Objectives: trials of antibiotics for symptoms of ARIs have developed in an environment in which antibiotic use was already established. Thus trials have been of "no-antibiotic" vs usual care (antibiotics). This might influence recruitment (less severely affected vs more). If severity influences response to treatment, then excluding more severely affected patients would yield a smaller effect size.

Method: we reanalysed the trials selected for the Cochrane meta-analysis of "antibiotics for acute otitis media in children", looking for evidence of patients excluded on the basis of being too ill. We also looked for evidence of severity influencing outcome.

Results: We could extract 'non-recruitment because the child was too ill' out of the total recruited from only two of the seven trials. They were 52/232 and 27/240 respectively, or 17%.

Conclusions: This figure has the potential to result in under-estimates of the effects of antibiotics for acute otitis media.