The iliotibial band, or IT band, is a thick strip of fibrous tissue on the outside of your thigh. It’s the fifth most common location of injury among runners, and it’s also a very weird anatomical structure.
Two completely different muscles, with seemingly different roles during running, both connect to the IT band: 100% of the fibers of the tensor fascia lata, on the side of your hip, connect to the IT band, but so do about 50% of the fibers of the gluteus maximus (one of the prime movers of the hip joint).
Another strange thing about the IT band is that it seems to be a uniquely human structure: chimpanzees and other primates do not have an IT band (almost all of their gluteus maximus fibers connect directly to the femur, and their TFL muscle inserts on a shorter fascia that terminates above the knee).
Despite its anatomical uniqueness, and its high rate of injury, we—meaning the biomechanical and sports medicine community—don’t really understand a whole lot about what the IT band is doing during running (or walking or cycling, for that matter). All of the most popular “musculoskeletal models” of the body used for gait analysis don’t have a true IT band: they just have a TFL that goes down the side of the leg, and a gluteus maximus that inserts along the femur.
I have just published a new preprint that addresses this problem. The title tells it all: “A full-body musculoskeletal model with an anatomically informed iliotibial band.”
You can read the full preprint below, if you want the technical details, or read on for some highlights on how this model of the IT band works and I think it will be very useful for studying IT band mechanics during running.
Something to understand: this paper is a beginning, not the end, for my research on IT band biomechanics. The big centerpiece—simulating five gait cycles of running at 8:30/mi pace for one runner—won’t blow you away, though it is a big technical accomplishment, and it does produce some nice visualizations. This study, and this IT band model, are a foundation for future research (some of which I’m working on already) into many different aspects of IT band biomechanics.
Building the IT band model
To build this IT band model, I thankfully did not need to start from scratch. Way back in 2014, Carolyn Eng—a PhD student at Harvard doing her dissertation on energy storage in the iliotibial band—painstakingly collected data on the anatomical, geometric, and mechanical properties of the IT band in a careful study of several human cadavers.[1]
That dataset was used to make a digital model of the IT band—and only the IT band, not the rest of the body—in a proprietary biomechanical modeling software called SIMM.
The only problem was that, within a few years, SIMM was hardly in use. Much of the biomechanics community had instead turned to a free and open-source successor, OpenSim, which fostered a rich culture of iterative improvements to full-body models of the body, which advanced hand-in-hand with new methods for studying biomechanics during walking and running.
Current state-of-the-art full-body musculoskeletal models look like this:
And feature anatomically accurate joints, geometrically correct muscle moment arms, realistic strength values for the individual muscles, and carefully tuned and validated parameters for muscle-specific factors you might’ve learned about in exercise science class, like optimal fiber length, tendon slack length, and the curves that dictate the force–length and force–velocity relationships.[2]
The key idea here is that we can use this sort of musculoskeletal model as a digital twin—given gait data from a runner, we can start with a “generic model” that gets scaled to your height, your weight, your limb dimensions, and your (predicted) muscular strength, then make that model move the way you move, and experience the same forces you experience when you hit the ground.
By doing so, we can estimate the forces inside your body, at key injury locations like the Achilles tendon, kneecap, or tibia.
The only problem? Current state-of-the-art models still don’t have a real IT band. So, the strategy to attack the problem was clear: first, I needed to do some digital archaeology: “rescue” the Eng et al. IT band model from the abandoned SIMM project and digitize their cadaver measurements.
Then, I needed to perform a digital Dr. Frankenstein operation: recreate the IT band in OpenSim, then graft it onto a full-body musculoskeletal model. And, of course, validate the model’s fidelity to real experimental data, and see whether it generates reasonable simulations of running.
The real step-by-step went something like this:
And if you want to read through the full details, they’re in the preprint.
Successes and shortcomings
The coolest part about this model is that it actually works—in that it “really runs” in a way that respects the laws of physics, the principles of physiology, and the individual anatomy of the exemplar subject (one adult male, running at about 8:30/mi pace).
The simulation I ran (literally) with the model essentially said “find the muscle activations that the real runner used to move this way,” and the nice part about the results is that we can compare them against the real muscle activations—measured with electromyography (EMG) and never used or seen by the model.
Here’s what the estimated vs. actual muscle activations looked like, for all muscle groups that I had EMG data for:
You’ll notice that the agreement with many of the muscle groups is pretty good.[3] For others, it’s not so good—and the one that’s actually concerning is the TFL.
That muscle is “boxed” in the plot because I don’t have EMG data from this exact subject for that muscle; I had to use EMG data reported by other studies. Since the two muscles connecting to the IT band are the gluteus maximus and the TFL, we’d ideally like to model the activity of both muscles very well.
The model nails it for the glute max, but not for the TFL (at least, to the extent the literature data are correct). The preprint has some extensive discussion on many factors that are not contributing here—it’s not the muscle geometry, or its length or velocity, or other pathological flaws in the modeling procedure. Instead, the optimizer that solves for the estimated muscle activations “wants” to use the TFL quite strongly in early swing—it has a large hip flexion moment arm and thus is a low-cost way to initiate hip drive.
