What The '15 Cigarettes A Day' Loneliness Figure Actually Measures

The 2023 U.S. Surgeon General advisory contains a sentence that has by now appeared in hundreds of news stories, conference slides, and op-eds: the mortality risk associated with social disconnection is comparable to smoking up to 15 cigarettes a day. It is a striking comparison, precisely calibrated to be quotable. It is also worth asking what it actually measures, because the figure has traveled much further than its original evidentiary basis was designed to support.

Where the number comes from

The comparison traces to Julianne Holt-Lunstad’s 2010 meta-analysis in PLoS Medicine, which pooled 148 studies covering 308,849 participants and found that stronger social relationships were associated with a 50% increased likelihood of survival over the follow-up periods studied. Holt-Lunstad and colleagues then situated that effect size against other well-established mortality risk factors, smoking among them, to communicate its magnitude to a clinical audience unfamiliar with treating social connection as a health variable at all. The Surgeon General’s office adopted a version of that framing in 2023, extending it into the cigarette-count language that has since become the figure’s public identity.

Two things are true about this number, and they pull in different directions. It is a real, well-powered synthesis of a genuinely large evidence base. And it is a comparison of relative risk magnitude, not a claim that social disconnection and smoking operate through the same physiological pathway or carry the same dose-response structure. Smoking’s harm scales with cigarettes smoked, a continuous and biologically direct exposure. The 2010 meta-analysis measured something categorical and considerably more heterogeneous: whether a person fell into a stronger or weaker category of social relationship, defined differently across the 148 source studies.

What “social relationships” meant in that pooled estimate

This is the part that gets lost in the retelling. The 2010 meta-analysis did not measure one thing called loneliness. It pooled studies that operationalized social relationships through marital status, frequency of contact, network size, participation in group activities, and subjective loneliness ratings, among other measures. Some of these are structural properties of a person’s network — whether they live alone, how many social ties they have. Others are subjective states — how connected a person feels regardless of the objective size of their network. Isolation and loneliness are independently predictive of health outcomes, and conflating them inside a single pooled effect size obscures which one is doing the work in any given context.

Holt-Lunstad’s own 2015 follow-up meta-analysis in Perspectives on Psychological Science makes this distinction explicit in a way the 2010 paper’s headline number does not. That later analysis separated the three constructs and reported different odds ratios for each: social isolation carried an odds ratio of 1.29 for early mortality, loneliness 1.26, and living alone 1.32. The effects held after adjusting for health status, and the paper found that social deficits were more predictive of death in samples averaging under 65 than in older cohorts, an age pattern that runs against the common assumption that connection matters most in old age. But the separation itself is the methodological contribution. It shows that the widely cited 50% figure from 2010 is an average across constructs that behave somewhat differently, at different odds ratios, and probably through different mechanisms — isolation plausibly operating through reduced practical and instrumental support, loneliness through chronic stress physiology.

What the comparison to smoking can and cannot support

None of this means the cigarette comparison is wrong as a communication device. It is defensible as a statement that the relative risk associated with weak social relationships sits in the same rough range as the relative risk associated with heavy smoking, drawn from comparably large and well-replicated observational literatures. Holt-Lunstad has made a related argument directly, in a 2021 review in the American Journal of Lifestyle Medicine, that social connection belongs alongside diet, exercise, and smoking cessation as a modifiable target for preventive medicine — a framing that depends on this magnitude comparison holding up.

What the comparison cannot support is a claim about mechanism, dose-response, or reversibility. Smoking cessation has decades of intervention trial evidence behind it, showing that quitting reduces risk in a measurable and time-dependent way. The 2022 American Heart Association scientific statement on social isolation, cardiovascular disease, and stroke — which independently found a roughly 30% increased risk of heart attack, stroke, or death from either associated with isolation and loneliness — is candid about the corresponding gap on the social side. It identifies the absence of intervention evidence as the central research gap in the field. There is a large observational literature establishing that disconnected people fare worse. There is comparatively little randomized evidence establishing that improving connection changes that trajectory, at the population-mortality level the smoking comparison implies.

This matters because all of the studies feeding the pooled estimates — the 2010 meta-analysis, the 2015 follow-up, the AHA statement — are observational cohort data. They establish association, not causation, and the direction of causation is not settled. Illness plausibly drives isolation as much as isolation drives illness: people with declining health withdraw from social contact, get excluded from groups organized around physical activity, and lose network ties through their own or others’ mortality. The National Academies’ 2020 consensus report on isolation and loneliness in older adults — which put roughly a quarter of adults 65 and older in the isolated category — calls for routine clinical assessment of isolation precisely because the health care system has no comparable infrastructure for it that it has for smoking status, but it stops short of claiming the causal chain runs cleanly in one direction.

The CDC’s 2024 surveillance report, using 2022 survey data, links loneliness and lack of social and emotional support to heart disease, stroke, dementia, type 2 diabetes, depression, anxiety, and premature mortality, and does so with the appropriate hedge: these are associations documented in surveillance data, not effects demonstrated through intervention.

What would actually settle the comparison

A study that would strengthen this specific claim would need three things the current literature mostly lacks. First, a design that separates isolation from loneliness rather than pooling them, following the 2015 meta-analysis’s approach rather than the 2010 headline. Second, longitudinal or intervention data that can distinguish disconnection as cause from disconnection as consequence of declining health, rather than the mostly cross-sectional and short-follow-up cohort data underlying most of these pooled estimates. Third, and most importantly, a body of randomized trials analogous to smoking-cessation trials, testing whether structured increases in social connection reduce mortality risk over time, at a scale comparable to the 308,849-person meta-analysis that produced the original number.

Until that evidence exists, the smoking comparison is best read as a statement about how large the association is, made by researchers trying to get a genuinely underweighted risk factor taken seriously in clinical and policy settings. It is not yet a statement about how reversible that risk is, or through what specific pathway it operates. Those are different, harder questions, and the headline figure was never built to answer them.