The U-shaped curve: BMI and healthy life expectancy

Body Mass Index is a crude measure — it doesn't distinguish muscle from fat, and it says nothing about where fat is distributed, which matters for metabolic risk. Despite its flaws, BMI remains one of the most extensively studied predictors of population-level mortality and morbidity, simply because it's cheap to measure and has been recorded consistently across millions of people over decades.

Why the relationship is U-shaped, not linear

Unlike smoking, where risk rises fairly steadily with exposure, BMI's relationship with mortality and healthy life expectancy traces a U-shape (some analyses find a shallow J-shape). Risk is lowest in a broad middle range — roughly BMI 22 to 25 in most population studies — and rises at both ends:

• Underweight (BMI below 18.5): associated with elevated mortality, driven partly by frailty, reduced physiological reserve, and in some cases underlying illness that caused the low weight in the first place rather than the weight itself.
• Overweight (25–30): a comparatively small risk increase in most studies — some analyses find little to no elevated risk in this range for otherwise healthy people.
• Obesity (30–35): a more clearly elevated risk, particularly for cardiovascular disease and type 2 diabetes.
• Severe obesity (above 35): the steepest portion of the curve, with substantially elevated risk across nearly every major cause of death.

This shape is why the Helixspan model doesn't simply reward "lower BMI" — a model that did would incorrectly score being underweight as good. Instead it centers the reward around the 22–25 range and reduces the score moving in either direction.

What BMI is a proxy for

The reason BMI predicts risk despite being a rough measure is that it correlates, at a population level, with things that more directly cause harm: visceral fat accumulation, insulin resistance, chronic inflammation, and cardiovascular strain at the high end; and muscle wasting, reduced bone density, and inadequate physiological reserve at the low end. Two people with the same BMI can have very different risk profiles depending on body composition and fitness — which is a real limitation of using BMI alone, in this model and in medicine generally.

The practical range to aim for

For most adults, the evidence supports aiming for the broad 20–25 range rather than chasing the exact center of it. The curve is relatively flat through the low-to-mid 20s, meaning the difference in risk between a BMI of 21 and 24 is small — the more meaningful risk changes happen at the tails, particularly moving into obesity or below the underweight threshold.