Executive Perspective
After decades of Alzheimer’s research, the challenge may be bigger than finding the next therapeutic target. The history of failed trials, a pipeline increasingly extending beyond amyloid and tau, advances in patient selection, and the emergence of AI raise a broader strategic question: Can we better identify which biology matters in which patient, at which stage of disease—and make better scientific, clinical and investment decisions as a result?
Drawing on more than three decades in pharmaceutical and biotechnology commercialization, including neuroscience and specialty therapeutics, I look at Alzheimer’s development from a broader strategic perspective: how disease biology, patient selection, clinical development, capital allocation and commercial viability come together to determine which programs ultimately reach patients.
Thirty Years Later
For decades, Alzheimer’s research has focused heavily on two biological hallmarks: amyloid plaques and tau tangles.
There are good reasons for that. Both are central features of Alzheimer’s pathology, and recent therapies have demonstrated that reducing amyloid can modestly slow cognitive and functional decline in some patients.
But after decades of research, billions of dollars invested and an extraordinary number of clinical trials, Alzheimer’s remains one of medicine’s most difficult diseases.
Donepezil was approved in 1996. Since then, our ability to study Alzheimer’s has changed dramatically.
We have PET imaging, increasingly sophisticated blood biomarkers, genetic information, better characterization of disease progression and therapies capable of substantially reducing amyloid.
Yet we still cannot stop or reverse Alzheimer’s.
A review of Alzheimer’s Phase II and III clinical development from 2004 through 2021 identified 98 failed compounds. Of those, 63 were intended to modify disease and 35 were symptomatic treatments. Among the disease-modifying failures, 23 targeted the amyloid pathway and seven targeted tau.
Perhaps the most interesting number is the remaining 33 disease-modifying failures involving other mechanisms.
| Category | Failed compounds | Strategic implication |
|---|---|---|
| Amyloid-related | 23 | Repeated failures occurred in the most heavily pursued disease-modifying pathway, although newer amyloid antibodies have subsequently demonstrated clinical benefit |
| Tau-related | 7 | Another central feature of Alzheimer’s pathology has proved difficult to translate into successful therapy |
| Other disease-modifying mechanisms | 33 | Moving beyond amyloid and tau has not, by itself, solved the problem |
| Symptomatic approaches | 35 | Improving cognition or symptoms without altering the underlying disease has also proved difficult |
| Total | 98 | Alzheimer’s drug-development failure has crossed many different mechanisms |
What the Failures Tell Us
The numbers challenge an overly simple explanation of the industry's history.
It isn't just that researchers spent decades pursuing amyloid and should now find another target.
Many other targets have already been tried.
That makes me question the assumption that Alzheimer's will ultimately yield to another single mechanism.
Perhaps the problem is more complicated.
More Than Amyloid and Tau
Alzheimer’s involves an interconnected set of biological processes: amyloid and tau pathology, neuroinflammation and microglial activity, synaptic dysfunction, oxidative stress, vascular and metabolic dysfunction, impaired protein regulation and declining neuronal resilience.
These processes don't necessarily operate independently. More importantly, they may not contribute equally in every patient or at every stage of disease.
That changes the question.
Instead of continuing to ask only which target causes Alzheimer’s, perhaps we should also be asking:
What is driving progression in this particular patient, at this particular stage of disease?
That distinction could have significant implications for how we develop drugs and design clinical trials.
The Current Pipeline Is Already Broader
The 2026 Alzheimer's pipeline reinforces this point. A published review identified 158 drugs being evaluated across 192 clinical trials as of January 1, 2026, spanning amyloid, tau, inflammation, synaptic plasticity, metabolism, proteostasis and numerous other mechanisms.
Several current programs illustrate the range.
| Program | Development status | Approach | Why it matters |
|---|---|---|---|
| AR1001 — AriBio | Phase III | PDE5 inhibition; synaptic plasticity and neuroprotection | Tests a non-amyloid approach centered on neuronal and synaptic function |
| Buntanetap — Annovis Bio | Phase III | Regulation of APP, tau and α-synuclein translation | Addresses production of multiple neurotoxic proteins rather than a single target |
| Zervimesine (CT1812) — Cognition Therapeutics | Phase II | Sigma-2 receptor modulation / synaptic protection | Attempts to interfere with toxic amyloid oligomers at neuronal synapses rather than simply remove plaque |
| Metformin — academic/NIA-supported research | Phase III prevention research | Metabolism / insulin sensitization | Examines whether metabolic biology can influence progression toward dementia |
None of these programs tells us which approach will succeed.
What interests me is what the pipeline itself is telling us.
Researchers are already looking well beyond a simple amyloid-versus-tau debate.
The Patient May Matter as Much as the Target
By the time Alzheimer's dementia becomes clinically apparent, substantial neuronal damage may already have occurred. Successfully modifying one component of the disease at that point may not be enough to restore neurological function that has already been lost.
This makes earlier intervention increasingly important.
But treating earlier creates its own challenges.
Who will progress? How quickly? Which biological processes are driving that progression? Which patients are most likely to respond to a particular mechanism?
