Large Trial Finds Test Scores Rose in All Groups of Older Adults With Low Cognitive Scores, With Slightly More Gain From a Website and Brief Support

That is the finding of the Active Brains trial, a randomised controlled trial published in The Lancet Primary Care in 2026. It enrolled 11,563 adults aged 60 to 85 from 670 GP practices in England, Wales and Scotland. All had low scores on a computerised verbal reasoning test, the Baddeley test. All needed internet access, and people with a dementia diagnosis were excluded, which limits how far the results can be generalised. The trial was open-label, so participants knew which group they were in. The statisticians and the research team making follow-up calls did not.

Participants were assigned by chance to one of three groups. The control group (3,927 people) got a one-page, evidence-based advice sheet on activities that protect cognitive health, plus usual care. The Active Brains group (3,169) got a website with support for physical activity, a Mediterranean diet and brain-training games. The third group (3,087) got the website plus three brief support sessions of about 10 minutes, by email or phone.

The main result was the verbal reasoning score at 12 months. Higher is better. The control group averaged 18.3. The website group averaged 18.7, an adjusted difference of +0.50 points (95% confidence interval 0.13 to 0.87). The website-plus-support group averaged 19.2, a difference of +0.97 (0.59 to 1.34). These are small gaps on the test scale. The paper does not present them as dramatic.

The more striking result came from the control group itself. The study looked at how many people no longer met the test-score criteria for age-associated cognitive decline, a measure of lower-than-expected thinking ability for age. At 12 months, 883 of 3,113 control participants with data (28.4%) no longer met it. The figures were 842 of 2,566 (32.8%) in the website group and 811 of 2,546 (31.6%) in the website-plussupport group.

Milder and more severe levels of impairment were also measured. Here, "mild" and "severe cognitive impairment" mean score cutpoints on one reasoning test, not clinical diagnoses. According to the authors, fewer people in the website-plussupport group, but not the website-only group, met the test criteria for mild or severe cognitive impairment. Across the trial population, the paper reports an 8% relative reduction in meeting the mild-impairment test criteria with online help plus support (risk ratio 0.92; 95% CI 0.87–0.97). For those who started with the mildimpairment test criteria, there was an 18% reduction in scores consistent with severe impairment (risk ratio 0.82; 0.70–0.96).

Several cautions matter. Everyone received some cognitive-health advice and cognitive testing, so there was no group left entirely untouched. The control group's improvement therefore cannot be pinned on the advice sheet alone. Test scores can also rise on retesting, and people can drift above or below a cut-off by chance. Data at 12 months were available for 79.3% of the control group but only about 67% of the other two groups. The authors used statistical methods to fill the gaps, but the difference in dropout is worth noting. Nearly all participants were White (95.8%). This is where the argument is weakest: without an untreated comparison group, we cannot say how much of any change came from the advice itself.

The trial measured a test score, not dementia. The authors say follow-up for dementia at five years is ongoing, and that research is needed on longer-term effects. Serious adverse events and deaths were similar across groups and judged unrelated to treatment.

The University of Southampton's press release went further than the paper. Its headline said puzzles can "reverse" cognitive impairment, and it said scientists had "proven" this. The trial measured test scores at 12 months and did not diagnose or track impairment over time, so that wording outruns the data. The release also described the 8% and 18% reductions as applying to those who had website access or extra support, which is looser than the paper's wording. Lead researcher Paul Little was quoted in the release saying, "Simple messages and simple things that people can do really make a difference," and that this need not cost the NHS much.

The paper tests brief primary-care interventions and offers no personal advice to readers, so what follows is our reading, not a finding. The study's advice centred on low-cost habits: staying physically active, eating a Mediterranean-style diet, and doing mentally engaging things. The Southampton release lists puzzles, online brain-training games, playing an instrument, singing in a choir, learning a language and social activities. The trial cannot say which of these matters most, or that any one of them caused the score changes.

What to watch next

What to watch next is the five-year follow-up, which will show whether these early gains last and whether they translate into fewer cases of dementia.

1. Age-Related Decline in a Cellular Recycling Pathway Lets 'Zombie' Cells Pile Up, Mouse Study Finds 2. Large Review Links Insomnia to 26% Higher Stroke Risk and More Hospital Stays 3. Large Trial Finds Test Scores Rose in All Groups of Older Adults With Low Cognitive Scores, With Slightly More Gain From a Website and Brief Support 4. Testosterone Therapy Linked to Higher Cardiovascular Risk in Postmenopausal Women 5. Report: White House Stalls Kennedy's Push for a Federal Definition of Ultra-Processed Food