Antibiotics for the prevention of acute and chronic suppurative otitis media in children

Abstract

Background

Acute otitis media (AOM) is a common childhood illness which may be frequent and painful. AOM may be associated with tympanic membrane perforation and can progress to chronic suppurative otitis media (CSOM).

Objectives

To determine the effectiveness of long‐term antibiotics (six weeks or longer) in preventing any AOM, AOM with perforation and CSOM.

Search methods

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2010, Issue 3) which includes the Acute Respiratory Infections Group's Specialised Register, MEDLINE (January 1966 to July Week 4, 2010), OLD MEDLINE (1950 to 1965) and EMBASE (1990 to August 2010).

Selection criteria

Randomised controlled trials of long‐term antibiotics versus placebo or no treatment for preventing AOM, AOM with perforation, or CSOM.

Data collection and analysis

Two authors independently extracted the data for: AOM; episodes of AOM; recurrent AOM; episodes of illness; side effects; antibiotic resistance; and outcomes at end of intervention (any AOM) and following cessation of intervention (any AOM). For dichotomous outcomes, we calculated the summary risk ratio (fixed and random‐effects models). For rate outcomes, we calculated the summary incidence rate ratio.

Main results

Seventeen studies (1586 children) were included. All studies enrolled children at increased risk of AOM. In seven studies the children were prone to otitis media. The majority were high‐quality studies and most (16 studies) reported data for our primary outcomes. One reported AOM with perforation or CSOM. Long‐term antibiotics reduced any episode of AOM (14 studies, 1461 children, risk ratio (RR) 0.65, 95% CI 0.53 to 0.79; random‐effects model) and number of episodes of AOM (13 studies, 1327 children, incidence rate ratio (IRR) 0.51, 95% CI 0.39 to 0.66; random‐effects model). Approximately five children would need to be treated long‐term to prevent one child experiencing AOM whilst on treatment. Antibiotics prevented 1.5 episodes of AOM for every 12 months of treatment per child. We explored statistical heterogeneity. Long‐term antibiotics were not associated with a significant increase in adverse events (12 studies, 817 children, RR 1.99, 95% CI 0.25 to 15.89; random‐effects model).

Authors' conclusions

For children at risk, antibiotics given once or twice daily will reduce the probability of AOM while the child is on treatment. In similar populations, antibiotics will reduce the number of episodes of AOM per year from around three to around 1.5. We believe that larger absolute benefits are likely in high‐risk children. These conclusions were not affected by sensitivity analyses.

Plain language summary

Antibiotics to prevent acute ear infections in children

Acute otitis media (AOM, infection in the middle ear space) is common in children, causing pain and deafness. Most children experience at least one episode and some children suffer recurrent AOM (more than three episodes in six months or four episodes in 12 months), and some progress to eardrum perforation. Interventions (antibiotics, vaccines or reduced exposure to bacterial and viral pathogens) that decrease the frequency and/or the severity of infection are needed. Antibiotics given once or twice daily may reduce episodes of bacterial AOM and their complications. This review included 17 studies (1586 children). Long‐term antibiotics (equal to or more than six weeks) almost halved the risk of further infections. There was not enough information to know if antibiotics reduced acute otitis media with perforation or chronic suppurative otitis media (chronic perforation), or improved long‐term outcomes. Antibiotics did not appear to be a frequent cause of significant side effects (for example, allergic reactions or diarrhea). Parents must balance these potential side effects plus the cost and inconvenience associated with antibiotics against the likely benefits of treatment. Antibiotic resistance from the long‐term use of these drugs is also an issue which should be considered, particularly for children with recurring infections.

Summary of findings

Summary of findings for the main comparison. Antibiotic versus control ‐ primary outcomes for the prevention of acute and chronic suppurative otitis media in children.

Antibiotic versus control ‐ primary outcomes for the prevention of acute and chronic suppurative otitis media in children
Patient or population: patients with the prevention of acute and chronic suppurative otitis media in children Settings: Intervention: Antibiotic versus control ‐ primary outcomes
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Antibiotic versus control ‐ primary outcomes
Prevention ‐ any AOM or CSOM during intervention	Low risk population		RR 0.65 (0.53 to 0.79)	1461 (14 studies)	⊕⊕⊕⊝ moderate1
	300 per 1000	195 per 1000 (159 to 237)
	Medium risk population
	550 per 1000	358 per 1000 (291 to 435)
	High risk population
	800 per 1000	520 per 1000 (424 to 632)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ Larger effect noted in small studies (see funnel plot). This may be due to publication bias.

