Prophylactic antibiotics to prevent pneumonia and other complications after measles: community based randomised double blind placebo controlled trial in Guinea-Bissau

Dear Sir

Measure of effect is not transferable

The study by Garly et al has important methodological flaws. The
measure of effect that was reported was the odds ratio of development of
low respiratory complications. This is an unusual measure for a randomised
controlled trial, where the relative risk and absolute risk reduction are
more appropriate. It might be understandable that odds ratio was chosen
over relative risk or absolute risk reduction because the study only
achieved 38% of the planned sample size [84/218]. The study then went on
to report the ‘Number Needed to Treat’. This requires the absolute risk
reduction. This mix of the weaker odds ratio and NNT is curious. Would the
results be different if the relative risk was reported?

In addition the 95% confidence interval around the odds ratio of the
primary outcome was reported as ‘odds ratio 0.08 (95% confidence interval
0 to 0.56). This result is clearly not possible because a ratio cannot be
zero. When there is no difference between a numerator and denominator, the
ratio is 1.

The authors acknowledge the final sample size as a serious limitation
to their study. Assuming that the null hypothesis (that there was no
difference in the outcome of lower respiratory infection with antibiotics
or placebo) was the basis of this study, the conclusion that prophylactic
antibiotics may have an important role in the management of measles
infection in low income countries is unsupported by the 38% sample size
and the faulty confidence interval. Further studies are required before
this paper is taken seriously.

Competing interests:
None declared