Selling an answer to Ozempic: the underpowered bet at the heart of Fractyl Health

The arithmetic was unforgiving. On January 29th 2026 Fractyl Health, a small Massachusetts maker of medical devices, published six-month results from the main study of Revita, its only product. Revita is an endoscopic procedure that ablates the lining of the duodenum, so that people who have lost weight on GLP-1 drugs can stop taking them without the pounds returning. Management called the data compelling. Shareholders, who until that morning had been willing to believe it, saw it differently: the shares fell 68% on the day, a further 21% the next, and closed near $0.46—roughly a quarter of their value two days earlier.

Revita answers a genuine and growing problem. The most successful drug class of the age, which includes Wegovy, Ozempic and Zepbound, has a retention problem: many patients quit within a year, and published third-party studies put the typical weight regain at about a tenth of the loss once they do. Fractyl's pitch is that a single outpatient procedure can preserve that loss without a lifetime of injections. Facing a market of millions expected to stop GLP-1 therapy in 2026, the company has promised a "definitional" year. Until the data landed, the Street treated the thesis as probable rather than possible: Morgan Stanley carried an $8 price target and an overweight rating.

A signal in search of significance

The trouble is how thin the evidence beneath that optimism turned out to be. The January results came from the "midpoint cohort" of the REMAIN-1 study, which the company itself conceded was not sufficiently powered for a definitive read-out. In the efficacy population of forty, patients who received Revita regained 4.5% of their weight against 7.5% for the sham arm, a difference carrying a p-value of 0.07 on a one-sided test—by any sober standard a trend, not a finding. Five participants had been excluded for diet-and-lifestyle non-compliance. Analysts had anchored on a bar of roughly a 50% relative reduction in regain; the data fell short of it.

Then came the wrench. Morgan Stanley cut its target from $8 to $2, downgraded the stock to equal-weight, lowered its subjective probability that Revita would succeed from 50% to 35% and trimmed its modelled risk-adjusted peak sales from about $700m to $490m. Canaccord Genuity raised the more awkward question: an "outlier site". One of the six trial sites—the first to enroll—had not yet set up a dietary-counselling programme, and Fractyl attributed that site's odd results to a "relatively less robust diet and lifestyle counseling program." In a trial of forty patients one wayward site can move the whole signal; nobody can yet say whether the effect is real or partly a sampling artefact.

Compensation, then the real question

Enter the lawyers. A securities class action filed on behalf of anyone who bought between January 13th 2025 and January 29th 2026 alleges that Fractyl overstated Revita's clinical, regulatory and commercial prospects and failed to disclose that operational problems at trial sites compromised the integrity of the results; investors have until October 20th to apply to lead the case. Be clear about what such an action is and is not. It is a mechanism for redistributing an already-realised capital loss—evidence that the loss was large and that counsel see exposure, not a verdict on the biology. No judge will decide whether Revita works; that is for a clinic, not a court.

The company, meanwhile, has quietly shrunk to its single bet. It held about $85.6m in cash early this year, enough on its own guidance to reach early 2027, but it has already paused its type-2-diabetes study and a real-world registry to preserve that money. The honest test is close: topline six-month data from the pivotal REMAIN-1 cohort is expected early in the fourth quarter, feeding a possible FDA marketing application later in the year.

None of this kills the thesis. A clean pivot read-out would reopen a genuine market in millions, and the biology may yet survive the arithmetic of one ill-prepared site. But it explains how the trade is priced either way: a whole company now rests on a single trial, funded by a balance sheet that reaches only to early 2027, in which one wayward site can decide the signal. The class action will, if it settles, return a small fraction of the loss to some of those who bore it. It will not decide whether Revita works. That answer costs nothing to wait for, and it is imminent.