FALLS AND EXERCISE: WHAT 108 RANDOMIZED TRIALS MEASURED

Falls are one of the few areas in the exercise literature with a high-certainty answer. A Cochrane review pooled 108 randomized trials and 23,407 people to get it — and the part usually skipped is that the answer is specific to which kind of exercise, not exercise in general.

What the Cochrane Review Measured

Sherrington and colleagues, Cochrane Database of Systematic Reviews 2019 (CD012424). 108 randomized controlled trials, 23,407 community-dwelling participants across 25 countries. Average participant age was 76, and 77% were women. The review's own note on quality: most trials had unclear or high risk of bias for one or more items.

Across 81 trials comparing exercise of any type against a control:

• Rate of falls — reduced 23% (rate ratio 0.77, 95% CI 0.71 to 0.83; 12,981 participants, 59 studies). The review rates this high-certainty evidence. Against an illustrative baseline of 850 falls per 1,000 people followed for a year, that is 195 fewer falls (95% CI 144 to 246).
• Number of people who fall at least once — reduced 15% (risk ratio 0.85, 95% CI 0.81 to 0.89; 13,518 participants, 63 studies), also high-certainty.

Subgroup analysis found no difference in effect based on whether trials selected participants who were already at increased risk of falling.

The Split by Exercise Type

This is the part that carries the practical information. The review found that different forms of exercise had genuinely different effects, and the test for subgroup differences was significant (P = 0.004, I² = 71%).

• Balance and functional exercises — rate of falls reduced 24% (RaR 0.76, 95% CI 0.70 to 0.81; 7,920 participants, 39 studies). High-certainty evidence.
• Multiple exercise types, most commonly balance and functional work plus resistance training — rate of falls probably reduced 34% (RaR 0.66, 95% CI 0.50 to 0.88; 1,374 participants, 11 studies). Moderate certainty.
• Tai Chi — rate of falls may be reduced 19% (RaR 0.81, 95% CI 0.67 to 0.99; 2,655 participants, 7 studies), rated low certainty for fall rate; the reduction in the number of people falling (RR 0.80, 95% CI 0.70 to 0.91) was rated high certainty.
• Resistance training on its own, without balance and functional work — the review states it is uncertain of the effect.
• Walking programs, and dance — also uncertain.
• Flexibility or endurance exercise — no trials compared these against a control at all.

The abridged version published in the British Journal of Sports Medicine (2020;54:885-891) adds one subgroup finding: the effect was larger in trials where the intervention was delivered by a health professional, usually a physiotherapist. No difference was found by age 75+, or by group versus individual delivery.

Where the Evidence Gets Thinner

The falls numbers are the strong part. Everything downstream of them is rated lower by the review's own authors:

• Fall-related fractures — exercise may reduce the number of people experiencing one or more (RR 0.73, 95% CI 0.56 to 0.95), but from only 10 studies and 4,047 participants, rated low-certainty.
• Falls requiring medical attention — RR 0.61 (95% CI 0.47 to 0.79), 5 studies, also low-certainty.
• Falls requiring hospital admission — the review states the effect is unclear (RR 0.78, 95% CI 0.51 to 1.18), 2 studies, very low-certainty.
• Health-related quality of life — the review reports exercise may make little important difference here (SMD −0.03, 95% CI −0.10 to 0.04), with the confidence interval smaller than the minimally important difference on both scales used.

So the fracture finding, which is the outcome most people actually care about, is supported by a fraction of the data behind the falls finding.

Adverse Events

Reported to some degree in 27 trials covering 6,019 participants, though the review notes adverse events were closely monitored in both arms in only one trial. Fourteen trials reported none. Aside from two serious events in a single trial — one pelvic stress fracture and one inguinal hernia surgery — the remainder were non-serious and primarily musculoskeletal, with a median of three events per exercise group.

A More Recent Analysis That Partly Disagrees

Cheng and colleagues, Ageing Research Reviews 2026;113:102924, ran a network meta-analysis on 21 RCTs and 3,387 participants to rank exercise programs against each other rather than against control.

Their SUCRA ranking: the Falls Management Exercise Programme (FaME) at 68.56%, then the Otago Exercise Program at 57.58%, aquatic exercise at 43.96%, Tai Chi at 41.52%, and balance training at 0.58%.

That last figure sits awkwardly against Cochrane's high-certainty finding for balance and functional exercise. The two analyses are not measuring the same thing — Cochrane pooled 108 trials against inactive controls, while this one ranked 21 trials of named programs against each other, and FaME and Otago are themselves multi-component programs built around balance work. Reporting both is more honest than picking the one that reads cleaner.

The same paper also reports a reverse U-shaped dose-response between total exercise dose and fall outcomes, with the modeled optimum at roughly 420 MET-minutes per week — and notes the optimal dose varied by modality.

Building It Into the Day Instead of Scheduling It

Clemson and colleagues, BMJ 2012;345:e4547 — the LiFE trial — tested a different delivery idea in 317 adults aged 70 or older who had already had two or more falls, or one injurious fall, in the previous 12 months.

Three arms: the Lifestyle integrated Functional Exercise approach, which taught balance and strength principles and embedded them into everyday routines rather than into workouts; a structured program of balance and lower-limb strength exercises three times a week; and a sham control of gentle exercise.

Over 12 months, fall incidence was 1.66 per person-year in LiFE, 1.90 in the structured program, and 2.28 in control. LiFE versus control was a 31% reduction (incidence rate ratio 0.69, 95% CI 0.48 to 0.99). The structured program versus control was not statistically significant (0.81, 95% CI 0.56 to 1.17).

Static balance, ankle strength, function and participation were significantly better in the LiFE group than control. Both LiFE and the structured group improved dynamic balance. Falls in this trial were collected by self-report.

Who These Studies Were Actually Run In

This matters more than usual here. The Cochrane population averaged 76 years old and was 77% women. The LiFE trial required participants to be 70 or older and to have already fallen. These are not trials of active people in their fifties.

I could not find a randomized fall-prevention trial run in healthy, active adults in their fifties or early sixties — I searched PubMed for trials in the 50-64 range, in middle-aged adults, and in masters athletes, and did not find one measuring fall outcomes in that population. What exists is a large, high-certainty body of evidence in adults who are on average two decades older than a fifty-year-old reading this, many of them already selected for fall risk. Whether the same effect sizes apply earlier is not something these trials measured.

The Protocol

Framed as what the studied interventions actually contained, not as a prescription:

• The intervention with the strongest evidence was balance and functional exercise, not general fitness. Standing on one leg, narrowed stance, heel-to-toe, controlled sit-to-stand, stepping and reaching — the patterns that challenge balance directly.
• The largest point estimate came from combining categories. Trials that added resistance training on top of balance and functional work produced the 34% figure; resistance training alone was the arm Cochrane called uncertain.
• Tai Chi was a studied intervention with its own evidence, at a smaller effect than the combined programs.
• LiFE tested embedding the work into daily routine rather than into scheduled sessions, and that arm was the one that separated from control in that trial.
• Named multi-component programs — FaME, Otago — ranked highest in the 2026 network analysis. Both are structured, published programs rather than general advice.
• Effects were larger in trials delivered by a health professional, per the BJSM subgroup analysis. That is a finding about the trials, and a reasonable thing to raise with a physical therapist.

The evidence here is unusually strong for the primary outcome and unusually specific about type. It is also drawn almost entirely from people considerably older than the audience this site is written for. Both of those are true at the same time.