Skip to content
Umbrella Aged Care
All topics

Emerging evidence: a clinical appraisal

GPs & clinicians 20 min read

This is the technical companion to the family-facing supplement page. It is written for GPs, nurse practitioners, pharmacists and aged care clinicians who want the trial data, the effect sizes, the Australian access position and the methodological caveats rather than a traffic-light summary — including the adjuncts and repurposed agents patients raise in consultations.

Numbers in the text link to the 45 sources listed at the foot of this page.

Anti-amyloid monoclonal antibodies: registered in Australia, not funded, and not for this population

The Therapeutic Goods Administration registered donanemab (Kisunla) in May 2025 58 and lecanemab (Leqembi) in September 2025, the latter for mild cognitive impairment and mild dementia due to Alzheimer’s disease with confirmed beta-amyloid pathology 57. Neither is PBS-listed. Out-of-pocket cost for the drug alone has been estimated at $40,000–50,000 per year in Australia, rising to $80,000–100,000 once amyloid confirmation and MRI surveillance are included. Recommended monitoring includes MRI three to four times in the first six months to detect amyloid-related imaging abnormalities 59.

The effect size deserves plain statement. In CLARITY-AD (n=1,795), lecanemab produced an 18-month difference on the Clinical Dementia Rating Sum of Boxes of −0.45 points on an 18-point scale — around 27% slowing of decline 79. In TRAILBLAZER-ALZ 2 (n=1,736), donanemab produced a CDR-SB difference of −0.67 points in the low-to-medium tau population, roughly a third 80. These are real, statistically robust effects. They are also small in absolute terms, they slow rather than halt decline, and whether the difference crosses the threshold of clinical meaningfulness for an individual patient remains genuinely contested among specialists, not merely among sceptics. Australian cost-effectiveness modelling has been published in the Medical Journal of Australia and is worth reading before fielding family questions 61.

The safety profile is the other half of the equation. Amyloid-related imaging abnormalities — oedema and microhaemorrhage — are common: ARIA with oedema or effusion occurred in 12.6% on lecanemab 79 and 24.0% on donanemab, against roughly 2% on placebo, and three donanemab deaths in TRAILBLAZER-ALZ 2 were considered treatment related 80. Risk is substantially higher in APOE ε4 homozygotes, and anticoagulation increases the risk of serious haemorrhagic events — relevant given how many older Australians take an anticoagulant for atrial fibrillation.

For residential aged care the practical implication is simple: essentially no current RACF resident is eligible. These drugs are for early symptomatic disease with confirmed amyloid pathology and no significant comorbid burden — the opposite of the typical resident profile. The clinical task in aged care is therefore mostly counselling: explaining to adult children who have read a headline why their 88-year-old mother with advanced dementia, atrial fibrillation and a prognosis measured in months is not a candidate, and why that is a clinical judgement rather than rationing.

GLP-1 receptor agonists: the EVOKE result

Oral semaglutide was tested in EVOKE and EVOKE+, two phase III trials enrolling 3,808 adults aged 55–85 with mild cognitive impairment or mild dementia due to Alzheimer’s disease, with a two-year primary treatment phase. Topline results reported at the Clinical Trials on Alzheimer’s Disease conference in December 2025 showed no statistically significant slowing of disease progression on the Clinical Dementia Rating Sum of Boxes, and no meaningful differences on secondary cognitive or functional outcomes 60. Reductions of up to around 10% were observed in some mechanistically consistent biomarkers, without translating into clinical benefit. The one-year extension phase was discontinued 60.

This matters beyond GLP-1s. It is another instance of a well-powered trial in which biomarker engagement did not produce clinical benefit — the same disconnect seen with resveratrol, nicotinamide riboside and several failed anti-amyloid programmes. Clinicians being asked about metabolic drugs for brain health should be clear that the current position is negative, not pending.

Blood-based biomarkers: where they help and where they do harm

Plasma p-tau217 has excellent diagnostic accuracy for Alzheimer’s pathology in specialist cohorts, approaching that of amyloid PET and CSF analysis. Work through the Australian Dementia Network is examining performance in primary care populations, where lower pre-test probability materially changes positive predictive value 73.

