REF-1575 · Sherrington et al., 2019 · Cochrane Database of Systematic Reviews 2019, issue · Paper · PubMed
Does exercise prevent falls in older people?
Yes, in trials at 60+: falls 23% fewer (RaR 0.77, 0.71-0.83), led by balance and functional work. Walking alone uncertain; endurance untested. Athletes: a transfer.
Main caveat: Abstract only; no full text is held, so doses, trial lists and the per-domain bias tables were not read.
The search ran to 2 May 2018 across CENTRAL, MEDLINE, Embase, three other databases and two registers. Eighty-one trials (19,684 participants) compared any exercise with control. On an illustrative 850 falls per 1000 people in a year, exercise meant 195 (95% CI 144 to 246) fewer falls; on 480 fallers per 1000, 72 (52 to 91) fewer fallers.
Subgroup analyses showed no evidence of a difference by whether trials selected people at raised fall risk. Effect differed by exercise type (P = 0.004, I² = 71%). Other outcomes were much less certain.
Protocol
Included: randomised trials (nine cluster) of any exercise as a single intervention, aged 60+, living in the community. Four trials in people recently discharged from hospital are not summarised in the abstract.
Primary outcome: rate of falls.
Funding and interests: NIHR support per the Europe PMC record; interests not in what was read.
Key Results
| Exercise vs control | Rate of falls, RaR (95% CI) | People falling, RR (95% CI) | Certainty |
|---|---|---|---|
| All types | 0.77 (0.71-0.83); 12,981 people, 59 studies | 0.85 (0.81-0.89); 13,518, 63 studies | high, high |
| Balance and functional | 0.76 (0.70-0.81); 7920, 39 studies | 0.87 (0.82-0.91); 8288, 37 studies | high, high |
| Multiple types, mostly balance plus resistance | 0.66 (0.50-0.88); 1374, 11 studies | 0.78 (0.64-0.96); 1623, 17 studies | moderate, moderate |
| Tai Chi | 0.81 (0.67-0.99); 2655, 7 studies | 0.80 (0.70-0.91); 2677, 8 studies | low, high |
| Primarily resistance, dance, walking | uncertain | uncertain | — |
| Flexibility, endurance | no trials against control | no trials against control | — |
| Other outcome, all types | Result (95% CI) | Certainty |
|---|---|---|
| People with a fall-related fracture | RR 0.73 (0.56-0.95); 4047, 10 studies | low |
| People with a fall needing medical attention | RR 0.61 (0.47-0.79); 1019, 5 studies | low |
| People with a fall needing hospital admission | RR 0.78 (0.51-1.18); 1705, 2 studies | very low, unclear |
| Health-related quality of life | SMD -0.03 (-0.10 to 0.04); 3172, 15 studies | low, little important difference |
Harms: reported to some degree in 27 trials (6019 participants), closely monitored in both arms in only one. Fourteen trials reported none. Two serious events in one trial (pelvic stress fracture, inguinal hernia surgery); the rest were non-serious and mainly musculoskeletal, median 3 events (range 1 to 26) in exercise groups. Harm is therefore poorly measured, not shown absent.
Mechanisms
Measured: none; the outcome is falls, and the review pools programmes.
Authors' interpretation: programmes that lower falls primarily involve balance and functional exercises.
Inference: falling depends on reacting to loss of balance while weight-bearing, which seated riding trains little. No trial here tests that.
Practical Application
Does exercise prevent falls in older people? Yes, in trials at 60+: falls 23% fewer (RaR 0.77, 0.71-0.83), led by balance and functional work. Walking alone uncertain; endurance untested. Athletes: a transfer.
Good for: adding balance and functional work to a programme for someone over 60, including someone who only rides, walks or swims (the last step is a transfer, Inference).
Good for: counting fracture carefully: RR 0.73 is low-certainty, so the authors write only that exercise may reduce fractures.
Not good for: claiming that lifting alone prevents falls. Primarily resistance programmes are uncertain; the combined ones carry the moderate-certainty result.
Not good for: quality of life, or hospital admissions.
Not good for: programme dose, which the abstract does not give.
Limitations
Abstract only; no full text is held, so doses, trial lists and the per-domain bias tables were not read.
Risk of bias: most trials unclear or high on one or more items, though the authors still reach high certainty for the main falls outcomes.
Falls by recall or diary is the usual method in such trials; the abstract does not specify (inference).
Illustrative rates: the 850 and 480 per 1000 come from control groups of community-dwelling older people, mean age 76, and will not fit a fit athlete.
Harm monitoring is weak.
How this entry was checked and changed
Why this level
(from the abstract) — SR/MA of controlled trials, with GRADE by the authors. A single review, so not A+. Most trials had unclear or high risk of bias on at least one item, and certainty varies by outcome, so quote the GRADE beside each number