Running shoes and injuries: what the evidence actually supports

With that said first rather than last, here is the state of play. It is a genuinely awkward set of findings for an industry that sells footwear on the promise of protection, and it is worth stating plainly rather than hedging.

• No running shoe has been shown to prevent running injuries. Not a stability shoe, not a maximal one, not a minimal one, not an expensive one.
• The pronation-matching model — measure the foot, prescribe the category — was tested in large randomised trials and did not reduce injuries.
• The strongest and most consistently observed predictors of a running injury are previous injury and training load, especially how quickly load increases.
• What shoes demonstrably do change is where load is distributed through the leg, and how comfortable running feels. The second of those has some evidence behind it as a filter worth using.

None of that means shoe choice is pointless. It means the claim has to be smaller and more accurate: a shoe that suits you makes running more pleasant and lets you do the training that actually determines how you get on. If you are picking a first pair, how to choose running shoes does the practical version of that decision; this guide is about what the evidence will and will not let you say about it.

"Studies show" is the phrase to distrust most in this subject, because the studies in question are three very different things doing three very different jobs. Almost every disagreement about running shoes turns out to be someone quoting a laboratory result as though it were a clinical one.

Two structural problems follow from that table. First, randomised trials with injury as the outcome need large numbers and months of follow-up, so there are few of them and most were run in military recruits — a population doing prescribed marching and drill on a fixed schedule, which is not recreational running. Second, the laboratory work is excellent and abundant, which makes it tempting to reason from a measured change in loading straight to a prediction about injury. That step has repeatedly failed when anyone has gone and checked it.

For roughly thirty years the organising idea of running shoe retail was this: the foot rolls inward as it loads, some feet roll further than others, excessive rolling causes injury, and a shoe built to resist it — chosen by looking at the arch, or by watching a treadmill video — reduces injuries. It was coherent, it was easy to explain across a counter, and enormous amounts of footwear were sold on it.

It was also testable, and it got tested. The largest tests were randomised trials in United States military recruits, run across Army, Air Force and Marine Corps basic training, because that is one of the few settings where you can randomise several thousand people into footwear groups and then count injuries systematically. Recruits had their foot type assessed at intake and were either issued the shoe the model prescribed for that foot type or issued a standard shoe regardless. Injury rates between the groups did not meaningfully differ.

The civilian evidence points the same way. A prospective study of novice runners who were all given the same neutral shoe, whatever their feet did, followed them for a year and found that those classified as pronators were not injured more than the rest — if anything, the trend ran the other way. That is observational rather than randomised, so it establishes an association and not a cause. But two different designs, two different populations, and the same direction of answer is about as clear as this field gets.

What collapsed was the prescription, not the product. Stability shoes are not bad shoes, and nothing here says otherwise: plenty of runners simply prefer how a broader, more guided platform feels, particularly late in a long run, and preference is a perfectly good reason to buy one. The claim that failed is the diagnostic one — that measuring a foot and matching a category to it lowers your risk. Stability versus neutral shoes covers what the stability features do and what they cost. In this catalogue they cost nothing in money and about 30 g in weight: the 10 stability shoes have a median price of €150 against €160 for the 68 neutral ones, and a median weight of 288 g against 258 g.

It would be wrong to swing from "shoes prevent injuries" to "shoes do nothing". Shoes demonstrably change things. The question is only which things, and how far a measured change can be pushed.

Laboratory studies show consistently that changing a shoe changes the mechanics of the stride. A lower heel relative to the forefoot shifts work down the chain towards the ankle and calf and away from the knee; a higher one does the reverse. A stiffer forefoot changes how much the toe joints bend and how much energy goes through them. A softer, taller midsole changes the shape of the impact peak and the way the foot sits over its base. These are real, repeatable measurements of where load is distributed.

What they are not is a prediction. Moving load somewhere else is not the same as making tissue fail or not fail, because that depends on how much load the tissue was already handling, how well conditioned it is, and — above all — how fast the amount changed. A shoe redistributes. Your training decides whether the redistribution is a problem.

Two randomised trials in recreational runners are worth naming, because they are rare and they were built properly. In the first, runners were given the same shoe model in two midsole hardnesses without being told which they had, and followed for months: the group in the softer version had a lower injury rate over the period. In the second, runners were given the same model in three different heel drops; there was no overall difference in injury rate between the groups, with a hint in the subgroup analysis that occasional and regular runners responded differently — a subgroup finding, which is the weakest kind of finding there is.

Read together, those are a reasonable basis for two modest statements: cushioning and comfort are worth taking seriously, and no particular drop number is safer than another. They are not a basis for believing that softer is always better, and they are certainly not a basis for a shoe wall sorted by risk.

The most useful idea to come out of running shoe research in decades is also the simplest. Benno Nigg and colleagues proposed that a runner has a movement pattern their body settles into — a preferred movement path — that a shoe which lets them keep it feels comfortable, and that comfort is therefore a usable proxy for whether the shoe is fighting them. Instead of measuring the foot and prescribing, you let the runner filter by feel.

