The Fifteen-Cigarettes Claim: Where Does It Actually Come From

A single sentence has done more to popularize the public health framing of loneliness than any other in the field. It holds that weak social ties are as dangerous to health as smoking fifteen cigarettes a day. It appears in health system briefings, in employer wellness materials, in journalism covering the subject, and in the introductions of books written by people with credentials to know better. It is rarely footnoted past a general gesture toward “research.” The question this piece asks is narrow: has anyone actually replicated that specific comparison, with that specific number, or does the whole edifice rest on one paper being paraphrased with increasing looseness over a decade.

What the 2010 paper actually measured

The source almost everyone is gesturing toward is Julianne Holt-Lunstad’s 2010 meta-analysis in PLoS Medicine, which pooled 148 studies covering 308,849 participants and found that people with stronger social relationships had a 50% greater likelihood of survival over the follow-up periods studied, compared with people with weaker relationships. That is a real and large effect. The paper’s discussion section does compare this effect size, in general terms, to other well-established mortality risk factors — the kind of comparison meta-analysts make routinely to give readers a sense of scale, noting that the magnitude of the association is comparable to that of established risk factors for mortality.

What the paper does not contain, as far as the published findings establish, is a quantified head-to-head calculation converting relationship strength into cigarettes per day. That number — fifteen, specifically — is not a result the meta-analysis reports. It functions, in nearly every downstream citation, as an illustrative gloss on “comparable to smoking,” not as a finding anyone measured directly. A gloss can be accurate in spirit and still not be a replicated empirical claim. The distinction matters because the entire rhetorical power of the fifteen-cigarettes line comes from its precision — it sounds like something someone calculated — and the precision is not there in the source usually cited for it.

The 2015 follow-up moved in a different direction entirely

Five years later, Holt-Lunstad published a second meta-analysis in Perspectives on Psychological Science, this time separating out three distinct predictors that the 2010 paper had treated together: social isolation, loneliness, and living alone. The odds ratios were 1.29 for isolation, 1.26 for loneliness, and 1.32 for living alone, for early mortality. These are meaningful effects — comparable in scale to a number of recognized clinical risk factors — but they are considerably smaller than a 50% survival advantage, and they are not expressed anywhere as a cigarette-equivalent either.

This is worth sitting with, because it changes what the “as bad as smoking” claim is actually standing in for. The 2010 estimate combined many different operationalizations of social relationship strength — network size, contact frequency, marital status, integration into groups — into a single pooled effect. The 2015 paper split that composite into isolation, loneliness, and living alone specifically because they are not the same construct and do not carry the same weight. Isolation is a structural fact about a person’s network: how many ties, how often activated. Loneliness is the subjective experience of feeling that those ties are insufficient. A person can be lonely with many social contacts, or isolated without feeling lonely at all. The 2015 paper found these two things independently predictive of mortality, at similar but not identical magnitudes, which is itself evidence that collapsing them into one number — as the cigarette comparison implicitly does — obscures more than it clarifies.

So when a document cites “loneliness” and reaches for the smoking analogy, it is often reaching for a number generated from a pooled 2010 estimate of social relationship strength broadly, applying it specifically to loneliness, which the researcher’s own later work treated as a distinct and somewhat smaller-effect construct. That is not fraud. It is the kind of compression that happens when a nuanced finding travels through several layers of paraphrase, but it is worth naming precisely because it happens.

Where the claim shows up next, and what it does there

The 2021 review by Holt-Lunstad in the American Journal of Lifestyle Medicine argues for treating social connection as a modifiable risk factor belonging alongside diet, exercise, and smoking cessation in preventive health frameworks. This is a reasonable argument, and it is the argument the fifteen-cigarettes line is usually deployed to support: not “loneliness causes X amount of harm,” but “loneliness belongs in the same category of thing medicine already takes seriously.” The 2021 paper does not, however, produce a new quantified comparison to smoking. It cites the existing body of work and makes a policy argument from it.

Vivek Murthy’s 2020 book, written while serving as U.S. Surgeon General, frames loneliness as a public health issue on a par with other major risk factors, and helped move the smoking comparison into mainstream political and media discourse. Cigna’s 2020 workplace loneliness report, which found 61% of U.S. adults reporting loneliness sometimes or always, sits in the same lineage of documents that treat the health comparison as established background rather than something requiring its own citation. Neither source, going by what they report, independently measured or replicated a cigarette-equivalent figure; both are working from the same underlying 2010 result, filtered through the same simplifying analogy that has by now detached from its original hedge.

The National Academies’ 2020 consensus report on isolation and loneliness in older adults takes a more careful path. It calls on the health care system to routinely assess isolation and loneliness, treating the underlying mortality evidence as sufficient to justify screening, without leaning on the cigarette framing at all. That absence is itself informative. A consensus report assembled by a panel with time to check primary sources chose not to repeat the analogy, even though it had every reason to want a vivid public-facing number if one were defensible.

What would actually settle this

No study identified here directly compares loneliness or isolation to smoking as risk factors within the same sample, using the same statistical model, in a way that would produce a genuine equivalence figure rather than a rhetorical one. Doing that would require a cohort study measuring both smoking history and validated loneliness or isolation scores in the same participants, following them long enough to observe mortality, and modeling both exposures side by side with comparable units — which is difficult, since “loneliness” does not have a dose scale analogous to cigarettes per day. The 2010 and 2015 meta-analyses come closer than most public health claims ever get to a rigorous quantitative basis, which is precisely why the looseness of what gets built on top of them is worth flagging. The underlying science is genuinely strong. The specific number attached to it in slide decks and press releases is not something anyone appears to have actually calculated.

None of this means the underlying claim — that weak social ties carry mortality risk on the scale of other major, medically recognized risk factors — is wrong. The 2010 and 2015 meta-analyses, taken on their own terms, support a claim close to that one. What is not supported, on the evidence available, is the specific and now ubiquitous cigarette figure, which appears to have originated as an illustrative aside and has since been repeated as though it were itself a measured finding. Anyone citing it should be citing the general comparability language in the 2010 paper, not a number, and should be prepared for the fact that the closest thing to a number the field has since produced — the separate odds ratios for isolation and loneliness in 2015 — is smaller and more specific than the phrase implies.