Few phenomena have created as divisive a rift recently among health professionals as the so-called “obesity paradox,” the repeated finding that obese people with certain health conditions live longer than slender people with the same ailments. And when a January meta-analysis involving nearly three million research subjects suggested that overweight people in the general population also live longer than their slimmer counterparts, the head of Harvard University’s nutrition department, Walter Willett, called the work “a pile of rubbish.” A few new studies suggest that these paradoxes may largely be artifacts of flawed research designs, but some experts disagree, accusing the new studies of being inaccurate. Among the biggest questions raised by this new research is the impact of age: whether obesity becomes more or less deadly as people get older and why.

The January meta-analysis, led by U.S. Centers for Disease Control and Prevention senior scientist Katherine Flegal, pooled data from 97 studies of the general global population and reported that, in sum, overweight individuals—those with a body mass index of 25 to 29.9—were 6 percent less likely to die over various short time periods than people of normal weight (with a BMI 18.5 to 24.9) were. For people over the age of 65, however, being overweight conferred a 10 percent survival advantage. Flegals' findings also suggest that obesity, which has always been considered a major health risk, is not always dangerous and that it becomes less so with age: Adults with grade 1 obesity (BMIs of 30 to 34.9), she found, were no more likely to die than were normal weight adults; for grade 2 obesity (BMI of 35 to 39.9), the increased death risk for adults of all ages was 29 percent, but restricting the analysis to adults over the age of 65, the increased death risk associated with grade 2 obesity was not statistically significant.. The older a person is, the analysis seemed to say, the safer extra pounds become.

Two new studies, however, challenge the notion that extra weight, whether a little or a lot, can be safe. They also suggest that age is closely related to weight-related health risk and that added pounds could be more, not less, dangerous in old age than in youth. In a March 2013 analysis published in the American Journal of Epidemiology and based on U.S. National Health Interview Survey data, Ryan Masters, a Robert Wood Johnson Foundation Health and Society Scholar at Columbia University, and his colleagues reported that grade 1 obesity (a BMI of 30 to 34.9) increases the death risk 45- to 54-year-old men by 10 percent but that among 75- to 84-year-olds, the increased death risk is 59 percent. (For women the percentages were even larger—15 percent and 72 percent, respectively.) Mortality risks for being overweight, they found, increased with age too. In a follow-up study published online last week in the American Journal of Public Health, Masters and his colleagues concluded that, overall, obesity was responsible for 18 percent of all deaths among 40- to 85-year-olds between 1986 and 2006—a figure that is more than four times higher than some previous estimates.

How can two groups looking at similar populations come to such different conclusions? In part, the discrepancy stems from how the researchers consider age. Although the CDC researchers separated out the impacts of obesity and overweight at younger and older ages, they didn’t account for potential age-related influences that, according to Masters, strongly affect the relationship between weight and mortality. One such factor is selection bias. When scientists invite individuals to participate in survey-based studies, those who sign up are usually the healthy ones. This effect is likely to be magnified in elderly populations and particularly in obese elderly populations, Masters says. The small percentage of obese elderly people who participate in studies are likely to be in much better condition than the ones who don’t, which can make it seem as if obesity is less dangerous in old age than it really is. When Masters and his colleagues controlled for this effect using statistical methods and then calculated the mortality risks associated with obesity in the U.S. National Health Interview Survey data, they found that obesity was far deadlier in old age than the CDC meta-analysis had suggested it to be.
Masters and his colleagues accounted for two other potentially important age-related effects as well. One is that people born in 1950 tend to live to older ages than people born in 1930. And the latter group tends to live to older ages than people born in 1910, because those born later spend more of their lives reaping the benefits of public health interventions such as sanitation, refrigeration and modern medicine. The second is that  people born later also spend more of their lives exposed to relatively new environmental influences that are thought to drive the obesity epidemic—ubiquitous processed foods and sugar-sweetened beverages, hormone-disrupting chemicals, and sedentary behaviors. “People who are born later get a greater duration of exposure to all those obesity-inducing environmental changes,” Masters says, and “this cohort-based exposure is the key component to trying to understand not only your likelihood of being obese in America, but the chance for obesity to negatively impact your health and thus impact your mortality risk.” Although grade 1 obesity accounted for 3.4 percent of deaths at age 66 among white men born from 1915 to 1919, it accounted for 5.8 percent of deaths at age 66 for white men born 20 years later, Masters found. Ultimately, according to the new studies, the research suggesting obesity to be “safe” in old age didn’t account for the fact that the seniors being followed had not been obese for very long; future seniors who gain significant weight at a younger age and keep it on will have a much greater mortality risk.

Willett—the Harvard nutrition professor who strongly opposes the notion of an obesity paradox—largely agrees with the new findings and the idea that birth period shapes obesity-related mortality risk. “This is likely to become even more important as the current cohort of adolescents and young adults gets older, as it is true that the adverse effects of overweight are cumulative over a lifetime,” he says. Willett thinks, however, that there are many other spurious causes for the obesity paradox, too, including the fact that as people get sicker, they often lose weight, a trend that associates slimness with poor health even though in these cases slimness hasn’t caused poor health. (Masters attempted to control for this so-called “reverse causation” by excluding people from his analysis who had a BMI under 18.5.) Other researchers try to account for this effect by excluding sick people from their analyses: In a study published online in August, for instance, Harvard nutrition researcher Chandra Jackson and her colleagues analyzed data from the general U.S. population and found that mortality risk appeared to decrease with increasing BMI among people with diagnosed diabetes; when they excluded those with diabetes, excess weight increased death risk. “It is imperative to address the methodological concerns that could lead to misleading conclusions,” says Jackson, who agrees that the obesity paradox is likely specious and that birth period could have important confounding effects, too.

Others are not convinced by the new findings. Flegal, the lead author of the CDC meta-analysis, says she does not think that her findings were biased by cohort effects, in part because many of the studies included in her analysis did not pool mortality estimates from wildly different age groups. She also says there’s little evidence that her findings were affected by selection bias. Neil Mehta, a professor of global health at Emory University who also studies the impact of obesity on mortality, says that the statistical method used by Masters and his colleagues “is questionable,” in part because it does not properly separate certain interrelated variables. The new findings ultimately inflate the risks of obesity in old age because older people, he says, suffer from many health problems and tend to die of a variety of causes. Yet Mehta agrees that “obesity duration matters a lot to mortality—the longer you are obese the higher the risk of dying,” a notion that his research supports, too.

Eric Reither, a sociologist at Utah State University and a co-author of the new studies with Masters, notes that their research is not necessarily at odds with Flegal’s work. “They also find that obesity contributes to premature mortality,” he says. “I would say that we build upon their research.” And although the new studies do little to settle the overall obesity paradox controversy, they do suggest that the impact of weight may be importantly dependent on age. “When people talk about the effect of [being] overweight or obesity on mortality risk, they often think it exists in a vacuum—that to be overweight at age 20 is to be overweight at age 60 is to be overweight at age 75,” Masters says. Instead of simply focusing on how age might influence the impact of obesity, he explains, scientists should consider that obesity might fundamentally shape the way people age—even if, right now, there’s no consensus on how it does so or even the nature of its influence.