Every row of the evidence table

This one reports light-to-moderate drinking as protective (HR 0.65) against non-drinking, and it flatly contradicts the Mendelian randomisation evidence. Infrequent drinkers were excluded, and non-drinkers include people who stopped for health reasons.

Direction is for QUITTING. Risk approached that of never-smokers and levelled off about seven years after quitting; weight gain after quitting narrowed the advantage. The authors flag residual confounding.

Direction is for more recent birth cohort: after controlling for age and period, later-born cohorts were less likely to have dementia in all three regions, more so among women; the point estimates for men in England were not significant.

Direction is for later birth cohort: each successive cohort had a significantly lower incident dementia rate, independent of sex, education and age.

Direction is for the later survey year: prevalence fell from 11.6% to 8.8%, with rising education associated with some of the decline despite a worse cardiovascular risk profile.

Direction is for later calendar epoch: incidence fell 44% from the first to the fourth epoch, but only among people with at least a high school diploma. The authors could not fully explain the decline.

Direction is for the later-born population: prevalence was 6.5% against the 8.3% expected from the earlier study (odds ratio 0.7), which the authors read as a cohort effect.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The estimated age at plasma %p-tau217 positivity predicted age at symptom onset (adjusted R2 0.337-0.612) with a median absolute error of 3.0-3.7 years, which the authors call acceptable for trials but not yet for individual clinical use.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Stand-alone plasma p-tau217 gave 81% accuracy and 79% PPV; confirming positives with PET or CSF raised both to 91% while cutting PET scans needed to find 100 amyloid-positive people from 536 to 124. Immunoassays matched mass spectrometry on PPV but had lower accuracy.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Combined plasma %p-tau217 and Aβ42/40 detected abnormal CSF amyloid status (AUC 0.949) and each predicted rising amyloid PET over time in people who started below the positivity threshold, replicated in two further cohorts; p-tau231 and GFAP added no clear independent value.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. All four assays (Lumipulse, Johnson & Johnson, ALZpath, NULISA) gave statistically equivalent AUCs of 0.84-0.90 for amyloid positivity, with cross-platform correlations of 0.40-0.86.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Among p-tau217-positive patients, a positive eMTBR-tau243 identified clinically established Alzheimer's disease with 81% accuracy and predicted faster cognitive decline and tau accumulation.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Subgroup-specific optimal cutoffs and a double-cutoff strategy both improved accuracy over the standard single cutoff, most markedly in chronic kidney disease (0.65 to 0.83) and anaemia.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The prespecified APS2 cut point gave 91% accuracy, 90% sensitivity and 92% specificity against amyloid PET.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma p-tau217 detected amyloid PET positivity with 82% sensitivity and 86% specificity, comparable to CSF p-tau217; less sensitive for tau PET.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. AUC 0.93-0.96; accuracy 89-91% in secondary care and 85% in primary care, lower (83%) at age 80 or older; a two-cutoff approach raised accuracy to 92-94% at the cost of 12-17% intermediate results.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma p-tau217 was the best-performing marker (88.1% sensitivity, 88.7% specificity, AUROC 91.1%, moderate certainty), ahead of p-tau181 and p-tau231; about 90% of studies were at high risk of bias.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The Pitt-p-tau217 assay separated PSEN1 carriers from controls (AUC 0.94) and amyloid PET-positive from -negative cognitively normal adults (AUC up to 0.84), equivalent to the ALZpath assay.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. All p-tau217 tests detected abnormal amyloid PET (AUC 0.91-0.96) and tau PET (0.94-0.97); the mass-spectrometry %p-tau217 ratio was best (accuracy 0.93 versus 0.83-0.88 for immunoassays) and is proposed as a stand-alone confirmatory test, with immunoassays better suited to triage.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The Lumipulse plasma p-tau217/Aβ42 ratio reached AUC 0.963-0.966 for amyloid PET and 0.947-0.974 for tau PET, equivalent to CSF tests and better than plasma p-tau217 alone, with fewer intermediate results.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma p-tau217/Aβ42 had the highest accuracy (0.88 in Black and 0.91 in White participants) and fewest intermediate results, with no difference in classification by race.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma p-tau217 identified symptomatic AD with 95% sensitivity and 82% specificity (86% with the Aβ42 ratio) and tracked CSF biomarkers closely (AUC 0.94-0.96); reduced kidney function raised levels in patients without AD.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The APS2 blood test reached AUC 0.96-0.97 and 88-92% accuracy, and its 91% accuracy for clinical AD beat primary care physicians (61%) and dementia specialists (73%).

