I want to point out this interesting analysis in Science of a counterintuitive trend noticed in the last few years, the fact that BMI is not a foolproof predictor of mortality and that being overweight can actually be better for you. The discussion illuminates the complex relationship between body weight and fat, demonstrates the limitation of using BMI as a health metric and underscores the appreciation that "fat", far from being a monolithic entity, is a complex and multifaceted beast whose exact nature, location and time-dependence can be far more consequential for mortality and health than simply its amount.

As the illustration above shows, while obese and unhealthy people suffer from the highest mortality, people with normal BMI can also be quite unhealthy and be near the upper or mid range of the mortality scale. Conversely, an obese person can be metabolically healthy. So why is this? The short answer is that for high-BMI individuals, the right fat in the right location might provide some benefits, like soaking up toxins or being a source of energy. In case of people with normal BMI it gets even more interesting; they often suffer from a poor nutritional and metabolic status in spite of their favorable BMI profile, and this can lead to worse mortality and health. From the article:

Although it is widely used, the BMI does not accurately measure fat content, reflect the proportions of muscle and fat, or account for sex and racial differences in fat content and distribution of intra-abdominal (visceral) and subcutaneous fat (11). Indeed, the body shape index (ABSI), a new index that quantifies abdominal adiposity relative to BMI and height is thought to be a better predictor of mortality than BMI (11). Excessive visceral fat in obesity predisposes to the “metabolic syndrome,” associated with insulin resistance, diabetes, hyperlipidemia, and cardiovascular diseases (12) (see the figure). In contrast, massive fat storage in peripheral adipose tissue has been shown to be metabolically inert in certain mouse models (12).

Likewise, it is possible that subcutaneous depots provide a safe harbor for potentially toxic lipids in obese individuals, thereby improving metabolic and cardiovascular health (12). The latter scenario may occur in some obese individuals with a healthy metabolic status, associated with a preponderance of subcutaneous fat, normal insulin sensitivity, absence of diabetes, and reduced risk of cardiovascular diseases (13). It is also possible that adipose tissue provides crucial energy reserves to meet metabolic demands during chronic illness, potentially decreasing mortality in obese patients. It must also be considered whether health care providers have increasingly adopted aggressive diagnostic and treatment strategies such as diet and exercise for obese diabetic patients, leading to better health outcomes and reduction in mortality.

How can a normal BMI be deleterious to health? Humans with genetic or acquired defects that prevent fat storage in adipose tissue are thin and yet develop severe fatty liver, insulin resistance, and diabetes (12). Furthermore, it is estimated that about 24% of adults in the United States with normal BMI have unhealthy metabolic profiles, even in the absence of major intercurrent illness (13). This “metabolically unhealthy/normal BMI” phenotype manifested by hyperinsulinemia, insulin resistance, hyperlipidemia, and increased risk of cardiovascular diseases is of greater concern for Asians, who have increased body fat at normal BMI values and are highly susceptible to developing diabetes (14). A low BMI may mask poor nutritional status and fail to detect crucial differences in fat and skeletal muscle content. Because skeletal muscle accounts for the majority of glucose disposal, loss of skeletal muscle mass (sarcopenia) owing to aging or physical inactivity, despite a normal BMI, can impair insulin sensitivity and negatively affect cardiovascular health and mortality (15). Relative insulin deficiency or poor control of blood sugar in diabetes also lead to sarcopenia, visceral adiposity, oxidative stress, and inflammation. These, as well as other factors, may plausibly predispose to morbidity and mortality in individuals with apparently normal BMI.

To me, the practice of boiling down something as complicated as health or mortality to a single number like the BMI says a lot about the human desire to simplify and to use what's readily available rather than what's important. The belief again reminds you of the drunkard and his keys; BMI is readily measurable and it's what we know, so why not use it? The truth is of course more convoluted. True metrics of mortality will have to take into account not just variables like fat distribution but - as the graphic illustrates - other biochemical and physiological indicators like insulin sensitivity and inflammation. It's very much a holistic approach, something that medicine is increasingly appreciating in both diagnosis and treatment. As the authors conclude:

The optimal weight that is predictive of health status and mortality is likely to be dependent on age, sex, genetics, cardiometabolic fitness, pre-existing diseases, and other factors. To quote Galileo, “Measure what can be measured, and make measurable what cannot be measured.” Clearly, there is an urgent need for accurate, practical, and affordable tools for assessing body composition, adipose hormones, myokines, cytokines, and other biomarkers to serve as predictive tools for phenotyping obesity and related metabolic disorders and assessing the risk of mortality. Advances in these areas will allow the examination of biological mechanisms and provide insights into the causal role of obesity in health and disease.

Reference: "The Health Risk of Obesity", Ahima and Lazar, Science, 2013, 341, 856.