Peptide Drugs for Alzheimer's After PRI-002's Phase 2 Miss and Semaglutide's EVOKE Failure
Priavoid's oral all-D-peptide missed its 48-week goal in a 304-person trial, with better numbers only in later, exploratory analyses. Novo Nordisk's 3,808-person semaglutide trials found no slowing at all. This guide explains both results, how their safety data relate to the approved antibodies, and how families can judge the next headline and find a trial.
Key points
- Priavoid's PRImus-AD Phase 2 trial of oral PRI-002 (304 people with MCI or mild dementia due to Alzheimer's) missed its primary efficacy goal, change in CDR-SB at week 48, according to topline results presented at CTAD in Boston on November 17, 2026
- The best-looking PRI-002 result, 35.9% less decline on the ADAS-Cog13 at week 72 (p=0.022), was exploratory and per protocol; the intention-to-treat result was 26.7% (p=0.102), and only one of ten reported secondary or exploratory comparisons reached p<0.05
- PRI-002 met its primary safety endpoint; ARIA-E occurred in 2% (300 mg), 0% (600 mg), and 1% (placebo), and ARIA-H in 8%, 11.7%, and 5%, but central MRI monitoring for ARIA was ended early, so these rates cannot be compared with antibody trials
- PRI-002, formerly called RD2, is an all-D-peptide from Forschungszentrum Jülich designed to break up amyloid-beta oligomers; most of its efficacy support comes from mouse studies, and the only earlier patient trial had 19 participants
- Novo Nordisk's evoke and evoke+ trials of oral semaglutide in 3,808 people with early Alzheimer's found no slowing on CDR-SB at 104 weeks, despite small changes in some biomarkers
- Records studies linking GLP-1 drugs to less dementia in people with diabetes remain hypothesis-generating; evoke tested people who already had Alzheimer's symptoms, most without diabetes
- No peptide drug is approved to slow Alzheimer's disease; ClinicalTrials.gov and the Alzheimer's Association's TrialMatch are the main ways to find trials
Priavoid's oral peptide missed its main goal at CTAD
On November 17, 2026, the German company Priavoid presented first results from PRImus-AD, a Phase 2 trial of its experimental Alzheimer's pill PRI-002, at the Clinical Trials on Alzheimer's Disease (CTAD) conference in Boston. The trial missed its main efficacy goal.
PRImus-AD enrolled 304 people aged 55 to 80 with mild cognitive impairment (MCI) or mild dementia due to Alzheimer's disease, at 38 sites in Czechia, Germany, Italy, the Netherlands, Poland, and Spain. It was double-blind, so neither patients nor study staff knew who got what. Participants were split evenly among 300 mg of PRI-002 a day, 600 mg a day, and a placebo.
The primary efficacy measure was the change after 48 weeks in the Clinical Dementia Rating Sum of Boxes (CDR-SB), an 18-point scale built from interviews with the patient and a care partner about memory, judgment, and daily life. Higher scores mean more impairment. PRI-002 did not beat placebo on it. The trial did meet its primary safety goal.
Priavoid's release leads instead with later, exploratory numbers that look better.
Why the week-72 numbers carry little weight
PRImus-AD used a common-close design: the study ended for everyone when the last person enrolled reached week 48. Early enrollees were treated for up to 96 weeks, so only they could contribute week-72 data. That smaller group was not what the trial was built to test.
Priavoid pooled the two doses, saying their blood levels overlapped heavily, and reported two analysis populations. The modified intention-to-treat (mITT) population counts randomized patients in their assigned group whether or not they followed the protocol, which keeps the groups comparable the way randomization made them. The per-protocol population keeps only people who followed the study rules closely. Because people who drop out or miss doses often differ from those who stay, per-protocol results can drift in either direction and are treated as supporting evidence.
On the ADAS-Cog13, a 13-item test of memory, language, and attention, PRI-002 showed 26.7% less decline than placebo at week 72 in the mITT population (p=0.102) and 35.9% less in the per-protocol population (p=0.022). Researchers usually treat p below 0.05 as unlikely to be chance alone, and the mITT result missed that bar.
The release gives p-values for ten comparisons beyond the primary endpoint: CDR-SB, ADAS-Cog13, and a daily-living scale (ADCS-ADL), at weeks 48 and 72, in both populations. One of the ten fell below 0.05. Test enough things and a few will cross that line by chance, which is why trials name one primary endpoint in advance. On the CDR-SB at week 72, PRI-002 showed 12% slowing in the mITT population (p=0.459).