Three cautions apply in Australian general practice. First, availability and funding are not settled, and assay standardisation across platforms is an active issue. Second, these tests answer the question “is Alzheimer’s pathology present?”, not “does this person have dementia?” or “when will they decline?” — pathology is present in a substantial minority of cognitively normal older adults. Third, a positive result in a person for whom no disease-modifying therapy is accessible or appropriate generates prognostic anxiety without a corresponding action.

The reasonable position is that biomarker testing belongs inside a structured diagnostic pathway with specialist involvement and a plan for what a positive or negative result will change. It does not belong on a general screening panel.

Why the supplement literature keeps producing false positives

The recurring failure mode is worth naming explicitly, because it explains almost every discrepancy on the family-facing page.

Confounding by health-consciousness. Supplement users differ systematically from non-users on education, income, physical activity, smoking, alcohol and health service engagement — every one of which independently predicts dementia risk. Residual confounding survives statistical adjustment. This is why an ADNI analysis can report a 64% reduction in Alzheimer’s risk among long-term omega-3 users 17 while randomised trials of the same intervention return a standardised mean difference of −0.02 16.

Reverse causation. Early dementia reduces appetite, dietary variety and self-care. Low serum nutrient levels may therefore be a consequence of preclinical disease rather than a cause. This affects the vitamin D, B12 and omega-3 literatures particularly.

Underpowered trials with slower-than-expected control decline. LipiDiDiet missed its primary endpoint largely because the control group declined far less than the power calculation assumed 24. MIND was well powered but found both arms improving 8. Where the control group does not decline, no intervention can demonstrate slowing.

Practice effects on repeated cognitive testing. Both arms of US POINTER improved 5. Both arms of MIND improved. Any single-arm, open-label study reporting cognitive improvement after eight weeks — the creatine pilot in Alzheimer’s disease is the clearest recent example 32 — cannot separate a treatment effect from a practice effect and expectation.

Post-hoc subgroup rescue. Negative trials routinely generate positive subgroups: high-homocysteine responders in VITACOG, high-omega-3 responders in the same trial, high-adherence responders in MIND, high-biomarker responders in MIND 8687. Some of these are real biological effect modification. Most are noise. The distinguishing test is prospective replication in a trial designed around the subgroup, and that has rarely been done.

Selective reporting of secondary endpoints. A trial that misses its primary endpoint but reports several significant secondary endpoints is hypothesis-generating. This is the correct reading of LipiDiDiet and of much of the EGb 761 literature.

Agents worth tracking, with the appraisal

These are the interventions where the underlying science is interesting enough that the position may change. Current status is stated plainly.

Creatine monohydrate

Promising, not yet proven

Mechanistically coherent, safe, cheap, and awaiting a controlled trial in cognitive impairment.

The evidence

Impaired cerebral bioenergetics precedes symptomatic Alzheimer’s disease, and brain creatine kinase systems are disrupted. The 2025 CABA pilot in 20 patients with Alzheimer’s disease achieved 11% increase in whole-brain total creatine on magnetic resonance spectroscopy with 20 g/day over eight weeks, with improvements in global and fluid cognition 32. Meta-analysis in healthy adults shows a memory effect concentrated in older participants (SMD 0.88 in the 66–76 year subgroup) 33.

In practice

Target engagement is demonstrated; efficacy is not. The pilot was single-arm and open-label. Independent musculoskeletal benefits — grip strength and muscle cross-sectional area improved in the same cohort 3275 — mean the risk–benefit calculation in sarcopenic older adults is favourable regardless of the cognitive question. Renal function should be checked before use in this population.

Low-dose lithium

Uncertain / conflicting

The mechanism paper is important science. The human evidence is genuinely conflicting, and the one adequately designed trial was negative.

The evidence

Aron and colleagues showed brain lithium is selectively reduced in MCI, that amyloid sequesters it, that dietary lithium depletion in mice accelerates amyloid deposition, tau phosphorylation, microglial activation and synapse loss via GSK3β, and that lithium orotate replacement prevents these changes 34. The prior clinical literature is thinner than the enthusiasm suggests. Forlenza’s 45-patient aMCI trial (lithium 0.25–0.5 mmol/L, 12 months) found reduced CSF phospho-tau and better ADAS-Cog and attention performance 93. A 13-year cross-sectional recall of that cohort reported substantially better MMSE and verbal fluency in the original lithium arm — from 36 of 61 participants, with 30.5% deceased and unequal group sizes, so differential survival and attrition are entirely unexcluded 95. The definitive result so far: a two-year randomised placebo-controlled trial of low-dose lithium carbonate in 80 adults with MCI, in which none of six co-primary outcomes met the prespecified threshold; verbal delayed recall was borderline (0.69 points/year, 95% CI 0.01–1.37, p=0.05) 35.

