How many steps a day lower the risk of dying?

Ten thousand steps is the most widely repeated number in personal health. It is printed on watches, built into phone software and used as a daily target by millions of people. It did not come from a study.

The 2022 meta-analysis that produced the best available figures on the question says so in its opening sentence: although 10,000 steps per day is widely promoted to have health benefits, there is little evidence to support this recommendation. The research that has since been done gives a different and more useful set of numbers.

What the pooled cohorts found

Paluch and colleagues assembled 15 studies — seven published, eight unpublished — with start dates between 1999 and 2018, and asked each study’s investigators to process their own participant data to a shared protocol. The total came to 47,471 adults, 3,013 deaths, and 297,837 person-years of follow-up, with a median follow-up of 7.1 years. The outcome was all-cause mortality, taken from death certificates and country registries.

Participants were sorted into quartiles by steps per day, with medians of 3,553, 5,801, 7,842 and 10,901. Compared with the least active quartile, the adjusted hazard ratio for death was 0.60 for the second quartile, 0.55 for the third, and 0.47 for the fourth.

The shape of that sequence is the finding. The largest single improvement comes from moving out of the least active group — from about 3,500 steps to about 5,800. Going from there to nearly 11,000 adds much less.

Modelling the dose-response curve rather than quartiles gave a point where the benefit levels off, and it differed by age: risk fell with increasing steps until 6,000 to 8,000 steps a day in adults aged 60 and over, and until 8,000 to 10,000 in adults under 60.

Where the benefit begins

A 2024 umbrella review in Preventive Medicine approached the same question from the other end, asking why published estimates vary so much and producing an updated pooled figure.

It found a nonlinear association with a protective threshold at 3,143 steps a day, and a pooled hazard ratio of 0.91 for each additional 1,000 steps. Sorted by activity category, the most active group — above 12,500 steps a day — had the lowest risk, at 0.35 compared with the least active.

The review also explains why different papers report such different numbers. Across eleven earlier systematic reviews it found considerable variability in how results were presented, driven by different ways of quantifying the exposure. Some studies report per-1,000-step increments, some quartiles, some categories, and these are not interchangeable even when the underlying data agree.

Two practical points survive that variation. The first is that the threshold for measurable benefit is around 3,000 steps a day, far below the popular target. The second is that the curve keeps descending above it, so 8,000 is better than 5,000 even though the gain is smaller than the first step up.

Does speed matter?

The 2022 meta-analysis tested this directly, by adjusting for the number of steps and then asking whether the rate of stepping still mattered.

It partly did. Comparing the highest quartile with the lowest, the association with mortality was attenuated but remained significant for peak 30-minute cadence (0.67) and peak 60-minute cadence (0.67). It was not significant for the time spent walking at 40 steps per minute or faster (1.12), nor at 100 steps per minute or faster (0.86).

Read carefully, that is a narrow finding. What survived adjustment was the best sustained half-hour and hour of the day, not the total time spent walking briskly. It is consistent with a period of continuous effort mattering, and it is not strong enough to support the advice to walk faster that usually accompanies it.

What this evidence cannot establish

Every study described here is a cohort study. People were observed, not assigned. That single fact sets the ceiling on what any of it can support.

People who take 11,000 steps a day differ from people who take 3,500 in ways that go far beyond walking: they are less likely to be ill, more likely to be employed, more likely to have somewhere safe to walk, and more likely to do other things associated with living longer. Researchers adjust for the differences they recorded. The ones they did not record stay in the result.

Reverse causation is the specific hazard in mortality research, and it points the same way each time. Illness reduces activity months or years before it kills. Some of the people in the lowest step quartile were not sedentary and then unwell; they were unwell and therefore sedentary. That alone produces an association between few steps and early death without walking having done anything.

None of this makes the finding useless, and none of it is a reason to walk less. It is the reason the conclusion is stated as an association by the researchers themselves, and the reason a randomized controlled trial would answer a question these studies cannot.

Where the numbers land

Question	What the evidence shows	Source
Is 10,000 steps the right target?	No basis in research; the benefit flattens earlier	Lancet Public Health, 2022
Where does benefit start?	About 3,143 steps a day	Preventive Medicine, 2024
Per 1,000 extra steps	Hazard ratio 0.91 (95% CI 0.87–0.95)	Preventive Medicine, 2024
Where does it level off?	6,000–8,000 steps at age 60+, 8,000–10,000 under 60	Lancet Public Health, 2022
Least active vs most active quartile	Hazard ratio 0.47	Lancet Public Health, 2022
Does cadence matter?	Peak 30- and 60-minute cadence yes; total brisk time no	Lancet Public Health, 2022
Is any of it causal?	Not established: all observational	Both

The useful version for a person rather than a population: the distance between doing very little and doing a moderate amount is where nearly all of the measured difference sits, and that distance is about four thousand steps, not ten thousand.