Intended for healthcare professionals

Rapid response to:


Randomised controlled trial of effects of early discharge after surgery for breast cancer

BMJ 1998; 317 doi: (Published 07 November 1998) Cite this as: BMJ 1998;317:1275

Rapid Response:

Not no null hypothesis

I am somewhat confused by some of the stastical measures used in this
paper. It seems that as there was not "significant" (ie p<0.05)
difference in the two groups (early and late discharge) that the authors
conclude that there was no real difference. But this is not how the null
hypothesis works.

To use the null hypothesis one says that there is no difference in
the two groups and then calculates a probability value for the samples
being drawn from the same population. A low probability (p<0.05
traditionally) suggests that the chances of the groups being the same is
small and that there is a difference.

However this cannot be used in reverse; one cannot say that as there
chance of them being from the same population is reasonable (p>0.05)
that they therefore did come from the same population (ie no difference
made by discharge date) as this negates the power calculations inherent in
the p-value. It may well have been that the study was too small to detect
any difference.

Both an insufficient sample and no difference in measured outcomes
would give p value>0.05 without any way of distinguishing the two with
these calculations.


Competing interests: No competing interests

12 November 1998
C O'Loughlin
Medical SHO
Northern General Hospital, Sheffield