Ultramarathon Running Research Project - Pilot Study

Towards the start of the year, I contacted James Elson (Race Director of Centurion Running) about an idea that I had had. I had recently read a few papers by Martin Hoffman's group from the Western States Research Committee looking at factors that affected a runner's ability to complete the Western States 100 mile race. I had also signed up to the Ultrarunners Longitudinal TRAcking project at Stanford, which aims to follow ultrarunners and track their health over many years. I thought that it might be quite interesting to do some of our own researchr here in this country, and thought that James' series of 100 mile races was a perfect opportunity to do something similar, as they are very well organised, they are quite homogenous in their organisation, and James has a particular interest in delving into the stats behind ultra running.

The plan that I proposed to him was to create a couple of surveys for runners of his events to fill in: one before the race, asking questions relating to the runners themselves (biometrics, professional status, running history, etc.); and another survey after the event asking questions relating to how the race went, what their strategy was for the race, their nutrition, etc. By putting this information together with the split times from the event itself, my hypothesis was that we may be able to see some trends dropping out from the data. Yes, this is the kind of thing that I do for "fun"!

In the short term, this approach would give us a good overview of the race itself, the sorts of people who entered, and which runners took a sensible approach to racing. But my main aim was for the long term, building up a database of survey results and race performances for ultrarunners that, over time, will allow us to look at more specific questions relating to performance: Do certain training choices improve your chances of finishing a 100 mile race? Are certain racing strategies more likely to result in a finish/win? Do people with green eyes run faster than people with blue eyes? These burning questions and more may well be answerable in the future from such a database.

To kick things off, we organised a pilot project for the South Downs Way 100 at the end of June. This went very well, and we found some (I think at least) very interesting things from this that bode very well for the future. James has created a research page on the Centurion Running website where you can download the report that I put together (using my official "Dr" title to make it seem to people that don't know me that I have some air of professionalism about me...). Alternatively, I have replicated the report below so that I have a record of it on my little blog. Hopefully you find it interesting, and any comments will be gratefully received. Hopefully we can get even more people interested in taking part in the full project at next year's races!



Introduction

Very little research is currently available regarding factors that may influence a runner's ability to complete a 100 mile event, and the optimum training plan and gear choice for ensuring maximum performance (the work of Martin Hoffman et al. at the Western States Endurance Run being a notable exception - see [1] to see a recent study that inspired our own survey). In many respects, I personally suspect that such answers simply do not exist; we are all different, and what works for one person may not work for another. However, there likely exist some underlying trends that may help guide us to shape our own training and to give us the best possible chances of achieving our goals.

Being a scientist by trade with a particular focus on bio-statistics (my current position is in the field of genetics and bioinformatics for Cancer Research UK), and having a particular personal interest in ultramarathon running, I decided to see if any such trends could be found in a systematic analysis of runners. I contacted the Centurion Running Race Director, James Elson, and together we put together a series of questions that would allow us to analyse certain trends and to see if anything of interest stood out to us.

We had several goals, both in the short term with an analysis of how the race itself went, but also in the long term. The Centurion Running events provide a fantastic opportunity to collect data from multiple independent but consistently organised races throughout the year. The hope is that, as we collect data from multiple races over time, we will develop a large database from which we can ultimately mine interesting results. And who knows; perhaps we will discover the secret to a perfect 100 miler!

This race was our first attempt at collecting these data - a beta test if you will. We had a fantastic response to our survey, and of course we welcome any feedback that people may have regarding what they thought of the survey itself (for instance whether there were any questions that they thought were worded badly or were ambiguous), or any suggestions for additional questions that could be asked. The complete analysis (so far - there is a lot of data to play with) can be downloaded from the Centurion Running website, but this is my selection of the most interesting findings thus far. As time progresses, we will build up this "research", and as the sample size increases we can conduct some really interesting (?!) statistical analyses.

I hope that you find this as interesting as I have, and please remember that we are always open to suggestions for questions of interest. This is research done by runners, for runners, and we all have the same ultimate goal - to have the best run that we possibly can, whether that be to finish as comfortably as possible or to gain that podium finish.

