The Boston Globe had an article on the front page of its magazine section last weekend: How much exercise is too much? Growing evidence shows that overdoing endurance training may damage your heart and shorten life expectancy. But don’t hang up your running shoes and wetsuits just yet.
The story cites two recently published studies and one review paper. Only one of the studies looks at mortality:
Some of his recently published papers reviewing the latest research suggest that regular marathon running increases the risks of an abnormal heartbeat, damage to heart tissue, and hardening of the arteries. Other research suggests that those who train hard every day don’t live as long as those who run at a more moderate pace a few days a week.
In a February study, Danish researchers followed nearly 1,900 runners for three decades and found that those who jogged slowly for up to 2½ hours a week lived about six years longer on average than those who ran longer and faster. Swedish researchers reported in June that elite cross-country skiers who had the fastest times in a 56-mile ski marathon or those who competed in the greatest number of those marathons were also 30 percent more likely than their fellow competitors to be hospitalized for an irregular heartbeat.
“When there’s enough smoke, there’s usually some fire,” Thompson said. “This may be a small fire, but I think most of us believe there’s cause for some concern.”
I had a vague memory of having looked at the Danish study when it came out and not having put up a blog post. But I couldn't remember what I had wanted to write. And I do have my wishful thinking to contend with, having just signed up for a 50 mile race and still hoping to put in more fast running once my recent running injury is further in the past. So I pulled an ungated version of the Swedish paper. And what do I find? This the Boston Globe version:
That doesn't look good for 'fast' or 'frequent' runners...but what is 'fast' or 'slow' --- for bad runners a 'fast' pace may be what a good runner considers 'slow'. Hey, my standard for fast is faster than 7:10 min/mile and I don't do most of my running at that pace, or even 10% or 5% (my injury may in part be due to trying to do one mile fast per day). So the vast majority of my running is slow, i.e. 8:30-10 min/mile. One the other hand, my 'slow' def. overlaps with my wife's 'fast'.
But what does the relevant part of the paper say? Here CI is 'confidence interval'.
Jogging pace and mortality
In a subanalysis of the fourth survey, the hazard ratios adjusted for sex and the confounders in model 2 were 0.37 (95% CI: 0.12, 1.17) for slow pace (178 joggers, 3 deaths), 0.53 (95% CI: 0.29, 0.95) for average pace (704 joggers, 12 deaths), and 1.22 (95% CI: 0.49, 3.04) for fast pace (201 joggers, 5 deaths), compared with those for nonjoggers. This analysis comprised only a few deaths among the joggers and should be met with caution, but the results suggest that a slow or average pace could be related to the lowest mortality.
Frequency of jogging and mortality
A subanalysis of the fourth survey yielded similar results for frequency of jogging with hazard ratios of 0.40 (95% CI: 0.15, 1.10) for ≤1 time per week (323 joggers, 4 deaths), 0.40 (95% CI: 0.16, 0.98) for 2–3 times per week (474 joggers, 5 deaths), and 1.24 (95% CI: 0.51, 3.02) for >3 times per week (84 joggers, 5 deaths), compared with values for nonjoggers. Hence, these data should also be interpreted with caution. A frequency of jogging of ≤3 times per week was associated with the lowest mortality, and we could find no increase in survival with >3 jogging sessions per week compared with nonjogging.
That's it. 178 slow joggers with 3 deaths, 201 fast joggers with 5 deaths. Reallocate 1 death and the difference between slow and fast jogging mortality hazards goes away...And they don't have enough 3+ times a week joggers to determine much of anything, even with 5 deaths. And I don't see any significance testing for trend effects, which is standard for this sort of paper, in jogging speed or frequency --- probably since it is obvious that they don't have any trend effects at any conventional level of signicance.
Come on. This sort of data shouldn't move your prior belief much.
The AF study also doesn't seem to show much...yes, endurance athletes get more AF and have heart remodelling, we already know this. The question is what effect that has. Some forms of AF are bad, but are these ones bad?
I do agree that jogging should be fun and not 'stress'. Beyond that, don't worry too much. If going 50 miles/week with some nice fast stretches is fun, so be it.
Update #1: also, anybody who thinks slow and occasional joggers have less than half the mortality of non-joggers, 90% of the population, including all other non-running leisure time athletes, is kidding themselves. Light jogging, and if I had to guess, heavier jogging, doesn't add 6.2 years to male life expectancy holding everything else constant. That's some combination of sampling error, unobserved confounders, reverse causation, and bad modeling.
The study just don't have enough power here to say much. And I'm concerned that their parametric risk adjustment is doing too much work given how few deaths they have, lots of population heterogeneity and strong functional form assumptions that aren't tested.
Ah, now I remember my first reaction to the paper: is there an age-graded reporting bias on pace? Say old runners say they run 'fast' since they run faster than other old guys, but younger runners say they run 'slow' since they run as fast as other young runners. Reference groups change.
Also, can you take a flat slope for 'hours per week' vs mortality seriously, while having a the same time having a steep slope for 'runs per week' vs mortality? One 4 hour run per week is safe, four half hour runs are dangerous?
Update #2: on a more technical note, did they have one young fast high milage runner die on them? That would explain everything, including the high weight on that one occurance (from the runner's low age) --- not sure how small sample confidence intervals work in this sort of Cox model. Not what I do at work...though I do feel I should put I foot down more when I see these sorts of models go by...