When someone asks me about something statistical I always panic, if only internally. It's not my strong point and that's why I'm a dentist and not a statistician.
Somehow on clinic on Friday, a student and I got to looking at the 5 year results on the Hall technique trial taking place in Scotland (link here). I was waxing on, as I do, about how great the results looked - major failures (i.e. irreversible pulpitis, loss of vitality, abscess, or unrestorable tooth) were 3% in teeth treated with the Hall technique and 16.5% for those with conventional treatment. The p value was 0.000488 - highly significant. And then it said "NNT 8". That's when the question came - what is NNT?
Let's say there's a drug that prevents heart attacks and in a trial comparing it to placebo, the results report an NNT of 8 over, say, 5 years. What this means is that 8 patients would have to take the drug for 5 years to prevent one heart attack.
Now, this is a useful indicator of the effectiveness of the drug. Say the NNT was 100. That would mean 100 patients would need to take the drug for 5 years to prevent a single heart attack. Now you might think that it's still worth taking the drug but if the cost were, say, £1000 a year for the drug, it starts to look like a pretty expensive game. And what if the side effects are significant. For example, the drug causes severe and debilitating muscle cramp in 20% of people. So one person in 100 doesn't get a heart attack but 20 people get laid low by leg cramp. Maybe it's still worth it, but the point is that you begin to be able to get a sense of the benefit of an intervention by using NNT.
What the Hall trial here was saying is that since there was an NNT of 8, this means that - compared to conventional treatment - 8 teeth would need to be treated with the Hall technique instead of that conventional treatment to prevent one major failure. If that NNT had been 100, the technique wouldn't have looked that much better than conventional. If the Hall technique is, say double the cost of conventional treatment, treating 100 teeth with it instead of the conventional treatment could seem extravagant.
If the NNT had been 2 then that would mean for every 2 teeth treated with the Hall technique compared to the conventional treatment there'd be one tooth saved. That would be very impressive.
But cost is only one issue. What are the consequences of a major failure? This is an important issue when deciding what the NNT needs to be to have significance clinically. Say you had a drug that stopped a lethal cancer but the NNT was 100. You might feel that this was good as you save a life for every 100 people given the drug. The consequence of the intervention is massive. But what about saving a tooth? Is an NNT of 100 still significant? An NNT of 8 probably is from a humble dentist's (rather than oncologist's) point of view given that this means fewer children needing extractions and the loss of space maintenance. I'd like to know, though, what the cost of doing this is to every deciduous tooth with caries as this would allow an economic evaluation of the technique too.
There are no right and wrong answers to what the NNT should be, but perhaps you can see how the NNT can help inform us as clinicians, our patients and policy people who have to decide how to allocate scarce NHS resources.
If you'd like to read some more about this there's a very accessible explanation here.