Walking speed: what counts as a meaningful change, and what the trials on speeding it up found

Walking speed at a usual, comfortable pace is one of the most reproducible simple measures in geriatrics. Why it carries information at all, and how it is measured, we covered separately. This piece answers the question that arrives right after the first measurement: what does a change in that number mean — and can it be acted on so that something other than the number changes.

The second half of the answer turned out to be unexpected, and more honest than this field usually writes.

What is known about the level

The anchor study is a pooled analysis of individual data from nine cohorts: 34,485 community-dwelling adults aged 65 and older, mean age 73.5, 59.6% women, 79.8% white, followed for 6 to 21 years (Studenski et al., JAMA, 2011, PMID 21205966). Mean baseline speed was 0.92 m/s. Each additional 0.1 m/s corresponded to a hazard ratio for death of 0.88 (95% CI 0.87–0.90), and the association held across the whole range of speeds, with no cliff edge.

This is an observational association. It says that people who walk slowly die earlier; it does not say that a person who starts walking faster will live longer. The distance between those two sentences is the subject of the rest of this article.

What counts as a meaningful change

The first question after a repeat measurement is how far the number has to move before it means anything.

The canonical estimate comes not from one sample but from three at once: older adults with mobility disability in a strength-training trial (n=100), people in the subacute phase after stroke (n=100), and a prospective cohort of community-dwelling older people (n=492) — Perera et al., J Am Geriatr Soc, 2006, PMID 16696738. Estimates of a "small meaningful" change in gait speed fell between 0.04 and 0.06 m/s, and of a "substantial" change between 0.08 and 0.14 m/s. Hence the familiar reference points: about 0.05 m/s and about 0.10 m/s.

An important caveat from the authors themselves: this is meaningfulness relative to self-reported mobility. That is, 0.05 m/s is the size of change at which a person begins to notice a difference in how they move. It is not the same as "the size at which prognosis changes".

Recalibration in a different population gave close numbers: 424 sedentary adults aged 70–89 with a low score on the physical performance battery (Kwon et al., LIFE-P, J Nutr Health Aging, 2009, PMID 19536422) — for the 4-metre test, minimally significant change was 0.03–0.05 m/s and substantial change 0.08 m/s. For 400-metre walk time in the same work the thresholds are their own and are measured in seconds (20–30 and 50–60), not in metres per second. Protocols do not convert into one another — there is a harder argument for that below.

And immediately, a limit that is rarely written about

The measurement itself carries error, and that error is comparable to those thresholds.

Start with the good news. Walk the course twice in a row and the spread is small: a study of the minimum detectable change in the 4-metre test (Goldberg and Schepens, Aging Clin Exp Res, 2011, PMID 22526072) used two small groups of community-dwelling older adults — 15 people each, mean ages 74.2 and 72.1 — and error between two consecutive trials was under 5% of mean speed. The smallest change that counts as real at the 95% level in that design was 0.108 m/s in intermediate-speed walkers and 0.144 m/s in fast walkers. But those are two trials at one visit, not a repeat measurement days later.

Once time passes between measurements, the picture changes. The NIH Toolbox normative work: 1,320 people aged 18 to 85, with retesting in 164 of them a mean of 8.6 days later (Bohannon et al., Arch Phys Med Rehabil, 2019, PMID 30092204). Mean usual speed was 1.12 m/s and mean maximum speed 1.61 m/s; women aged 80–85 walked at 0.95 m/s. Reproducibility, however, turned out modest: the intraclass correlation for usual speed was 0.406, and the minimum detectable change was 0.50 m/s, about 45% of the mean. The authors wrote it plainly: the limited reliability restricts judgements about change.

Putting this together: the 0.05 m/s threshold was born as a group-level quantity and works in studies of groups, where sample size carries it. In one person, between two visits, a shift of that size can consist entirely of measurement noise. The practical consequence needs nobody's recommendation: it makes sense to think of your own speed as a series of several measurements in one and the same protocol, rather than as a pair of points.

What is known about change over time

There is a lot of data here, and all of it is observational.

Trajectories of slowing were mapped in the Health ABC study: 2,364 adults aged 70–79, well-functioning at baseline, 20-metre walk, eight years of follow-up (White et al., J Gerontol A Biol Sci Med Sci, 2013, PMID 23051974). Three trajectories emerged; roughly a quarter of participants fell into the fast-decline group — 0.030 m/s per year, about 2.4% of baseline. That group had a 90% higher risk of death than the slow-decline group. Women, Black participants, people with obesity, weaker knee extensors and low physical activity were more likely to be in it.

