The fable of the superiority of randomized studies no longer holds up against the reality of observational studies

[GuyGadeboisTheBack]

How would randomization be superior?

Nothing, but he had to say it.

[ABC2019]

But that does not prevent that if there is no fraud, the randomized study is superior.

The observational study, that is to say the observation of the reality of the population, does not automatically include randomization ... by nature?

How would randomization be superior?

because it neutralizes all the confounding factors: by drawing lots in the SAME control group, all other parameters are statistically equal except the one you want to test.

While if you compare the mortality of young people who came on foot to be tested at the IHU with that of old subclaquants hospitalized in Paris, well you don't just have the treatment that changes, you have plenty of confusing factors.

If you haven't understood this basic principle, or don't admit it, it can be argued forever ....

[Janic]

The observational study, that is to say the observation of reality, does not automatically include randomization ... by nature?

of course!

[GuyGadeboisTheBack]

If you haven't understood this basic principle, or don't admit it, it can be argued forever ....

Except that the reality is quite different. Reread (if you have read it, which I very much doubt) carefully this paper:
https://www.cochranelibrary.com/cdsr/do ... anguage=fr

[Obamot]

Of course we can do a randomized study, especially since it must be tempting, they are more expensive to carry out than the others. But it's at least as easy to pipeaut an observational study, and easier to screw up, because generally one will not be in control of the whole environment or of the method of acquiring the data (see the Lancet cases). .
What is pipeau is the argument, which is HS. If we compare these two types of studies, we must assume that they are honest. The interest of the observational is that it has a larger field than the other. We cannot test the effect of a harmful substance on health by injecting it into patients (unless we are called Mengele), while we can do an observational study on those who have been. accidentally victims.

Agree with one detail. Everything is debatable but I do not know if you will agree with the fact that observational studies are generally done with little means (generally academics not subsidized by the big labs who want randomized ones) and at the end of the day. the study, only data but not big interests to the key. So no interest in defrauding (or considerably less, except in losing your reputation ...?)

[ABC2019]

The observational study, that is to say the observation of reality, does not automatically include randomization ... by nature?

of course!

Obviously not, unless we do not understand what "randomization" means ...

[ABC2019]

If you haven't understood this basic principle, or don't admit it, it can be argued forever ....

Except that the reality is quite different. Reread (if you have read it, which I very much doubt) carefully this paper:
https://www.cochranelibrary.com/cdsr/do ... anguage=fr

what the paper says is that in 11 out of 14 studies, there is no significant difference between randomized and observational studies.
It definitely doesn't mean that the randomized ones are worse than the observational ones, it just means that they are often useless - that's exactly what I said.

But if there is a divergence of results, that can't apply, and I don't see how a randomized study could not show an effect when there is one.

[Obamot]

No, no and no! You know you're a smart guy, you pretend that clinical tests can be equivalent regardless of the approach you choose: “observational VS randomized”. But this is not the case and you know it very well, the trick here is that the NEEDS are not the same ... If it is in a hospital, the randomization is there automatically since it is dictated by the pathologies (yes you quibble but you will see further that if you have overwhelming results in one direction or the other then the interference created by this type of uncertainty becomes very weak). And since the placebo aspect also applies to those to whom we give the active molecule, or not (you can always try to challenge Tonton'lol), there is no bias. Why? Well, because in general we do a double-blind randomized study with molecules with mixed to poor results at the base, not when you have success rates on a large scale, in the whole spectrum of generations (the cases where the molecule matches almost all the time with convincing results in all respective samples if you prefer) I can give concrete examples. So your remark has no scientific interest since all this is already known and archi known from Methuselah

[ABC2019]

you are right

[GuyGadeboisTheBack]

Oh yeah...