The Science of Sport test range

Donate to The Science of Sport

2009-01-25T19:03:00.001-08:00

Donate now!

Coyle analysis sample post

2008-09-15T02:32:00.000-07:00

The Coyle study on Armstrong: A "minor error" or a scientific "hoax"? Analysis and insight

As promised, we turn our attention to this story, which broke last week, co-inciding with the news that Lance Armstrong is coming out of retirement and will try to race the 2009 Tour de France. It's quite an intricate story, and technical, so forgive the longish post, but we try to go back to the beginning and then work through the sequence of events in order.

The story, which was reported in the New York Times, reports that Ed Coyle, physiologist at the University of Texas, admitted to making what he calls a "minor error" when calculating Lance Armstrong's efficiency during his research. That "minor error" happens to have a major impact on the study's findings, since his main finding was that Armstrong became more efficient between 1993 and 1999.

The paper was, it must be said, widely criticized from the beginning. It drew two separate letters, criticizing the methods, debating the scientific stringency of testing, and questioning the conclusions. It became something of a "shining light" to research without quality control, a running joke of sorts within sections of the scientific community.

I recall attending a conference in the USA soon after its publication - it was the hot topic, of course, because not only was Coyle doing research on an elite subject, it was THE elite subject - the "greatest physiological specimen in the world". Just like roadies wait years to see rock stars or musicians, any exercise physiologist would leap at the chance to publish data on a record-breaking Tour de France cyclist!

So predictably, the paper was something of a conversation starter at scientific conferences. Conversation was not positive, however, with many dismissing it as trivia, rather than science. They were being kind...Few would have expected the next two years to keep the paper quite as much in the public and legal eye, since it became a legal defence vehicle for Armstrong, as we shall see...

First things first: The Coyle study from 2005. What was found?

To begin with, we have to look back and report on the findings of Coyle in the research study that is now the focal point of the "error".

The paper was called "Improved muscular efficiency displayed as Tour de France champion matures", which kind of reveals the paper's hand from the very first line. Here's a breakdown of what Coyle did (note that we're focusing only on the efficiency part, and not some of the other measurements made, but we'll gladly send you the paper if you'd like - just drop us an email address)

The figure below demonstrates just what Coyle did, and what he found.

The research began in November 1992, when Coyle did his first battery of tests on Armstrong. He then did a second test a few months later, followed by a third in 1993. Then a long break interrupted the testing, and it was in that period that Armstrong was diagnosed with and treated for testicular cancer.

Testing resumed in August of 1997, and the final test took place in November 1999. This was the only testing session that co-incided with Armstrong's Tour de France dominance (1999 to 2005), although it must be pointed out that it was done in November, four months AFTER Armstrong won the 1999 Tour. This has relevance for Coyle's conclusions, as we shall see.

The test: Explaining efficiency measures

Testing consisted of a VO2max test, during which time, Coyle measured gases (oxygen in, CO2 out) and did a blood lactate measurement at the end of the test. He calculated two important variables:

Gross Efficiency - the ratio of work done to energy expended to do the work. The work done is taken from the power output, while the energy expended to do the work is calculated using the respiratory gases and calculations we won't get into here. But for example, if a cyclist is riding along at 200 W, and their respiratory gases are used to calculate that their energy consumption is 1000 W (or Joules per second), then that cyclist is 20% efficient, according to this method.
Delta Efficiency - this is a more comprehensive method, because it is calculated as the ratio of the change in work done per minute to the change in energy expended per minute. It is considered a better measure of efficiency because it takes into account the use of oxygen (and energy) at rest and when no work is being performed.

This gets technical, but the simple way to think of this is that as you do more work, your oxygen use rises. We can use that oxygen use to calculate how much energy you are using, and then say that energy use is proportional to work rate (that's fairly obvious, hopefully).

Now, if we take the inverse of the slope of that line (in otherwords, work done vs. energy use), then we can work out delta efficiency. However, it's critical that this slope take into account what the energy use was when you were not doing any work - the resting energy use, and also the energy use when cycling at zero load. The graph below is schematic, but I use it illustrate the point - the energy use rises with increasing work rate, but must take into account energy use when work rate is zero.

Studies as far back as 1975 (Gaesser & Brooks) have shown that gross efficiency tends to skew the results, because of the failure to account for energy use at zero load. Therefore, delta efficiency is considered the better method, but only if used properly, as we'll see!

Coyle's study found that Armstrong's efficiency increased progressively over the 7 years in which he was tested, as shown in the figure above. His delta efficiency improved from 21.37% in 1992 to 23.12% in 1999.

