Thread by @GidMK

Headlines: Diet soft drinks linked to cancer, death. DRINK THEM AND DIE

Study:

- observational research
- small increased absolute risk even when increasing drink consumption by >6000%
- LOWEST RISK WAS FOR PEOPLE WHO DRANK 100-200ML OF DIET DRINKS A DAY

Study is here: jamanetwork.com/journals/jamai…

Basic story - scientists looked at ~a lot~ of people (450,000) for a long time, and split them up depending on how many soft drinks they drank

At the end of 16 years of follow-up, those who drank >2 soft drinks a day had an increased risk of many bad things (cancer, death, cardiovascular disease) than those who completely abstained

This was true for sugar and artificially sweetened drinks

The authors used standard Cox regression models, which allowed them to control for a range of factors

They concluded that there's a likely association between soft drink consumption and BAD THINGS

However

Firstly, the absolute risk increase was pretty small across the board

https://twitter.com/justsaysrisks/status/1168995180797747201?s=20

In other words, if someone went from drinking 1 can of Diet Coke a YEAR to 2 cans A DAY they would increase their risk of, say, cancer, by less than 1%

Secondly, THE RISK INCREASE WAS NON-LINEAR

In practical terms, this means that the people who were at the lowest risk were people who drank roughly one can of Coke/Diet Coke every 2 days

If we're going to be consistent in using this research, we should encourage modest soft drink consumption, which I find absolutely hilarious

This brings us to arguably the most important point - this research was OBSERVATIONAL

Now, observational research can be incredibly useful and meaningful for public health, but there are also some real difficulties in drawing causal conclusions from studies like this

In other words, we don't know if the increased risk of Bad Things was due to soft drink consumption or something that the study didn't measure

For example, we know that ethnicity and income can change someone's risk of death, but the study didn't control for these things at all

Could be that people who drink more soft drinks are just less well off than those who don't!

Another hilarious thing about the study was that for some diseases, soft drinks were REALLY protective at low doses

For example, drinking an extra Diet Coke a week REDUCED your risk of breast cancer by 21% 😂

Realistically, it's more likely that there are some significant confounders still remaining so we just don't know whether soft drinks = death/cancer from this study

Also, for the epi nerds, the study authors attribute the reduction in risk at low doses to reverse confounding, but I find this argument unconvincing

Firstly, they did a specific sensitivity analysis to try and control for reverse causality which didn't change the results, so their own results don't agree with this interpretation

Secondly, this is absolutely a non-differential effect, so you'd expect it to apply equally across all groups of the study rather than only reducing the observed risk in one particular group of people