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. APS2 gave AUC 0.94 and 88% agreement with amyloid PET, with similar performance by ethnicity, sex, age and APOE4 status.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The ALZpath plasma p-tau217 immunoassay reached AUC 0.92-0.96 for amyloid and 0.93-0.97 for tau across three cohorts, comparable with CSF, and a three-range cut-off cut confirmatory testing by about 80%.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma p-tau217 identified amyloid positivity with AUC 0.94; a two-cut-point rule sent only 17% for CSF or PET (sensitivity 0.94, specificity 0.86), and among amyloid-positive people it identified high tau load (AUC 0.92), potentially avoiding 56.9% of tau-PET scans.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. A blood model of p-tau217, age and APOE ε4 (AUC 89.3% development, 94.3% validation) used to triage patients gave 88-92% overall accuracy while cutting confirmatory CSF tests by 61-86%.

Direction is for disclosing the blood panel during multidisciplinary meetings: median diagnostic confidence rose from 80% to 90%, though the suspected etiology changed in only 5% of patients and the syndrome diagnosis in 1%. Against CSF/PET, high- and low-probability results identified 76% of amyloid positives and 85% of negatives.

Direction is for the blood test's prognostic performance: plasma p-tau217 predicted cognitive decline and progression to MCI as well as tau PET (R2 0.33 vs 0.34; HR 1.57 vs 1.61), and using the blood test first cut the number needed to screen for a trial by 94%. A prognostic-performance result, not a health outcome.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The two-marker plasma model matched PET-based stages with a C index of 0.91 in both the derivation and validation cohorts, tracked clinical stage (C index 0.84-0.86) and aligned with autopsy ADNC (AUC 0.96), improving on %p-tau217 alone at the intermediate stage.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Plasma Aβ42/40, p-tau217, p-tau181 and NfL classified AD with AUC 83% and FTLD with 88%, consistent across countries, rising to 89% and 95% when combined with neuroimaging patterns and cognitive measures.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Serum GFAP tracked post-mortem tau pathology independently of amyloid, correlated negatively with brain weight and cognition, correlated with astrocyte reactivity in tissue, and separated neuropathologically confirmed AD from non-AD dementias with AUC 0.91.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Smell testing followed by CSF seed amplification only in those with reduced smell predicted brain Lewy body pathology with 94% accuracy and cut CSF testing by 43%; in a living cohort it predicted CSF assay status with 79% accuracy.

Benefit was larger for younger, male and non-spousal carers. The largest single effect (training the person with dementia) rests on 10 observations and the authors call it promising rather than definitive.

A pilot designed to test feasibility; 68 carers and a 2-week outcome.

Digital psychological interventions reduced caregiver burden and improved self-efficacy and quality of life; the effect on depressive symptoms was not statistically significant (P = .06).

At the 1-month comparison the automated self-guided programme reduced depression and burden and increased positive gains versus waitlist; anxiety did not differ. Group differences mostly disappeared once controls received the programme.

A factorial trial of components: the mindfulness component carried the benefit and amplified the others.

Direction is for NIDUS-Family: an 89% probability of cost-effectiveness from a health and social care perspective (87% societal) and £8,934 (37%) lower average costs, though the cost difference's confidence interval includes zero.

The outcome is how prepared carers felt to decide about firearm access, not whether any injury was avoided.

Direction is for psychosocial and pharmacological interventions, which the review reports as showing mild to modest efficacy in high-quality meta-analyses. A narrative clinical review summarising other meta-analyses, not a new pooled estimate.

Benefit in Hispanic and White caregivers and in Black spousal caregivers; not in Black non-spousal caregivers.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Tablet tests correlated strongly with MMSE, MoCA and Trails A (r > 0.7) and the Match test separated amyloid-positive from -negative patients (AUC 0.82) better than MMSE or MoCA (0.75).

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Pooled sensitivity 76% and specificity 83% for dementia, with 73% / 84% in primary care; the authors call it a practical screening tool while noting high or unclear risk of bias in patient selection and timing.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. The two-stage GPCOG had sensitivity 0.85 and specificity 0.86 against DSM-IV dementia, and was superior to the Abbreviated Mental Test.

Direction is for value for money: the intervention was modelled as dominant, saving JPY 452,826 per person and adding 0.08 QALYs over 35 years, with sensitivity analyses described as confirming robustness. A model extrapolating an 18-month trial.

The community health worker programme achieved verified linkage to medical care in 16.7% of dyads versus none in the control arm. Caregiver outcomes were largely unchanged, with self-efficacy favouring control.

