Skip to 0 minutes and 14 seconds So how do we go about specifically doing these in a trial and you know I love steps and you know I love boxes already. So like to have things that are lockstep that are easy for you to remember and quick to be able to do it. So you want to make sure you’re calculating these NNT’s with a primary outcome or any significant ADRs because we know these are the ones that are the focus of the trial. So we want to make sure we look at these. And really calculating NNT and NNH when there isn’t a significant difference between the groups will give you information that probably isn’t applicable because the numbers will won’t mean anything at that point.
Skip to 0 minutes and 52 seconds Remember four measures you have to have nominal data and again, because you’re looking for the difference between the group you’re going to do them with superiority studies not with non-inferiority and we’ll go over that more in the next session on non-inferiority. And then you need to make sure is the primary outcome you’re looking at is it bad? Or is it good will depend on how you list the statement And so then you will list it is in favor of the study drug or not in favor of the study drug.
Skip to 1 minute and 18 seconds So let’s just briefly before we go on for further let’s just talk about again the confidence intervals and the touching or crossing aspects So you can see the examples on the left, You can let tell from looking at it that one’s A1c and ones weight-loss that these are continuous variables. There’s no RR there’s no HR listed so those are just general CIs. So we’re looking to see if it touches or crosses zero.
Skip to 1 minute and 43 seconds So first example with hemoglobin A1c that the sample had a 0.5 it was negative 0.62 to 1.2 in the population is this considered a good or bad confidence interval and you would say it’s not good because it does cross 0 the weight loss one that had -3.2 kilograms with the range of -9.7 to -2.5 is good because it doesn’t touch across 0. Looking on the right hand side with those two examples you can see they both have RR listed by their confidence intervals so those are going to be measured. So the number we’re looking to not touch your cross is one. So you can see with the first one 0.8 with 0.6 to 1.0.
Skip to 2 minutes and 27 seconds That’s considered not good because you can see it touches one. And then you can see the GI bleed below 0.3 to 0.2 to 0.89 is that good or not and that one is good because it doesn’t touch across one. Now I know that from teaching this from years that students often have problems visualizing the touching or crossing so we’ll talk more about this non-inferiority but it’s often good if it’s not graphed in the trial that you graph that just so you can see for yourself that the line of the confidence interval does not touch across either zero or in depending on what you’re looking at And then as it says at the bottom of the slide.
Skip to 3 minutes and 4 seconds Remember to always check the width of the confidence interval no matter what you’re looking at. Because if the sample is not reliable then it’s going to be even worse in the population because of the fact that you’re estimating something from a value that’s not very reliable. So I always get asked what is considered a good NNT and NNH Well I already told you that a good NNT is obviously small and a good NNH is large. Because you want to have fewer patients that you need to treat to get an additional benefit. And you want fewer th or more patients that you treat before you see some harmful effect.
Skip to 3 minutes and 37 seconds There’s also as I mentioned other things that you need to take into consideration depends on the primary outcome and what’s the treat definition of treatment of success for that outcome. I already discussed the duration of observation it makes a huge difference if it’s six months versus six years. Comparison either active treatment versus placebo. Obviously you’re going to want to have a large or smaller NNT with placebo than you do with an active comparison so it would make a difference . And then we already talked about severity of illness and that’s why it’s good to look at NNT versus NNH because that’ll take that into account.
Skip to 4 minutes and 13 seconds So if you don’t have any idea in the studies that say you’re evaluating a study you have no idea what’s considered a good or about NNT for that disease state you’re looking at you need to pull similar trials and see if NNT were listed there to see how yours in this study compares to those. There’s also a few websites available there’s really nothing that’s all that complete comprehensive but I did you’ll see when you get to the end of this presentation. There’s slide that lists several places that you can go and try to find NNT for different disease states to give you some idea of what your NNT should a good NNT would be in that situation.
Skip to 4 minutes and 52 seconds But again that it all depends on all the other pieces of NNT the duration and all those types of things and so just the number doesn’t always tell you very much. Now NNT for treatment is going to be a much smaller number with for prevention and this makes sense you have to treat a lot more patients to prevent something than you do to treat it. So just to give you an example in NNT if somewhere less than 300 is really good for prostate cancer prevention whereas NNT of less than or equal to five is good for the treatment of migraines.
Skip to 5 minutes and 26 seconds Now in addition to looking at the NNT in a trial you also have to compare benefits versus risks So that’s why you need to compare the NNT in light of the NNH anytime you’re looking at these in a clinical trial.
What is a Good NNT/NNH number?
Prof. Mary Ferrill clarifies when to calculate a NNT/NNH number and what is a good NNT/NNH number in this video.
First, there are two major steps to determine when a NNT/NNH number should be calculated. Make sure you understand them and move on.
Besides, she also demonstrates the crossing and touching of confidence intervals (CIs) by four examples. The most important thing is that, remember to check the width of the CI no matter what you are looking at to ensure the reliability of the data.
What is a good NNT/NNH number? A good NNT is obviously small and a good NNH is large.If you don’t have any idea in the studies, you need to pull similar trials and see if NNT were listed there.
Finally, NNT for treatment is a much smaller number for prevention because you have to treat lots of patients to prevent something. If NNT is less than 300, it is good for prostate cancer prevention; if NNT is less than or equal to five, it is good for the treatment of migraines.