You have seen the line. It appears in newspaper headlines, on charity posters, in podcast intros, and — in a slightly more careful form — on this website. Is loneliness as bad for you as smoking?
The honest answer is: the underlying research is real and reasonably strong, the comparison is legitimate as a piece of scientific benchmarking, and almost every popular version of it says something the original data does not support. All three of those are true at once, and a claim we repeat on our own site deserves to be examined rather than just repeated.
Where the figure comes from
The source is a 2010 meta-analysis published in PLOS Medicine by Julianne Holt-Lunstad, Timothy Smith and J. Bradley Layton: Social Relationships and Mortality Risk: A Meta-analytic Review.
It pooled 148 studies covering 308,849 participants. The headline result was an odds ratio of 1.50 — a 50 percent greater likelihood of survival over the follow-up period for people with stronger social relationships. That finding held across age, sex, initial health status, cause of death and length of follow-up.
The paper then did the thing that made it famous. It placed that effect size alongside effect sizes for other well-established mortality risk factors, each drawn from its own separate meta-analysis: smoking, obesity, physical inactivity, air pollution, excess alcohol. Lacking social connection landed in the same range as smoking up to 15 cigarettes a day.
It is worth being precise about what that chart is, because the precision is the point of this whole piece. It is a benchmark, not a league table. The measures of social connection on it do not all score the same — broader, more complex measures of social integration carry larger effects than narrower ones like living alone — and reading a ranking into it is the first step down the road that produces the headlines.
That is the whole provenance. It is not a single dramatic study, and it is not a marketing figure. It is a benchmark chart in a peer-reviewed meta-analysis.
What the comparison actually says
Holt-Lunstad has since published a page correcting the ways her own statistic gets mangled, which is a fairly unusual thing for a researcher to have to do. Two corrections matter most.
It is not about loneliness alone. The original comparison was based on an aggregate across many measures of social connection — social network size, social support, social integration, marital status, living alone, participation in social or religious groups. Loneliness was one of them, not the whole thing. Every version of the claim that begins "loneliness is..." has already narrowed the finding.
It is "up to 15 cigarettes", not "two packs". The smoking effect size used was for moderate daily smoking, defined as up to 15 cigarettes. Presumably 40 a day is worse; the original source does not estimate that. The accurate sentence, in her words, is: "Lacking social connection is comparable to smoking up to 15 cigarettes per day."
And "comparable" means comparable in magnitude of association with earlier death. It is not a clinical equivalence. Nobody has established that a quiet year does to your arteries what a year of Marlboros does.
So, is loneliness as bad for you as smoking?
Here it gets more interesting, because the sharpest critique of the comparison was co-written by the person who created it.
In 2023, Robert W. Smith, Julianne Holt-Lunstad and Ichiro Kawachi published a commentary in the American Journal of Epidemiology: Benchmarking Social Isolation, Loneliness, and Smoking. Their conclusion is that the comparison "has been helpful for raising awareness of robust evidence linking social relationships and health", but that "the analogy often oversimplifies the evidence".
Four things it leaves out.
Relative risk is not absolute risk. Odds ratios tell you how much a risk factor multiplies your chances relative to someone without it. They tell you nothing about how many people it actually kills. Two risk factors can have identical odds ratios and wildly different population impact, depending on how common they are and who has them.
On population impact, the numbers diverge sharply. The commentary cites a study across 20 countries which estimated that 2.4 percent of deaths might not have occurred without social isolation, against 20.6 percent without tobacco smoking. On that measure smoking is roughly an order of magnitude worse. The authors are careful to note that this is not a like-for-like test of the original chart, which compared moderate smoking specifically — and when they ran their own crude comparison using U.S. cohort data, the population attributable fractions for social isolation and for smoking up to a pack a day came out close to each other, at about 7 and 6.5 percent. Which comparison is fair depends on which question you are asking, and the honest position is that both figures are defensible and they point in different directions.
Causation is not settled. These are observational studies. Sick, poor and depressed people have smaller social networks, and smaller social networks may then make them sicker. Untangling which way the arrow runs is genuinely hard, and the commentary is explicit that the evidence defining causal mechanisms is not yet at the standard we have for tobacco. With smoking, we know what the smoke does to the tissue. With disconnection, we have plausible pathways — stress physiology, inflammation, sleep, health behaviors — and much less certainty.
The measure changes the answer. In the 2010 meta-analysis the association was strongest for complex measures of social integration, at an odds ratio of 1.91, and weakest for the simple binary of living alone versus living with others, at 1.19 — with a confidence interval that crossed 1, meaning it was not statistically significant. Living by yourself is not the risk factor. The texture and range of your relationships is.
What the comparison gets right
None of this makes the claim junk. It makes it a benchmark rather than an equation.
Smoking is a reasonable thing to benchmark against: it is one of the most researched exposures in medicine and one of the great public health successes. And the direction of the underlying evidence really is consistent. The Surgeon General's 2023 advisory reports that loneliness and social isolation are associated with a 26 percent and 29 percent increased risk of premature death respectively, that poor social connection carries a 29 percent higher risk of heart disease and a 32 percent higher risk of stroke, and that chronic loneliness and isolation can raise the risk of dementia in older adults by approximately 50 percent. Those are separate findings, not restatements of the cigarette line, and they are not seriously disputed.
There is also a point in the commentary that has nothing to do with statistics and has stayed with me. The authors argue that the smoking analogy has quietly pushed us toward treating disconnection as an individual condition — something to be fixed with therapy, programs, befriending schemes, even medication — when the tobacco story was actually won at the population level, through the design of buildings, prices, laws and public space.
If that reading is right, then the useful response to the evidence is not to tell people to feel less lonely. It is to change the number of places where connection can happen at all.
What we do with it
I run 5 Degrees because of the second half of that argument, not the first. Not because a statistic frightened me, but because the settings that used to put strangers in front of each other have thinned out, and building new ones is something an ordinary person can actually do.
So when this site cites the 15-cigarettes figure — and it does — take it as what it is. A serious finding, carefully benchmarked, about the size of an association between social connection and dying earlier. Not a diagnosis, not an equivalence, and not a reason to panic.
It is a reason to take the problem seriously enough to build something. That is a lower bar than the headline implies, and a more useful one.