Direction is for the extended nursing role intervention: 74% fewer unmet needs and higher quality of life at six months, with a small QALY gain at about 1,425 EUR added cost; no effect on caregiver burden.

Direction is for the Care Ecosystem programme: total cost of care was $475.80 per member per month lower than matched controls at 12 months, with fewer emergency, outpatient and professional visits.

Direction is for collaborative dementia care management: fewer neuropsychiatric symptoms, lower caregiver burden, better mental health and 0.137 QALYs gained at no significant added cost (3186€ per QALY) over 36 months; no difference in antidementia drug use, overall quality of life or inappropriate medication.

The outcome is how many diagnoses were made, not whether anyone was better off. The passive digital marker alone did not raise detection (AOR 0.84, 95% CI 0.63-1.11).

Direction is for the benefit redesign: projected out-of-pocket drug spending fell 55% ($56 to $25) for low-income-subsidy beneficiaries and 25% ($772 to $576) for others, with larger reductions for people with 6-8 comorbidities.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Driving features distinguished drivers with MCI from those with normal cognition (AUC 0.82), rising to 0.87 with demographics, APOE4 and cognition; drivers with MCI showed greater declines in trip frequency, night driving and route variety.

Direction is for the test's accuracy against the reference standard; an accuracy result, not a health outcome. Cross-validated balanced accuracy of 70-95% and AUC above 78% across the six conditions; the model is a development-stage tool with one external memory-clinic validation sample.

The outcome is how prepared carers felt to decide about firearm access, not whether any injury was avoided.

Direction is for DREAMS START against treatment as usual: a 78% probability of being cost-effective at £20,000 per QALY, with cost and QALY differences individually non-significant. The clinical sleep result is in the table under its own row.

Which intervention helped depended on severity: stimulation for cognition at any severity, training for global cognition in mild dementia, rehabilitation for function in mild dementia and for carer burden. Heterogeneity was high.

The association was strongest at moderate intake rather than the highest.

882 cases across five quintiles and several interaction strata, so the subgroup estimates rest on small numbers.

The association is reported in people who already have elevated Alzheimer's blood markers, which is the point of the study.

Direction is for a healthier plant-based diet: higher overall and healthful plant-based diet scores were associated with lower dementia risk, while a higher unhealthful plant-based diet score was associated with higher risk (pooled HR 1.42).

Independent of overall diet quality. Substituting olive oil for margarine or mayonnaise was also associated with lower risk; substituting for butter or other vegetable oils was not.

Pooled observational studies only.

Observational. Points the opposite way to the randomised MIND diet trial.

Significant on one adherence score (MEDAS, HR 0.77) but not the other (PYRAMID, HR 0.86, 95% CI 0.73-1.02).

Controls declined while the olive-oil arm held steady; the nut arm fell between the two. Cognitive follow-up was available for 334 of 447 enrolled.

Observational. Points the opposite way to the randomised MIND diet trial.

Pooled observational studies only.

Significant on one adherence score (MEDAS, HR 0.77) but not the other (PYRAMID, HR 0.86, 95% CI 0.73-1.02).

Controls declined while the olive-oil arm held steady; the nut arm fell between the two. Cognitive follow-up was available for 334 of 447 enrolled.

Dietary intake and blood levels, pooled from observational cohorts. The much larger supplement estimate (HR 0.36) comes from one cohort's self-selected supplement users, not from a trial.

Compared with flexible health advice, not with nothing; both groups improved.

Both groups improved; the structured programme improved more. The trial compares two active interventions, so there is no untreated comparison.

Concentrating activity into one or two days worked about as well as spreading it across the week.

Direction is for 12 weeks of cycle-ergometer exercise: putative glymphatic influx in the putamen and meningeal lymphatic vessel size and flow increased, and inflammation-related plasma proteins fell in step with the imaging changes. A mechanistic imaging result, not a cognitive outcome.

The authors call the association weak (RR 0.97), and it was absent beyond 10 years of follow-up.

Pooled across 29 cohorts. The association weakened to HR 0.85 after maximal confounder adjustment and was not significant for low-intensity activity.

The authors state the association held in longer follow-ups, which is their answer to the reverse-causation objection raised by the Whitehall II cohort.

866 cases over a short follow-up for a disease with a long prodrome, so some of the association may be people already declining moving less.

The largest effect for physical activity anywhere in this table by a wide margin. On 678 cases over 6.3 years, in a disease with a long prodrome, reverse causation is the more likely explanation for an effect that size.

