[music]
Speaker 1: Good afternoon. Thank you for coming to this afternoon's masterclass, titled, How Can the Number Needed to Treat and Number Needed to Harm Inform My Clinical Decision When Treating Patients with Bipolar Depression. I'm Parsh Sachdeva a medical science liaison with Sunovion Pharmaceuticals. Here today, I have with me Dr. Leslie Citrome, Clinical Professor of Psychiatry and Behavioral Science at New York Medical College in Valhalla, New York.
Dr. Citrome was the consultant of Sunovion Pharmaceuticals. In today's program, we'll be evaluating the story of how the analysis of number needed to treat and number needed to harm tell us a story in terms of evaluating therapies available for treatment of bipolar depression. We'll get started. Dr. Citrome, can you start and can you tell us how do we calculate what is NNT and what is NNH?
Dr. Citrome: I'll be happy to. Thank you for having me here today. This is one of the things I'm really excited about because it's a way of determining how important a clinical trial result may be. We're given graphs that something is better than something else that, let's say drug better than placebo, and we're given a p-value less than 0.05, and we're expected to say, "Okay, this is important." What you really need to determine is the number needed to treat and this is effect size, it determines how important that result can be.
Number needed to treat answers the question, how many patients needed to be treated with one intervention versus the other before expecting to encounter one additional responder, for example, or one additional remitter. The lower the number needed to treat value, the better drug A looks versus the alternative, and the number needed to harm is looking at outcomes we would rather avoid. Let's say how many patients needed to be randomized to drug versus placebo, for example, before expecting to encounter one additional patient who has excessive weight gain or an additional adverse event of somnolence, for example. Number needed to treat number needed to harm are effect size measures and helps us figure out what these percentages mean, what this difference between drug and placebo means in human terms and patient units.
Speaker 1: Thanks for that overview, Dr. Citrome. Can you tell us how we calculate NNT.
Dr. Citrome: It actually sounds pretty mysterious, but it's very straightforward. All you need to know to calculate the number needed to treat for, let's say, A versus B on an outcome that you're interested in like remission, you need to know the percentage of patients on drug A versus drug B who were remitted. Two percentages. Drug B can be, by the way, placebo, it doesn't have to be another drug.
What you want to do is subtract these two percentage points, and then see how many times it goes into the number 100. You take one divided by the difference in percentages. You'll also round up, because the lower the number needed to treat, the more powerful it is, and you don't want to exaggerate a difference. A number needed to treat a 3.1, if we round it down to three, that sounds better, but it may be inaccurate.
We'll round it up to 4, err on the side of caution. Let me give you an example. Drug A results in remission 50% of the time, drug B, let's say, the placebo control results in remission 20% of the time. Relatively speaking, it's two and a half times better, but that doesn't answer the question, how often will a clinician encounter a difference in day-to-day care of people with that disorder? Well, very simple calculation, you have two percentages, 50 and 20, subtract them, you get 30.
That goes into number 100, 3.3 times rounded up to 4. For every four patients who are given A versus B, you'll encounter one additional remitted patient, and that will probably lead you to select A over B under most situations, unless, of course, A is not tolerated well or not accessible, not on the formulary or whatever, but otherwise, A is better than B. This is going to be clinically relevant because you're going to encounter this difference in day-to-day care of people.
Speaker 1: I'll pick it up, Citrome. You did mention about the power of that NNT. Can you tell us what's the magnitude of NNT and what's a good NNT number to look for?
Dr. Citrome: Any number needed to treat value less than 10 represents a single digit and represents an effect size that is substantial enough to be able to be observed in day-to-day practice. If it takes five patients, six patients, seven patients eight patients on one drug versus the alternative to encounter a difference, you will encounter that difference in day-to-day practice.
Now, the most powerful NNT that we can see in medicine is two you can't have an NNT of one because nothing is perfect. Two is the best NNT you can see and it's very unusual. NNT of three represent a strong effect size. NNT that is, let's say, eight or nine is a weaker effect size. Any number needed to treat value that's a double-digit or triple-digit, generally, means you will see that outcome very often in day-to-day clinical practice and is less important.
[00:05:28] [END OF AUDIO]
