PDAC · Landscape Series · Chapter14

The Stroma Problem: Why PDAC's Microenvironment Kills Every Drug Before It Arrives

He came in complaining of weight loss, fatigue, and just not feeling well. The soft B
symptoms that every oncologist recognizes instantly as a cancer somewhere, except we
could not yet say where. His CT scan showed fullness in the head of the pancreas. Not a
discrete mass. Just fullness. His CA 19-9 was mildly elevated. In the range that could be
pancreatitis, could be something else.

Ronald Matteotti, MD · Surgical Oncologist · PDAC Landscape Series

Stroma as percentage of
tumor volume

0 %

Of tumor that is actually
cancer cells

< 0 %

PEGPH20 Phase 3
stroma-degrading strategy

Failed

Months OS TTFields first to
bypass the barrier

A patient who still sits with me

Igor was a frequent flyer in my practice. Every clinician who saw him knew what was happening. None of us could prove it.

The case that defines this chapter

He came in complaining of weight loss, fatigue, and just not feeling well. The soft B symptoms that every oncologist recognizes instantly as a cancer somewhere, except we could not yet say where. His CT scan showed fullness in the head of the pancreas. Not a discrete mass. Just fullness. His CA 19-9 was mildly elevated. In the range that could be pancreatitis, could be something else.

Our GI colleagues took him to endoscopic ultrasound and did a fine-needle aspiration. The pathology came back as stromal tissue. Fibrous stroma. No malignant cells. They repeated it. Same result. The clinical picture was unambiguous. Every experienced clinician thought the same thing, that this was a stroma-dense, cellularly depleted pancreatic adenocarcinoma. The problem was that we could not prove it.

Over the next several months, we were stuck. No major institution starts systemic chemotherapy without tissue diagnosis. We watched him lose weight. We repeated scans. We repeated the biopsy. We requested more imaging, more biomarkers, more consultation. Nothing gave us what we needed to proceed.

By the time the pathology finally yielded malignant cells on a third attempt, the window had closed. The desmoplastic reaction around his vessels was advanced. He was no longer a surgical candidate. He never really had been.

He started systemic treatment when the diagnosis was finally confirmed. The response was minimal. The stroma that had hidden his cancer from our biopsies was now hiding it from our drugs. He did not live long after that.

Igor’s case sat with me for a long time. It still does. The frustration was not diagnostic nihilism or institutional rigidity. The frustration was that everything we have built in oncology assumes we can access the tumor. We can biopsy it to confirm the diagnosis. We can deliver drugs to it once the diagnosis is confirmed. We can measure response when the drugs reach the cells. Igor’s stroma defeated every one of those assumptions in sequence.

The question his case forced on me, and the question the rest of this article is built around, is what it would look like if we had a modality that could actually penetrate the stroma and reach stroma-dense pancreatic cancer cells selectively. Not degrade the stroma, which has been tried and has failed. Not destroy the fibroblasts, which has been tried and has sometimes made the cancer worse. Something different.

What the stroma actually is

Pancreatic cancer is one of the few solid tumors where cancer cells are not the majority of the tumor.

In a typical PDAC specimen, the desmoplastic stroma accounts for up to 80% of the tumor volume. The cancer cells themselves may be less than 20% of what the pathologist sees under the microscope. This is not an incidental feature of the disease. It is the defining feature, and the feature that separates PDAC from almost every other cancer in terms of how drugs, immune cells, and even biopsy needles interact with it.

The stroma is a dense mat of extracellular matrix proteins, primarily collagen and hyaluronic acid, interlaced with cancer-associated fibroblasts and activated pancreatic stellate cells. Hyaluronic acid binds water and expands, generating interstitial fluid pressures markedly higher than those found in normal tissue. That pressure collapses the tumor vasculature. Microvessels are compressed to the point where blood flow becomes erratic. Drugs circulating in the blood simply cannot reach the cancer cells in meaningful concentration.

“Pancreatic cancer is effectively a drug-delivery disease disguised as a drug-resistance disease. Many chemotherapeutic agents kill pancreatic cancer cells just fine in vitro. In the patient, they never arrive.”

The cellular side of the story is more complicated than the field thought ten years ago. Cancerassociated fibroblasts are not a single cell type. They exist in functionally distinct subtypes. Myofibroblastic CAFs sit directly adjacent to cancer cells and deposit the collagen that physically restrains tumor growth. Inflammatory CAFs sit further away and secrete cytokines that actively promote tumor progression and immune suppression. Different origins, different functions, different responses to therapy.

This heterogeneity matters because the naive therapeutic instinct, which was to deplete the stroma as thoroughly as possible, turns out to be wrong. Some CAFs restrain the cancer. Some promote it. Depleting them indiscriminately can make the cancer more aggressive, not less. This is not a theoretical concern. It is exactly what happened in clinical trials and in carefully designed mouse models when CAF populations were broadly ablated.

