First things first. The study in question is a paper published yesterday (6th December 2011) in the European Heart Journal by a group in Belgium entitled "Exercise-induced right ventricular dysfunction and structural remodelling in endurance athletes". Unfortunately, it requires a subscription, but you can see the abstract here. Being a scientist myself, I have looked over the paper and thought that I would share my thoughts on the implications.
The study measures various aspects of cardiac function for 40 endurance athletes, each a specialist in one of 4 events which are, in order of duration; marathon, endurance triathlon (possibly Olympic length?), alpine cycling, or ultra-triathlon (Ironman distance). Measurements of cardiac function were taken using cardiac magnetic resonance imaging (cMRI) and echocardiograms, as well as analysis of biochemical markers of function such as B-type natriuretic peptide (BNP) and cardiac troponin I (cTnI), both proteins involved with contractions of muscle cells in the heart and hence good predictors of cardiac risk. For each athlete, three measurements were taken; the first was taken as a baseline 2-3 weeks prior to the event, the second was taken within an hour of crossing the finishing line at the event, and the final 'delayed' measurement was taken 6-11 days after the event.
Broadly speaking, they found the following key results:
- The function of the right ventricle (RV; the chamber of the heart that pumps deoxygenated blood to the lungs) decreases immediately following intense endurance exercise
- Similar decreased function is not seen in the left ventricle (the chamber of the heart that pumps oxygenated body through the circulatory system)
- This decrease in function returns to normal after resting
- The decrease in function is more severe for athletes in the longer endurance events
- Reduction in RV function correlates with increased presence of biomarkers for myocardial injury
- A small number of athletes showed signs of permanent damage to the RV and arrhythmias that may lead to further health problems
- This typically occurred in the older athletes who had been training for longer
First of all, let's put these results into context. The overall crux of the paper seems to be that problems with the RV occur when we do extreme exercise, and the longer the activity the worse the problem. This may seem odd at first - isn't exercise supposed to be good for our hearts? Well yes, but here we are talking about putting our bodies through extreme stresses, not just a short jog on a Sunday afternoon. Your heart has to work incredibly hard to keep up with what you are trying to get out of it, and it is perhaps unsurprising that this results in serious stress. RV fatigue in response to activity has also been suggested by several other recent papers, and this study adds to the mounting evidence that it is RV degradation (and not left ventricular degradation as was previously thought) that is the more clinically relevant in proponents of intense activity.
However, the important thing is that this degradation is only detected immediately following activity. The problems correct themselves and our hearts return to normal function pretty quickly afterwards. So the idea that "running a marathon is bad for your heart" is an overstatement to say the least.
Now the more worrying result is that 5 of the athletes were found to have more permanent cardiac problems (based on delayed gadolinium enhancement, an imaging method that enhances the difference between infarcted and normal myocardium). The data seem to suggest that these 5 individuals were older and had more experience at training. This suggests that some people may be susceptible to suffering fibrosis in the RV walls following prolonged training for larger endurance events (ultramarathons anybody?!).
So the question is; should we be worried? Well first of all I should say that linking this to "marathon runners" is just plain wrong. The paper clearly points out that, of the athletes tested, the effects are least evident in marathon runners. It is the Ironman athletes that are most at risk. However, ultrarunning is probably more comparable with the longer endurance sports studied here, but without looking specifically at this cohort there is not really any way of knowing for sure.
But let's look at the study itself. Statistical analyses are kind of my expertise I suppose, given that my job is analysing data for Cancer Research UK on a daily basis, so my first thoughts were about how these conclusions were reached from the data. Here are a couple of things that I noticed about the study which may affect the conclusions. Apologies if this gets a bit technical - ignore this if it all gets too much!
- The biggest problem with the study is obviously that this is a very small sample size. 40 athletes is small enough, but actually there were only 7 marathon runners. The paper claims that this sample size gives 90 % power to detect a change of 10 % given a significance threshold of 0.05. Basically this means that it will only call 1 out of every 10 real changes non-significant (where we only call a change significant if the probability of getting a value at least as large as what we see is 5 %... You know what, it's not important!). 90 % is very good, but it seems surprising to me given that the sample set is so small. Anyway, drawing conclusions from such a small sample size, particularly where the variation is so high, is likely to be problematic.
- The baseline was taken 2-3 weeks prior to the event. However the variability in the level of training of different athletes at this time is likely to be very high. Whilst some will be tapering, others may still be training hard. For many athletes, this may not be a suitable baseline.
