A Finding That Sounds Like Good News
Plot mortality against BMI in a large population study and you rarely get a straight line. You get a U — or more often a J. Risk is elevated at low BMI, drops to a minimum somewhere in the middle, and climbs again at high BMI.
The awkward part is where the bottom of that U sits. In a number of large analyses, the lowest observed mortality has fallen not in the "normal" band but slightly above it, in what the standard categories call overweight. Findings like this got labelled the obesity paradox, and they've been used to argue that BMI thresholds are set too low, or that carrying extra weight is protective.
The reality is more interesting than either the headline or the dismissal. The curve is real. The interpretation is where things go wrong.
What the Curve Actually Shows
First, be clear what these studies measure. They're overwhelmingly observational: take a large cohort, record BMI at one point, follow them for years, count deaths. That design tells you about association. It cannot, on its own, tell you what would happen if a given person's weight changed.
Second, the U shape has two arms, and they have completely different explanations.
The left arm — elevated risk at low BMI. This is the part almost everyone misreads. Being underweight is genuinely associated with higher mortality, but a large share of the signal comes from people who are thin because they're ill, not ill because they're thin.
The right arm — elevated risk at high BMI. Well established, and consistent across study designs including those less vulnerable to the confounders below.
The debate is almost entirely about how much of the left arm and the flat middle is artefact.
Three Confounders That Bend the Curve
Reverse causation
Serious illness causes weight loss, often years before diagnosis. Cancer, COPD, heart failure, dementia — all of them tend to reduce body weight during their preclinical phase.
If you measure BMI once and follow people for five years, some of your "low BMI" group is people already carrying undiagnosed disease. They die at higher rates, and the low-BMI band looks dangerous. The weight loss was a symptom, and the study recorded it as an exposure.
Researchers address this by excluding deaths in the first several years of follow-up, on the theory that anyone dying early was probably already sick at baseline. When studies apply longer exclusion windows, the left arm typically shrinks and the curve's minimum shifts downward.
Smoking
Smokers tend to weigh less than non-smokers and die at substantially higher rates. That combination alone can manufacture an apparent protective effect of higher weight, because the lean group is enriched with smokers.
Statistical adjustment for smoking helps but is imperfect — self-reported smoking status is crude, and "former smoker" covers everything from quit-last-year to quit-thirty-years-ago. Analyses restricted to lifelong never-smokers generally show a steeper, more monotonic relationship between BMI and mortality than analyses of the full population.
BMI can't distinguish tissue
This is BMI's structural limitation showing up in a new place. Two people at BMI 27 might be a sedentary person carrying substantial visceral fat and a physically active person carrying substantial muscle. Their metabolic risk profiles are not similar.
In older populations this cuts particularly hard. Muscle mass declines with age, and low muscle mass independently predicts poor outcomes. An older adult with a BMI of 22 might have less lean tissue than one at 26. Some of the apparent benefit of higher BMI in elderly cohorts is really the benefit of not having lost muscle.
Where the Paradox Genuinely Holds Up
Not every instance of the paradox dissolves under scrutiny. In certain specific clinical populations, higher BMI has been associated with better outcomes even after careful adjustment: patients with advanced heart failure, those on dialysis, some cancer cohorts, and older adults recovering from major surgery or acute illness.
The proposed mechanism is metabolic reserve. When the body is under sustained catabolic stress, having energy stores and muscle mass to draw on may improve survival odds. That's a plausible explanation for a sick population and a poor argument for a healthy one.
The distinction matters enormously and is routinely lost. "Higher BMI is associated with better survival in dialysis patients" and "carrying extra weight is good for you" are not the same claim, and only the first has decent support.
What This Means for Your Own Number
If you calculate your BMI and land at 26 or 27, the honest reading is not "studies say I'm in the safest band." It's closer to: BMI at this level, on its own, is a weak predictor of your individual risk, and you should look at other measures.
Here's how to use the tool sensibly:
- Enter your height and weight in metric or imperial units.
- Read the BMI value and the category it falls in.
- Treat the category as a rough population-level screening signal, not a diagnosis.
- Pair it with at least one measure of fat distribution and, ideally, with actual clinical markers.
Waist circumference and waist-to-height ratio capture something BMI cannot: where the fat sits. Visceral fat around the organs behaves very differently from subcutaneous fat, and abdominal measures track it far better than total weight does. A widely used rule of thumb is keeping waist circumference under half your height.
Beyond that, blood pressure, fasting glucose or HbA1c, and a lipid panel tell you more about metabolic health than any anthropometric measure. Cardiorespiratory fitness is another strong independent predictor — and someone can improve it substantially without the scale moving.
Reading Studies Without Getting Fooled
A few questions to ask whenever you see a BMI-and-mortality headline:
Who was in the cohort? General population, or people already diagnosed with a condition? The paradox behaves very differently in each.
Did they exclude early deaths? If not, reverse causation is unaddressed and the low-BMI arm is inflated.
How was smoking handled? Never-smoker subgroups are the cleanest comparison available in observational data.
How old were participants? Age changes body composition enough that the same BMI means different things at 35 and 75.
What was the outcome? All-cause mortality, cardiovascular events, and quality-adjusted life years can point in different directions. Survival isn't the only thing worth measuring — years lived with mobility limitations or joint disease don't show up in a mortality curve at all.
FAQ
Does the obesity paradox mean BMI thresholds should be raised? Most researchers say no for the general population. The current thresholds are imperfect, but the arguments for raising them lean heavily on studies with unresolved confounding.
Why does the curve differ for older adults? Body composition changes with age, and low muscle mass independently raises risk. The optimum band tends to shift upward in elderly cohorts partly for that reason.
Is the left arm of the U entirely artefact? No. Genuinely low body weight carries real risks including reduced bone density and poorer resilience to illness. The debate is about magnitude, not existence.
Should I aim for a specific BMI number? BMI is a screening tool, not a target. Discuss individual goals with a healthcare professional who can factor in your body composition, fitness, medical history and lab results.
What's a better single measure than BMI? There isn't one measure that dominates. Waist-to-height ratio adds fat-distribution information cheaply, and combining it with BMI is more informative than either alone.
The Takeaway
The U-shaped curve is a real feature of observational data and a poor guide to individual decisions. Most of what makes the left arm rise is illness and smoking rather than thinness itself, and most of what makes higher BMI look protective in older cohorts is muscle mass that BMI can't see. The paradox is less a discovery about weight than a demonstration of how much a single crude number hides.
Calculate your BMI free at the BMI Calculator on sadiqbd.com — no sign-up, instant results. For interpretation of your own health, speak with a qualified healthcare professional.