What has been tried

A decade of stromal strategies, and what the data actually show.

PEGPH20 was supposed to be the answer. It failed in Phase 3, and the reason is uncomfortable.

PEGPH20 was the most highly anticipated stroma-targeting agent of the 2010s. A pegylated recombinant human hyaluronidase designed to enzymatically degrade hyaluronic acid in the tumor stroma, it promised to reduce interstitial pressure, restore vascular function, and allow chemotherapy to finally reach pancreatic cancer cells. Preclinical models were compelling. Early-phase trials were encouraging. The Phase 3 HALO-301 trial tested PEGPH20 with gemcitabine and nab-paclitaxel against chemotherapy alone in patients with high hyaluronic acid tumors.

HALO-301 failed. No overall survival benefit. Halozyme discontinued development in 2019. The PDAC community, cautiously optimistic about stroma-first therapeutic strategies, had to reckon with a failure that could not be easily dismissed. The drug did what it was designed to do. It degraded hyaluronic acid. It reduced interstitial pressure. And the patients who received it did not live longer.

What PEGPH20 Actually Taught The Field

The stroma is not just a physical barrier to drug delivery. It is also a biological restraint on tumor spread. Remove the barrier without concurrent treatment of the cancer cells, and the cancer cells, which were partly contained by the matrix surrounding them, become freer to disseminate. Broad stromal depletion can increase metastasis. The stroma is doing some of the work of keeping the cancer in place. Destroy it naively and the cancer escapes.

The full competitive landscape of stromal strategies, honest about what works and what does not.

TARGET AND AGENT	STATUS	WHAT THE DATA SHOW
PEGPH20 (HA degradation) Halozyme	Failed Phase 3	HALO-301 missed OS endpoint. Development discontinued 2019. Taught the field that stroma depletion alone makes things worse, not better.
Defactinib (FAK inhibition) Verastem	Phase 1-2	Reprograms CAFs rather than killing them. Modest PFS gains with gem/nab-P and with checkpoint inhibitors. Real but incremental.
Plerixafor (CXCR4 block) Sanof	Phase 1-2	Disrupts CXCL12-CXCR4 axis. Early data showed increased CD8 T-cell infiltration with checkpoint combinations.
Vactosertib (TGF-beta) MedPacto	Phase 2 ongoing	Blocks the central signaling pathway CAFs use to maintain the fibrotic matrix. Active evaluation in combinations.
FAP-directed agents Multiple developers	Early-phase	Target fibroblast activation protein on CAFs selectively. Promising in principle. CAF heterogeneity complicates which subtype you are hitting.
TTFields (Optune Pax) Novocure	FDA-approved 2026	Does not degrade the stroma. Delivers through it. Electric fields bypass vascular delivery entirely. PANOVA3: OS 16.2 vs 14.2 months, HR 0.82.
Size-switching nanoparticles Academic and industry	Preclinical	Shrink in response to stromal enzymes. Navigate the matrix before releasing payload. Years away from the clinic but the most biologically sophisticated approach.

“The lesson of 20 years of stromal failure is not that the stroma cannot be beaten. It is that the stroma cannot be beaten by a single agent. It has to be attacked from multiple directions at once.”

The approval nobody framed this way

Optune Pax was approved for pancreatic cancer in 2026. The framing most oncologists missed is that it is the first validated bypass of the stroma problem.

Tumor Treating Fields do not attempt to degrade, deplete, or remodel the stroma. They deliver a therapeutic effect through it. Alternating electric fields at the right frequency disrupt mitosis in dividing cancer cells. The delivery mechanism, transducer arrays on the skin, does not depend on vascular access. It does not care how compressed the microvessels are. It does not care how dense the collagen matrix is. The field passes through both.

PANOVA-3 MEDIAN OS, TTFIELDS ARM

16.2 mo

vs. 14.2 months control
HR 0.82, p = 0.039

DELIVERY MECHANISM

Non-vascular

Bypasses compressed microvessels entirely. Works through stroma.

CONCEPTUAL FRAMEWORK

Bypass

Not degrade, not deplete, not remodel. Reach through instead of break down.

This is a conceptual shift the field has underappreciated. For 20 years, the assumption was that the answer to the stroma problem was to defeat the stroma. TTFields is the first clinically validated intervention that sidesteps the problem. The benefit is modest. The delivery mechanism is clumsy, requiring patients to wear a device 18 hours a day. But the conceptual proof is important. The stroma can be bypassed rather than destroyed, and a therapy that bypasses it can extend life in patients whose cancers are otherwise inaccessible.

The part that does not get talked about

The stroma does not just stop drugs. It stops biopsy needles. That is how Igor's diagnosis got delayed six months.

