Skip to 0 minutes and 14 seconds So let’s look at again comparing superiority to non inferiority when you’re going to interpret the p-values. So a p-value greater than 0.05 in a superiority study says it’s not statistically significantly and there’s no difference. Whereas a non inferiority not statistically significance means that NI was not met. Right? Remember doesn’t mean it’s inferior. A p-value of less than or equals 0.05 in superiority studies tells you that one treatment is statistically significantly better than another whereas a non-inferiority, it just tells you that the experimental treatment is as good or not inferior to the control. Based on an a priori margin. So it tells you it’s non inferior.
Skip to 0 minutes and 59 seconds So again just to remember that NI study does not if it an NI study cannot show non-inferiority this doesn’t mean that it’s the treatment is inferior just like superiority trials can’t show non inferiority non-inferiority trials cannot find something to be inferior. All right so now we’d also just tell you to take a minute to look at power and how it differs and some and how it’s similar between what you see in superiority study. Now the power that is chosen or than at least the number of patients in power depends on the margin that’s chosen Now remember we talked about that margin is it’s not the same thing as the effect size it’s seen as superiority study.
Skip to 1 minute and 40 seconds It’s going to be smaller than that. It’s considered the minimally important effect that is used to calculate power and superiority studies. So if you know what the effect size is then that’s what the author’s will determine what could be the minimum amount that you will see. So you can see that wider margins can lead to a type 1 error. Narrower margins can lead to a type 2 error. So you want to make sure the margin is as potentially accurate as possible. So you’re not getting misleading results.
Skip to 2 minutes and 9 seconds And because you’re using a smaller the margin is smaller than the effect size, you’re going to need larger sample size It’s a normally an acceptable power and a non-inferiority trial is greater than equal to 90%. Remember greater than equal to 80% in superiorities studies. Although many would tell you for superiority studies greater than 90 is is better than 80% as well.
Skip to 2 minutes and 31 seconds But you want to make sure that you’re not the authors aren’t setting it up for them to be able to find it’s non inferior because it isn’t because of the way the study was designed Now just with superiority you can look at those four main components of power to see if maybe power is it might be in question of what they did. So here let me give you an example of a power calculation from a study. So if you look at the four components of power that are here that we talked about before. that you’re going to want to look at sample size at the endpoint.
Skip to 3 minutes and 7 seconds And now you can remember the power calculation it’s not going to give you the endpoint numbers so you’re going to have to look at that to see if it meets it in the final analysis of the patients. It needs to be linked to the primary outcome and then power has to be greater than 90 percent and then you need to check and make sure that the NI margin is appropriate about what’s listed in the literature.
Skip to 3 minutes and 27 seconds And again, I see the below that you need to make sure that power is higher in non-inferiority studies in superiority because you’re using a smaller Delta than the effect size and unfortunately you can see I told you that non-inferiority trials often have more discrepancies with their study design if there have been several studies that have checked different clinical trials based on the consort for both superiority and for non inferiority and there’s even more errors in non-inferiority studies than there are in superiority studies. So that’s why it’s important to make sure you analyze those key results because it’s easier to find out inferiority than it is to find superiority.
Skip to 4 minutes and 8 seconds All right so assessing the NI margin this is something that you should ensure that the authors did it’s nothing that you can do you’re just going to assess from what you found there but the author should assess it based from statistical reasoning standpoint as where as clinical judgment and balance these because if you just go with the chemical judgment that it’s probably not going to be statistically relevant and if you go with statistically relevant it probably is going to hinder the clinical judgment because you’re going to need so many patients it’s going to be almost impossible to do.
Skip to 4 minutes and 39 seconds So we remember we talked about zero and one being our number not to touch across when we were doing measures of association for superiority studies. Now those numbers will still be in our chart but now the new number or magic number or threshold that you don’t want to touch your cross to ensure non-inferiority is this Delta margin and again it should be preset specified all of these things have to be done a priori otherwise, once you see your results it’s going to be difficult to then go back and reset those because it’s going to bias you since you’ve seen what the results are So assessing the NI margin again you’re not going to necessarily do all of this as the reader but you need to make sure that they discuss the authors discuss is the NI margin appropriate and where did it come from.
