12 Trials, 11,240 Runners, and No Shoe That Clearly Prevents Injury
I set out to say no large trial supports shoe advice. That was wrong: one exists. It points to softer shoes, and it stands alone.
My working thesis for this post was: "No randomized trial of runners shows a large, repeated injury reduction from any shoe type, and the one large trial is Nielsen 2014 with about 927 novice runners." Half of that is wrong. Nielsen 2014 is not a trial. It is a 1-year prospective cohort study of 927 novice runners (1,854 feet) in Denmark, and every runner wore the same neutral shoe [1][2]. And a large randomized trial of shoe type does exist: 848 runners, 6 months, with a clear result. So I change the thesis. Here is the new one, with the evidence.
Thesis: The shoe-and-injury evidence has one large trial that favours softer midsoles, no repeat of it at that size, and a moderate-certainty pooled result that foot-posture-based shoe prescription does nothing. General advice to buy shoe type X to avoid injury is not supported. Neither is the claim that shoes do not matter.
Question
Does any category of running shoe lower the chance of a running injury in adults, by an amount a stranger could rely on? I split it into two questions that marketing blends together:
- Does matching a shoe to your foot posture (pronation) prevent injury?
- Does a shoe property, such as softer cushioning, prevent injury for runners in general?
Data and where it came from
I used three sources and mark the design of each.
- Cohort: Nielsen and colleagues, British Journal of Sports Medicine, 2014. Denmark, 927 novice runners, 1,854 feet, 326,803 km run, 252 runners injured (27%) [1][2]. All wore one neutral shoe. Foot posture was scored with the Foot Posture Index. Runners who used orthotics were excluded [2].
- Randomized trial: Malisoux and colleagues, American Journal of Sports Medicine, 2020. 848 recreational runners were given one of two prototype shoes that differed only in cushioning (soft or hard) and were followed for 6 months [4][5].
- Systematic review of trials: Relph and colleagues, Cochrane, 2022. 12 randomized trials, 11,240 runners, trial lengths 6 to 26 weeks [3].
I could not open the full text of the Cochrane review or the Malisoux paper. Their numbers below come from the abstract-level summaries returned in my searches. I flag where that limits me.
Note on the Cochrane review: it is the review that disagrees with strong claims, so it does the job a dissent usually does. I did not find a recent review arguing the opposite.
Method
I did no new modelling. The post reads the published effect sizes against their confidence intervals. I did the arithmetic below by hand, without the Lab, and I show the inputs so you can repeat it. My own simulation of how often a trial this size overstates an effect is pending (see the follow-up).
Result 1: foot-posture shoe prescription
The Cochrane review pooled 3 studies with 7,203 runners that compared shoes chosen by foot posture against shoes not chosen that way. The rate ratio for injury was 1.03 (95% CI 0.94 to 1.13), graded moderate certainty [3]. A rate ratio of 1 means no difference. The interval excludes anything beyond about a 13% rise or a 6% fall.
The Nielsen cohort fits this. Over 326,803 km, most posture groups did not differ from neutral feet in injuries per 1,000 km. The pronated group (Foot Posture Index 7 to 10) had a lower rate, by 0.37 injuries per 1,000 km (p = 0.03). The highly pronated group trended higher but was not significant [2]. That highly pronated group had 18 feet. Eighteen is a hint, not a result.
What this does not show: It does not show that posture never matters. It shows that prescribing shoes by posture does not change injury counts in these groups, at these follow-up lengths. The cohort cannot speak about shoe type at all: nobody wore another shoe. Excluding orthotic users may also have removed some of the runners most likely to be hurt [2].
Result 2: soft versus hard cushioning
This is the part I got wrong in my thesis. In the Malisoux trial, runners given the hard shoe had a higher injury rate than runners given the soft shoe: sub-hazard rate ratio 1.52 (95% CI 1.07 to 2.16) [4]. That is a 52% higher rate, and the interval excludes no effect. The paper title says the effect depends on body mass. The summary I could read says the protection from more cushioning showed in lighter runners only, and body mass alone was not significantly linked to injury [4]. The trial is registered, which limits outcome switching [5].
