Fifteen cigarettes a day: where the loneliness number comes from

Loneliness as deadly as 15 cigarettes a day? The number was read off a 2010 bar chart, where the main result is shorter than the smoking bar.

You have probably met the sentence. Loneliness is as bad for you as smoking fifteen cigarettes a day. It has been in newspapers, in charity leaflets and in at least one article on this site, our own piece on the loneliness epidemic. In May 2023 it reached the top: the US Surgeon General, Vivek Murthy, put it in his advisory Our Epidemic of Loneliness and Isolation. "Lacking social connection can increase the risk for premature death as much as smoking up to 15 cigarettes a day."

A number that precise sounds measured. It was not measured. It was read off a bar chart.

The chart

The chart is Figure 6 of a paper published in PLoS Medicine in July 2010 by Julianne Holt-Lunstad of Brigham Young University and two colleagues, Timothy Smith and Bradley Layton. They pooled 148 studies that had followed 308,849 people for an average of seven and a half years and asked a plain question: did the people with stronger social relationships die less often?

They did. The pooled odds ratio was 1.50, which the authors described as a 50 percent greater likelihood of survival. How the relationships were measured mattered a great deal. Studies that used complex measures of social integration found an odds ratio of 1.91. Studies that only asked whether a person lived alone found 1.19, and that one was not even clearly different from no effect.

Figure 6 then lines up the result against other risk factors, taken from other people's meta-analyses published between 1994 and 2008. Alcohol, flu vaccination, cardiac rehabilitation, physical activity, obesity, blood pressure drugs, air pollution. And one bar labelled "Smoking < 15 cigarettes daily".

Here is the detail almost nobody repeats. The bar for the overall finding of the meta-analysis, the famous 50 percent, is shorter than the smoking bar. Only two subgroups reach past it: studies contrasting high and low social support, and studies with complex measures of integration. So "as bad as fifteen cigarettes" is true of the strongest slice of the evidence. For the headline result, a fairer summary would be "not quite as bad as smoking fewer than fifteen a day".

The authors were careful about this in their own text. They wrote that the effect was "comparable with" well-established risk factors and exceeded obesity and physical inactivity. The cigarettes came later. The Surgeon General's version draws its figure from a 2017 review by Holt-Lunstad, Theodore Robles and David Sbarra in American Psychologist, and a small note under the advisory's chart says: "This graph is a visual approximation."

What 50 percent means

A second trap sits inside the word "likelihood". An odds ratio is not a probability, and the two drift apart when an outcome is common. Survival over seven years is common.

Try it with invented round numbers. Say 80 of every 100 socially isolated people in a study are still alive at the end. Their odds of survival are 80 to 20, or 4. Multiply by 1.5 and you get 6, which is about 86 survivors per 100. That is a real difference, six more people alive in every hundred. It is not what most readers picture when they hear "50 percent more likely to survive".

Does the link hold up?

The link itself has held up better than the slogan. In 2023 a team at Harbin Medical University in China published a new meta-analysis in Nature Human Behaviour: 90 cohort studies, more than 2.2 million people. Social isolation, meaning few contacts you can count, went with a 32 percent higher risk of death from any cause. Loneliness, the feeling, went with 14 percent. In 2024 three researchers working with the Institute for Replication, Fabio Molo, Samuel Pawel and Gorka Fraga González, re-ran that analysis and tried other reasonable methods. The main claims were robust, they wrote, though most corrections for publication bias gave "somewhat smaller effects".

Two things in those numbers are worth holding on to. Being alone and feeling alone are not the same, and the first is the stronger signal. And all of this is observational. Holt-Lunstad and her colleagues said so in 2010: you cannot randomly assign people to be isolated. People who are already ill withdraw from others, and poverty, depression and disability push the same way. The studies adjust for what they can. They cannot adjust for everything.

What to do with it

We are going to stop using the fifteen cigarettes. Not because the harm is invented. Because the precise comparison is a picture of a picture, and the honest version is strong enough on its own: across dozens of cohorts and millions of people, those with few relationships die sooner, the effect sits in the same range as other risk factors doctors take seriously, and it is bigger when you measure relationships properly.

The practical part is the same either way. The measure that predicted survival best in 2010 was not whether someone had a spouse or lived with somebody. It was how woven in they were: how many kinds of contact, how often, with whom. That is the sort of thing a street can change. A neighbour who knocks is not a treatment and has never been tested as one. But the evidence points in their direction, and it does not require the number to be exact. The same pattern turned up when we looked at the famous rat experiments on addiction, in Rat Park never replicated. Rats still choose company over heroin: the neat story fell apart, and something smaller and sturdier was left.

Sources

• J. Holt-Lunstad, T. B. Smith, J. B. Layton, "Social Relationships and Mortality Risk: A Meta-analytic Review", PLoS Medicine 7(7): e1000316, 2010 — https://doi.org/10.1371/journal.pmed.1000316 (Figure 6: https://journals.plos.org/plosmedicine/article/figure?id=10.1371/journal.pmed.1000316.g006)
• Office of the U.S. Surgeon General, "Our Epidemic of Loneliness and Isolation", 2023, Figure 4 and note — https://www.hhs.gov/sites/default/files/surgeon-general-social-connection-advisory.pdf
• J. Holt-Lunstad, T. F. Robles, D. A. Sbarra, "Advancing social connection as a public health priority in the United States", American Psychologist 72(6), 517–530, 2017 — https://doi.org/10.1037/amp0000103
• F. Wang, Y. Gao et al., "A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness and mortality", Nature Human Behaviour 7, 1307–1319, 2023 — https://doi.org/10.1038/s41562-023-01617-6
• F. Molo, S. Pawel, G. Fraga González, "A Robustness Reproduction of 'A Systematic Review and Meta-Analysis of 90 Cohort Studies of Social Isolation, Loneliness and Mortality'", I4R Discussion Paper 169, 2024 — https://ideas.repec.org/p/zbw/i4rdps/169.html