Pools 27 small trials in people who already have dementia, and reports a much larger effect (SMD 0.69) than the single largest such trial, DAPA, which found none.

The authors call the effect small (-0.18 points). Significant in women, not in men, on 91 randomised and 76 completing.

Seven small trials; working memory did not improve (P=0.14).

Direction is for carrying the deletion, present in 60% of African American ε4 homozygotes: it reduced the AD odds ratio and delayed onset in ε4/ε4 cases with local African ancestry at APOE, confirmed in All of Us, and it abolishes SPI1 repression at the site.

Direction is for carrying APOE2, which is associated with longer life irrespective of Alzheimer's status. The lifespan figures come from gene-targeted mice, not from the human cohort.

Direction is for one copy of the APOE3 Christchurch variant: median onset of cognitive impairment at 52 versus 47 years in matched PSEN1 E280A carriers without it, with preserved brain metabolism and fewer vascular amyloid features in the few imaged or autopsied. Not modifiable.

Direction is for carrying the rare FN1 variant rs140926439 in APOE ε4 homozygotes: odds of AD 0.29 and onset delayed 3.37 years in the 7185-person validation cohort, with zebrafish work suggesting fibronectin loss of function reduces gliosis. Not modifiable.

Confined to people who said the aid actually improved their hearing (HR 0.86); those reporting poor improvement showed nothing (HR 0.98).

Mostly observational (25 of 31 studies), so people who take up hearing aids may differ from those who do not. The authors call for randomised trials — and ACHIEVE, the trial that followed, was null overall.

Against untreated cataract. People who had surgery ended up at about the same risk as controls who never had cataracts (HR 0.84, 95% CI 0.66-1.06), and there was no short-term gain in those already cognitively impaired.

Regression discontinuity across a birth-date eligibility threshold, which makes confounding by health-seeking behaviour less likely than in a plain cohort comparison. Eligibility did not change uptake of other preventive services or other chronic-disease diagnoses, which is the check that makes the design convincing.

Regression discontinuity across the 2 September 1933 eligibility threshold; stronger in women than men.

Pooled observational studies, so open to the confounding the natural-experiment studies were designed to avoid.

Direction is for antisense oligonucleotides lowering APP in neurons derived from people with an APP duplication or trisomy 21: reduced APP, rescued endolysosome and autophagy dysfunction and fewer intracellular and extracellular amyloid-β aggregates. Cell-culture work with no human outcome.

Phase 1 target-engagement result in 19 people; no clinical outcome measured.

Prespecified and post hoc re-analyses of a trial already reported elsewhere, so this is not independent evidence.

Direction is for lecanemab in mice: it significantly reduced amyloid pathology and neuritic damage, an effect absent with Fc-silenced antibody or without microglia despite intact plaque binding. Mouse work with no human outcome.

An initial negative opinion was reversed on re-examination, with use restricted to APOE4 heterozygotes and non-carriers. The EPAR describes ARIA as affecting more than 1 in 10 people without giving a percentage.

Phase Ib/IIa biomarker result in 13 people; no cognitive outcome measured.

Direction is for the slower titration schedule against standard dosing, on the drug's own side effect. Amyloid reduction was comparable.

An open-label extension: after 18 months everyone is on the drug, so the comparison is early-start versus delayed-start, not drug versus placebo.

Less decline than placebo, alongside infusion reactions and ARIA.

Phase 1 dose-finding. The outcome is amyloid clearance; no cognitive outcome was measured.

Benefit is on the surrogate marker only: the approval rested on amyloid reduction, not on confirmed clinical benefit, and the confirmatory trial was later terminated.

Phase 1b safety and target-engagement result: adverse events were all mild or moderate with no serious events on MAPTRx, and CSF total tau fell dose-dependently by more than 50% in the two highest-dose groups. No clinical outcome was measured.

Significant for MCI and for the MCI-or-dementia composite; not significant for probable dementia alone (HR 0.86, 95% CI 0.72-1.02).

A large drop in blood pressure (22/9 mmHg) and fewer serious adverse events in the intervention villages.

Individual participant data from five double-blind placebo-controlled trials, which is the strongest design in this table for this exposure.

Consistent across 14 trials (I-squared 0%) but small in absolute terms: 0.39 percentage points over about 4 years.

Direction is for STOPPING the antipsychotic: discontinuers were hospitalised less than matched continuers.

Direction is for discontinuation: tapering after 12 weeks lowered the 24-month absolute risk of delirium and fracture versus continuing, with no difference in stroke, pneumonia or death. Results were similar for abrupt discontinuation and after 24 weeks.

