4D Path’s QPOR platform has been selected as an exploratory biomarker in the Phase II DAD-IO study (NCT04724018) in locally advanced or metastatic urothelial carcinoma, led by Dana-Farber and supported by Gilead. The trial evaluates a double–antibody-drug conjugate regimen—sacituzumab govitecan and enfortumab vedotin—with or without pembrolizumab, in both treatment-naïve and previously treated populations. QPOR will prospectively analyze pre-treatment H&E-stained biopsy slides to generate image-derived biomarkers intended to predict response to these combination regimens.
The move advances QPOR from retrospective validation to a prospective clinical setting in a tumor type with few effective predictive tools. The platform applies physics-inspired algorithms to routine pathology images to infer features tied to immune activity, proliferation, and tumor architecture. For sponsors and investigators testing ADC plus IO strategies, this is a direct attempt to identify responders and understand resistance mechanics using material sites already collect, without layering on new tissue requirements, IHC panels, or broad genomic sequencing. The Phase II DAD-IO program builds on earlier Phase I signal from the double-ADC approach and widens the context to include triplet therapy, creating a relevant proving ground for predictive stratification.
Strategically, embedding a digital pathology biomarker prospectively—while keeping it exploratory—reflects a pragmatic path for AI tools seeking clinical credibility without putting trial timelines at risk. It also mirrors a broader shift: as ADC combinations proliferate and toxicity management becomes a gating factor, sponsors are searching for scalable predictors that can improve risk-benefit and control trial size and cost. H&E-based prediction, if reproducible across scanners and sites, offers operational appeal relative to biomarker architectures that depend on fresh tissue, multiplex assays, or complex logistics. For 4D Path, association with a high-visibility ADC/IO program expands use beyond breast cancer datasets where prior retrospective performance was shown and aligns the company with a dominant development theme in solid tumors.
This selection has practical implications for trial operators. Sites will need reliable whole-slide imaging, standardized staining, and data transfer workflows to support centralized analysis, adding to the digital pathology footprint already creeping into multicenter studies. CROs and central labs will be asked to harmonize image quality, version-control algorithms, and manage provenance, creating new SOPs and audit trails that satisfy both GCP and software-as-a-medical-device expectations. For sponsors, a validated image-based predictor could inform enrichment or adaptive designs in subsequent phases, potentially reducing exposure to double-ADC toxicity among likely nonresponders. Regulators are watching the maturation of image-derived biomarkers; prospective evidence that is analytically locked, scanner-agnostic, and clinically correlated will be needed before any companion diagnostic or labeling ambitions are realistic.
Analysis of baseline biopsies is slated to begin in early 2026, with the trial expected to run 12 to 18 months. Key signals to watch are pre-specified performance metrics for the algorithm in predicting response across the double-ADC and triplet arms, subgroup consistency in treatment-naïve versus previously treated patients, and robustness across sites and scanners. Equally important will be evidence of assay lock, batch-effect control, and reproducibility on independent holdouts. If the biomarker shows durable, prospectively confirmed utility, expect sponsors to test it for cohort enrichment in later-stage studies and for 4D Path to pursue a defined regulatory pathway. Risks include limited generalizability beyond study conditions, variability in real-world staining and imaging, and the evolving regulatory stance on AI-derived H&E biomarkers. The near-term question for the field is whether an image-only predictor can meaningfully optimize ADC/IO trial execution and patient selection without incurring new operational friction.
Jon Napitupulu is Director of Media Relations at The Clinical Trial Vanguard. Jon, a computer data scientist, focuses on the latest clinical trial industry news and trends.
