BruceBlogMiscellaneous: ASCO Post on TROP2 Advances. CANO on Dig Path & Pathologist "Auditing" Role (foots to AMA App S revisions and CMAA)

Please provide a graduate level summary of this TROP article. I am particularly interested in the gateway test for TROP2 drug access, a special AI or ML digital imaging process of IHC. Thanks.

###

Below is a graduate-level analytic summary of the ASCO Post article, with explicit emphasis on the “gateway test” problem for TROP2 drug access and the emerging role of AI/ML-enabled digital pathology for IHC, framed in a way that should be directly useful for a range of life-science or diagnostics clients and stakeholders.

Graduate-Level Summary: TROP2 ADCs and the Emerging “Gateway Test” Problem

The ASCO Post article synthesizes late-breaking Phase III data from ASCENT-03 and TROPION-Breast02, establishing TROP2-directed antibody–drug conjugates (ADCs) as a new first-line standard of care for patients with metastatic triple-negative breast cancer (TNBC) who are ineligible for immune checkpoint inhibitors . Both trials demonstrate substantial and consistent progression-free survival (PFS) benefits over chemotherapy (hazard ratios ≈0.6), with datopotamab deruxtecan (Dato-DXd) additionally showing a statistically significant overall survival (OS) advantage—a landmark result in this historically aggressive disease setting.

From a therapeutic standpoint, the article positions TROP2 ADCs not merely as incremental improvements but as paradigm-shifting agents, particularly given the historically short survival of metastatic TNBC. Importantly, these benefits appear broadly distributed across prespecified subgroups, reinforcing the notion that TROP2 is a widely expressed and clinically actionable target rather than a niche biomarker. This creates both opportunity and tension around how patients will be selected for treatment as these agents move rapidly into routine clinical practice.

The “Gateway Test” Issue: From Binary IHC to Quantitative Access Control

Although the article does not yet describe a formal companion diagnostic requirement, its implications for TROP2 testing as a gateway to drug access are substantial. As TROP2 ADCs move into first-line use—and potentially earlier disease settings—payers, regulators, and manufacturers will inevitably demand more precise patient stratification, especially if multiple TROP2-targeted agents compete within the same indication.

Historically, IHC biomarkers have functioned as binary or semi-quantitative gatekeepers (positive vs negative; low vs high). However, TROP2 biology complicates this paradigm: expression is continuous, heterogeneous, and spatially variable, and clinical benefit in the reported trials appears not strictly confined to “high expressers.” This creates a classic access dilemma:

Too permissive a test risks payer skepticism and uncontrolled utilization.
Too restrictive a cutoff risks denying benefit in a population with few alternatives.

The article implicitly signals that traditional pathologist-scored IHC may be insufficient to manage this balance as TROP2 ADC utilization scales.

Role of AI / ML-Enabled Digital Pathology as the Next-Generation Gateway

Against this backdrop, AI-driven digital pathology emerges as a natural technical solution to the TROP2 gateway problem—even though it is not yet explicit in the trial reporting. Machine-learning–based image analysis can convert conventional TROP2 IHC slides into quantitative, reproducible, and spatially resolved biomarkers, enabling:

Continuous scoring rather than coarse ordinal categories
Tumor–stroma discrimination and regional expression mapping
Reduction of inter-observer variability, a known weakness of manual IHC
Algorithmic alignment with clinical outcomes, allowing post-hoc optimization of cutpoints based on real-world response data

For manufacturers—such as AstraZeneca and its partners—this approach supports a “soft companion diagnostic” strategy: not a hard exclusionary test at launch, but a progressively refined access tool that can evolve with HEOR data, payer requirements, and label expansions.

Strategic Implications

The article’s clinical message is clear: TROP2 ADCs are here to stay in TNBC. The less explicit—but strategically critical—message is that diagnostic infrastructure must now catch up. As treatment shifts earlier, lasts longer, and competes within class, AI-enabled TROP2 IHC is poised to become the de facto gateway technology, balancing access, evidence, and cost control.

In this sense, TROP2 may follow the trajectory seen with HER2-low and PD-L1: initial enthusiasm driven by broad benefit, followed by increasing diagnostic sophistication once therapeutic success forces the system to ask not whether to treat, but how precisely to decide who gets treated.

Bottom line for your client

The article documents a therapeutic inflection point, but its downstream impact will hinge on who controls the TROP2 “on-ramp.” AI/ML-based digital IHC is well positioned to become that gate—scientifically credible, operationally scalable, and economically defensible—well before a formal FDA-mandated companion diagnostic emerges.

