The BBC reports on new findings published in in Archives of Internal Medicine from a longitudinal study of running and health. In 1984, the authors recruited members of a national club of over-50 runners and a control group of similar non-runners. Based on health outcomes observed in the years since, the authors conclude that "Vigorous exercise (running) at middle and older ages is associated with reduced disability in later life and a notable survival advantage." And -- more good news for runners -- the runners did not disproportionately suffer from knee and ankle injuries.
With studies like this, of course, I approach the paper wondering how much of the observed difference is due to the "treatment" (running) and how much is due to other differences between the treatment and control groups. Here I think that unmeasured confounding is significant; I would guess that selection accounts for maybe half of the difference in outcomes between the runners and non-runners. People who are in a running club at age 55 are unusual for many reasons that are not easily observed and controlled for. For one thing, they've usually run for a long time, meaning that the study really compares lifelong exercise regimens and not just whether the subjects run late in life. More importantly, older runners tend to have unusually fortunate genetic inheritances; anyone who didn't would be unable or unwilling to keep up a running regimen at that age. In my experience dedicated older runners also tend to be people with a fierce determination to conquer challenges and stick to a regimen. These are people who eat well and see their doctor and have friends and have all sorts of other health advantages. It's not surprising that people with such genetic and lifestyle advantages live longer, but attributing it to their running -- ie suggesting that anyone could start running and have similar outcomes -- would be to overlook a lot of these confounding factors.
The authors of the study recognize some of the selection problems they face, and they do what they can to address them, conditional on the study design. I was reminded in reading the paper that public health and medical researchers tend to be considerably more diligent than researchers in other social sciences in choosing words that differentiate association and causation. These distinctions tend to be lost in media coverage of the research but in this case the authors do a careful job of recognizing the obstacles to drawing inferences about the effects of running based on their study.
Posted by Andy Eggers at August 12, 2008 7:36 AM