Reversing a precancerous stomach lining, and the organoid assay built on the pathway the drug blocks

What the work claims

This is a randomized, placebo-controlled, triple-masked early phase II trial, registered as a treatment study rather than a modeling study, which is precisely what makes its organoid component interesting. The clinical claim: pyrvinium, at 2 mg/kg/day by mouth for 14 days, reduces the burden of gastric intestinal metaplasia (GIM) with high-risk features compared with placebo. The primary endpoint is the change in the number of clinical biopsies showing GIM between a baseline research endoscopy and a follow-up endoscopy 4 to 8 weeks later; one-year histology secondaries include complete versus incomplete GIM, return of parietal cells as a marker of reversal of atrophy, and inflammation.¹

The modeling claim rides along as a secondary aim: generate gastric organoids from the same patients, test pyrvinium in vitro in patient-derived tissue, and use the organoids to identify new drugs that reverse GIM.¹ That aim converts a chemoprevention trial into a paired in-vitro and in-vivo design, where the organoid result and the biopsy endpoint can be compared within the same donor.

The structure is staged. Roughly 60 participants enter an observational cohort for standardized gastric mapping; from those, 32 with confirmed extensive high-risk GIM are randomized, 16 to pyrvinium and 16 to placebo. The drug is an FDA-approved agent no longer in active use, given as a liquid suspension, and the protocol warns participants about sun sensitivity, a sign the team expects real pharmacology at this dose.¹

How it works

GIM is a transdifferentiation event: gastric mucosal cells adopt an intestinal phenotype, and the intestinalized, atrophic stomach is the substrate from which intestinal-type gastric cancer usually arises. Reversal, if achievable, means pushing differentiated cells back toward a gastric fate, or removing the field that keeps regenerating the wrong one.

Pyrvinium is an anthelmintic that was found, in a well-known 2010 chemical-biology study, to inhibit Wnt signaling by activating casein kinase 1alpha.² Wnt is not incidental to the stomach; it is the maintenance signal of the gastric stem-cell compartment. The Lgr5-positive stem cells that build the gastric epithelium, and from which gastric organoids are derived, were defined by their capacity to drive self-renewal and to build long-lived gastric units in vitro, in Wnt-supplemented culture.³ In other words, the pathway pyrvinium suppresses is the same pathway the culture system feeds on.

That overlap is the mechanism story of this trial from an organoid perspective. In the stomach in vivo, a transient Wnt block might prune an aberrant stem-cell niche while leaving enough reserve for repair. In a gastric organoid, Wnt pathway activation is a culture requirement, so the drug meets no reserve, only dependency. The dose that reverses metaplasia in a patient and the dose that simply kills the assay are closer together in a dish than in an organ.

Where a skeptic should push

First, the biology of the endpoint. Fourteen days of drug, measured as the change in biopsy count 4 to 8 weeks later, is a large bet on plasticity. Intestinal metaplasia develops over years; whether a two-week exposure can measurably shrink a field, as opposed to suppressing a marker, is an open question. The one-year secondaries, especially parietal-cell return, are the outcomes that would actually evidence reversal of atrophy, and they will take patience to read.¹

Second, sampling. The primary endpoint is a count of biopsies showing GIM, and biopsy counts are exquisitely sensitive to where the endoscopist samples. The registry itself notes that clinical care is variable because mapping and surveillance guidelines are lacking, which is exactly why the trial builds a standardized research mapping protocol into its observational cohort. Even so, with 16 patients per arm, a modest true effect can drown in sampling noise; the randomization and triple masking are what give the trial its power, and they deserve credit for that.¹

Third, the in-vitro arm carries its own confound, and this is the point a modeling audience should hold onto. If pyrvinium kills gastric organoids through on-target Wnt inhibition, then the organoid screen will read toxicity as efficacy, and any new drug identified by the same screen will be selected for the same property: the ability to poison a Wnt-dependent culture. A reversal screen that scores cell identity rather than viability would dodge this, but the registry does not say the screen will be scored that way. What is registered is the intent, not the scoring logic.¹

What this changes for organoid chemoprevention

For organoid models of human organs, REVISE is a template and a warning at once. The template: a chemoprevention trial with same-patient organoids is the rare design in which the clinical endpoint and the in-vitro readout share a denominator, so a concordance analysis becomes possible within donors. If the trial reports it, that pairing will be one of the cleanest tests yet of whether an organoid can predict a histologic reversal that takes a year to appear in biopsies.

The warning generalizes beyond pyrvinium. Any organoid screen of a drug that touches a pathway the culture system depends on, Wnt, R-spondin, EGF, Notch, BMP, measures a shifted therapeutic window. The dish is not a neutral window into the tissue; it is a tissue plus its feeding regime, and the feeding regime is a drug target more often than people like to admit. Screens for stemness-pathway inhibitors are the most exposed, because stemness pathways are, almost by definition, what keeps organoids proliferating in Matrigel.

The opportunity, if the team reads it correctly, is to use the organoid as a cell-identity instrument rather than a killing assay: score gastric versus intestinal markers, mucin phenotypes, and parietal-cell lineage return, which is what the clinical trial itself measures. The threat is that a viability-based organoid readout quietly redefines reversal as cytotoxicity, produces numbers that look predictive, and teaches the field a false lesson about a class of drugs that were never meant to be poisons.

The bottom line

Established: a well-masked, placebo-controlled phase II trial exists, is recruiting at Duke, and pairs 14 days of pyrvinium with standardized mapping and a one-year histologic follow-up, plus a patient-organoid arm.¹ Hypothesis: GIM is pharmacologically reversible within weeks, and organoids can identify the next reversal drug. What would confirm the claim: a significant between-arm change in GIM biopsy burden at 4 to 8 weeks that persists, and ideally strengthens, at one year, with parietal-cell return. What would break it: a null primary endpoint, or an organoid arm whose readout cannot distinguish reversal from simple toxicity. The trial's completion is estimated for late 2026, so this is a near-term, checkable bet, and the organoid community should be watching how the in-vitro arm is scored more closely than what it finds.¹