Comment by carbocation | Hacker News Reader

Comment by carbocation | Hacker News Reader

The (2015) label is very important, because people have had concerns like this, and consequently have directly studied things like:

Do statins (cholesterol lowering drugs) increase suicidality and depression? No (in fact, the signal was for benefit)[1].

In a causal genetic analysis, does a decrease in cholesterol cause an increase in depression? No (in fact, the only significant finding from the study was that an increase in triglycerides is causal for increased depression risk)[2].

Does driving cholesterol to extremely low levels with powerful new medications called PCSK9 inhibitors cause change in neurocognition? No, and while there were no significant differences between the PCSK9i group and the placebo group, the PCSK9i group had numerically superior results, so there is really no signal for harm[3].

The concerns raised in the article don't really stand the test of time. I think this is probably because they were cherry-picked to begin with.

1 = https://www.thelancet.com/pdfs/journals/lanpsy/PIIS2215-0366...

2 = https://www.cambridge.org/core/journals/psychological-medici...

3 = https://www.nejm.org/doi/full/10.1056/nejmoa1701131

While trials can determine that on average there is no adverse effect it does not necessarily imply that no individual would experience any negative effect. It is possible that the negative effects experienced by some are cancelled out by positive effects experienced by others. If the effect is reversible, one can prove causality by giving an individual both placebo and the drug, in sequence, without the subject knowing which is which.

Yes, a crossover trial design would be one way to address that. For example, various side effects have long been attributed to statins. This was tested in a trial published very recently in a placebo controlled crossover design, which found that the effect attributed to statins has probably been overestimated[1].

(They found no significant difference between placebo and statin; but, by eye, I would be willing to agree with a skeptic that it looks like there is probably a difference that is just too modest to identify at statistical significance at this sample size.)

I should add that the original article being discussed is making claims that would not require any crossover trial to identify. If lowering cholesterol doesn't cause numerically higher suicide or depression, then the author's point is invalidated.

1 = https://www.jacc.org/doi/10.1016/j.jacc.2021.07.022

I disagree. These drugs do a lot more than change cholesterol.

Reply on news.ycombinator.com