An Alzheimer's trial's published results lead with a subgroup, not the full sample

A peer-reviewed paper on Annovis Bio's Alzheimer's drug buntanetap was published April 27 in npj Dementia, part of the Nature portfolio [s1]. The company's own announcement, issued the next day, frames the results around a specific subgroup: patients who tested positive for pTau217, a blood biomarker associated with earlier-stage disease, showed "statistically significant, dose-dependent improvements in cognition," as measured by the ADAS-Cog11 scale [s2]. What the announcement does not state is how the drug performed against its own pre-specified primary endpoints across the trial's full population — a gap worth noting on its own terms.

The trial

The randomized, double-blind, placebo-controlled study enrolled 351 patients with mild to moderate Alzheimer's disease and tested three doses of buntanetap — 7.5 mg, 15 mg, and 30 mg — over 12 weeks [s2]. The trial's primary endpoints, established before enrollment, were ADAS-Cog11 (a standard cognitive assessment scale) and ADCS-CGIC (a clinician-rated global impression of change) [s2].

What was reported, and what wasn't

The company's characterization of the published results centers entirely on the pTau217-positive subgroup of patients with mild disease specifically, describing dose-dependent cognitive improvement on ADAS-Cog11 as statistically significant within that group [s2]. Neither the publication announcement nor Annovis's press materials specify the corresponding result for ADAS-Cog11 or ADCS-CGIC in the trial's full 351-patient population — the population the primary endpoints were originally defined to test [s2]. No p-values, effect sizes, or percentage improvements for the subgroup finding are given in the materials available at publication [s2].

That omission does not mean the full-population result was negative — this article cannot confirm that either way from the sources verified here. But a 12-week trial's own announcement of its peer-reviewed publication leading with a subgroup analysis, rather than the pre-specified primary-endpoint result in the full enrolled population, is itself informative about which result the company considers its strongest finding to lead with.

The biomarker angle

pTau217 has emerged in recent years as one of the more promising blood-based biomarkers for identifying Alzheimer's pathology, and positivity for it is generally associated with more clearly established amyloid and tau pathology than a diagnosis based on clinical symptoms alone. A finding that benefit concentrates in biomarker-confirmed patients is a plausible and increasingly familiar pattern in Alzheimer's drug development — several other therapies in the field have likewise shown clearer effects in biomarker-defined subgroups than in clinically diagnosed populations overall. That pattern lends some biological plausibility to the subgroup finding here, even without the full population data in hand.

Safety

Buntanetap was reported as safe and well-tolerated across all three doses and both disease stages studied [s2]. Safety was assessed separately in ApoE4 carriers — the genotype associated with the highest genetic risk for Alzheimer's disease — and showed no increase in adverse events compared with non-carriers or placebo [s2]. That is a meaningful reassurance for a drug intended for long-term use, independent of how the efficacy question ultimately resolves.

Why the framing matters beyond this one drug

Alzheimer's drug development has increasingly moved toward biomarker-defined populations rather than diagnosis by symptoms alone, on the theory that a drug's true effect can be diluted or masked when trial cohorts include patients whose cognitive symptoms stem from something other than the amyloid and tau pathology a given drug targets. That shift is scientifically defensible, and it is also commercially convenient: a positive subgroup finding, reported without the corresponding full-population result, is a pattern that can make an ambiguous or disappointing trial read as a success in press coverage. Readers cannot tell, from the announcement alone, which of those two explanations — meaningful biological subgroup effect, or selective emphasis of the more favorable result — better describes this trial without reading the full-population data in the underlying paper.

What would resolve the open question

The full-population primary endpoint results should be contained within the peer-reviewed paper itself, even where the company's own summary does not restate them. Readers and clinicians evaluating buntanetap's development program have reason to look at that underlying publication directly, rather than at summary announcements, before drawing conclusions about how the drug performed against the endpoints the trial was designed to test.

This article is informational and is not medical advice.

Sources

• [s1] "Buntanetap treatment in mild to moderate Alzheimer's disease: phase 2/3 study." npj Dementia, 27 April 2026. https://doi.org/10.1038/s44400-026-00073-z
• [s2] Annovis Bio, "Annovis Publishes Phase 2/3 Alzheimer's Trial Results in Nature Portfolio," 28 April 2026. https://www.annovisbio.com/press-release/annovis-publishes-phase-2-3-alzheimers-trial-results-in-nature-portfolio