All those egg-bashing scientists are going to have egg on their face when this study,
which is no yolk, cracks the case on the dietary benefits of those calcium-carbonate covered
comestibles.
This ends the egg humor part of the presentation.
I need to joke a bit though, because honestly, it sort of pains me to talk about studies
like this.
But I need to talk about them, because other people talk about them, and I feel like as
a physician part of my job is to say no, don't take these studies seriously.
But before I rant, here's what you need to know: a huge study of 500,000 Chinese people
found that those who ate more eggs had less cardiovascular death.
So, are eggs back on the menu?
I don't know – these studies really don't tell us whether adding eggs or removing eggs
from your diet will make any difference.
I'll put it really plainly: Studies that use responses to a food-frequency questionnaire
to link to some health outcome are not worth the paper they are printed on.
And so let me use this recent egg study as an object lesson in the problems with dietary
epidemiology research.
First – no one eats randomly.
Except my two year old who asked for pasta this morning.
The rest of the world chooses what they eat based on a variety of social, economic, practical,
and gustatory factors.
These confounders can not be controlled for with simple statistical adjustment.
If I told you that American eaters of foie gras live longer than those who don't, would
you attribute that to the foie gras, or to the fact that these 1 percenters have access
to quality healthcare and other good things?
And, as I've discussed before, adjustment for things like "income range" does not
fully account for the complex socioeconomic web we weave.
Second – Eggs, like coffee, marijuana, wine, chocolate, and many other exposures are not
really one thing.
While the macronutrient composition of eggs is somewhat stable, the micronutrient composition
is all over the map.
When researchers say coffee protects against colon cancer, are they referring to black
coffee or a double-tall mocha Frappuccino?
When the exposure is muddy in this way, inferences about effects become much less reliable.
Third – multiple comparisons.
There are over 130 food items on the dietary health questionnaire.
The chance that one of those 130 items will appear to be statistically linked to any health
outcome is near 100% - I just have to try them all.
In fact, given no true relationship, there will be, on average, 6 or 7 items on the questionnaire
that nevertheless fall below our conventional statistical significance threshold of 0.05.
Fourth – I told you there are 130 items on the food frequency questionnaire.
But I can make even more.
I can combine items to calculate your total calorie intake, or fat intake, or magnesium
intake, or even your pesticide intake.
More potential exposures!
Fifth – many of these dietary studies use huge datasets, like the China Kadoorie biobank
with its 500,000 participants in the egg study.
This means it is trivially easy to find statistically significant effects that are not remotely
clinically interesting.
Even if you believe the primary findings of this study – that consuming an egg per day
reduces your risk of cardiovascular death by 18% compared to rarely eating eggs, the
absolute effect is tiny.
You'd need to treat nearly 800 people with daily eggs to prevent one cardiovascular death
per year.
That's a lot of umm…
The bottom line: what you put into your body matters, but you put a lot of stuff into your
body.
No one thing is going to keep you alive, and conversely, no one thing is going to kill
you.
So when your friend tells you that you should eat more eggs based on this study, remember
how Homer Simpson handled it: So the next time you see a study that uses
a food frequency questionnaire to make some inference about something you eat, remember:
it's all a shell game.
No comments:
Post a Comment