Skip to content

Diagnostic criteria and the biological definition

TL;DR — Between 1984 and 2024 Alzheimer's disease was redefined from a clinical syndrome diagnosable only by exclusion, to a syndrome plus supportive biomarkers, to a biological process defined by biomarkers alone. The 1984 NINCDS-ADRDA criteria stated flatly that "the diagnosis cannot be determined by laboratory tests"; the 2024 Alzheimer's Association revision states that an abnormal Core 1 biomarker "is sufficient to establish a diagnosis of AD" and that the disease begins with Alzheimer's neuropathologic change while people are asymptomatic (McKhann 1984, PMID 6610841; Jack 2024, PMID 38934362). The International Working Group rejects the clinical application of that step, recommending that amyloid-positive cognitively normal people be described as at risk for AD rather than as having it, on the ground that most biomarker-positive cognitively normal individuals will not become symptomatic on a proximate timeline (Dubois 2024, PMID 39483064). The empirical core of the disagreement is quantified: amyloid positivity rises from 10% (95% CI 8–13) at age 50 to 44% (37–51) at age 90 in cognitively normal people, whereas 5-year progression to symptomatic AD in cognitively normal people was 11% at NIA-AA stage 1 and 56% at stage 3, and 57% (45–71) when both amyloid and tau PET were positive (Jansen 2015, PMID 25988462; Vos 2013, PMID 24012374; Moscoso 2025, PMID 40522652). Which criteria set is used changes the measured prevalence of preclinical AD in the same 70-year-olds by a factor of nearly four (Kern 2018, PMID 29653987).

The four generations of criteria

Generation Document What defines AD Biomarkers Intended setting
1984 NINCDS-ADRDA (McKhann 1984, PMID 6610841) Insidious progressive amnestic dementia, other causes excluded; probable / possible / definite None; laboratory tests serve exclusion only Clinical and research
2011 NIA-AA three-workgroup set: dementia (McKhann 2011, PMID 21514250), MCI (Albert 2011, PMID 21514249), preclinical (Sperling 2011, PMID 21514248), with an introduction (Jack 2011, PMID 21514247) Clinical syndrome across three stages; biomarkers raise or lower certainty Amyloid and neuronal-injury markers, research use; MCI criteria have four certainty levels Split: core clinical criteria for practice, biomarker criteria for research
2014 (parallel track) IWG-2 (Dubois 2014, PMID 24849862) An appropriate clinical AD phenotype (typical or atypical) plus a pathophysiological biomarker Amyloid PET or CSF as diagnostic; MRI/FDG-PET demoted to staging/monitoring Research diagnosis
2018 NIA-AA Research Framework, AT(N) (Jack 2018, PMID 29653606) Underlying pathologic processes documented by biomarkers or autopsy; explicitly not by symptoms A (amyloid), T (pathologic tau), N (neurodegeneration), each ±; cognition staged separately Research only — the framework says using it in general medical practice is "premature and inappropriate"
2024 Alzheimer's Association revised criteria (Jack 2024, PMID 38934362) A biological process beginning with Alzheimer's neuropathologic change while asymptomatic Core 1 (amyloid PET, approved CSF, accurate plasma assays, especially p-tau217) sufficient for diagnosis; Core 2 (tau PET, later-changing biofluid) for prognosis Bridge between research and clinical care
2024 (counter-position) IWG clinical-biological construct (Dubois 2024, PMID 39483064) Clinical phenotype and biology together; amyloid-positive cognitively normal = at risk, not diseased Same assays, different interpretive rule; "presymptomatic AD" reserved for a specific biomarker pattern near symptom onset Clinical use

The 2011 set is the hinge: it preserved a clinical definition for practice while creating a research vocabulary for preclinical disease, and it stated explicitly that the preclinical recommendations were "solely intended for research purposes and do not have any clinical implications at this time" (PMID 21514248). Every subsequent controversy is about whether that firewall should have been dismantled.

The ordering that the 2018 AT(N) framework made operational was proposed two years earlier as a hypothetical cascade: Aβ biomarkers (low CSF Aβ42, amyloid PET) become abnormal while people are still cognitively normal; after a variable lag, neuronal-injury biomarkers (CSF tau, FDG-PET hypometabolism, MRI atrophy) dominate; neurodegeneration tracks symptoms more closely than amyloid does (Jack 2010, PMID 20083042). The 2010 paper is a model, not a dataset; DIAN later supplied the timeline in years (Bateman 2012, PMID 22784036). Its load-bearing claim — that amyloid is initiating and neuronal injury is proximate — is exactly what the 2024 Core 1 / Core 2 split encodes, and exactly what the IWG disputes when Core 1 is applied to asymptomatic people.

