Will a few extra pounds save you or kill you? Science can’t tell us, yet news stories project certainty

Carolina Branson, PhD is an associate editor with HealthNewsReview.org. She tweets as @BransonCarolina.

Doctor Weighing PatientTo diet, or not to diet? That is the question. And if you’re like some health care journalists who have covered the phenomenon called the “obesity paradox” over the last several years, you may have suggested to readers that a little extra fat is a good thing. Simply stated, the obesity paradox refers to some studies that have shown BMI or Body Mass Index that is slightly higher than the “normal” range is associated with lower mortality.

Health news consumers might have read such stories and concluded that a few extra pounds wouldn’t hurt them after all. Fast forward to today, however, mere months after a spate of stories suggested that “Obesity may not take years off your life,” and a new batch of stories based on research from The Lancet are swinging the other way again. They report that being overweight (defined as having a BMI over 25 and less than 30) increases mortality substantially and that “fat kills.”

So which study and which stories are accurate? What are health consumers to think? And should this information guide people’s decision to diet or attempt to lose weight? The answers, unfortunately, are not clearly explained by some news outlets because they don’t provide a comparative look at what recent studies found and how to make sense of them. As we’ve pointed out before the caveats in the obesity paradox stories have not been adequately explored, nor are they addressed in the new round of “fat kills” stories that we reviewed.

Key question: Why is this study different from previous research?

Let’s start by clarifying that this isn’t really an “obesity” paradox at all but rather a “mildly overweight paradox.” Obesity is still clearly linked to health risks and so there’s no paradox about it.

This new Lancet study tries to move us forward by addressing certain methodological problems in past research. For example, previous studies didn’t exclude people who had chronic diseases that may have made them lose weight and increased their risk of dying. Including such people would obviously skew any study attempting to link body weight to risk of death. The new study does exclude people with existing chronic conditions and also those who died within five years of the start of the study.

However, Andrew Stokes, PhD, Assistant Professor of Global Health at Boston University, doesn’t believe taking these measures is enough to eliminate problems from the analysis. One big reason is that the researchers failed to take into account a patient’s past weight history.

“The problem is most clear through an analogy to smoking,” Stokes says. “Imagine using a snapshot of smoking behavior to study the effects of smoking on mortality. This approach would group never smokers and former smokers into the same category, although their risks of death are very different. Similarly, people who were formerly obese and then lost weight should not be pooled with people who remained non-obese throughout their lifetime. Evidence from the literature suggests this is a major problem for several reasons: first, the effects of past obesity may persist into the present even after weight is lost; second, some people who have experienced weight loss have done so as a result of a health condition, a source of bias commonly referred to as reverse causality.”

Although the authors of the new study attempted to account for reverse causality by excluding subjects with chronic diseases, Stokes says this isn’t sufficient because “it doesn’t capture undiagnosed or subclinical illnesses.” Moreover, excluding such subjects eliminates a large amount of data, “thereby reducing power and generalizability,” Stokes adds.

Reporting a certainty that doesn’t exist

Like the obesity paradox stories, the “fat kills” stories are based on observational studies and suffer from the shortcomings mentioned above. Yet the conclusions have been inappropriately framed as definitive by some news outlets. Take this quote in a Medical Daily story from Dr. Richard Peto, professor of medical statistics and epidemiology at the University of Oxford: “If you could lose about 10 percent of your weight, a woman would knock 10 percent off the risk of dying before she was 70, and for a man it would knock about 20 percent off.”

That’s a pretty bold claim, but since observational studies such as this one can’t prove cause and effect, it’s simply not accurate to suggest that losing 10 percent of your weight will “knock 10 percent off the risk of dying.” There’s no way these numbers can be translated into straight percentages that apply to individual people.

Similarly, an article from Time headlined “How Body Fat Can Shorten Your Life” paraphrases one of the co-authors, Dr. Shilpa Bhupathiraju, suggesting that the results “should put to rest confusion over whether overweight can lower risk of premature death.” Really? Had the story tapped Stokes, or indeed any independent expert who was not connected to the study, it might have left some room for doubt as to whether the question is finally settled.

A story from Reuters headlined “Study finds fat kills, casting doubt on ‘obesity paradox'” also fails to quote any independent expert or discuss any potential limitations of the new study. But its headline will certainly grab attention, as will its claim that every 5 units higher BMI above 25 kg/m2 is “linked to a 31 percent higher risk of premature death.”

Academic reputations at stake

These stories also don’t delve into the academic sniping that’s occurred in this field and how that might affect the framing of the evidence. Dr. Walter Willett, the eminent Harvard nutrition expert and co-author on the most recent study, reacted angrily to a 2013 study from the CDC that appeared to support the obesity paradox. He called the study “a pile of rubbish” and said “no one should waste their time reading it.”

Considering Willett’s firmly entrenched position, is it any surprise that his colleagues at Harvard, including Dr. Bhupathiraju who was quoted in TIME, would aggressively assert that their findings are now the last word on the subject? And in the interest of balance, wouldn’t it have been a good idea for journalists to seek out the opinion of one of the CDC authors whose results cut the other way on this question?

Those opinions weren’t hard to find; they were published in an editorial accompanying the Lancet paper. And not surprisingly, they suggest that conclusions from this type of study can never be free from bias. And yet that editorial, unfortunately, wasn’t referenced or discussed by any of the news stories we saw apart from a piece at MedPageToday. Regardless of whether being overweight limits or extends life, or has no substantial bearing on it, it is the journalist’s responsibility to report the limitations in any study so that they provide a thorough and balanced perspective to readers.

You might also like

Comments (2)

Please note, comments are no longer published through this website. All previously made comments are still archived and available for viewing through select posts.

Maggie Herbert

July 25, 2016 at 10:08 am

Since many of these articles are used to ‘inspire’ the overweight to lose weight, it seems like it would be most accurate to compare overweight who successfully reach ‘normal’ BMI with overweight who don’t diet at all. Oops, that’s an issue because there aren’t very many overweight people who can actually lose the excess weight and keep it off, around 95% regain the weight, most end up heavier! So, I guess we also need to compare the longevity of those overweight who diet and regain (most of us!) with those who don’t diet….

Tim Olds

July 29, 2016 at 5:22 pm

The key thing abut The Lancet meta-analysis is that it excluded people who have ever smoked. Along with this exclusion criterion and the others (existing disease etc), about 60% of the potential pool of participants were excluded. Now we know from previous studies that (a) people with pre-existing diabetes are CVD are likely to live longer if they are fatter, and (b) when we exclude ever-smokers, the BMI associated with lowest mortality is shifted towards slower value. The latter effect is probably because people who have never smoked live unusual lives — they’re likely to exercise more, eat better, be wealthier and be leaner, so that their reduced mortality may have nothing to do with fatness. So the most we can say from The Lancet study is that if you have have never smoked, don’t have any chronic disease, and are not going to die in the next five years — and you don’t live in Australia — then it may be better to have a BMI of 24. Otherwise, it may be better to have a BMI of 27.