Smartwatch Prompts Show Small, Uncertain Cognitive Benefit in At-Risk Older Adults

A six-month randomised trial in suburban Japan has found that a fully automated lifestyle programme, delivered almost entirely through a wrist-worn fitness tracker and smartphone alerts, produced a small improvement in global cognition among older adults at elevated risk of decline. The INSPIOR trial, published in eClinicalMedicine, is being hailed as a proof of concept for scalable dementia prevention, but its investigators are unusually candid that the cognitive signal is fragile, statistically unconfirmed, and of uncertain clinical meaning.

The trial enrolled 355 community-dwelling adults aged 65 and older in Kashiwa, Japan, all of whom showed at least one marker of physical vulnerability: low body-mass index for their age, pre-frailty or frailty on a standard questionnaire, or sarcopenia-related decline such as weak grip strength, slow gait speed, or reduced skeletal muscle mass. Roughly half of the participants scored below the cut-off for suspected mild cognitive impairment on the Montreal Cognitive Assessment, or MoCA, at baseline. Participants were randomly assigned in equal numbers to an automated intervention or to an assessment-only control group that received no advice, education, or prompting of any kind.

The intervention itself was strikingly lean on human input. Participants in the active arm wore a Fitbit Charge 5 continuously for six months, including during sleep. After an initial 28-day baseline window with no prompts, the system began issuing text-message alerts generated entirely from each participant’s own averaged lifelog data, with no counsellor, coach, or clinician shaping the content. Human contact was limited to a group orientation, device setup, and technical support by telephone. The prompts were purely threshold-based goal setting: daily step targets climbed in steps from 3,000 to 8,000; weekly exercise targets rose from 45 to 150 minutes, matching the World Health Organization recommendation for moderate-intensity activity; sleep targets aimed for a six-to-nine-hour nightly band; and an optional nutrition component, tracked through a food-logging application, encouraged participants to raise the share of energy they obtained from protein to 15 to 20 percent.

What makes INSPIOR distinctive is that no previous randomised trial of a multidomain lifestyle intervention has generated and delivered its behavioural content automatically from objective daily data while also measuring cognition as a primary outcome. Earlier landmark efforts, including the Finnish FINGER trial and more recent programmes in the United States, Australia, Japan, and elsewhere, depended on intensive face-to-face coaching or human-led sessions, which makes broad implementation expensive and difficult. A Cochrane review concluded that multidomain interventions confer small and inconsistent cognitive benefits, and the evidence base for the nutritional component of such programmes has been particularly weak, with a randomised trial of the MIND diet showing no advantage for global cognition over an active control diet.

On the trial’s three co-primary outcomes, only one showed a between-group difference. Global cognition, measured by the Japanese version of the MoCA, favoured the intervention by an adjusted 0.58 points on the 30-point scale, a difference that reached nominal statistical significance in the intention-to-treat analysis. However, because the trial registered three co-primary outcomes without a pre-specified hierarchy or multiplicity strategy, the authors also analysed all three jointly with Holm adjustment, and under that stricter accounting the MoCA result was no longer statistically significant. Executive function, measured by the Trail Making Test, showed no between-group difference at all, and skeletal muscle mass was essentially unchanged, with an adjusted difference of just 0.01 kilograms per square metre.

The behavioural data told a clearer story. In per-protocol analyses of the continuously recorded device logs, intervention participants moved an average of 696 more steps per day, accumulated 6.6 more minutes of active time per day, and slept 11.6 minutes longer per night than controls, relative to their own baselines. Notably, for steps and sleep both groups declined over the six months, and the difference reflects a smaller decline in the intervention group rather than an absolute improvement. Protein intake rose by nearly 8 grams per day in participants who opted into the dietary component, and a modest difference in the frailty index also favoured the intervention. Yet the authors caution that only the intervention group wore the device continuously, so some of these behavioural differences may reflect reactivity to being measured rather than a true effect of the prompts.

