Sunday 13 May 2018

Misleading labelling

This is what the public health people want to put on your next bottle of beer or wine:
As always, it's all-source mortality that will matter most where alcohol protects against some stuff and increases the harm of other stuff. And the best estimate on that is still a small reduction in all-source mortality among those drinking about a standard drink per day.

The health warning labels are then highly misleading. It's true that most people will get little or no health benefit from low levels of drinking. But it's also true that most people will not get cancer from drinking, even at high levels of drinking. Small changes in things that are low overall risk are like that.

Even that study that just came out that everyone was trumpeting as showing that there is no benefit from light drinking showed that there is in fact a benefit from light drinking. The researchers were a titch misleading about it in the press release and in the headline charts, but they did put the real result in an online appendix. Where they'd made the J-curve disappear by truncating the x-axis, it comes back if you show all the data.

It's illegal to advertise things on products that aren't true. If you put a label on saying that something is New Zealand made, you'll get in trouble if the thing wasn't actually New Zealand made. But what about things that are highly misleading? That set of labels together provides a pretty misleading overall perception about the health risk of drinking.

If we wanted truth in labeling, we'd mandate this chart on every bottle instead. Compared to not drinking, there is a reduction in overall mortality risk with light drinking. Drinking about 3 drinks per day is as risky as not drinking. And drinking more than about 3 drinks per day is riskier than not drinking. And if you were really keen, you could flatten the curve a bit for whatever lifestyle confounding you think is still in there, making the base of the J a bit closer to the zero-line, but also flattening out the other end of the J-curve too.

No comments:

Post a Comment