2012 saw the inaugural running of the Centurion Running "South Downs Way 100" (SDW100) race, a 100 mile continuous trail race along the South Downs National Trail from Winchester to Eastbourne. The route takes in the beautiful rolling hills of the South Downs, with a total climb of about 12,700 feet. 

210 runners (including myself) signed up to the South Downs Way 100 race, of which 162 (77.14 %) made it to the starting line on the day. Of these, 117 runners successfully completed the 100 mile course, resulting in a finishing rate on the day of 55.71 %. 66 runners (40.74 % of the starters) opted to take part in our survey. The finishing rate for those taking part in the survey was 53 finishers, giving a finishing rate of 80.3 % amongst those that took part.

Pacing

The decision for how best to pace a 100 mile race is not simple, and will be highly dependent on the individual runner's strengths and weaknesses. Some people aim to run as even a pace as possible, whilst others prefer to set off at a good pace and slow down over the course of the race. Both methods have their advantages and disadvantages - e.g. running a slower even pace allows you to conserve energy for the latter sections, but you may end up running too slow overall or missing your goals; setting off fast will break you away from the pack and avoid getting congested, but may result in you blowing up early. A recent analysis that I conducted on the last 26 years of racing profiles from the Western States Endurance Run suggests that an even pace seems to lead to slightly improved finishing times, whilst heading out too fast often leads to dramatic drop off in times in the latter stages of the race. So how did runners at the SDW100 pace themselves?

Figure 1 shows the distribution of speeds (in miles per hour) between the 15 checkpoints of all 117 runners who completed the race. The speed is calculated based on the distance between successive checkpoints and the time taken between them. Male (blue) and female (red) runners are shown separately. The white dot in the center of each "violin" represents the median value (the middle value if you rank them from lowest to highest), and the width of the violin represents the proportion of runners showing that speed. So essentially, most runners ran at the speed where the violin is fattest.

Figure 1: Distribution of speed (miles/hour) of runners between checkpoints

There are several interesting results from this figure:

1. Generally, runners appear to have set off fast at the start, and slowed down in the second half. However, pacing in the second half of the race was very even rather than continuing to deteriorate.
2. The women were more sensible and held off at the start and were thus able to maintain a more even pace throughout the entire course than the men.
3. The sections from Queen Elizabeth Country Park to Harting Downs, and from Cocking to Bignor Hill seem to be the two hardest sections of the SDW100, based on the fact that the average speed of all runners seems to be consistently lower compared to the other sections.
4. Most people sped up in the final section - probably due to the final downhill section into Eastbourne combined with the fact that it was nearly over! 

Overall, given the course profile, I am quite impressed by the pacing of the runners. It seems that the majority of people were sensible in tackling the course, particularly the women. Figure 2 shows the same data, but this time shows the pacing profile for the individual runners by connecting the points between the aid stations with a straight line. The top 5 male finishers (blue) and the top 5 female finishers (red) are highlighted. The colour of the lines has been graded such that the top finishers profile has the darkest colour. Here we can see that top 5 male runners ran a pretty even pace throughout the event, slowing down over the tougher sections mentioned above, but otherwise holding the pace quite well through to the end. In general, there wasn't a huge difference in the checkpoint splits, but over 100 miles this was enough to lead to a 2 hour gap between first and fifth place. As a slightly selfish point, I can see that I lost most of my time on Ryan Brown (winner of the event) between Housedean Farm and Southease, but otherwise we were quite similar in our pacing. The women's race started very evenly matched, but Claire Shelley was able to hold onto her pace throughout the entire course, ultimately giving her a clear victory, and a 6th place finish overall. The remaining top 5 women were all relatively similarly paced, but slowed more substantially in the latter half of the race.