Based on this finding, Coyle named his study, and the theory was published that Lance Armstrong had seen a progressive increase in his efficiency over the years, and this was to become a key part of this "armour" in 2005, when this study would provide some support for his claims that his ascendancy in the world of cycling was "natural". Coyle speculated on a number of physiological factors explaining this finding (changes in enzyme activity, muscle fibre switches etc.). However, the first responses to Coyle's paper were swift...

The first response: Criticism of methods and overinterpretation of data

The first response and criticism was swift, and came from two sources. First, David Martin and colleagues from Australia wrote a letter titled "Has Armstrong's cycle efficiency improved?". This was accompanied by a letter from Yorck Olaf Schumacher and his colleagues titled "Scientific considerations for physiological evaluations of elite athletes".

Essentially, these letters criticized the study design, the method and the scientific process follwed, including the conclusions. The raised the following points:

Timing of testing sessions - Coyle very clearly concluded that his measurements of muscular efficiency were of paramount significance to Armstrong's Tour victories. He denies this, but the title and his conclusions make very clear that his view is that Armstrong's success is a function of this improved efficiency. Coyle would go on to testify in court that Armstrong's rise could have been achieved without doping, so it's quite clear that his finding was intended for support of Armstrong's Tour performance. Yet remarkably, NOT A SINGLE testing session co-incided with the Tour. All the testing happened out of season, and only the 1999 test even overlapped with the Armstrong Tour victories.
Issues around equipment - calibration, reliability, validity etc., which we won't get into here, other than to say that over a period of seven years, the control of equipment is obviously crucial. Coyle responded to these queries, and they do not seem to have huge influence over the current debate
The conclusion - Coyle's was really one of the first papers to even suggest that muscular efficiency improves over time and with training. While this would seem intriguing, it also disagrees with many other findings, which are that extensive endurance training does not improve cycling efficiency. Also, efficiency is not a factor that seems to be associated with performance in elite cyclists, and so the conclusions are 'liberal', to say the least.

The next steps: Re-analyzing the data and digging up errors

These issues are primarily behind my earlier observation that the paper was widely criticized, even early on. However, what transpired next is even more significant, because Christopher Gore, Michael Ashenden, Ken Sharpe and David Martin continued their quest for the "truth", and eventually managed to get hold of (some) data from Coyle's testing.

Their analysis, and some "between the lines" reading of the latest round reveals some surprsing, startling facts about the research. But in the name of time (and length!) I'm going to call it for today's post, and leave you with that teaser, which we'll pick up on tomorrow, when we look the "minor error" and what impact it has on the results.

Join us then!

Ross

Bolt 9.55 seconds? No chance!

2008-09-12T10:30:00.001-07:00

Could Usain Bolt have run 9.55 s without his celebration? Not a chance...

Yesterday, the news wires were buzzing with the news that scientists in Oslo predicted that Usain Bolt, Jamaica's triple Olympic champ, would have run 9.55 seconds had he not celebrated prematurely in his 100m final in Beijing.

It was on the radios, internet, TV news, all over. A while back, just after that race, we speculated that a 9.61s time was about the limit, given the split times that were available.

So this 9.55 s is quite different from that. And far be it from me to criticize the physicist's assumptions, and calculations, but what we have here is a classic case of losing sight of the wood for the trees. Their method involved looking at the final 2 seconds of the race, where Bolt began his celebrations, and compared his acceleration to that of Richard Thompson, who finished second. They looked at two possible outcomes: One is that he maintained the same acceleration as Thompson (that is, slowed down, because all athletes slow down at the end of a 100m race), and the second is if he maintained an acceleration 0.5m/s2 greater than Thompson.

It was in the second of these scenarios that they worked out that he'd run 9.55 s. But the problem with this emerges when you consider the official 10m splits from the race, courtesy the IAAF analysis and a website in which they discuss the race.

So let's look at the analysis, and let me start by asking a simple question: Where in this race are you going to find 0.14 seconds to help Bolt run 9.55 seconds? The answer, as you'll see, is that you won't find it at the end of the race, in the celebrations. It's just not physiologically possible...

The splits:

These are the split times from Bolt's race, according to the IAAF analysis. The graph below it shows the times making some basic assumptions (apologies for the lack of integration and physics equations, but I wish to make a point using simplicity as the vehicle). Again, ask the question: Where are you going to help Bolt knock 0.14 seconds off his time?

