Every row of the evidence table

A burden-of-disease decomposition: age-standardised rates rose slightly and population growth was the largest driver of the rise in cases. It compares no exposure, so no direction is recorded.

A lifetime-risk estimate (42% after age 55, higher in women, Black adults and APOE4 carriers) and a projection of new cases. It compares no exposure, so no direction is recorded.

A pooled incidence and prevalence estimate with no exposure compared, so no direction is recorded.

A prevalence estimate and forecast: cases are projected to rise from 57.4 million to 152.8 million while age-standardised prevalence stays flat, with the rise attributed to population growth and ageing. It compares no exposure, so there is no direction to record.

A pooled prevalence estimate with no exposure compared, so there is no direction to record.

A position paper from an expert network. It sets out 12 priority research areas on sex and gender in Alzheimer's disease and reports no effect estimate.

A threshold-derivation meta-analysis: it converges on 18 Centiloids (data-driven) and 27 Centiloids (visual reads) and flags 11-26 as a range to read with caution. It compares no exposure, so no direction exists to record.

Appropriate use criteria rated by consensus: for amyloid PET seven of 17 scenarios were rated appropriate, two uncertain and eight rarely appropriate; for tau PET five, six and six. No effect estimate is reported.

An editorial reviewing the FDA clearance of the Lumipulse plasma ratio test. It relays performance figures from the FDA summary and other studies and reports no data of its own.

A clinical practice guideline. It sets performance thresholds for blood tests (at least 90% sensitivity and 75% specificity for triage; at least 90% for both to substitute for amyloid PET or CSF) and reports no effect estimate of its own.

A price-threshold model: it estimates the test price at which diagnostic cost per true-positive case equals standard care ($290-$1,950 depending on use), and compares no actual price against it, so there is no direction to record.

A workgroup recommendations document. It sets target sensitivity and specificity for blood biomarker tests (≥90% / ≥85% for triage in primary care; ~90% / ~90% as a confirmatory test) and reports no data of its own.

A pooled prevalence among carers (50.8% lonely, 37.1% socially isolated) with no non-carer comparison, so no direction for the effect of caregiving can be recorded.

A cost estimate ($818 billion, 80% of it borne by families through unpaid care, earnings losses, out-of-pocket costs and quality-of-life losses) with no exposure compared, so no direction is recorded.

A descriptive spending analysis with no exposure compared, so there is no direction to record: Part D spending rose with dual-eligibility duration, Part B drug spending varied widely by state, and ACO enrollment made only a modest difference.

A cost projection: total burden is estimated at $344 billion in 2020 rising to over $3 trillion in 2060, with African American and Latino families projected to bear a disproportionate share. It compares spending shares between groups but tests no exposure, so no direction is recorded.

A cost estimate with no exposure compared, so there is no direction to record. Higher unpaid-care costs went with lower medical spending on nursing facility care.

A cost projection with no exposure compared, so there is no direction to record. Dementia's share of replacement cost is projected to rise from 44% to 53%.

A cost estimate with no exposure compared, so there is no direction to record.

A cost estimate and projection with no exposure compared, so there is no direction to record.

A cost estimation across 204 countries with no exposure compared, so there is no direction to record.

A threshold analysis estimating what each country should be willing to pay (from $4-$18 a year in low-income countries to $254-$9,434 for lecanemab and $387-$13,964 for donanemab in high-income countries). It sets no price against an actual list price, so there is no direction to record.

A cost-effectiveness model reporting ICERs of JPY 8.5 million and 7.9 million per QALY without stating a threshold or a verdict; the authors argue that narrow value assessments understate social value. No direction is recorded.

A criteria document. It defines Alzheimer's disease biologically and sets out staging, and reports no effect estimate.

An editorial. It argues that biomarker positivity alone should not define Alzheimer's disease in people without symptoms, and reports no data of its own.

A research framework defining Alzheimer's disease by biomarkers (the AT(N) system) for research use. It states its own intended limits and reports no effect estimate.

Consensus guidance for primary care on disclosing and communicating a dementia diagnosis, developed with patients, care partners and clinicians. It reports no effect estimate.

A clinical practice guideline. It sets out a structured, patient-centred process for evaluating, diagnosing and disclosing suspected Alzheimer's disease or related dementias in any practice setting, and reports no effect estimate.

A single-programme implementation report with no comparison group. It counts enrolment and service use and estimates no effect.

Implementation guidance drawn from six earlier comprehensive dementia care programmes. It reports no effect estimate of its own.

A classification-validation study without a neuropathological reference standard, so accuracy cannot be judged. The groups behaved as expected: patients labelled LATE had mild amnestic impairment, limbic atrophy and stable MMSE over 30 months, while AD and AD/LATE groups declined.

A Delphi consensus among 60 international expert panellists, reaching agreement on 15 major and 42 minor referral criteria. No patient outcome is measured and no effect is estimated.

A prognosis meta-analysis: it pools time to death and to nursing home admission and compares no exposure. Women's shorter survival was attributable to later age at diagnosis.

