Over at the Examining Room of Dr. Charles, one of the newer ScienceBlogs, there’s a post reminding me that I want nothing to do with medical research. I mean, how do you sort out what’s a cause, and what’s an effect in data like this:
Another recent study, published in Diabetes Care this past March, looked at the relationship between hours slept and the risk of developing diabetes. Researchers followed 1,100 middle age men starting in 1987 up through 2004. The men who slept 6 hours or less per night had double the risk of developing diabetes as compared to those who slept 7-8 hours per night. The men who slept more than 8 hours per night had triple the risk.
I’d probably lean toward “effect” rather than cause, as I could easily imagine that the effects of diabetes would make it difficult for some people to sleep, and would make some other people sleep more. That seems more plausible than a couple of hours of sleep a night being enough of a stress to cause diabetes.
But how you would sort that out, I have no idea.
Sure, one of the symptoms of diabetes in frequent urination, so you’ve got a mechanism for less sleep right there. But we also know that sleep deprivation increases inflammatory mediators in the bloodstream, so it is possible that sleep deprivation could accelerate other pathological processes, such as the development of diabetes. It would not be a direct cause, but it could increase the risk.
I was thinking something similar just yesterday. The stress hormone cortisol appears to encourage the deposition of belly fat. The belly fat pattern seems to correlate with higher than usual incidence of heart attacks. But so does stress…..
MKK
actually it seems fairly causal. check out this related article for more information on the direct hormonal/metabolic effects of sleep deprivation:
We empirical economists have gotten used to arguments about what “causality” even means when applied to situations like this. I take a fairly hard line: X can’t be said to “cause” anything unless we have some way of manipulating X directly.
So I’d be willing to say that sleep deprivation can cause diabetes, because we could invite some participants into the lab and decide to randomly assign them to be deprived or not. (Though even then one could argue that its the particular method of accomplishing sleep deprivation that causes things–poking you with a stick when you get tired might have different effects than playing loud music.)
But the study Chad describes is not looking at sleep deprivation, but at sleep. I can’t think of any way that sleep could be manipulated without having its own effects–sleeping pills may prevent diabetes, but that isn’t the same thing as saying that sleep does.
I believe this is why doctors normally talk about Risk Factors for most things. They know that causation isn’t necessarily provable, but correlation is common.