The mirror observation concerns people whose speed increased: 439 adults aged 65 and older, measured quarterly over a year, with mortality followed for eight years (Hardy et al., J Am Geriatr Soc, 2007, PMID 17916121). Of six measures of health and function, only improved gait speed was associated with survival: eight-year mortality was 31.6% in those who sustained a 0.1 m/s gain, 41.2% in those who improved transiently and 49.3% in those who never improved; the adjusted hazard ratio was 0.42 (95% CI 0.29–0.61).

This study is often cited as proof that speeding up is good for you. It does not show that, and the authors themselves wrote that interventional data are needed. What was observed were people whose speed rose on its own — because a flare settled, because they recovered from illness, because treatment changed. Here a faster step is a sign that the person got better, not the cause of it.

For disability the picture is equally dense and equally observational: a pooled analysis of seven cohorts, 27,220 adults aged 65 and older, outcomes over three years (Perera et al., J Gerontol A Biol Sci Med Sci, 2016, PMID 26297942). Each 0.1 m/s of speed corresponded to a relative risk of dependence in bathing or dressing of 0.68 in men and 0.74 in women. The share of people with difficulty walking a quarter to half a mile ranged from 47% in the slowest subgroup to 4% in the fastest among men, and from 40% to 6% among women.

Walking while doing something else

A separate strand is walking while the person is occupied with something else. The difference between usual speed and speed under a cognitive load is called dual-task cost.

The Gait and Brain study followed 112 people with mild cognitive impairment, mean age 76.6, for up to six years with twice-yearly visits; 27 progressed to dementia (Montero-Odasso et al., JAMA Neurol, 2017, PMID 28505243). A high dual-task cost while counting backwards corresponded to a hazard ratio of 3.79 (95% CI 1.57–9.15), and while naming animals to 2.41 (1.04–5.59). Usual speed below 0.8 m/s on its own, however, did not predict progression to dementia: 3.41 with a confidence interval of 0.99–11.71.

It is an interesting result, and its boundaries are visible in its own description: one cohort, 112 people, 27 events, and these are people with established cognitive impairment, not older adults in general.

And straight away the honest counterweight, because the dual-task test is often sold as universally better: for predicting falls it is no better than the ordinary one. A systematic review with meta-analysis (Menant et al., Ageing Res Rev, 2014, PMID 24915643) found single-task and dual-task gait speed tests equivalent in predicting falls in older people, including in subgroups with cognitive impairment and in slow walkers.

Marker or target — the central question

Everything above describes walking speed as an indicator. Whether acting on it changes anything else is a separate question.

The first half of the answer is encouraging. A meta-analysis of 42 studies, 2,495 healthy older adults, mean age 74.2, baseline speed 1.22 m/s (Hortobágyi et al., Sports Med, 2015, PMID 26286449): resistance, coordination and multimodal programmes together added 0.10 m/s over controls, about 8.4%. That figure pools habitual and fast walking; for habitual comfortable pace alone, which is what this article is about, the gain is smaller — 0.07 m/s, about 5.8%: resistance 0.09 m/s, coordination 0.08, multimodal 0.05. So in relatively healthy older adults the step does get faster, by an amount that lands between the "small meaningful" and the "substantial" change above.

The second half of the answer is where confidence usually ends.

The best trial in this field is LIFE: a multicentre randomised study of 1,635 sedentary adults aged 70–89 with a low physical performance battery score but able to walk 400 metres, followed for an average of 2.6 years (Pahor et al., JAMA, 2014, PMID 24866862). Structured physical activity versus a health education programme: the share of people who lost the ability to walk 400 metres was 30.1% versus 35.5%, hazard ratio 0.82 (95% CI 0.69–0.98).

And now the part that usually disappears from retellings of this trial. Within the same randomisation, the investigators examined which measures actually moved (Santanasto et al., J Am Geriatr Soc, 2017, PMID 28221668; the same 1,635 participants, mean age 78.9, 67.2% women). The result:

• the groups did not separate on the 4-metre gait speed test — overall p = 0.78;

• they did not separate on grip strength (p = 0.62) or balance (p = 0.12);

• on 400-metre walking speed the activity group was faster (p < 0.001);

• on chair stands the activity group did better (p < 0.001), and that component accounted for 39% of the intervention's effect on loss of mobility, with the total battery score accounting for 29%.

All of these comparisons are between groups, not "better than before": both arms could move in the same direction, and they parted only where stated. The same conclusion held when 4-metre speed in that randomisation was analysed in metres per second rather than as a score: the between-group difference came to a few thousandths of a metre per second at every time point (Custodero et al., J Am Med Dir Assoc, 2023, PMID 36878264).

