Picture a laptop propped on a kitchen table on a weekday afternoon, somewhere in England, in the second winter of the pandemic. On the screen are six or seven other people in their seventies, all of whom have noticed something slipping: a name that will not come, a reason for standing in the hall. Leading the hour is not a neurologist or a memory clinic consultant but a trained non-clinical facilitator, supervised at a distance by a clinical psychologist. Between the formal sessions there are informal video calls the researchers actually called tea breaks, and individual phone calls to set small goals. This was the APPLE-Tree programme, and between October 5, 2020 and December 31, 2022 it enrolled 746 older adults across 11 sites in England, mean age 74.4 years, 350 of them women and 392 men, all living with subjective cognitive decline or mild cognitive impairment.
The design, published in The Lancet Healthy Longevity in October 2025, was a single-masked, multicentre randomised controlled trial, registered in advance as ISRCTN17325135. Half were assigned to twelve months of the programme: ten one-hour group video sessions over six months, then monthly tea breaks. Half got usual care plus brief written information about dementia prevention. The primary outcome was cognition on a Neuropsychological Test Battery at 24 months, and 635 of the 746 participants (85 percent) were included in that analysis. Scores rose in both groups, a little more in the programme group: mean 0.33 at 24 months against 0.21, an adjusted mean difference of 0.06, with a 95 percent confidence interval of minus 0.001 to 0.128 and a p value of 0.055. The trial had been powered to find a difference of 0.15. What it found was smaller, and by the conventional line, not statistically significant. The confidence interval brushes zero.
So what did hold up? Attendance, strikingly. Of 374 people offered the programme, 305 (82 percent) came to at least five of the main sessions, the minimum dose the team had set in advance. The cognitive training app, by contrast, was used by only 49 (13 percent), a reminder that people take up the parts of a programme that feel like company and leave the parts that feel like homework. Across the secondary outcomes there was little to report: no significant between-group differences in day-to-day function, sleep, anxiety and depression, or informant-rated general health. Safety was unremarkable, which matters for anything meant to be rolled out widely: serious adverse events occurred in 35 intervention participants (9 percent) and 30 control participants (8 percent), and none were judged related to the intervention.
The reason this trial is worth a family's attention is not the number, which is small and uncertain. It is the question the trial chose to answer. The past decade of dementia prevention research has been mostly about content: which risk factors, which diet, how much exercise. APPLE-Tree asked the unglamorous next question, which is who delivers it, to whom, and at what cost. Remote group calls, run by people who are not clinicians, do not need a memory clinic's diary or a specialist's hour. The trial authors' own conclusion is deliberately unshowy: they describe an accessible intervention linked to small cognitive improvements that were not statistically significant, and note that "further work is needed to optimise the intervention for delivery in routine settings".
The money followed in June 2026, when the team published an economic evaluation embedded in the same trial in Age and Ageing. Over two years, mean adjusted costs were 2966 pounds per person in the programme group against 2551 pounds in the control group, and quality-adjusted life years were almost identical, 1.511 against 1.520, an adjusted difference of minus 0.010. Judged on quality-adjusted life years at a 30,000 pound threshold, the probability of cost-effectiveness was 12 percent; judged on cognition, where the cost worked out at 6809 pounds per one-point change in test score, it was 96 percent. Post hoc analyses, which are exploratory and were not the question the trial was built to answer, suggested more favourable results among participants from non-White ethnic groups and among those who did not own their homes. The authors are collecting longer-term data in a trial extension.
For someone sitting with a recent note in their file that says mild cognitive impairment, none of this is an instruction. Nobody in APPLE-Tree was told to stop a medicine or start one; the thing being tested was a way of organising support, not a drug. What the trial shows is that several hundred seniors with memory worries will reliably turn up to a group on a screen, that the programme did them no measurable harm, and that two years of it left cognition a fraction better than written leaflets, by an amount the statistics cannot separate from chance. Whether a service like this appears near you is a commissioning decision, and whether it would suit you is a conversation for your own clinician.