Introduction: Australian Aborginal children experience extremely high rates of bacterial respiratory infections. In contrast to non-Aboriginal children, adverse outcomes are frequent.

Objectives: To describe an explicit approach to the presentation and interpretation of randomised controlled trials (RCTs) within Cochrane reviews. The review of antibiotics for acute otitis media (AOM) is used as an example.

Methods: All RCTs included within reviews are categorised according to the prior probability of the adverse outcome of interest. This can be assumed to be the mean proportion of participants affected from both the intervention and control groups. Trials should then be grouped according to their underlying level of risk. The following cut-points are suggested: low-risk (<10%); moderate risk (10-30%); and high-risk (>30%). Differences in inclusion criteria and the estimated effect sizes (relative risk, risk difference and likelihood ratios) should be systematically described.

Results: For children with AOM in high-risk populations, perforation of the eardrum is the primary outcome of interest. Only two of eligible studies within the meta-analysis reported this outcome. Even in these trials the overall risks were low to moderate. The differences in perforation rate were not statistically significant due to their small number. However, the estimated probability that antibiotics will reduce perforation rates (calculated from the available data) is more than 85%.

Discussion: The failure to present results in a way that facilitates use by high-risk populations limits the applicability of evidence for disadvantaged groups. Greater attention to the prior probability of adverse outcomes and the likelihood of benefit are strongly recommended. This will also help resolve disagreement over the conclusions drawn from Cochrane reviews.