Clinical Practice in NI studies

So let’s quickly talked about can superiority be determined in non-Inferiority studies. And the answer is yes. but again it has to be pre-specified. and it should look at the ITT data. So it should provide the effect size and we already talked about that power is going to be need to be higher in a non inferiority trial. So if they specify superiority with the effect size the power should be met because you need more patience in the non inferiority studying it than you do in there. Now if they do find superiority then, it’s good to calculate in NNT.

So you’ll have that information just go back and look at the session on measures of Association on how to calculate that and then determine what that means based on the the six key components within that study. Now can NI be assumed if the superiority study was not significant and hopefully you remember the answer to that is no. A superiority study can either find that one drug is superior to the other or inconclusive which is the not. The not does not mean that it’s non inferior. All right. So can authors try to prove NI posthoc after non significant results are concluded in a superiority study?

Because I just told you above that not finding significant difference does not mean it’s not inferior but you don’t want to do that because once you’ve got your study results it really biases you to determine what’s the most clinically relevant difference so unless it stated a priori for Delta and powered because remember you’re going to need more patience n an NI study than a superiority. So this is rarely done and it’s not as well accepted. So just remember if you test for non inferiority as long as it’s listed a priori you can then test for superiority if non-inferiority met.

You can also do it the other way but it’s more controversial even if it’s pre specified so usually the not and I followed by superiority is more acceptable Alright now let’s try to put this into practice and you go through a few examples to better help you to be able to visualize these and please make sure you go back and look at the different slides so that you can completely fill these out.

So here’s an excerpt from a clinical trial and I’ve tried to give you the key pieces of information so that you can graph this so remember you want to make sure that you have the Delta margin which is listed on the left you want to make sure you have the results and you want to make sure it what type of confidence interval it is. So you can see it’s grafted here you can see that it’s a measures so 1 goes here instead of a 0. You can see the Delta margin is positive so it goes on the right which means you list the favors drug versus control the way I have it listed there.

And then you graph the results and again I put them the zone non-inferiority. So hopefully you can clearly see from this that non-inferiority was met. And so is superiority and this is indicated also and confirmed by the two p-values that they gave you for both non inferiority and superiority. Now again looking at this it looks like it’s a fairly small confidence interval. So you can be more assured that the information was more reliable from the sample.

Now this doesn’t mean we’re not looking at other parts of the study but just looking at and determining whether or not inferiority was met from the way it set it up it appears that they did as well as superiority in this study but again we got a look at everything else in its entirety to be able to come up with conclusions All right now here’s a second one where we have the margin we have listed what the Delta is and we have the results. So we can see that there’s no RR there’s no HR don’t let CR confuse you because that’s clinical rate for cure.

So you want to make sure that you understand the clinical reoccurrences It’s not a form of HR or so our number that’s going to be there 0. Our Delta margin is positive so it’s going to go here so you can see then we set up the table favors agent on the left favors control on the right and you can see that this margin doesn’t doesn’t touch your cross 0 but it does cross the Delta. So this one is a little bit confusing because it’s obvious from and we put the zone of non-inferiority there.

So you can tell it’s clearly seen that it doesn’t meet non-inferiority but if you look at their results that they have listed there they have it listed as P is equal to 0.02. Now it claims superiority of the control agent which was the sulfa combination drug to the Cranberries but the p-value was listed with the non-inferiority. But hopefully also you noticed at this point that the confidence interval is really wide. So I really wouldn’t wanted to conclude too much from this but to show that Cranberries was really not non-inferior. I would have a little bit more difficult time saying that it was superior just because of the wide confidence interval. All right let’s look at one more.

This looked at vancomycin versus linezolid for infections. so you can see that they gave us the non-inferiority margin here and this one’s a negative value. So we’re going to set up and they did state that they look at superiority if if non-inferior was met. So this one is going to be listed on the opposite side. Remember I told you the Delta is going to be on the left if it’s negative maybe right if it’s positive and then you’re gonna have to flip your favors control favours test drug at the bottom. So you can see that it’s the zero because there’s isn’t measures.

There’s nothing listed anywhere that says CHR or RRR the Delta again is on the left because it’s a negative value and then you graph the results. Now they didn’t give us the difference which is the circle in the middle but you can just minus that was it was 11% and then you could see that the confidence interval is very wide 0.5 to 21.6%. Now technically non-inferior was shown but again that confidence interval is extremely wide. Again you’d want to compare it to what you see in other studies but I think you know irrespective of other studies that that’s probably a pretty wide one.

Technically superiority was also shown but they didn’t provide a p-value which you think was interesting that that they’d want to provide that. So you can see that this this chart was set up totally backwards than the other ones because it was a negative Delta Delta so you would have had a difficult time interpreting it. if you wouldn’t have set it up correctly.