STATS ARTICLES 2005
Will a Few Extra Pounds Lead to a Longer Life?
May 24, 2005
A mathematician explains why two recent studies on obesity could produce radically different numbers - and how the media ignored the real issue.
With America engaged in a protracted and costly war on obesity, a recent and shockingly counter-intuitive report from the Centers for Disease Control (CDC) was seized upon by the media as the medical equivalent of the Pentagon papers: Not only was obesity not as deadly as people had been led to believe, but being overweight actually extended your life, compared to people of normal weight.
Though these findings differed significantly from a previous CDC study, which attributed a much deadlier cost to being obese or overweight, most journalists decided that the second study was more authoritative (perhaps because it was more in line with their waistlines).
From the stately repose of the New York Times Op-ed page, John Tierney claimed the original CDC estimates were steeped in bias. “The supposedly deadly consequences of fat provided the scientific rationale for the last politically correct form of prejudice,” he wrote. And along with fellow columnist David Brooks and science writer Gina Kolata, Tierney touted the conclusion that normal people die at a higher rate than overweight people.
Here’s the problem: even if you discount the much higher mortality rates in the first study, the difference in mortality rates in the second study were not conclusive. In two of three data sets used, the results were not statistically significant.
NPR missed the controversy altogether, claiming that the CDC simply “revise(d) its death rate from obesity.” The CDC actually did no such thing; rather, they presented conflicting studies without concluding which is correct. And this is the key point that was missing from so much of the coverage: The scientific community and the CDC have not yet agreed on which study is better, and why.
The first study (co-authored by the director of the CDC, Julie Geberding) concluded that 400,000 deaths per year were attributable to being obese and overweight in the United States. The authors subsequently revised this number, citing mathematical and other careless errors, to about 365,000 deaths per year.
Controversial as the definition may be, normal weight is defined as having a Body Mass Index (BMI) between 18.5 and 24.9, overweight by BMI 25 to 29.9, and obese as BMI equal or over 30. (Your BMI is calculated by your weight in pounds multiplied by 703, divided by the square of your height in inches.) While BMI may incorrectly categorize a small percentage of people with a lot of muscle mass, such as athletes, it is embraced by the CDC as a reasonable way to judge the majority of the population, in part because height and weight measurements are easy to obtain and consistent (in contrast to body-fat percentage, for example).
The second study -- released last month, and authored by the Katherine Flegal, a Senior Research Epidemiologist at the CDC’s National Center for
Health Statistics -- concluded that only 26,000 deaths per year were attributable to obesity and being overweight, less than eight percent of the first estimate. Even more counter-intuitive was their observation that being overweight is associated with a lower death rate than normal weight.
The media glommed on to the second study as if it was a free pass to the cookie jar. Isn’t it just great that being overweight is actually good for you? We’re not all about to die from fat – in fact, a little too much is good for us!
But would the media’s response have been different if the first study had come out second? This is an important question, because the second study did not convincingly refute the first (nor vice-versa). The second study claimed to use more sophisticated statistics and more recent data to obtain its numbers. But by the very nature of “recent data,” they were unable to follow the long term effects of being overweight and obese on the recent data set. Furthermore, the Geberding study’s authors have not yet had time to rebut the criticism offered by the more recent work.
How can scientists come to such drastically different conclusions?
So which study should we believe? Unfortunately, scientists who study obesity are not in agreement about the “correct” method to employ when assessing deaths attributable to obesity and being overweight. As a consequence, even though the Flegal study was sponsored by the CDC, Geberding has refused to endorse its conclusions or publicly refute them.
The Causal Pathway
For one, being either obese or overweight does not kill you directly, the way a heart attack might. Obesity leads to other conditions, such as diabetes, cardiovascular disease, sleep apnea, hypertension and a variety of other problems that can subsequently lead to death.
Obese people have a higher rate of diabetes – and this higher rate can be measured. But should the deaths associated with a higher rate of diabetes be attributed to the obesity? It depends on your assumptions. If you believe that obesity causes diabetes, then it seems fair to attribute deaths caused by diabetes to obesity.
It is possible, however, that diabetes is not caused by obesity, but is just associated with it. Perhaps certain people are genetically disposed to both obesity and diabetes. Then diabetes would be considered a separate cause of death, and deaths associated with it should not be attributed to obesity. The same analysis can be done with cardiovascular disease, or any of the other complicating and life-threatening illnesses associated with overweight and obese.
The problems with assumptions about causality might be resolved by simply looking at the whole population and comparing death rates of obese (or overweight) people with those of normal weight people, without regard to cause. This is the approach taken by the Flegal study.
However, the science here is not so clear either. It is extremely important for any epidemiological study to adjust for confounding factors, or take into account bias that might be introduced inadvertently. Suppose that there are higher rates of smoking among overweight people than among obese people. Then the results about death rates could be misleading: smoking would make the rates of death higher among the overweight than it should be. For this reason, the data is adjusted to make the two groups (overweight and obese) have the same smoking rate. Smoking is considered a confounding factor.
But should smoking be considered a confounding factor? It depends on the causal assumptions. Perhaps smoking leads to decreased appetite, which is why overweight people smoke at a higher rate than obese people. Perhaps it leads to a more active social life, promoting weight loss, or preventing weight gain. Perhaps smoking changes the metabolic system so that the same caloric intake will not cause as much weight gain. If so, deaths due to smoking might counteract deaths due to obesity. And if you adjust for smoking as a confounding factor, you undermine the “true” data about deaths due to overweight.
Will an extra helping a day keep the doctor away?
Even if we were to completely understand which factors are causal and which factors are confounding, we would have a hard time understanding what to take out of these studies. The second study claims that overweight people die at a lower rate than people of normal weight (though this conclusion is only statistically significant for one of three data samples considered, so we should be careful not to over-egg this idea). Should we stop trying to shed the pounds, or even indulge a little more?
Hard to say. While being overweight is associated with a lower death rate, no cause and effect relationship has been established. It is possible that overweight people have more intervening medical attention than normal weight people, lowering the death rates. Examples of such intervention might include statins for high cholesterol, bypass surgery to remove heart stoppages or preventative screens for cancer. The general improvement in heart disease prevention through surgery and medication (as opposed to weight loss) might have greater impact on overweight people, as doctors concentrate their efforts on those they perceive to have high risk. Possibly being at a normal weight doesn’t flag a doctor’s intervention through prescriptions, preventative testing, and minor surgery meant to offset risk of cardiovascular death.
Both studies agree that obesity is associated with higher death rates. As for being overweight, top scientists at leading institutions across the country are in disagreement. It may not be wise to equate being overweight with increased longevity just because the media has plumped for one study over another.