In an exploratory model entering changes in exercise, sleep, and protein intake simultaneously, only the change in exercise was associated with the change in MoCA score, with a standardised coefficient of 0.24. The investigators are careful to insist that this association cannot be read as proof of a mechanism. The model is observational, the behavioural measures were recorded under different conditions in the two groups, and the activity, sleep, and nutrition components were delivered together, making it impossible to attribute any cognitive change to a single ingredient. A post-hoc responder analysis found that 60 percent of intervention participants improved by at least one MoCA point, compared with 41 percent of controls, but the thresholds used were never pre-specified or established as clinically important.

The trial also carries methodological caveats that the authors lay out with unusual transparency. It was open-label, and the staff who assessed outcomes at six months had also set up devices and managed the study, so they could often recognise participants and infer their group. Allocation concealment was not implemented as a separate procedure, and the authors note that a 0.58-point difference is small enough that even modest bias could matter. A procedural problem with the baseline Trail Making Test, which was administered under a non-standard procedure and re-administered after allocation, means the executive-function estimates must be treated as non-confirmatory. The absence of corroboration on a second cognitive instrument, they write, weakens the case that the MoCA difference reflects a broad cognitive benefit.

What, then, does a 0.58-point shift on the MoCA actually mean? No universally accepted minimal clinically important difference exists for the instrument, but a distribution-based estimate puts it at 1.0 point, a figure compatible with the upper bound of the trial’s confidence interval but not with the point estimate itself. The authors therefore read the finding not as a clinically meaningful improvement for any individual but as a small shift in the population distribution of cognitive scores, of uncertain significance. They note that even a one-point difference on a 30-point screening instrument has been associated with a higher rate of recorded dementia diagnosis in observational cohorts, which is precisely why larger trials are warranted rather than why the result should be celebrated.

The broader implication is about delivery rather than effect size. INSPIOR demonstrates that a multidomain prevention programme can run at community scale with human contact limited to setup and troubleshooting, reaching a risk-enriched population that previous coaching-heavy trials struggled to serve efficiently. No serious adverse events related to the device, the dietary application, or the protein-balanced food supplied to some participants were reported, and mild skin irritation and the burden of daily dietary recording were the only intervention-related complaints. The authors conclude that wearable-centric, automated programmes may offer a scalable route to supporting healthy ageing, but that this trial does not establish that they prevent cognitive decline. Larger, longer, multicentre trials with masked outcome assessment are now needed to determine whether the small cognitive signal observed in Kashiwa is durable, clinically meaningful, and attributable to any particular component of the automated package.

Subject of Research: A wearable-based automated multidomain lifestyle intervention for cognitive and physical function in older adults at risk of decline

Article Title: Effect of a wearable-based individualised multidomain lifestyle intervention on cognitive and physical function in older adults at risk of decline (INSPIOR): a single-centre, open-label, randomised controlled trial

Article References: Sakurai, K., Tanaka, R., Ledsam, J., Shiraishi, I., Kasahara, H., Anzai, S., Watanabe, M., Funakawa, M., Shimada, S., Goto, T., Kawabata, R., Murasakino, K., Kawaura, T., Inamura, N., Kaneta, K., Kobayashi, S., Nomura, T., Karasawa, N., Matsumoto, T., … Hisatsune, T. (2026). Effect of a wearable-based individualised multidomain lifestyle intervention on cognitive and physical function in older adults at risk of decline (INSPIOR): a single-centre, open-label, randomised controlled trial. eClinicalMedicine, 100, Article 104212. https://doi.org/10.1016/j.eclinm.2026.104212

Image Credits: AI Generated

DOI: 10.1016/j.eclinm.2026.104212

Keywords: wearables, cognitive decline, dementia prevention, multidomain intervention, randomised controlled trial, older adults, physical activity, sleep, nutrition, MoCA, digital health, frailty