Perturbation training topped 19 exercises tested for balance in 423 trials of 33,901 older adults

The exercise that did most for balance in 423 trials was not standing on one leg. It was being shoved.

Perturbation training means someone destabilizes you on purpose, with a push, a tug on a harness, a treadmill belt that jolts underfoot, and you practice the recovery. Across 33,901 older adults, the largest effects were for perturbation training, at 0.66 on a standardized scale, ahead of eighteen other types.

Stretching finished last, and not in a close way. Compared with control, 18 of 19 modalities improved overall balance; stretching showed no clear benefit.

Why falls justify a review this size

More than one in four people age 65 years or older fall each year, and the National Institute on Aging is blunt about what follows: a simple accident like tripping on a rug or slipping on a wet floor can change your life. If you fall, you could break a bone, which thousands of older adults experience each year, and for older people a broken bone can also be the start of more serious health problems and can lead to long-term disability.

Exercise is the intervention everyone recommends, which is exactly the problem. “Exercise” covers tai chi and treadmills and swimming and squats, and until now the guidance has largely treated them as interchangeable. Five databases were searched to 22 December 2025, and 423 trials with 33,901 participants and 2,546 effect sizes were included. The reviewers then used a Bayesian network meta-analysis to rank the modalities against each other rather than one at a time.

That method is worth a sentence of its own. A network meta-analysis compares treatments that were never tested head to head, by chaining them through the trials that share a common comparator. It is how you get a league table out of a literature where almost nobody ran the direct comparison you wanted.

The balance league table

Four modalities separated from the pack. The largest effects were for perturbation training (0.66), sensory training (0.55), dance (0.54) and agility stepping (0.53).

Dance at third is the one that will surprise people, and it makes sense on inspection: a dance class demands weight shifts, direction changes, turning the head while moving, and doing all of it to an external tempo you do not control. That is a balance challenge dressed as a social evening, and it comes with an adherence advantage no clinic program can match, because people turn up to dancing for reasons that have nothing to do with their hip. Sensory training at second is less familiar. Balance runs on three inputs, the inner ear, vision, and the position sense coming up from your feet and joints. Sensory training degrades one on purpose, standing on foam or closing your eyes, and forces the others to compensate.

The balance finding that changes advice

Where the paper earns its length is in refusing to name a single winner. Balance is not one skill, and the ranking flips depending which kind you measure.

Tai Chi, perturbation training, Pilates, and agility stepping or aquatic exercise had the largest estimates for global, static, dynamic and proactive balance, respectively. Static balance is holding still, dynamic is staying upright while moving, and proactive is adjusting before a known challenge such as stepping over a curb you can see. Those are three different jobs for the same body, and there is no obvious reason one exercise should be best at all of them. Reactive is the emergency recovery after something unexpected, and there the signal was weaker: balance-specific training, aquatic exercise and conventional resistance training showed tentative reactive-balance signals.

Setting mattered too. Perturbation training showed larger estimates in increased-fall-risk and supervised groups, whereas sensory training produced the largest home-based estimate. The best exercise for a frail person in a clinic is not the best exercise for someone practicing alone in a kitchen, which is an unusually practical thing for a meta-analysis to be able to say.

What a balance ranking cannot do

The reviewers write the brake into their own conclusion: these estimates do not establish superiority.

That caution is warranted by how network meta-analyses work. Ranking probabilities are notoriously unstable, and when credible intervals overlap, as several of these do, the order can be reshuffled by one new trial. Perturbation training at 0.66 and sensory training at 0.55 are not reliably distinguishable from each other.

Dose is also unresolved. Dose-response associations were nonlinear and exploratory, and the paper closes with a warning about measuring the wrong thing: dose should reflect task-specific challenge, safety and feasibility, because METs capture metabolic cost rather than balance-specific challenge. Standing on a wobble board burns almost no calories and is hard in the way that counts here, which is a problem for every guideline that counts exercise in minutes and intensity. The unit the field uses cannot see the thing this review is about.

And balance measured in a trial is not a fall prevented. These outcomes are balance test scores, not fracture rates, and the step from one to the other is an assumption rather than a result.

What to take to the gym for balance

The practical message survives every caveat: eighteen of nineteen things worked, so the choice that matters most is doing one of them rather than which one you pick.

If you have already fallen or feel unsteady, the evidence points toward supervised perturbation work, which is exactly the kind of thing that needs a professional and a harness rather than a YouTube video. If you are managing at home, sensory training carried the largest home-based estimate. If you want to actually keep going, dance came third out of nineteen and is the only entry on the list anybody attends for fun.

Stretching remains worth doing for other reasons and is not a balance intervention. That is the single clearest thing in 423 trials, and it contradicts a lot of what gets handed out on a photocopied sheet.

Physical activity contributes to the prevention and management of noncommunicable diseases such as cardiovascular diseases, cancer and diabetes, so none of this is the only reason to move. It is a reason to move in a way that occasionally makes you wobble.