Relative to abstinence from sweeteners, not relative to an equivalent sweetness of sugar or other sweeteners.
If I had to resolve today I would do 75%.
Background:
"The sugar substitute erythritol shortens the lifespan of Aedes aegypti potentially by N-linked protein glycosylation"
https://pubmed.ncbi.nlm.nih.gov/32277123/
"Elevated Erythritol: A Marker of Metabolic Dysregulation or Contributor to the Pathogenesis of Cardiometabolic Disease?"
https://www.mdpi.com/2072-6643/15/18/4011
"Higher blood levels of the artificial sweetener erythritol were associated with increased risk of heart attack and stroke."
https://www.nih.gov/news-events/nih-research-matters/erythritol-cardiovascular-events
"The top 25% had twice the risk of heart attack, stroke and death compared to the bottom 25%,’ says Stanley Hazen, a cardiologist at Cleveland Clinic, Ohio, who led the research. ‘That places it on par with the strongest other cardiovascular risk factors, such as diabetes.’"
https://www.unmc.edu/healthsecurity/transmission/2023/03/07/study-shows-artificial-sweetener-erythritol-linked-to-increased-stroke-and-cardiac-risks/
"The sugar-free paradox: cardiometabolic consequences of erythritol"
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10275890/
There are no relevant RCTs of erythritol on human cardiovascular risk that I can find. Probably no one can get an RCT of increased erythritol consumption past an IRB if they suspect it of increasing cardiovascular risk, even though the potential harm of such a trial is in expectation infinitesimal compared to the ongoing harm of not doing the trial.
I think one should have a high prior probability that anything which is a known agonist of taste-sweetness receptors might also be an agonist of other glucose receptors throughout the body, and that the latter could be problematic. In the case of erythritol it looks like there is negligible uptake into cells so that limits the potential for harm.
Had a convo with gpt4o about this. So I was thinking the long-term studies are just observational and selection effects could be masking a small harm. It’s pretty standard the stuff that has hype about being healthy gets more health-obsessed people using it and then it shows a fake benefit on the observational studies even if it’s a complete placebo. GPT couldn’t really contradict that. However, the big update was that erythritol doesn’t have any aldehyde or ketone groups, which are the most reactive sites on other sugars, so it has less potential for cross-linking reactions.
@JonathanRay Otoh the average human has only 4.5g of glucose in their blood at any time and you can get 4x that from one erythritol product and the elimination half life is circa 8 hours (extrapolated from Wikipedia saying 80-90% elimination after 24 hours). So if one uses a couple erythritol products a day then average plasma erythritol levels could be higher than average plasma glucose levels. One would need to quantify how much less reactive it is
@JonathanRay Reactivity ranged between 1% and 10% as much in studies according to gpt4, and it’s not significantly taken up into cells which simplifies the plasma concentration math. 14g/day erythritol would result in average plasma levels identical to that of glucose. If one was consuming 100g/day it could be very problematic if we take the upper bound of that reactivity range