Claudin-18.2 imaging when one biopsy says negative
Stomach cancer patients now get a drug only if a biopsy shows a protein called claudin-18.2. One biopsy core finds it only about six times in ten when the whole tumour has it, so patients are being refused a treatment on a test that misses.
Overview
Zolbetuximab is approved in the US, EU, Japan, China and elsewhere for CLDN18.2-positive, HER2-negative gastric and gastro-oesophageal junction adenocarcinoma, with patient selection by the VENTANA CLDN18 (43-14A) RxDx immunohistochemistry assay. Roughly 38% of screened patients in the pooled SPOTLIGHT and GLOW populations were positive at the 75%-of-tumour-cells threshold. In June 2026 China's regulator approved satricabtagene autoleucel, a CLDN18.2 CAR-T, in the same biomarker-defined population, so the number of decisions resting on this one assay is rising.
The assay's weakness is measurable. Against whole-tumour-section ground truth, a single representative biopsy gave 58.8% sensitivity for CLDN18.2, rising to 76.5% with six to eight biopsies, and a minimum of six endoscopic biopsies was needed to keep false negatives down; 18% of tumours were discordant between the primary and the matched nodal metastasis. A multi-laboratory ring trial found a 12% false-negative rate with 91% overall accuracy.
The imaging work is almost entirely Chinese and early: 68Ga-PMD22, 68Ga-NC-BCH and 124I-18B10(10L) each have a published pilot of 11 to 17 patients, and none has entered a multicentre or registration trial. The opportunity is therefore unusually well specified for a crossover case, because the incumbent test's failure rate is published and the decision it gates is binary.
When a gastric cancer biopsy reads CLDN18.2-negative, no test distinguishes a truly negative tumour from a false negative caused by sampling, so patients may be denied an approved antibody or CAR-T on a result that a single core gets right only about six times in ten.
- Regulators may not accept an imaging agent as an alternative route to a therapy whose label names a specific tissue assay, leaving the scan as an adjunct that nobody reimburses.
- A scan threshold equivalent to the 75%-of-cells histology cut-off has not been defined, and defining it is a study in itself.
- Three Chinese groups are ahead in the clinic, so a later entrant needs a better construct, not just a programme.
- Whether any of the three published tracers moves to a multicentre trial or reports agreement against the approved companion diagnostic.
- Label expansions of CLDN18.2 therapies into earlier lines, which would multiply the number of selection decisions.
- Any regulatory statement on non-tissue routes to establishing CLDN18.2 positivity.
The five domains of a radiopharmaceutical due-diligence review, adapted from the Fryback-Thornbury hierarchy of diagnostic efficacy. Each carries the state of the public evidence and the fact behind it, not a score out of five: a single total hides the reason and stops telling candidates apart. Domains marked as carrying most weight are the ones that decide the case at this asset's stage.
CLDN18.2 positivity at the 75% threshold was 38.4% across 4,507 patients screened for SPOTLIGHT and GLOW, and Astellas reports approximately 35% positivity in Chinese patients with advanced disease.
Three separate first-in-human tracer studies of 11 to 17 patients each, all single-centre and all in China; no comparative or multicentre data.
The case is built on the incumbent's measured failure: 58.8% sensitivity from one representative biopsy against whole-section truth, 76.5% with six to eight biopsies, 18% primary-versus-node discordance, and a 12% false-negative rate in a multi-laboratory ring trial.
No adverse events reported in the small published cohorts; the iodine-124 construct carries a higher radiation burden than the gallium-68 ones, and no formal comparison has been published.
An approved companion diagnostic already defines the claim: the VENTANA CLDN18 (43-14A) RxDx assay. An imaging agent would have to be positioned against it, most plausibly as an adjunct where tissue is inadequate rather than as a replacement.
Red, amber or green on each review category, each with its reason, so an obvious no-go takes seconds rather than a meeting.
Gastric wall and gastro-oesophageal junction imaging sits next to physiological uptake, and the published agents are antibody or antibody-fragment based with the kinetics that implies.
A binary treatment decision gated by a test with a published 12% laboratory false-negative rate and a 41% single-core miss rate against whole-section truth.
The binders come from therapeutic antibody programmes and the published imaging work sits with Chinese academic centres and their partners.
Companion-only by construction, but tied to two approved therapies in a disease with an estimated 980,000 new cases a year, and the CAR-T approval raises the cost of a false negative.
Gallium-68 works for two of the three published agents; the iodine-124 version would be harder to distribute.
Three groups are already in humans, all in China, so a Western programme would be following rather than leading the science.
- Chinese academic PET programmesfirst-in-human
68Ga-PMD22, 68Ga-NC-BCH and 124I-18B10(10L), each with a published first-in-human pilot of 11 to 17 patients.
https://pmc.ncbi.nlm.nih.gov/articles/PMC11149600/ - Astellasapproved
Zolbetuximab (Vyloy), the approved CLDN18.2 antibody, selected by the VENTANA 43-14A RxDx immunohistochemistry companion diagnostic.
https://www.prnewswire.com/news-releases/chinas-national-medical-products-administration-nmpa-approves-vyloy-zolbetuximab-for-first-line-treatment-of-advanced-gastric-or-gastroesophageal-junction-adenocarcinoma-302342446.html - CARsgen Therapeuticsapproved in China
Satricabtagene autoleucel, a CLDN18.2 CAR-T reported approved in China in June 2026 for the same biomarker-defined population.
https://chinamedwaypoint.com/resources/cancer-care/cancer-care-in-china/car-t-cell-therapy/ct041-satri-cel-cldn18-2-gastric-cancer-nmpa-approval - Roche Diagnosticsapproved companion diagnostic
The VENTANA CLDN18 (43-14A) RxDx assay, the incumbent companion diagnostic an imaging agent would be measured against.
https://www.cancerbiomed.org/content/22/3/212
Similar pages
not linked directly; found by shared links- OpportunityHER3 imaging, and why three tracers stalled
Shares Immuno-PET, PET/CT and the tags target crossover, companion diagnostic.
- OpportunityDLL3 imaging for small-cell lung cancer
Shares Immuno-PET, PET/CT, Gallium-68 and the tag companion diagnostic.
- OpportunityTROP2 imaging where no tissue test is required
Shares Immuno-PET, PET/CT, Gallium-68 and the tag target crossover.
- OpportunityGPC3 imaging of the indeterminate liver nodule, led from China
Shares Immuno-PET, PET/CT, Gallium-68.
- OpportunityCEACAM5 imaging for occult colorectal recurrence
Shares Immuno-PET, PET/CT, Gallium-68.
- TrialInvestigating the Safety and Diagnostic Performance of 68Ga-NYM032 Injection PET/CT in Men With Newly Diagnosed Prostate Cancer.
Shares PET/CT, Gallium-68.
- TrialThis Study Evaluates the Diagnostic Performance and Safety of 68Ga-NYM032 PET/CT Imaging in Patients With Suspected Recurrence of Prostate Cancer.
Shares PET/CT, Gallium-68.
- TargetHER2
Shares Gastric and gastro-oesophageal junction adenocarcinoma, Claudin-18.2, Immuno-PET.