The nucleus of the falls / fracture-prevention cluster — a large, high-lever, patient-important outcome for the older stratum that was previously a named-but-empty gap. Falls are not a surrogate: at least one-third of community-dwelling people over 65 fall each year, and fall-related fracture (hip above all) is a direct driver of loss of independence and death. Source: one gold-tier Cochrane SR, 108 RCTs / 23,407 participants, mean age 76 (Sherrington et al., 2019).
The headline — exercise prevents falls, and this is HIGH-certainty
Exercise (all types) versus a control not thought to reduce falls (Sherrington et al., 2019):
- Rate of falls: RaR 0.77 (0.71-0.83), 59 RCTs, HIGH certainty (GRADE). In absolute terms, at an illustrative control risk of 850 falls per 1000 people per year, that is 195 (144-246) fewer falls per 1000 per year.
- People who fall (>=1 fall): RR 0.85 (0.81-0.89), 63 RCTs, HIGH. At 480 fallers/1000/yr control, 72 (52-91) fewer fallers per 1000 per year.
This is a rare thing in this wiki’s domain: a HIGH-certainty, RCT-based, patient-important-outcome effect. Falls are directly observed and hard to game, exercise-vs-control is randomisable, and removing the high-risk-of-bias trials left the result essentially unchanged (so non-blinding did not downgrade it) (Sherrington et al., 2019). Contrast the physical-activity mortality evidence, which is observational and far softer -> Physical Activity Dose and Mortality.
Which exercise — the decision is NOT exercise, it is balance
The types differ (subgroup test for rate of falls P=0.004), and the ordering is a decision object (Sherrington et al., 2019):
| Exercise type | Rate of falls | Certainty | Verdict |
|---|---|---|---|
| Balance + functional (gait / balance / task training) | RaR 0.76 (0.70-0.81), 39 RCTs | HIGH | the evidenced workhorse |
| Multiple categories (usually balance+functional + resistance) | RaR 0.66 (0.50-0.88), 11 RCTs | moderate | largest point effect (34%) |
| Tai Chi (3D) | RaR 0.81 (0.67-0.99), 7 RCTs | low (rate); HIGH for fallers RR 0.80 | works, mixed certainty |
| Resistance training ALONE | RaR 1.14 (0.67-1.97), 5 RCTs | very low | no clear fall benefit standalone |
| Dance / walking alone | uncertain | very low | insufficient evidence |
| Flexibility / endurance | no trials vs control | — | unstudied (not no effect) |
- The active ingredient is safely challenging balance, not exercise volume as such. The authors’ earlier metaregression and an independent network meta-analysis (Tricco 2017) converge on this (Sherrington et al., 2019). «Exercise programmes that reduce falls primarily involve balance and functional exercises, while programmes that probably reduce falls include multiple exercise categories (typically balance and functional exercises plus resistance exercises).» (Sherrington et al., 2019)
- Resistance training alone does not reduce falls here — its point estimate sits above 1. This is not a contradiction of resistance training’s independent mortality/diabetes signal (Muscle-Strengthening Activity and Mortality) or its role in defending muscle (Protein and Resistance Training for Muscle and Strength): different outcome, different mechanism. Falls are a balance/gait problem; strength is a component of the effective multiple-category programmes but not the lever on its own. (inferred from Sherrington et al., 2019)
- Walk more is not fall-prevention advice. Walking programmes have insufficient evidence and are not what reduces falls — a substitution worth naming, because it is the intuitive default.
Fractures and injurious falls — the leg to the bone side, honestly lower-certainty
The outcomes people most fear are the worst-measured (streetlight effect on the outcome side) (Sherrington et al., 2019):
- Fall-related fractures: RR 0.73 (0.56-0.95), 10 RCTs, LOW certainty — downgraded two levels for imprecision and likely publication bias. Absolute: 47 vs 64 per 1000 -> ~17 fewer fractures/1000/yr. For balance+functional specifically RR 0.44 (0.25-0.76), 7 RCTs, still LOW — a larger point estimate on thin data.
