Breaking the Biomarker Bottlenecks: Why CDx and Digital Pathology Just Got Much More Strategic
Below, Chat GPT 5.2 summarizes a pair of articles in The Pathologist. The experts are Jorge Reis-Filhos (AstraZeneca), Rob Monroe (Danaher-Leica) and Karan Arora (Leica). For policy background, see this blog back on Feb 10 on AMA CPT struggles to code whole slide imaging AI diagnostics (here). As advanced CDx shift more and more to digital pathology, it gives the lab one more way to collide dysfunctionally with the CMS 14 Day Rule. (See an interview with Reis-Filhos alone, a month earlier, here.)
CDx and Digital Pathology:
Infrastructure, Not Ornament
Companion diagnostics (CDx) and digital pathology are no longer niche technical topics. They are moving toward the center of oncology strategy. That shift matters even more given recent AMA CPT developments around the codability of whole slide imaging (WSI) and AI-based pathology services. As I discussed on February 10, the AMA CPT Editorial Panel appears prepared to route WSI/AI services into Category III, after years of ambiguity and effective moratoria on codability [DIHP Blog 2-10].
That coding posture does not guarantee payment. But it removes a structural barrier. And when you read recent industry discussions from AstraZeneca, Danaher, and Leica, it becomes clear that CDx and digital pathology are being repositioned as therapy infrastructure—not optional lab enhancements.
The timing is not coincidental.
CDx Is Becoming Computational Biology
Historically, companion diagnostics were binary gatekeepers. One biomarker, one cutoff, one therapy.
What is now being described is very different. CDx is evolving into a computational modeling layer that attempts to quantify tumor biology continuously, not categorically. AI-based quantitative scoring systems measure target expression, membrane localization, spatial relationships, and biologically relevant correlations. The example of AI-quantified Trop-2 expression for ADC selection is emblematic: the biomarker is no longer simply “present or absent,” but a mechanistically informed, quantitatively normalized signal intended to predict drug internalization and response.
That is a shift in kind, not degree.
Instead of:
“Does the patient meet threshold X?”
The emerging model asks:
“What is the biological state of this tumor, and how does that map to therapeutic response?”
This transforms CDx from a lab service into a decision model.
AI as Reproducibility Mandate
One of the more pragmatic themes in the discussion is not about scientific discovery, but about reproducibility. AI is being framed as a solution to long-recognized interpretive variability in IHC:
Inter-pathologist variability
Inter-laboratory variability
Instability around binary cutoffs
That framing matters enormously. If AI becomes the accepted method for stabilizing interpretation and ensuring consistent patient selection across sites, then AI-enabled digital pathology becomes foundational. It stops being an innovation accessory and starts looking like a quality-control necessity.
That intersects directly with the CPT conversation. If WSI and AI services are required to deliver reproducible CDx results at scale, then codability is not academic—it becomes a deployment prerequisite [DIHP Blog 2-10].
Digital Pathology as Vertical Infrastructure
Perhaps the most revealing aspect of the discussion is how Danaher and Leica describe their positioning. This is not just scanner sales or software add-ons. It is vertical integration:
Antibody sourcing (via Abcam)
Assay development
Computational pathology systems
AI-enabled scoring platforms
Global clinical trial lab networks
Quality systems aligned with pharma
In other words, digital pathology is being integrated into the drug-development stack.
If AI-enabled IHC interpretation becomes integral to ADCs, IO agents, and combination regimens, then WSI scanners and computational platforms are no longer peripheral. They become embedded in the therapeutic value chain.
That elevates the strategic importance of CPT pathways for WSI-AI services [DIHP Blog 2-10].
Multimodal Biomarkers — With Parsimony
Another notable shift is the emphasis on multimodal AI biomarkers. The vision includes combining:
H&E histology
IHC markers
Genomics
Transcriptomics
Selected EHR parameters
Potentially radiology
However, what is striking is the explicit emphasis on “maximum parsimony.” The goal is not maximal data aggregation. It is identifying the minimal sufficient parameter set needed to guide therapy decisions.
That language reflects awareness of real-world deployment constraints:
Regulators resist black-box opacity.
Payers resist open-ended cost expansion.
Health systems resist workflow complexity.
Designing multimodal CDx models around minimal deployable parameter sets is a strategic move to avoid triggering reimbursement and adoption backlash.
Inferring Genomics from H&E
One of the more provocative claims is the ability to infer genomic alterations directly from H&E slides using AI. If validated at scale, this has major implications.
It would mean:
Sequencing becomes triaged rather than universal.
Lower-resource settings could access biomarker stratification without NGS.
Digital pathology platforms gain leverage in global markets.
In economic terms, some CDx value could migrate from molecular sequencing labs toward computational pathology platforms. That is not an incremental adjustment—it is a reallocation of value.
Regulators as Partners, Not Obstacles
Another tonal shift is worth noting. The panel describes regulators as increasingly collaborative participants in CDx development. That aligns with what we have seen in breakthrough designations for AI-driven diagnostics and biomarker tools.
Meanwhile, at the CPT level, the movement toward Category III for WSI/AI suggests institutional adaptation rather than outright resistance [DIHP Blog 2-10].
Taken together, the regulatory posture appears less adversarial than it might have five years ago.
Competitive Implications
This is not a neutral academic discussion. It has strategic consequences.
First, vertical integration is emerging as an advantage. Companies that control antibodies, assays, AI platforms, digital hardware, and pharma partnerships will have structural leverage in CDx development.
Second, pharma is embedding AI-CDx thinking earlier in drug pipelines. Late-stage bolt-on diagnostics may become less viable if AI-driven biomarker strategy is built into early R&D.
Third, CDx may become platform-dependent. If computational scoring and multimodal integration require specific hardware/software stacks, interoperability becomes a strategic and political question.
Four Important Takeaways
1. CDx Is Shifting from Binary Testing to Continuous AI Biology Modeling.
This changes validation paradigms, regulatory conversations, and payer narratives.
2. AI Is Being Positioned as a Reproducibility Requirement, Not a Novelty.
If AI becomes essential for consistent patient selection, codability of WSI/AI services becomes strategically necessary [DIHP Blog 2-10].
3. Multimodal Biomarkers Will Be Designed for Parsimony.
The industry is explicitly trying to balance innovation with deployability and reimbursement realism.
4. Digital Pathology Is Becoming Therapy Infrastructure.
WSI and AI are being integrated into drug development and companion diagnostic deployment. That makes the AMA’s Category III direction more consequential than it might first appear [DIHP Blog 2-10].
In short, CDx and digital pathology are converging into a unified AI-driven therapeutic infrastructure. The CPT coding shift does not settle reimbursement. But it removes a structural barrier at exactly the moment when industry players are building vertically integrated ecosystems around AI-enabled biomarker selection.
That alignment is not accidental—and it is unlikely to be temporary.
