> Unless a treatment is both miraculous in effect and works for everyone, it's hard to figure out if it works.
Yup. It's worth noting that "all or none" evidence is still considered category 1 evidence on many scales. (If you treat a group where all would be expected to die, and some survive... or a group where many would be expected to die, and all survive). It's only valid for the most dramatic effects, but you don't need randomization. During a safety trial you might come up with "all or none" evidence if your effect is strong enough.
But otherwise, you're going to need to compare the treatment to something else. There's no ability to magically draw the exact same population from some earlier trial.
Yup. It's worth noting that "all or none" evidence is still considered category 1 evidence on many scales. (If you treat a group where all would be expected to die, and some survive... or a group where many would be expected to die, and all survive). It's only valid for the most dramatic effects, but you don't need randomization. During a safety trial you might come up with "all or none" evidence if your effect is strong enough.
But otherwise, you're going to need to compare the treatment to something else. There's no ability to magically draw the exact same population from some earlier trial.