Alzheimer's disease statistics — quick reference¶

Last curated: 2026-09-03 (evidence-breadth deepening)

Use note. Every row states its estimate year or follow-up, its population and its method. Counts, rates, proportions, hazard ratios, odds ratios and modelled attributable fractions are not interchangeable. Uncertainty intervals (UI) accompany Global Burden of Disease model estimates; confidence intervals (CI) accompany sampled studies. Conflicting estimates are shown side by side and never averaged. PMIDs added in the evidence-breadth section were retrieved through live PubMed E-utilities on 2026-09-03; earlier records retain their documented 2026-08-31 retrieval provenance.

Counting caution that applies to the whole file. Most population statistics count dementia, a syndrome. Only autopsy and biomarker studies count Alzheimer's neuropathologic change. Where a row says "dementia", it is not a count of Alzheimer's disease.

1. Global burden¶

Measure	Estimate (95% UI)	Year	Population / method	Source
People living with dementia	57.4 million (50.4–65.1)	2019	GBD forecasting model, 204 countries	PMID 34998485
Projected people living with dementia	152.8 million (130.8–175.9)	2050	Same model with risk-factor and education predictors	PMID 34998485
Change in age-standardised prevalence	+0.1% (−7.5 to 10.8)	2019→2050	Same	PMID 34998485
Female-to-male prevalence ratio	1.69 (1.64–1.73)	2019	Same	PMID 34998485
Projected female-to-male ratio	1.67 (1.52–1.85)	2050	Same	PMID 34998485
People living with dementia	43.8 million (37.8–51.0)	2016	GBD 2016 dementia analysis, 195 countries	PMID 30497964
Increase in count	+117% (114–121)	1990→2016	Same	PMID 30497964
Change in age-standardised prevalence	+1.7% (1.0–2.4); 701 → 712 per 100,000	1990→2016	Same	PMID 30497964
Dementia deaths	2.4 million (2.1–2.8); 5th leading cause globally	2016	Same, mortality modelled from prevalence and excess mortality	PMID 30497964
Dementia DALYs	28.8 million (24.5–34.0)	2016	Same	PMID 30497964
DALYs attributable to 4 modelled risks (high BMI, high fasting glucose, smoking, sugar-sweetened beverages)	6.4 million (3.4–10.5)	2016	Comparative risk assessment within GBD 2016	PMID 30497964
Group burden of 37 nervous-system conditions (AD and other dementias in the top 10 by age-standardised DALYs)	443 million DALYs (378–521); 3.40 billion people affected	2021	GBD 2021 nervous system disorders	PMID 38493795

Do not compare 43.8 million (2016) with 57.4 million (2019) as a time trend. They come from different GBD vintages with different models; the second is a forecasting analysis.

Regional heterogeneity, projected change 2019→2050¶

Region	Projected change in cases (95% UI)
High-income Asia Pacific	+53% (41–67)
Western Europe	+74% (58–90)
Eastern sub-Saharan Africa	+357% (323–395)
North Africa and the Middle East	+367% (329–403)

Source: PMID 34998485. Growth is attributed principally to population growth and ageing, with population growth dominating in sub-Saharan Africa and ageing dominating in east Asia.

2. National and regional prevalence¶

Population	Estimate	Year	Method	Source
US adults 65+	Dementia 10% (95% CI 9–11); MCI 22% (20–24)	2016	HRS-HCAP, n=3,496, neuropsychological battery + informant, population-weighted	PMID 36279130
US adults 65+	5.0 million with ADRD (1.6% of population)	2014	Medicare fee-for-service prevalence applied to census	PMID 30243772
US adults 65+	13.9 million (3.3%) projected	2060	Same model, census projections	PMID 30243772
US adults 65+	6.9 million with Alzheimer's dementia; 13.8 million projected by 2060	2024	Alzheimer's Association compiled estimates	PMID 38689398
India, 60+	Major neurocognitive disorder 7.2%; mild 17.6%	2024 report	LASI-DAD, n=4,096, DSM-5 algorithm, nationally representative	PMID 38324518
Indonesia, 65+	27.9% (95% CI 25.2–28.9)	2021 data	STRiDE, 10/66 short schedule, n=2,110, weighted	PMID 37278200
South Africa, 65+	12.5% (95% CI 9.5–16.0)	2021 data	STRiDE, 10/66 short schedule, n=408, weighted	PMID 37278200
Japan (Hisayama), crude prevalence 65+	6.7% (1985), 5.7% (1992), 7.1% (1998), 12.5% (2005), 17.9% (2012), 15.8% (2017), 12.1% (2022)	1985–2022	Seven cross-sectional community surveys	PMID 41466306
Gothenburg 85-year-olds	29.8% (1986–87) → 21.7% (2008–10); OR 0.66 (95% CI 0.50–0.86)	—	Two population samples, identical DSM-III-R methods	PMID 28733627

Young-onset dementia¶

Measure	Estimate	Population / method	Source
Age-standardised prevalence, ages 30–64	119.0 per 100,000 (~3.9 million people worldwide)	Meta-analysis, 95 studies, 74 in meta-analysis, 2,760,379 participants	PMID 34279544
Prevalence by age band	1.1 per 100,000 (30–34) → 77.4 per 100,000 (60–64)	Same	PMID 34279544
Age-standardised incidence, ages 30–64	11 per 100,000 (~370,000 new cases/year)	Meta-analysis, 61 articles	PMID 35715891
Incidence by age band	0.17 per 100,000 (30–34) → 5.14 per 100,000 (60–64)	Same	PMID 35715891

