In 2012, neuroscientist Gary Landreth and his team at Case Western Reserve University published a finding in Science that electrified the Alzheimer’s research world: bexarotene, a pill already FDA-approved for a rare skin cancer, appeared to strip beta-amyloid plaques from the brains of Alzheimer’s-model mice within 72 hours and restore their memory. Four years later, immunologist Michal Schwartz’s laboratory at the Weizmann Institute of Science reported in Nature Medicine that anti-PD-1 checkpoint inhibitors, the blockbuster cancer immunotherapy drugs nivolumab and pembrolizumab, could rally the immune system to clear those same plaques in a different set of transgenic mice. Both discoveries carried the same tantalizing implication: drugs with known safety profiles, already manufactured at scale, might leapfrog years of early development and reach Alzheimer’s patients far sooner than any purpose-built molecule.
As of June 2026, neither drug has come close to fulfilling that promise. Independent laboratories failed to reproduce the core plaque-clearing results for both compounds, a small human pilot trial of bexarotene showed no amyloid reduction on brain imaging, and the FDA has since approved two entirely different anti-amyloid antibodies, lecanemab and donanemab, that now define the competitive landscape. The story of these two repurposed cancer drugs is less a tale of breakthrough than a case study in how the distance between a striking mouse experiment and a working human therapy remains stubbornly wide.
The original findings and why they mattered
Landreth’s 2012 paper described a surprisingly clean mechanism. Bexarotene activates retinoid X receptors, which boost production of ApoE and the cholesterol transporter ABCA1. Together, these proteins help package soluble beta-amyloid for removal from the brain. In multiple transgenic mouse strains engineered to accumulate amyloid, the drug reduced both soluble amyloid-beta levels and visible plaques, and the treated mice performed better on standard memory tests such as the Morris water maze. An NIH Research Matters summary highlighted the speed of the effect, noting that an existing cancer drug had reversed amyloid accumulation in days rather than the months typical of experimental Alzheimer’s therapies.
The checkpoint-inhibitor work, published in 2016, took a different biological route. Schwartz’s group showed that blocking PD-1, a receptor that tumors exploit to shut down immune attack, triggered an interferon-gamma-dependent immune response in Alzheimer’s-model mice. That response recruited monocyte-derived macrophages into the brain, where they appeared to engulf and clear amyloid deposits. The treated animals showed reduced plaque burden and improved cognition. Because anti-PD-1 antibodies were already transforming oncology, with pembrolizumab alone approved for more than a dozen cancer types, the notion of redirecting them toward neurodegeneration attracted immediate attention from both researchers and pharmaceutical companies.
Both findings drew their urgency from the same structural advantage. Drug repurposing sidesteps the years and hundreds of millions of dollars required to establish basic human safety, because toxicity data, dosing ranges, and manufacturing protocols already exist. For Alzheimer’s disease, where novel drug candidates have failed in late-stage clinical trials at rates exceeding 99 percent over the past two decades, starting with a compound that has already cleared the FDA’s safety bar is not just convenient. It is, in principle, a way to compress a 12-to-15-year development timeline into something closer to five or six years.
Where the evidence broke down
The bexarotene story fractured within a year of publication. In 2013, Science published results from four independent laboratories that attempted to replicate Landreth’s findings. As Nature reported, some groups observed modest reductions in soluble amyloid-beta, but none could reproduce the dramatic plaque clearance that had generated headlines. Researchers at the University of Chicago, who participated in the replication effort, stated that plaque number and plaque area were unchanged in their follow-up experiments. The teams explored whether differences in mouse strains, drug formulations, or imaging techniques might explain the discrepancy, but no technical variable rescued the original effect.
A small human trial added another layer of disappointment. Researchers at the Cleveland Clinic conducted a pilot study (registered as NCT01782742) in which Alzheimer’s patients received bexarotene and underwent amyloid PET imaging. The drug did not reduce brain amyloid signal. While the trial was small and not powered to detect subtle changes, the result removed the most direct piece of translational evidence that bexarotene’s mouse findings might carry over to people.