What the 2024 revision actually claims

The revised criteria define AD as a biological process; assert that an abnormal Core 1 biomarker is sufficient to establish the diagnosis and to inform clinical decision-making across the continuum; assign Core 2 biomarkers (tau PET, later-changing biofluid markers) a prognostic role that increases confidence AD is contributing to symptoms; and set out an integrated biological and clinical staging scheme built to accommodate the fact that copathologies, cognitive reserve and resistance modify the mapping between biological and clinical stage (Jack 2024, PMID 38934362). That last clause matters: the criteria themselves concede that biological stage does not determine clinical stage. The document also declines to be a workflow guideline — it offers "general principles", leaving who to test, when, and what to disclose to other documents (see guidelines).

The IWG objection, stated in numbers

The IWG's 2024 recommendation makes an empirical rather than a philosophical argument: most biomarker-positive cognitively normal individuals will not become symptomatic along a proximate timeline, so labelling them as having a disease is unwarranted "without a clear knowledge of when or whether symptoms will ever develop"; it recommends the term at risk for AD for amyloid-only positivity and reserves presymptomatic AD for a biomarker pattern indicating proximity to symptom onset (Dubois 2024, PMID 39483064).

The relevant quantities:

Quantity Estimate Population / method Source
Amyloid positivity, cognitively normal, age 50 → 90 10% (8–13) → 44% (37–51) IPD meta-analysis, 2,914 normal-cognition participants, 55 studies Jansen 2015, PMID 25988462
Amyloid positivity, subjective cognitive impairment, 50 → 90 12% (8–18) → 43% (32–55) Same PMID 25988462
Amyloid positivity, MCI, 50 → 90 27% (23–32) → 71% (66–76) Same PMID 25988462
Age at which 15% are amyloid positive, by APOE ~40 y (ε4ε4), 55 y (ε3ε4), 65 y (ε3ε3), 95 y (ε2ε3) Same PMID 25988462
Implied interval, amyloid positivity → dementia 20–30 years Inference from the same model PMID 25988462
5-year progression to symptomatic AD, cognitively normal normal 2%; stage 1 11%; stage 2 26%; stage 3 56%; SNAP 5% Washington University volunteers, n=311, CSF-based staging Vos 2013, PMID 24012374
Pooled risk of progression by preclinical stage stage 1 20% (10–34); stage 2 38% (21–59); stage 3 73% (40–92) Meta-analysis of preclinical AD studies Parnetti 2019, PMID 30646955
Tau PET positivity, cognitively unimpaired 349/3,487 (9.8%); 3% (2–4) at 60 y → 19% (16–24) at 90 y 21 cohorts, 6,514 participants, visual read for Braak V–VI Moscoso 2025, PMID 40522652
5-year progression, cognitively unimpaired, by PET profile A+T+ 57% (45–71); A+T− 17% (13–22); A−T− 6% (5–8) Same PMID 40522652
5-year progression to dementia, MCI with A+T+ 70% (59–81) Same PMID 40522652

Read together, these support both sides of the argument and neither position wholly. Amyloid positivity alone is a weak short-horizon predictor — 11–20% over five years — which is the IWG's point. Amyloid plus tau is a strong one — 57% over five years in cognitively unimpaired people — which is the case for a staged biological definition rather than a purely clinical one. The 2024 criteria anticipate this by separating Core 1 (diagnostic) from Core 2 (prognostic), so the disagreement narrows to whether "diagnosis" should be reserved for states with proximate clinical consequence.

Criteria choice changes the count

In a population sample of 259 cognitively unimpaired 70-year-olds from Gothenburg with CSF measured once, the prevalence of "preclinical AD" was 9.7% by IWG-2, 9.7% by Dubois 2016, and 13.1% (stage 1) plus 9.7% (stage 2) by NIA-AA 2011; under AT(N), 13.1% were A+/T−/N−, 7.3% A+/T−/N+ and 2.3% A+/T+/N+, while 18.9% were A−/T−/N+ and 4.6% A−/T+/N+ — that is, 23.5% had neurodegeneration or tau abnormality without amyloid. Overall 46% of this unimpaired sample had some pathological marker (Kern 2018, PMID 29653987). Two consequences follow. First, epidemiological statements about "preclinical Alzheimer's disease" are not comparable across criteria sets. Second, the A− categories — historically labelled suspected non-Alzheimer pathophysiology — are numerically large and are not explained by the amyloid-centred framework; Vos found 23% SNAP with a 5-year progression rate of 5%, similar to biomarker-normal participants (PMID 24012374).

Discordance between pathology and syndrome

The biological definition presumes that neuropathologic change is the disease and symptoms are its consequence. Autopsy data complicate this in two directions.

