Social isolation is one of the fourteen modifiable risk factors on the Lancet Commission's list. We have cited that list repeatedly — in what the 45% figure actually means, in the factor-by-factor breakdown, and in the piece on hearing loss — without ever writing the post on isolation itself. This is that post.
It is a harder subject than hearing: the exposure is harder to define, the studies disagree with each other, and the biggest objection to the whole literature is one the literature cannot answer.
Two different things, routinely treated as one
Social isolation is objective: how many people you see, how often, whether you live alone, whether you belong to anything. It can be counted from the outside.
Loneliness is subjective: the gap between the contact you have and the contact you want. It is a feeling, and it is reported rather than observed.
They overlap only loosely. People with full diaries are lonely. People who live alone and see almost no one are often not. These are two separate exposures, measured with different instruments, and — this is the part that matters — the evidence for each points in a different direction depending on which study you read. An article that runs them together has already lost the thread.
What the cohort data shows
The observational literature is large and, at first glance, consistent.
A 2015 systematic review and meta-analysis in Ageing Research Reviews pooled 19 longitudinal cohort studies and found that low social participation (RR 1.41, 95% CI 1.13–1.75), less frequent social contact (RR 1.57, 95% CI 1.32–1.85) and greater loneliness (RR 1.58, 95% CI 1.19–2.09) were each associated with incident dementia. The authors judged the strength of these associations comparable to low educational attainment, physical inactivity and late-life depression.
Two findings in the same paper get dropped from most retellings, and both belong here. Results for social network size were inconsistent across studies. And low satisfaction with one's social network was not significantly associated with dementia onset (RR 1.25, 95% CI 0.96–1.62) — a confidence interval crossing 1.00, which is a null result, not a weak positive.
A larger 2018 meta-analysis in the Journal of Alzheimer's Disease covered 31 cohort and 2 case-control studies comprising 2,370,452 participants. Poor social network was associated with dementia risk (RR 1.59, 95% CI 1.31–1.96), as was poor social support (RR 1.28, 95% CI 1.01–1.62), and in studies with at least ten years of follow-up good social engagement was modestly protective (RR 0.88, 95% CI 0.80–0.96). Loneliness, in that analysis, was not significantly associated with dementia risk: RR 1.38, 95% CI 0.98–1.94. The authors reported it as non-significant. It is regularly quoted as though it were not.
Where the studies stop agreeing
Then the two largest individual cohorts contradict each other, and the contradiction is instructive.
A 2022 UK Biobank analysis in Neurology followed 462,619 participants (mean baseline age 57.0) for a mean of 11.7 years, during which 4,998 developed dementia. After full adjustment — demographic, socioeconomic, biological including APOE genotype, cognitive, behavioural and psychological factors — social isolation was associated with a 26% higher risk of dementia (HR 1.26, 95% CI 1.15–1.37), while loneliness was not (HR 1.04, 95% CI 0.94–1.16). The authors calculated that 75% of the apparent loneliness association was attributable to depressive symptoms. On this reading, objective isolation is the exposure and loneliness is mostly depression wearing a different name.
A Framingham analysis published in the same journal the same year reached almost the opposite conclusion. Among 2,308 dementia-free participants (mean age 73), 144 were lonely at baseline and 329 developed dementia over ten years. Loneliness carried a higher risk (HR 1.54, 95% CI 1.06–2.24), and in participants under 80 without an APOE ε4 allele the risk was roughly tripled (HR 3.03, 95% CI 1.63–5.62). Neurology graded this Class I evidence — the highest grade available to a prognostic observational study, which is not the same as a trial.
An important qualification on that comparison: the Framingham hazard ratio was adjusted for age, sex and education only. The UK Biobank figure was adjusted for depression, among much else. When one analysis controls for depression and the other does not, they are not answering quite the same question — and depression is the variable doing most of the work.
Earlier work adds a third answer. A 2014 analysis of the Amsterdam Study of the Elderly followed 2,173 people for three years and found the reverse of UK Biobank: feelings of loneliness predicted dementia (OR 1.64, 95% CI 1.05–2.56) while social isolation did not.
Three large studies, three verdicts on which exposure is the real one. That is where the evidence sits.
The problem none of this can solve
Here is the crux, and it is the reason to be careful with all of the numbers above.
