Davunetide missed both main endpoints in a 313-patient PSP trial

What is davunetide?

Davunetide (NAP) is an eight-amino-acid fragment of activity-dependent neuroprotective protein (ADNP). The PSP trial report gives the rationale for testing it: in preclinical studies, davunetide promoted microtubule stability and reduced tau phosphorylation, and progressive supranuclear palsy (PSP) is linked to tau pathology ^[5].

The hypothesis is coherent and testable: stabilize microtubules, reduce the hyperphosphorylated tau that drives PSP pathology, and slow functional decline. The cited studies never tested it end to end. None directly measured NAP in human CSF or brain tissue after intranasal dosing, so the step between administration and target engagement is inferred, not measured.

Did davunetide slow progressive supranuclear palsy?

Davunetide did not slow PSP: across 313 patients, it did not differ from placebo on either co-primary endpoint after 52 weeks ^[5]. The randomized, double-blind, placebo-controlled Phase 2/3 trial ran across 48 centers, gave intranasal davunetide (30 mg twice daily) or placebo, and analyzed patients by intention to treat. Neither the PSP Rating Scale (PSPRS) nor the Schwab and England Activities of Daily Living scale (SEADL) changed differently from placebo ^[5].

The PSP trial is the highest evidence grade the intranasal neuroprotective peptide category has produced, and it came back flat.

Did davunetide show any benefit in earlier trials?

Davunetide produced scattered secondary-endpoint signals in amnestic MCI and schizophrenia, and none of them is confirmatory ^[1]⁠^[6]⁠^[7].

Amnestic mild cognitive impairment. Before PSP, davunetide was tested in 144 people with amnestic MCI across eight dose and placebo arms. A later sex-stratified reanalysis of that dataset found dose-dependent improvement in men on a delayed visual matching-to-sample test, and a high-dose improvement over placebo in women on a digit span test of working memory and attention ^[1]. The result is a real, reported statistical finding, not vendor marketing. It comes from a secondary sex-stratified analysis rather than a pre-registered confirmatory endpoint, and no large trial has tried to replicate it.

Schizophrenia. A trial in 63 subjects found no significant difference between davunetide and placebo on the primary cognitive battery (p=.45). It did find a significant effect on a functional-capacity measure, the UPSA (p=.048), with effect sizes of .74 and .48 at the 5 mg and 30 mg doses ^[6]. A magnetic resonance spectroscopy substudy of 18 of those subjects found a non-significant trend increase in a neuronal-integrity marker, alongside a significant increase in a choline marker at the high dose (p=.040) ^[7].

None of these results is confirmatory-grade evidence, and none is nothing. Davunetide has scattered secondary-endpoint signals from three smaller analyses in two populations, and one large, clean, negative primary result. For comparison, intranasal insulin shows no cognitive benefit in pooled analysis.

Was PSP the right disease for a davunetide trial?

PSP was a defensible choice: a tauopathy whose decline is fast enough to measure within a one-year trial window, which matters for trial feasibility.

The same AL-108-231 dataset was later used to calculate that a 5.7-point worsening on the PSPRS over six months is the minimal clinically meaningful change, which confirmed that the primary endpoint tracked real functional decline rather than noise ^[2]. The dataset also supported later volumetric MRI progression models, in which third ventricle, midbrain, and frontal lobe volumes showed the largest, most sample-efficient effect sizes for future trials ^[3]. CSF neurofilament light chain and phosphorylated tau were also shown to predict PSPRS and SEADL decline in PSP more broadly ^[4].

Trial infrastructure sound enough for the field's later biomarker work argues against "bad endpoint" as the explanation for davunetide's failure.

Has any drug slowed PSP?

No disease-modifying trial in atypical parkinsonism, including PSP, had produced a convincing positive result when a review of interventional trials assessed the category ^[8]. The review places davunetide's PSP failure alongside a separate GSK-3 inhibitor trial that also failed in PSP ^[8].

That record reframes the davunetide result. Davunetide joined a list of PSP candidates that have not moved outcomes, which is a different claim from "davunetide's underlying biology does not work." For a disease-modification trial in Parkinson's disease itself, see why exenatide does not slow Parkinson's disease.

Does the PSP failure prove davunetide's mechanism does not work?

The PSP failure does not prove davunetide's microtubule-stabilization mechanism is inert; it shows that this peptide, dose, route, and population did not beat placebo ^[5]. CNS exposure was never confirmed, so mechanism failure and delivery failure are confounded in the result ^[5].

The route's actual brain exposure remains unconfirmed rather than proven. How nasal peptides reach the brain covers the narrow routes from nose to brain.

Treating the trial as proof that the mechanism is inert, rather than proof that one trial in one disease did not separate from placebo, is the error nearly every consumer summary of davunetide makes.

Is davunetide still being developed?

Davunetide development is limited: the most recent cited study is a 2024 reanalysis of the old MCI dataset for sex-specific effects ^[1]. None of the cited studies reports a confirmatory Phase 3 program in a new population.

What is still unknown about davunetide?

Three questions about davunetide remain open:

• Brain delivery. Whether intranasal dosing delivered pharmacologically meaningful NAP concentrations to the relevant brain regions during the PSP trial was not measured in any of the cited studies.
• The MCI signal. Whether the sex-stratified MCI signal would replicate in a properly powered, pre-registered confirmatory trial is untested since the original analysis ^[1].
• Other tauopathies. None of the cited studies addresses whether a different tauopathy, a different tau isoform burden, or an earlier disease stage would respond differently to microtubule stabilization.

The topline PSP verdict is not open. Davunetide was flat on both co-primary endpoints, in the largest and most rigorous trial the molecule ever received ^[5].