3. Incidence and lifetime risk¶

Measure	Estimate	Population / method	Source
All-cause dementia incidence, ages 65–69	~4 per 1,000 person-years	7 cohorts, US + Europe, 49,202 individuals	PMID 32611641
All-cause dementia incidence, ages 85–89	~65 per 1,000 person-years	Same	PMID 32611641
Incidence trend	−13% per calendar decade (95% CI 7–19%); men −24% (14–32), women −8% (0–15)	Same, 1988–2015	PMID 32611641
Framingham 5-year cumulative hazard, four epochs	3.6 → 2.8 → 2.2 → 2.0 per 100 persons (declines of 22%, 38%, 44%); decline only with ≥ high-school diploma (HR 0.77, 0.67–0.88)	5,205 people ≥60, consistent criteria since 1975	PMID 26863354
Pooled incidence change, high-income countries	0.82 (95% CI 0.51–1.33), I²=94.9%; 0.69 (0.47–1.00) excluding Hisayama	Meta-analysis of 5 high-quality cohort studies	PMID 30271219
Incidence above age 90	18.2% per year (95% CI 15.3–21.5)	The 90+ Study, 330 non-demented participants, 2003–2007	PMID 20186856
— 90–94 / 95–99 / 100+	12.7% / 21.2% / 40.7% per year; Poisson doubling time 5.5 years	Same	PMID 20186856
Incidence ≥90 by race/ethnicity	Asian 89.9, White 96.9, Latino 105.8, Black 121.5 per 1,000 person-years	2,350 health-plan members ≥90, 2010–2015	PMID 30797730
US lifetime risk from age 55	42% (95% CI 41–43); ~45–60% in women, Black adults and APOE ε4 carriers	ARIC, n=15,043, death as competing risk	PMID 39806070
US annual incident dementia cases	~514,000 (2020) → ~1,000,000 (2060)	Lifetime-risk estimates applied to census projections	PMID 39806070

Incidence and 25-year risk by race/ethnicity (US health plan, 274,283 members aged 64+, 2000–2013)¶

Group	Incidence per 1,000 person-years	25-year cumulative risk from age 65
African American	26.6	38%
American Indian/Alaska Native	22.2	35%
Latino	19.6	32%
Pacific Islander	19.6	25%
White	19.3	30%
Asian American	15.2	28%

African American vs Asian American hazard ratio 1.65 (95% CI 1.58–1.72). Source: PMID 26874595.

4. Mortality and survival¶

Measure	Estimate	Population / method	Source
All-cause mortality, any dementia vs none	HR 5.90 (95% CI 3.53–9.86)	Meta-analysis, 78 studies, 63,125 with dementia vs 152,353 controls	PMID 36097997
Mean survival from AD diagnosis	5.8 years (SD 2.0)	Same	PMID 36097997
Non-AD dementia vs AD, mortality	HR 1.33 (1.21–1.46)	Same	PMID 36097997
Survival from diagnosis, non-AD vs AD	−1.12 years (−1.52 to −0.72)	Same	PMID 36097997
Highest-mortality subtype	Lewy body dementia HR 17.88 (5.87–54.46)	Same	PMID 36097997
Median survival, typical AD (memory clinic)	7.2 years (95% CI 7.0–7.5)	Amsterdam Dementia Cohort, 2,081 biomarker-confirmed patients	PMID 40294367
Median survival, atypical AD	6.3 years (5.8–6.9); HR 1.31 (1.10–1.56) vs typical	Same	PMID 40294367
5-year survival after dementia onset (Japan)	47.3% (1988 cohort) → 65.2% (2002) → 58.9% (2012, ns)	Hisayama, three 10-year cohorts	PMID 41466306
US Alzheimer's deaths recorded on death certificates	119,399	2021, US vital registration	PMID 38689398
Change in reported AD deaths	>+140%	2000→2021, US	PMID 38689398
18-month mortality, advanced dementia in nursing homes	54.8%	323 residents, 22 nursing homes	PMID 19828530

5. Disease duration and stage¶

Measure	Estimate	Population / method	Source
Total AD duration from age 60	~24 years	Multistate model, 6 cohorts, n=3,268, death as end state	PMID 31164314
Total AD duration from age 80	~15 years	Same	PMID 31164314
Preclinical / prodromal / dementia duration, entering preclinical at 70	~10 / ~4 / ~6 years	Same	PMID 31164314
Annual MMSE change, population-based incident AD	−1.53 points (SD 2.69)	Cache County, 328 incident cases, mean 3.80 y follow-up	PMID 21606896
Annual CDR-SB change	+1.44 (SD 1.82)	Same	PMID 21606896
Annual NPI change	+2.55 (SD 5.37)	Same	PMID 21606896
Proportion progressing <1 point/year 5–7 years after onset	30–58% of survivors	Same	PMID 21606896
MCI → dementia annual conversion, specialist settings	9.6% (AD 8.1%, VaD 1.9%)	Meta-analysis, 41 inception cohorts	PMID 19236314
MCI → dementia annual conversion, population studies	4.9% (AD 6.8%, VaD 1.6%)	Same	PMID 19236314

