Alzheimer’s Follows Three Distinct Paths, and Scientists Are Starting to Tell Them Apart

Seven in ten. That’s roughly the share of people who have amyloid plaques accumulating silently in their brains, one of Alzheimer’s earliest biological signatures, who showed no measurable cognitive decline over nearly six years of careful monitoring. Not gradual decline. Not subtle, hard-to-catch slippage. None at all. It’s a finding that should probably recalibrate the way most of us picture what early Alzheimer’s actually looks like, long before symptoms arrive.

The preclinical phase of Alzheimer’s has become medicine’s most fiercely contested terrain. This is the period, often spanning a decade or more, before the disease announces itself to the person living with it, when biomarkers betray what the mind hasn’t yet revealed. Drugs are trialled here, predictions made, hopes raised and dashed. And now, published this month in Alzheimer’s & Dementia, researchers at the University of Southern California have made a rather inconvenient argument: the conventional picture of this phase, everyone sliding slowly downhill at roughly the same rate, is almost certainly wrong.

Three Trajectories, Not One Slow Slope

The team analysed data from two linked studies. The A4 trial enrolled participants who tested positive for amyloid beta buildup but had no cognitive symptoms yet; LEARN, a companion observational study, tracked amyloid-negative individuals for comparison. Across several years of follow-up, participants completed batteries of memory, attention and reasoning tests. Rather than averaging those results together, as most studies do, the researchers applied latent class mixed-effects models, statistical machinery that searches for hidden subgroups with genuinely different trajectories. Three groups emerged. About 70% of amyloid-positive participants stayed cognitively stable across the study period. A second group declined slowly. A third, smaller still, deteriorated at a significantly faster clip.

“Most studies look at the average across participants, which can make it seem like everyone is slowly getting worse at the same rate,” said Michael Donohue, professor of neurology at the Keck School of Medicine of USC (Keck School of Medicine faculty directory) and the study’s corresponding author. “But we found that this approach masks major differences between people, suggesting that Alzheimer’s disease is more variable than often depicted.”

What distinguished the declining groups from the stable majority wasn’t just behaviour; it showed up in blood and brain tissue. Participants who declined, whether slowly or quickly, started the study with higher levels of phosphorylated tau 217 (p-tau217 for short), a protein fragment that has become one of the more promising blood-based markers in the field. They also showed more tau burden on PET brain scans and smaller hippocampi. That seahorse-shaped structure, tucked deep in the temporal lobe and essential for forming new memories, is among the first casualties of Alzheimer’s pathology, and its shrinkage apparently tracked with who was heading for trouble.

Using those biomarkers in combination, the team could correctly sort participants into stable or declining categories roughly 70% of the time. Which is genuinely promising, at least as a proof of concept. “P-tau217 was one of the strongest signs of which participants would decline, but we still cannot predict exactly how an individual person’s disease will progress,” Donohue acknowledged.

Why the Average Has Been Misleading Clinical Trials

The implications for drug development are, to put it plainly, rather a lot to absorb. Prevention trials for Alzheimer’s typically enrol people with amyloid buildup and then measure average cognitive change across the whole group over the study period. But if around 70% of those participants aren’t declining anyway, any genuine drug effect gets diluted, swamped in the collective stability of people who would have been fine regardless. Runpeng Li, a postdoctoral scholar at Keck and the study’s first author, put the practical problem clearly: “Many people with Alzheimer’s remain stable over the course of a study, which can make it hard to tell if a treatment is working. Identifying those who are more likely to decline could make trials more efficient and more informative.” The suggestion, implicit but pointed, is that some high-profile Alzheimer’s prevention trials may have returned negative or ambiguous results partly because of this design flaw, enrolling too many stable participants and then wondering why the signal was so weak.

Li also noted, separately, that the findings call into question how prevention trials are structured from the ground up: the field, he said, may need to rethink its approach to trial design in preclinical disease more broadly. That’s a significant thing to say. The A4 trial itself, which forms the backbone of this dataset, tested the monoclonal antibody solanezumab in amyloid-positive adults and ultimately found no cognitive benefit. Whether better participant stratification would have changed that result is now, arguably, an open question.

Refining the predictive model is the obvious next step, probably by incorporating additional blood biomarkers or more granular imaging data. Seventy percent accuracy leaves a meaningful margin for error, and the researchers are candid about that.

The People Who Defy the Model

Then there are what the team is quietly calling the misfits. People flagged as likely to decline who stayed stable. People marked as probably stable who got worse. They represent the model’s failures, yes, but also its most scientifically interesting territory. Donohue said he plans to investigate what makes certain patients more resilient and whether those insights could somehow be leveraged to protect others from faster progression.

That question has weight beyond basic science. Cognitive reserve, the brain’s capacity to compensate for accumulating damage, varies considerably between people and is thought to be at least partly modifiable, through education, physical activity, social engagement, though the mechanisms remain murky. If the stable 70% are demonstrating some version of active resilience rather than simply lagging behind on the amyloid clock, understanding what they’re doing differently could be the most useful thing to come out of this research. The real prize isn’t just cleaner trial design. It’s the possibility, still distant but no longer unimaginable, of moving people from the declining categories into the stable one.

https://doi.org/10.1002/alz.71366

Frequently Asked Questions

Does having Alzheimer’s plaques in the brain mean you will definitely develop dementia?

Not necessarily, and that’s precisely what makes this research striking. Around 70% of people with confirmed amyloid buildup in this study showed no cognitive decline over nearly six years. Scientists now think the relationship between amyloid accumulation and clinical disease is far more complicated than a simple cause-and-effect chain, and identifying what protects the stable majority is a major focus of current research.

What is p-tau217 and why does it matter for Alzheimer’s prediction?

P-tau217 is a form of the protein tau, detectable in blood, that becomes elevated when Alzheimer’s pathology is active in the brain. In this study it was among the strongest predictors of which amyloid-positive participants would go on to decline cognitively. Blood-based biomarkers like this are valuable because they offer a far less invasive and costly way to monitor disease progression than PET brain scans, and their predictive power is improving rapidly.

Why have so many Alzheimer’s prevention drug trials failed, and could better patient selection help?

That is exactly the question this research raises. When a large proportion of trial participants aren’t declining anyway, any genuine drug benefit gets diluted into statistical noise, making it look like the drug isn’t working even if it is for the subgroup that matters. Better biomarker-guided selection of participants most likely to decline could, in principle, make future trials substantially more informative and give promising drugs a fairer test.

What does preclinical Alzheimer’s disease actually mean?

It refers to the stage when biological markers of the disease, particularly amyloid plaques and tau tangles in the brain, are detectable but the person has no symptoms. This phase can last a decade or longer before, if ever, cognitive decline begins. It has become the main focus of prevention research because intervening before symptoms appear is thought to offer the best chance of altering the disease course.