In practice

Two things are worth knowing that rarely make the summaries. First, the epidemiology does not point one way: the Danish drinking-water study is non-linear, with lower dementia above 15 µg/L but a significantly *higher* incidence rate ratio of 1.22 at 5.1–10.0 µg/L 96, and a Scottish cohort of 37,597 found no protective association and a trend to increased risk in women 97. A genuine dose-dependent neuroprotective effect should not look like that. Second, an FDG-PET substudy of patients four years into low-dose lithium found reduced glucose metabolism in both hippocampi and the cerebellum, not increased 94 — a finding of uncertain significance, but not the direction the neurotrophic hypothesis predicts. Patients with a longevity-medicine interest will have read the Nature paper and not any of this. The reasonable position: mechanistically serious, clinically not actionable, and the appropriate next step is a trial of lithium orotate specifically rather than off-label use of an unregulated supplement with a narrow therapeutic index and interactions with diuretics, ACE inhibitors, ARBs and NSAIDs.

Multivitamin–mineral supplementation

Promising, not yet proven

The only supplement with replicated randomised benefit on cognition. Effect small; dementia prevention not demonstrated.

The evidence

Three independent cognitive sub-studies within COSMOS, using telephone, web-based and in-person neuropsychological assessment respectively, each favoured multivitamin over placebo 202122. Pooled mean difference on global cognition 0.07 SD units (95% CI 0.03–0.11, p=0.0009) and episodic memory 0.06 SD (95% CI 0.03–0.10, p=0.0007) 22. Cocoa extract, tested in the same factorial design, showed no effect — a useful internal negative control 20.

In practice

Prespecified analysis of MCI and dementia incidence in COSMOS-Mind was null but severely underpowered 23. The population was predominantly white, educated and North American; generalisability to a frail Australian RACF cohort is unestablished. Nonetheless, cost, safety and replication make this the most defensible supplement recommendation available.

Fortasyn Connect (Souvenaid)

Uncertain / conflicting

Missed primary endpoint at 24 months; consistent secondary and 36-month findings. Industry-funded.

The evidence

LipiDiDiet (n=311, prodromal AD, IWG-1 criteria): NTB primary endpoint difference 0.098 (95% CI −0.041 to 0.237, p=0.166) at 24 months, with control-group decline far below the assumed −0.4 24. At 36 months, 60% reduction in decline on the NTB 5-item composite (between-group difference 0.212, 95% CI 0.044–0.380, p=0.014), 45% on CDR-SB, and reduced hippocampal atrophy, with Cohen’s d 0.25–0.31. Only 81 participants contributed 36-month efficacy data 25.

In practice

Not subsidised in Australia. Safe and well tolerated over eight years of follow-up 25. Reasonable to support a motivated family in early disease; not to recommend routinely, and no role in moderate or advanced dementia.

B vitamins in hyperhomocysteinaemia

Uncertain / conflicting

A biologically coherent subgroup hypothesis that has not been prospectively confirmed.

The evidence

VITACOG: whole-brain atrophy 0.76%/year on active treatment versus 1.08% on placebo (p=0.001), with 53% relative slowing in those with baseline homocysteine >13 µmol/L 9. Cognitive benefits concentrated above the median homocysteine of 11.3 µmol/L 10. Retrospective analysis found the atrophy benefit confined to the highest tertile of baseline plasma omega-3 (>590 µmol/L) 11. The Hong Kong replication at lower doses, in a population selected for homocysteine ≥10 µmol/L, was null at two years, with a significant negative interaction with aspirin use 13.

In practice

The trial that would settle this — combined high-dose B vitamins plus omega-3 in people with both elevated homocysteine and low omega-3 status — has been called for since 2016 and has not been adequately done 12. Meanwhile, measuring and correcting B12 and folate deficiency remains standard care and is a different proposition entirely.

Herpes zoster vaccination

Promising, not yet proven

The strongest quasi-causal evidence in dementia prevention outside a randomised trial — but for a vaccine no longer available in Australia.

