[aatd86]

It's a well-known post-operation side effect that affects a few people transiently.

Maybe it's my fault for not being specific but in context, I was talking about long-term, possibly irreversible complications. While what you point out exists (every psychoactive drug has undesirable side-effects in general, even coffee can be dangerous), it is incomparable to the rate of occurences of side-effects such as tardive dyskinesia for neuroleptics for instance.

That's why I also added "in general" ...

[MichaelZuo]

> Maybe it's my fault for not being specific but in context, I was talking about long-term, possibly irreversible complications.

How is significantly increased rate of dementia among the affected population not long-term? It seems irreversible too?

[aatd86]

It clearly says that the link has not been established and that it is a risk. Not only that, it's more of a risk in older patients.

There is no formal causality link here.

Unless you can find other studies. And even then, methodology matters.

There is a realm of difference between that and the type of side-effects that were the subject of the initial conversation, to the point that I'd be tempted to say it is a red-herring.

[MichaelZuo]

I'm fairly certain it's possible to discuss the possibility and likelihood of things in the medial context without there being a 'formal causality link'.

From a random sample of HN user's comments in the last year in some context or another this is a well established usage in a supermajority of them. So if you must insist on it, you likely will always be outvoted, at least with other HN users.

[aatd86]

I don't understand what you mean respectfully.

If it's about statistics and sampling methodology, I am afraid we are veering toward being quite off-topic.

I'll say that my comment was toward paying attention to the different kind of study methodologies, such as metanalysis vs randomized controlled trials and even within the latter, there are differences in qualities depending on how the control group is ensured. As opposed to experimental groups as well.

Then of course sample size.

All that because cross-correlations are possible and correlation is not causation in general.

So yeah, in the end one can ask on hackernews but that will not be statistically relevant. Worse there is the bias that people who have undergone such procedures without issue won't report most likely. So really that's not a way to collect scientific evidence.