Evidence Against Excessive Cardio?

This is an old argument and is basically baloney. The ultimate takedown is here, written by an exercise physiologist for Runner's World:

http://www.runnersworld.com/health/too-much-running-myth-rises-again

Money quote:

"But here, from the actual abstract, is the part they never mention:

"Cox regression was used to quantify the association between running and mortality after adjusting for baseline age, sex, examination year, body mass index, current smoking, heavy alcohol drinking, hypertension, hypercholesterolemia, parental CVD, and levels of other physical activities.

"What this means is that they used statistical methods to effectively “equalize” everyone’s weight, blood pressure, cholesterol, and so on. But this is absurd when you think about it. Why do we think running is good for health? In part because it plays a role in reducing weight, blood pressure, cholesterol, and so on (for more details on how this distorts the results, including evidence from other studies on how these statistical tricks hide real health benefits from much higher amounts of running, see my earlier blog entry). They’re effectively saying, 'If we ignore the known health benefits of greater amounts of aerobic exercise, then greater amounts of aerobic exercise don’t have any health benefits.'"

I am a long distance runner. Love it and won't stop!

It's funny how there are such distinct lines drawn between cardio lovers and lifting lovers... I have heard both ends of it and decided to do my own thing. I eat a lot of protein like a lifter, but focus mostly on cardio and less on weight training like a runner. I think both ends of the argument would just tell me I'm all wrong

This is an old argument and is basically baloney. The ultimate takedown is here, written by an exercise physiologist for Runner's World:

http://www.runnersworld.com/health/too-much-running-myth-rises-again

Money quote:

"But here, from the actual abstract, is the part they never mention:

"Cox regression was used to quantify the association between running and mortality after adjusting for baseline age, sex, examination year, body mass index, current smoking, heavy alcohol drinking, hypertension, hypercholesterolemia, parental CVD, and levels of other physical activities.

"What this means is that they used statistical methods to effectively “equalize” everyone’s weight, blood pressure, cholesterol, and so on. But this is absurd when you think about it. Why do we think running is good for health? In part because it plays a role in reducing weight, blood pressure, cholesterol, and so on (for more details on how this distorts the results, including evidence from other studies on how these statistical tricks hide real health benefits from much higher amounts of running, see my earlier blog entry). They’re effectively saying, 'If we ignore the known health benefits of greater amounts of aerobic exercise, then greater amounts of aerobic exercise don’t have any health benefits.'"

I think either you or the author of that Runner's World article are misunderstanding what a statistical regression does. A multivariate regression models the variance in the data with respect to the dependent variable (here, mortality rates). Let me break this down.

Suppose you want to know whether mortality rates are affected in any way by running. You happen to know from prior research that "baseline age, sex, examination year, body mass index, current smoking, heavy alcohol drinking, hypertension, hypercholesterolemia, parental CVD, and levels of other physical activities" (the other predictor variables mentioned in the abstract) contribute to mortality. There are two ways to answer your main research question: (1) match for all of these variables in the two populations you are studying, one of which runs and the other does not. This is going to be hard to impossible, given the number and nature of these variables. (2) Find out what the levels of these variables are in the population, and include them as predictors in your model. If, after including them as predictors, marathon running STILL accounts for some variation, then marathon running has an effect on mortality rates.

You might also be a bit confused about what it means when your predictor variables are correlated with each other. Runners may in fact be overall eat better, drink less, etc., etc. But they might not be. If you want to know the effect of running independent of the variables which are correlated with running, you have to regress running against these other variables and take the residuals and put them in your model as predictors.

What the study in question found was this (quote from the abstract):

"Running distances of 0.1-19.9 miles/week, speeds of 6-7 miles/hour, or frequencies of 2-5 days/week were associated with a lower risk of all-cause mortality, whereas higher mileage, faster paces, and more frequent running were not associated with better survival."

What this means in plain English is that the study found no benefits to running more than 20 miles a week. It doesn't mean that there are no benefits, and it certainly does not mean that there are dangers.

"To reiterate, I'm not flipping to the other extreme and arguing that there's no point of the diminishing returns for exercise, or even that there's no possibility of heart damage associated with extreme ultraendurance exercise. These are open and legitimate questions. But this scaremongering about relatively modest amounts of exercise in favor of "hunter-gatherer" exercise is silly. We can speculate all we want about "potential" risks and benefits, but the real-world epidemiology is crystal-clear: if you exercise for an hour a day, you're likely to live longer than if you exercise less than an hour a day."

