More information about text formats
Sir, the fine clinical study of Bradshaw and Sen on the usefulness of
antiemetics when administering morphine in emergency situations provides
interesting data. In my view their data do not support the conclusions of
the authors completely.
They reported a more than twofold increase of side effects
(nausea/vomiting) in patients without antiemetic prophylaxis as non-
The study is too small to rule out that the observed relative risk
increase from 1,6% to 3,7% is due to chance or not. Calculating the 95%
confidence interval from their data, the real rate for the antiemetic-free
group could have been somewhere between 1,2% - 8,3%.
Of course if we accept a cumulative incidence of up to 8% side
effects in acute pain patients, and we do not want to bring it down, then
the study gives a satisfying answer. If we want to prove statistically
that the observed effect of a reduction from around 4% to 2% is really due
to chance, then a study size of around 2500 is necessary (usual
assumptions: 95% confidence level, 80% power, 1:1 balance of unexposed:
exposed). Smaller sample sizes will always produce non-significant results
when rates are so small.
Alas, for clinical practice, even if a larger study proves the reduction
to be robust, the number needed to treat would be quite large as rates and
their differences are small.
For calculating the confidence intervalls I used STATA 9.0, (cii
command, assumption of binomial distribution) and for the sample size
estimate the STATCALC program of EPI Info Version 6, CDC, Atlanta.