So the numbers mostly leaned toward the drug without reaching significance, and only a larger trial could show whether any benefit is real. Priavoid plans to present the full dataset in 2027 and says it will use the results to decide on a possible next study. It also described preliminary changes in p-tau181, a blood marker of tau, as consistent with the drug's proposed action, without giving figures. Semaglutide's trials, covered below, show how far biomarker shifts can sit from patient benefit.
A mirror-image peptide aimed at oligomers
Amino acids come in two mirror-image forms, L and D, like left and right hands. Human proteins use the L form, and the enzymes that digest proteins are shaped to fit it. PRI-002 is built entirely from D-form amino acids. In lab tests under its earlier name, RD2, it resisted breakdown in simulated gut fluid, blood plasma, and liver enzyme preparations, which is how a peptide can work as a pill.
The compound came out of Dieter Willbold's lab at Forschungszentrum Jülich and Heinrich Heine University Düsseldorf. In 2008 the group described D3, a D-peptide found by mirror-image phage display that reduced amyloid plaque in mice; RD2 was a redesigned D3. Willbold co-founded Priavoid, co-owns it, and is a co-inventor of PRI-002.
Its target is amyloid-beta oligomers, small soluble clumps of amyloid protein that the Jülich group and others have proposed as the main cause of synapse damage. PRI-002 is meant to pull those clumps apart into single, harmless protein units. The approved antibodies work differently: lecanemab binds soluble clumps called protofibrils, and donanemab was designed to clear amyloid plaque.
Most of the efficacy case comes from mice. In a 2017 study, 28 days of RD2 improved learning in transgenic Alzheimer's mice without changing plaque load, and a 2019 study found oral RD2 reversed memory deficits and lowered brain oligomers in old mice with advanced disease.
The human record before PRImus-AD was small. Phase 1 trials in 64 healthy young volunteers found doses up to 320 mg a day for 28 days well tolerated. A single-center Phase 1b trial led by Oliver Peters of Charité in Berlin, now Priavoid's chief medical officer, randomized 19 people with early Alzheimer's to 300 mg a day or placebo for 28 days. No serious adverse events or ARIA occurred, and spinal-fluid amyloid and tau markers did not change. The 9 people on the drug did better than the 10 on placebo on a word-list memory test four weeks after stopping, a lead too small to stand alone.
ARIA: brain swelling and bleeding on MRI
ARIA stands for amyloid-related imaging abnormalities. ARIA-E is brain swelling (edema), and ARIA-H is small bleeds or iron deposits. Both are known risks of lecanemab (Leqembi) and donanemab (Kisunla), two anti-amyloid antibodies the FDA has approved for early Alzheimer's. The Leqembi label says ARIA usually causes no symptoms but can cause headache, confusion, vision changes, or seizures, and serious or fatal events can occur.
Priavoid reported these rates by the end of PRImus-AD:
| PRI-002 300 mg | PRI-002 600 mg | Placebo | |
|---|---|---|---|
| ARIA-E | 2% | 0% | 1% |
| ARIA-H | 8% | 11.7% | 5% |
The FDA labels, from each antibody's Phase 3 trial, list ARIA-E in 13% on lecanemab versus 2% on placebo, and 24% on donanemab versus 2%. ARIA-H was 17% versus 9% for lecanemab and 31% versus 13% for donanemab. The donanemab label reports lower ARIA with a newer, slower dose schedule.
Those figures do not show PRI-002 is safer. The trials enrolled different people and scanned them on different schedules. According to Priavoid's release, the unblinded safety monitoring board recommended ending close ARIA monitoring by central MRI reading after the first 90 patients reached 24 weeks; after that, ARIA counts came from investigators' adverse-event reports. Fewer scans can mean fewer detected cases. The antibody labels require a baseline MRI and repeat scans in the first months. PRI-002's ARIA-H rate at 600 mg was also more than double placebo's, a gap the full data should explain.
Efficacy can't be compared either. Lecanemab slowed CDR-SB decline by 0.45 points over 18 months in a 1,795-person trial (1.21 versus 1.66), and donanemab by 0.7 points over 76 weeks in a 1,736-person trial. Each met its primary goal in a trial built to test it. PRI-002 has not done that yet.
Semaglutide's 3,808-person test found no slowing
GLP-1 drugs are peptides too. Records studies linking them to less dementia in people with type 2 diabetes made them a leading peptide candidate for Alzheimer's. Our overview of GLP-1 drugs beyond weight loss covers that early rationale; the trial results here supersede its outlook.
Novo Nordisk tested oral semaglutide, the drug in the diabetes pill Rybelsus, in two Phase 3 trials, evoke and evoke+. They randomized 3,808 people aged 55 to 85 with amyloid-confirmed MCI or mild Alzheimer's dementia, at 566 sites in 40 countries, to up to 14 mg a day or placebo. Novo announced on November 24, 2025, that neither trial slowed progression, and the full results appeared in The Lancet on March 19, 2026.
