Tuesday, June 17, 2003

By Ira Dreyfuss, The Associated Press

New research labels as an "illusion" a major study's conclusion that couch potatoes who take up at least moderate regular exercise can reduce their risk of dying early.

The apparent benefit "can be entirely attributed to measurement error," said researcher Paul T. Williams, a biostatistician in the Life Sciences Division of Lawrence Berkeley National Laboratory in Berkeley, Calif.

However, the lead scientist in the original study says additional data from the research project can prove him right. And other experts say that even if Williams' analysis is correct, other studies have shown so many health benefits from exercise that it must extend lifespan.

Williams examined a landmark study published in 1995 in the Journal of the American Medical Association by scientists at the Cooper Institute, a Dallas organization that studies exercise and lifestyle.

The Cooper Institute team looked at data on 9,777 men who had taken two treadmill exercise tests almost five years apart. The scientists then followed the men for more than five years. The researchers adjusted statistically for age and other risk factors, so they could focus on seeing if exercise affected the risk of death.

Men in the least fit 20 percent on both treadmill tests were most likely to die, the study found. However, men whose times had improved enough on the second test to pull themselves out of the least fit group had a lower risk of death.

Men who were unfit on the first test and fit on the second had a 44 percent reduction in their risk of death, compared with men who were unfit on both tests.

The researchers concluded that getting out of the least fit group could pull people out of the group at highest risk of early death. The finding is commonly cited to support current federal guidelines on physical activity, which call for doing at least 30 minutes of moderate activity on most, if not all, days of the week.

But Williams contends the researchers did not account sufficiently for the fact that the treadmill test is not a perfect measurement of physical ability. The test has a good day-bad day problem: A person might go longer in one test and shorter in another while having the same underlying fitness.

In his experiment, Williams ran numbers on a computer model. He created two hypothetical treadmill tests, and varied the scores according to his assumptions of measurement error.

His results were the same as the Cooper scientists described in the JAMA article, Williams said.

And if the article's results can be explained by measurement error, scientists must reject the conclusion that there were improvements due to physical activity, Williams said. "It hasn't been proved that changing to moderate exercise would affect your life expectancy," he said.

Williams believes physical activity can improve health, but that vigorous exercise is needed to change life expectancy. "If you are unfit and you become substantially fit, I believe that will change your life expectancy."

Williams' challenge is itself challenged by Steven N. Blair, president and CEO of the Cooper Institute, who led the JAMA article study team as a scientist, before his promotion.

Men whose fitness improved on the treadmill tests also reported a corresponding change in their physical activity, he said. Those self-reports are a sign that the improved life-spans were the result of improved living, not a glitch in the methodology, he said.

A colleague of Blair, digging deeper into the Cooper Institute data, said he was seeing physical changes which also argue that the lifespan improvements are real.

"The idea that people don't change on repeated measures is absolute nonsense. They do, and we've got the data to prove it," said Tony Jackson, a professor of health and human performance at the University of Houston.

At Stanford University, medical professor and physical activity researcher William Haskell felt Williams made a point in criticizing the design of the Cooper study. "It does raise an issue about how much weight we should be putting on those studies," he said.

Another institute board member, I-Min Lee of the Harvard School of Public Health, noted that some measurement error is a normal part of science. But, she said, other studies give lots of reasons to think that moderate exercise should improve health enough to reduce the risk of an early death.

"If [Blair's] study were the only one, I probably wouldn't put as much weight on it, but it is a piece of the puzzle that fits into the larger picture," Lee said.