6. Biomarker frequencies and progression risk¶

Measure	Estimate (95% CI)	Population / method	Source
Amyloid positivity, cognitively normal, age 50 → 90	10% (8–13) → 44% (37–51)	IPD meta-analysis, 55 studies, 2,914 normal-cognition participants	PMID 25988462
Amyloid positivity, MCI, age 50 → 90	27% (23–32) → 71% (66–76)	Same, 3,972 MCI participants	PMID 25988462
Age at which 15% are amyloid positive, by APOE	~40 y (ε4ε4), 50 y (ε2ε4), 55 y (ε3ε4), 65 y (ε3ε3), 95 y (ε2ε3)	Same	PMID 25988462
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 reads for Braak V–VI	PMID 40522652
Tau-PET positivity, MCI / AD dementia at age 75	43% (41–46) / 79% (77–82)	Same	PMID 40522652
5-year progression, cognitively unimpaired A+T+	57% (45–71)	Same	PMID 40522652
— A+T− / A−T−	17% (13–22) / 6% (5–8)	Same	PMID 40522652
5-year progression to dementia, MCI A+T+	70% (59–81)	Same	PMID 40522652
5-year progression to symptomatic AD, by NIA-AA preclinical stage	normal 2%; stage 1 11%; stage 2 26%; stage 3 56%; SNAP 5%	Washington University volunteers, n=311, CSF-based staging	PMID 24012374
Pooled progression risk by preclinical stage	stage 1 20% (10–34); stage 2 38% (21–59); stage 3 73% (40–92)	Meta-analysis of preclinical AD studies	PMID 30646955
Preclinical AD prevalence in the same 70-year-olds, by criteria set	IWG-2 9.7%; Dubois 2016 9.7%; NIA-AA stage 1 13.1% + stage 2 9.7%; any pathological marker 46%	Gothenburg H70, 259 cognitively unimpaired, CSF	PMID 29653987

7. Diagnostic test performance¶

Test	Performance	Population / method	Source
Plasma p-tau217	Sensitivity 88.1% (95% CI 86.7–89.5), specificity 88.7% (87.4–89.9), AUROC 91.1% (88.9–92.4), DOR 50.7 (40.6–63.4)	Meta-analysis, 113 studies, 29,625 individuals, biological reference standards	PMID 40818474
Plasma p-tau212 / p-tau205 / p-tau181 / p-tau231 (AUROC)	90.3% / 85.1% / 81.5% / 80.2%	Same	PMID 40818474
APS2 (%p-tau217 + Aβ42:Aβ40), primary care, prospective	AUC 0.96 (0.94–0.98); PPV 88%; NPV 90%; accuracy 88–92% across 4 cohorts	1,213 patients, Sweden 2020–2024, predefined external cutoffs, CSF reference	PMID 39068545
Primary care physician accuracy vs APS2	61% (53–69) vs 91% (86–96)	Same	PMID 39068545
Dementia specialist accuracy vs APS2	73% (68–79) vs 91% (88–95)	Same	PMID 39068545
Mass-spec %p-tau217 vs immunoassays, Aβ-PET status accuracy	0.93 vs 0.83–0.88 (P<0.007)	BioFINDER-2, 998 participants, 5 assays head-to-head	PMID 39468767
Plasma p-tau217/Aβ42 (Lumipulse)	AUC 0.963–0.966 vs amyloid PET; 0.947–0.974 vs tau PET	Clinic cohort n=391, community cohort n=121	PMID 40156286
— intermediate zone, ratio vs p-tau217 alone	clinic 10.6% vs 13.0%; community 16.5% vs 31.4%	Same	PMID 40156286
CSF Aβ42/40 vs Aβ42 alone (vs amyloid PET)	Sens/spec 0.94/0.82 vs 0.93/0.57; AUC 0.90 (0.83–0.97) vs 0.78 (0.68–0.88)	103 memory-clinic patients with known PET status	PMID 35473631
Amyloid PET (florbetapir) vs autopsy	Binary agreement 96%; ρ 0.78 (0.58–0.89) vs immunohistochemistry	29 end-of-life participants, mean 99 days ante-mortem	PMID 21245183
Tau PET (flortaucipir) vs autopsy, Braak V–VI	Sensitivity 92.3–100%; specificity 52.0–92.0% across 5 readers	64 primary-cohort autopsies, 28 sites	PMID 32338734
AD cortical thickness signature, MCI → mild AD dementia	Sensitivity 83%, specificity 65% over ~2.5 years	49 participants with CDR 0.5	PMID 19109536
CSF t-tau / p-tau / Aβ42 fold-change vs controls	2.54 (2.44–2.64) / 1.88 (1.79–1.97) / 0.56 (0.55–0.58)	Meta-analysis, 231 articles, 15,699 AD vs 13,018 controls	PMID 27068280
Plasma %p-tau217 vs FDA-approved CSF tests, Aβ-PET AUC	0.95–0.97, equivalent; tau-PET AUC 0.95–0.98 (superior to CSF)	BioFINDER-2 n=1,422, Knight ADRC n=337	PMID 38382645
Two-step plasma p-tau217 workflow, Aβ-PET accuracy	88.2–92.0%; CSF tests reduced by 61–86%	MCI, n=348, BioFINDER-1/2	PMID 37653254

Threshold caveat. Approximately 90% of studies in the plasma p-tau meta-analysis were rated high risk of bias for not using predefined or externally derived thresholds (PMID 40818474). The prospective, predefined-cutoff study (PMID 39068545) is the exception, not the norm.