Figure 2: Pacing profile of the top 5 male (blue) and female (red) runners

Non-Finishers

DNFs (“Did Not Finish”) are an unfortunate reality for all runners. A lot of things have to go right on the day for a runner to complete a 100 mile race with zero issues, and unfortunately over the course of a race the chances of this happening are slim. Everybody suffers bad patches, and often these can be so bad as to necessitate pulling out from the race. This section examines factors associated with the need to pull out.

Figure 3 shows the number of runners who dropped out at each of the 15 checkpoints along the way to Eastbourne. The height of each bar indicates the number of runners who dropped out at that aid station for male (blue) and female (red) runners. The number at the top of each bar represents the cummulative number of DNFs up to that aid station.

Figure 3: Number of runners who dropped out at each aid station

We can see from this figure that no runners dropped out at Jevington (4.3 miles from the end), suggesting that by that point runners felt that they were too close to the end to give up. A surprising number of people had dropped by Queen Elizabeth, and these seemed to be runners suffering from ongoing injuries and illnesses so may have felt by 20 miles in that it was not a good day to run. A number of runners dropped out at Cocking, which may represent runners going above the marathon distance for the first time and struggling, and at Washington, which was the halfway point so may represent a landmark point in the race. The majority of runners dropped out at Clayton Windmills, Housedean Farm and Southease, where runners were a long way into the race and struggling, but far enough from the end that it would be a tough slog to finish.

Figure 4 shows the proportion of the 13 runners who took part in our survey and suffered a DNF at the SDW100. This is only a small proportion of the total number of runners who dropped from the race, and in particular only 1 female runner who took part in our survey DNF'd, but may give us an idea of the factors that may cause a runner to be unable to continue.

Figure 4: Reasons for DNF amongst the runners who took part in our survey

This figure suggests that there was not one single reason for runners to drop, although muscular pain and previous ongoing injuries were both amongst the most common reasons. It is fairly typical for runners to push themselves despite injuries, and this can lead to more problems in the long term (as I have learnt from experience). Muscular pain is also expected given the nature of the event.

What Sort of People Run 100 Mile Events?

It takes immense commitment and mental fortitude to complete a 100 mile race, but I maintain that with the right training, commitment, and determination anyone can do it. But what sort of people put themselves in for these incredible feats of endurance? This section analyses the characteristics of the runners of the South Downs Way 100 race who took part in our survey, and looks at whether or not any of those were significantly different between finishers and non-finishers or between males and females. The survey included a large number of questions regarding the characteristics of the runners taking part in the SDW100, but I will not include all of those here (surprisingly, eye colour appeared to have no bearing on a runner's ability to complete the race, and men are significantly taller than women...). Instead I will look at a few of the more interesting results. The full report containing all of these comparisons can be downloaded from the Centurion Running website.

Figure 5 shows the distribution of the age of all runners who took part in the survey, split between the male (blue) and female (red) runners, and also by finishers and non-finishers. There was a good spread of ages in this event, with runners aging from their early 20s to their late 50s. Most runners were around their 40s, and the distribution was similar for both women and men. Typically it seemed to be the younger runners who DNF'd, so the older generation can obviously teach the younger a thing or two! There's no substitute for experience. However, this difference was not significant in a statistical sense.

Figure 5: Distribution of the age of runners

Figure 6 shows the distribution of the weight of runners who took part in the survey, split between the male (blue) and female (red) runners, and also by finishers and non-finishers. Weight is measured by both the standard weight in kg (left panel) and also by the body mass index (BMI; right panel), which is a measure of your weight scaled by your height squared to normalise between individuals. Generally there was no difference in weight/BMI between finishers and non-finishers, with about a third of people with BMI > 25 ("overweight" and "obese") DNF'ing. One of the "obese" runners (BMI > 30) DNF'd at 70 miles (a fantastic achievement) and the other finished in sub-24. So this seems to suggest that weight isn't a particularly good indicator of whether or not you will complete a 100 mile race, and also that BMI isn't a great measure of fitness (professional rugby players would be "obese" by this measure for instance). We see that, as we would expect, there is a significant difference between males and females (as measured by a Wilcoxon rank sum test).

