Picture the moment a trial begins for one person: not a scanner humming in an academic medical center, but a chair, a cuff, a tube of blood drawn at a private neurology practice an hour from anywhere. For the pivotal Phase 3 study of buntanetap in early Alzheimer's disease, that tube is the gate. Among the inclusion criteria is a plasma pTau217 result consistent with Alzheimer's pathology, and the two design abstracts Annovis Bio presented at the 2025 Alzheimer's Association International Conference, published in the Alzheimer's & Dementia supplement on December 23, 2025, say the quiet part out loud: the company built patient selection around a blood draw rather than imaging.
The design itself is plainly stated in those abstracts, which are conference reports rather than peer-reviewed full papers. The trial plans to randomize 760 participants across up to 100 sites in the United States, academic and private, recruited through site databases, physician referrals and word of mouth. Participants are assigned one to one to placebo or 30 mg of buntanetap once daily. It runs as two linked studies: a six-month stage testing symptomatic effects on cognition, then an eighteen-month stage testing for potential disease-modifying effects, with ADAS-Cog13 and ADCS-iADL as the primary endpoints. Buntanetap is an oral molecule that inhibits translation of several neurotoxic aggregating proteins, including amyloid precursor protein and tau. It is investigational, approved nowhere, and nothing about its enrollment machinery speaks to whether it works.
Here is the honest shape of the funnel story. The operations abstract, authored by Annovis clinical staff, lists what a poster would display as of July 2025: a map of recruiting sites, pTau217 selection inclusion and exclusion figures, enrollment against the 760 target, monthly accrual against projection, and a screen failure rate broken down by reason. Those are exactly the numbers a field starved of real-world biomarker screening data wants. The published abstract contains the list, not the figures. So the funnel has been described in unusual detail, and not yet disclosed. Readers should hold that distinction firmly: a promised table is not a reported result.
Why build the door this way at all? The answer sits in the company's completed Phase 2/3 study, published in npj Dementia on April 27, 2026. That three-month, randomized, double-blind, placebo-controlled trial randomized 351 people with mild to moderate Alzheimer's to 7.5 mg, 15 mg, 30 mg or placebo. It missed both primary endpoints, and the authors attribute that in large part to the fact that about 40 percent of participants lacked amyloid pathology; the trial had enrolled on clinical criteria, with no biomarker gate. Only later, once commercial plasma testing became available, did the sponsor sort participants by a pTau217 to total-tau ratio at or above 4.2 percent, as an earlier conference abstract from January 2025 describes. In that biomarker-positive mild subgroup the drug showed dose-dependent cognitive gains on ADAS-Cog11. Those analyses were post hoc, the p values nominal, and a subgroup rescued after the fact is a hypothesis, not a finding.
The Phase 3 is that hypothesis wearing a seatbelt, and it filled up fast. Annovis announced on July 7, 2026 that the trial had reached full enrollment with 850 patients at 83 United States sites, above the registered target of 760, with top-line six-month data guided to the first quarter of 2027 and eighteen-month data to the first quarter of 2028. That is a company statement, not peer-reviewed data, and the trial remains blinded. What it does suggest is logistical: when the entry test is a blood draw a community practice can order, recruitment stops being the exclusive business of PET-equipped centers.
For a family weighing a trial, the practical meaning is narrow and worth saying precisely. A blood biomarker is being used here to decide who is eligible for a study, which is not the same as a diagnosis, a prognosis, or a reason to expect benefit. Screening in means a research team believed your biology matched the question they are asking. It does not mean the drug will help, and the people who screen out are, in the published record so far, an unreported number. Whether blood screening really is the efficient front door the sponsor describes is a claim that can only be settled by the tables that have not been printed yet, and by two readouts still to come.