The Number on the Scale
Step on a scale at your doctor's office, and the verdict follows fast. Body mass index above 25? Overweight. A flag in your chart. Maybe a conversation about diet and exercise, or a referral to a nutritionist. The message from public health authorities, medical associations, and government dietary guidelines has been consistent for decades: anything above "normal" weight is a health risk you should work to correct.
In 2013, a CDC senior scientist named Katherine Flegal tested this assumption against the largest evidence base ever assembled, as her team at the National Center for Health Statistics systematically reviewed 97 prospective studies covering more than 2.88 million individuals across four continents. Their outcome measure was simple: death from any cause.
What came back from the data inverted the expected gradient. Adults classified as overweight, with BMI between 25 and 30, had a hazard ratio of 0.94 for all-cause mortality compared to those at "normal" weight (BMI 18.5–25). The overweight group was 6% less likely to die during study follow-up periods, and the 95% confidence interval of 0.91 to 0.96 excluded 1.0, confirming the finding was statistically significant.
Grade 1 obesity (BMI 30–35) showed a similar pattern: HR 0.95 (95% CI 0.88–1.01), meaning no statistically significant increase in mortality. Only at BMI 35 and above did mortality climb, with a hazard ratio of 1.29.
The Paradox Has a Name
Cardiologists had noticed this pattern for years in heart failure patients, where sicker patients with higher BMIs sometimes survived longer — a phenomenon they called the "obesity paradox." But Flegal's meta-analysis went further, covering the general population, healthy and sick alike, using the same standard BMI categories that doctors and insurers apply every day.
The finding held when Flegal limited the analysis to studies with measured rather than self-reported height and weight, and it held across age groups, with the protective association growing even stronger among adults 65 and older (HR 0.90, 95% CI 0.86–0.94). The meta-analysis drew on more than 270,000 deaths across studies spanning decades, was published in JAMA, and has been cited more than 5,000 times.
The Strongest Case Against It
Harvard's Department of Nutrition pushed back within weeks, as Frank Hu, Walter Willett, and JoAnn Manson convened a public panel and called the findings "flawed and misleading," arguing that the core issue was reverse causation.
People in the "normal" BMI category include those who are thin because they are already dying: undiagnosed cancer, chronic lung disease worsened by smoking, the wasting that accompanies frailty in old age. By failing to exclude these individuals, the critics argued, Flegal's meta-analysis inflated the mortality of the normal-weight group, making overweight look artificially protective.
The Global BMI Mortality Collaboration put this hypothesis to the test in 2016, using individual-level data from 10.6 million adults across 239 studies and restricting their analysis to never-smokers without chronic disease at enrollment, then excluding the first five years of follow-up. Under these conditions, overweight was associated with an 11% increase in mortality (HR 1.11, 95% CI 1.10–1.11). There is also no well-established biological mechanism for a protective effect of moderate overweight, since extra adipose tissue increases insulin resistance, blood pressure, and inflammatory markers. "Even as you get near the upper reaches of the normal weight range, you begin to see increases in chronic diseases," Manson noted in the panel discussion.
Why the Debate Isn't Settled
Flegal responded in a 2019 paper co-authored with Stanford's John Ioannidis, documenting what they called methodological flaws in the Global BMI Mortality Collaboration's approach and noting that the collaboration's restrictions excluded roughly 60% of participants and 75% of deaths. The resulting sample, consisting of healthy never-smokers alive at least five years after enrollment, is not representative of the population that actually walks into a doctor's office. A meta-analysis that excludes the majority of its data is answering a different question: not "is overweight dangerous for people in general?" but "is overweight dangerous for a narrow subset of exceptionally healthy people?"
A 2016 study of 12 million Korean adults by Yi and colleagues lent support to Flegal's findings. It applied the same restrictive criteria used by the Global BMI Mortality Collaboration and still found overweight associated with lower mortality (HR 0.85). Flegal characterized the collaboration's analytical approach as introducing selection bias rather than correcting for it.
A separate 2016 meta-analysis by Aune and colleagues in the BMJ may point toward a partial resolution. Covering 230 cohort studies and 30.3 million participants, it found the lowest mortality at BMI 23–24 in never-smokers, with a J-shaped curve. This suggests the problem may lie in the categories themselves, since the "normal" BMI range of 18.5–25 spans a wide band whose lower end includes borderline underweight individuals whose elevated mortality drags down the group average, potentially making the overweight category look protective by comparison.
What We Didn't Prove
This meta-analysis does not demonstrate that gaining weight improves your health, because observational data can identify associations but cannot establish causation. The hazard ratio of 0.94 does not mean that eating more to move from BMI 23 to BMI 27 would reduce your mortality risk. The overweight adults in these studies may differ from normal-weight adults in unmeasured ways: muscle mass, physical activity, socioeconomic status, genetics.
BMI itself is a crude instrument that cannot distinguish between muscle and fat, does not account for fat distribution, and was never designed as an individual clinical tool.
Heterogeneity across the 97 included studies was high (I² = 85%), and while the overall effect pointed consistently in one direction, the magnitude varied enough to warrant caution about applying the summary hazard ratio to any individual person.
All-cause mortality is only one outcome, and being overweight is consistently associated with higher rates of type 2 diabetes, hypertension, and osteoarthritis. Lower mortality does not necessarily mean lower morbidity.
The Bottom Line
The largest meta-analysis of BMI and mortality ever conducted found that the weight category your doctor calls "overweight" was not associated with higher death rates — it was associated with lower ones. This does not invalidate the medical risks of excess body fat, but it does suggest that the clean boundary at BMI 25 is a policy convention, not a biological threshold, and that the relationship between weight and longevity is far messier than a single number implies.
What You Can Do
Stop fixating on BMI 25 as a cliff edge. The data from the largest meta-analysis ever conducted on this question suggest that the risk profile around that number is considerably flatter than decades of public health messaging have implied, and that a BMI of 27 carries a fundamentally different risk profile than a BMI of 40.
Focus on metabolic health markers. Blood pressure, fasting glucose, lipid panels, and waist circumference reveal more about cardiometabolic risk than BMI alone. Ask your doctor to discuss these alongside the scale.
Maintain muscle mass. Some of the "protective" effect of overweight may reflect lean mass rather than fat, and resistance training preserves muscle as you age while independently reducing mortality risk regardless of your BMI category.
Don't gain weight on purpose. The association between overweight and lower mortality is observational, not prescriptive, and no controlled trial has shown that intentional weight gain improves health outcomes.
Contextualize medical advice. If your doctor flags your BMI, ask: "What specific risk factors am I showing?" The answer matters more than which side of an arbitrary line you fall on.