Consider a standard treatment called Drug X that has efficacy in some Disease called A. Lets assume that if people with Disease A take Drug X that 50% live to 5 years and 50% will die from the disease.

Now consider a new treatment for Disease A called Drug Y that you think is similar in efficacy to Drug X but is perhaps cheaper, has less side effects, or is in some other way desirable. Your goal as the manufacturer of Drug Y is to get it approved for use in Disease A so you can make lots of money. How do you do it?

Option 1: Do a clinical trial of superiority of Drug Y over Drug X.

H0: Drug Y = Drug X (null hypothesis, both have 50% response rates)

Ha: Drug Y > Drug X by some margin, say 10% (alternative hypothesis, what you hope for))

This is a bad option since your trial will fail because the drugs work about the same.

Option 2: Do a clinical trial of non-inferiority of Drug Y against Drug X

H0: |Drug Y - Drug X| >10% (null hypothesis)

Ha: |Drug Y - Drug X|

It is used in trials which try to prove that the studied intervention is not unacceptably worse than the control. Look at the paper below:

http://www.trialsjournal.com/content/pdf/1745-6215-12-106.pdf

It's an unsavory, unacceptable means of adopting "placebo fraud" to clinical trials of new xenobiotics.

Where's the placebo arm in the safety analysis? There's no "true" placebo arm under the non-inferiority margin. Efficacy is moot if the safety has been wilfully "marginalized"!

I recommend this article on non-inferiority active controlled trials. It has been cited a lot in the work of some statisticians.

Thnx

Look up the CONSORT statement for equivalence/non-inferiority RCTs. There is a good explanation.

Consider a standard treatment called Drug X that has efficacy in some Disease called A. Lets assume that if people with Disease A take Drug X that 50% live to 5 years and 50% will die from the disease.

Now consider a new treatment for Disease A called Drug Y that you think is similar in efficacy to Drug X but is perhaps cheaper, has less side effects, or is in some other way desirable. Your goal as the manufacturer of Drug Y is to get it approved for use in Disease A so you can make lots of money. How do you do it?

Option 1: Do a clinical trial of superiority of Drug Y over Drug X.

H0: Drug Y = Drug X (null hypothesis, both have 50% response rates)

Ha: Drug Y > Drug X by some margin, say 10% (alternative hypothesis, what you hope for))

This is a bad option since your trial will fail because the drugs work about the same.

Option 2: Do a clinical trial of non-inferiority of Drug Y against Drug X

H0: |Drug Y - Drug X| >10% (null hypothesis)

Ha: |Drug Y - Drug X|

Thanks a lot Sir

Brant A Inman I believe that the following:

H0: |Drug Y - Drug X| >10% (null hypothesis)

Ha: |Drug Y - Drug X|

Good Explanation