These aren't theoretical questions. They affect trial enrollment, study duration, sample size, development cost and ultimately whether a promising therapy ever reaches patients.
Buntanetap provides an interesting example. Its earlier Phase II/III study did not meet its primary endpoints in the overall population. Approximately 40% of participants lacked amyloid pathology. In a biomarker-positive subgroup with mild disease, investigators reported signals that supported further evaluation. The current Phase III study therefore uses p-tau217 to help confirm Alzheimer’s pathology.
Whether buntanetap ultimately succeeds is unknown.
But the development experience illustrates something larger:
Patient selection can change what a clinical trial is actually testing.
Having spent much of my career evaluating what happens between clinical development and commercialization, I see these decisions as part of the same continuum.
Scientific rationale, trial design, patient selection, development time, capital requirements and eventual clinical adoption are not separate questions. Each affects the probability that an investigational therapy ultimately becomes a viable medicine.
A compelling mechanism is essential, but it is only the beginning. The right patients still have to be identified and enrolled, the clinical effect has to be meaningful, and the investment required to reach an answer has to remain supportable.
Where AI Could Change Alzheimer’s Development
This is where AI becomes particularly interesting.
I don't think the important story is that AI is going to “solve Alzheimer’s.” Its more immediate value may be helping researchers make sense of a disease that produces far more information than humans can easily integrate.
Genetics, blood biomarkers, PET and MRI imaging, cognitive testing, medical history, vascular and metabolic factors and potentially digital measures of speech and behavior can increasingly be considered together rather than as separate pieces of information.
That could have very practical consequences.
Better patient identification could mean fewer screening failures, faster enrollment and more biologically consistent study populations. AI may also help researchers understand why a therapy appears to work in one group of patients but produces only a modest effect when those patients are averaged together with everyone else.
An NIA-highlighted 2025 machine-learning study of screening for early Alzheimer’s trials estimated that its approach could reduce the number of people requiring additional screening by 55%, shorten recruitment by 13 months, and reduce amyloid PET screening scans by 72%.
Thirteen months matters.
It matters first to patients waiting for new treatments.
But it also matters to the company developing the drug. Thirteen months can affect development expense, cash runway, financing requirements, patent life and the amount of capital at risk before a meaningful clinical or regulatory milestone is reached.
For a smaller biotechnology company, the difference can be substantial.
A promising program can have compelling biology and still become extremely difficult to finance if development takes too long, requires repeated capital raises or fails to identify the right patients early enough.
That is why the potential return from AI should not be viewed only in scientific terms.
If AI can improve patient selection, reduce unnecessary screening, shorten recruitment, produce cleaner trials and help companies reach better decisions earlier, it could improve the productivity and return on R&D investment.
AI may eventually help with something even more fundamental: identifying patterns across biomarkers, genetics, imaging and clinical characteristics that tell us which patients are most likely to progress and which biological processes matter most in those patients.
It isn't going to establish causality or rescue a poor drug. Clinical trials still have to demonstrate that a therapy works. AI models are also only as reliable as the data used to build and validate them.
But better information earlier in development has value—even when the answer is that a program should not proceed.
That is an important part of capital allocation too.
A Different Way to Think About Alzheimer’s Development
After decades of searching for the next Alzheimer’s target, perhaps the bigger opportunity is figuring out which biology matters in which patient, and when.
The future may become much more patient-specific: understand the individual's disease biology and stage first, then determine which intervention makes the most sense.
Amyloid matters. Tau matters.
So may inflammation, synaptic function, vascular and metabolic health, oxidative stress, protein regulation and neuronal resilience.
Their relative importance may also change as the disease progresses.
The future of Alzheimer's treatment may therefore be less about discovering the Alzheimer's drug and more about determining which intervention, or eventually which combination of interventions, is appropriate for which patient and at which point in the disease.
There is also an investment implication.
For companies, boards and investors evaluating Alzheimer's programs, the question cannot be limited to whether the science is compelling. It also has to include whether the patient population can be identified, whether the trial can generate a meaningful answer, how much capital and time will be required to reach that answer, and whether the potential clinical value justifies that investment.
That isn't science versus commercialization.
It is development strategy.
Thirty years after the approval of donepezil, we have become remarkably good at measuring individual components of Alzheimer's disease.
Perhaps the next major advance will come from understanding how those pieces fit together—and using that understanding to make better scientific, clinical and investment decisions.
Are we still looking for one answer to a multi-process disease? And could AI help us ask a better question?
I would be interested in hearing how scientists, clinicians, drug-development leaders, executives and investors see it.
Joseph McCoy is Founder & Principal Consultant of PharmaKonsult.com, advising biotechnology and pharmaceutical companies on commercialization strategy, clinical-to-commercial planning, market development, organizational readiness and capital-informed development decisions.
About Joseph McCoy
Joseph McCoy is Founder and Principal Consultant of PharmaKonsult and an executive commercial strategy advisor with more than three decades of pharmaceutical and biotechnology commercial leadership experience across neuroscience, rare disease, endocrinology, institutional healthcare and specialty pharmaceuticals.
