The relationship isn't linear, you can see in https://bjsm.bmj.com/content/54/20/1195#sec-16 that above 2200 MET-minutes/week there is increased cancer risk which outweighs the cardiovascular benefits. Now admittedly it is a very weak correlation, and MET measurement is imprecise, but I think it is safe to say that there are risks associated with over-exercising and while being in shape is good, being a ultra-marathon runner or other extreme fitness enthusiast is unhealthy long-term, and the risks start sooner than you might expect. (2200 MET-minutes/week is about 3 hours of 8mph running per week)
I remember reading somewhere that, to maximize your health, run as much as you can each week, you won't hit the upper bound when it gets bad.
In the example you cited, that's 36 hours (!) of runnning in a week: meaning I'd have to run 6 hours a day and only take Sunday off each week before I was officially running too much. And I've noticed I can't run more than 4 days a week anyway, so even that's off limits. 6 hours a day, 6 days a week, even under optimal conditions, wouldn't be something I'd approach even if I was retired.
So practically speaking, "run as much as you can before your body aches stops you" seems like a good rule of thumb.
I'm not sure what math you're using. 12.9 MET for running * 3 hours * 60 minutes/hour is 2322 MET-minutes, so just 3 hours of running is sufficient to exceed the 2200 threshold. 36 hours of running would be 27864 MET-minutes, of course it is too much. But 3 hours of running in a week is quite easy to exceed, e.g. if you run half an hour every morning.
There are some studies which suggest that the artery buildup from frequent running is not bad, but these don't take cancer or other factors into account - in contrast all-cause mortality is one of the most comprehensive statistics available. The drawback is that although it gives you a number, that number might be affected by unknown variables.
MET-minutes seem to be "Metabolic equivalent of task" minutes [1]. In the paper, they give running an MET of 8 (paragraph 3 of Data Analysis section), so 2200 MET-minutes of running would be 2200/8 = 275 minutes, if I understand correctly.
I'd disagree slightly with your characterization of the author's findings on over-exercising; quoting from the discussion section:
"Our findings do not provide evidence for increased mortality risk with physical activity amounts as high as seven times above the current recommended target range. All cause, CVD and CHD mortality risks were lower at physical activity levels up to approximately 5–7 times the recommended level but the additional reduction in risk of mortality with engagement in activity at levels beyond the recommendations was modest and with increasing uncertainty, as reflected by the wide confidence intervals. Thus based on all of the available studies, we did not identify a higher mortality risk at any level of physical activity above the recommended level, although the lowest point estimate for all cause mortality was approximately 2000 MET min. Our analyses suggest that 10–12 hours of weekly vigorous physical activity cannot be considered harmful to longevity."
My reading is that the authors did not draw any conclusions about increased risks at higher MET levels due to the broad confidence intervals caused by the cancer risk data.
Right. I think that statement is based on Supplement eTable 5, where they analyzed numbers up to 6,000 MET-minutes/week.
Using confidence intervals in this case is a bit misleading though. What would be most useful IMO would be the prediction interval for the lowest point estimate, as that could be used to determine if individual activity amounts are appropriate.
If the authors don't draw conclusions from that, I wouldn't either.
For example, suppose the main impact of exercise on cancer risk were simply the correlation with time spent outdoors in the sun. Then you would expect to see a chart like this. But a conclusion like 'limit running to 3 hours/week, while walking can be up to 12 hours/week' would be totally wrong.
They do draw that conclusion - "the lowest point estimate for all cause mortality was approximately 2000 MET min". What their analysis did not identify is what level of exercise beyond that results in a significant increase in mortality risk. If you look at the individual studies that have categories above 2200, (Slattery 1989, Arraiz 1992, Besson 2008, Kvaavik 2010, Paganini-Hill 2011, Brown 2012, Arem 2015, Huerta 2016, Klenk 2016, Celis-Morales 2017, Lear 2017, Zhou 2017, and Liu 2018), there is some variation - almost every study shows a U-shape, but the lowest point differs by gender, age, measurement method, BMI, and so on. And the effects of too much exercise are small compared to the effects of too little, so it is not surprising that their basic univariate analysis with Stata glst is unable to resolve it. But given the number of U-shape studies I think it is reasonable to conclude that some amount is too much - in the absence of reliable statistics I would say to trust one's body and if it feels like you are pushing yourself hard every week then cut back.
> 2200 MET-minutes/week is about 3 hours of 8mph running per week
Wow, that is a lot of exercise for the average person. That would be 30 mins per day, six days a week. Again, I don't think 99% of people can sustain this for a more than a few months.
