Stokes analyzed national survey data in two ways: The standard method calculated death rates using BMI at time of survey. It found a slight benefit to being overweight.
His alternative method used BMI based on the maximum weight that participants reported in their lifetime. It found that being overweight (BMI 25 to 29.9) raised mortality risk 28 percent; lower-level obesity (BMI 30 to 34.9) increased it 67 percent; and higher-level obesity (BMI 35 and above) more than doubled it. Two-thirds of American adults are in the first two categories.
The study, published in Population Health Metrics, surveyed adults ages 50 to 84 who had never smoked. Smoking is so deadly that it would overpower any other finding; it also has odd effects on weight.
Stokes is confident of his method and his finding of increased death risk, but said that future research may refine the size of the risk.
It is far from the last word on the obesity paradox, and Stokes is far from alone in his belief that excess weight is riskier than is commonly thought. The issue can have obesity researchers sound like paleontologists fighting over the meaning of their dinosaur fossil finds.
Katherine Flegal, a senior scientist at the U.S. Centers for Disease Control and Prevention, has published several papers about her finding that being overweight reduces mortality risk, and said she sees no reason to reevaluate it.
Agency policy prevents her from commenting on other researchers' papers, she said, but she noted that she had controlled for sickness in her own research and found no effect.
A huge analysis she led of previous studies, in the Journal of the American Medical Association last year, confirmed her overall findings. It got a lot of news coverage - and sharp criticism from some others in the field.
"It's kind of funny," Flegal said. "Here are all these papers. All published in the peer-review literature. And then they get upset." (Critics say she left out some papers with different results.)
One of her sharpest detractors is Walter Willett, the nutrition department chair at the Harvard School of Public Health. Willett called Stokes' method of excluding sick people from the analysis "simple and intuitive."
He called Flegal's work "complete rubbish." Worse, he said, is the message that it sends.
If you are asking what is a healthy weight, he said, "you should not have that answer be influenced by data from someone who has lost 20 pounds and is dying of cancer."