuring PD-L1 alone is not enough to describe the immune environment.

We need to know whether CD8 T cells are present, where they are located and whether they show evidence of activation or dysfunction.

We need to examine PD-L1 expression on cancer cells and relevant non-cancer cells, using clearly defined testing methods.

We should also examine PD-L2 expression and evidence of interferon-gamma signalling.

Where appropriate, tumour antigen presentation and other mechanisms affecting T-cell recognition should be investigated.

A PD-L1 score tells us about one measurement. Understanding an immune response requires examining the system around it.

Second, Examine What Changes During Treatment

A baseline biopsy provides a starting point.

It cannot show what happens after treatment begins.

Where clinically appropriate and ethically justified, researchers could compare tumour samples collected before treatment with samples collected during treatment.

These are called paired biopsies.

They allow investigators to examine changes within the same patient's cancer over time.

The questions are specific.

Did more CD8 T cells enter the tumour?

Did interferon-gamma signalling increase?

Did cancer cells or surrounding cells increase PD-L1 expression?

Was PD-L2 present, and did its expression change?

Did the tumour show evidence of an effective immune attack, or only an inflammatory response without meaningful cancer-cell destruction?

The distinction is essential.

A biological change during treatment is not automatically evidence that the treatment is controlling the cancer.

Third, Compare the Biology With Actual Responses

The next step is to connect the laboratory observations to clinical outcomes.

Which patients achieved an objective response?

Which patients experienced stable disease, progression or serious treatment toxicity?

Were the biological changes different between responders and non-responders?

Did patients whose tumours increased PD-L1 expression respond more often, or was that change also observed in patients whose cancer continued progressing?

Did PD-L2 expression or T-cell location help distinguish the outcomes?

Were responses durable?

These comparisons are necessary because finding a mechanism in a tumour does not establish that the mechanism caused the patient's response.

The investigation must connect what changed inside the tumour with what happened to the patient.

Fourth, Investigate the Post-Durvalumab Question Directly

The question raised in Section 07 requires its own clinical investigation.

We need evidence from patients with cholangiocarcinoma whose cancer progressed during or after durvalumab treatment.

Researchers would need to record previous treatment, disease progression, tumour characteristics, biomarkers and the patient's clinical condition.

The next treatment must be clearly identified.

Pembrolizumab alone, nivolumab alone and nivolumab combined with ipilimumab are different treatment approaches.

Their outcomes cannot simply be combined and attributed to the difference between PD-L1 and PD-1 blockade.

We would need to measure objective responses, response duration, disease control, survival, quality of life and adverse effects.

Where possible, prospective studies with appropriate comparison groups would be needed to establish whether the strategy provides a meaningful clinical benefit.

The question is not whether PD-1 blockade is molecularly different from durvalumab. We already know it is. The question is whether that difference changes patient outcomes after progression.

What Would Count as an Answer?

A useful answer would identify a reproducible biological or clinical characteristic associated with benefit from PD-1 blockade.

It would distinguish that characteristic from changes that occur without benefit.

It would also establish whether the expected benefit is sufficient to justify the risks of treatment.

That evidence could improve patient selection, guide future trials and help clinicians make better-informed treatment decisions.

It could also demonstrate that the hypothesis is incomplete or incorrect.

Both outcomes would advance understanding.

For the patient, the objective is not to defend a preferred explanation.

It is to discover which decisions genuinely improve the opportunity for survival.

A plausible mechanism gives us a question. A measured patient outcome gives us evidence. Survival decisions require the two to be connected.

10 | What Does Today's Patient Do With This Understanding?

A negative PD-L1 result does not prove that PD-1 blockade cannot work. But it does not establish that treatment will work either. The patient's task is to understand what the evidence supports, what remains uncertain and which next steps could preserve a legitimate treatment opportunity.

I began this investigation with a question.

How could patients classified as PD-L1 negative respond to medicines designed to block inhibitory signalling through PD-1?

The 41.3% response rate in the PD-L1-negative nivolumab subgroup of CheckMate 067 established that the responses occurred.

It did not establish why.

We followed that question through the checkpoint biology, the limitations of tumour biopsies, dynamic PD-L1 expression and the immune environment of cholangiocarcinoma.

We also examined the molecular difference between durvalumab, pembrolizumab and nivolumab.

The investigation has not established that a patient with PD-L1-negative cholangiocarcinoma will benefit from PD-1 treatment.