Skip to 5 minutes and 27 seconds Now they first should start looking at meta analyses or systematic reviews of the drug that they’re comparing it to to placebo controlled trials to see what the effect sizes and the confidence intervals to see what they were and if those are available they should look at previous studies comparing the standard treatment for placebo and again checking the effect sizes and the confidence interval and again the margin should be clinically relevant and it’s considered the largest difference that is clinically acceptable and so they just need to provide their rationale a lot of times they’ll say it’s should be no more than 50% less than the effect size it is a general rule that is used but just need to make sure that it makes sense that it’s listed by the authors and they provide the rationale as to how they got that margin.
Skip to 6 minutes and 16 seconds So again it should be prestated or used a priori in the methods section and again that rationale needs to be there otherwise that could be a confounder in your study if they don’t discuss why they chose it how they chose it and why it’s appropriate Now again we always know clinical practice guidelines is a place to go back and see if any margins are listed there and also previous studies but you have to make sure that if you’re looking at previous studies this is even more important and not inferiority trials than superiority trials is you need to make sure the trials compare a similar patient population the primary outcome measures for the same the disease state severity was the same and the study treatment with the dose needs to all be the same otherwise again you’re setting up this trial could be setted setting up to find non inferiority when it’s actually not there.
Skip to 7 minutes and 6 seconds All right, just a couple more things that you need to keep in mind as you’re reading this and assessing whether the investigators set their design appropriately. So the NI margin again must be the maximum increase in risk that your patients would on average be willing to accept in exchange for the novel treatments reduction of harm or benefit.
Skip to 7 minutes and 29 seconds Remember I said that there’s really no reason to do a non-inferiority study if a placebo control trial could be done and there’s really no point in looking at a new drug if there isn’t some benefit that a patient would receive from it because the treatment is going to be potentially just as good so why would you want as good if a standard treatment has been out there longer and then you also have to consider where patients willing to accept some loss and effectiveness because remember it’s going to be as good which means that there may be a margin then it’s not as good but it’s acceptable so you want to make sure that good enough is something that your patients are willing to accept if there are fewer risks associated with it All right let me give you a couple examples of that and how NI margins are listed in clinical trial.
Skip to 8 minutes and 20 seconds So look at let’s look at the one on the left hand side it says that in the rationale for where they got their margin from that in FDA lists that for HIV drugs at ten to twenty ten - excuse me, ten to twelve percent it’s considered acceptable so they set their margin at twelve percent. So my question to you is we’re choosing a twelve percent margin rather than the 10 percent be easier? or more difficult to find non-inferiority? so let you think about that for a second and then I’ll answer it for you and the answer would be easier.
Skip to 8 minutes and 57 seconds So again this is something that you’re want to want to keep in mind because it’s going to be easier for the study to not find that the treatment is non-inferior than if it chose the 10% margin that the FDA considers as acceptable Now example number 2 lists the non-inferiority margin but it doesn’t provide a rationale as to how that margin was chosen so that means you’re gonna have to go back and do a little bit more legwork to be able to find out if that was considered appropriate.
How to Assess the NI Margin?
Prof. Mary Ferrill explains how to evaluate P values, power, and mainly the NI margin in this video.
When interpreting P values in NI, we have to keep in mind that if an NI study cannot show non-inferiority, this doesn’t mean the treatment is inferior. NI studies can only find NI or NOT.
When interpreting power in NI, we need to know that NI margins are generally smaller than the minimally important effect used to calculate power in superiority studies. Wider margins can lead to a type I error, while narrower margins can lead to a type II error.
Then we are given an example on NI power calculation. Pay attention to the NI Power Criteria.
As a reader, when we assess the NI margin, we need to make sure if the NI margin is appropriate and where did it come from. Besides, it should be clinically relevant and PRESTATED in methods selection.
Ultimately, we are given two examples. Would choosing a 12% margin rather than 10% be “easier” or more difficult to meet NI? Please share your thoughts and answers below.