Skip to content
clinicaltrialshub.com
Clinical Trials Hub

Clinical trials, drug development and research news

Neurology

Daily Alzheimer's pill slowed decline by 31%, but only for some

In the 1,535-person POLARIS-AD trial, once-daily AR1001 slowed early Alzheimer's by 31% in people on no other dementia drugs, yet missed in the whole group.

The patterned concrete facade of Stanford University Medical Center, home of POLARIS-AD global principal investigator Sharon Sha.
Photo: Noahedits, via Wikimedia Commons · CC BY-SA 4.0
Contents
  1. What did POLARIS-AD test?
  2. How much did AR1001 slow decline?
  3. A worked example with the CDR-SB numbers
  4. Why did it only work in one group?
  5. How is AR1001 supposed to work?
  6. How does it compare with Leqembi and Kisunla?
  7. What about side effects?
  8. What happens next for AR1001?
  9. Can you get AR1001 or join a trial?

A once-daily tablet called AR1001 slowed the decline of early Alzheimer's disease by 31% over a year in people who were taking no other Alzheimer's medicine, according to topline results from the phase 3 POLARIS-AD trial. In the full trial of 1,535 people, which also included patients on standard symptom drugs, the benefit was small and could have been down to chance.

The South Korean company AriBio, which makes AR1001 and paid for the study, presented the results at the Clinical Trials on Alzheimer's Disease (CTAD) conference in Boston on 16 November 2026. The trial is counted as a success because it was designed with two primary groups, and the drug only had to win in one of them. That design detail matters for how much weight the 31% can carry, and we come back to it below.

AR1001 is mirodenafil, a drug from the same family as Viagra. It has been sold in South Korea for erectile dysfunction since 2007, according to Fierce Pharma, and AriBio is testing a lower daily dose of 30 mg for Alzheimer's. It is not approved for Alzheimer's anywhere.

What did POLARIS-AD test?

POLARIS-AD (NCT05531526) enrolled people aged 55 to 90 with mild cognitive impairment or mild dementia caused by Alzheimer's disease. Everyone had to have proof of amyloid in the brain, from a spinal fluid test or an amyloid PET scan, and a Mini-Mental State Examination (MMSE) score of 20 or more out of 30. That is an early-stage group, broadly similar to the people treated with the antibody drugs Leqembi and Kisunla.

Participants were randomly assigned to AR1001 30 mg or a dummy tablet once a day for 52 weeks. Neither they nor their doctors knew which they were getting. AriBio says the study ran at about 230 sites in 13 countries, including the United States, EU member states, South Korea, China, Canada and the United Kingdom, and that 87.9% of participants completed the year. Of those, 95.2% chose to continue into a 52-week extension in which everyone takes the drug.

The main measure was the Clinical Dementia Rating Sum of Boxes (CDR-SB). It scores six areas of everyday life, including memory, orientation, judgement, community affairs, home and hobbies, and personal care, on a scale from 0 to 18. Higher scores mean more impairment. Change in early Alzheimer's is slow on this scale, so treatment effects are measured in fractions of a point.

The trial had two prespecified primary populations:

  • everyone randomised, called the overall population (1,535 people)
  • the 604 people who were not taking a cholinesterase inhibitor such as donepezil, or memantine, when they joined, called the monotherapy population

Fierce Pharma reports that the statistical plan split the allowance for false positives between the two, so the overall group needed a p-value below 0.02 and the monotherapy group below 0.03. A win in either one would count.

How much did AR1001 slow decline?

Measure at week 52 Monotherapy group (604) Overall group (1,535)
CDR-SB difference versus placebo -0.44 points (p=0.0266) -0.11 points (p=0.4015)
Slowing of decline 31% about 7% (not significant)
MMSE difference versus placebo 0.7 points (p=0.0232) numerically favoured AR1001
ADAS-Cog 13 difference versus placebo -1.16 points (p=0.0798) numerically favoured AR1001

Figures from AriBio's press release, except the 7% overall slowing, which was reported by BioSpace and Fierce Pharma. AriBio did not give exact values for the overall group's secondary measures.

The monotherapy result met its bar of 0.03 with little room to spare. The overall result was nowhere near significance. On the MMSE and the ADAS-Cog 13, two tests of memory and thinking, the monotherapy group again leaned towards the drug, with only the MMSE crossing the usual 0.05 threshold. AriBio calls that MMSE result nominally significant, which means it was not part of the formal chain of tests that protects against chance findings.

A worked example with the CDR-SB numbers

Percentages like 31% can sound larger than they feel. Here is a rough way to turn them back into points, using only the two numbers AriBio published for the monotherapy group.

If a 0.44-point difference is 31% of the decline seen on placebo, then the placebo group worsened by about 0.44 divided by 0.31, which is roughly 1.4 points over the year. People on AR1001 would then have worsened by about 1.0 point. On an 18-point scale, that is a gap of less than half a point after 12 months.

This is our back-calculation, and AriBio has not yet published the raw group averages, so treat it as an estimate. It does show why researchers argue about whether effects of this size are noticeable to families. Our guide to reading trial results in absolute terms explains the same issue for risk figures.