The supporting evidence is thinner than the idea's popularity, and it is worth being precise about which parts are strong. On the mechanism, laboratory work is fairly consistent: runners in footwear they rate as comfortable tend to use slightly less oxygen at the same pace and show somewhat lower muscle activity, which is what you would expect if the shoe is not being fought. On injuries, the most-cited result is a controlled trial in military recruits who were allowed to choose inserts by comfort from a range and who went on to report fewer injuries than a control group. That is one study, in a marching population rather than a running one, using inserts rather than shoes. Encouraging; not settled.

Used as a filter rather than a promise, it is still the best practical advice available, and it has the advantage of being free. Of two shoes that both suit your surface, your session and your budget, take the one that feels better on your foot — and weight that more heavily than the category printed on the box or the description on the wall.

Comfort is assessed standing and moving, in the afternoon when feet are at their largest, in the socks you run in. Getting the fit right covers what to check and in what order, including the one measurement that matters most and is almost never taken.

If there is a footwear risk the evidence does support, it is not any particular value of drop, stack or stiffness. It is changing one of them abruptly.

The minimalist boom demonstrated this at scale. Randomised trials that moved experienced runners into substantially less shoe reported more injuries in the transitioning groups than in the controls over the following weeks, and the effect tracked how fast the transition was run. The shoes were not defective, and the runners were not doing anything foolish. Tissue that suddenly had to do considerably more work had not been given time to adapt to doing more work.

That is a general physiological point rather than a footwear one. Fitness — the ability to keep going — improves over weeks. Tendon and bone adapt to new loading patterns over months. Any change that alters how load is shared through the leg is asking the slow system to catch up, and the slow system does not care that the shoes were expensive.

1. 1

   Change one variable at a time

   Drop, or stack, or a plate, or a category — not three of them in one purchase. If something goes well or badly you want to know which change did it, and a shoe that differs in four ways from your last pair tells you nothing.

2. 2

   Start short, easy and flat

   Put the new pair on for the first two or three runs of twenty to thirty minutes, on a surface you know, at a pace you could hold a conversation at. Hills and speed both amplify whatever the new geometry is doing.

3. 3

   Keep the old pair working

   Alternate for the first few weeks rather than retiring the old shoes the day the new ones arrive. You get a reference for how the new pair feels, and your legs get a break from the new loading pattern every other run.

4. 4

   Give it four to six weeks

   This is a rule of thumb and labelled as one — no trial hands you a number here. But it sits at the right end of the timescale on which tendon and bone adapt, and it is long enough that a genuine mismatch will have made itself known.

5. 5

   Judge it on the next day, not the run

   New shoes almost always feel interesting during the run. What matters is how you feel the following morning, and whether anything that appears is settling down run by run or building up. Building up is the signal to stop and, if it does not resolve, to get it looked at.

A rotation makes all of this considerably easier, because it turns every new shoe into a gradual introduction by default rather than an all-or-nothing swap. Building a shoe rotation covers how to put one together without buying three pairs at once, and when to replace running shoes covers the other end of the same problem — how to tell a pair is finished without waiting for a mileage figure to say so.

If footwear is a small dial, it is fair to ask what the large ones are. The observational literature is unusually consistent here across many cohorts, many countries and many levels of runner — and all of it is observational, so read these as strong, repeated associations rather than as proven causes.

• Previous injury. A previous running injury is the single most consistently observed predictor of the next one, in cohort after cohort. It is also the one nobody can sell you a fix for, which may be why it gets the least airtime.
• Training load, and above all sudden increases in it. Sharp jumps in weekly distance or in hard running show up repeatedly in prospective studies. The famous "ten per cent a week" ceiling is a rule of thumb rather than a trial result — when it has been tested head to head it has not clearly beaten other progressions — but the principle underneath it, that the size of the jump matters more than the size of the week, is well supported.
• Being new to running. Novice runners get injured at higher rates than experienced ones in prospective cohorts, which is one more argument for adding distance slowly in the first year and for treating early enthusiasm as the risk factor it is.
• Everything around the running: sleep, total life stress, other sport, and how much genuine recovery a week contains. Hard to study, routinely ignored, and obvious to anyone who has trained through a bad month at work.

None of that is as satisfying as a purchase, which is exactly why shoes get blamed. A shoe is something you can fix on a Saturday afternoon. A training week is something you have to fix on a Tuesday, and then again the Tuesday after that, for three months.

This guide has deliberately not named a single injury, because naming one invites you to match it to yourself and that is not something a website can do responsibly. What it can do is say clearly when the next step is a person rather than a purchase.

• Pain that is still there after a few days of rest or easy running, or that comes back every time you return to normal training.
• Pain that is present when you are not running at all — sitting, sleeping, walking to work.
• Anything sharply localised to one small spot that you can put a fingertip on, particularly if it is getting worse week on week.
• Swelling, obvious loss of strength, anything that makes you limp, or anything that has been quietly recurring for months.

A physiotherapist or sports doctor can examine you, watch you run, ask what your training actually did over the last month, and rule things in and out. None of that is available to a website and none of it is available to a shoe. If they do suggest a change of footwear as part of a plan, that is a considered recommendation made about you — take it over anything written here.

What a shoe purchase is legitimately for is making the running you want to do more pleasant: the right shoe for your surfaces, a fit that leaves your toes alone, a ride you like enough to go out in on a wet Tuesday. That is a smaller claim than the one on most shoe walls, and it is a real one. The finder ranks the whole catalogue against your surfaces, your sessions, your fit shape and your budget, and tells you why each shoe placed where it did.