Wednesday, March 18, 2009

Obesity Oxford findings

As per my previous entry on GMTV a blip passed through about research that being overweight was as harmful as smoking. A quick look at the BBC and I decided to look at the sources. In this instance the Press Association which is where so many papers dredge their 'facts' that they regurgitate with padding.

So in pure terms what can we see - well the study dealt with individuals predominately from Western Europe and North America average age 46 average BMI 25.

Let's stop there for a second, the "ideal" (their quote marks too) BMI is between 22.5 and 25 so we're already dealing with the top end of the index. Secondly geographically we're looking at people who may well eat the same types of food. Okay continue.

Each additional five BMI units corresponded with a 40% increase in deaths from heart and artery disease and strokes, an increase of 60% to 120% from diabetes and liver or kidney disease, a 10% increase from cancer, and a 20% increase from lung disease.
Note the use of percentages rather than absolutes, I've commented on this several times, but to reiterate if the chances of something occurring are 1 in 10,000 and you do something that increases these odds to 2 in 10,000 that's a 100% increase. Saying that would you change your lifestyle to drop a 2 in 10,000 chance to 1 in 10,000?

Just on a quick Google I find the Odds on Dying this gives the chance of heart disease as 1 in 5 and stroke as 1 in 23. For both this is 140 in 575, from this report that increases to 196 in 575. For cancer it's 10 in 70 which increases to 11 in 70.

Anyway let's take these facts apart and connect them the way that the news media have.

BMI is a measure of ideal weight
If your BMI is above the ideal weight you are more likely to have health problems
Therefore being overweight gives you health problems.

Anyone nodding at that? Now let's connect it up a different way.

BMI is a measure of ideal weight
Eating certain foods can increase your BMI index
If you increase your BMI index you are more likely to have health problems
Therefore eating certain foods gives you health problems.

Same results different conclusion. In other words it may not be the 'fact' that you're overweight that causes health problems it may be that the food that you eat that coincidently makes you overweight causes health problems.

To quote XKCD "Correlation doesn't imply causation, but it does waggle its eyebrows suggestively and gesture furtively while mouthing 'look over there'." So am I saying weight isn't a factor? No I'm not, but by the same token I can make an educated stab that those who showed up as more likely to have health problems ate more fatty foods, or even performed less exercise none of which is linked by the media.

But what about the headline of the BBC story
Obesity danger 'rivals smoking'
With quotes no less, in the PA story we get
as dangerous as smoking
So hey let's do something radical and look at the Press Release (DOC format) from CTSU the ones who, you know, did the actual research. In bold right at the top
Moderate obesity takes years off life expectancy, though not as many as smoking
Um? Ah but wait further on
severe obesity, which is still uncommon, can shorten a person’s life by 10 years. This 10 year loss is equal to the effects of lifelong smoking.
See the media were right after all, well if you ignore the bit about it only applying to the uncommon severe obesity and that it's equivalent to lifelong smoking.

Oh and applause for
Though not perfect, BMI is useful for assessing the extent to which fatty tissue causes ill health
which is more than the media indicate who seem to suggest that BMI is the accurate method of calculating obesity. Okay the next bits confirm the whole 'overweight be bad' thing, but then we get to something not just glossed over, but non-existent in the media reporting.
There was also a higher death rate among those who had a BMI well below 23 to 24. This hazard was much greater for smokers than for non-smokers. The exact reason for this remains unknown.
So people who are overweight have a higher death rate than those of ideal weight, but also in those who are underweight. But whereas we know the causes of ill-health in the 'fatties' we're flummoxed as to the cause for the 'sticks'; Um again diet anyone?

Also worth noting that mention of 23 to 24 because that's where the lowest mortality rate was found despite this being the top end of the ideal 22.5 to 25 range. Oh wait if you check out the abstract from The Lancet
mortality was lowest at about 22·5—25
BMI was associated inversely with overall mortality, mainly because of strong inverse associations with respiratory disease and lung cancer
So the lowest mortality rate was between 23 and 24, which can be accurately said to be between 22.5 and 25; and we don't know why the underweight have higher mortality rates, except we do.

So to sum up - selective reporting of facts leading to a story that suggests information that isn't supported by the very study they're reporting. Think of this next time you get an ' X leads to Y' story in the news.


Anonymous said...

Heh. And to think I was under the misguided impression that it's illegal to report things that aren't true…

FlipC said...

No lying involved simply "selective reporting of facts"; white lies if you must, lies of omission.

Now if you want out and out lies in a paper you can turn to [removed due to libel issues] which fields the most complaints issued to the PCC and has had three times as many complaints upheld as its nearest competitor.

Anonymous said...

Oh, they're competing for PCC complaints now?

FlipC said...

Heh I don't think competing is the right word ;-)