Claudin 18.2, B7-H3 & TROP2: Biomarker Strategies for the Next ADC Wave

Claudin 18.2, B7-H3 & TROP2: Biomarker Strategies for the Next ADC Wave

The first generation of antibody-drug conjugates taught oncology an uncomfortable lesson: the target on the label is not always the biomarker that predicts benefit. HER2-low broke the old link between overexpression and eligibility, and TROP2 ADCs reached the market with no TROP2 test at all.

The next wave sharpens that question. Claudin 18.2 already has an approved companion diagnostic, built for a naked antibody, and the ADCs now reading out in phase 3 are using a very different cut-off. B7-H3 ADCs are producing survival benefits in small cell lung cancer without testing anyone for B7-H3. TROP2 is moving towards quantitative, computational scoring. Three targets, three different answers to the same question: does target expression need to be measured, and if so, how?

This article is for translational, biomarker and companion diagnostic teams. It covers where each target stands as of October 2026 and what a sensible biomarker strategy looks like when the payload, not the target, may be doing most of the work.

Key takeaways

• Zolbetuximab was approved by the FDA on 18 October 2024 with the VENTANA CLDN18 (43-14A) RxDx Assay; positivity means at least 75% of viable tumour cells with moderate-to-strong membranous CLDN18 staining, about 38% of patients.
• The leading claudin 18.2 ADC, sonesitatug vedotin, met its overall survival endpoint in CLARITY-Gastric01 using a much lower threshold: at least 25% of tumour cells at any intensity, about 60% of gastric/GEJ cancers, scored with a different Ventana assay (SP455).
• B7-H3 ADCs ifinatamab deruxtecan and risvutatug rezetecan enrolled small cell lung cancer patients without B7-H3 selection. Whether B7-H3 needs testing at all is an open question.
• TROP2 is shifting from conventional IHC towards computational membrane scoring; our TROP2 ADC and immunotherapy patient selection analysis covers it in depth.
• The cross-cutting rule: target expression is most predictive when payload delivery depends on it, and least predictive when a bystander payload hits a target that almost every tumour expresses.

Claudin 18.2: biology and the first companion diagnostic

Claudin 18 is a tight junction protein with two splice isoforms. CLDN18.1 is expressed in lung; CLDN18.2 is normally restricted to differentiated gastric mucosa, where it sits largely buried within tight junctions. In malignant transformation, cell polarity is disrupted and the epitope becomes more accessible on the tumour cell surface. Expression is retained in many gastric and gastro-oesophageal junction (GEJ) adenocarcinomas and occurs in pancreatic, biliary and other cancers, which is why the target is attractive for antibodies, ADCs, bispecifics and cell therapies.

Zolbetuximab and the 43-14A assay

The FDA approved zolbetuximab-clzb (Vyloy) on 18 October 2024 with fluoropyrimidine- and platinum-based chemotherapy for first-line HER2-negative gastric or GEJ adenocarcinoma whose tumours are CLDN18.2-positive by an FDA-approved test. Japan approved it in March 2024 and the European Commission in September 2024. The VENTANA CLDN18 (43-14A) RxDx Assay from Roche was approved the same day as the companion diagnostic (PMA P230018).

Approval rested on two phase 3 trials, both of which used the assay to enrol patients:

Three assay details matter for anyone building on this precedent:

• The cut-off is high. Roche defines positivity as at least 75% of viable tumour cells with moderate-to-strong membranous CLDN18 staining above background, which identifies about 38% of patients with advanced gastric/GEJ cancer.
• The clone is not isoform-specific. 43-14A detects both CLDN18 isoforms. In gastric tissue the signal is read as CLDN18.2, because that is the predominant isoform there, but the inference depends on tissue context. Assays for lung metastases or non-gastric primaries need to address this directly.
• Staining patterns need training. Pathologists must judge intensity and membrane localisation across often heterogeneous tumours, with apical and incomplete patterns adding interpretive difficulty. That is a standard semi-quantitative IHC problem, and a natural candidate for digital pathology validation.

The claudin 18.2 ADC, bispecific and CAR-T pipeline

The 75% cut-off was set for a naked antibody whose mechanism (antibody-dependent and complement-dependent cytotoxicity) plausibly needs dense target. ADCs carrying membrane-permeable payloads may not. The pivotal ADC trials have tested exactly that.

Sonesitatug vedotin: a lower bar, a new assay

On 27 July 2026, AstraZeneca reported that CLARITY-Gastric01 met its dual primary endpoint of overall survival in third-line and later gastric, GEJ or oesophageal adenocarcinoma, and a key secondary OS endpoint across second line and later. The other dual primary endpoint, PFS by blinded central review, showed a trend but did not reach significance. Detailed hazard ratios have not yet been presented.

The biomarker story is the more important one for diagnostics teams. The trial enrolled patients with CLDN18.2 on at least 25% of tumour cells at any staining intensity, which AstraZeneca estimates covers about 60% of gastric/GEJ cancers, against roughly 38% at the zolbetuximab threshold. The efficacy analyses will also support the clinical performance of the Ventana SP455 assay. If approved, gastric cancer would have two CLDN18 assays with two scoring algorithms for two drugs against the same target. Laboratories would need to run, or at least report, both.

