In this RCT, three of six schools were randomly allocated to an asthma education program, the other three to no education program. Hence this is a cluster-randomised trial. There may also have been some stratification, because each arm contained 2 government schools and 1 independent school. The primary outcome is change in QOL.

The results have not been directly adjusted for clustering. The authors have calculated intra-class correlation coefficients (ICC) for the QOL domains and simply reported them. The ICC appear to have been small (<0.002), but what are the cluster inflation factors? To be consistent with the design it would be more appropriate to incorporate the clustering adjustment in the statistical analysis.

I have just a few comments re the rest of the paper:

How was the number of schools (and required estimated number of wheeze identified students) decided?

Table 2 should be more clearly labelled – it is not immediately clear that %’s (relating to improvement) are presented here.

Figures 2-4. Are all these histograms necessary?

There are several numerical discrepancies that need to be addressed – most have already been spotted by the referee.
 
 
 

Julie Morris