- Falls requiring medical attention: RR 0.61 (0.47-0.79), 5 RCTs, LOW.
- Falls requiring hospital admission: RR 0.78 (0.51-1.18), 2 RCTs, VERY LOW — CI crosses 1, unclear.
- Quality of life: SMD -0.03, little important difference (LOW).
So the chain exercise -> fewer falls is HIGH-certainty; the downstream exercise -> fewer fractures / hospitalisations chain is promising but LOW-to-very-low certainty, because trials under-report these harder outcomes, not because the effect is absent (expectancy test: fractures are rarer events needing far larger trials). This is the exercise leg of fracture prevention.
The two channels of exercise -> fracture, kept distinct (2026-08-27). Exercise reaches fracture by two largely independent routes, and this page measures only one. The falls channel above (balance/functional -> fewer falls -> fewer fractures) is BMD-independent and carries the larger, HIGH-certainty upstream signal. The bone-loading channel (loading -> higher BMD -> stronger bone) is a separate, smaller effect on a surrogate -> Exercise and Bone Mineral Density (Mohebbi 2023: BMD SMD ~0.29-0.41, moderate at best, fracture transmission not shown). A BMD meta-analysis is structurally blind to the falls channel, so BMD under-reads exercise’s fracture value — while a small BMD gain does not itself guarantee a fracture reduction (Surrogate Outcomes). Frame the recommendation on the composite of both channels, not on either number alone. (Mohebbi et al., 2023; inferred from Sherrington et al., 2019)
The other leg — vitamin D / calcium supplements do NOT prevent fractures in the community-dwelling
The intuitive second lever for the older adult — supplement vitamin D and/or calcium for the bones — fails in the community-dwelling, non-deficient stratum, and this is the decision-relevant contrast to the exercise result above. The USPSTF evidence review (Kahwati 2018, gold SR, 11 RCTs, N=51,419, community-dwelling adults >=50 with the deficient / institutionalized / osteoporotic / prior-fracture strata excluded) found (Kahwati et al., 2018):
- Vitamin D + calcium: no fracture benefit. Total fracture (WHI, n=36,282) ARD -0.35% (-1.02 to 0.31), HR 0.96; hip ARD -0.14% (-0.34 to 0.07). SoE Low for no benefit.
- Vitamin D alone: no clear benefit (hip pooled ARD -0.01%, RR 1.08; a single-trial total-fracture signal ARD -2.26% not replicated). Calcium alone: insufficient.
- «Vitamin D supplementation alone or with calcium was not associated with reduced fracture incidence among community-dwelling adults without known vitamin D deficiency, osteoporosis, or prior fracture» (Kahwati et al., 2018).
- A harm to weigh: vitamin D + calcium raised kidney stones — pooled ARD 0.33% (0.06 to 0.60), RR 1.18 (1.04-1.35), SoE Moderate (the review’s only above-low grade); calcium alone did not, and vitamin D alone was unstudied on stones (Kahwati et al., 2018).
This is NOT “supplements don’t work” — it is the enhancement arm of a two-arm curve, and the review says so itself. The benefit lives in the excluded strata: «The fracture benefits overall appear to be largely attributable to benefits among the high-risk populations, with little to no benefit in lower-risk populations (1 fewer hip fracture per 1000 community-dwelling adults per year [95% CI, 0 to 2])» (Kahwati et al., 2018) (Kahwati reporting the 2014 Cochrane review, which included institutionalized/deficient/osteoporotic populations). So the supplement leg is a deficiency-repletion lever, not a community-wide one -> Deficiency Repletion vs Enhancement, Vitamin D and Calcium Supplementation for Fracture Prevention (the page that owns the fracture-benefit-net-of-harms verdict, where a broader D+Ca hip-fracture signal sits as a population DISTINCTION from Kahwati’s community-dwelling null).
