← More or Less
More or Less
Does playing tennis make you live 9.7 years longer?
00:08:58
Should we all take up tennis? We delve into the evidence on its life-extending powers.
Loading summary
Transcript29 lines
- [00:00]
A
Hello and thanks for downloading the More or less podcast. We're the programme that helps you take a swing at suspicious numbers, challenge data that seems clearly out and return statistical sanity. I'm your umpire, Tim Harford. Recently a post did the rounds on Twitter that claimed that if you do this, you might live 9.7 years longer. First things first, the claim does come from a peer reviewed paper by a group of doctors and scientists from the Mayo Clinic. The team used data from Denmark to follow a group of people and try and work out which sport produced the most gains in years lived. They looked at eight different sports and found that people could expect to gain
- [00:47]
B
9.7 years of life from playing tennis, 6.2 for badminton, 4.7 for soccer, 3.7 for cycling, 3.4 for swimming, 3.2 for jogging, 3.1 for calisthenics and 1.5 for health club activities.
- [01:05]
A
That last one sounds ominous. So tennis could extend your life five years more than jogging could. Or could it? People on the Internet were quick to point out that doing this. Is in many countries an expensive and elitist sport. Now, it is a well established fact that people who are wealthier and have a higher level of education tend to live longer. So the commenters argued, are our balls in the wrong court? Is longevity just an artifact of socioeconomic status or is tennis truly responsible for longer lives? To answer this, we need to look into the data. So in the spirit of Wimbledon, we've done just that.
- [01:56]
C
So I'm Adam Jacobs, I'm a medical statistician. I've been working in clinical trials for about 30 years.
- [02:02]
A
As far as these types of studies go, Adam says the one this claim came from is quite a good one.
- [02:08]
C
So there are several good things about this study. See, it's a large study. They had over 8,000 participants in it and they had a long follow up time, which is really good. They followed people for 25 years, which means that a lot of people are going to die in that time. So you get good data on all cause mortality, which was their primary outcome measure. And that's a good primary outcome measure because it's unambiguous and everybody knows it's relevant.
- [02:37]
A
But what of the twit's earlier point that tennis is often played by people with university degrees and exceptional healthcare, who by the way, tend to live longer. How do we make sure that someone's socioeconomic status isn't clouding things?
- [02:53]
C
Yeah, so that's called confounding. So that may explain some of the association and it probably does. I don't think it's as big a deal as I first thought it was when I saw the paper. They do an unadjusted analysis where they do not adjust for socioeconomic status.
- [03:11]
A
They do this through witchcraft and wizardry, AKA hazard regression models.
- [03:16]
C
Well, that's statistical magic. Basically what we do is we analyse it in a statistical model. You can do a statistical model, and this one, the thing is called Cox Proportional hazards regression, and that's a very widely used statistical model for analyzing survival data. And that allows you to just plug these confounding variables into the model. And, and what it tells you is what would the effect be if all the confounding variables were the same between the people who play tennis and people who don't play tennis? And in that analysis, they found that the people who played tennis got 10.4 more years of life expectancy than the people who didn't play tennis. They then adjusted that for education and income and a couple of other things, and that went down to 9.7.
- [04:07]
A
Now, the model wasn't perfect. They lost almost a third of people from the tennis group. We don't really know why this happened, but it's likely to do with insufficient data. Still, if you look at the group before and after the attempted flattening of people's education levels, wealth and so on, the numbers are surprisingly similar. The group of tennis players, without adjustments, lived an extra 10.4 years. And as opposed to 9.7. Not a huge difference.
- [04:36]
C
That tells me maybe it's not a massively strong confounding here, which surprised me.
- [04:44]
A
This could be for two reasons. The first being it's very difficult to truly adjust for confounders.
- [04:50]
C
And when I say they've adjusted for education and income, they have, and that's a good thing to have done. But they are quite difficult things to adjust for accurately. They tend to come in quite broad categories. So, you know, you might have somebody who's got a degree, somebody who's got postgraduate education, might be two of your categories. Well, if you've got a degree, you might have just scraped a third in a Mickey Mouse subject from a not very good university, or you might have a first class degree in physics from a top university, and those would be in the same category. And similarly for the income levels, they're divided into quite broad categories. You're not really capturing the full range of variation, but, you know, you're certainly capturing some of it. And that's better than nothing.
- [05:39]
A
Better than nothing, but still a relatively blunt tool. But when they didn't adjust for confound us, the number was 10.4. Not hugely different. So can we say that playing tennis makes you live 9.7 years longer?
- [05:54]
C
No, we cannot say that.
- [05:55]
A
Oh.
- [05:56]
C
So there are a number of reasons why we can't say that, but I think the biggest one is, is the sample size.
- [06:02]
A
The bigger the sample size, the less likely results are just down to something random or luck. Larger groups make it easier to spot traceable patterns and have smaller margins of error.
- [06:13]
C
Now, I just said we had 8,000 participants plus in the study, but most of those didn't play tennis. There are only 167 participants who played tennis and then in their adjusted analysis, they only had 111 people who played tennis and only 16 of those died.
- [06:33]
A
That is not a big number. The deaths were also all cause mortality. So we don't know how they died.
- [06:40]
C
So we're talking about quite small numbers. So you could get a lot of random variation. It may just be that they were lucky. Now, in that study, the participants who played tennis did well, but that doesn't mean to say that they would always do that. They may just have been lucky.
- [06:58]
A
The people who played tennis also skewed younger, which they did try to account for, but it's difficult to ignore the full effect. There is another thing going on, one that might explain the difference even more. The study looked at eight different sports. They asked people how many sports they did. The people who were in the tennis category also reported playing other sports.
- [07:23]
C
It may possibly be, and I'm speculating here, that it's having a good mix of sports, because this study was done in Denmark where a lot of people cycle and in fact, something like 80% of the people who played tennis also cycled.
- [07:41]
A
So perhaps it's the combination of exercise that's important or the fact that they're generally sportier people that's important.
- [07:48]
C
And maybe it's the variety in that kind of exercise that's important because they're two very different sportiers sports. With cycling, you're generally doing sustained but not hugely intense effort over a period of times, unless you're in the Tour de France and cycling up Mont Ventou or something. But for most people it's sustained, moderate activity, whereas tennis is quite short bursts of rather intense activity. So it's a different kind of exercise.
- [08:17]
A
And the people who played tennis spent more time doing other sports than tennis. So is Adam going to come away from analysing this study keen to put on his whites to extend his life
- [08:28]
C
no, no, but I might do more cycling.
- [08:31]
A
Thanks to Adam Jacobs and also thanks to statistical Hawkeyes James Thomas, Dylan Neal and Alison o' Mara Eaves. New balls, please. And that's all we have time for this week, but if you see anything you want us to check, please write in to more or lessbc.co.uk until next week. Goodbye.