The evidence

Regression discontinuity design in Wales exploiting a date-of-birth eligibility threshold: 3.5 percentage point absolute reduction in dementia diagnosis over seven years (95% CI 0.6–7.1, p=0.019) 44. Australian replication across 65 general practices (n=101,219) using the November 2016 Zostavax program threshold: 1.8 percentage point absolute reduction over 7.4 years (95% CI 0.4–3.3, p=0.01), with no effect on other preventive service uptake or unrelated chronic disease diagnoses 45. Both studies used the live attenuated vaccine. Propensity-matched observational data suggest the recombinant AS01-adjuvanted vaccine is associated with at least equivalent or lower dementia risk, and possibly an adjuvant-mediated effect 46.

In practice

Zostavax was deregistered in Australia in December 2024; Shingrix is the NIP-funded vaccine (65+, Aboriginal and Torres Strait Islander people 50+, eligible immunocompromised adults 18+) 47. The dementia evidence for the recombinant vaccine is observational, not quasi-experimental. Vaccinate for the registered indication. A randomised trial powered for dementia outcomes has been called for and has not been done.

Other adjuncts and repurposed agents

The agents below come up in consultations, usually because a patient or their adult child has read about one. Grouped here because they share a pattern: a plausible mechanism, a body of small positive trials, and no adequately powered independent replication. That pattern is not the same as evidence of benefit, and it is not the same as evidence of absence either.

Saffron (Crocus sativus)

Uncertain / conflicting

The best-performing botanical in the field on paper, and the one with the most serious replication problem.

The evidence

A 22-week multicentre randomised double-blind trial in 54 patients with mild-to-moderate Alzheimer’s disease found saffron 30 mg/day comparable to donepezil 10 mg/day on ADAS-Cog and CDR-SB, with significantly less vomiting 98. Subsequent trials have reported non-inferiority to donepezil and to memantine, over 16 to 52 weeks, and meta-analyses report superiority to placebo without serious adverse events 103.

In practice

The replication picture is the limitation. The Alzheimer’s saffron trials come overwhelmingly from one research group in Iran, the samples are small, and there is no large independent multicentre replication with standardised extract and biomarker confirmation. Active-comparator non-inferiority in a 54-patient trial is a weak design for establishing efficacy — it can equally reflect an underpowered study in which neither agent did much. Worth knowing about; not worth recommending in place of a registered medicine.

Huperzine A

Uncertain / conflicting

An unregulated cholinesterase inhibitor. The Cochrane conclusion is the one to quote.

The evidence

Cochrane pooled six trials in 454 patients: benefit on MMSE (WMD 2.81, 95% CI 1.87–3.76), ADAS-Cog at 6 and 12 weeks, CDR, CIBIC-plus and ADL, with adverse events no different from placebo. The authors nonetheless concluded that most included trials were of low methodological quality, only one was of adequate quality and size, and there is inadequate evidence to make any recommendation 99.

In practice

Effect sizes that large from a cholinesterase inhibitor should prompt scepticism rather than enthusiasm — they exceed what donepezil achieves in far larger trials. Pharmacologically this is a cholinesterase inhibitor sold as a supplement: same mechanism, same adverse-effect profile, none of the dose standardisation or oversight. If a patient wants cholinergic augmentation, prescribe the registered drug.

Citicoline (CDP-choline)

Uncertain / conflicting

Short-term signal, largely in vascular cognitive impairment, on an evidence base two decades old.

The evidence

Cochrane reviewed 14 randomised trials and found no effect on attention, but evidence of benefit on memory and behaviour and a stronger signal on global clinical impression. Seven of the 14 trials ran only 20–30 days and just one extended to 12 months; populations ranged from subjective memory complaint to vascular dementia 100.

In practice

Well tolerated, widely prescribed in parts of Europe, and never subjected to the long trial in a defined population that the Cochrane authors called for twenty years ago. If considering it at all, the population with any supporting rationale is vascular cognitive impairment, not Alzheimer’s disease.

Benfotiamine

Uncertain / conflicting

A serious trial is running. There is no efficacy result yet, and nothing to act on until there is.