The decision-change for the elderly stratum: the two legs of fracture prevention run opposite ways. For the generally-nourished, non-osteoporotic community-dwelling older adult, exercise (balance + functional) carries a fracture signal (RR 0.73, LOW) sitting downstream of its HIGH-certainty fall reduction, while vitamin-D/calcium supplementation shows a fracture null — so the better-evidenced lever is the movement one, reversing the common intuition that the pill is the bone-protective move. Supplementation earns its place only in the confirmed-deficient / institutionalized / osteoporotic stratum (a prescriber act needing this person’s status). (inferred from Kahwati et al., 2018; Sherrington et al., 2019)
Who, and how much — the stratification
- Route (a), stated by the source: the relative effect does not vary by baseline fall-risk, age (>=75 vs <75), or group-vs-individual delivery — every subgroup falls within the headline CI. So there is no effect modification to personalise on; but «there is likely to be a greater absolute impact in people identified at increased risk of falling, but there is benefit also for those who are at more general risk in the community» (Sherrington et al., 2019), because absolute benefit scales with baseline risk -> Baseline Risk and the Relative-Absolute Split. The high-risk elderly get more falls averted per person; the general community still benefits.
- Delivery: effective whether delivered in a group or individually; by a health professional or a trained non-health instructor — though a health professional gave a somewhat larger effect (RaR 0.69 vs 0.82, P=0.04) (Sherrington et al., 2019). This widens implementable options (home, group class, supervised).
- Dose/duration: a sustained habit, not a course — most programmes ran 12 weeks or more and nearly a third lasted a year or more; head-to-head data were too underpowered to set an optimal dose, but higher-challenge, higher-dose balance work trended better (Sherrington et al., 2019).
- Harms: where reported, predominantly non-serious musculoskeletal; 2 serious events across all 108 trials (a pelvic stress fracture and a hernia). Adverse-event monitoring was poor (very low certainty), but nothing signals net harm — the intervention is low-risk relative to a large benefit.
Ranking the intervention TYPES against each other — the NMA verdict (Pillay 2024)
Sherrington ranks exercise types; it does not rank exercise against the whole menu of fall-prevention options (vision treatment, home-hazard assessment, multifactorial and multicomponent programs, CBT, education, vitamin D). Pillay’s network meta-analysis for the Canadian Task Force does — 219 RCTs, 167,864 participants, 59 intervention nodes (incl. usual care), 8 NMAs, one per outcome, with certainty by CINeMA and ranking driven by moderate/high certainty on critical outcomes rather than by SUCRA point ranks (Pillay et al., 2024). This is a type-F refinement of the Sherrington verdict, not independent (type-E) corroboration — Pillay «included 125 of 283 studies included in the previous review» and benchmarks its exercise findings against the Sherrington exercise Cochrane review [19], so the two share a trial base and the second cannot confirm the first as though independent (author lists do not overlap, but shared trials defeat independence anyway) (Pillay et al., 2024).
Of 21 interventions with moderate-or-high certainty for benefit on >=1 outcome, 14 (67%) had a structured-exercise focus (Pillay et al., 2024). The ordering confirms and sharpens the lever is balance:
- Top of the ranking: supervised, long-duration balance/resistance (group or individual) and group tai chi — these carried benefit «across more outcomes than the other exercise and single-component interventions.» Supervised = >2 sessions beyond instruction; long-duration = >3 months. This adds the dose/supervision dimension Sherrington left underpowered: the effective program is supervised and sustained, not a one-off class (Pillay et al., 2024).
- Also beneficial on single outcomes (moderate certainty): whole-body vibration (fallers; but the studied dose — 20 min/day, 5 days/week, 18 months — is onerous), high-intensity/dose education (low-dose education did not qualify), long-duration CBT for fear-of-falling, and mixed programs adding home-hazard assessment (HHA) to a multifactorial core (Pillay et al., 2024).
