----
If I'm reading correctly, that 2015 MR paper establishes that LDL and triglycerides play a role in heart disease by filtering populations used in prior studies for genetic blood lipid factors. But it doesn't make any claims about alcohol. Is there already a well established link between alcohol and LDL / triglycerides?
My impression is that the relationship between LDL, HDL, and triglycerides and what we eat/drink is not well understood. But I'm not involved in biology or medicine, so maybe you can help me understand this better?
Do you happen to know if rs1229984 is linked to any particular group that could make this result problematic? I noticed that the paper mentions that "There was also some evidence for a difference in years of education, and, while the size of the effect was small, this requires further investigation." That sounds like a decent-sized red flag.
Take a look at the population frequencies on the right. While the minor allele is infrequent in Europeans, it is extremely common in East Asians, for example.
It seems to me that we would not have to know about the link in order for the conclusions to be unfounded. These MR analyses rest on the assumption that the instrument is not linked to any potential confounders. This seems like a bold and unwarranted assumption when we are talking about an instrument linked to behavior, given the complexity of the chain of events (chemical, biological, and psychological) between gene expression and observable human behavior.
That education red flag involves just one possible confounder out of many that were never considered.
That's a very convenient way to socially forbid replication attempts. I don't buy it. The argument basically amounts to a claim that we answered the question definitively in 2014 and anyone questioning that result now is unethical. Today's scientific research, especially in any area pertaining to people, is too unreliable to blindly trust this way.
Declaring this level of risk to be outside the bounds of permissible research is an attitude incompatible with scientific progress. It's elevating short-term harm prevention to an unjustifiable extreme that makes it impossible to determine whether we have an accurate model of the world.
My reading (not just [1], although [1] is a lucid examination) suggests to me that IRBs are net detrimental to research. There's an incentive misalignment that pushes them too far in the "harm" prevention camp and that makes it too difficult to answer scientific questions.
That's too bad. The rest of us try to learn something from Nuremberg.
So I guess what I'd say is that it looks like the one thing you know about IRBs is that Scott from SlateStarCodex once had a run-in with them and generalized wildly from it.
I think neither of will profit from continuing this discussion, though. You can safely sum up my point as "the problems people in psychology, and the social sciences in general, have with IRBs do not necessarily generalize to biomedical research". One of the basic complaints scientists have about IRBs is the "mission creep" from human medical research to the rest of the sciences. The premise of that argument doesn't line up with your argument here on this thread.
Also, SlateStarCodex is bad.
My point is that IRBs have an incentive structure that encourages excessive risk aversion. My citation is an instance of this phenomenon in action. It has nothing to do with the particulars of psychology. You have not attempted to address this incentive alignment argument in any way. Blocking an attempt to replicate an alcohol study due to weak evidence for a very small negative effect is another example, albeit hypothetical, of this same excessive risk aversion. You have not made a similarly substantial argument. Instead, you've just asserted that institutional review is embedded in biomedical research. It is, but I think it shouldn't be.
> Also, SlateStarCodex is bad.
It's amazing that immediately after a paragraph in which you claim to be making a substantive argument, you make a bald and unsupported assertion that an influential and widely-read blog, one that Pinker has called "wonderful", is in fact "bad".
This is not the kind of intellectual conversation that HN was meant to host.
Maybe an interaction with the prevalence of leaded gasoline?
2. While we like to claim that a study randomizing people to alcohol wouldn't be ethical, pre-2014 I don't think that would have been true. I think there was enough equipoise. Post-2014, I would need to see some strong counter-evidence to make me think there is equipoise once again.
3. The environment changes. A trial came out last week showing that ASA may be harmful for older healthy people [2]. This probably reflects that our bloodstream is no longer nicotine-filled, etc., so the heart attack risk is low enough that the bleeding risk may outweigh it. There may be something about alcohol that was helpful in the 1950s that is no longer so today, due to these environmental changes.
1 = (Free full link): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4091648/
2 = https://www.nejm.org/doi/full/10.1056/NEJMoa1800722?query=fe...