8. Treatment effects¶

Disease-modifying (anti-amyloid antibodies)¶

Trial	Primary result	Amyloid change	ARIA-E	Source
CLARITY AD (lecanemab, n=1,795, 18 mo)	CDR-SB 1.21 vs 1.66; difference −0.45 (95% CI −0.67 to −0.23), P<0.001	−59.1 Centiloids (−62.6 to −55.6)	12.6%	PMID 36449413
TRAILBLAZER-ALZ 2 (donanemab, n=1,736, 76 wk, low/medium tau)	iADRS difference 3.25 (1.88–4.62), P<0.001; CDR-SB −0.67 (−0.95 to −0.40)	not reported in primary abstract	24.0% (52 symptomatic; 3 treatment-related deaths)	PMID 37459141
EMERGE (aducanumab high dose, 78 wk)	CDR-SB −0.39 (−0.69 to −0.09), P=0.012	dose-dependent reduction	most common AE	PMID 35542991
ENGAGE (aducanumab high dose, 78 wk)	CDR-SB +0.03 (−0.26 to 0.33), P=0.833	dose-dependent reduction	most common AE	PMID 35542991
GRADUATE I / II (gantenerumab, 116 wk)	CDR-SB −0.31 (−0.66 to 0.05), P=0.10 / −0.19 (−0.55 to 0.17), P=0.30	−66.4 / −56.5 Centiloids; amyloid-negative in 28.0% / 26.8%	24.9%; symptomatic 5.0%	PMID 37966285
A4 (solanezumab, preclinical AD, 240 wk)	PACC difference −0.30 (−0.82 to 0.22), P=0.26	amyloid increased: +11.6 vs +19.3 Centiloids	<1% both arms	PMID 37458272
TRAILBLAZER-ALZ 4 (donanemab vs aducanumab, 148 participants)	Amyloid clearance at 6/12/18 mo: 37.9%/70.0%/76.8% vs 1.6%/24.6%/43.1% (P<0.001)	median time to clearance 359 vs 568 days	23.9% vs 34.8%	PMID 40390253
TRAILBLAZER-ALZ phase 2 (donanemab, n=257, 76 wk)	iADRS difference 3.20 (0.12–6.27), P=0.04	−85.06 Centiloids vs placebo	ARIA-E (mostly asymptomatic)	PMID 33720637
EXPEDITION3 (solanezumab, mild dementia, n=2,129, 80 wk)	ADAS-cog14 difference −0.80 (−1.73 to 0.14), P=0.10	not a plaque-clearing antibody	ARIA-E 1 vs 2 patients	PMID 29365294

Non-amyloid disease-modification trial¶

Trial	Primary result	Safety	Source
evoke / evoke+ (oral semaglutide, 104 wk; n=1,855 / 1,953 randomised)	CDR-SB difference −0.08 (95% CI −0.35 to 0.20), P=0.57 / +0.10 (−0.17 to 0.38), P=0.46 — both null	Treatment-emergent adverse events 91.2% semaglutide vs 84.8% placebo; 5 investigator-attributed treatment-related deaths (1 vs 4)	PMID 41865758; NCT04777396; NCT04777409
INVOKE-2 (AL002 TREM2 agonist, n=381, week 96)	CDR-SB LS differences −0.31 (−1.61 to 0.98) / +0.13 (−1.18 to 1.43) / −0.17 (−1.49 to 1.15) vs placebo, all P>0.05	ARIA-like MRI the most frequent TEAE; target engaged	PMID 41787076; NCT04592874
ELAD (liraglutide, n=204, 52 wk)	Cerebral glucose metabolic rate difference −0.17 (−0.39 to 0.06), P=0.14	Generally safe in non-diabetic AD	PMID 41326666; NCT01843075
EPOCH (verubecestat, n=1,958, 78 wk)	ADAS-cog 7.9 / 8.0 vs 7.7 placebo (P=0.63 / 0.46) — null, stopped for futility	More rash, falls, sleep disturbance, suicidal ideation, weight loss, hair-colour change	PMID 29719179; NCT01739348
APECS (verubecestat prodromal, n=1,454, 104 wk)	CDR-SB 1.65 / 2.02 vs 1.58; 40 mg worse (P=0.01); dementia HR 1.38 (1.07–1.79) at 40 mg	Stopped for futility; worse outcome at higher dose	PMID 30970186; NCT01953601
TANGO (gosuranemab, n=650 treated, 78 wk)	CDR-SB 1.85 high dose vs 1.85 placebo; CSF N-terminal tau reduced at all doses	SAE 10.3–12.3% vs 12.1% placebo	PMID 38012285; NCT03352557

Minimal clinically important differences, for comparison¶

Stage	ADAS-Cog	CDR-SB	iADRS	MMSE
MCI	+2 to +3	+1	−5	−1 to −2
Mild AD	+3	+2	−9	−2
Moderate–severe AD	—	+2	—	−1.4 to −3

Source: rapid systematic review of 10 articles (PMID 38561021). The review's own conclusion is that average anti-amyloid treatment effects are lower than these thresholds.