The stroma literature almost never addresses the diagnostic dimension of the problem. Every review focuses on drug delivery. Every therapeutic strategy is framed in terms of how a drug gets to a cancer cell. The question of how a pathologist gets to a cancer cell, in the first place, is treated as a solved problem. It is not.

In tumors with extreme desmoplastic reaction and relatively sparse cancer cell populations, a needle placed through the stromal mass has a meaningful probability of returning only fibrous tissue. The cancer is there. The sampling just missed it, because most of what the needle encountered was the stroma rather than the tumor. Published diagnostic yield of EUSguided FNA in pancreatic masses is 75% to 85% in experienced hands. That rate drops in stroma-dense tumors. Patients like Igor can require two, three, sometimes four biopsy attempts before malignant cells are captured.

Every attempt takes time. Every attempt delays treatment. In a disease where median survival from diagnosis is under a year, time spent in diagnostic limbo is time that cannot be recovered. The patients whose biology is most aggressive and most stromally dense are the patients in whom the diagnostic system is most likely to fail. They are also the patients who can least afford the failure.

This is why the liquid biopsy question covered in Chapter 12 of this series is not separate from the stroma problem. It is part of the stroma problem. A blood-based diagnostic that could confirm PDAC in a patient with a suspicious imaging study and inadequate tissue sampling would, in principle, unlock treatment for patients like Igor. The current generation of liquid biopsies does not yet have the sensitivity to do this reliably. The next generation might. The connection between those two chapters of this book is not accidental.

What real solution looks like

The stroma cannot be beaten by a single agent. It has to be attacked from multiple directions at once, with therapies that do different things, delivered by different mechanisms.

A therapy that actually solves the stroma problem would need to do several things at once. Reach the cancer cells at therapeutic concentrations despite the vascular collapse. Kill the cancer cells selectively without broadly ablating the supporting stroma. Work in patients whose baseline biology is cellularly depleted. And do all of this without waiting on a diagnostic confirmation that the stroma itself often prevents.

The TTFields approval is one fragmentary answer to part of that requirement set. Other nonvascular delivery strategies are in development, including ultrasound-focused drug release, radiofrequency-enhanced permeabilization, and direct intratumoral injection. The nanotechnology approach is the one most likely to eventually deliver a truly targeted therapy for stroma-dense PDAC. It is also the approach furthest from the bedside. The development timeline is years to a decade. For patients being diagnosed today, nanotechnology is not yet a treatment option.

The realistic near-term path is the systematic combination of existing strategies. Targeted therapy against KRAS for patients whose cancers will become accessible to the drug as they evolve. TTFields delivery of therapeutic effect through the stroma in parallel. Chemotherapy through whatever vascular access remains. Stromal remodeling agents like defactinib to modestly improve drug penetration. Immunotherapy with combinatorial approaches that account for the immunosuppressive CAFs. None of these individually solves Igor’s problem. In combination, with patient selection informed by stromal subtype analysis, they might.

Back to IGOR

If Igor walked into my clinic today, some things would be different. Most of them would still not have saved him.

I could offer him faster liquid biopsy workup, even if the current test would not definitively diagnose him, at least it would add another data point. I could refer him for molecular imaging more specific for PDAC than standard CT. I could request repeat EUS with largerbore needles and more aggressive sampling strategies. If his tumor finally yielded cancer cells, he would have access to daraxonrasib if he carried a RAS mutation, to Optune Pax if he were still locally advanced, to combination regimens that did not exist when I took care of him. Some of those therapies might give him more time.

None of them would have gotten him past the fundamental problem, which was that his cancer was hidden from us by its own surroundings. What Igor needed, and what the next generation of PDAC patients needs, is a set of tools that do not depend on the tumor cooperating with our current methods. Diagnostic tools that do not require successful tissue sampling. Therapeutic tools that do not require an intact vascular delivery system. Surveillance tools that can track disease progression without scanning for imageable masses.

“The stroma is not just a drug-delivery problem. It is a diagnosis problem, a treatment-sequencing problem, and a time problem. Solving it means solving all four at once. That is why it is hard. That is also why it matters.”

I do not know if we will solve the stroma problem in my career. I do know that Igor taught me why it matters, and why the clinical version of the problem is harder than the literature usually admits. When a patient in front of you is obviously dying of pancreatic cancer, and you cannot prove it, and the proof is the key that unlocks treatment, the stroma is not an abstraction. It is the wall between that patient and the care they deserve.

The progress we make against it will be measured not just in median survival numbers, but in how many more Igors get diagnosed in time to do something about it.

This is Chapter 14 of the PDAC Landscape Series

A chapter by chapter examination of the science, the data, the economics, and the care delivery failures that define pancreatic cancer in 2026.

For unbiased, physician-led oncology second opinions:

Ronald Matteotti, MD · Trusted Telehealth®