Several people cited the AHS-2 as a pseudo-RCT that supported veganism (EDIT 2023-10-03: as superior to low meat omnivorism).
[…]
My complaint is that the study was presented as strong evidence in one direction, when it’s both very weak and, if you treat it as strong, points in a different direction than reported
I think this is wrong in a few ways:
1. None of the comments referred to “low meat omnivorism.” AHS-2 had a “semi-vegetarian” category composed of people who eat meat in low quantities, but none of the comments referred to it
2. The study indeed found that vegans had lower mortality than omnivores (the hazard ratio was 0.85 (95% CI, 0.73–1.01)); your post makes it sound like it’s the opposite by saying that the association “points in a different direction than reported.” I think what you mean to say is that vegan diets were not the best option if we look only at the point estimates of the study, because pescatariansim was very slightly better. But the confidence intervals were wide and overlapped too much for us to say with confidence which diet was better.
Here’s a hypothetical scenario. Suppose a hypertension medication trial finds that Presotex monotherapy reduced stroke incidence by 34%. The trial also finds that Systovar monotherapy decreased the incidence of stroke by 40%, though the confidence intervals were very similar to Presotex’s.
Now suppose Bob learns this information and tells Chloe: “Alice said something misleading about Presotex. She said that a trial supported Prestotex monotherapy for stroke prevention, but the evidence pointed in a different direction than she reported.”
I think Chloe would likely come out with the wrong impression about Presotex.
3. My comment, which you refer to in this section, didn’t describe the AHS-2 as having RCT-like characteristics. I just thought it was a good observational study. A person I quoted in my comment (Froolow) was originally the person who mistakenly described it as a quasi-RCT (in another post I had not read at the time), but Froolow’s comment that I quoted didn’t describe it as such, and I thought it made sense without that assumption.
4. Froolow’s comment and mine were both careful to notice that the study findings are weak and consistent with veganism having no effect on lifespan. I don’t see how they presented it as strong evidence.
[Note: I deleted a previous comment making those points and am re-posting a reworded version.]
I think this is wrong in a few ways:
1. None of the comments referred to “low meat omnivorism.” AHS-2 had a “semi-vegetarian” category composed of people who eat meat in low quantities, but none of the comments referred to it
2. The study indeed found that vegans had lower mortality than omnivores (the hazard ratio was 0.85 (95% CI, 0.73–1.01)); your post makes it sound like it’s the opposite by saying that the association “points in a different direction than reported.” I think what you mean to say is that vegan diets were not the best option if we look only at the point estimates of the study, because pescatariansim was very slightly better. But the confidence intervals were wide and overlapped too much for us to say with confidence which diet was better.
Here’s a hypothetical scenario. Suppose a hypertension medication trial finds that Presotex monotherapy reduced stroke incidence by 34%. The trial also finds that Systovar monotherapy decreased the incidence of stroke by 40%, though the confidence intervals were very similar to Presotex’s.
Now suppose Bob learns this information and tells Chloe: “Alice said something misleading about Presotex. She said that a trial supported Prestotex monotherapy for stroke prevention, but the evidence pointed in a different direction than she reported.”
I think Chloe would likely come out with the wrong impression about Presotex.
3. My comment, which you refer to in this section, didn’t describe the AHS-2 as having RCT-like characteristics. I just thought it was a good observational study. A person I quoted in my comment (Froolow) was originally the person who mistakenly described it as a quasi-RCT (in another post I had not read at the time), but Froolow’s comment that I quoted didn’t describe it as such, and I thought it made sense without that assumption.
4. Froolow’s comment and mine were both careful to notice that the study findings are weak and consistent with veganism having no effect on lifespan. I don’t see how they presented it as strong evidence.
[Note: I deleted a previous comment making those points and am re-posting a reworded version.]
more discussion of this comment on LW.