Early Detection

For a century, Alzheimer's disease could be confirmed only after death. The past decade changed that: a blood draw can now detect the disease's molecular signature years before symptoms appear. This page explains what the new biomarkers actually measure, when testing makes sense — and why, for most healthy readers, the honest answer is "not yet."

The Biomarker Era

For most of its history, Alzheimer's was a clinical diagnosis — memory complaints, cognitive tests, and a definitive answer only at autopsy. The past decade changed the ground rules. Blood tests can now detect the disease's molecular signature — amyloid plaques and the tau pathology that follows them — years before symptoms, and the star of that class is p-tau217: in research cohorts, a plasma p-tau217 assay identified Alzheimer's pathology with roughly 95% accuracy against amyloid PET imaging (Palmqvist et al., JAMA, 2020), a performance confirmed for commercial-grade immunoassays (Ashton et al., JAMA Neurology, 2024). In 2024, the field's revised diagnostic criteria made the shift official, defining the disease biologically through biomarkers rather than symptoms alone (Jack et al., Alzheimer's & Dementia, 2024).

But "the biomarker era" is not the same thing as "everyone should get tested." The same 2024 criteria that built biomarkers into the diagnosis caution against using them in people without symptoms. The tests have outrun the answers about what to do with the results — and that gap is the subject of this page. For the biology behind what these molecules are doing, the how the brain ages topic is the baseline.

The Preclinical Long Game

The biomarker era exists because the disease gives a long warning: amyloid accumulates for fifteen to twenty years before symptoms, and tau pathology follows with roughly a decade of lead time. That gap — between "detectable" and "symptomatic" — is what makes early detection conceptually possible, and it is also what makes it practically thorny. A test can now tell a forty-five-year-old something about a disease that may arrive at eighty; what it cannot yet tell anyone is what to do with that information.

The prevention trials aimed at this phase have been humbling. The A4 trial treated amyloid-positive but cognitively normal adults with an anti-amyloid antibody and found no cognitive benefit over several years (Sperling et al., NEJM, 2023) — a reminder that detecting early and fixing early are different achievements, and that the second has not happened yet. Until the treatment side catches up with the detection side, early detection is a clinical tool for people with symptoms and a research tool for everyone else.

How Well Each Test Sees the Disease

Two honest glosses. First, these numbers come from research cohorts enriched with people who already had concerns — real-world accuracy in unselected populations runs a bit lower, which is precisely why the field advises testing only when symptoms make Alzheimer's genuinely likely. Second, notice what the chart leaves out: accuracy at detecting pathology says nothing about what a positive result means for any person's future. That is the next section's problem.

The Test Menu

The verdict column grades accuracy for Alzheimer's pathology, not general usefulness. NfL is genuinely valuable in neurology — it just cannot tell Alzheimer's from a dozen other brain problems, which is exactly the limitation that makes it a poor screening answer. And note what the menu implies about cost and access: PET scans run to several thousand dollars in the US; the blood tests are cheap in principle and unevenly available in practice.

When Testing Makes Sense

• ✅ Persistent, clinician-confirmed cognitive symptoms. The standard path when memory or thinking changes are real and persistent: history, labs (thyroid, B12, and the rest), cognitive assessment, MRI — and then biomarkers to resolve whatever remains ambiguous. Testing follows symptoms, not curiosity.
• ✅ Atypical presentations. Early onset — symptoms before sixty-five — or unusual symptom patterns deserve biomarker workup, because the differential includes other diseases (frontotemporal degeneration, vascular damage) with different management.
• ✅ Trial participation. Prevention trials use biomarkers to enroll the right people and measure whether interventions work; in that context, results serve the research and are handled by researchers.
• ❌ Asymptomatic screening. For a cognitively normal person with no clinical concern and no plan, a positive amyloid result answers a question with no proven response and invites distress — the 2024 criteria explicitly discourage this. The quarterly audit is the site's testing framework, and it deliberately contains no amyloid markers: it tracks the modifiable factors from the twelve, not the disease signature.

The Amyloid-PET Caution

Amyloid PET is highly accurate at what it detects — and what it detects is not destiny. Roughly a third of people in their eighties carry significant brain amyloid without dementia; a positive scan reads a risk factor, not a fate, and a negative scan in a symptomatic person is genuinely useful information that redirects the workup. The trial-and-error record agrees: in the IDEAS study, amyloid PET changed clinical management in about 60% of 11,409 Medicare beneficiaries with MCI or dementia (Rabinovici et al., JAMA, 2019) — the test earns its cost when symptoms frame the question.

The caution is for everyone else. A positive scan in an asymptomatic person means elevated risk plus, in disclosure studies, measurable short-term distress even when results are delivered carefully. And a positive scan is not a treatment plan: the anti-amyloid antibodies — lecanemab, whose Clarity AD trial slowed decline modestly on a clinical scale while causing brain swelling or microbleeds in a meaningful minority (van Dyck et al., NEJM, 2023) — are a specialist decision for early symptomatic disease, weighed case by case. Nothing about that sequence starts at a shopping-mall test kit.

What a Result Does Not Tell You

The precision of the new tests invites a category error: treating a molecular snapshot as a personal forecast. Four things a biomarker result does not tell you, stated plainly:

• 📅 Timing. A positive p-tau217 result says pathology is present; it says nothing about whether symptoms are years away, decades away, or never. The spread across individuals is enormous.
• 🛡️ Destiny. Amyloid is a risk factor, not a sentence — and the modifiable factors from the twelve keep mattering after a positive result, not despite it.
• 🧬 Genes vs pathology. An APOE e4 genotype is inherited risk — constant from birth, and not a disease state — while p-tau217 is a live read of current pathology. They answer different questions; a genetic report is not a biomarker result.
• 🔁 Reassurance through re-testing. Serial testing to watch a number move has no established benefit and invites exactly the anxiety-driven monitoring loop this site warns about elsewhere. One result, one clinician conversation — not a tracking habit.

The tools are genuinely good at what they measure. The discipline the era demands is older than the tools: ask what a result would change before ordering the test. If the answer is nothing, the test is a souvenir, not a medical act.

Related Topics