The evidence

BenfoTeam is a seamless phase 2A–2B randomised double-blind placebo-controlled trial of benfotiamine, a thiamine prodrug, in 406 participants with early Alzheimer’s disease over 72 weeks, with CDR-SB and ADAS-Cog13 as co-primary endpoints 101. The rationale is impaired thiamine-dependent glucose metabolism in Alzheimer’s disease — the same bioenergetic thesis underlying the creatine work.

In practice

Currently a protocol, not a result. Benfotiamine is available as a supplement and is being promoted on the strength of the hypothesis. The correct answer to a patient asking is that a properly designed trial is under way and it would be reasonable to wait for it.

Valiltramiprosate (ALZ-801, oral homotaurine prodrug)

Not supported

Missed its primary endpoint. The positive result is a prespecified subgroup, which is a hypothesis rather than a finding.

The evidence

The phase III APOLLOE4 trial randomised 325 APOE ε4 homozygotes with early Alzheimer’s disease to oral valiltramiprosate or placebo for 78 weeks. The overall efficacy population showed no significant effect on ADAS-Cog13 (11% slowing, p=0.607), though hippocampal atrophy slowed significantly (18%, p=0.017). In the prespecified MCI subgroup (n=125) there were nominally significant effects on ADAS-Cog13 (52%, p=0.041) and hippocampal atrophy (26%, p=0.004) 102.

In practice

Notable for producing no ARIA, which is the main liability of the injectable anti-amyloid antibodies, and for being oral. But a negative primary endpoint with a nominally positive subgroup is precisely the pattern that has repeatedly failed to replicate in this field. Not available in Australia, and not something to raise with families as an option.

What this means for how you counsel patients and families

Three positions are defensible and worth stating out loud. First, no supplement currently available prevents or treats dementia, and any claim to the contrary is ahead of the evidence. Second, a standard multivitamin has replicated randomised evidence of a small cognitive benefit and is cheap and safe 202122, which is a reasonable thing to say to a family who want to do something. Third, the interventions with the best evidence in this space are not supplements at all — they are hearing correction, blood pressure control, physical activity, deprescribing, treating depression, and preventing delirium.

It is also worth being explicit about uncertainty rather than defaulting to dismissal. Creatine and the lithium repletion hypothesis are serious science that may yet change practice. Saying “we do not know yet, here is what would need to be shown, and here is why I would not start it today” maintains credibility in a way that blanket scepticism does not — particularly with patients who read primary literature.