There seems to be a bit too much positioning here from both sides and not enough good information.

I think you're missing the point of this a bit. What Hutchinson is saying is that obviously there is some point at which cardio becomes "too much," but that level almost certainly quite high. If I run an hour a day, that's 50 miles a week -- that's a lot of running! And that's much better than not running at all. If people are doing more than that -- 70, 100+ miles a week -- they are generally not doing it to optimize health, they're doing it because they love running and want to get better at it. Much like skydivers and rockclimbers don't do it for their health. True, we don't know the point at which cardio becomes "too much," but the warnings that 20-30 miles a week is the threshold are very likely wrong.

Make sure you read this post, where Hutchinson gives more information about the research behind the 30-mile myth:

http://www.runnersworld.com/running-tips/second-thoughts-too-much-running

I think either you or the author of that Runner's World article are misunderstanding what a statistical regression does. A multivariate regression models the variance in the data with respect to the dependent variable (here, mortality rates). Let me break this down.

Suppose you want to know whether mortality rates are affected in any way by running. You happen to know from prior research that "baseline age, sex, examination year, body mass index, current smoking, heavy alcohol drinking, hypertension, hypercholesterolemia, parental CVD, and levels of other physical activities" (the other predictor variables mentioned in the abstract) contribute to mortality. There are two ways to answer your main research question: (1) match for all of these variables in the two populations you are studying, one of which runs and the other does not. This is going to be hard to impossible, given the number and nature of these variables. (2) Find out what the levels of these variables are in the population, and include them as predictors in your model. If, after including them as predictors, marathon running STILL accounts for some variation, then marathon running has an effect on mortality rates.

I can certainly understand why you would want to try to control as many unrelated factors as possible, but it's disingenuous to assume that factors such as BMI and blood pressure are unrelated to running. If a person is running 60 miles a week and still has high BMI and hypertension, I'm not surprised that they have high mortality -- but I'm not going to say that's due to the running!

What Hutchinson and I are saying is that it's very, very difficult to determine conclusively where the point of diminishing returns of cardio is, but ignoring positive effects of cardio such as lower BMI and lower blood pressure when measuring the returns of cardio is probably not going to help much in answering the question.

I have SO MANY studies...each conflict with others. So I decided that everyone should just do the workout that makes you the happiest.

If you love runners high...run
If you love your big beefy muscles...lift
If you want like both...do both.
If you love Yoga but hate everything else...Do yoga!

My point is that everybody pretty much agrees that some exercise (no matter what kind) is better then none.
So do what you love :-)

I personally hate running and there is a horrible family history of knees dying young, so I try to keep it under 45 minutes if I jog.

While research and studies are awesome and all that jazz.. I will just run whenever I feel like it. Running is addictive as is heavy lifting but I listen to my body. I think I've found the medium that works - it changes every week but I really don't listen to what the lifters or cardio bunnies say.. my doctor says I am healthy. I just do my own thing

I think either you or the author of that Runner's World article are misunderstanding what a statistical regression does. A multivariate regression models the variance in the data with respect to the dependent variable (here, mortality rates). Let me break this down.

Suppose you want to know whether mortality rates are affected in any way by running. You happen to know from prior research that "baseline age, sex, examination year, body mass index, current smoking, heavy alcohol drinking, hypertension, hypercholesterolemia, parental CVD, and levels of other physical activities" (the other predictor variables mentioned in the abstract) contribute to mortality. There are two ways to answer your main research question: (1) match for all of these variables in the two populations you are studying, one of which runs and the other does not. This is going to be hard to impossible, given the number and nature of these variables. (2) Find out what the levels of these variables are in the population, and include them as predictors in your model. If, after including them as predictors, marathon running STILL accounts for some variation, then marathon running has an effect on mortality rates.

I can certainly understand why you would want to try to control as many unrelated factors as possible, but it's disingenuous to assume that factors such as BMI and blood pressure are unrelated to running. If a person is running 60 miles a week and still has high BMI and hypertension, I'm not surprised that they have high mortality -- but I'm not going to say that's due to the running!

What Hutchinson and I are saying is that it's very, very difficult to determine conclusively where the point of diminishing returns of cardio is, but ignoring positive effects of cardio such as lower BMI and lower blood pressure when measuring the returns of cardio is probably not going to help much in answering the question.

I realize my post was in the TL;DR category, but if you read the second part of it, you'll see that I explain how statisticians deal with collinearity among predictor variables in regression analysis.