The Risk Factor Nobody Checks
At your annual physical, someone checks your blood pressure, your cholesterol, your blood sugar, and your weight. You get asked about smoking and drinking. You might get a questionnaire about depression. Nobody asks how many close friends you have, how often you see them, or whether you ate dinner alone last night.
Julianne Holt-Lunstad, a psychologist at Brigham Young University, thought this was a problem. She knew from a 1988 review in Science that social relationships predicted mortality, but the finding hadn't entered clinical practice. Physicians didn't screen for it. Public health agencies didn't campaign against it. The claim felt soft, she suspected, because social relationships were measured in too many ways across too many small studies for any single result to command attention. So she decided to synthesize all of them.
What 148 Studies Found
Holt-Lunstad, along with Timothy B. Smith and J. Bradley Layton, searched PsycINFO, MEDLINE, CINAHL, Sociological Abstracts, and the Social Sciences Citation Index for every prospective study that measured some aspect of social relationships and tracked participants long enough to record deaths. They found 148 studies covering 308,849 people across North America, Europe, Asia, and Australia, with an average follow-up of 7.5 years. The results, published in PLOS Medicine in 2010, were striking.
Participants with stronger social relationships had 50% higher odds of survival (OR = 1.50, 95% CI 1.42 to 1.59). The effect was consistent across age, sex, initial health status, cause of death, and follow-up period. Studies using complex, multidimensional measures of social integration yielded an even larger effect: 91% higher odds of survival (OR = 1.91, 95% CI 1.63 to 2.23).
How large is that? The paper's Figure 6 placed the effect alongside other established risk factors for mortality, drawn from published meta-analyses of each. Social relationships exceeded the influence of physical inactivity (based on cardiac rehabilitation trials), obesity and overweight, and flu vaccination. The effect was comparable to smoking cessation and exceeded the protective effect of pharmaceutical interventions for cardiac patients. Holt-Lunstad later contextualized the finding in a way that stuck: lacking social connection carries a mortality risk equivalent to smoking up to 15 cigarettes per day.
The Confirmation
Five years later, Holt-Lunstad's team published a second meta-analysis in Perspectives on Psychological Science with more than ten times the participants: 70 studies and 3.4 million people from North America, Europe, Asia, and Australia. This time they focused specifically on the absence of social connection. The results: social isolation increased mortality odds by 29% (OR = 1.29), loneliness by 26% (OR = 1.26), and living alone by 32% (OR = 1.32). All three effects survived adjustment for health status, socioeconomic factors, and depression. The findings did not vary by gender or world region.
Independent teams have since piled on. Rico-Uribe et al. (2018) found loneliness predicted mortality in a meta-analysis of 35 studies. Shor and Roelfs (2015) examined social network size across 91 cohorts and reported a 13% mortality increase per standard deviation decrease in network size. A 2024 comprehensive meta-analysis and meta-regression confirmed the core finding: social isolation raised all-cause mortality in older adults by 35%.
In May 2023, U.S. Surgeon General Vivek Murthy issued a formal advisory declaring loneliness and social isolation "an underappreciated public health crisis," citing Holt-Lunstad's work by name and stating that social disconnection "increases the risk for premature death to the same levels as smoking up to 15 cigarettes a day."
A Statistical Caveat Worth Stating
The "50% increase" language requires a precision check. Published commentary on the study noted that an odds ratio of 1.50 is not the same as a 50% increase in the probability of survival. With the study's average mortality rate of 29%, an OR of 1.50 translates to roughly a 32% increase in the risk of dying for the socially disconnected compared with the socially connected (risk ratio β 1.32). Still a large effect. Still exceeding obesity. But the magnitude depends on which metric you use, and odds ratios overstate absolute differences when baseline rates are high. Holt-Lunstad has acknowledged this distinction in subsequent interviews while noting that even the more conservative estimate places social disconnection among the strongest modifiable risk factors for death.
The Strongest Counterargument
The most serious critique is reverse causation: sick people become isolated because they are sick, not the other way around. Physical disability restricts mobility. Chronic illness drains the energy needed to maintain friendships. Depression, a known mortality risk factor, corrodes social motivation. If the studies are simply detecting illness masquerading as loneliness, the causal arrow points the wrong direction.
This objection has weight. The 2010 meta-analysis included both community samples and clinical populations, and while initial health status did not moderate the overall effect, the 2015 follow-up acknowledged that effect sizes shrank when more covariates were controlled. Critically, though, the effect did not disappear. In fully adjusted models accounting for health status, depression, and socioeconomic factors, social isolation and loneliness remained independent predictors of mortality. When physically ill individuals were excluded entirely, the association persisted. The biological plausibility has grown stronger too: social isolation is now linked to elevated inflammation (C-reactive protein, interleukin-6), impaired immune function, heightened cortisol reactivity, and accelerated telomere shortening, all through experimentally supported pathways.
What We Didn't Prove
Meta-analyses of observational studies cannot establish causation. You cannot randomly assign people to have no friends for twenty years and measure the result. The included studies varied widely in how they measured social relationships, from binary marital-status variables to validated multi-item inventories of social integration, and the heterogeneity across studies was large (I2 = 81%). The data come overwhelmingly from North America and Western Europe; generalization to collectivist cultures or the Global South remains untested. The "smoking 15 cigarettes" comparison, while directionally supported by the effect size data, is a rough benchmark rather than a precise equivalence, since smoking dose-response curves and social connection operate through entirely different biological mechanisms. The 2015 follow-up found that the mortality effect of loneliness and social isolation was weaker in adults over 65, possibly because competing causes of death dominate at older ages.
The Bottom Line
For decades, public health has focused on the risks you can see: the cigarette in your hand, the number on the scale, the minutes you spend on the treadmill. Meanwhile, a risk factor at least as powerful has gone unscreened. Two meta-analyses covering more than 3.7 million participants, confirmed by independent teams and endorsed by the U.S. Surgeon General, show that weak social connections predict death as reliably as the factors your doctor already asks about. Social isolation exceeds obesity as a mortality risk factor. It is comparable to smoking. And no one checks for it at your annual physical.
What You Can Do
Treat your social connections as a health behavior, not a luxury. Schedule time with friends the way you schedule exercise. If you have moved to a new city, joined remote work, or retired, recognize that you have lost structural sources of daily social contact and need to replace them deliberately. Evidence-based strategies include joining groups organized around shared activities (not forced socializing), volunteering, and simply talking to the people you encounter during routine errands. If you are a physician or health worker, consider adding a brief social connection screen to annual checkups: the three-item UCLA Loneliness Scale takes under a minute. If you are a manager, recognize that return-to-office policies and remote-work policies both carry social health trade-offs that deserve explicit attention. The research does not prescribe a minimum number of friends. It says that the quality and variety of your connections matter more than the raw count.