Standing desks cut sitting by 100 minutes and left the pain unchanged

The Cochrane review of workplace interventions to reduce sitting pooled 34 studies and 3,397 people. Sit-stand desks cut workplace sitting by about 100 minutes a day in the first three months, with a confidence interval running from 116 minutes down to 84. Over three to twelve months that fell to 57 minutes, and the interval widened to a range from 99 minutes down to 15. Both results are graded low-quality evidence, and past twelve months there is nothing at all.

So the desks do move the needle on how long you sit, and the effect roughly halves within a year. Cochrane’s own press release put the health question in its headline: still unproven.

A separate Cochrane review looked specifically at whether standing or walking more reduces musculoskeletal symptoms in people who already have them. Ten studies, 955 participants. On low back symptoms the pooled effect was a standardised mean difference of −0.35 with a confidence interval from −0.80 to 0.10. That interval crosses zero, which means the result is consistent with no effect. Pain-related disability came out at a mean difference of −0.4 with an interval from −2.70 to 1.90, which is about as null as a result gets.

One genuinely good finding sits alongside those: standing more did not cause harm. No excess musculoskeletal pain, no varicose veins, no productivity drop. If you want to stand because sitting all day feels wrong, the evidence says go ahead. It does not say your back will thank you.

The best-evidenced thing you can buy is an arm support

This one surprised me. The Cochrane review of ergonomic interventions for upper limb and neck disorders in office workers, 15 randomised trials and 2,165 workers, found moderate-quality evidence that an arm support combined with an alternative mouse reduced the incidence of neck or shoulder disorders. The risk ratio was 0.52, with a confidence interval from 0.27 to 0.99. The upper bound sits a hair under 1, so it just clears significance, and it is still the strongest single equipment result in the literature.

The same review found the alternative mouse on its own did nothing measurable, at a risk ratio of 0.62 with an interval from 0.19 to 2.00. And it says plainly that workstation adjustment and sit-stand desks “demonstrated no measurable benefit”. Fourteen of the fifteen studies carried a high risk of bias, so nobody should be smug about any of it.

A 2025 meta-analysis in the Journal of Clinical Medicine pooled 24 trials and 4,086 workers and did find ergonomic interventions reduced pain. The effect was 0.28 points on a 0 to 10 visual analogue scale. The threshold usually cited for a change a patient would actually notice is around 2 points. And the disability outcome, which is the one that decides whether someone can work, came out at a standardised mean difference of −0.13 with an interval from −0.28 to 0.03. Not significant.

OSHA does not tell you how high your chair should be

Search for chair dimensions and you will find dozens of articles citing an OSHA seat-height range. Open OSHA’s chair page and there are no dimensions on it. The only number is that a chair should let you recline at least 15 degrees from vertical. Height gets a qualitative test instead: the chair is right when the whole sole of your foot rests on the floor and the back of your knee sits slightly higher than the seat.

That is not an oversight. The United States has no ergonomics standard, and the eTool is advisory guidance rather than regulation. The UK is the mirror image. HSE legally requires employers to assess display screen workstations and then publishes almost no numbers at all, with its checklist running entirely on words like comfortable and adequate. The one hard dimension HSE gives that nobody else does is a 2 to 3 cm gap between the front of the seat and the back of your knee.

Where OSHA does get specific is the space under the desk, and that page is the one worth reading. It asks for clearance 20 in wide, 17 in deep at knee level, 24 in deep at foot level, and 4 in high at the foot, with a worksurface adjustable between 20 and 27 in near the user.

Be careful quoting OSHA against itself, though. The postures page says elbows bent between 90 and 120 degrees. The evaluation checklist says 90 to 100. The monitors page says a viewing distance of 20 to 40 in, and the checklist says 18 to 20. Canada’s CCOHS says 40 to 74 cm, which is a narrower and closer range than either. These are guidance documents written by different people at different times, and averaging them produces a number nobody published.

Why adjustability is the argument that actually holds

Popliteal height is the distance from the floor to the underside of your thigh just behind the knee, sitting. It is what sets the seat height that lets your feet reach the floor without your thighs being compressed.

The FAA’s published anthropometry puts the fifth percentile woman at 13.8 in and the ninety-fifth percentile man at 18.7. That is nearly five inches of spread between two perfectly ordinary adults, and it is the reason a chair with a fixed seat height is wrong for most of the people who sit in it. No trial is needed to see the problem.

Even designing for that range leaves people out. NASA’s current human-systems guidance is blunt about it: impose fifth to ninety-fifth percentile limits across all the criteria at once and 37 percent of the population will not fit the hardware, because nobody is at the same percentile on every dimension.

Two things I could not find, and will not pretend otherwise. There is no systematic review isolating lumbar support as an intervention for low back pain in office workers. Searches turn up rehabilitation trials in people with chronic back pain and back-belt studies in manual handling, which answer different questions. And the ANSI/HFES 100 standard that furniture marketing loves to name is, according to the Human Factors and Ergonomics Society itself, no longer maintained as an American National Standard. Redevelopment was announced for 2025 and I found no confirmation that a replacement was approved.

How common the injuries are

Three statistical agencies, three different definitions, so they will not add up.

In the UK, 511,000 workers reported musculoskeletal disorders caused or made worse by work in 2024/25, a rate of 1,470 per 100,000, accounting for 7.1 million lost working days and 27 percent of all work-related ill health. That is self-reported survey data from the Labour Force Survey, not clinical diagnosis.

In the US, private-industry employers reported 946,290 cases under the heading of overexertion, repetitive motion and bodily conditions across 2023 and 2024, at 44.7 per 10,000 full-time workers. Of those, 492,140 involved days away from work, with a median of 14 days. That heading is broader than musculoskeletal disorders, and BLS changed its classification system in 2023, so it will not line up with older figures.

Across the EU in 2020, 6.0 percent of workers said they had bone, joint or muscle problems caused or made worse by their job in the previous year.

I have left the cost figures out. The dollar totals that circulate for this come from an insurer’s index rather than a statistical agency, and no government body I could find breaks out a cost for musculoskeletal disorders alone.

What I would actually do with all this

Fit the furniture to the body first, because that part is measurable and the anthropometry is unambiguous. Get the seat height right against your own popliteal height, get the clearance under the desk to OSHA’s 20 by 17 by 24 in envelope, and get the monitor top at or slightly below eye level.

Then treat everything else as comfort rather than medicine. A standing desk will get you out of the chair, and after a year it will get you out of the chair about half as much. An arm support has the best trial evidence of anything you can buy and it is also the cheapest thing on the list. On sit-stand ratios, CCOHS is refreshingly honest: recommendations run from standing five minutes an hour to swapping every 20, 30, 45 or 60 minutes, and there is no consensus.

Our measuring guides cover the dimensions themselves, and the desk converter guide goes through what actually fits when the desk is the constraint.

Questions people ask

No link on this page pays me anything. Read how I check numbers.