Figure 6: Weight (left) and BMI (right) distribution for all runners

Further observations (data not shown): 

1. Runners are split roughly half and half between university graduates and non-graduates, and almost all are employed
2. Income is very varied amongst runners
3. Most men are meat eaters, but women are much more diverse in their diets
4. Generally men and women drink about the same amount (~5 units a week), but the heavier drinkers are all men
5. Most runners are in a relationship
6. Most runners have no religious proclivities

Training 

To date there has been no real research into the optimum way to train for a 100 mile event, and different runners will adopt completely different training regiments and gear choices to ensure that they run to their best abilities. In some ways, I believe that for a particular runner, the optimum way to train is whatever works. We all find our own training styles (which may be largely dependent on our personal, family and social lives), our own racing style (whether that is to try and complete the event, to beat a personal best, or to win), and our own gear selection choices (which, if you're anything like me, may be entirely dependent on your current financial situation). Moreso than with races like 10Ks and marathons, there does not seem to be a single school of thought on the best ways of doing things (for example, do you need to do speedwork for a 100 mile race?). However, there are likely to be certain factors that may be correlated with improved race performance. In this section, we look at the type of training that different runners undertook prior to the start of the race.

One thing seems to irrefutable - the best way to train for a running event is to run, and it is no doubt important that for success in ultramarathons running long distances in preparation for the race is important. But there is a balance between teaching the muscles in your legs to efficiently deal with the stresses and strains of running ultramarathons, and the dangers of over-training such as fatigue and injury. Figure 7 shows the distribution of the average weekly mileage for all competitors, and also for finishers and non-finishers. A Wilcoxon rank sum test shows that, for the men, there is a statistically significant difference between the number of miles run per week by finishers compared to non-finishers (as shown by the small p-value). Unfortunately (or fortunately depending on how you look at it) there is only a single non-finisher for the women's race so we do not have enough information for an adequate test. But this seems to suggest that there may be a significant correlation between the amount of training that we do and our chances of completing the 100 mile race. It will be interesting to see the results of this analysis in the future when we have more data from future races.

Figure 7: Distribution of the average weekly mileage of all runners

Figure 8 shows the average of these values, plotted for all finishers and all non-finishers, as well as for the top 5 runners in the survey and for the front and back half of the remaining finishers. Note that here we have only plotted the data for the male runners, since there were only 7 female runners taking part in the survey. This figure suggests that there may be a correlation between total miles run and your position in the race, since the top 5 ran more than the rest of the front pack, who ran more than the back pack, who ran more than non-finishers. This will be an interesting effect to look into in the future.

Figure 8: Average weekly mileage for all runners compared to the top 5 finishers, front of the pack finishers, back of the pack finishers, and non-finishers

These figures seem to suggest that there is definitely benefit to running a good weekly mileage, but the sample size is too small to make any major generalisations. Also, there is another side to this which is not taken into consideration, and that is the problem of over-training. If we train too much without allowing our body to recover adequately, this may also have negative effects on our race performance. In the future, when we have a larger data set from multiple races, it will be interesting to delve into this in more detail.

Figure 9 shows the highest weekly mileage run by runners, whilst Figure 10 shows the number of weeks before the race at which this highest mileage week was run. The majority of runners seemed to run mileage of 60-70 miles (for males) and 70-80 miles (for females), and this was consistent for both finishers and non-finishers. A few runners ran 120+ miles in a week, but these were the minority. It would be interesting to correlate the weekly mileage with ultimate finishing time when the data set has increased in size. Typically runners ran their highest weekly mileage about 4 weeks outside of the race start, but it seems that those that did not finish actually ran their highest mileage closer to the race (2-4 weeks). Again, the sample size is small and this does not appear to be statistically significant, but it would be interesting to see if this fact stands up in future races. If true, this may indicate that runners should leave a suitable length of time for tapering before the race (although I confess, I do not personally do this).