- Two null/negative results that redraw the menu:
- «No multifactorial intervention showed moderate certainty for any benefit» where components were delivered only on individual needs assessment (Pillay et al., 2024);
- «Adding other interventions to exercise does not appear to substantially increase benefits» (Pillay et al., 2024). Together these say the exercise core does the work — stacking assessment/home-hazard/education on top of it is not where the benefit lives, and a multifactorial program without exercise at its center is not evidenced. The USPSTF reached the same split independently (below).
- Walking is not fall prevention here either — «None of the interventions with moderate certainty evidence focused on walking», and single-component walking carried potential harm signals warranting caution, especially at elevated risk (Pillay et al., 2024). This reinforces the walk-more is not fall-prevention advice substitution above, now from a second trial base.
Two interventions that INCREASED falls — harm at the type level
The NMA is one of the few sources that catches fall-prevention interventions doing net harm (Pillay et al., 2024):
- Universal vision assessment + treatment (VAT) — screening-and-treating a general (non-visually- impaired-selected) population — had moderate certainty it increased both fallers and fractures. The plausible mechanism — a change in refraction destabilising gait before adaptation — is not stated by the source; what the finding licenses is the decision-change that universal vision-correction is not a safe blanket fall-prevention move. (inferred from Pillay et al., 2024)
- A comprehensive multifactorial program (exercise + HHA when indicated) had moderate certainty for increasing fallers — but its direct-evidence trials targeted very high-risk elders (mean ages 83, 85; prior fallers; on fall-risk-increasing drugs), so the harm is confounded with frailty and does not transport to the general older adult (Pillay et al., 2024).
- The workhorse is low-harm: «Vitamin D and most single-component exercise interventions are probably associated with minimal harm» (two exercise nodes: moderate certainty for small ~5-8/100 any-AE harms, little-to-no serious harm) — consistent with Sherrington’s 2-serious-events-in-108-trials safety profile (Pillay et al., 2024).
What patients actually value — the outcome-valuation layer (Pillay KQ2)
The falls cluster has a rare asset: empirical patient outcome-valuation, the values-and-preferences evidence GRADE usually has to substitute panel judgement for -> Rating Outcome Importance. Pillay’s KQ2 review (44 studies, mostly EQ-5D, mostly from people who experienced the event) puts disutilities on a 0 (no HRQoL loss) to 1 (equal to death) scale (Pillay et al., 2024):
| Health state | Disutility (95% CI) | Certainty |
|---|---|---|
| LTC admission | 1 (median, TTO) | LOW |
| Non-hip fracture, <3 mo | 0.57 (0.43-0.71) | LOW |
| Hip fracture, <3 mo | 0.53 (0.44-0.62) | MODERATE |
| Non-hip fracture, 12 mo | 0.19 (0.10-0.28) | MODERATE |
| Hip fracture, 12 mo | 0.16 (0.12-0.20) | HIGH |
| Functional impairment (>=1 ADL) | 0.12 (0.05-0.19) | LOW |
| Fall (within 12 mo) | 0.09 (-0.04-0.22) | LOW |
| Injurious fall | no evidence | — |
- The ranking (not the point values) is the decision object: «Based on the much higher disutility, fracture (of any type) is probably more important than either falls (0.09 over 12 months) or functional status (0.12 for impairment in at least 1 ADL)» (MODERATE), and admission to long-term care may top them all — in the single TTO study «80% of participants said they would rather be dead» (Pillay et al., 2024). This empirically orders the falls cluster’s outcome menu: a fall itself is a mild disutility; what patients weight heavily is the fracture and the loss of independence the fall can trigger. It validates the nucleus’s framing that falls are a route to the feared outcomes, not the feared outcome — and it warns against reading a “fewer falls” result as if it captured the value.