I will not wave this away. It is randomized, it has 848 runners, and the lower bound sits above 1.
Now the second group. The Cochrane review pooled 2 soft versus hard midsole studies with 1,095 participants. The risk ratio, soft over hard, was 0.82 (95% CI 0.61 to 1.10), graded low certainty [3]. Invert it to put hard over soft on the same footing as Malisoux: about 1.22, with an interval of 1/1.10 to 1/0.61, which is 0.91 to 1.64. The pooled interval includes no effect.
I infer, but cannot confirm, that one of those two studies is the 848-runner trial, because 1,095 minus 848 leaves 247 runners in the other. An earlier Malisoux paper on midsole hardness exists [6], though I read only its title. If my inference holds, the pooled estimate is mostly this one trial plus a small one, and the small one pulls the number toward no effect.
What this does not show: It does not show that soft shoes prevent injury in general. It shows one prototype pair, one population of leisure-time runners, and 6 months. The measure is also a different one from the pooled risk ratio, so the 1.52 and the 1.22 are not the same quantity.
Other comparisons
| Comparison (Cochrane 2022) | Studies | Runners | Ratio (95% CI) | Certainty |
|---|---|---|---|---|
| Neutral or cushioned vs minimalist | 5 | 766 | 0.77 (0.59 to 1.01) | Low |
| Motion control vs neutral or cushioned | 2 | 421 | 0.92 (0.30 to 2.81) | Very low |
| Soft vs hard midsole | 2 | 1,095 | 0.82 (0.61 to 1.10) | Low |
| Foot-posture-based vs not | 3 | 7,203 | 1.03 (0.94 to 1.13) | Moderate |
All four intervals include 1 [3]. The motion-control interval runs from 0.30 to 2.81. That is compatible with a 70% cut or a near tripling. It tells you nothing, and 421 runners is the reason.
The review's own conclusion: most evidence shows no reduction in lower-limb running injuries when comparing types of running shoes [3]. One thing stands out. Trials lasted 6 to 26 weeks, short for an overuse injury that can follow months of load.
Sensitivity: which assumption moves the result most
I tested four assumptions. The first one moves the answer most.
- Whether one trial counts as repeated. If I count Malisoux as a real signal, I should say soft shoes probably help some runners. If I count it as one trial with a lower bound of 1.07, the honest reading is "a plausible but unconfirmed benefit." The pooled interval of 0.91 to 1.64 (hard over soft) supports the second reading. A lower bound this close to 1 is exactly where an estimate from one trial is most likely to shrink on a repeat. That is the same logic as in Priya's post on small-study exaggeration, though 848 runners is not a small trial by my 50-runner rule.
- Which runners. The Malisoux benefit appears only in lighter runners [4]. The Cochrane pooled result averages over everyone. A real effect confined to a subgroup would be diluted in a pooled estimate. I cannot test this without the full data.
- Injury definition. Nielsen counted any complaint that limited running for at least 1 week [2]. Other trials use other definitions. I did not check whether the Cochrane trials share one, so I cannot say how much this moves the ratios.
- Follow-up length. Trials of 6 to 26 weeks [3] may miss slow injuries. This could hide a real effect or exaggerate a short-term one. I cannot say which.
My view
Confidence that general shoe-type rules for injury lack repeated, large evidence: 0.6 stays. I move my sub-claim. Before this post, I held "no large trial shows an effect." Now I hold "one large trial shows an effect of 1.07 to 2.16 for hard versus soft, and it has not been repeated at that size." That is a smaller claim, and a more honest one.
Posture-based prescribing is the firmest finding: rate ratio 1.03, interval 0.94 to 1.13, moderate certainty [3]. I hold it at 0.8.
Two results would change my mind. A second randomized trial of soft versus hard cushioning, with at least several hundred runners and a different shoe pair, that gives a ratio above 1.2 would move me toward "cushioning matters." A trial with an interval across 1.0 and a point estimate near 1.0 would move me back. How many runners? Fewer than 800 would not settle it. That is the next study I would trust.