Figure 9: Highest weekly mileage in the 2 months leading up to the race

Figure 10: Number of weeks before the SDW100 when runners ran their highest mileage

Figure 11 shows the distribution of the longest single training run in the 2 months leading up the SDW100 for all competitors, and also for finishers and non-finishers. Interestingly, it seems as if the runners who DNF'd had generally run the longest training runs leading up to the race (median of around 50 miles). Whilst this is not statistically significant, it is an interesting point that having not run a long single stage race previously does not seem to be a major reason for not finishing a 100 mile race. The majority of male finishers seemed to have finished runs of around 30 to 40 miles, whilst female runners were running around 40 to 50 miles at a time.

Figure 11: Longest single training run in 2 months leading up to the race

One interesting point is that the women seemed to have run more than the men in their training. In fact, the women seemed to be generally more prepared than the men (more training, longer running history, etc.). This may indicate that men are more likely to jump straight into something like a 100 miler whereas women are more likely to enter only when they have a lot of experience beforehand. Although we must bear in mind that we have a much smaller subset of women than men. 

Use of NSAIDs

The use of non-steroidal anti-inflammatory drugs (NSAIDs) such as Ibuprofen is very common in ultramarathon events. The stress put on the body in a 100 mile run is high, and inflammation can arise following muscular damage. In particular, it is common for runners to use NSAIDs following more acute injuries, such as sprained ankles and the like to allow them to continue. In some regards NSAIDs are a good idea, since they reduce the swelling and allow you to continue to run without altering your gait (which itself may cause more problems). However, pain is your body's method of telling you that something is wrong, and if it hurts then you probably shouldn't be doing it. It is up to the individual runner whether they are willing to risk further damage in exchange for being able to continue to run.

Some runners however use NSAIDs more regularly, taking them throughout the entire race to prevent muscular damage and to aid with recovery. Interestingly, research by the Western States scientific board has shown that the use of NSAIDs in ultramarathons does not actually reduce muscle damage and soreness at all following the race, and in fact appears to result in increased markers for inflammation and endotoxemia in the blood [6, 7].

More worryingly, there is additional research connecting the use of NSAIDs to renal problems due to a reduction in the filtering rate of the glomerulus (the first step in filtering unwanted waste from the blood to be excreted as urine) [2, 3]. Intense prolonged activity (such as, say, running 100 miles) can lead to a build up in the blood of myoglobin, a protein released into the bloodstream from damaged muscle tissue. If this occurs, reduced filtration can be disastrous, as myoglobin may build up in the kidneys leading to rhabdomyolysis and potentially kidney failure. Other studies have shown that the use of NSAIDs may increase the chances of developing exercise-associated hyponatremia (EAH) - an imbalance in the levels of electrolytes in the blood (usually caused by over-hydration), leading to reduced blood serum sodium concentration and impaired cellular transport. This may further increase the chances of developing rhabdomyolysis [4, 5].

Whilst these studies are relatively small, they do suggest that more should be done to increase the awareness of the effects of using NSAIDs in ultramarathons, beyond them simply being "magical pills" that allow us to complete the race.

We asked runners taking part in the SDW100 about their use of NSAIDs to see how prevalent their use was. Figure 12 shows the proportion of runners who took NSAIDs at some point in the race, and how often they took them. Amongst the competitors that completed our survey, 42.4 % used NSAIDs at some point during the race, with the majority of those using them only once or twice. There appears to be a higher proportion of NSAID usage for female runners as compared to males, and also interestingly a higher use between finishers as compared to non-finishers. In both cases, the sample size is too low to draw any conclusions from this, but it will be interesting to see if this trend is also seen in future races.

Figure 12: The use of NSAIDs amongst competitors

Conclusions

Working together with James Elson and Centurion Running, we have been able to collate a large amount of data for runners who took part in a 100 mile ultramarathon race. In the future, we hope to collect additional data from runners in all future Centurion Running events and build a database of information which can be linked to race performance, with the ultimate goal of using this to look for potential trends that may indicate the optimal way to train for and race in a 100 mile event. Hopefully this will add to the growing research being conducted by others to better understand the sport that we all love.