Sources cited on this page

  1. 5 Baker LD, et al. Structured vs self-guided multidomain lifestyle interventions for global cognitive function: the US POINTER randomized clinical trial. JAMA. 2025. View source
  2. 8 Barnes LL, et al. Trial of the MIND diet for prevention of cognitive decline in older persons. N Engl J Med. 2023;389:602–11.
  3. 9 Smith AD, et al. Homocysteine-lowering by B vitamins slows the rate of accelerated brain atrophy in mild cognitive impairment: a randomized controlled trial (VITACOG). PLoS One. 2010;5:e12244.
  4. 10 de Jager CA, et al. Cognitive and clinical outcomes of homocysteine-lowering B-vitamin treatment in mild cognitive impairment: a randomized controlled trial. Int J Geriatr Psychiatry. 2012;27:592–600.
  5. 11 Jernerén F, et al. Brain atrophy in cognitively impaired elderly: the importance of long-chain ω-3 fatty acids and B vitamin status in a randomized controlled trial. Am J Clin Nutr. 2015;102:215–21.
  6. 12 Oulhaj A, et al. Omega-3 fatty acid status enhances the prevention of cognitive decline by B vitamins in mild cognitive impairment. J Alzheimers Dis. 2016;50:547–57.
  7. 13 Kwok T, et al. A randomized placebo-controlled trial of using B vitamins to prevent cognitive decline in older mild cognitive impairment patients. Clin Nutr. 2019.
  8. 16 Mulargia R, et al. Exploring the preventive effects of omega-3 polyunsaturated fatty acids supplementation on global cognition: a systematic review and meta-analysis of cognitively unimpaired older adults. Clin Transl Neurosci. 2025. (SMD −0.02, 95% CI −0.07 to 0.04.)
  9. 17 Wei BZ, et al. The relationship of omega-3 fatty acids with dementia and cognitive decline: evidence from prospective cohort studies of supplementation, dietary intake, and blood markers. Am J Clin Nutr. 2023.
  10. 20 Baker LD, et al. Effects of cocoa extract and a multivitamin on cognitive function: a randomized clinical trial (COSMOS-Mind). Alzheimers Dement. 2022.
  11. 21 Yeung LK, et al. Multivitamin supplementation improves memory in older adults: a randomized clinical trial (COSMOS-Web). Am J Clin Nutr. 2023.
  12. 22 Vyas CM, et al. Effect of multivitamin-mineral supplementation versus placebo on cognitive function: results from the clinic sub-cohort of the COSMOS randomized clinical trial and meta-analysis of three cognitive studies within COSMOS. Am J Clin Nutr. 2023.
  13. 23 Sachs BC, et al. Impact of multivitamin-mineral and cocoa extract on incidence of mild cognitive impairment and dementia: results from COSMOS-Mind. Alzheimers Dement. 2023.
  14. 24 Soininen H, et al. 24-month intervention with a specific multinutrient in people with prodromal Alzheimer’s disease (LipiDiDiet): a randomised, double-blind, controlled trial. Lancet Neurol. 2017;16:965–75.
  15. 25 Soininen H, et al. 36-month LipiDiDiet multinutrient clinical trial in prodromal Alzheimer’s disease. Alzheimers Dement. 2020.
  16. 32 Smith AN, et al. Creatine monohydrate pilot in Alzheimer’s: feasibility, brain creatine, and cognition. Alzheimers Dement (N Y). 2025. (Single-arm, open-label, n=20, 8 weeks — no control group.)
  17. 33 Prokopidis K, et al. Effects of creatine supplementation on memory in healthy individuals: a systematic review and meta-analysis of randomized controlled trials. Nutr Rev. 2022.
  18. 34 Aron L, et al. Lithium deficiency and the onset of Alzheimer’s disease. Nature. 2025.
  19. 35 Gildengers AG, et al. Low-dose lithium for mild cognitive impairment: a randomized clinical trial. JAMA Neurol. 2026. (None of six co-primary outcomes met the prespecified significance threshold.)
  20. 44 Eyting M, et al. A natural experiment on the effect of herpes zoster vaccination on dementia. Nature. 2025. (Wales; 20% relative reduction in new dementia diagnoses over 7 years.)
  21. 45 Pomirchy M, et al. Herpes zoster vaccination and dementia occurrence. JAMA. 2025. (Australian quasi-experiment; 1.8 percentage point absolute reduction over 7.4 years.)
  22. 46 Taquet M, et al. The recombinant shingles vaccine is associated with lower risk of dementia. Nat Med. 2024.
  23. 47 National Centre for Immunisation Research and Surveillance (NCIRS). Zoster (shingles) vaccine: frequently asked questions. View source
  24. 57 RACGP newsGP. Lecanemab gains TGA approval for Alzheimer’s treatment (registered 25 September 2025 for mild cognitive impairment and mild dementia due to Alzheimer’s disease). View source
  25. 58 ABC News. Donanemab approved in Australia for treatment of Alzheimer’s disease (TGA registration, May 2025). View source
  26. 59 Forward with Dementia (Australia). Update on anti-amyloid therapies — not PBS-listed; out-of-pocket cost estimated at $40,000–50,000 per year for the drug, and $80,000–100,000 per year including monitoring; MRI surveillance 3–4 times in the first six months. View source