Why did it only work in one group?

Nobody knows yet, and the trial was not built to answer that question. The monotherapy group was a planned primary population, which gives it more standing than a subgroup picked after the results came in. It is still a slice of the trial defined by what other medicines people happened to be taking at the start.

People who are already on donepezil or memantine may differ from those who are not, for example in how long they have had symptoms or how they were diagnosed. AriBio has not released baseline details for the two groups, so it is not possible to say whether that explains the gap. A second possibility is that the standard drugs blunt AR1001's effect in some way. A third is chance, since a p-value of 0.0266 still leaves room for a fluke.

Dr Richard Oakley of Alzheimer's Society, in the charity's response, said the drug "doesn't seem to be as effective for people taking other drugs for Alzheimer's symptoms, so this needs further investigation". Dr Sheona Scales of Alzheimer's Research UK told The Guardian the benefit "was not significant across the overall trial group". Prof Tara Spires-Jones of the University of Edinburgh described the results as "very promising".

Sharon Sha of Stanford University, the trial's global principal investigator, said in AriBio's release that the secondary endpoints and biomarker findings "warrant further evaluation alongside the overall population outcome". If you want more background on how mixed results like this are judged, see our explainer on what it means when a clinical trial fails.

How is AR1001 supposed to work?

PDE5 inhibitors stop an enzyme from breaking down a chemical messenger called cyclic GMP (cGMP). More cGMP relaxes the muscle in blood vessel walls. AriBio's case is that a version that crosses into the brain can raise cGMP there, improve blood flow and act on nerve cell survival, inflammation and tau, the protein that forms tangles inside brain cells.

The spinal fluid results gave partial support. At week 52, cGMP in spinal fluid was higher on AR1001 than on placebo in both groups (p<0.0001 overall). A marker of tau, p-tau217, fell slightly more on AR1001, with a difference of -5.34 pg/mL in the monotherapy group (p=0.0402) and -3.52 pg/mL overall (p=0.1784). The amyloid ratio in spinal fluid did not change, which fits a drug that does not target amyloid plaques.

Interest in this drug family partly comes from observational research. A 2024 study in Neurology of 269,725 men in UK primary care with erectile dysfunction found that those prescribed PDE5 inhibitors had an 18% lower rate of later Alzheimer's diagnosis than those who were not. Studies like that cannot prove cause and effect. POLARIS-AD was the first large randomised test of the idea.

How does it compare with Leqembi and Kisunla?

The monotherapy result is in the same range as the antibody drugs on paper. BioSpace notes that Leqembi (lecanemab) gave a 0.45-point CDR-SB difference, or 27% slowing, at 18 months in its phase 3 Clarity AD trial, and Kisunla (donanemab) gave 0.7 points, or 29%, at 76 weeks in TRAILBLAZER-ALZ 2. Fierce Pharma points out that about 60% of people in Clarity AD were taking symptom drugs at baseline, and the antibody results held across the whole trial.

There are big practical differences too. Leqembi and Kisunla are given by infusion or injection, need MRI scans to watch for brain swelling and bleeding called ARIA, and remove amyloid from the brain. AR1001 is a tablet taken at home. AriBio's release does not mention ARIA, and BioSpace notes that none was seen in the earlier phase 2 study.

In the UK, the MHRA licensed both antibody drugs in 2024. NICE said in June 2025 final draft guidance that their benefits were too small to justify the cost to the NHS, and that appraisal has since been reopened, The Pharmaceutical Journal reports. A cheap daily pill would change those cost sums, but only if its benefit holds up in a further trial.

What about side effects?

Over the year, side effects were reported by 69.0% of people on AR1001 and 66.4% on placebo, and serious adverse events were close at 10.0% and 9.8%. Most were mild or moderate. The differences showed up in a few specific complaints:

Side effect AR1001 Placebo
Diarrhoea 7.7% 4.4%
Falls 7.3% 5.9%
Dizziness 6.1% 3.1%
Stopped the drug because of a side effect 5.2% 1.7%

Four people on AR1001 (0.5%) and two on placebo (0.3%) died during the study. AriBio has not said what caused those deaths or whether investigators judged any of them related to the drug.

Falls and dizziness deserve attention in a drug for older adults, since PDE5 inhibitors can lower blood pressure. The trial excluded people with recent heart attacks, unstable angina, serious heart failure or very low blood pressure, and anyone already taking another PDE5 inhibitor such as sildenafil, according to the ClinicalTrials.gov record. Real-world patients would be less carefully selected.

What happens next for AR1001?

AriBio says it will review the complete data and discuss them with the US Food and Drug Administration. It has not said whether it will file for approval on these results. The outside experts quoted by The Guardian all called for further trials before the drug could be prescribed.

There is a commercial decision pending as well. Fosun Pharma paid AriBio $60 million in May for an option to license AR1001 in the US, Europe and Japan, and Fierce Pharma reports that it has 90 days after receiving the full POLARIS-AD data to decide. The open-label extension is due to finish in the third quarter of 2027, which should show whether any effect lasts into a second year.

