Is sitting really the new smoking? Running the actual numbers

Someone said it to me at a party in Glockenbach a few weeks back, with the calm confidence of a person quoting scripture: sitting is the new smoking. Then they sat back down on the sofa, looking mildly haunted, and asked whether they should get a standing desk.

I’ve heard the line so often I’d stopped questioning it. Which is embarrassing, because I write about this for a living and the claim is unusually easy to check — both sides of it have been quantified, in the same units, by people who then wrote the comparison down. So I went and looked.

The short version: the analogy fails, and it fails by about a factor of ten. But the reason it caught on is more interesting than the debunk, and there’s one version of it that’s defensible. Let me do this properly.

The claim, as everyone states it

The strong form goes like this. Sitting is an independent risk factor for early death — independent meaning it predicts mortality even after you account for how much someone exercises. Modern life has us sitting nine or ten hours a day. Therefore prolonged sitting is a mass-scale environmental hazard on the order of tobacco, and your desk job is quietly killing you regardless of whether you go to the gym.

The really strong form, which genuinely appeared in headlines, is that sitting is worse than smoking.

The first part of that is true and worth taking seriously. Sedentary time does show up as an independent predictor in a lot of cohorts — I went through that evidence in the post on sedentary behaviour. The problem is entirely in the second step: the leap from “this is a real independent risk factor” to “this is comparable in size to smoking.” Nobody ever showed that. It got assumed.

Where it came from

The phrase is journalism, not epidemiology. It spread in the early 2010s off the back of a wave of cohort studies finding that self-reported sitting time predicted mortality after adjustment for leisure-time exercise — which was a legitimately surprising finding, because until then the assumption was that an hour at the gym bought you the rest of the day.

That result deserved coverage. What it got instead was a slogan, and slogans don’t carry effect sizes. “Sitting is an independent risk factor of modest magnitude” is accurate and unshareable. “Sitting is the new smoking” is neither of those things but travels beautifully.

Fair enough — it’s a good line. It’s just not a measurement.

What the numbers actually show

In 2018 a group led by Jeff Vallance did the obvious thing that somehow nobody had done in public: put the two risks on the same axis. Their commentary in the American Journal of Public Health is short and slightly withering, and it’s the single best document on this question. (According to PubMed: Vallance et al., DOI.)

Here’s what they found, in relative risk:

A 22% increase against a 180% increase. And relative risk always flatters the smaller number, so the authors also converted both into absolute terms — excess deaths per 100,000 people per year, which is the version that actually means something:

• Heaviest sitters vs lightest: about 190 excess deaths per 100,000 per year.
• Heaviest smokers vs never-smokers: more than 2,000 excess deaths per 100,000 per year.

Same units, same population scale, roughly a tenfold gap. That’s the whole debunk, really. Everything after this is nuance.

For the smoking side, the reference numbers are if anything harsher than the ones Vallance used. Jha and colleagues followed over 200,000 American adults in the New England Journal of Medicine in 2013 and found current smokers aged 25–79 died at about three times the rate of never-smokers. The survival numbers are the ones that stayed with me: the probability of making it from 25 to 79 was 70% for women who’d never smoked and 38% for women who smoked; 61% versus 26% for men. Life expectancy shortened by more than a decade. (According to PubMed: Jha et al., DOI.)

There is no sitting study anywhere that looks remotely like that.

The million-person study that reframes it

The other reason the analogy breaks is that sitting, unlike smoking, appears to be substantially cancellable.

Ulf Ekelund’s team harmonised individual-level data from 16 cohorts — 1,005,791 adults, 84,609 deaths, follow-up from 2 to 18 years — and published it in The Lancet in 2016. They crossed sitting time against physical activity, which is the analysis everyone actually needs and almost nobody had run. (According to PubMed: Ekelund et al., DOI.)

Two results from that table matter, and it’s worth being precise about which one does which job — they’re not the same kind of evidence.

The clean comparison. People who sat more than 8 hours a day but fell in the most active quartile — roughly 60–75 minutes of moderate activity daily — had a hazard ratio of 1.04, confidence interval 0.99 to 1.10. That crosses 1. And critically, the referent group there sat under 4 hours a day and was also in the most active quartile. So that comparison holds activity constant and varies only the sitting. It’s the within-stratum test, the one that actually isolates the chair, and among active people it comes back null. The authors’ own conclusion is precisely that: about 60–75 minutes of moderate activity a day “seem to eliminate the increased risk of death associated with high sitting time.”

The other one, which I nearly over-read. People who sat the least (under 4 hours a day) but were in the least active quartile had a hazard ratio of 1.27. Significantly elevated — worse, on the face of it, than the eight-hour sitters above.

It’s tempting to put those side by side and declare the case closed. I shouldn’t, and neither should anyone else: those two groups differ in both sitting and activity, so the ordering doesn’t prove sitting is harmless. A genuine harmful effect from sitting could easily hide behind a larger protective effect from activity and still produce exactly that ranking. Confounded comparisons don’t stop being confounded when they agree with you.

What the 1.27 does establish is narrower and still worth having: sitting less bought those people nothing. Low sitting was not protective in the absence of movement. Put the two results together properly — the null within the active stratum, and the elevated risk among inactive low-sitters — and the picture is that the movement is carrying the weight, not the chair position.

One genuine exception, which I’ll not skate over: TV time behaved worse than general sitting. Three or more hours a day was associated with higher mortality at almost every activity level, and even in the most active quartile, five-plus hours showed a hazard ratio of 1.16. Watching television clearly comes bundled with other things — snacking, later bedtimes, lower income — that a desk job doesn’t. Sitting on your sofa is not epidemiologically identical to sitting at your desk, which is quietly hilarious given how much of the panic targets offices.

