Do running apps cause injuries? A Knee Surgeon reads the evidence
Associate Professor Chinmay Gupte
Consultant Orthopaedic Knee Surgeon | Associate Professor in Musculoskeletal Science, Imperial College London | President-Elect, British Association for Surgery of the Knee
The claim that app-generated training plans are driving a rise in running injuries has now reached the national press. The Times has reported coaches and physiotherapists who say they are seeing runners arrive injured or burnt out after following algorithmic plans, with the criticism centring on Runna, the London company bought by Strava last year.
I read this with interest, because the runners in question end up in clinics like mine. I also read it with some caution, because the evidence being offered does not support the claim being made, and because there is a much better piece of evidence that nobody in the coverage has mentioned.
WHAT THE COACHES ARE RIGHT ABOUT
Start with what holds up, because a good deal of it does.
An algorithm that builds a plan from your recent race time is measuring your performance, not your tissue tolerance. Those are different things, and the gap between them is where most running injuries occur. Cardiovascular fitness improves within weeks. Tendon, bone and cartilage adapt over months. A runner whose lungs are ready for the session in front of them may have a tibia that is not, and no personal best in a spreadsheet will reveal that.
The coaches quoted are also right that a plan cannot see the things that most often matter. It does not know you slept badly, that you are ill, that work has been punishing, or that you are six months postpartum. That last point is the sharpest observation in the coverage and the one least discussed elsewhere. Postpartum return to running involves changes in pelvic floor function, connective tissue laxity and sleep that no pace calculator can model.
And the criticism of optional strength work lands. Strength training is the best-evidenced injury prevention measure available across sport. Building it into a product as something the user can skip is a design decision, and most users will skip it.
WHY THE REPORTING CANNOT SETTLE THE QUESTION
Now the problem. The two figures doing the heaviest lifting in the coverage are a coach reporting that around one in five of her clients are former app users, and a physiotherapist reporting that about half her new running clients were following app plans.
Neither figure can tell us anything about causation, because both are missing a denominator.
A running coach’s clientele consists, almost by definition, of people for whom something else did not work. Finding former app users in that group is close to a tautology. The physiotherapy figure has the same problem in a more acute form. If the app has around two million monthly users, then half of a physiotherapist’s new runners using it may simply reflect how popular it is. If half of all runners in a city use a product, you would expect half of injured runners to use it, whether the product helps, harms or does nothing at all.
To answer the question you would need something quite different: the injury rate among app users compared with the injury rate among runners following a magazine plan, a club schedule or a human coach, over the same training block, with similar runners. Nobody has published that. Until somebody does, we have a hypothesis rather than a finding.
Two further points a reader deserves. Almost every critical voice in the coverage is a running coach, which is to say a provider of the service the app displaces, and at least one is developing a competing product. That does not make them wrong, and coaches are precisely the people positioned to notice a pattern first. But it is relevant context. And the case that triggered the debate, a well-known runner who sustained a stress fracture, is one the runner herself attributes to her existing intensive training rather than the app. The coverage also includes a physiotherapist who found the app’s post-injury plan useful in holding him back.
I would add that the baseline here is brutal regardless of method. Marathon training injures a large proportion of those who attempt it. Studies of marathon and half-marathon populations have reported injury or illness symptoms in the large majority of participants across a training block. Against that background, any popular training product will accumulate injury stories.
THE EVIDENCE NOBODY IN THE COVERAGE CITED
Here is what frustrates me about the debate. There is a large, recent, directly relevant study, and it goes unmentioned.
The Garmin-RUNSAFE study followed 5,205 runners across 588,071 running sessions over 18 months, with training data captured from their watches rather than self-reported. Thirty-five per cent sustained a running-related injury. Its central finding overturns how almost everyone, coaches included, has been thinking about training load.
Injury risk tracked the length of a single session relative to the runner’s longest run in the previous 30 days. A run 10 to 30 per cent longer than that carried a 64 per cent higher rate of overuse injury. More than double carried 128 per cent. Meanwhile the metrics actually built into watches and apps performed poorly: the week-to-week ratio showed no relationship with injury, and the acute to chronic workload ratio showed an inverse one.
The authors were explicit that this calls into question how the technology industry has implemented load guidance. That criticism is sharper and better supported than anything in the newspaper coverage, and it applies to the whole category rather than one company.