Symptomatic therapy¶

Intervention	Effect (95% CI)	Population / method	Source
Cholinesterase inhibitors	−2.7 ADAS-Cog points (−3.0 to −2.3) over 6–12 months	Cochrane, 13 RCTs	PMID 16437532
— withdrawal for adverse events	29% vs 18% placebo	Same	PMID 16437532
Donepezil vs placebo (AD2000)	MMSE +0.8 (0.5–1.2); BADLS +1.0 (0.5–1.6) over 2 years; institutionalisation 42% vs 44% at 3 y (P=0.4)	565 community patients, mild-moderate AD	PMID 15220031
Continuing vs stopping donepezil (DOMINO-AD)	SMMSE +1.9 (1.3–2.5); BADLS −3.0 (1.8–4.3) over 52 weeks	295 patients, moderate-severe AD	PMID 22397651
— nursing-home placement, year 1 after discontinuation	HR 2.09 (1.29–3.39); no difference years 2–4 (0.89, 0.58–1.35)	Same cohort, 4-year follow-up	PMID 26515660
Memantine, moderate-severe AD	SIB +3.11 (2.42–3.92); CIBIC+ 0.21 (0.14–0.30); ADL19 1.09 (0.62–1.64); NPI 1.84 (1.05–2.76)	Cochrane, up to 14 studies, ~3,700 participants	PMID 30891742
Memantine, mild AD	ADAS-Cog 0.21 (−0.95 to 1.38) — no benefit; discontinuation for AEs RR 2.12 (1.03–4.39)	Cochrane, post-hoc subgroups, ~600 participants	PMID 30891742
Donepezil 23 mg vs 10 mg	SIB +2.2 (P<0.001); CIBIC+ no difference; withdrawal 30.2% vs 17.9%	1,467 randomised, moderate-severe AD, 24 weeks	PMID 20678673
Cognitive stimulation	Cognition SMD 0.40 (0.25–0.55); MMSE 1.99 points (1.24–2.74); communication SMD 0.53 (0.36–0.70)	Cochrane, 37 RCTs, 2,766 participants	PMID 39804128
Cognitive training vs control	Global cognition SMD 0.42 (0.23–0.62); verbal semantic fluency 0.52 (0.23–0.81)	Cochrane, 33 RCTs	PMID 30909318
Cognitive training vs alternative treatment	SMD 0.21 (−0.23 to 0.64), low certainty	Same	PMID 30909318
Exercise (DAPA)	ADAS-Cog adjusted difference −1.4 (−2.6 to −0.2, P=0.03) favouring usual care	494 participants, 4 months supervised training, 12-month outcome	PMID 29769247
Donepezil 10 mg vs placebo, 24–26 wk	ADAS-Cog MD −2.67 (−3.31 to −2.02); MMSE +1.05 (0.73–1.37); withdrawal 24% vs 20%	Cochrane, 13 studies / 3,396 in main analysis	PMID 29923184
Galantamine 16–24 mg vs placebo, 6 mo	ADAS-cog MD −2.86 (−3.29 to −2.43); no MCI benefit (−0.21, −0.78 to 0.37)	Cochrane, 6 studies / 3,049 (AD); 2 studies / 1,901 (MCI)	PMID 39498781
Rivastigmine 6–12 mg oral or 9.5 mg patch, 26 wk	ADAS-Cog MD −1.79 (−2.21 to −1.37); withdrawal OR 2.01 (1.71–2.37)	Cochrane, 6 studies / ~3,200	PMID 25858345
Vitamin E 2,000 IU/day vs placebo (TEAM-AD)	ADCS-ADL decline 3.15 units less (0.92–5.39); ~6.2-month delay over 2.27 years; memantine no benefit	613 VA patients, mild-moderate AD on a ChEI	PMID 24381967

Neuropsychiatric symptoms¶

Intervention	Effect (95% CI)	Source
Brexpiprazole 2–3 mg, agitation	CMAI difference −5.32 (−8.77 to −1.87), P=0.003, Cohen d 0.35	PMID 37930669
Citalopram 30 mg, agitation	NBRS-A −0.93 (−1.80 to −0.06), P=0.04; mADCS-CGIC OR 2.13 (1.23–3.69); MMSE −1.05 (−1.97 to −0.13); QTc +18.1 ms (6.1–30.1)	PMID 24549548
Escitalopram ≤15 mg, agitation	Primary endpoint not significantly improved; PubMed abstract reports an internally inconsistent point estimate and CI, so no numeric effect is reproduced	PMID 40133524
Methylphenidate, apathy	NPI apathy −1.25 (−2.03 to −0.47), P=0.002; HR for no apathy 2.16 (1.19–3.91)	PMID 34570180
Sertraline / mirtazapine, depression	Cornell difference 1.17 (−0.23 to 2.58) / 0.01 (−1.37 to 1.38) — no benefit; adverse reactions 43% / 41% vs 26%	PMID 21764118
Suvorexant, insomnia	Total sleep time +28 minutes (11–45), P<0.01 at 4 weeks	PMID 31944580
WHELD person-centred care	QoL +2.54 (0.81–4.28); CMAI −4.27 (P=0.0076); NPI-NH −4.55 (P<0.001); cost saving	PMID 29408901
Atypical antipsychotics, agitation / psychosis	SMD −0.21 (−0.30 to −0.12) / −0.11 (−0.18 to −0.03); somnolence RR 1.93 (1.57–2.39); SAE RR 1.32 (1.09–1.61)	Cochrane, 24 RCTs, 6,090 participants
Pimavanserin discontinuation (HARMONY)	Relapse 13% vs 28% (HR 0.35, 0.17–0.73) among open-label responders	392 open-label, 217 randomised; stopped early
Stepwise pain protocol vs usual care	CMAI −7.0 (−3.7 to −10.3); 17% reduction	Cluster RCT, 352 nursing-home residents

