How Big Is the Effect, Not Just Whether It Exists

Every major report on loneliness now leads with a comparison to smoking. The Surgeon General’s 2023 advisory put the mortality risk of social disconnection on a par with smoking up to 15 cigarettes a day. That comparison has done a great deal of work in moving loneliness from a personal complaint to a public health category. It has also flattened a body of evidence in which the actual effect sizes vary by a factor of two or more, depending on which study, which outcome, and which construct — isolation or loneliness — is being measured.

It is worth going back to the numbers themselves.

Three effect sizes, not one

Julianne Holt-Lunstad’s 2010 meta-analysis, pooling 148 studies and 308,849 participants, found that people with stronger social relationships had a 50% increased likelihood of survival over the follow-up periods studied. That is a large effect, and it is the number most often cited when someone wants to establish that social connection matters for mortality at all.

Her 2015 meta-analysis, however, is more useful for comparing constructs against each other, because it reports three separate odds ratios rather than one pooled figure: social isolation carried an odds ratio of 1.29 for early mortality, loneliness 1.26, and living alone 1.32. These are not small effects, but they are also not the “50% increased likelihood” figure. The 2010 and 2015 papers are frequently cited interchangeably as if they say the same thing. They do not. One reports a composite protective effect of strong relationships; the other decomposes risk into three distinct and only moderately correlated exposures, each with a materially smaller effect size.

The 2022 American Heart Association scientific statement, led by Crystal Cene, adds a fourth data point specific to cardiovascular outcomes: social isolation and loneliness were associated with roughly a 29% increased risk of heart attack or death from heart disease, and a 32% increased risk of stroke. Notice the pattern. General mortality risk clusters in the 25–30% range across most recent estimates once the analysis is disaggregated by construct. The 50% figure from 2010 sits well above that range, and it is the one that survives in advocacy language and government advisories, likely because it is the largest and most quotable number, not because it is the most representative one.

Why the range matters more than the headline

None of this is a case for dismissing the 2010 finding. A meta-analysis of 148 studies is a substantial evidence base, and a 50% figure for a composite exposure is not implausible given how many pathways — behavioral, physiological, and access to care — connect social relationships to survival. The point is that policy communications routinely borrow the largest available number to establish urgency, then apply it to a much narrower claim, such as the effect of loneliness alone, or of a specific intervention aimed at loneliness alone.

The AHA statement is unusually candid about this gap. It explicitly names the absence of intervention evidence as the central research gap in the field: strong observational effect sizes for isolation and loneliness as risk factors, but very little trial evidence establishing that reducing isolation or loneliness reduces cardiovascular risk by any comparable margin. An association of roughly 30% between loneliness and stroke risk says nothing about what fraction of that risk a successful intervention could recover.

The intervention evidence, by contrast, is thin and small

This is where the RCT literature is most exposed by comparison. The HEAL-HOA trial, published in the Lancet Healthy Longevity in November 2024, is one of the few randomized controlled trials testing a loneliness intervention against a control group among lonely older adults, in this case in Hong Kong. Most of the intervention literature consists of uncontrolled programme evaluations that cannot produce an effect size comparable to the observational odds ratios above, because there is no counterfactual. Where a controlled effect size exists, it should be reported as such and not blended with the risk-factor literature, which measures a different thing: the strength of an association at baseline, not the size of a treatable effect.

The 2023 BMC Public Health review of the state of loneliness research names this directly as a structural problem: inconsistent measurement across studies makes effect sizes difficult to compare, let alone pool into confident population-level estimates. A 2024 study in Scientific Reports adds a further complication specific to age: the relationship between isolation and loneliness is not constant across the life course, which means a pooled effect size across all ages may misstate the risk for any particular group, older adults especially.

What the size of the number should be used for

Effect sizes in this range — odds ratios clustering around 1.25 to 1.3 for loneliness and isolation individually, higher when relationship strength is measured as a composite — are large enough to justify treating social disconnection as a legitimate population health exposure, on the same order as several established risk factors and unambiguously suitable for surveillance. That is a defensible finding to publicize.

What is not yet supported is any published estimate of how much that risk is reduced by a specific policy, program, or clinical intervention. Until controlled trials on the scale of the observational meta-analyses exist, that number simply is not available. Reports that borrow the largest observational effect size and imply it as an achievable reduction from action are making an inference the evidence does not support.