The RED line represents the ACTUAL PERFORMANCE. It adds up to a time of 9.685 seconds, considering also that Bolt's reaction time was 0.165 seconds. You'll note that Bolt's fastest 10m interval was from 60 to 70m, taking 0.82 seconds. I must point out that no one has ever measured a human being running a 10m interval faster than this. In our analysis of the race, we got a lot of interesting discussion and data from people, and of all the recorded 10m split times, this is the fastest ever measured. To speed up for the remaining 30 m would represent not only the fastest splits ever run, but also the longest period for which they are ever run.

That's not to say, of course, that Bolt would not be able to run a 0.80s segment. But the key is the pacing strategy - nobody speeds up progressively all the way to the finish line. Nobody. There are mechanical and metabolic reasons for this, but the point is that even holding that speed would be unusual, and speeding up would be highly, highly unlikely. Note, for example, that Bolt has already started slowing down BEFORE he starts celebrating. According to the splits, Bolt slows from 70m to 80m. The celebrations started at 80m. So speeding up? I don't think so.

The BLUE line, to simplify, represents Bolt's projected splits if he continues to accelerate. This is effectively the assumption made by the physicists when the calculate his 9.55 second time. I must emphasize that if you want to find 0.14 seconds at the end of the race (and answer my simple question), then you HAVE TO project that Bolt continues to speed up.

According to this assumption, Bolt would run faster and faster - he has to, in order to do what was projected by the analysis. Again, I must stress that this has never been done - I believe it to be impossible to speed up this much after 70m, and even Bolt would have slowed, or at the very best, held his speed. The whole basis for the argument by the physicists is flawed because there is no reason to believe that Bolt would continue to run faster than Thompson, or accelerate.

The GREEN line represents what I would in fact consider a more likely scenario. In this case, Bolt maintains that top speed that he hits between 60 and 70m. He thus runs the final 30m at 0.82 seconds/10m speed. If he does this, then he run 9.605 seconds.

In reality, I suspect that Bolt would slow down at the end anyway, even without his celebrations. His most likely performance is thus somewhere between 9.61 seconds and 9.69 seconds.

Now, I know there's no fancy physics here, no integration. Just split times, and a very simple question: Where in this race are you going to find 0.14 seconds to help Bolt run 9.55 seconds?

Answer, you can't find that time at the end of the race. Unless you assume that Bolt is going to run a 0.79 second 10m interval somewhere in the race. But that, I'm afraid, is not possible, and therefore, you cannot conclude that he would have run 9.55 seconds without celebrating.

What is possible? There is still time to be made up, but it wasn't the celebration

Having said this, I make the suggestion that Bolt's CELEBRATIONS cost him only about 0.05 seconds. However, that's not to say he cannot still run under 9.60 seconds.

One area for improvement is the start - a reaction time of 0.165 seconds can easily be cut down. Asafa Powell, for example, had a reaction time of 0.134 seconds in Beijing. Therefore, we can estimate that Bolt might get a 0.140 second reaction time.

If that happens, then suddenly he's down to 9.66 seconds. Add to this the fact that there was no tail-wind in Beijing, and it has been estimated that a tailwind of 1m/s improves 100m times by 0.05 seconds. Therefore, on an ideal day, with a tailwind of 1m/s (it could be as much as 2m/s, recall), a super fast reaction time, Bolt could run 9.61 seconds, and still celebrate. Take away those celebrations (another 0.05 seconds, in my estimation), and we have a 9.56 seconds.

But there is no way the 9.55 second time would have come without those celebrations - the trees just got in the way.

Preview of forthcoming attractions. The Coyle-Armstrong debate:

A big debate has flared up in the last few days, ignited by Lance Armstrong's comeback. It turns out that Ed Coyle, he who published a paper that "proved" why Lance Armstrong was superior without doping, has admitted that he may have made "some mistakes" in that paper. He only did so under pressure from the University of Texas after fellow scientists lodged a formal complaint of scientific misconduct against him.

The paper, which you can find here, showed that Armstrong improved his muscular efficiency over the years, but it was fraught with problems. In fact, it became a running joke within sections of the scientific community. That didn't stop Coyle from using his data to testify at a legal hearing that Armstrong had a physiological reason to have dominated without using drugs. It was a shameful display of science meets money meets tacky indulgence, and loses all credibility.

In response to the latest "attacks", Coyle had this to say: “This is a minor waste on my time. However, I don’t understand how they can afford to spend so much time on this. Don’t they have real jobs?”

Well, yes, Ed Coyle, they do have jobs. They are credible scientists, who search for the truth. But then aren't we all?

So the announcement, and the challeges to the Coyle paper are more than welcome. We'll look at the issue early next week. So join us then!

Ross