Be very wary of headlines like this. They draw the line at 6 hours, but that includes everyone, even people that only get 3 hours. What we want is the raw curves. If you get exactly 5 hours and 55 minutes a night, you might only be 1.5 times as likely as someone who gets 9 hours a night, while the people getting 4 hours a night are 15 times more likely.
What you want to see is the study, how the analysis variables were constructed, and what models were tested.
If a simple binary "subject slept 6 hours or fewer" variable is constructed, and there's no other discriminant, then if there is a significant positive correlation, it will show.
Testing multiple conditions, exclusively (3 or fewer, 4 or fewer, 5 ..., etc.) would get around this.
By way of comparison: I've used the Harris-Benedict basal metabolic rate formula to estimate caloric expenditure by individuals. The question came up: how well does this scale to extreme weights. It turns out that the original study is online as a PDF, including full stats on all the study subjects. My surprises were that 1) the formula is a lot older than I'd imagined (dating to 1919), involved a relatively small study size (a few score participants), and the weight range did not include high-end extremes unfortunately rather too common in current populations. I've taken to treating the values generated for extremely overweight individuals with a bit more suspicion as a result.
If one sleeps less than 6 hours a night, maybe there's other reasons for that, that lead to stroke, like anxiety over being deep in debt. So reducing such risk of stroke may not be as simple as getting more sleep.
Reminds me of the study that concluded that people living together (vs. married) leads to ~30% greater chance of alcholism. Maybe alcoholics just aren't as likely to get married?
Surely there's a Wiki on this type of logical fallacy somewhere. Anyone know what it's called?
Great, I'll know now. I see a lot of such conclusions in major media. Wish they'd point out the fallacy, but that probably wouldn't sell the news as well!
Love the example in the 2nd link: 'One class of examples is sometimes called the "Rooster Syndrome": "believing that the rooster’s crowing causes the sun to rise".'
Sleeping very little -indicates-> Stress and Anxiety -builds to-> Hypertension -leads to-> Stroke
Major disturbances in sleeping habits - Insomnia or Excessive Sleeping - are themselves strongly indicative of Depression, another major marker in terms of the Anxiety->Hypertension->Stroke vector.
So, to conclude: Yes, sleeping very little is probably correlated with health problems, but mostly because it indicates something else being very wrong.
Is the effect reversible? How long does it take for a persons stroke risk to go back to normal if they start sleeping more? What if I get 5 hours of sleep five nights a week, but 10 hours each of the other two nights?
It does depend on what you need, but in general, not sure many people recommend to sleep less. I thought the world suffers from overload. The number of hours we sleep has fallen.
I think the best sleeping habits reflect those of Churchill and humans before the 19th century: sleep 6 hours and then sleep 2 to 4 in the middle of the day.
Doesn't appear to be published yet, just presented at the SLEEP2012 conference.
Ruiter M, Howard VJ, Letter AJ, Kleindorfer D. Short sleep predicts stroke symptoms in persons of normal weight. SLEEP 2012; June 11, 2012; Boston, MA. Abstract 0829.