After 104 weeks, CDR-SB scores had risen 2.3 points on semaglutide and 2.3 on placebo in evoke, and 2.2 versus 2.1 in evoke+. Adverse events were reported by 91.2% of people on semaglutide and 84.8% on placebo, a safety profile the authors called consistent with other uses.
Some biomarkers did move. Semaglutide lowered hsCRP, a blood marker of inflammation, and in a spinal-fluid substudy several markers of Alzheimer's biology fell. Jeffrey Cummings and Philip Scheltens, both on the trials' steering committee, later wrote that those spinal-fluid differences were 10% or less, lost significance after correction for multiple comparisons, and came with no slowing of decline.
Why did the records studies point the other way? They counted new dementia diagnoses in people with diabetes. Evoke measured progression in people who already had symptoms, and 86% of them did not have type 2 diabetes. Prevention remains an open question. A September 2026 records study of 1,320 matched adults with type 2 diabetes and MCI found new dementia in 15.5% of those starting a GLP-1 drug versus 20.4% of those starting a DPP-4 inhibitor, another diabetes drug class, at an average of 3.9 years (hazard ratio 0.74, meaning a 26% lower rate). The authors call it hypothesis-generating, since records studies can't rule out differences between the people given each drug.
An older GLP-1, liraglutide, did little better in the ELAD Phase 2b trial. In 204 people with mild to moderate Alzheimer's and no diabetes, a year of daily injections missed the primary goal, a change in brain glucose use on PET scans, and left CDR-SB unchanged, though one executive-function test favored the drug.
Semaglutide is not approved to treat or prevent Alzheimer's. People taking it for diabetes or weight shouldn't stop because of this news without talking to their prescriber.
Peptide vaccines came first
Earlier peptide approaches tried to train the immune system against amyloid or tau. Dosing of an early amyloid vaccine, AN1792, was halted in a Phase 2 trial after reports of brain inflammation (meningoencephalitis), which eventually affected 18 of 298 vaccinated patients (6%) and none of 74 on placebo; six still had disabling problems when the 2003 report was written. AADvac1, a tau peptide vaccine from Axon Neuroscience, was safe and raised antibodies in the 196-person ADAMANT trial but had no significant effect on cognition or function across the whole group. UB-311, an amyloid vaccine from Vaxxinity, raised antibodies in 97% of recipients in a 43-person Phase 2a trial in Taiwan, with ARIA microbleeds in 14% of participants.
PRI-002 takes a different route. Priavoid says it is designed to avoid treatment-driven immune responses, which the company links to ARIA, and the full scan data due in 2027 will be the first real test of that claim.
Checking the next Alzheimer's headline
A few questions sort most Alzheimer's results:
- Did it meet the primary endpoint? That is the goal researchers committed to before seeing data. PRI-002 and semaglutide both missed. A miss followed by a strong subgroup or later time point justifies another trial and nothing more.
- Intention-to-treat or per protocol? Trust the first more.
- How many people, for how long? The approved antibodies proved their effect in trials of roughly 1,700 to 1,800 people over about 18 months.
- Patients or biomarkers? A lower tau or inflammation marker is not slower memory loss, as evoke showed.
- Press release or journal? Topline results come from the company before peer review. Our guide to reading drug-company comparisons goes further.
- Mice or people? Much of the oligomer case, and many brain claims for wellness peptides, rests on animal data. Our neuropeptides overview grades Cerebrolysin, Semax, Dihexa, and others, and the tesamorelin article covers small cognition trials of growth-hormone-releasing hormone.
No peptide drug is approved to slow Alzheimer's disease. PRI-002 is available only inside clinical trials, and no version of it is approved for sale.
Where families can look for trials
PRI-002's registry record lists no expanded-access program, so a trial is the way to get drugs like it. ClinicalTrials.gov lets you search "Alzheimer Disease," filter to recruiting studies, and add a location; each record lists eligibility, sites, and the main outcome. PRImus-AD is NCT06182085, and evoke and evoke+ are NCT04777396 and NCT04777409. The Alzheimer's Association runs TrialMatch, a matching service for dementia studies, and a 24/7 helpline at 800-272-3900.
Questions for a study team: What phase is it, and what are the odds of placebo? How often are MRI scans done? What will it cost, including travel? What happens when the study ends? Involve the person's doctor, since eligibility can hinge on current treatment; PRImus-AD, for example, excluded people using an anti-amyloid antibody at the start.
Ongoing coverage lives on our Alzheimer's disease and semaglutide tag pages.