A descriptive proteomic mapping: APOE-linked protein changes were present before amyloid pathology and stable across age and disease stage, with ε2 tied to cellular-maintenance and anti-inflammatory pathways and ε4 to cell-cycle, oligodendrocyte-precursor and downstream vascular, immune and proteostatic proteins. It estimates no effect on a clinical outcome.

A transcriptome-wide association study prioritising nine genes (including BIN1, PTK2B and DMPK) with broadly consistent effects across populations; it reports no effect estimate for any exposure, so no direction is recorded.

Locus-discovery genome-wide association study: it reports which genetic regions reached significance and no effect estimate for any exposure, so no direction is recorded. Nine signals (HP1BP3, SLC50A1, PTPRC, NPAS3, DDHD1, CHST9, SMYD2, PRAMEF1, GFRA1) appeared only in APOE-stratified analyses; DDHD1 was protective in ε4 carriers.

Locus-discovery genome-wide association study: it reports which genetic regions reached significance and no effect estimate for any exposure, so no direction is recorded. The study reports 14 new loci for clinically diagnosed AD and two from AD-by-proxy.

Locus-discovery genome-wide association study: it reports which genetic regions reached significance and no effect estimate for any exposure, so no direction is recorded. The study reports eight new loci (six ancestry-specific, two trans-ancestry) and nominates eight genes in microglial activation, glutamate production and signalling.

Locus-discovery genome-wide association study: it reports which genetic regions reached significance and no effect estimate for any exposure, so no direction is recorded. Beyond APOE, the study reports TREM2 R47H and variants near RAB3B, SPOCK3/ANXA10 and SMAP2, and a gene-based association with GIPR.

Locus-discovery genome-wide association study: it reports which genetic regions reached significance and no effect estimate for any exposure, so no direction is recorded. Two signals at 17p13.2 and 18q21.33 are reported; the authors state that African ancestry at 17p13.2 went with lower AD risk and that overall African ancestry did not differ between cases and controls, without an effect estimate.

A single case report: one man with the PSEN1 E280A mutation stayed cognitively intact until 67 despite very heavy amyloid, and carried a gain-of-function RELN variant. With one case and no comparison group beyond the earlier Christchurch case, no direction can be recorded.

Among 37 heterozygous carriers (ages 56.6-82.1) none had developed AD or MCI by the March 2022 data freeze and apolipoprotein B was lower than in non-carriers, but the authors say continued follow-up is necessary and no homozygotes were found, so protection could not be tested. No incidence comparison is reported, so no direction is recorded.

A narrative review arguing that Alzheimer's disease is a brain-specific form of diabetes; it states that type 2 diabetes alone "fall[s] far short of mimicking AD" and reports no effect estimate.

A retraction notice. It withdraws the 2006 Nature paper on Aβ*56 over figures showing "signs of excessive manipulation, including splicing, duplication and the use of an eraser tool", and reports no effect estimate.

An investigative news feature reporting signs of image fabrication in papers on the Aβ*56 amyloid oligomer. It is not a study of an outcome and reports no effect estimate.

A perspective review of arguments for and against the amyloid hypothesis, concluding that amyloid-beta is likely necessary but not sufficient and acts early as a trigger of tau-mediated neurodegeneration. It reports no effect estimate.

Mechanistic multi-omic profiling of blood from three ARIA-positive patients and three matched controls on lecanemab. It describes a clonally expanded cytotoxic CD8+ T-cell signature and estimates no clinical effect.

A single-clinic safety series with no comparison group. It reports how often ARIA and infusion reactions happened in practice, which is useful, but it cannot estimate an effect of the drug in either direction.

A mechanistic tissue study describing which microglial states accompany amyloid removal (TREM2 and APOE upregulation correlating with antibody response and clearance; complement signalling contributing). It estimates no effect of immunisation on a clinical outcome.

Corporate announcement discontinuing a product and terminating its confirmatory trial. It reports no clinical outcome, so no direction exists to record.

A coverage determination. It sets the conditions under which Medicare pays and reports no effect estimate.

A descriptive transcriptional atlas: it maps expression changes, isoform usage, co-expression modules and disease subtypes in microglia from post-mortem brains and estimates no effect of an exposure.

A regulatory enforcement action, not a study. The SEC's order finds the reported Phase 2 biomarker data were manipulated by an unblinded consultant and its complaint alleges the full patient data "showed no measurable cognitive improvement in the patients' episodic memory"; it estimates no effect.

Retraction notices for five PLOS ONE papers by Wang and colleagues, two co-authored with a Cassava Sciences employee. They withdraw the papers over blot-image integrity and estimate no effect.

A feasibility trial in 29 residents: designed to test whether the trial can be run, not powered for efficacy, and it reports no efficacy estimate.

A mechanism study: it shows Arc packages tau into extracellular vesicles and that cell-to-cell tau transmission is almost absent in Arc knockout mice, with Arc levels correlating with phosphorylated tau in human AD brain vesicles. It tests no intervention and estimates no effect on an outcome.

A model-building and screening study: it identifies over 500 genetic modifiers of tau propagation (including VPS29 and the UFMylation cascade) and shows UFMylation inhibition suppresses seeded propagation in vitro and in vivo. It estimates no effect on a clinical outcome.

A narrative review proposing glymphatic failure as a final common pathway to dementia. It reports no effect estimate.