You don't see many really old, obese people whereas you see a lot of old thin people. It is reasonable to assume, exceptions aside, that obesity kills. 

Unless you reach a certain age, it has been said. When it comes to seniors, research has reported an "obesity paradox" concluding that, at age 65 and older, having an elevated BMI won't shorten your lifespan, and may even extend it. A new study took another look at the numbers, finding the earlier research flawed. The paradox was a mirage: As obese Americans grow older, in fact, their risk of death climbs.

The researchers argue that past studies of longevity and obesity were biased due to limitations of the National Health Interview Survey, or NHIS, which provides information on obesity. The survey excludes individuals who are institutionalized, such as in a hospital or nursing home—a group largely made up of seniors. Consequently, the data is overrepresented by older respondents who are healthy, including the relatively healthy obese. What's more, many obese individuals fail to make it to age 65—and thus do not live long enough to participate in studies of older populations.

"Obesity wreaks so much havoc on one's long-term survival capacity that obese adults either don't live long enough to be included in the survey or they are institutionalized and therefore also excluded. In that sense, the survey data doesn't capture the population we're most interested in," says Dr.Ryan Masters of Columbia University's Mailman School of Public Health, the study's first author.

They matched NHIS data on obesity with corresponding records in the National Death Index using data from close to 800,000 adults surveyed between 1986 and 2004. Next, they performed statistical adjustments to account for the survey selection biases. The result: risk for death from obesity increases with age.

The finding supports studies that document how obesity takes a cumulative, even compounding toll on the human body. Exponents of the obesity paradox have explained their counterintuitive results by suggesting that obesity's extra padding protects seniors from fall-related injuries and provides energy reserves during illness.

"This study should put to rest the notion that it's possible to 'age out' of obesity risk, and provides a powerful counterfactual against those who say concern over obesity is overhyped," says co-author Dr. Bruce Link, a professor of Epidemiology and Sociomedical Sciences at the Mailman School.

Going forward, they have set their sights on another possible reason behind the obesity paradox; that some of the older obese captured in NHIS data only put on extra weight later in life. 

"The recent obesity epidemic hit all age groups at the same time, meaning many of the elderly obese only gained their excess weight in the last 10 years or so," Masters says. "To account for this fact, I will take a page from studies of cigarette smokers by looking at 'life years' to measure how long someone has been obese rather than whether or not they happen to be obese at the time of a single snapshot survey."

Published in the American Journal of Epidemiology.