Can This Simple Treadmill Test Predict Your Marathon Time?

A study in this month’s Journal of Strength and Conditioning Research tests a simple way to predict your marathon time, using a treadmill test two weeks before the marathon. Does it work? Yes and no.

The idea is pretty simple. Researchers recruited runners before the Melbourne, Canberra, and Gold Coast marathons. Each runner did a treadmill test that started slowly and got progressively faster until they reached exhaustion and had to step off. The tests lasted about 30 minutes on average.

Then, two weeks later, they ran their marathon, and the researchers checked to see if the treadmill test was a good predictor of how fast the 40 subjects who completed the study ended up running.

Here’s what the data looks like (“Actual MPT” is their marathon time in minutes; “Predicted MPT” is the predicted marathon time based on the treadmill test):

IMAGE COURTESY OF JOURNAL OF STRENGTH AND CONDITIONING RESEARCH

There’s no doubt that the dots cluster around the diagonal line, indicating that there’s a good relationship between predicted and actual time. The statistical analysis suggests that about 45 percent of the variation in marathon time can be predicted by the treadmill test.

Is that useful? The authors argue that this is a “simple, accessible, and cost-effective method” for predicting marathon finishing time, with “many potential benefits including enhanced mental preparation, goal setting, nutrition planning, and fluid planning.”

Perhaps so, but there are a couple of caveats worth considering. First, take a look at the scatter. It’s fair to say that most (though far from all) of the dots are within about 20 minutes in either direction of the line. I would guess that most people who are taking marathon preparation at all seriously (i.e., those who would consider doing this kind of treadmill test) can predict their time to within a 40-minute range without much effort.

In theory, you can estimate marathon performance based on three parameters that can be measured in exercise physiology labs, as Michael Joyner and others have pointed out: VO2max (the maximum rate at which you can use oxygen), lactate threshold (a proxy for how close to that maximum you can stay over the duration of a marathon), and running economy (how much oxygen you need to complete a marathon).

The treadmill test essentially offers a combined rough estimate of VO2max and running economy, similar to taking the speed at the end of a VO2max test—an approach that Tim Noakes and others used for marathon prediction back in 1990. The advantage here is that no lab equipment is required.

In fact, it’s pretty much the same as running a short race, like a 5K, while removing possible confounding factors like course variations and pacing errors. (Even though the average duration of the treadmill test was 30 minutes, it starts off at a near-walk of 24-minute miles at a 4-percent incline, so you’re only really pushing for the last half of the test at most.)

So the interesting question, to me, is: What causes the scatter in the data? Is it mostly intrinsic, or mostly extrinsic?

To put in another way, if you ran the same study with the same subjects a few more times, would the subjects who underperformed (or overperformed) in the marathon relative to their treadmill prediction be likely to underperform (or overperform) again? Or is the scatter just demonstrating the fundamental unpredictability of the marathon, thanks to the difficulty of getting all the details (fueling, pacing, taper, etc.) right on the day?

As you can probably guess, I think there’s a bit of both involved. There’s always going to be some unpredictability, so no prediction method will be perfect. My first reaction in seeing the graph above is that I was surprised at how well such a short and artificial test did at predicting times. In contrast, age, sex, and training hours failed to have any significant predictive value in this particular sample. It’s a testament to the fact that, while marathons are “different,” they’re not that different from other distances: the size of your aerobic engine, and to a lesser extent its efficiency, really matters.

Still, there are some important and reproducible differences in how well some people will be able to extrapolate their 5K ability to a marathon compared to others. A 10K will offer a better prediction than a 5K; a half marathon will offer a better prediction than a 10K; and a marathon will offer the best “prediction” of all. A short one-shot test can’t capture these differences.

So is there any utility in this test? Personally, I’d be more inclined to do a half marathon four weeks out (and then plug the result into one of the many prediction calculators out there) than an arbitrary test on a progressively accelerating inclined treadmill two weeks out. It just seems more relevant and realistic.

But it’s also true that it can be fun to try something different, and to compare your result with how others fared. If you’d like to do that, here’s the treadmill protocol:

Stay on the treadmill until you can no longer maintain the required pace, then step off (safely!). Take your time in minutes, multiply it by (-3.85) and add 351.57. Bingo, that’s your predicted marathon time in minutes! No guarantees and no refunds, though.