Metabolic research writing has a habit of treating mechanism and outcome as roughly the same thing. If the target is right and the compound hits it, the result is supposed to follow. PXL770 is a tidy counterexample: a compound that did everything it promised at the mechanistic level and still failed to carry a clinical endpoint. Working out why is more useful than reading another description of a "promising metabolic drug".

First: it is not a peptide

PXL770 routinely turns up in lists next to semaglutide or tirzepatide, and that is misleading. It is a small molecule, a thienopyridone derivative of 424 g/mol, developed by France's Poxel SA for oral administration. For comparison, tirzepatide is a peptide of roughly 4,800 g/mol that has to be injected precisely because gut enzymes would digest it before it did anything. The difference is not cosmetic: it determines route of administration, metabolism, the character of side effects, and which experiments are even possible.

What "direct" AMPK activation actually means

AMPK is the enzyme a cell switches on when fuel runs low. ATP falls, AMP rises, AMPK activates and flips metabolism over: it stops building fat (suppressing de novo lipogenesis), starts burning it, and restrains growth programmes through mTORC1. The idea of throwing that switch pharmacologically is old, much of the enduring interest in metformin rests on it.

But metformin does not activate AMPK directly. On the dominant model it inhibits mitochondrial respiratory complex I; the cell's energy balance deteriorates, and AMPK switches on in response. You are not pressing the button, you are engineering a small energy deficit and hoping the button presses itself. The consequence is a wide spread of collateral effects, because what you changed is not the enzyme but the cell's entire energetics.

PXL770 was built differently: the molecule binds the AMPK complex itself and stabilises its active conformation without changing AMP levels. That is what made it interesting, "first-in-class direct activator". A clean pharmacological question: if you switch AMPK on precisely and selectively, do you get metabolic benefit?

The answer: mechanism yes, endpoint no

In the phase 1b study published in Cell Reports Medicine, PXL770 did reduce de novo lipogenesis in people with insulin resistance and fatty liver disease. The mechanism engages in humans. That is not trivial, most compounds die well before this point.

Then came phase 2a: the STAMP-NAFLD (non-alcoholic fatty liver disease) trial, published in Lancet Gastroenterology & Hepatology. 120 participants, 15 US sites, 12 weeks, three dosing regimens against placebo. Primary endpoint: hepatic fat content.

  • Placebo: −1.1%
  • 250 mg once daily: −1.0% (no effect at all)
  • 250 mg twice daily: −14.3%, p=0.084
  • 500 mg once daily: −14.7%, p=0.076

The trial did not meet its primary endpoint. And this is where it gets interesting as a lesson in reading science news. A 13–14 percentage-point separation from placebo looks substantial. But p=0.076 means that at this sample size the result cannot be confidently distinguished from chance. Press releases at the time called the data "positive", leaning on secondary measures, liver enzymes, glycaemia in the diabetic subgroup. That is not technically false. But the primary endpoint is what the investigators committed to in advance, and failing it is the result of the trial.

Adverse events: diarrhoea in 17–23% versus 0% on placebo. No life-threatening events.

What a company does when the mechanism is right and the indication is not

What happened next is more common in biotech than people assume: Poxel did not discard the compound, it changed the disease. The NASH (non-alcoholic steatohepatitis) programme was wound down and PXL770 was redirected to rare disease.

Kidney International (2023) published preclinical data in autosomal dominant polycystic kidney disease: in a Pkd1 knockout mouse model, blood urea fell 47% and cystic index 26%, alongside reductions in proliferation, macrophage infiltration and fibrosis. The Journal of Pharmacology and Experimental Therapeutics (2022) reported X-linked adrenoleukodystrophy data: very-long-chain fatty acid C26:0 down roughly 90% in patient-derived cells. The FDA (U.S. Food and Drug Administration) granted Orphan Drug Designation for ADPKD and Fast Track for X-ALD.

The logic is sound. In NAFLD the metabolic disorder has a dozen causes, and a modest shift in lipogenesis is not enough. In a monogenic disease where one specific pathway is broken, that same modest shift may be decisive. But none of these data come from humans. They are cells and mice. Orphan designation is an accelerated regulatory pathway, not evidence of efficacy, and the two should not be conflated.

The takeaway

PXL770 is a good teaching case for reading any literature on research compounds. Target engagement and clinical efficacy are two different claims, and the first does not guarantee the second. "Positive press release" and "primary endpoint met" are also two different claims. And preclinical data in a mouse model, however striking the percentages, remain a hypothesis until a randomised human trial exists.

The full breakdown with primary-source citations is in the PXL770 monograph. A related case, a small molecule with a metabolic target that has not reached the clinic at all, is SLU-PP-332. The metabolic compound catalogue is here.

See also: a general overview of PXL770 as a direct AMPK activator.

PXL770 is a research-use-only laboratory reagent. It is not a medicine, is not approved by any regulator for human treatment, and is not intended for human consumption or veterinary use. This material is reference information, not medical advice.