Risk reduction and brain health

The evidence

Network meta-analysis of 71 trials in 5,606 people with MCI or dementia found benefit across exercise types, with resistance training ranking highest for global cognition in dementia and multicomponent exercise for MCI ⁴³. The EXERT trial (n=296, 18 months) found no difference between moderate-high intensity aerobic training and low-intensity stretching and balance work — but neither group declined cognitively over 12 months, and hippocampal volume loss was low in both ⁴¹.

In practice

EXERT is often reported as a negative trial. The more interesting reading is that both supervised exercise groups remained stable when decline was expected, which is a signal that regular structured movement of almost any kind may matter more than intensity.

The evidence

ACHIEVE randomised 977 adults aged 70–84 with untreated hearing loss to hearing intervention or health education. There was no difference in three-year cognitive change in the total cohort ³⁸. In the prespecified higher-risk ARIC sub-cohort, cognitive decline was 48% slower with hearing intervention ³⁸; in a secondary analysis, decline was around 60% slower in the top quartile of predicted risk ³⁹. Hearing intervention also reduced falls by 27% over three years and improved self-reported communicative function ⁴⁰.

In practice

Treat hearing loss. Even setting cognition aside, the falls and communication benefits are real, and untreated hearing loss makes every other aspect of dementia care harder.

The evidence

Hypertension is one of the largest contributors in the Lancet Commission’s population-attributable model ⁴. Trial evidence for cognitive endpoints is more mixed than for stroke, but the direction is consistent and the cardiovascular case is independent and strong.

In practice

Targets must be individualised in frail older adults. In advanced dementia and frailty, postural hypotension and falls are the immediate risk, and aggressive targets stop making sense.

The evidence

A natural experiment in Wales exploited a date-of-birth eligibility cut-off for the live zoster vaccine and found a 3.5 percentage point absolute reduction in new dementia diagnoses over seven years (20% relative reduction) ⁴⁴. The design is quasi-randomised and far less prone to healthy-vaccinee bias than ordinary cohort studies. An Australian replication using the same design across 65 general practices (n=101,219) found a 1.8 percentage point absolute reduction over 7.4 years ⁴⁵. Separate observational work suggests the recombinant vaccine is associated with at least as low a risk ⁴⁶.

In practice

Australia’s National Immunisation Program now funds Shingrix (recombinant) for adults 65 and over, Aboriginal and Torres Strait Islander people 50 and over, and eligible immunocompromised adults 18 and over. Zostavax, the live vaccine used in the Welsh and Australian natural experiments, was deregistered in Australia in December 2024 ⁴⁷. So the vaccine with the strongest quasi-causal dementia evidence is no longer the one available here — a genuine evidentiary gap. Vaccinate for shingles prevention, which is the established indication; treat any dementia benefit as a plausible bonus, not the reason.

The evidence

Social isolation and low education are both Lancet Commission risk factors ⁴. Computerised cognitive training was a component of the positive FINGER and US POINTER programmes ^5,6, and adding cognitive training to exercise improved ADAS-Cog scores in the Australian-relevant SYNERGIC trial ⁴². Isolating cognitive training as a standalone intervention produces much weaker results, and gains tend not to transfer beyond the trained task.

In practice

Brain-training apps sold on a dementia-prevention claim are not supported by this evidence. Genuine social participation, meaningful activity and learning are worth pursuing on their own merits.

The evidence

Depression carries a population-attributable fraction in the Commission’s model ⁴. Whether treating it reduces subsequent dementia incidence has not been established by trial. Depression is also a common prodrome of dementia, which complicates the causal picture.

In practice

Half of Australian residents with dementia have a recorded depression or mood disorder ¹. It is under-treated, and treating it improves quality of life whatever it does to dementia trajectory.

The evidence

This is the one row on this page not anchored to a trial, because there isn’t one. The reasoning is mechanistic and observational: sleep is when the brain clears metabolic waste, and untreated obstructive sleep apnoea travels with cognitive impairment. Neither observation establishes that treating sleep prevents dementia, and no randomised trial has tested it. We have included the row because the question is asked often, and the honest answer is “we don’t know”.

In practice

Diagnose and treat OSA for the daytime function, cardiovascular and safety benefits. Do not promise a dementia benefit.

The evidence

Excessive alcohol and smoking each carry population-attributable fractions in the Commission model ⁴. Alcohol-related brain injury is also a directly reversible cause of cognitive impairment in its own right.