Burke 1991.

Methods	Randomised ‐ yes, computer‐generated random numbers Concealment of allocation ‐ adequate Double‐blind ‐ yes Intention‐to‐treat (ITT) ‐ yes Loss to follow‐up ‐ not described Design ‐ parallel
Participants	N ‐ 232 children Age ‐ between 3 and 10 years Setting ‐ general practice; 48 general practitioners in 17 general practices in Southampton, Bristol and Portsmouth (UK) Inclusion criteria ‐ acute earache and at least 1 abnormal eardrum Exclusion criteria ‐ antibiotic treatment or acute otitis media (AOM) < 2 weeks prior to randomisation, strong indication for antibiotic treatment according to general practitioner, contraindication for amoxicillin, serious chronic conditions Baseline characteristics ‐ slight imbalance in gender (boys treated with antibiotics versus boys treated with placebo = 52% versus 42%) and figure 1 appears to demonstrate that fewer children were crying at baseline (0 hours) in the amoxicillin arm compared with the placebo arm, suggesting a failure of randomisation
Interventions	Tx ‐ amoxicillin 250 mg 3 times daily for 7 days; N = 114 (N = 114 included in analysis for short‐term outcome) C ‐ matching placebo 3 times daily for 7 days; N = 118 (N = 118 included in analysis for short‐term outcome) Use of additional medication ‐ analgesics (paracetamol 120 mg/5 mL) for pain as needed
Outcomes	Main outcomes were divided into short‐term, middle‐term and long‐term: Short‐term ‐ (a) duration of symptoms; (b) use of analgesics (assessed by weighing bottles); (c) clinical signs at 1 week; (d) incidence of complications; (e) treatment failure (i.e. second‐line antibiotics were required) Middle‐term ‐ (a) tympanometry findings at 1 and 3 months Long‐term ‐ (b) number of AOM episodes in 12 months; (b) number of specialist referrals Home visits by researcher at day 1, days 4 to 6 and general practitioner visit at day 7 Symptom diary kept by parents for 21 days
Notes	It is not clear whether the "discharging ears" in Table 1 should be included as perforations, we now included the number of perforations as summarised in Table 2 in our analysis
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers
Allocation concealment (selection bias)	Low risk	Randomisation was carried out independently of the investigators; randomisation code was kept sealed and was unknown to any of the participants in the study
Other bias	Unclear risk	ITT analysis ‐ yes; baseline characteristics ‐ imbalance for gender and crying
Blinding of participants and personnel (performance bias)	Low risk	Each bottle was identified only by number
Incomplete outcome data (attrition bias)	Unclear risk	Loss to follow‐up ‐ not described; all randomised patients included in short‐outcome analysis