I think that these analyses have highlighted some interesting things. The most interesting thing to me was the pacing strategies used by runners throughout the race, shown in Figure 1. I believe that this could be interesting for anybody looking at taking part in SDW100 2013 (look out for the sections leading to Harting Downs and Bignor Hill!). Once we have a larger number of runners, I will look at how people that did not finish (particularly those that dropped for reasons of fatigue) paced the race to see if there is anything to learn for future runners (e.g. did they set out too fast).

I believe that the number of runners who dropped out at each aid station, shown in Figure 3, says a lot about the psychology of an ultrarunner. There certainly appear to be certain landmarks where people are more likely to drop out. From personal experience I can completely see a runner pushing to at least get to the halfway point rather than drop out at 40 miles. And in the latter stages of the race, when the effects of the race are really being felt, the chance of dropping out seems to decrease as the distance to the final goal decreases. When you can smell that finish, you'll push through anything to get there!

A lot of the characteristic data (such as age, height, weight, education, etc.) will ultimately be incredibly interesting to get a feeling for the type of person who would willingly do this kind of thing for fun. Whether there exists an "average ultrarunner" I highly doubt - in fact I would go so far as to say that the very concept is oxymoronic! But it certainly seems as if the decision to push yourself to "go long" is universal, regardless of your age, gender, body shape, nationality, ethnicity, education, or religious beliefs. The commonality is a mental aspect; the part of the brain that says "that's too hard" or "why?" doesn't seem to be engaged. As Henry Ford said, "Whether you think that you can, or that you can't, you are usually right".

There is a lot of data already, and a lot of additional analyses that we can perform particularly once the size of the database increases. In the meantime, I have written this report to "wet your whistle" for what may come in the future. I hope that you have enjoyed reading through some of these analyses (and I hope that I have kept things relatively non-technical to avoid scaring people off). I certainly have found it to be quite enlightening and I thank everybody that took part in the survey for the time that they spent in filling in the questionnaires. We got a great response, and hopefully the results of that work shown here will encourage even more people to take part in subsequent surveys in future races.

As a shameless plug, I am finding myself looking into running-related data more and more these days. If you are interested in reading more, please keep a lookout on my blog at constantforwardmotion.blogspot.com.

Congratulations once again to everybody that took part in the SDW100! Now I am looking forward to being on the other side of the aid station at the North Downs Way 100 in August. If you see me, say hi! Just look out for the sideburns.

References

1. Hoffman MD and Fogard K, Factors Related to Successful Completion of a 161-km Ultramarathon, Int J Sports Phys and Performance, 2011; 6:25-37.
2. Murray MD and Brater DC, Adverse effects of nonsteroidal anti-inflammatrory drugs on renal function, Ann Intern Med, 1990; 112:559–560.
3. Baker J, Cotter JD, Gerrard DF, Bell DF and Walker RJ, Effects of indomethacin and celecoxib on renal function in athletes, Med Sci Sports Exerc, 2005; 37:712–717.
4. Bruso JR, Hoffman MD, Rogers IR, Lee L, Towle G and Hew-Butler T, Rhabdomyolysis and hyponatremia: a cluster of five cases at the 161-km 2009 Western States Endurance Run, Wilderness Environ Med, 2010; 21:303–308.
5. Ellis C, Cuthill J, Hew-Butler T, George SM and Rosner MH, Exercise-associated hyponatremia with rhabdomyolysis during endurance exercise, Phys Sportsmed, 2009; 37:126–132.
6. McAnulty S, McAnulty L, Nieman D, Morrow J, Dumke C and Henson D, Effect of NSAID on muscle injury and oxidative stress, Int J Sports Med, 2007; 28:909–915.
7. Nieman DC, Henson DA, Dumke CL, et al. Ibuprofen use, endotoxemia, inflammation, and plasma cytokines during ultramarathon competition, Brain Behav Immun, 2006; 20:578–584.