Thinking about the Benefit-Risk Ratio Posted on February 20, 2006 David J. Kupfer (bio) discusses Number Needed to Treat vs. Number Needed to Harm.

So, when I think about clinical trials now, I think about, regardless of what the treatment is or the modality that we're using for intervention, I'm thinking much more about the benefit-risk ratio of that treatment. So I'm thinking about, "Does this treatment really make a difference, and if so, how much of a difference?" I'm less concerned about whether there's a statistic of .05 or .01, than is it clinically meaningful. In other words, am I going to change the way I would have a treatment strategy, say, for treatment resistant depression or severe anxiety or phobias, phobic disorders which are not responsive to any kinds of behavioral interventions? What would I do? If I can be convinced that this is a new, good strategy, then what that would help me decide is that there's a big benefit there. That benefit can be measured clinically if you think about NNT coming from evidence based medicine or Number Needed to Treat. The number needed to treat very simply is, "How many people do I have to treat to really make a difference with the next person?"

So, for example, if you think about an NNT of 4, it means that I have to treat 4 people and that fourth person will do better than the other treatment or better than placebo. It's like 1 out of 4 that is really improving. That's pretty good actually. We don't see that in most of medicine and you often see 1 out of 10 or NNT of 10 or an NNT of even 15. Yet depending on whether you're talking about a life-saving procedure, or really making a huge difference say in the treatment of heart disease or a tumor, that is certainly acceptable.

On one hand the equation that talks about number needed to treat is a good way of thinking about benefit. But one of the things that strikes me as well as some others in this area now thinking about clinical trials is that doesn't tell me how risky the procedure is. It may have a very good benefit ratio but what if it's causing all kinds of problems, and how can I measure that and how can I measure that in such a way that I can compare the two? That's where the issue comes up that people have begun to use more frequently Number Needed to Harm or NNH, which is really the same way of thinking about it. How many people get treated with this new procedure compared to say the old procedure who have a risk? And we decide what the adverse risk is.

So, for example, the recent set of ambiguities about well, how do I best treat a depressed kid who has suicidal feelings. Am I going to increase those suicidal feelings and possibly lead to a suicide attempt if I use the anti-depressants that presumably now says wait a minute, there's a risk. I need to be able to balance that out. That's why it's important when you look at clinical trials to see both of those things that is, what's the benefit in terms of clinical efficacy and probably even reducing the likelihood that I'm going to think about suicide versus, what's the activation possibilities of that particular drug in terms of so called adverse effect.

« Back to Article