9. Safety¶

Exposure	Harm	Population / method	Source
Lecanemab, pooled ARIA	19% (95% CI 16–23); symptomatic 3% (2–4); infusion reactions 26% (19–34); discontinuation for ARIA 5% (3–7)	2 RCTs + 5 real-world studies, 1,576 patients	PMID 41478817
— APOE4 gene dose, ARIA risk	RR 1.45 (heterozygotes), 3.54 (homozygotes) vs non-carriers	Same	PMID 41478817
Lecanemab, real-world clinic	ARIA 22% of 194 at risk; ARIA-E ± H 15%; isolated ARIA-H 6.7%; symptomatic ARIA 5.7%; severe 1.0%; no macrohaemorrhage or death; 4.3% withdrew for ARIA	234 consecutive patients, mean 6.5 months	PMID 40354064
— symptomatic ARIA by stage	27% (mild dementia) vs 1.8% (MCI/very mild dementia)	Same	PMID 40354064
Atypical antipsychotics, 10–12 weeks	Death 3.5% vs 2.3%; OR 1.54 (1.06–2.23); risk difference 0.01 (0.004–0.02)	15 trials, 3,353 on drug vs 1,757 placebo	PMID 16234500
Antipsychotic continuation vs withdrawal	12-month survival 70% (58–80) vs 77% (64–85); 24-month 46% vs 71%; 36-month 30% vs 59%	DART-AD, 165 randomised care-home residents	PMID 19138567
Cholinesterase inhibitors	Syncope HR 1.76 (1.57–1.98); bradycardia 1.69 (1.32–2.15); pacemaker 1.49 (1.12–2.00); hip fracture 1.18 (1.04–1.34)	19,803 exposed vs 61,499 unexposed, Ontario	PMID 19433698
Strong anticholinergics >1,095 TSDDs	Dementia OR 1.49 (1.44–1.54); PAF 10.3%	58,769 cases vs 225,574 controls, QResearch	PMID 31233095
Benzodiazepines	Pooled OR 1.33 (1.19–1.49); 1.14 (0.82–1.58) after lag-period correction	Meta-analysis, 35 articles	PMID 34679196
Delirium superimposed on dementia	In-hospital mortality 32% vs 8% (neither); adjusted HR 2.14 (1.33–3.45)	1,409 hospitalisations, geriatric ward	PMID 28350792
Suicide, first year after dementia diagnosis	Rate 26.42 per 100,000 py; SMR 1.53 (1.42–1.65); ages 65–74 SMR 3.40 (2.94–3.86)	2,667,987 Medicare beneficiaries	PMID 34036738
Suicide attempt, recent MCI / recent dementia	HR 1.73 (1.34–2.22) / 1.44 (1.17–1.77)	147,595 propensity-matched US veterans	PMID 33760039

10. Prevention and risk factors¶

Population attributable fractions (meta-analysis of 74 studies; 12 factors meta-analysable)¶

Factor	Unweighted PAF (95% CI)	Weighted PAF (95% CI)
Low education	17.2% (14.4–20.0)	9.3% (6.9–11.7)
Hypertension	15.8% (14.7–17.1)	7.1% (5.4–8.8)
Hearing loss	15.6% (10.3–20.9)	7.2% (5.2–9.7)
Physical inactivity	15.2% (12.8–17.7)	7.3% (3.9–11.2)
Obesity	9.4% (7.3–11.7)	5.3% (3.2–7.4)
Seven-factor model (9 studies)	55.0% (46.5–63.5)	32.0% (26.6–37.5)
Norton 2014 combined worldwide PAR, unadjusted → overlap-adjusted	49.4% (25.7–68.4) → 28.2% (14.2–41.5)	PMID 25030513

Source: PMID 38824956 unless noted. Weighted values adjust for communality between factors and are the appropriate figure for policy; PAFs were higher in low- and middle-income countries than in high-income countries. The ~30% overlap-adjusted combined figure has been stable across Barnes 2011, Norton 2014 and Stephan 2024.

Prevention trial results¶

Trial	Effect	Source
FINGER (n=1,260, 2 y)	NTB difference 0.022 z/year (0.002–0.042), P=0.030	PMID 25771249
US POINTER (n=2,111, 2 y)	Structured vs self-guided 0.029 SD/year (0.008–0.050), P=0.008; both arms improved	PMID 40720610
MAPT (n=1,680, 3 y)	Multidomain + omega-3 0.093 (0.001–0.184), adjusted P=0.142 — null	PMID 28359749
preDIVA (n=3,526, 6 y)	Dementia HR 0.92 (0.71–1.19), P=0.54 — null	PMID 27474376
SPRINT MIND (n=9,361)	Probable dementia HR 0.83 (0.67–1.04); MCI HR 0.81 (0.69–0.95); MCI or dementia 0.85 (0.74–0.97)	PMID 30688979
ACHIEVE (n=977, 3 y)	Difference 0.002 (−0.077 to 0.081), P=0.96 — null primary; cohort interaction p=0.010	PMID 37478886
Ginkgo biloba (GEM, n=3,069, median 6.1 y)	All-cause dementia HR 1.12 (0.94–1.33); AD HR 1.16 (0.97–1.39) — null	PMID 19017911
ADAPT + follow-up (~7 y)	AD: celecoxib HR 1.03 (0.72–1.50); naproxen HR 0.92 (0.62–1.35) — null	PMID 23562431
Herpes zoster vaccination (natural experiment, Wales)	−3.5 percentage points over 7 years (0.6–7.1), P=0.019; 20.0% relative reduction (6.5–33.4)	PMID 40175543

