Now, a new study led by researchers at the University of Bristol in the UK is the latest attempt to show why those kinds of conclusions are flawed, and it all comes down to misreading the data we've been getting.
"This study demonstrates that correlation is not causation and that when it comes to public health recommendations we need to be cautious interpreting data based on associations alone," explains statistical epidemiologist David Carslake.
"We found that previous studies have underestimated the impact of being overweight on mortality and our findings support current advice to maintain a BMI of between 18.5 and 25."
BMI divides an adult's weight in kilograms by their height in metres squared - the idea is to approximate whether your weight is healthy.
While the system has a lot of shortcomings, generally speaking it comes to this: a BMI of under 18.5 is taken to mean you're underweight; between 18.5 and 24.9 means you're healthy; between 25 and 29.9 means you're in the overweight range; and between 30 and 39.9 means you're obese.
While doctors recommend that most of us try to stick in between 18.5 and 24.9, research has shown that people with an overweight BMI actually live longer, leading some to conclude that being "overweight" is the "healthiest and most 'normal' weight of all".
But Carslake and his team say that previous research insisting you can be 'fat but fit' has misread something vital in the data.
Basically, in the large studies so far conducted on BMI and mortality, people falling in the healthy range BMI aren't necessarily healthy – and could have had their weight reduced by things like early stages of illness or health-damaging behaviour such as cigarette smoking.
"One of the main arguments against this association is that ill health can result in weight loss," explains endocrinologist Nick Finer from University College London, who wasn't involved with the study.
"[S]o favouring those who have not lost weight and may be modestly overweight [results in] so-called reverse causation."
In other words, to account for the fact that some people who are thin may in fact be unhealthily thin due to diseases and other things that may ultimately kill them, you need to look beyond their individual BMI.
But not too far beyond: what Carslake and fellow researchers did was compare the BMI and mortality of parents with the BMI of their adult children – analysing the health records of approximately 30,000 mother-and-child pairs and 30,000 father-and-child pairs contained in a large longitudinal study from Norway.
But why children, you might wonder?
"There is known to be a strong correlation of body weight and BMI between parents and their offspring due to heredity and a shared environment," explains Finer, "so using the child's BMI as a proxy for their parents' BMI can be used as an instrument unbiased by reverse causality."
In short, the children's BMI acts as a kind of guide for what the parent's BMI might be, but without being subject to things like individual illnesses that could bear on the parent's own mortality.
When they looked at the data that way, the apparent harmful effects of lower BMIs seen in other conventional studies were reduced, while the harmful effects of higher BMIs were greater.
Of course, from one perspective, it's another far-reaching statistical analysis – and one that relies on some fairly broad assumptions to make its point – so we shouldn't necessarily assume this new study has all the answers.
But at least it gives us another new insight into how the link between being overweight and longevity could be possible at all, and reminds us we always need to be careful in terms of interpreting associations like this.
"We are used to seeing conflicting studies purporting to show that something is either good or bad for our health," says one of the team, clinical epidemiologist George Davey Smith.
"More robust approaches for identifying the causal effects of factors influencing health [are required] if we are to make recommendations for public health based on reliable evidence."
More about: #Weight