Withdrawal is an early symptom of dementia. Years before a diagnosis, people stop going to things. Conversation becomes effortful. Names and threads get lost, and the social world quietly contracts to avoid the embarrassment. Apathy and loss of initiative are core features of the prodromal phase, and they are features of the disease, not lifestyle choices.
So when a cohort study finds that isolated or lonely people go on to develop dementia more often, there are two readings, and they are almost indistinguishable in observational data. Either the isolation contributed to the decline, or the decline produced the isolation and the study caught it on the way down. The authors of the Amsterdam study said so themselves: feelings of loneliness "may signal a prodromal stage of dementia."
Researchers push back with long follow-up, adjustment for baseline cognition, and excluding cases diagnosed in the first few years. Those help. They do not settle it, because a dementia diagnosable at 75 has been developing since the person's fifties, and no realistic follow-up window sits entirely outside the disease process.
The honest position is that the observational evidence cannot separate cause from early sign, and no trial has been run that would. This is the same structural problem we described with hearing loss, where a large randomised trial did eventually test the question and returned a null primary result. For isolation there is no equivalent trial, and there may never be one — you cannot randomise people to a decade of friendship.
What interventions have and have not shown
This is where the field is weakest, and where the claims get loosest.
A meta-analysis of loneliness-reduction interventions found something that should be read as a warning about every optimistic headline in this area: single-group before-and-after studies and non-randomised comparisons produced larger effect sizes than randomised comparisons. That is the signature of a result generated by the design rather than the intervention. Among the randomised studies, the most effective approach was not arranging more contact — it was addressing maladaptive social cognition, the pattern of expecting rejection and reading neutral social signals as hostile. Cognitive approaches beating befriending schemes is a consistent and counter-intuitive finding, and it sits alongside what we have written about the evidence base for psychological therapy in anxiety.
On the specific question this post is about — whether reducing isolation or loneliness protects cognition — a 2024 systematic review found nine studies with 1,025 participants in total, with a high risk of bias in most of the randomised trials. Six reported improved cognitive function, five of those involving technology. The reviewers also noted that the interventions did not reliably change social isolation itself, which makes the cognitive findings hard to interpret. Nine small, mostly biased studies is not an evidence base. It is a starting point.
The scale of the problem is not in doubt. The WHO Commission on Social Connection reported in 2025 that around one in six people worldwide are affected by loneliness and social disconnection, and linked it to roughly 871,000 deaths a year — a modelled attribution rather than a counted total. Knowing a problem is large is not the same as knowing what fixes it.
What follows from this
Not much that is specific, which is itself the finding.
Social contact belongs on the same list as blood pressure, hearing, alcohol and movement — worth attending to, with the caveat that the causal evidence is weaker than for the vascular factors. Sustained conversation is good for reasons that do not need a dementia argument, and the dementia argument is the least secure reason to make it.
Two things are worth acting on more firmly. If someone has withdrawn noticeably over a year or two, treat that as a symptom worth asking about rather than a preference — it may be depression, it may be hearing, it may be early cognitive change, and all three are things a doctor can work with. And if the underlying worry is memory rather than sociability, the more useful article is this one. Withdrawal following a bereavement is its own situation, and grief has a research literature of its own.
Nothing Proco sells addresses this. There is no supplement for loneliness, none on the Commission's list of fourteen, and no mechanism by which a capsule substitutes for people. We have written about this because it matters, not because it leads anywhere.
This article is not medical advice and cannot assess anyone's individual risk.
Sources: Kuiper JS, et al. Social relationships and risk of dementia: meta-analysis. Ageing Res Rev. 2015 · Penninkilampi R, et al. Social engagement, loneliness and risk of dementia: meta-analysis. J Alzheimers Dis. 2018 · Shen C, et al. Associations of social isolation and loneliness with later dementia. Neurology. 2022 · Salinas J, et al. Loneliness and 10-year dementia risk. Neurology. 2022 · Holwerda TJ, et al. Feelings of loneliness, but not social isolation, predict dementia onset. J Neurol Neurosurg Psychiatry. 2014 · Masi CM, et al. A meta-analysis of interventions to reduce loneliness. Pers Soc Psychol Rev. 2011 · Baptista C, et al. Cognitive effects of interventions to reduce social isolation and loneliness. Aging Ment Health. 2024 · Livingston G, et al. Dementia prevention, intervention and care: 2024 Lancet Commission report. Lancet. 2024 · WHO Commission on Social Connection. From loneliness to social connection. 2025