POLARIS-AD sat alongside other CTAD 2026 news, including two-year data on Biogen's tau-targeting drug diranersen, covered in pharmaphorum's round-up. We have also reported on other attempts to treat Alzheimer's without targeting amyloid, such as the cromolyn and ibuprofen COGNITE trial.

Can you get AR1001 or join a trial?

Not at the moment. AR1001 is unlicensed for Alzheimer's in every country, and POLARIS-AD is no longer recruiting. Alzheimer's Society has warned that people should not take Viagra in an attempt to treat Alzheimer's, because AR1001 is being developed and dosed specifically for the disease and Viagra has not been tested for it.

If you or a relative has early Alzheimer's and wants to know about trials, these are useful questions for a memory clinic:

  • Has the diagnosis been confirmed with an amyloid scan, spinal fluid test or blood test, which most trials now require?
  • Would starting donepezil or memantine affect which trials I could join later?
  • Are there trials of tablets or non-amyloid drugs recruiting near me?
  • If I am offered a place, how often would I need to visit, and could a family member come with me?

Our guide to finding and joining a clinical trial explains how to search UK and international registers, and our explainer on placebo and blinding covers what happens if you are given the dummy tablet. For how a drug like this would reach the NHS, see how new medicines get approved by the FDA, MHRA and NICE.

This article reports early trial results for general information. It is not medical advice. Do not start, stop or change any medicine without talking to the doctor who manages your care.

Frequently asked questions

What is AR1001?

AR1001 is mirodenafil, a PDE5 inhibitor from the same drug family as Viagra. It has been sold in South Korea for erectile dysfunction since 2007. AriBio is developing a once-daily 30 mg tablet for early Alzheimer's disease, on the theory that it raises a brain messenger called cGMP, improves blood flow and protects nerve cells. It is not approved for Alzheimer's in any country.

Did the POLARIS-AD trial succeed?

Technically yes, by its own rules. The trial had two primary populations and needed a win in only one. In 604 people taking no other Alzheimer's drug, AR1001 slowed decline on the CDR-SB scale by 31% over 52 weeks. In all 1,535 participants the difference was about 7% and not statistically significant, so the result is mixed and needs confirming.

Should I take Viagra or sildenafil to protect against Alzheimer's?

No. Alzheimer's Society has warned against this. AR1001 is a different drug at a dose chosen for Alzheimer's, and even its trial result is not yet confirmed. Observational studies linking erectile dysfunction drugs with lower Alzheimer's rates cannot show cause and effect. These drugs also lower blood pressure and interact with some heart medicines, so only take them as prescribed.

Is AR1001 better than Leqembi or Kisunla?

There is no head-to-head trial, so nobody can say. The monotherapy result, a 0.44-point CDR-SB difference at one year, looks similar on paper to the antibody drugs, but they showed benefit across their whole trial populations over 18 months or longer. AR1001 would be easier to take, as a tablet with no infusions or ARIA monitoring scans.

When could AR1001 be available in the UK?

Not soon. AriBio first plans to discuss the data with the FDA, and has not announced any filing in the US, UK or Europe. A confirmatory trial may be needed. In the UK, the MHRA would have to license it and NICE would then judge whether the NHS should pay for it, a process that usually takes well over a year after a filing.

Sources

  1. AriBio's Oral AR1001 Meets Primary Efficacy Objective in Prespecified Monotherapy Population in Phase 3 POLARIS-AD Trial. AriBio via GlobeNewswire, 2026-11-16
  2. Phase 3 Trial to Evaluate the Efficacy and Safety of AR1001 in Participants With Early Alzheimer's Disease (Polaris-AD), NCT05531526. ClinicalTrials.gov, 2026
  3. New Alzheimer's drug AR1001 shows promising results in Phase 3 clinical trial. Alzheimer's Society, 2026-11-17
  4. AriBio's oral Alzheimer's drug meets monotherapy goal in phase 3 as Fosun weighs licensing option. Fierce Pharma, 2026-11-17
  5. CTAD 2026: AriBio's Alzheimer's candidate slows cognitive decline as monotherapy. BioSpace, 2026-11-17
  6. Pill that mimics Viagra can slow Alzheimer's progression, study suggests. The Guardian, 2026-11-18
  7. AriBio's oral Alzheimer's drug nets Phase III win. Clinical Trials Arena, 2026-11-17
  8. Mixed results for experimental Alzheimer's therapies at CTAD. pharmaphorum, 2026-11-19
  9. Phosphodiesterase Type 5 Inhibitors in Men With Erectile Dysfunction and the Risk of Alzheimer Disease. Neurology, 2024-02
  10. Final draft guidance finds benefits of 2 Alzheimer's treatments remain too small to justify the additional costs to the NHS. NICE, 2025-06

Medical disclaimer: For information only, not medical advice. Speak to a qualified healthcare professional about your care.

The Briefing

Trial results that matter, every Friday.

One email. The week's readouts, approvals and research, explained without hype.

Related reading