But there is plenty of well-documented, conflicting evidence that running is linked to longer life expectancy. And isn't obesity killing us? (Or maybe high BMI isn't so bad if you exercise, too.)
Is your head spinning yet?
Allow me to switch hats to my "day job" for a moment, wherein I get paid to play with data and write articles about correlation...
Believe none of the headlines.
Not the good stuff or the bad stuff.
And here's why...
Each of the headline-grabber articles shows correlation, not causality.
Correlation is a purely statistical relationship. For example: ice cream consumption and bicycle accidents rise and fall at about the same time of year.
The best analysis will try to show both correlation and causality. Unfortunately, in the messy realm of real-world statistics, that's often easier said than done.
To be fair, the 2008 study did attempt to identify exercise as the cause for lower disability and higher life expectancy among runners by controlling for factors including age, sex, body mass index, smoking, and disability level. But maybe the investigators failed to capture some key explanatory variable like the "running buddy effect?" After all, research does show a positive correlation between social networks and health.
And in the new studies that show higher rates of heart malfunction among endurance athletes, maybe the analysis fails to account for those with extremely low body weight? Research shows that both "underweight" (low BMI) and obesity are correlated with higher mortality than "normal" weight. Or maybe ultra-runners are more likely to eat a certain type of food, or are more likely to go through bouts of dehydration...?
The point is that no study can control for all factors.
In fact, the American Journal of Cardiology article even admits to this weakness:
"Recent studies have shown that the prevalence of AF [atrial fibrillation] is higher in individuals who are involved in intense short-term training and long-term sports participation compared to general population of the same age although clear evidence about the causal relation between these conditions is lacking." (emphasis added)To look at the importance of understanding both correlation and causality, let's take a step back and consider alcohol.
We've all heard that moderate alcohol consumption is "good for our health," right?
Well... it's not that simple.
The alcohol/health correlation is based on the U-shaped "mortality risk" curve shown below. In this case, much like with golf scores, a lower number is better because it means less likelihood of dying at any given point in time.
Those who drink 0 drinks per day (left side of chart) have a higher mortality risk (i.e. are more likely to kick the bucket) than those who drink, say, 20 grams (slightly more than 1 drink) per day. After that first drink, risk of death starts going up. (By the far right hand side of the chart we're drinking ourselves into instant liver failure.)
So pour me a beer, right?
Not so fast.
We've forgotten about that little problem of causality.
When we stop to think about the "reasons" - the causal factors - that make those abstainers avoid alcohol, a few causes stand out:
- People who are non-drinkers might abstain because they have another health problem (think of prescription labels that warn about adverse drug interactions with alcohol).
- Non-drinkers may have been problem drinkers in the past (thus the damage may already be done).
- And last, but not least, people who are moderate drinkers probably are drinking with friends (again with that pesky "friend" benefit).
Moderate alcohol consumption isn't the cause of better health, it's a symptom of it!
When we look at statistics, we must keep in mind both the correlation and the causal factors. The statistical relationship is important, but the reason for the relationship is more important.
(You thought I was done? Well, almost...)
Even if there was a proven link between exercise and longer (or shorter) life there will still be a sizeable minority of the population that bucks the trend.
Correlation predicts average results.
Some of us are a-typical.
And (unfortunately) we have no idea who is typical and who is unique until we test the theory in that great "experiment of one."
The moral of this story:
Talk to your doctor.
Listen to your body.
Read the literature, but do so with a critical eye.
(And if you don't believe me, try Amby Burfoot and George Sheehan's take on the conflicting headlines.)