In other words, an intervention that did reduce the rate of mobility loss in this group never separated from the control arm on the short walking-speed test. The benefit travelled past it — through leg strength.

From which follows something rarely said out loud: walking speed on a short test is a very strong marker and a poorly supported target. The claim "add 0.1 m/s and you will live longer" is carried over from observational cohorts, not produced by a trial. As of today there is no randomised study in which speeding up the usual step, by itself, changed mortality, hospitalisations or falls. That is not the same as "exercise does not affect falls": trials of exercise programmes against falls do exist, but they do not isolate a faster short usual step as the cause. The absence of that trial is not a detail: this is exactly the point at which the longevity field normally writes a confident "proven".

Where the 0.8 and 1.0 thresholds came from

Both numbers exist, both have an origin, and neither is a boundary between health and disease.

1.0 m/s was derived and validated inside a single cohort: 3,047 well-functioning older adults, mean age 74.2, a six-metre test, split into a derivation sample (n=2,031) and a validation sample (n=1,016), mean follow-up 4.9 years (Cesari et al., J Am Geriatr Soc, 2005, PMID 16181165). In those walking slower than 1.0 m/s, the relative risk of persistent lower-extremity limitation was 2.20, of death 1.64, of hospitalisation 1.48. The key phrase is "well-functioning": the threshold was derived in people without marked limitation and validated in the same kind of people.

0.8 m/s is not the output of a calculation on one cohort but a consensus of an expert group after reviewing 27 longitudinal papers (Abellan van Kan et al., IANA Task Force, J Nutr Health Aging, 2009, PMID 19924348). The group was looking for an easy-to-remember point for the 4-metre test, and confirmed that usual speed over a short distance predicts disability, cognitive impairment, institutionalisation, falls and mortality at least as well as composite tools.

From there the numbers travelled into guidelines — and travelled differently. The European sarcopenia consensus uses a speed of ≤0.8 m/s as a marker of the severe form (Cruz-Jentoft et al., EWGSOP2, Age Ageing, 2019, PMID 30312372). The Asian working group in its 2019 update took 1.0 m/s on a six-metre test (Chen et al., AWGS, J Am Med Dir Assoc, 2020, PMID 32033882) — meaning that one and the same notion of "slow" differs by a quarter between two consensus documents.

This is not an error by one of the groups. A threshold is a convenient point on a continuous curve, and the curve, as the 2011 pooled analysis showed, has no step in it. Where the point goes depends on who is being measured and why.

There is also a factor the thresholds do not account for at all — height. In the Dijon-Three-City cohort (4,011 people aged 65–85, up to five measurements over 11 years; Elbaz et al., Sci Rep, 2018, PMID 29367642) the difference in fast walking speed between the top and bottom height quartiles was 0.100 m/s at age 65 and had vanished by age 80 — 0.012 m/s, p = 0.57. Speed fell faster in taller participants: −0.183 m/s versus −0.121 m/s over ten years. A fixed threshold at 65 therefore represents a deeper deficit for a tall person than for a short one, and by eighty that correction is no longer needed.

What to do with all this

Nothing written above is an instruction. But several frames follow from it, and inside them a conversation with a clinician becomes concrete.

• One measurement is a point, not a trend. Test error at a repeat visit is comparable to the "meaningful change" threshold, so a series of several measurements says more than the difference between two.

• The protocol has to stay the same. Four metres, six metres, ten metres, 400 metres and the six-minute test are different measurements; a threshold derived on one does not transfer to another. LIFE showed this directly: the 4-metre and the 400-metre tests answered the same intervention differently.

• Marked slowing is discussed, not interpreted at home. Sustained loss of speed, new unsteadiness and falls are a reason for an in-person assessment: behind them sits a long list of causes, from joints and vision to neurology and medication side effects.

• Your own number is worth seeing next to the others. Walking speed reads together with muscle mass, grip strength and cardiorespiratory fitness — separately each describes its own slice of reserve. A Lonevi record shows such measures over time, and that is the only thing worth looking at: the trend, not a single number. What a record summary fundamentally cannot tell you is covered in a separate piece.

And finally, the reason this piece exists. Walking speed is one of the few measures where there is so much data that the edge of knowledge is visible too: an exceptionally strong association with outcomes, and at the same time no trial that would turn that association into influence. This is the normal condition of science, and it is more honest than a confident promise.

This material is for information only and does not replace a consultation with a doctor. Decisions about treatment, investigation and physical activity are made together with your treating clinician.