🧬 New in Nature Medicine: Mosele, André and colleagues lay out where ADCs go next. Worth reading in full, and worth reading critically from the pathology side. 🔬 The core diagnosis is hard to argue with: ⚠️ ADC benefit is mostly modest. T-DXd in HER2-high is the exception, not the rule. SG in TNBC: 4.8 vs 1.7 months PFS. EV(P) in mUC and MIBC is a great example, too. 💧 Dependent on compound minor fractions of the administered dose reaches the tumor. ❓ The bystander effect is compelling preclinically. Its quantitative contribution in humans remains unknown - but seems to be relevant (TDM1 versus TDXD). 🚧 Chemistry innovation now outpaces trial capacity. Every single tweak is minor in isolation. Their fix: PDO/PDX banks, master protocols, phase 0 designs, and composite predictors combining AI-assisted IHC that scores target expression and subcellular localization with payload-sensitivity assays. Here is where I would push back. 🧠 That composite predictor is not a future concept. For Nectin-4 it already exists. With @niklas_kluemper and colleagues we showed that membranous Nectin-4 drops during metastatic spread (Clin Cancer Res 2023), that NECTIN4 amplification predicts EV response with >90% vs ~30% ORR by a simple FISH assay (JCO 2024), and at #ASCOGU26 a four-tier membranous score adapted from CAP HER2-gastric criteria in 183 EV-treated patients (publication coming soon). So the bottleneck is not that we lack the assay. 🎯 1️⃣ EV is still approved and given biomarker-unselected. The field accepted "Nectin-4 is ubiquitous" as a premise and did not revisit it initially. Membranous expression, the actual prerequisite for binding, was probably overestimated from the start. 2️⃣ Which block you stain decides the answer. If target expression falls between primary and metastasis, biopsy site is not a logistical detail. It is relevant for the relevant assay result. 3️⃣ Localization is not a refinement, it is the readout. H-scoring in many (clinical) trials treats cytoplasmic and membranous signal as interchangeable. For an ADC that must bind the surface, that is a category error. HER2 taught us this twenty years ago - and that is probably the reason why anti-HER2 ADCs with membranous (!!!) HER2 scoring are successful with selection. 4️⃣ Tissue-agnostic ADC development, which the authors rightly push, presupposes a validated, transferable scoring system. Without it, "target-expressing" means whatever each lab's antibody and cutoff say it means. The uncomfortable version: the reason ADC benefit looks modest may partly be that we keep measuring the target inappropriate and then treat everyone anyway. Pathology is not the downstream validation step in this story. It is the rate-limiting one. 🔬 Curious what @FAndreMD and others think about moving membranous scoring standards ahead of the next wave of constructs rather than behind it. 🔗 doi.org/10.1038/s41591-026-0…