This is probably my biggest outstanding question, but I’m confident in my results in the sense that this TFL activation isn’t “wrong” as per the model’s specifications; to the extent it doesn’t match reality, it means the optimizer itself is missing something about the human body’s strategy for recruiting muscles during running.
The obvious answer here is to get more EMG data, not just on the TFL but on more muscles in the lower body, and use them to improve or validate the strategy (i.e. the “objective function”) used in musculoskeletal simulations of running.
Here is a nice animation of what the right-side muscle activation patterns look like visually (the model has left-leg muscles too, just not shown in this video):
Practical implications of IT band anatomy for runners
Here’s the biggest upshot from this paper: the IT band gets tensioned by two completely different muscle groups—the TFL and the gluteus maximums. And these muscles have fundamentally different “jobs” during running: the TFL is a hip flexor, and the glute max is a hip extensor. What’s more, it might not even make sense to talk about “the” IT band as one monolithic strip of tissue, for a few reasons.
The first is that the IT band has quite different material properties on its anterior side versus its posterior side. The anterior side, which is more directly connected to the TFL, is stiffer, while the posterior side (connected to the gluteus maximus) is much more compliant.
Is this just an artifact of how my IT band model was implemented? No—a conference abstract from a few years ago showed some very clever experimental data suggesting the same thing. The experiment involved sending an electrical stimulus directly to either the TFL or the gluteus maximus, then measuring how the resulting force propagated down the IT band using ultrasonic transducers.
Just as my model predicts, stimulating the TFL generates a separate force transmission pathway than the gluteus maximus—and the gluteus maximus pathway involves a more compliant sub-tract of the IT band.
In running, these sub-tracts experience very different force patterns:
Now, this next part is speculation, but it might be the case that “anterior” IT band injuries involve a different mechanism—or require a different style of treatment—than “posterior” IT band injuries, since the stress involved is coming from a different muscle. I look forward to further research in this area, as well as feedback on this hypothesis from clinicians who have experience treating IT band syndrome in runners.
One other anatomical thread worth pulling on is the fact that not all of the glute max fibers insert on the IT band. One potential mechanism for excessive force being directed down the IT band might be insufficient strength or insufficient muscle activation of the deeper fibers of the gluteus maximus, which insert on the femur.
Do these findings suggest you should work on glute max strength to treat IT band syndrome? I’m not sure—too much glute max work could just be putting more strain into an already-injured area. But also, it could be the case that heavy, controlled glute max work could be helpful in the same way that heavy, controlled calf work is helpful for Achilles tendon injuries, via stimulating the involved tissue. This, too, requires more research.
Some questions I plan on pursuing in future work
My whole motivation for making a “modernized” IT band model was to unlock some of the exciting new techniques available in computational biomechanics. Using a “digital twin” model, we can potentially get real answers to questions like:
- Does strengthening your gluteus medius (or other abductors/external rotators) reduce IT band forces?
- What kind of gait modifications can you make to reduce IT band strain?
- How do IT band forces change as a function of speed?
- Does the loading of the IT band differ qualitatively across different runners?
Which are all questions I plan on using this new model to answer!
Learn more about the biomechanics of training
If you enjoy this kind of content, subscribe to my email list! It's the best way to find out when I have new material coming out.
You can also follow me on Instagram at @jdruns, and on YouTube as well, but the best way to stay in touch is definitely my email list.
If you want to support my work, check out my comprehensive book on marathon training, Marathon Excellence for Everyone. It's the single best resource out there for modern marathon training, and it includes tons of details on the science behind modern marathon training.
Learn more about Marathon Excellence here, or get the book now on Amazon. If you live outside of the United States, Marathon Excellence is also available in a Metric Edition with all workouts in kilometers!
I also have a shorter book from back in 2013 that covers a simple, scientifically based approach to training for distances from the 800m to the 10k, Modern Training and Physiology. Check it out!
Footnotes
[1] She also collected data from several chimpanzee cadavers for another study on the comparative anatomy of the human and the chimpanzee hip anatomy! Eng’s dissertation was published in 2014 which means she was probably collecting her data circa 2010–2013.
[2] These parameters come from a hodgepodge of cadaver, MRI, and in vivo studies on the various properties of the human body, and many clever techniques have sprung up for properly scaling the size and strength of these models. Each new “generation” of model often comes with a new set of observations—for example, a new model was released in 2016 after a paper came out with comprehensive MRI-based estimates for the muscle volumes of all of the muscles in the lower body in healthy young adults.
[3] You’ll notice also there is some lag between the modeled muscle activation and the real EMG signal; this is a pretty well-known phenomenon called electromechanical delay and is attributed to various things we are not modeling—non-homogenous elastic properties in the tissue, for example.
Related articles
New web app: Predicting LT1 pace and Zone 2 pace from 5k time
Designing a plyometrics program for improving bone strength in young runners
Three theories of tissue damage accumulation during running
Lecture: The science behind modern marathon training
What assumptions are baked into your race prediction model?
Tendons do not store energy for free
In windy conditions, running at a constant effort is usually better than running at a constant speed
