Question: When doing pharmeceutical testing, what percentage is considered "effective"?

[Taverner]

As in, I have two groups who have malady X. I give one group a sugar pill, and the other the treatment in question, drug Y.

How many more people in the second group need to have been cured versus the ones cured in the control group for drug Y to be considered effective?

[ddeclue]

1. better than sugar pill outside of MOE.

[Taverner]

2. So if the margin of error is say 6%

Would that mean drug Y is sucessful, even if its only 1% or % over the MOE?

[ddeclue]

4. no outside means outside so it would need to be >6%

and MOE is a function of sample size - how many subjects were tested. Google "student's T"

[damntexdem]

9. This is the question: could the difference have occurred by pure chance?

Then some confidence level is used: 5%, or 1% -- that is, the difference could occur by chance in 1 out of 20 cases (5%), or 1 out of 100 cases (1%). Then, based on the sample size, how big a difference would be needed is computed to ensure that the difference would likely only occur by chance at whatever level is chosen.

Alternately, a desired difference can be chosen (e.g., 10% less or more of something than with a sugar pill, or with the standard treatment), and given the confidence level, the needed sample size computed. In other words, select a sample large enough that if the desired difference were to show, it would be statistically significant -- would only have happened by chance at the chosen level.

To take it to a different arena, notice the margins of error for different polls. Some are wider and some narrower, depending on the sample size.

[NC_Nurse]

3. I don't remember, but it's shockingly small. LOTS of drugs fail in the 3rd stage

because the standards are higher at that point and placebo will tie or beat them. Even if they make it - it's not by as much as you might think.
Sadly, I can't remember exactly where they draw the line. But I know it doesn't stop them from then advertising as if the drug DEFINITELY works.

[postulater]

5. Good question

and what about those who are made worse by the drug.

There should be a worst case risk factor assigned to all drugs and medical procedures so you know what you are gambling for when you have something done or take a drug.

If the odds are 99 to 1 that you will get better, but the 1 is dead, you might re-think your decision.

[BlooInBloo]

6. There's nothing remotely close to as simple and general an answer...

as you're looking for.

[Speck Tater]

7. I once read that...

The statistical significance of many ESP tests is over twice as high as the statistical significance of aspirin's effect on heart attacks, yet the aspirin result is accepted and the twice as strongly supported ESP is not. (Dean Radin PhD in ISBN 0061778990)

Go figure. It's almost like scientists aren't really as objective as they claim to be.

[Taverner]

8. Extraordinary claims demand extraordinary proof

Asprin has been used for centuries in folk medicine, so it wasn't too far a stretch