It also cuts against the coaches. The traditional advice to increase weekly mileage by no more than ten per cent, which underpins how most human coaches build a block, was not supported by these data either. If we are going to criticise algorithms for using the wrong load metric, we should be equally willing to notice that the profession has been using the wrong one too.
Two caveats, since I have just spent several paragraphs on the importance of caveats. This is observational work, so it identifies a strong association rather than proving that following the rule prevents injury. And staying below ten per cent is not a guarantee of safety, since smaller progressions carried some increase in risk as well.
WHAT THIS MEANS IN PRACTICE
For runners using an app, the useful adjustment is not to abandon it. It is to override it in one specific way.
Look at your longest run in the last 30 days. Keep your next long run within roughly ten per cent of that distance, whatever the plan says. If you have had a fortnight off through illness, travel or work, your longest recent run has reset, and the plan will not know that. This is the single most actionable thing to come out of the research, and it is a change any user can make without cancelling a subscription.
Beyond that: do the strength work rather than skipping it, treat prescribed paces as a target rather than an obligation on days when you feel poor, and stop at the first sign of a niggle rather than at the point where it changes how you run.
THE GROUP I WORRY ABOUT
My concern as a surgeon is narrower than the one in the newspapers, and I think more defensible.
An app knows your recent race time. It does not know that you had a partial meniscectomy four years ago, how much meniscal tissue was removed, what your cartilage looked like at the time, or whether your running mechanics have recovered after an ACL reconstruction. Those factors materially change what a knee can tolerate, and they change the advice I would give.
After ACL reconstruction in particular, biomechanical asymmetries during running commonly persist well past the point at which a runner has been cleared on time and strength criteria. A plan generated from a 10km time has no way of detecting that. This is not an argument that apps are dangerous for the general population. It is an argument that a subset of users need individualised clearance that no onboarding questionnaire can provide, and that this subset does not always know who they are.
The company’s own position, that running carries irreducible injury risk influenced by sleep, nutrition, stress and prior injury, and that runners should seek professional input early when something does not feel right, is a reasonable one. The recent addition of injury history prompts at sign-up is a sensible move. The question is whether a set of questionnaire responses can substitute for examination in the people for whom it matters most, and I do not think it can.
WHERE I COME OUT
The claim that running apps are causing an epidemic of injuries is not established, and the evidence being used to support it in the press cannot establish it. The claim that algorithmic plans have structural blind spots, particularly around tissue readiness, illness, life stress and surgical history, is well founded and worth taking seriously.
For most runners the app is not the problem. The long run that jumped too far is. Fix that, do your strength work, and see somebody if your knee swells, locks, gives way or stops you running normally.
If you want the fuller picture on how running knee injuries actually arise and what to do about them, I have written about that separately in my guide to running injuries and the knee.
Associate Professor Chinmay Gupte is a Consultant Orthopaedic Knee Surgeon and Associate Professor in Musculoskeletal Science at Imperial College London, and President-Elect of the British Association for Surgery of the Knee. He treats runners and sports injury patients in London.
This article is for general information and professional education. It does not constitute individual medical advice. If you have specific symptoms or post-operative concerns, please seek personalised assessment from a qualified clinician.
Key References
- Alexander JLN, Culvenor AG, Johnston RRT, Ezzat AM, Barton CJ. Strategies to prevent and manage running-related knee injuries: a systematic review of randomised controlled trials. Br J Sports Med. 2022;56(22):1307-1319.
- Frandsen JSB, Hulme A, Parner ET, et al. How much running is too much? Identifying high-risk running sessions in a 5200-person cohort study. Br J Sports Med. 2025;59(17):e109380.
- Knurr KA, Kliethermes SA, Stiffler-Joachim MR, Cobian DG, Baer GS, Heiderscheit BC. Running biomechanics before injury and 1 year after anterior cruciate ligament reconstruction in Division I collegiate athletes. Am J Sports Med. 2021;49(10):2607-2614.
- Lauersen JB, Bertelsen DM, Andersen LB. The effectiveness of exercise interventions to prevent sports injuries: a systematic review and meta-analysis of randomised controlled trials. Br J Sports Med. 2014;48(11):871-877.
- Sayegh ET, Dib AG, Lowenstein NA, Collins JE, Breslow RG, Matzkin E. Up to one-half of runners return to running one year after arthroscopic partial meniscectomy. Arthrosc Sports Med Rehabil. 2022;4(4):e1505-e1511