IBI343 and satri-cel

Innovent’s June 2026 announcement confirmed that G-HOPE-001 met its PFS endpoint at the first interim analysis; no efficacy figures were released. The phase 1 data in Nature Medicine are informative on cut-offs: expansion cohorts admitted patients with any membranous staining on at least 1% of tumour cells, yet responses clustered in higher expressers. ORR was 32.6% in patients with 2+/3+ staining on at least 40% of cells, and there were no responses among 10 patients in the 1–39% band. Even with a bystander-capable payload, expression mattered.

Satri-cel became the first CAR-T approved for a solid tumour when China’s NMPA cleared it in June 2026. Its randomised phase 2 trial (CT041-ST-01, published in The Lancet in 2025) required at least 40% of tumour cells at 2+ or above and reported a PFS hazard ratio of 0.37 against physician’s choice.

The pattern across modalities is plausible: CAR-T and immune-engaging formats need enough antigen to trigger killing, the naked antibody needs the most, and ADCs with membrane-permeable payloads may tolerate the least. That hypothesis still has to be confirmed drug by drug.

B7-H3: a target that may not need a test

B7-H3 (CD276) is a B7-family immune checkpoint molecule whose receptor is still not definitively identified. It is overexpressed across a wide range of solid tumours, including small cell lung cancer (SCLC), prostate cancer, oesophageal squamous cell carcinoma, head and neck cancer, and osteosarcoma, with relatively limited expression in normal tissue. Merck describes moderate-to-high B7-H3 expression in about 65% of SCLC tumours. High prevalence plus a potent payload is the recipe for an ADC that works in unselected patients.

Ifinatamab deruxtecan (I-DXd)

Daiichi Sankyo and Merck’s I-DXd pairs an anti-B7-H3 antibody with the deruxtecan topoisomerase I payload. In the phase 2 IDeate-Lung01 trial, 137 patients with previously treated extensive-stage SCLC had a confirmed ORR of 48.2%, median PFS of 4.9 months and median OS of 10.3 months. The FDA granted Breakthrough Therapy designation in August 2025 and accepted the BLA with priority review in April 2026, with a target action date of 10 October 2026. At the time of writing (13 October 2026) we had not seen a decision announced. The confirmatory phase 3 IDeate-Lung02 trial against physician’s choice is ongoing.

Neither study selects for B7-H3. The IDeate-Lung01 eligibility criteria require a biopsiable lesion but no expression threshold. In the earlier phase 1/2 study, Daiichi Sankyo reported tumour shrinkage across a broad range of B7-H3 levels with no apparent correlation between efficacy and expression. The caveat: all 17 evaluable patients had moderate-to-high expression, so the dataset could not test the low end.

Risvutatug rezetecan (HS-20093, GSK’227)

Hansoh’s B7-H3 ADC, licensed to GSK outside China, produced the first phase 3 survival win in the class. ARTEMIS-008 randomised 461 Chinese patients with relapsed SCLC to risvutatug rezetecan or topotecan. At the interim analysis presented at the 2026 World Conference on Lung Cancer, median OS was 18.5 versus 10.3 months (HR 0.46). The trial eligibility excludes prior B7-H3 therapy but sets no expression requirement. A separate China phase 3 in osteosarcoma (ARTEMIS-011) met its PFS endpoint, and GSK is running global pivotal studies, including EMBOLD SCLC-301 in relapsed extensive-stage SCLC.

Other programmes, such as DualityBio and BioNTech’s BNT324/DB-1311, which is in a phase 3 trial against docetaxel in taxane-naive metastatic castration-resistant prostate cancer, are taking the class into new tumour types.

So does B7-H3 need testing?

In SCLC, the current answer appears to be no, for the same reason TROP2 ADCs launched without a test: when nearly every tumour expresses the target and the payload kills neighbouring cells, a threshold may only exclude patients who would have benefited. Three reasons not to close the question entirely:

• No trial has properly sampled B7-H3-low tumours. Absence of correlation in a population that is mostly high-expressing is not evidence that expression is irrelevant.
• Prevalence varies by tumour type. As B7-H3 ADCs move into prostate, oesophageal, head and neck and other cancers, the fraction of low expressers may be large enough to matter.
• Payload resistance may be the better biomarker. With topoisomerase I payloads, markers of payload sensitivity, efflux or DNA damage repair may predict more than antigen density. That is a hypothesis worth building into translational plans now.

TROP2: brief, because the question has moved on

Sacituzumab govitecan and datopotamab deruxtecan were approved without a TROP2 test, and conventional TROP2 IHC has not reliably predicted benefit. The field’s answer has been to change how TROP2 is measured rather than whether: AstraZeneca’s quantitative continuous scoring (QCS) uses computational pathology to calculate a normalised membrane ratio, and Roche’s VENTANA TROP2 (EPR20043) RxDx Device received FDA Breakthrough Device designation in April 2025. We cover this, together with PD-L1 and the ADC plus checkpoint inhibitor combinations, in our analysis of ADC and immunotherapy patient selection.

