The risk of sitting

Last week, everyone was blogging and tweeting about this New York Times article about the health effects of sitting for long periods every day. And I think everyone got it wrong.

Let’s start with a disclaimer. Although I have both formal training and professional experience dealing with probability and statistics (it was part of my Ph.D. thesis), epidemiology isn’t my thing. It may be me that’s wrong here. If someone with a professional background in the topics covered in this post writes in to tell me I’m all wet, I’ll be happy to eat my words.

The Times article covers two recent studies. The first, done in Finland, showed that people who exercise sit around, on the average, just as much as people who don’t exercise. The key line from the abstract is

In conclusion, exercise for fitness, regardless of its duration, does not decrease the inactivity time during normal daily life.

This study is interesting as much for its methodology as for its results: rather than keeping a diary of their activity, the test subjects wore instrumented shorts that monitored their quadriceps and hamstring muscles. Pretty cool.

But this wasn’t the study that bloggers and tweeters focused on.

The second study was a much larger and longer investigation into the relationship between sitting and mortality. Done in Australia, this work covered more than 200,000 people over about three years. According to the Times,

They found that the more hours the men and women sat every day, the greater their chance of dying prematurely. Those people who sat more than eight hours a day — which other studies have found is about the amount that a typical American sits — had a 15 percent greater risk of dying during the study’s three-year follow-up period than people who sat for fewer than four hours a day.

That increased risk held true in the Australian study even if the people sitting eight hours a day spent at least part of that day exercising.

It’s that second paragraph that led people astray, I think. The common interpretation of it was “It doesn’t matter how much you exercise if you sit a lot.” But that isn’t what the study says.

The key section from the abstract includes this sentence:

All-cause mortality hazard ratios were 1.02 (95% CI, 0.95-1.09), 1.15 (1.06-1.25), and 1.40 (1.27-1.55) for 4 to less than 8, 8 to less than 11, and 11 or more h/d of sitting, respectively, compared with less than 4 h/d, adjusting for physical activity and other confounders.

Let’s go through these numbers slowly.

The researchers took the mortality rate of people who sit less than 4 hours per day as their baseline and compared the mortality rates of longer-sitters to it. The mortality of people who sit 4-8 hours per day was, on the average 2% higher than the baseline. But this is just an average; the 95% confidence interval of the mortality ratio was 0.95 to 1.06. Because this interval includes 1, there was no statistically significant difference between 0-4 hour sitters and the 4-8 hour sitters.

The mortality of people who sit 8-11 hours per day was, on the average, 15% higher than the baseline. The 95% confidence interval of the mortality ratio was 1.06 to 1.25. This shows a statistically significant difference between the 0-4 hour sitters and the 8-11 hour sitters.

The mortality of people who sit more than 11 hours per day was, on the average, 40% higher than the baseline. The 95% confidence interval of the mortality ratio was 1.27 to 1.55. This shows a statistically significant difference between the 0-4 hour sitters and the 11+ hour sitters.

So far, so good. In choosing to report the 15% figure, the Times was picking the average number associated with a sitting 8-11 hours per day, which is fair. But where did that second paragraph come from, the one everyone picked up on? It came from this part of the abstract:

The association between sitting and all-cause mortality appeared consistent across the sexes, age groups, body mass index categories, and physical activity levels and across healthy participants compared with participants with preexisting cardiovascular disease or diabetes mellitus.

In other words, the 15% mortality increase that comes from longer sitting is as true for women as it is for men, as true for younger people as it is for older, as true from thinner people as it is for fatter, as true for exercisers as it is for non-exercisers, and so on.1

But that doesn’t mean that men who sits less than 4 hours per day have a lower mortality rate than women who sit 8-11 hours per day. It doesn’t mean that older people who sit less than 4 hours per day have a lower mortality rate than younger people who sit 8-11 hours per day. It doesn’t mean that fat people who sit less than 4 hours per day have a lower mortality rate than thin people who sit 8-11 hours per day. It doesn’t mean that diabetics who sit less than 4 hours per day have a lower mortality rate than non-diabetics who sit 8-11 hours per day.

And, most pertinently, it doesn’t mean that non-exercisers who sit less than 4 hours per day have a lower mortality rate than exercisers who sit 8-11 hours per day.

The consistency mentioned in the abstract is within groups, not between groups. The study says that sitting less reduces your risk, even if you’re a big time exerciser, even if you’re thin, even if you’re young. But it doesn’t (necessarily) trump exercising, being trim, or being young.

So, by all means, keep your standing desk, but don’t cancel your gym membership.


  1. Actually, I’m sure the 15% figure isn’t the average mortality ratio for all of these subgroups, but it is approximately the ratio for all of them. 


2 Responses to “The risk of sitting”

  1. Lri says:

    The amount of time the participants spent sitting or inactive wasn’t allocated randomly in the second study though. The differences would be much smaller if other (or more) factors that correlate with both inactivity and mortality were eliminated.

    Even if the time spent sitting was (somehow) randomized, you still couldn’t say that it’s sitting that increases mortality, but the kind of physiological states associated with average sedentary activities. The average physiological state measured during participants watching TV passively would be completely different from working actively on a computer or sitting in meditation or whatever.

  2. Boerma says:

    Nieuwenhuis et al. reviewed articles in the top neuroscience journals and found that half of the papers that use statistics, do it wrong. This stuff is hard even for (a subset of) the people that have had formal training in it, I wouldn’t expect NYT to get it right — much less the general ‘blogosphere’.