If you’d like, I can also:

Map this to HER2-low precedent
Frame it as a payer-facing HEOR narrative
Or draft a diagnostics-partner strategy memo focused on AI IHC positioning
##

####

I believe the TROP2 test is similar to the NASH test described by Cano. In the TROP2 test, we beleive that pathologists will be required to "check" or "verify' the test - is tumor being stained, etc. what perspectives does CANO, doing a deep dive on digital pathology, bring to people in the oncology R&D world running clinical trials. Who may not know much about dig pathology except what they know "inside the bubble" of NASh or TROP testing. The digital pathology ecosystem is not so healthy, says Cano.

###

Your instinct is exactly right: the emerging TROP2 “gateway test” problem in oncology looks structurally very similar to the AIM-NASH situation Cano analyzes, and Cano’s essay is valuable precisely because it forces people outside the digital-pathology bubble to confront what they are implicitly assuming about AI, validation, and clinical responsibility.

Below is a focused synthesis of what Cano brings to oncology R&D and clinical-trial leaders, especially those thinking about TROP2 IHC + AI primarily as a means to an endpoint (drug access, stratification, enrollment efficiency), rather than as a complex clinical infrastructure.

1. The Core Parallel: AIM-NASH and TROP2 Are Not Autonomous Tests

Cano’s central insight—grounded in the FDA’s qualification of AIM-NASH as a Drug Development Tool, not a diagnostic—maps directly onto your TROP2 scenario .

In both cases:

The algorithm does not replace pathology judgment
The human pathologist must verify:
- that the correct tissue is stained
- that tumor (not stroma or artifact) is being analyzed
- that staining quality, fixation, and context are adequate
The output is only valid once “checked” and accepted by a professional who retains responsibility

Cano reframes this as a calibration model, not automation:

AI reduces variability across observers; it does not eliminate the observer.

For oncology trial teams used to thinking in terms of central labs and readouts, this is a crucial shift. The “gateway test” is not a black box—it is a supervised measurement system whose legitimacy depends on documented human oversight.

2. What Oncology R&D Often Misses: Digital Pathology Is Infrastructure, Not Software

Cano’s most uncomfortable message for non-pathologists is that digital pathology does not fail at the algorithmic level—it fails at the ecosystem level.

Clinical trial leaders working on TROP2 may assume:

Slides can be digitized “once”
Algorithms generalize across sites
Variability can be averaged out statistically
AI scores are portable across scanners, stains, and labs

Cano argues that all four assumptions are empirically false in real-world deployment .

Key ecosystem fragilities he highlights:

Pre-analytical variability (fixation, ischemia time, staining protocols)
Scanner and platform heterogeneity
Weak interoperability between LIS, IMS, and AI tools
Silent failure risk, where outputs look plausible but are wrong
No stable reimbursement model, making digitization economically fragile

For TROP2 trials, this means:

The reliability of your “gateway test” is bounded not by model accuracy, but by the weakest lab in your network.

3. The Pathologist’s Role Is Expanding—Unfunded and Unrecognized

One of Cano’s most important contributions is reframing the pathologist as an auditor of algorithmic behavior, not just a tissue reader.

In TROP2 testing, this implies that pathologists are being asked to:

Verify biological relevance (tumor vs non-tumor)
Detect out-of-distribution behavior
Resist automation bias
Assume legal and ethical responsibility for AI-assisted outputs

Yet Cano notes a structural contradiction:

Regulators demand human-in-the-loop supervision
Payment systems do not compensate for that supervision
Trial budgets often ignore this labor entirely

For oncology R&D teams, this creates a hidden risk:

You are relying on expert labor that is essential for validity—but not explicitly supported, standardized, or incentivized.

This is particularly acute for gateway biomarkers like TROP2, where a single checkmark determines access to life-extending therapy.

4. Why Cano Says the Digital Pathology Ecosystem Is “Unhealthy”

Cano’s diagnosis is not anti-AI—it is anti-naïveté.

He argues the ecosystem is unhealthy because:

Regulatory progress (e.g., AIM-NASH) outpaces reimbursement
Clinical trials rely on digital infrastructure that routine care cannot afford
Adoption is concentrated in elite centers, creating access inequities
Responsibility is pushed downward to individual clinicians without systemic support

For oncology, this matters because trial-validated tools inevitably leak into clinical practice. A TROP2 test that works under trial conditions may:

Be economically unsustainable at scale
Be unevenly available across care settings
Increase variability rather than reduce it if oversight weakens

5. The Strategic Takeaway for Oncology R&D Leaders

Cano’s essay delivers a clear warning to teams running TROP2-based trials:

If you treat digital pathology as a plug-in rather than as regulated clinical infrastructure, you will underestimate risk, cost, and friction.

Practically, this means:

Design TROP2 workflows assuming mandatory human verification
Budget for pathologist oversight as a core component, not an afterthought
Expect regulatory acceptance to favor assistance, not autonomy
Recognize that algorithmic performance ≠ clinical robustness
Plan for ecosystem fragility, not idealized scalability

Bottom Line (in your framing)

Cano gives oncology R&D leaders something they rarely hear inside the TROP2 or NASH “bubble”:

The hardest part of AI-enabled pathology is not proving it works—it is making it safe, reproducible, supervised, and economically survivable once it leaves the trial sandbox.

