Responses
Other responses
Jump to comment:
- Published on: 29 April 2020
- Published on: 29 April 2020Understanding NNTs
Roe et al have written a useful article on the continuing misuse of relative risk, and the importance of understanding relative risk and absolute risk difference in injury risk outcomes in randomised controlled trials. In describing the Number Needed to Treat (NNT) they miss out an important word- the NNT is the number needed to treat to prevent one _extra_ adverse event, not to prevent a single adverse event. To see thus suppose the NNT was m. In their notation the risk in the intervention group is IG and the Control group is CG. The number of events expected in the intervention group if we treated m of them is mIG. To prevent one event we have mIG=1 and so we have to treat m=1/IG subjects to prevent one event. However we would expect mCG events in the control group. To prevent one _extra_ event in the intervention group we would require mCG-mIG =1 (assuming CG>IG) . Thus m=1/(CG-IG) which is the definition of the NNT. They could also, perhaps, have mentioned the problems in using the NNT, such as differing baselines leading to it being uninterpretable as described, for example by Stang, A., Poole, C., & Bender, R. (2010). Common problems related to the use of number needed to treat. Journal of Clinical Epidemiology, 63(8), 820–825
Conflict of Interest:
None declared.