- Similarly, the delayed time-point is taken 6-11 days following the event, but recovering for 11 days rather than 6 is going to have a big difference on recovery! Again, this leads to high variability in the time-course data.
- It would have been good to have used a non-endurance athlete group as a control set. For instance sprinters or 5K/10K runners to check that they do not suffer the same problems as the endurance athletes.
- In fact, it would have been better to focus on cycling and running separately. Perhaps there are some confounding issues here. Maybe the problem is specifically with cycling or swimming and not running? Maybe there's something in the water?! A more informative study (at least from my perspective as a runner...) would look at runners at 100 m, 400 m, 5 Km, 10 Km, 1/2 marathon, marathon, 50 mile, and 100 mile events and see if we see the same correlations.
- There is a level of selection bias in how they recruited their athletes. Whilst they were careful to select only athletes that were highly trained and within the top 25 % of their respective disciplines, all were recruited through advertisements in local triathlon groups. Therefore, marathon runners will have been chosen from the subset that also have a vested interest in triathlons. This may mean nothing, but it's always best to avoid any such biases.
- Also, I'm interested why they selected only the first 40 athletes, rather than accepting as many enrollments as they could get (providing they met the selection criteria). Interesting...
- All of the measurements were taken three times and averaged, with the mean being used. It would have been better to treat these measurements individually rather than combining them prior to analysis.
- I wonder if the results are a little too reliant on p-values. Whilst I do not have access to the raw data, the group distributions that are being compared appear to overlap by a very large amount, and have very high standard deviations as compared to the differences between the means. Calling some of these differences "significant" because they have a p-value less than 0.05, even though the differences do not look real by eye, may mean that the statistical test is not suitable. For instance using paired t-tests on non-Gaussian distributions may bias the data in such a way to make p-value calculations meaningless. The authors use a Kolmogorov-Smirnov test to confirm that the data follow a Gaussian distribution, but with such small sample sizes, who knows...
- It is not clear that the authors have corrected for multiple testing in this experiment. If you compare a large number of measurements using some threshold for what you consider to be "significant", you will see some of these passing your tests due to random chance (so called "false positives" - you declare this to be a statistically significant difference in the two sets even though the difference is purely due to variation in the data). Table 2 mentions that Bonferroni correction was used (which basically scales p-values by the number of tests performed), but nothing is mentioned in the methods section.
- How prevalent is right ventricular fibrosis in the general population? It could just be that some proportion of the population suffer this problem as they get older.
The author makes a good point to dissuade against sensationalist reporting of these results; some people are susceptible to tennis elbow, but people still play tennis! The problem is that some people are going to see this as another argument against running; "oh you don't want to run, it's bad for your knees - and now your heart". Unfortunately, that seems to be exactly what is happening. The comments on the Daily Mail website for instance are quite frightening (WARNING: May induce anger...). Sigh. Still, it could be worse. They could have got hold of this study...
Just in case you were wondering, this is not a real claim from the author that smoking is good for endurance training! The author is making a point that, if you cherry pick your arguments, you can prove anything. This is not "science" by the way. This is (intentionally in this case) "bad science". Don't let the media persuade you that all scientific studies are like this.
Science is all about skepticism - we come up with hypotheses and immediately come up with experiments and tests to refute them. Science doesn't "prove" things, and never has. When suitable evidence exists to quash a theory, we come up with a new one that incorporates the new observations. Here the hypothesis was "RV function is reduced by longer endurance exercise". They tested this by taking a number of measures of cardiac function, and their conclusions suggest that there is no compelling evidence that contradicts this hypothesis. Thus we will keep testing it to make sure it stands up to scrutiny.
Anyway, I will be interested to see what comes out of this in the future. For now, I am quite happy to continue to run without worrying about my heart giving out on me - the benefits far outweigh any potential problems. I have a heart of stone anyway apparently, so I don't need to worry! However it seems that there is an interesting novel effect here that may be more profound in ultra-endurance athletes. Time will tell. Interestingly, one of my colleagues is good friends with the authors of this paper, so I am hoping to have a chat to them about their findings in person.
In the meantime, I urge you to look into the Ultrarunner Longitudinal TRAcking (ULTRA) study being conducted at Stanford University. They are looking for as many ultrarunners as possible to take part. I've done my bit, and I hope that you will too.
Edited: To add the smoking story which I found kind of funny!