That perspective is essential if TROP2 testing is going to function as a credible, defensible gateway to drug access, rather than a brittle technical dependency that breaks under real-world pressure.

If you want, next we can:

Explicitly map AIM-NASH → TROP2 CDx strategy
Draft a trial-design checklist informed by Cano’s critique
Or translate this into a payer or FDA-facing narrative on why human-verified AI IHC is not a weakness—but a regulatory strength

I think CANO is concenred the payment system for pathology requires diagnosis, and the TROP2 system requires only "pathologist confirmation" which has no CPT coding.

You are reading Cano exactly the right way, and your framing gets to what is arguably his deepest—if implicit—concern.

Below is a clean articulation of Cano’s perspective, translated into the oncology / TROP2 gateway-test context, with explicit attention to payment, CPT logic, and the structural mismatch you’re pointing to.

Cano’s Core Concern (Reframed for TROP2)

Dr. Luis Cano is not primarily worried that AI pathology won’t work. He is worried that the U.S. payment system only recognizes pathology work when it culminates in a diagnosis, while emerging AI-enabled biomarker workflows—like TROP2 gateway testing—explicitly do not produce a diagnosis .

This creates a structural reimbursement paradox:

Pathology payment logic (CPT) assumes:

A pathologist renders a diagnosis → a billable professional service exists.
TROP2 (and AIM-NASH–like) workflows require:

A pathologist confirms technical and biological validity → but does NOT issue a diagnosis.

In CPT terms, that confirmation is medically essential but economically invisible.

Why This Matters More Than It First Appears

1. “Confirmation” Is Not Clerical — It Is Clinical Responsibility

Cano emphasizes that human involvement in AI pathology is not optional oversight; it is where legal, clinical, and ethical responsibility resides.

In TROP2 testing, the pathologist is being asked to confirm:

Correct tissue and tumor content
Adequate staining and fixation
Absence of artifacts or out-of-distribution behavior
Appropriateness of algorithmic interpretation

That is medical judgment, not QA—but it does not end in a diagnosis, so it does not map cleanly to existing CPT constructs .

2. CPT Is Diagnosis-Centric; TROP2 Is Access-Centric

Cano’s underlying critique is that CPT codes encode a worldview:

Pathology = diagnosis
Diagnosis = professional work
No diagnosis = no billable act

But TROP2 functions as a therapeutic gatekeeper, not a diagnostic endpoint.

The pathologist’s role is to say:

“Yes, this sample is valid for drug access decisions.”

That is neither an anatomic pathology diagnosis nor a laboratory test interpretation in the traditional sense.

Cano sees this as a category error baked into payment policy.

3. The AIM-NASH Precedent Makes the Problem Visible

Cano uses AIM-NASH to show how regulators already recognize this new role:

FDA qualified AIM-NASH as a Drug Development Tool
Explicitly not a diagnostic
Explicitly requires pathologist-in-the-loop confirmation
Explicitly assigns responsibility to the human reviewer

But qualification ≠ payment.

So we now have a situation where:

Regulators require human confirmation
Sponsors rely on that confirmation
Liability attaches to that confirmation
No CPT code exists to pay for it

This is the crux of Cano’s concern .

Why Cano Calls the Digital Pathology Ecosystem “Unhealthy”

From Cano’s perspective, the ecosystem is unhealthy because it externalizes unpaid labor onto pathologists while extracting value elsewhere:

Pharma gets cleaner endpoints
Trials get standardized enrollment
AI vendors get validation
Regulators get risk containment

But the pathologist absorbs:

Time
Cognitive load
Liability
Audit responsibility

—with no corresponding reimbursement mechanism.

This is not sustainable, especially when scaled beyond elite trial centers.

Implications for TROP2 Strategy (Your World)

Your observation leads to a hard but unavoidable conclusion:

If TROP2 access depends on “pathologist confirmation” rather than diagnosis, then the payment system is structurally misaligned with the clinical reality.

Practically, this means:

Expect fragility in real-world adoption
Expect institutional resistance outside trials
Expect variation in how confirmation is performed
Expect unpriced labor to become a bottleneck

Cano would argue that this is not a temporary glitch, but a signal that:

The pathology payment framework has not yet caught up with therapy-gated, AI-assisted medicine.

Bottom Line (Cano, distilled)

Cano is not warning that TROP2 digital pathology will fail technically.

He is warning that:

A system that requires expert medical confirmation but only pays for diagnosis is structurally incoherent.

Until that gap is addressed—by new codes, new payment models, or explicit sponsor support—the TROP2 gateway model will rest on unrecognized, uncompensated professional judgment.