The PD-1 checkpoint story followed a parallel arc. A replication study published in the journal Glia in 2018, involving researchers from multiple laboratories including teams at three pharmaceutical companies, found that systemic anti-PD-1 treatment did not alter brain amyloid pathology in several transgenic mouse models. The investigators also reported no evidence of monocyte-derived macrophage infiltration into the brain, directly contradicting the mechanism Schwartz’s group had proposed. These negative results did not rule out every possible immune-mediated benefit of checkpoint blockade in neurodegeneration, but they undercut the specific plaque-clearance claim that had driven the excitement.
Safety concerns compounded the scientific uncertainty. Anti-PD-1 antibodies at oncology doses can provoke severe immune-related adverse events: colitis, hepatitis, pneumonitis, and endocrine dysfunction among them. Administering such agents chronically to older adults with cognitive impairment, a population already vulnerable to infection and inflammation, would demand a risk-benefit calculation that only strong efficacy data could justify. Without reproducible evidence of plaque reduction or cognitive improvement in animals, that justification does not exist.
The changed landscape around these drugs
While the repurposing debate played out, the broader Alzheimer’s field moved. The FDA granted accelerated approval to lecanemab (marketed as Leqembi) in January 2023 and full traditional approval later that year, based on Phase 3 data showing both amyloid plaque clearance on PET imaging and a statistically significant slowing of cognitive decline. Donanemab received approval in 2024 after its own Phase 3 trial demonstrated similar plaque removal and clinical benefit. Both are monoclonal antibodies designed specifically to bind aggregated amyloid-beta, and both require intravenous infusion and regular MRI monitoring for side effects such as amyloid-related imaging abnormalities.
These approvals reshape the context for any repurposed plaque-clearing therapy. A candidate entering clinical development now would need to show not just that it reduces amyloid, but that it does so as effectively as, or more conveniently than, drugs already on the market. Bexarotene, an oral pill, would have a practical advantage over infused antibodies if it worked, but the replication failures and the negative pilot trial make that “if” very large. Anti-PD-1 antibodies, which are themselves infused and carry significant side-effect profiles, would face an even steeper comparative hurdle.
One nuance deserves attention. The ApoE pathway that bexarotene targets intersects with the strongest known genetic risk factor for late-onset Alzheimer’s: the ApoE4 allele, carried by roughly 25 percent of the population and present in about 60 percent of Alzheimer’s patients. If a future drug could modulate ApoE-mediated amyloid clearance without bexarotene’s replication problems, it might address a biologically distinct patient subgroup. Similarly, the broader question of whether immune checkpoint modulation can influence neurodegeneration has not been closed by the PD-1 replication failures. Observational studies in cancer patients receiving checkpoint inhibitors have produced mixed signals about dementia incidence, and those data are too preliminary and confounded to draw conclusions from.
What this means for patients and families now
Neither bexarotene nor anti-PD-1 antibodies can be considered evidence-based treatments for Alzheimer’s disease. No professional guideline endorses their off-label use for dementia, and physicians prescribing them today do so exclusively for cancer indications. Patients or caregivers who encounter claims about these drugs online should weigh the full evidence record: initial positive mouse studies that independent groups could not replicate, a negative human pilot for bexarotene, and no human Alzheimer’s data at all for anti-PD-1 therapy.
The most honest reading of this research is that it illuminated two potentially important biological pathways, lipid-mediated amyloid transport and immune-mediated plaque clearance, without delivering a usable therapy through either one. Drug repurposing remains a legitimate and sometimes powerful strategy; the antidiabetic drug semaglutide, for instance, is now being studied in a large Alzheimer’s prevention trial after epidemiological signals suggested cognitive benefit. But the bexarotene and PD-1 episodes illustrate a hard truth about the approach: a known safety profile is only half the equation. Without reproducible efficacy, the shortcut leads nowhere.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