  • Pathology without symptoms. Substantial neuritic plaque and tangle burden is commonly found in people with normal cognition assessed shortly before death — the "asymptomatic AD" phenotype — and in the Nun Study these brains showed CA1 neuronal hypertrophy (+44.9% cell body volume versus MCI) and higher early-life idea density, evidence of a resilience process that the biomarker definition does not capture (Iacono 2009, PMID 19587326).
  • Symptoms with more than one pathology. In community autopsies, among people with dementia 38.0% had AD plus infarcts versus 30.0% with pure AD, and more than half had multiple diagnoses; multiple pathologies raised the odds of dementia nearly threefold (OR 2.8, 95% CI 1.2–6.7) (Schneider 2007, PMID 17568013). A Core 1 biomarker identifies the AD component; it does not measure the rest of what is producing the syndrome. See vascular and metabolic contributions and imaging and neuropathology.

The 2024 criteria explicitly acknowledge copathology and reserve within their staging scheme (PMID 38934362); the operational question they leave open is what a clinician should tell a person whose Core 1 biomarker is positive but whose syndrome is being driven by something else.

Disclosure: the consequence that is measurable

If a biomarker result is a diagnosis, it will be disclosed. Two datasets bound the psychological effect in cognitively unimpaired people.

  • In the EARLY trial (143 centres, 14 countries), among 3,378 cognitively unimpaired participants disclosed an amyloid result, pre-disclosure depression (GDS), anxiety (STAI) and suicidality (C-SSRS) did not differ between elevated and not-elevated groups. After disclosure, intrusive-thought scores (Impact of Events Scale, 24–72 h) were higher in the elevated group (9.6 [SD 10.8] vs 5.1 [8.0]), as was concern about AD; the pattern held across regions and was milder after CSF than after PET disclosure (Grill 2024, PMID 38585443).
  • In a longitudinal observational cohort of 99 cognitively unimpaired adults (28% elevated amyloid), intrusive thoughts and avoidance were elevated initially and declined over six months (result × time β = −0.82, p<0.001), while AD test-related distress showed a persistent main effect of result (β = 12.09, p<0.001) and concern about AD dementia rose (Clark 2024, PMID 39129396).

The signal is therefore "mild, mostly transient distress with a persistent worry component" in research volunteers who consented to biomarker testing — a population selected for willingness to know. Neither study establishes what happens when the same result is returned opportunistically in primary care, which is the setting that plasma biomarkers make possible. Insurance, employment and driving consequences of a biological diagnosis in an asymptomatic person are not addressed by either dataset and remain an evidence gap as of this build; searches for empirical work on insurance discrimination after AD biomarker disclosure returned no eligible study.

Practical reading of the disagreement

  1. For research classification, AT(N) and its 2024 successor are the operative vocabulary; they were designed to make biomarker-defined groups comparable across studies (PMID 29653606).
  2. For trial eligibility, biomarker positivity is now a requirement rather than an enrichment strategy, which is why trial populations are biomarker-purer than the treated population will be — see clinical trials landscape.
  3. For clinical practice, the two 2024 documents give different answers for the same person. Both agree that a symptomatic person with a positive Core 1 biomarker has AD. They disagree only about the asymptomatic biomarker-positive person — which is precisely the group that screening with plasma assays will generate at scale.
  4. For epidemiology, no population estimate of "Alzheimer's disease" is interpretable without stating which criteria produced it (PMID 29653987).

Management utility is not outcome utility

New IDEAS enrolled 5,757 Medicare beneficiaries, including 21.7% Black and 20.3% Latinx participants. A composite management plan changed within roughly 90 days of amyloid PET in 59.0% (95% CI 57.6–60.5), exceeding 30% in every ethnoracial and presentation subgroup but varying from 45.5% in atypical MCI to 64.8% in typical MCI (Windon 2025, PMID 40728069). The study establishes that the test changes diagnosis, medication or counselling; it does not establish that those changes improve cognition, function, quality of life or survival. That distinction is central when “clinical utility” is used to justify biological diagnosis.

The legal consequence is no longer wholly evidence-free. Interviews with 17 dementia experts found three documentation practices for a preclinical biomarker result: record it as active AD, as non-informative, or as increased susceptibility. Clinicians using the active-disease framing were more willing to disclose to employers and insurers, creating differential discrimination exposure from the same laboratory result (Vaishnav 2024, PMID 38393896). This is clinician-behaviour evidence, not documented discrimination outcomes, but it closes the narrower question of whether documentation practice is uniform: it is not.

