Belapectin in MASH cirrhosis: a subgroup that did not survive its own confirmatory trial

What was tried

Galectin Therapeutics ran NAVIGATE (NCT04365868), a seamless adaptive Phase 2b/3 trial of belapectin, an intravenous galectin-3 inhibitor, in patients with MASH cirrhosis and portal hypertension who had no esophageal varices at baseline. The study randomized 357 patients to belapectin 2 mg/kg, 4 mg/kg, or placebo, with the proportion developing new esophageal varices at 78 weeks as the primary endpoint. The registry records termination with the note that "Based on FDA feedback, Stage 1 of NAVIGATE was analyzed as a stand-alone trial," and that the study "was not prematurely terminated for safety reasons." The design was built to confirm a subgroup finding from the earlier Phase 2b trial.

The biological hypothesis

Galectin-3 is a carbohydrate-binding protein that promotes hepatic stellate cell activation and macrophage-driven fibrosis. Belapectin is a complex polysaccharide that binds and inhibits it (ChEMBL CHEMBL4297577). In rodents, GR-MD-02 reduced liver fibrosis and portal pressure, which framed the clinical bet: inhibit galectin-3, slow fibrosis, and lower the portal hypertension that drives varices (Chalasani et al, 2020). The human genetics never matched the mouse story. Open Targets scores the LGALS3 to MASH association at just 0.12, carried almost entirely by literature with no genetic association evidence. The target was a fibrosis correlate with strong preclinical data and thin human anchoring.

What actually happened

The prior Phase 2b trial had already missed its primary endpoint. In 162 patients, belapectin produced no significant change in hepatic venous pressure gradient (2 mg/kg minus 0.28 mm Hg versus placebo plus 0.10, p=1.0; 8 mg/kg minus 0.25 versus 0.10, p=1.0) and no effect on fibrosis (Chalasani et al, 2020). The one positive note came from a subgroup of 81 patients without varices at baseline, where 2 mg/kg reduced portal pressure (p=0.02) and new varices (p=0.03). NAVIGATE enrolled exactly that subgroup. It did not replicate. New varices at 78 weeks occurred in 56 percent of placebo patients against 45 percent on 2 mg/kg and 51 percent on 4 mg/kg, a favorable trend that did not reach significance in the intent-to-treat population (NCT04365868 results). A nominal per-protocol signal in the 2 mg/kg arm survived, but the confirmatory primary did not.

Failure mechanism, best guess

This is a textbook subgroup-chase failure layered on a target with weak human validation. The Phase 2b primary endpoint was flatly negative on portal pressure, the most direct readout of the mechanism. The positive subgroup arose by partitioning a failed trial, and a confirmatory study built on it inherited the selection. The biology compounds the risk. Galectin-3 is upregulated in fibrotic liver, but upregulation of a marker does not prove it is a driver that a single antagonist can switch off. With no human genetic support linking LGALS3 to MASH or its complications, the program leaned entirely on preclinical pharmacology and a 81-patient subgroup. The drug was safe and the trend was in the right direction, which is precisely the pattern that keeps marginal antifibrotic programs alive longer than the evidence warrants.

How to prevent this next time

Two quantitative tools would have set the bar higher before NAVIGATE was funded.

First, a Bayesian predictive probability gate that accounts for selection. The probability that a subgroup effect replicates is

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

When the interim data is a subgroup drawn from a trial whose primary endpoint failed, the prior must shrink the apparent effect toward the null, and the resulting predictive probability is low. Second, an honest power and interval calculation. The observed 11-point varices difference (56 versus 45 percent) at roughly 118 patients per arm carries a Wald 95 percent confidence interval running from about minus 2 to plus 24 percentage points, which includes zero. The trial was sized for a subgroup-inflated effect, not the modest difference that real biology supports. Layer on the base rate that drug targets with human genetic support succeed about twice as often as those without (Nelson et al, 2015), and a galectin-3 program scoring 0.12 at Open Targets needed a much steeper threshold of proof.

The single highest leverage change would have been to treat the no-varices subgroup as a hypothesis rather than a result, power NAVIGATE for the realistic 10-point difference, and gate the program on a Bayesian predictive probability that penalized the selection.

What this means for similar programs

Antifibrotic targets that are abundant in diseased tissue but absent from human genetics are a recurring trap, and galectin-3 in MASH is now a documented example. The Claidex Mechanism Risk Score for LGALS3 is 42 (yellow), with most of the weight coming from genetic deficit (Open Targets 0.12) rather than phase or saturation. A new galectin-3 fibrosis program would need either human genetic evidence or a mechanistic biomarker that tracks target modulation to portal pressure, not a tissue correlate and a salvaged subgroup. Surrogate endpoints like new varices are slow and noisy, which makes underpowered confirmatory trials especially costly.

Open questions

• Does the per-protocol 2 mg/kg signal reflect a real responder population, or adherence-driven confounding.
• Is there any genetically supported fibrotic indication for galectin-3 inhibition, or is the target a marker throughout.
• Would a direct portal-pressure endpoint, rather than new varices, have given a faster and cleaner read on mechanism.

Sources

1. Effects of Belapectin, an Inhibitor of Galectin-3, in Patients With Nonalcoholic Steatohepatitis With Cirrhosis and Portal Hypertension. Chalasani N, Abdelmalek MF, Garcia-Tsao G, et al. Gastroenterology, 2020.

2. Study Evaluating the Efficacy and Safety of Belapectin for the Prevention of Esophageal Varices in NASH Cirrhosis (NAVIGATE). ClinicalTrials.gov NCT04365868, posted results. https://clinicaltrials.gov/study/NCT04365868.

3. Galectin Therapeutics Announces Top-Line Results of NAVIGATE. Galectin Therapeutics Inc, 2024. https://investor.galectintherapeutics.com/news-releases/news-release-details/galectin-therapeutics-announces-top-line-results-navigate.

4. Nelson MR, Tipney H, Painter JL, et al. The support of human genetic evidence for approved drug indications. Nature Genetics, 2015.

5. Open Targets Platform. LGALS3 to metabolic dysfunction-associated steatohepatitis association evidence. https://platform.opentargets.org/evidence/ENSG00000131981/MONDO_0007027.

6. ChEMBL. Belapectin compound record (CHEMBL4297577), galectin-3 inhibitor. https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL4297577/.