Because there's side effects to this, right?

And we allow it.

The fact is that we allow this to happen.

So this is a very interesting trial

where they looked at positive trials

versus negative trials by year.

So before year 2000,

what you see is that a lot of trials were positive.

57% of trials were positive.

After year 2000, only 8% were.

And what was the difference?

That was the year that everybody had to register

what their endpoints were.

So that means there was no more cheating

because what used to happen before 2000

is that you'd finish a trial

and say your outcome was heart attacks.

But you look at the data and says,

oh, the heart attack data's negative.

So let's make it heart attacks and strokes and hospitalization.

Does that work?

And it's like, no.

...