Open questions

  • What is the lifetime, not 5-year, risk of symptomatic AD for an amyloid-positive, tau-negative 70-year-old, with death as a competing risk? The available progression estimates are short-horizon and cohort-selected (Vos 2013, PMID 24012374; Moscoso 2025, PMID 40522652).
  • Does adding Core 2 tau positivity to Core 1 change management decisions enough to justify the additional testing burden, given that A+T+ carries 57% 5-year risk versus 17% for A+T− (PMID 40522652)?
  • What proportion of the 23.5% of unimpaired 70-year-olds who are A− but T+ or N+ progress, and to what — the SNAP category is large and its 5-year progression was indistinguishable from normal (Kern 2018, PMID 29653987; Vos 2013, PMID 24012374)?
  • Do the mild, largely transient psychological effects observed in consented research volunteers generalise to opportunistic plasma-biomarker results in primary care (Grill 2024, PMID 38585443; Clark 2024, PMID 39129396)?
  • Can criteria be written that classify the dominant contributor to a syndrome rather than the presence of one pathology, given that mixed pathology is the modal community case (Schneider 2007, PMID 17568013)?
  • What are the insurance, employment, licensing and legal-capacity consequences of a biological AD diagnosis in an asymptomatic person? No empirical study was retrieved in this build's searches.

References

  1. McKhann G, et al. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group. Neurology. 1984;34:939-944. PMID 6610841.
  2. Jack CR Jr, et al. Introduction to the recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement. 2011;7:257-262. PMID 21514247.
  3. McKhann GM, et al. The diagnosis of dementia due to Alzheimer's disease. Alzheimers Dement. 2011;7:263-269. PMID 21514250.
  4. Albert MS, et al. The diagnosis of mild cognitive impairment due to Alzheimer's disease. Alzheimers Dement. 2011;7:270-279. PMID 21514249.
  5. Sperling RA, et al. Toward defining the preclinical stages of Alzheimer's disease. Alzheimers Dement. 2011;7:280-292. PMID 21514248.
  6. Dubois B, et al. Advancing research diagnostic criteria for Alzheimer's disease: the IWG-2 criteria. Lancet Neurol. 2014;13:614-629. PMID 24849862.
  7. Jack CR Jr, et al. NIA-AA Research Framework: toward a biological definition of Alzheimer's disease. Alzheimers Dement. 2018;14:535-562. PMID 29653606.
  8. Jack CR Jr, et al. Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup. Alzheimers Dement. 2024;20:5143-5169. PMID 38934362.
  9. Dubois B, et al. Alzheimer disease as a clinical-biological construct — an International Working Group recommendation. JAMA Neurol. 2024;81:1304-1311. PMID 39483064.
  10. Jansen WJ, et al. Prevalence of cerebral amyloid pathology in persons without dementia: a meta-analysis. JAMA. 2015;313:1924-1938. PMID 25988462.
  11. Vos SJ, et al. Preclinical Alzheimer's disease and its outcome: a longitudinal cohort study. Lancet Neurol. 2013;12:957-965. PMID 24012374.
  12. Parnetti L, et al. Prevalence and risk of progression of preclinical Alzheimer's disease stages: a systematic review and meta-analysis. Alzheimers Res Ther. 2019;11:7. PMID 30646955.
  13. Moscoso A, et al. Frequency and clinical outcomes associated with tau positron emission tomography positivity. JAMA. 2025;334:229-242. PMID 40522652.
  14. Kern S, et al. Prevalence of preclinical Alzheimer disease: comparison of current classification systems. Neurology. 2018;90:e1682-e1691. PMID 29653987.
  15. Iacono D, et al. The Nun Study: clinically silent AD, neuronal hypertrophy, and linguistic skills in early life. Neurology. 2009;73:665-673. PMID 19587326.
  16. Schneider JA, et al. Mixed brain pathologies account for most dementia cases in community-dwelling older persons. Neurology. 2007;69:2197-2204. PMID 17568013.
  17. Grill JD, et al. Immediate reactions to Alzheimer biomarker disclosure in cognitively unimpaired individuals in a global truncated randomized trial. Neurol Clin Pract. 2024;14:e200265. PMID 38585443.
  18. Clark LR, et al. Psychosocial implications of learning amyloid PET results in an observational cohort. Alzheimers Dement. 2024;20:6579-6589. PMID 39129396.
  19. Jack CR, et al. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol. 2010;9:119-28. PMID 20083042.
  20. Bateman RJ, et al. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. N Engl J Med. 2012;367:795-804. PMID 22784036.
  21. Windon CC, et al. Amyloid PET and clinical management in a diverse, cognitively impaired population: the New IDEAS Study. Alzheimers Dement. 2025;21:e70504. PMID 40728069.
  22. Vaishnav N, et al. Creating an unprotected class: addressing legal risks in the era of biologically-defined Alzheimer's disease. J Alzheimers Dis. 2024;98:187-195. PMID 38393896.