The strongest counter-argument

I want to steelman this properly, because there’s a version of the sedentary case that’s much scarier than 1.22 and it deserves airtime.

When you stop asking people how long they sat and strap an accelerometer to them instead, the numbers get considerably worse. Ekelund’s 2019 harmonised meta-analysis in the BMJ pooled device-measured data from eight studies — 36,383 adults, mean age 62.6 — and found a steep gradient for sedentary time: hazard ratios of 1.00, 1.28, 1.71 and 2.63 across quartiles from least to most sedentary. (According to PubMed: Ekelund et al., DOI.)

2.63 is smoking territory. So why don’t I think this rescues the analogy?

First, let me knock down the objection I wanted to make, because it doesn’t survive contact with the paper. The obvious move is to cry reverse causation: in a cohort with a mean age approaching 63, the most sedentary people are plausibly sedentary because they’re already ill, not ill because they sat. Ekelund’s team tested exactly that — they re-ran the analysis excluding everyone who died in the first two years of follow-up. The sedentary association was slightly attenuated and otherwise intact. They still list residual reverse causation as a live limitation, and it surely contributes something, but I can’t hand-wave 2.63 away on those grounds and I’m not going to pretend I can.

Two things genuinely do defuse it, though.

The first is scale. In that same dataset, any physical activity at any intensity was powerfully protective: hazard ratios of 0.48, 0.34 and 0.27 across the second, third and fourth activity quartiles. Even light activity — pottering, standing, walking to the kitchen — ran 0.60, 0.44, 0.38. Those effects are larger than the sedentary penalty, and the sedentary numbers are already mutually adjusted for moderate-to-vigorous activity (they slide from 2.63 to 2.18 in the fully adjusted model). The authors’ own reading, in their words: given how strong the total-activity association is, “it is plausible that findings for sedentary time reflect the ‘flip side’ of a coin.”

The second is that the threshold is higher than the panic implies. The risk curve in that paper is flat-ish until about 9.5 hours a day of total sedentary time, then climbs steeply: 10 hours gave a hazard ratio of 1.48, 12 hours 2.92. That’s total daily stillness — desk, commute, sofa, everything — not your working day. The quartile carrying 2.63 is people who are barely moving at all.

So the accelerometer data doesn’t say the chair is poison. It says total daily movement is one of the strongest mortality signals we have, which is the same thing the ranked-habits post keeps arriving at from other directions — and it’s why the public health line that came out of that paper was “sit less and move more and more often,” not “sitting is the new smoking.”

The honest verdict: wrong by a factor of ten, right about the dose

So: analogy dead. But there’s one sense in which the people repeating it have hold of something real, and it’s worth separating out.

Smoking has a huge risk per person and a shrinking number of people exposed. Sitting has a small risk per person and near-universal exposure. Multiply risk by prevalence and the population-level burdens are much closer than the per-person numbers imply — which is precisely why sedentary behaviour is a legitimate public health target even though it’s nothing like tobacco individually.

That distinction is the whole thing. “Sitting is the new smoking” is a defensible policy statement about aggregate burden, dressed up as a personal statement about your risk. The person on the sofa in Glockenbach heard the personal version, because that’s the version that gets printed.

And the personal version does real damage, in a direction people don’t expect. If you believe your desk job is equivalent to a pack a day, the rational response is despair — the problem is structural, you can’t quit your job, so why bother. Whereas the actual finding is almost cheerful: 30–40 minutes of daily movement neutralises most of it, and an hour neutralises it entirely. That’s a solvable problem. The slogan converts a solvable problem into a hopeless one, which is a strange outcome for a piece of health advice.

What to do instead

Stop optimising the sitting. Optimise the moving. The evidence is overwhelmingly about accumulated activity, not about posture, and the two are not the same lever.

Target the daily dose, not the chair. The Lancet threshold was roughly 60–75 minutes of moderate activity a day to fully erase the association, which is more than most guidelines ask for. But the dose-response is steep at the bottom: the biggest jump in the accelerometer data is between the least active quartile and the second, not between the third and the fourth. Going from nothing to something is where nearly all the benefit lives.

Swap sedentary minutes for light ones — it counts. In the REGARDS cohort, replacing 30 minutes of sedentary time with light activity was associated with 8% lower cancer mortality risk, and with moderate-to-vigorous activity, 31% lower. Same study found bout duration wasn’t significantly associated with risk once total volume was accounted for. (According to PubMed: Gilchrist et al., DOI.) Total volume is the thing. The 20-minute pomodoro-timer stand-up ritual is fine, but it’s not the mechanism.

Walk places, and walk them briskly. Easily the highest-yield habit for a desk worker, partly because pace itself carries signal — see walking pace and longevity and the steps-per-day evidence. I don’t own a car and Munich makes this trivial; the U-Bahn plus twenty minutes of walking at each end does most of my daily dose before I’ve decided to exercise.

Cap the TV, not the desk. If you’re going to be neurotic about one category of sitting, the data says make it the evening sofa, not the office chair. Genuinely the opposite of what the panic targets.

Get a standing desk if you like standing. I’ve got one and I use it perhaps a third of the day, mostly because my back prefers it. It is not a longevity intervention. Standing still is not exercise, it’s just a different way of not moving, and treating it as a solution is how you end up feeling virtuous while your daily activity stays exactly where it was.

Look — sitting isn’t good for you, and I’m not arguing you should do more of it. I’m arguing that the size of a risk is part of the fact, not an optional detail, and that a health claim inflated tenfold isn’t a harmless exaggeration in a good cause. It’s a claim that makes people fatalistic about something they could actually fix in forty minutes a day.

The person at the party did buy the standing desk, by the way. They also started walking home from work. Only one of those was doing anything, and it wasn’t the furniture.