Selected risk-factor associations¶

Exposure	Effect	Source
APOE ε4/ε4 vs ε3/ε3 (white clinic/autopsy samples)	OR 14.9 (10.8–20.6)	PMID 9343467
APOE ε3/ε4 vs ε3/ε3 by ancestry	East Asian 4.54 (3.99–5.17); White 3.46 (3.27–3.65); Black 2.18 (1.90–2.49); Hispanic 1.90 (1.65–2.18)	PMID 37930705
APOE ε2 protective effect by ancestry	White 0.53 (0.48–0.58); Black 0.69 (0.57–0.84); Hispanic 0.89 (0.72–1.10, ns); East Asian 0.97 (0.77–1.23, ns)	PMID 37930705
TREM2 R47H	OR 2.92 (2.09–4.09) Iceland; 2.90 (2.16–3.91) combined	PMID 23150908
ABCA7 rs115550680 (African American)	OR 1.79 (1.47–2.12)	PMID 23571587
Heritability of AD	58% (full model) to 79% (best-fitting model)	PMID 16461860
Prevalent stroke → dementia	HR 1.69 (1.49–1.92), 36 studies, 1.9 million participants	PMID 30177276
Incident stroke → dementia	RR 2.18 (1.90–2.50), 12 studies, 1.3 million participants	PMID 30177276
Diabetes → incident AD	HR 1.65 (1.10–2.47)	PMID 15148141
≥2 midlife vascular risk factors → elevated late-life amyloid	OR 2.88 (1.46–5.69); late-life risk factors not associated (1.66, 0.75–3.69)	PMID 28399252
Midlife obesity → elevated late-life amyloid	OR 2.06 (1.16–3.65)	PMID 28399252
Midlife SBP ≥160 mm Hg → later AD	OR 2.3 (1.0–5.5); plus cholesterol ≥6.5 mmol/L, OR 3.5 (1.6–7.9)	PMID 11408299
Midlife obesity (BMI ≥30) → later dementia	HR 1.74 (1.34–2.26); overweight 1.35 (1.14–1.60)	PMID 15863436

11. Cost, care and health-system burden¶

Measure	Estimate	Year	Source
Global societal cost of dementia	US$1,313.4 billion; US$23,796 per person	2019	PMID 36617519
— direct medical / direct social / informal	16% / 34% / 50%	2019	PMID 36617519
People with dementia in LMICs vs share of costs in HICs	61% vs 74%	2019	PMID 36617519
Direct health spending attributable to ADRD (204 countries)	US$260.6 billion (95% UI 131.6–420.4)	2019	PMID 39150827
Informal care cost (same model), 57% of total	US$354.1 billion (190.0–544.1)	2019	PMID 39150827
Projected direct spending, 9.4% of world health spending	US$1.6 trillion (0.6–3.3)	2050	PMID 39150827
US unpaid caregivers / hours	>11 million / 18.4 billion hours	2023	PMID 38689398
Value of US unpaid care	US$346.6 billion	2023	PMID 38689398
US payments for health, long-term and hospice care, 65+ with dementia	US$360 billion	2024	PMID 38689398
QALY gain vs usual care, lecanemab / donanemab	0.38 / 0.51	modelled	PMID 42125995
Value-based price range, lecanemab / donanemab, high-income countries	US$254–9,434 / US$387–13,964	modelled, 174 countries	PMID 42125995
— low-income countries	US$4–18 / US$9–32	Same	PMID 42125995
Emergency admissions in England, adjusted excess length of stay (men / women with dementia)	+17% (15–18) / +12% (10–14)	2016/17	PMID 36888666
— 30-day post-discharge mortality	approximately doubled	Same	PMID 36888666
Caregiver vs non-caregiver, depression / stress / well-being	g = 0.58 / 0.55 / −0.40	meta-analysis, 84 articles	PMID 12825775
Pooled anxiety prevalence, informal dementia carers	32.1% (20.6–46.2)	meta-analysis, 10 studies	PMID 31409196
Missed/delayed dementia diagnosis (non-Hispanic White / Black / Hispanic)	41% / 46% / 54%; mean delay 31.2 / 34.6 / 43.8 months	HRS linked to claims, n=3,966	PMID 34091580
Previously diagnosed among community cases (Indonesia / South Africa)	0.2% / 0.5%	STRiDE	PMID 37278200
Eligible for lecanemab / donanemab among amyloid-positive memory-clinic patients	24.6% / 28.0% (9.1% / 10.3% of all clinic visitors)	Korean memory clinic, n=1,005 amyloid-positive of 2,726	PMID 41225766

12. Neuropathology at autopsy¶

Measure	Estimate	Population / method	Source
≥1 neuropathology	94%	1,079 autopsies, two longitudinal cohorts	PMID 29244218
≥2 / ≥3 / ≥4	78% / 58% / 35%	Same	PMID 29244218
AD present / AD in isolation	65% / 9%	Same	PMID 29244218
Distinct pathology combinations	>230, each in <6%	Same	PMID 29244218
AD share of person-specific cognitive loss	~50% mean; range 22–100%	Same	PMID 29244218
Among community dementia cases: AD + infarcts / pure AD / VaD / AD + PD-LBD	38.0% / 30.0% / 12% / 12%	First 141 Rush MAP autopsies (50 with dementia)	PMID 17568013
Multiple pathologies vs one, odds of dementia	OR 2.8 (1.2–6.7)	Same	PMID 17568013
LATE-NC at cognitively relevant levels	~25% of community autopsy brains	LATE consensus report	PMID 31039256
Neuropathologic AD subtypes	hippocampal-sparing 11%, typical 75%, limbic-predominant 14%	889 autopsy cases, tangle-density algorithm	PMID 21802369
Neocortical p-tau burden within Braak stage V	0.2% to 53.7%	61 brains, quantitative pixel assessment	PMID 42184025
Posterior cortical atrophy: AD at autopsy	94% (90–97); CAA 71%, Lewy body 44%, cerebrovascular injury 42%	145 autopsies within a 1,092-participant IPD meta-analysis	PMID 38267189