But it has established why a single negative PD-L1 result should not be asked to answer questions beyond what the test measured.

First, Understand What Your PD-L1 Result Means

Ask your treating team which PD-L1 assay and scoring method were used.

Ask whether the result was based on tumour-cell staining, immune-cell staining or another scoring system.

Ask what threshold defined a positive or negative result.

Most importantly, ask whether PD-L1 status is actually required to determine eligibility for the treatment being considered.

Before accepting what a negative result rules out, establish exactly what the test ruled in or ruled out.

Second, Confirm the Other Relevant Biomarkers

PD-L1 is not the only biomarker relevant to immunotherapy decisions.

Ask whether your tumour has been assessed for mismatch repair deficiency (dMMR) or microsatellite instability-high status (MSI-H).

Ask whether tumour mutational burden has been assessed where clinically appropriate and whether that result has any established treatment significance in your circumstances.

Comprehensive molecular profiling may also identify other actionable alterations that lead to treatment options unrelated to immunotherapy.

These include genomic alterations for which targeted therapies or relevant clinical trials may exist.

Do not allow the investigation of one treatment to distract from other treatments supported by stronger evidence for your tumour's biology.

Third, Understand What Previous Immunotherapy Actually Targeted

If you have received durvalumab, establish exactly what treatment you received and why it stopped.

Did the cancer progress during treatment, or was treatment discontinued because of toxicity or another clinical reason?

Durvalumab blocks PD-L1. Pembrolizumab and nivolumab block PD-1.

That molecular distinction is real.

But it does not establish that switching treatment after durvalumab progression will help.

Ask your oncologist what clinical evidence supports or argues against another checkpoint inhibitor in your circumstances.

Also ask whether previous immune-related toxicity could make further immunotherapy unsafe.

Knowing that two drugs are different is the beginning of the question, not proof that switching will work.

Fourth, Investigate the Evidence and Available Pathways

If established treatment options have been exhausted, ask whether a suitable clinical trial is available.

A trial may offer access to a treatment strategy while collecting the evidence needed to determine whether it works.

Where no suitable trial exists, ask whether there is a lawful access pathway that could be considered by the treating clinician.

Depending on the country, this may involve off-label prescribing, special access arrangements or other regulated pathways.

Availability, funding and prescribing requirements differ between jurisdictions.

Willingness to pay does not establish that treatment is medically appropriate.

A clinician must still assess the evidence, potential benefit, treatment risks and the patient's overall circumstances.

Access is a pathway to investigate. It is not evidence that a treatment will produce a response.

Fifth, Ask the Question That Changes the Conversation

The patient does not need to become an immunologist.

But the patient should understand enough to distinguish a test result from a biological conclusion.

For a patient with advanced cholangiocarcinoma, particularly after progression on durvalumab, the question could be:

"I understand that my PD-L1 result is negative and that the evidence for PD-1 treatment in my circumstances may be limited. But I also understand that PD-L1 expression can change and that PD-1 inhibitors block a different target from Durva. What evidence supports or argues against this treatment for me, and have we investigated every appropriate biomarker, clinical trial and treatment pathway?"

This question does not demand an unproven treatment.

It demands a clear explanation of the evidence behind the decision.

The answer may be that another checkpoint inhibitor is unlikely to help, that its risks outweigh the potential benefit, or that no appropriate access pathway exists.

Those may be clinically sound conclusions.

But the patient should understand how those conclusions were reached.

What Has This Investigation Established?

We know that some patients classified as PD-L1 negative respond to PD-1 inhibitors.

We know that PD-L1 expression can change in response to immune activity.

We know that cholangiocarcinoma frequently presents barriers to effective immune-cell activity.

We know that PD-1 blockade has produced responses in some patients with biliary tract cancer.

And we know that durvalumab and PD-1 inhibitors do not block identical molecular targets.

What we have not established is that dynamic PD-L1 expression explains the responses observed in PD-L1-negative patients.

Nor have we established that switching from durvalumab to pembrolizumab or nivolumab improves outcomes after progression in cholangiocarcinoma.

These are questions requiring direct clinical investigation.

The purpose of this investigation is not to turn uncertainty into a treatment promise.

It is to prevent uncertainty from being mistaken for proof that no opportunity exists.

Today's patient needs the evidence explained early enough to act on what it supports.

That is how understanding becomes an Earlier Right Action.

A negative test is a finding, not a verdict. Understand what it proves, investigate what it does not, and keep every evidence-supported option open.