Updates and corrections
No corrections yet.
Limitations
- PRImus-AD results come from a company press release and conference presentation; the full dataset and peer-reviewed paper are not yet available
- The release gives percentages and p-values but not the number of participants in each analysis population or at week 72
- Priavoid gave no figures for its p-tau181 biomarker findings
- ARIA rates from different trials are not directly comparable because of differences in MRI schedules, populations, and how ARIA was recorded
- Most mechanistic evidence for PRI-002 comes from mouse models and lab studies by the developing group, whose lead scientist co-founded the company
- Observational studies of GLP-1 drugs and dementia cannot rule out differences between people who received different diabetes drugs
- This article does not cover every peptide or peptide-like approach under study in Alzheimer's disease
Citations
- 1. Oral PRI-002 treatment in patients with MCI or mild AD: a randomized, double-blind phase 1b trialRandomized Controlled Trial Nature Communications 2025
- 2.
- 3. Liraglutide in mild to moderate Alzheimer's disease: a phase 2b clinical trialRandomized Controlled Trial Nature Medicine 2026
- 4. Lecanemab in Early Alzheimer's DiseaseRandomized Controlled Trial New England Journal of Medicine 2023
- 5. Donanemab in Early Symptomatic Alzheimer Disease: The TRAILBLAZER-ALZ 2 Randomized Clinical TrialRandomized Controlled Trial JAMA 2023
- 6. ADAMANT: a placebo-controlled randomized phase 2 study of AADvac1, an active immunotherapy against pathological tau in Alzheimer's diseaseRandomized Controlled Trial Nature Aging 2021
- 7. Safety, tolerability, immunogenicity, and efficacy of UB-311 in participants with mild Alzheimer's disease: a randomised, double-blind, placebo-controlled, phase 2a studyRandomized Controlled Trial eBioMedicine 2023
- 8. Evoke (+) trials of semaglutide for early alzheimer's disease: Innovations, lessons, and implicationsReview Journal of Prevention of Alzheimer's Disease 2026
- 9.
- 10.
- 11. Safety and pharmacokinetics of the orally available antiprionic compound PRI-002: A single and multiple ascending dose phase I studyclinical-trial Alzheimer's & Dementia: Translational Research & Clinical Interventions 2020
- 12.
- 13. The Aβ oligomer eliminating D-enantiomeric peptide RD2 improves cognition without changing plaque pathologyPreclinical Scientific Reports 2017
- 14. Aβ Oligomer Elimination Restores Cognition in Transgenic Alzheimer's Mice with Full-blown PathologyPreclinical Molecular Neurobiology 2019
- 15. Metabolic resistance of the D-peptide RD2 developed for direct elimination of amyloid-β oligomersPreclinical Scientific Reports 2019
- 16.
- 17. Dementia progression with GLP-1 receptor agonists in people with mild cognitive impairment and type 2 diabetes: target-trial emulation studyOriginal Research Alzheimer's & Dementia: Translational Research & Clinical Interventions 2026
- 18. LEQEMBI (lecanemab-irmb) injection: prescribing informationregulatory DailyMed 2026
- 19. KISUNLA (donanemab-azbt) injection: prescribing informationregulatory DailyMed 2026
- 20. Subacute meningoencephalitis in a subset of patients with AD after Abeta42 immunizationclinical-trial Neurology 2003
- 21. TrialMatch: Find Clinical Research for Alzheimer's and Other Dementiareference Alzheimer's Association 2026
Peptides in this article
Full peptide profiles with evidence levels, dosing data, and safety notes live on peptidelist.org.
Related insights
Maryland Caps What Public Health Plans Pay for Ozempic at $274: How State Drug Affordability Boards Work
Maryland's drug affordability board set payment limits on Ozempic and Jardiance for state and local government plans starting in 2027. Who is covered, why copays may not change, why Colorado's Enbrel limit is stuck in court, and what to ask your plan.
GLP-1 Drugs and Eating Disorders: Who May Benefit, Who Is at Risk, and the New Screening Advice
A 60-week Penn trial found semaglutide eased loss-of-control eating and body-image worries in adults with obesity, while a 45-member expert panel says people with active anorexia or restrictive bulimia should generally avoid these drugs. Semaglutide is not approved to treat binge-eating disorder.
Federal Employees and GLP-1s in 2027: Required Behavioral Therapy, Higher Costs, and How to Pay Less
Every FEHB and PSHB plan will require documented intensive behavioral therapy before and during weight-loss GLP-1 treatment in 2027, and many plans are raising member costs to 50%. A guide for federal employees, postal workers, and retirees before Open Season ends December 14.