Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Antibiotics for acute otitis media


The Achilles' heel of pooled analysis is that we concentrate on averages. No individual is average, and we could do with much less concentration on whether interventions work on average, but which patient characteristics determine where the intervention works best. No does it work, but in whom does it work? How can this be done? The answer is individual patient meta-analysis. Theoretically this can provide really useful information. Such an analysis for acute otitis media in children [1] holds out little indication that antibiotics are useful in any children.


Investigators of trials were approached for individual patient data if their trials randomised children aged 0-12 years with acute otitis media, compared antibiotics with no treatment or placebo, and had pain and fever as an outcome. Of 10 such trials, six provided data.

Outcomes calculated were presence of pain (yes/no), fever (greater or less than 38˚C), or both at 3-7 days. A series of pre-defined subgroup analyses were planned, together with logistic analysis to identify important correlates of treatment efficacy.


The six trials essentially tested amoxicillin versus delayed treatment or placebo. These six trials randomised 1,633 children. Overall, antibiotics reduced the incidence of an extended course of acute otitis media at 3-7 days by 13%, with an NNT of 8 (95% confidence interval 6-11).

Table 1 shows the overall result in more detail, together with those subgroups where there was a lower (better) NNT. The analyses indicated that the effect of antibiotics was modified by age, bilateral disease, and otorrhoea.

Table 1: Results of sub group analyses for antibiotics vs placebo in AOM

Extended course of acute otitis media (%)
Number in analysis
Relative risk
(95% CI)
(95% CI)
Overall result
0.83 (0.78 to 0.89)
8 (6 to 11)
<2 years with bilateral AOM
0.64 (0.62 to 0.80)
4 (3 to 7)
<2 years with unilateral AOM
0.92 (0.76 to 1.1)
not calculated
Otorrhoea present (any age)
0.52 (0.37 to 0.73)
3 (2 to 5)
Otorrhoea absent (any age)
0.80 (0.70 to 0.92)
8 (4 to 20)
Outcome was pain, fever, or both at 3-7 days


The authors of this analysis go to great pains to describe possible limitations, despite their individual patient analysis, and the great care they have taken in a detailed and sophisticated analysis. The take-home message, though, is that antibiotics seem to be most beneficial in younger children with bilateral acute otitis media, and where there is otorrhoea.

How much weight should we place on this? Not much, because differences between antiobiotics and placebo disappeared by five or six days, numbers were small, and what differences there were came from differences with placebo (look at Table 1 carefully). This analysis nails down that there is no subgroup of children for whom antibiotics can be really useful in acute otitis media, unless there are complications or other consideration. It proves the utility of individual patient analysis.


  1. MM Rovers et al. Antibiotics for acute otitis media: a meta-analysis with individual patient data. Lancet 2006 368:1429-1435.

previous story