Observational: 80% of residents were on no anti-dementia drug, so who gets prescribed and who does not may itself track prognosis.

Cholinesterase inhibitors, not memantine: -0.39 MMSE points a year against -2.50 with no treatment. The mortality benefit held only in the first year. Registry data with weighting, not a trial.

Seven separate comparisons against seven other antidiabetic drugs, so the comparator is always another drug rather than no treatment.

Direction is for AAVrh.10 delivery of APOE2 with the Christchurch mutation in mice: both vectors suppressed amyloid, but only the Christchurch variant also suppressed tau tangles and glial activation in tau mice and outperformed plain APOE2 behaviourally. Mouse work with no human outcome.

Heterogeneity was near-total (I-squared 98.9%) and the estimate went to null in people newly diagnosed with diabetes (HR 1.01), which is where prevalent-user bias would be smallest.

Comparator is other oral antidiabetic drugs, not no treatment.

The exposure actually studied is STOPPING metformin early, which carried 21% higher dementia incidence than staying on it. Read as indirect support for metformin, not as a trial of it.

Direction is for peripheral cancer in AD mice: tumour-secreted cystatin-C bound amyloid oligomers and activated microglial TREM2, enabling degradation of existing plaques and rescuing cognition; the effect was abolished by microglial deletion or mutation of TREM2 or cystatin-C. Mouse work with no human outcome.

Direction is for lowering PU.1 in microglia, which reduced the severity of amyloid pathology in mice; microglia-specific CD28 deficiency did the opposite and increased plaque load. Mouse work with no human outcome.

Direction is for deleting TIM-3 in microglia: it increased phagocytic activity, shifted microglia to the disease-associated state, reduced amyloid pathology and improved cognitive impairment in 5xFAD mice. Mouse work with no human outcome.

Direction is for blocking the ZBP1-RIPK1 axis: genetic deletion of Zbp1 or inhibition of RIPK1 attenuated neuroinflammation, amyloid pathology and behavioural deficits in AD mice, after showing that amyloid-induced oxidised mitochondrial DNA forms Z-DNA sensed by ZBP1 in microglia. Mouse work with no human outcome.

Direction is for removing the granzyme K-producing CD103-negative CD8+ T cells: their ablation ameliorated cognitive decline and reduced neuropathology in 3xTg-AD mice, and granzyme K caused neuronal dysfunction and tau hyperphosphorylation via PAR-1. Mouse and cell work with no human outcome.

Direction is for inhibiting C1q, C3 or the microglial receptor CR3, which reduced phagocytic microglia and early synapse loss; C1q was necessary for soluble amyloid-beta oligomers to damage synapses. Mouse work with no human outcome.

Direction is for NLRP3 or caspase-1 deficiency, which largely protected the mice from spatial memory loss and reduced amyloid deposition; the human data are limited to raised active caspase-1 in post-mortem MCI and Alzheimer brains.

The NPI-C agitation/aggression measure did not reach significance (p=0.094).

The authors describe the reduction as small (-1.32 points).

A post hoc dementia subgroup of a randomised-withdrawal trial that was not designed or powered for it; 46 randomised responders. The authors say to read it cautiously.

Both doses beat placebo. Treatment-emergent adverse events were more common than placebo (84.6% at 2 mg, 76.8% at 1 mg, 73.8% placebo).

For SGLT2 inhibitors specifically: HR 0.86 (95% CI 0.79-0.94) against DPP4 inhibitors, and HR 0.70 in the as-treated analysis.

Benefit was confined to APOE4 carriers (HR 0.60) and absent in non-carriers (HR 0.96). Observational, and it disagrees with the larger target-trial emulation that found nothing.

Direction is for MORE education. Up to a quarter of the effect ran through midlife vascular risk factors, and the authors say treating those factors would not close the educational gap.

Education and job complexity each contributed independently; about 28% of the education effect ran through occupation.

Effect held over and above measured Alzheimer's pathology, which is the point of the reserve idea.

Improves depressive symptoms (moderate certainty) and overall behavioural problems (low certainty). No effect on agitation (SMD -0.05, 95% CI -0.27 to 0.17) and no evidence of long-term effects. Sits against the null MIDDEL trial on the same exposure.

Direction is for normal sleep versus sleep deprivation, on a clearance measure — not on cognition, which this trial did not measure.

Significant only for contact measured at age 60, not at 50 or 70. The authors note that keeping up contact may itself be a marker of cognitive reserve rather than a cause of it.