  27. 60 Alzheimer Europe. Results of EVOKE and EVOKE+ trials show no effect of oral semaglutide on Alzheimer’s disease progression (n=3,808; reported at CTAD, December 2025). View source
  28. 61 Gao L, et al. Cost-effectiveness of donanemab for early Alzheimer disease in Australia. Med J Aust. 2026. View source
  29. 73 Australian Dementia Network (ADNeT). Blood-based biomarkers for Alzheimer’s disease in primary care. View source
  30. 75 Marshall S, et al. Creatine and cognition in aging: a systematic review of evidence in older adults. Nutr Rev. 2025.
  31. 79 van Dyck CH, et al. Lecanemab in early Alzheimer’s disease (CLARITY-AD). N Engl J Med. 2023;388:9–21. (n=1,795; 18-month CDR-SB difference −0.45, 95% CI −0.67 to −0.23; ARIA with oedema or effusion 12.6%; infusion reactions 26.4%.)
  32. 80 Sims JR, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA. 2023. (n=1,736; 76-week CDR-SB difference −0.67 in the low/medium tau population; ARIA with oedema or effusion 24.0%; three treatment-related deaths.)
  33. 86 Dhana K, et al. Adherence to the MIND diet and longitudinal changes in global cognition: a post-hoc analysis of the MIND trial. Am J Clin Nutr. 2026.
  34. 87 Dhana K, et al. Dietary intervention and cognition across Alzheimer’s disease biomarker levels: the MIND clinical trial. J Alzheimers Dis. 2026.
  35. 93 Forlenza OV, et al. Disease-modifying properties of long-term lithium treatment for amnestic mild cognitive impairment: randomised controlled trial. Br J Psychiatry. 2011;198:351–6. (n=45, 12 months, lithium 0.25–0.5 mmol/L; reduced CSF phosphorylated tau and better ADAS-Cog and attention scores.) View source
  36. 94 Forlenza OV, et al. Long-term lithium treatment reduces glucose metabolism in the cerebellum and hippocampus of nondemented older adults: an [18F]FDG-PET study. ACS Chem Neurosci. 2014;5:484–9. (n=19; four years of low-dose lithium was associated with reduced, not increased, hippocampal and cerebellar glucose uptake.) View source
  37. 95 Damiano RF, et al. Revisiting global cognitive and functional state 13 years after a clinical trial of lithium for mild cognitive impairment. Braz J Psychiatry. 2023;45:46–9. (Cross-sectional recall of the 2011 trial cohort; only 36 of 61 traced and 30.5% had died.) View source
  38. 96 Kessing LV, et al. Association of lithium in drinking water with the incidence of dementia. JAMA Psychiatry. 2017;74:1005–10. (73,731 cases, 733,653 controls; non-linear — IRR 0.83 above 15 µg/L but 1.22 at 5.1–10.0 µg/L.) View source
  39. 97 Duthie AC, et al. Low-level lithium in drinking water and subsequent risk of dementia: cohort study. Int J Geriatr Psychiatry. 2023;38:e5890. (37,597 Scottish participants; no protective association, with a trend to increased risk in women.) View source
  40. 98 Akhondzadeh S, et al. A 22-week, multicentre, randomised, double-blind controlled trial of Crocus sativus in the treatment of mild-to-moderate Alzheimer’s disease. Psychopharmacology (Berl). 2010;207:637–43. (n=54; saffron 30 mg/day comparable to donepezil 10 mg/day, with less vomiting.) View source
  41. 99 Li J, et al. Huperzine A for Alzheimer’s disease. Cochrane Database Syst Rev. 2008;(2):CD005592. (6 trials, 454 patients; apparent benefits across several scales, but only one trial of adequate quality and size — “inadequate evidence to make any recommendation”.) View source
  42. 100 Fioravanti M, Yanagi M. Cytidinediphosphocholine (CDP-choline) for cognitive and behavioural disturbances associated with chronic cerebral disorders in the elderly. Cochrane Database Syst Rev. 2005;(2):CD000269. (14 heterogeneous trials, most 20–30 days, largely vascular cognitive impairment.) View source
  43. 101 Feldman HH, et al. Protocol for a seamless phase 2A–2B randomised double-blind placebo-controlled trial of benfotiamine in early Alzheimer’s disease (BenfoTeam). PLoS One. 2024;19:e0302998. (406 participants, 72 weeks; protocol only — no efficacy results yet.) View source
  44. 102 Abushakra S, et al. Clinical efficacy, safety and imaging effects of oral valiltramiprosate in APOE ε4/ε4 homozygotes with early Alzheimer’s disease: the phase III APOLLOE4 trial. Drugs. 2025. (n=325; primary ADAS-Cog13 endpoint not met overall, 11% slowing, p=0.607; nominally significant in the prespecified MCI subgroup.)
  45. 103 Kehtari T, et al. From mood to memory: unlocking saffron’s potential in brain health. Cureus. 2025;17:e82924. (Narrative review of the saffron trial literature.) View source

See every source cited across the dementia section →

General information only, reflecting the interpretation of Umbrella Aged Care’s GPs of the published evidence as at September 2026. It is not individual medical advice, does not create a doctor–patient relationship, and must not be used to start, stop or change any treatment. Evidence and Australian regulatory and PBS arrangements change — always confirm current advice with the treating GP, pharmacist or specialist.

Enquire about care