This ridiculous nonsense is going to be a blog soon, but a brief thread on why this new study does not show that "wine and chocolate can help you live longer"pic.twitter.com/2HCHweEdJB
You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
So, first we find the study itself (and get past the paywall...)https://onlinelibrary.wiley.com/doi/abs/10.1111/joim.12823 …
The first things I do when I see large epi studies of this nature with seemingly ridiculous headlines are: 1. Check to see what confounders are controlled for 2. Compare groups on baseline characteristics 3. Calculate absolute risks
If you go to the study, you'll see that quite a few confounders are controlled for - this is good news! However, there are few measures of socio-economic status, and none for ethnicity, which is a potential issue
Granted, this study was conducted in Sweden, so ethnicity may be less of a bias, but it's still concerning not to see it mentioned
On the second point - we look at baseline characteristics, and now it seems that we may have a problem People in the high-scored group seem to be MUCH healthier than people in the low grouppic.twitter.com/qfAsRqcmsk
They exercise more, they smoke less, they have lower BMIs, take more dietary supplements, are more educated, etc This means that there is almost certainly residual confounding that this study cannot address
Now on to our final quick check And it is a biggie Calculate those absolute risks!
For reference: RELATIVE risk - as reported most often by the media - is (risk of eventA)/(risk of eventB) ABSOLUTE risk - which is more useful, generally - is (risk of eventA) - (risk of eventB)
In this case, the RELATIVE risk reported on in most media stories is 18%, which sounds big The ABSOLUTE risk? Well...
The highest quartile of intake in this study had 2523 cases in 180896 person-years, or a rate of 1.4% The lowest quartile of intake had 7308 cases in 472726 person-years, or a rate of 1.5% That is an ABSOLUTE risk difference of 0.1% Not very impressivepic.twitter.com/4rHgOZDQfy
So, to recap: 1. Decent controlling for confounders, but some issues 2. Likely residual confounding 3. Very large difference between absolute and relative risk Even without further analysis, this is enormously problematic
If you add on some of the other issues - for example, the score that this study calculated to determine 'anti-inflammatory food intake' was a bit slim - the media noise becomes that much harder to believe
As with so many other studies, this is an interesting piece of work that may have implications at the population level - although I'm skeptical - but definitely is not very applicable to everyday life Blog to come!
P.S. worth mentioning that the 'quartiles' of intake were super weird: 0-5, 6, 7, 8-13 With such large bottom and top groups, I'd guess that the analysis was problematic although not my area of expertise
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.