A cross-cutting biomarker strategy for ADC targets

When does target expression predict ADC benefit?

Claudin 18.2 sits on the left: a substantial negative population, and phase 1 data showing little activity below 40%. B7-H3 in SCLC and TROP2 sit on the right. HER2 shows that the line moves: what once required overexpression now extends to HER2-low and ultralow disease.

Assay and scoring choices

• Choose the scoring metric for the drug, not the target. Percentage of positive cells at any intensity, percentage at 2+/3+, H-score and membrane ratios answer different questions. The CLDN18.2 example shows two pivotal programmes choosing different metrics.
• Think about isoform and epitope. A pan-CLDN18 clone is acceptable in gastric tissue; it may not be elsewhere.
• Plan for low-end precision. Lower ADC cut-offs move the decision boundary into the range where intensity calls are least reproducible. Inter-observer concordance studies should be weighted there.

Setting cut-offs

Enrol broadly in early phases, ideally including expression-low patients, and analyse response against a continuous score before fixing a threshold. A cut-off derived retrospectively from the pivotal trial is common but needs prospective confirmation, and regulators will expect the analytical validation to match the clinical cut-off. Whether a test is a companion or complementary diagnostic depends on whether benefit exists below the cut-off; our guide to companion versus complementary diagnostics sets out that distinction, and the FDA companion diagnostic approval guide covers the regulatory path.

Tumour heterogeneity

Claudin 18.2 expression is heterogeneous within tumours and can differ between primary and metastatic sites and between biopsy and resection. Programmes should specify which specimen is acceptable, the minimum number of viable tumour cells, and whether archival tissue is valid after intervening therapy. For later-line ADCs, where patients may have had prior CLDN18.2-directed treatment, re-biopsy questions will become harder.

What assay and pathology vendors should offer

• Multi-cut-off reporting from one stained slide, so a laboratory can report against more than one drug-specific algorithm.
• Validated image analysis for membrane intensity and percentage scoring, with documented concordance against pathologist reads near the clinical cut-off.
• Continuous expression data from trials, not just positive or negative calls, so sponsors can test threshold hypotheses.
• Payload-sensitivity and resistance panels alongside target IHC, for targets like B7-H3 where antigen alone may not be enough.
• Global ring studies and proficiency schemes for new assays, which matter most when two assays for one target reach the clinic.

Discuss ADC biomarker strategy at ImmunoMark Boston

Antibody-drug conjugates are now one of the main engines of new targeted therapy in oncology, and the biomarker questions above sit where drug developers, pathologists and diagnostics companies have to agree. ImmunoMark Summit Boston, the East Coast Edition in June 2027, brings 300+ senior pharma, biotech and diagnostics leaders together across two parallel streams: Stream 1: Next-Gen Immuno-Oncology, which covers ADCs and other new modalities, and Stream 2: Biomarkers & Companion Diagnostics. The format is keynotes, panels, case studies, pre-arranged 1:1 meetings and a Day 1 drinks reception.

If you want to start the conversation sooner, the immuno-oncology conference in London takes place on 3–4 March 2027.

Register for ImmunoMark Boston 2027

If your company wants to meet this audience, whether you offer IHC assays, digital pathology, central laboratory services or ADC development support, see the Boston 2027 sponsorship opportunities or our overview of oncology conference sponsorship.

FAQs

What is the claudin 18.2 cut-off for zolbetuximab?

Tumours are CLDN18.2-positive for zolbetuximab when at least 75% of viable tumour cells show moderate-to-strong membranous CLDN18 staining on the VENTANA CLDN18 (43-14A) RxDx Assay, the FDA-approved companion diagnostic. About 38% of patients with advanced gastric or GEJ adenocarcinoma meet this threshold.

Do claudin 18.2 ADCs use the same test as zolbetuximab?

Not necessarily. The phase 3 CLARITY-Gastric01 trial of sonesitatug vedotin enrolled patients with CLDN18.2 on at least 25% of tumour cells at any intensity, about 60% of gastric/GEJ cancers, and its data will support the Ventana SP455 assay. Other programmes have used different thresholds, so cut-offs are drug-specific.

Do B7-H3 ADCs require B7-H3 testing?

Not in the trials so far. IDeate-Lung01 (ifinatamab deruxtecan) and ARTEMIS-008 (risvutatug rezetecan) enrolled small cell lung cancer patients without a B7-H3 expression requirement. B7-H3 is expressed in most small cell lung cancers, but low expressers have not been studied rigorously, and the answer may differ in other tumour types.

When does target expression predict ADC response?

Target expression tends to predict benefit when the target has a clear negative population, when normal tissue also expresses it, or when the payload has limited bystander activity. It predicts less when nearly all tumours express the target and a membrane-permeable payload kills neighbouring cells, in which case payload sensitivity may matter more.

Where is TROP2 biomarker testing heading?

Conventional TROP2 IHC has not reliably predicted benefit from TROP2 ADCs, which were approved without a test. Development has moved to computational pathology, notably AstraZeneca’s normalised membrane ratio scoring, alongside a Roche TROP2 assay with FDA Breakthrough Device designation.