Discussion about this post

User's avatar
Wayner's avatar

The article should have begun with "What they found was a rather modest increase in cardiovascular events among those who consumed artificial sweeteners. Indeed, it barely reached statistical significance for the composite measure:" At that point the article should have ended before the good doctor damned Diet Coke because he didn't like the taste. The article is mostly piffle.

Expand full comment
Zach's avatar

I certainly say agree with your view on refined sugar and starch: it’s important that medical professionals can be honest about how unnecessary and often harmful so many of these foods are when consumed regularly in the diet.

I’m less certain about the impact of artificial sweeteners. Without a doubt swapping refined sugar for sweeteners lowers both the total calories and amount of refined sugars in the diet, which is of plausible benefit.

In contrast, any negative health effects of artificial sweeteners is largely speculative. None of the basic research studies demonstrate a definitive mechanism. And the data in these associative epidemiology studies is simply not that convincing.

Take this study for instance. The trial protocol lists 16 (sixteen!) different primary endpoints: https://clinicaltrials.gov/ct2/show/NCT03335644

Whereas the main findings of this paper refer to just one of those primary endpoints (“CVD”), and the selection of artificial sweeteners is not even included in the trial protocol (because they collected data on many different foods).

Statistically this type of study design is largely a “fishing expedition” which is prone to produce a lot of random “statistically significant” effects. I just don’t see how you can get around the multiple comparisons issue when the authors can mine such a vast data set and there are literally thousands of possible comparisons that are eligible and there is no one policing how they are adjusting for this statistically.

So artificial sweeteners might be a problem, but I don’t currently see a plausible reason why they are worse than straight up pure sugar. And, it doesn’t really matter how many epidemiology studies people publish, the data simply point to the reason why we use RCTs in the first place: the ability to differentiate random effects from real ones.

Expand full comment
15 more comments...

No posts