- Fracture disutility is front-loaded: ~0.53-0.57 acutely, falling to ~0.16-0.19 at 12 months — the weight is in the first months after injury.
- Where value meets the ranking: the interventions with some certainty for reducing fractures (>=0.6 fewer per 100 treated) were mixed multifactorial+HHA/education (moderate/high) and — at low certainty — supervised balance/resistance, group tai chi, long-duration CBT, and whole-body vibration (Pillay et al., 2024). The most-valued outcome (fracture) is moved best by the same exercise top tier, but the fracture-specific certainty stays low — the honest gap the nucleus already flags on its fracture leg.
Preference is not efficacy — the delivery-format distinction (Pillay KQ3)
Sherrington finds group vs individual delivery does not modify the fall-reduction effect. Pillay’s KQ3 finds patients prefer individual balance/resistance: «For exercise programs in general, preferences for group or individual delivery probably vary, though individual balance/resistance programs were clearly preferred over group programs with high certainty» (Pillay et al., 2024). These are not in tension (not-joined check (ii): different construct — effect vs preference). The layer-3 consequence is concrete: because format does not trade away effect, following the preference (individual, where feasible) is a free adherence gain — «If implementing new balance/resistance programs, offering individual rather than group sessions when feasible may be most acceptable» (Pillay et al., 2024). Group delivery remains a legitimate, effective, lower-resource option. (inferred from Pillay et al., 2024)
Guidance proximity — where this sits against the guidance set
These reviews are inputs to guidance, which sharpens the guidance-null test. They fed the CTFPHC and informed the World Guidelines for Falls, which stratify by risk: low -> safe-mobility advice; intermediate -> supervised exercise (>=3 sessions/week, >=12 weeks) on functional balance and resistance, or tai chi; high -> multidomain intervention after multifactorial assessment; and single components (medication review, VAT, HHA, CBT) «are not recommended unless part of a multidomain intervention» (Pillay et al., 2024). The USPSTF independently reached the same exercise-first split: «With a B grade, they recommend exercise among those at increased risk for falls, and with a C grade, they recommend clinicians individualize the decision to offer multifactorial interventions» — the C grade because routine multifactorial net benefit «is small» (Pillay et al., 2024). The families converge (exercise for the at-risk older adult, weakly on multifactorial), so the guidance null is strong here and this wiki’s job is calibrated agreement-with-better-warrant — the fabric adds the mechanism (balance is the lever), the harm exceptions (universal VAT, high-risk multifactorial), and the outcome-valuation ordering (fracture >> fall), none of which the guideline statements carry on their face. (inferred from Pillay et al., 2024)
Where this sits in the ranking
For the functionally-drifting or at-risk older adult, this is a big rock: a large effect on a patient-important outcome, known with HIGH certainty, on a low-risk and widely-implementable intervention. It is the strongest outcome class the activity evidence carries for the elderly stratum -> Physical Activity Dose and Mortality. The loop stays open: no operation here grades this against a realised outcome, and appraise, do not prescribe binds — programme selection, progression and fall-risk titration in a specific person are prescriber/coaching acts. (inferred from Sherrington et al., 2019)
Caveat — the HIGH-certainty result is for the GENERAL older adult, not the already-frail
(inferred from Racey et al., 2021) Sherrington’s population is community-dwelling older adults (mean age 76), not selected for frailty. In the identified prefrail/frail stratum, Racey’s SR-MA finds physical activity does NOT significantly reduce falls (RR 0.80 [0.51-1.26], very low certainty, 7 studies) even while it improves mobility, ADLs and frailty status (Racey et al., 2021). This is a distinction, not a contradiction (not-joined check (ii): different population + far thinner evidence): balance+functional exercise is the best-evidenced fall lever before deep frailty; once frailty is established, the falls signal is uncertain, the intervention must be dosed down to avoid injury, and the payoff shifts to function/mobility rather than falls -> Frailty.