13. Evidence-breadth additions — 2026-09-03¶

Assay and imaging performance¶

Measure	Estimate	Population / method	Source
p-tau217, Lumipulse vs ALZpath Simoa, AD vs other neurodegeneration	AUC 0.952 (0.927–0.978) vs 0.955 (0.928–0.982)	392 participants, head-to-head	PMID 39679606
Ten plasma p-tau assays, abnormal amyloid	Best AUC 0.947; assay range 0.642–0.947	135 MCI, mean follow-up 4.9 y	PMID 36087307
Blood-biomarker evidence base	Sensitivity 49.3–91.4%; specificity 61.5–96.7%; certainty moderate to very low	49 studies, 31 tests	PMID 41193403
Tau PET added to base model, all-cause dementia	AUC 0.71 → 0.75 (discovery cohort)	331 MCI discovery + 117 external validation, mean follow-up 2.0 y	PMID 38857029
FDG-PET vs ASL-MRI, MCI subgroup	AUC 0.92 vs 0.80; sensitivity 0.90 vs 0.75; specificity 0.91 vs 0.73	Seven-study direct-comparison meta-analysis	PMID 40898829
Management-plan change after amyloid PET	59.0% (57.6–60.5)	New IDEAS, 5,757 Medicare beneficiaries	PMID 40728069

Treatment, care and safety additions¶

Measure	Estimate	Population / method	Source
Lecanemab core + extension: ARIA-E / ARIA-H microhaemorrhage / infusion reaction	13.6% / 16.0% / 24.5%; ARIA-E 34.5% in APOE ε4 homozygotes	1,612 exposed participants	PMID 38730496
Care Ecosystem: quality of life / caregiver depression / burden	+0.53 (0.25–1.30) / −1.14 (−2.15 to −0.13) / −1.90 (−3.89 to −0.08)	780 dyads, 12-month RCT	PMID 31566651
New psychotropic class after hospital discharge / continued >90 days	6.6% / 51% of new users	117,022 Medicare beneficiaries with dementia	PMID 36514208
Memantine vs placebo: cognition / behaviour	SMD −0.24 (−0.34 to −0.15) / −0.16 (−0.29 to −0.04)	30-trial meta-analysis, 7,567 participants	PMID 28922160
Brexpiprazole 1 mg / 2 mg, CMAI difference	−3.7 (−6.8 to −0.7) / −7.2 (−10.0 to −4.3)	Japanese 10-week RCT	PMID 39369280
Suicide within 1 year of dementia diagnosis	adjusted HR 2.57 (1.49–4.44); AD 2.50 (1.41–4.44)	36,541 newly diagnosed older adults, South Korea	PMID 33119492

Copathology and costs¶

Measure	Estimate	Population / method	Source
CAA prevalence / dementia with vs without CAA	38.0%; 53.7% vs 40.1%, adjusted OR 1.57 (1.18–2.10)	ACT community autopsy cohort, n=848	PMID 39481068
CAA without AD / AD without CAA / both	11% / 16% / 20%	Beijing brain-bank autopsies, n=483	PMID 40201593
Annual AD cost per person, mild → severe	US$468.28 → US$171,283.80	12-study AD-specific systematic review	PMID 37972428

14. Known conflicts and caveats¶

Two GBD vintages, two global counts. 43.8 million (2016) and 57.4 million (2019) are not a time series; they come from different models. Neither is a count of Alzheimer's disease specifically.
Indonesia at 27.9% versus India at 7.2%. Both are recent, nationally or regionally representative surveys, but they used different instruments (10/66 short schedule versus DSM-5 algorithm on a neuropsychological battery). This knowledge base keeps both and does not reconcile them.
Incidence is falling and rising. −13% per decade across seven Western cohorts (PMID 32611641) coexists with a rise then fall in Hisayama (PMID 28424272; PMID 41466306). Pooled meta-analysis of high-income countries found the decline non-significant (0.82, 0.51–1.33) (PMID 30271219).
Prevalence trends confound incidence with survival. Hisayama 5-year survival after onset rose from 47.3% to 65.2% between the 1988 and 2002 cohorts, which raises prevalence independently of incidence (PMID 28424272).
Death-certificate counts measure coding practice. US reported AD deaths rose >140% between 2000 and 2021 while stroke and heart disease deaths fell (PMID 38689398); GBD models dementia mortality from prevalence and excess mortality precisely because of this instability (PMID 30497964).
Biomarker accuracy is optimistic. ~90% of plasma p-tau studies used data-derived rather than predefined thresholds (PMID 40818474), and assays are not interchangeable (0.93 vs 0.83–0.88 accuracy) (PMID 39468767).
Two cost models, two totals. Wimo's US$1,313.4 billion and Lastuka's US$614.7 billion (direct plus informal) use different cost bases and informal-care valuation methods; both agree informal care is roughly half (PMID 36617519; PMID 39150827).
The benzodiazepine–dementia association does not survive lag correction (OR 1.14, 0.82–1.58) and should not be quoted as 1.33 without that qualification (PMID 34679196).
Trial effect sizes are below published MCIDs for the same instruments (PMID 38561021); reporting the treatment difference without the MCID misrepresents its interpretation.
Memory-clinic survival figures are left-truncated. Median 7.2 years from first clinic visit (PMID 40294367) is not comparable with 5.8 years mean from diagnosis in a broad meta-analysis (PMID 36097997).
Race and ethnicity categories are social, not biological. The incidence differences in section 3 co-occur with 3–13-month longer diagnostic delays in the same groups (PMID 34091580), so measured incidence and true incidence may differ by group.
PAFs must not be summed. Individual factors overlap; the weighted seven-factor figure of 32.0% is the correct combined estimate, not the sum of the individual unweighted values (PMID 38824956). Norton 2014 made the same correction (49.4% → 28.2%) a decade earlier (PMID 25030513).
TREM2 agonism engaged target without clinical benefit. INVOKE-2 (AL002) reduced CSF sTREM2 and raised osteopontin but missed CDR-SB at every dose, with ARIA-like MRI as the leading TEAE (PMID 41787076). Observational sTREM2–protection associations (PMID 31462511; PMID 41605308) therefore do not license this antibody, at this stage, as a treatment.