Summary of findings 2. Antibiotic versus control ‐ secondary outcomes for the prevention of acute and chronic suppurative otitis media in children.

Antibiotic versus control ‐ secondary outcomes for the prevention of acute and chronic suppurative otitis media in children
Patient or population: patients with the prevention of acute and chronic suppurative otitis media in children Settings: Intervention: Antibiotic versus control ‐ secondary outcomes
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Antibiotic versus control ‐ secondary outcomes
Side effects ‐ any clinical side effects during intervention	8 per 1000	16 per 1000 (2 to 127)	RR 1.99 (0.25 to 15.89)	817 (12 studies)	⊕⊕⊕⊝ moderate1
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹ Small number of adverse events noted.

Background

Description of the condition

Otitis media is a major health problem in young children. Otitis media (OM) is a broad term for any infection or inflammation occurring in the middle ear space (behind the tympanic membrane). It represents a spectrum of disease: i) otitis media with effusion (OME), fluid behind an intact tympanic membrane without the symptoms or signs of acute otitis media; ii) acute otitis media (AOM), fluid behind the tympanic membrane with the symptoms or signs of an acute infection; and iii) chronic suppurative otitis media (CSOM), persistent discharge of pus through a perforated tympanic membrane for more than six weeks. The criteria used in definitions of AOM are highly variable. Ear pain (Rosenfeld 2003), bulging of the tympanic membrane (Karma 1989) and recent discharge of pus are regarded as the most diagnostically reliable. In populations with high rates of CSOM, it is helpful to distinguish between those children with AOM without perforation and those children with perforation of the tympanic membrane, since AOM with perforation always precedes CSOM. Children who have had three episodes of AOM within the last six months or four episodes within the last 12 months are classified as having recurrent acute otitis media (rAOM) (Rosenfeld 2003).

A comprehensive search of The Cochrane Library and MEDLINE identified the following systematic reviews of interventions for AOM: antibiotics for treatment (Del Mar 1997; Kozyrskyj 2000; Sanders 2009); decongestants or antihistamines, or both for treatment (Coleman 2008); antibiotics for prevention (Bonati 1992; Williams 1993); and pneumococcal vaccines for prevention (Jansen 2009). The systematic review of antibiotics to prevent AOM by Williams et al (Williams 1993) found that antibiotics reduced the rate of AOM by 50%. However, AOM was infrequent (the AOM recurrence rate in the control groups was 0.062 to 0.488 per patient‐month) and each child required around 10 months of treatment to prevent one episode of infection. Similar overall results were reported in a more recent evidence summary (Rosenfeld 2003). In this meta‐analysis, studies with placebo groups having higher AOM recurrence rates (more than 0.2 episodes per month) showed a rate difference of 0.18 episodes per month (95% CI 0.08 to 0.28) compared to a rate difference of 0.06 episodes per month (95% CI 0.04 to 0.09) for studies where placebo groups that had lower recurrence rates. The studies included in these reviews were largely conducted in the USA or other industrialised countries. In these settings the pain or fever associated with an episode of AOM is the principle concern of the care‐giver. However, in disadvantaged populations the potential progression to CSOM is more important. In addition to constant offensive discharge, the hearing loss associated with this severe disease is likely to have a great impact on the language development and behaviour of young children. Not surprisingly, the World Health Organization's hearing preservation group has stated that populations with children having a 4% rate of CSOM have a massive public health problem.

Indigenous children living in Australia, North America, Northern Europe and New Zealand appear to be at particularly high risk of developing CSOM (WHO/CIBA 1996). In a recent cross‐sectional survey of children aged six months to 2.5 years and living in 29 remote communities in the Northern Territory, rates of perforation varied from less than 10% to 60% (Morris 2005). In longitudinal birth cohort studies in this population the onset of disease has been shown to commence within weeks of life, following nasopharyngeal colonisation with Streptococcus pneumoniae (S. pneumoniae), non‐capsular Haemophilus influenzae (H. influenzae) (NCHi) and Moraxella catarrhalis (M. catarrhalis) (Leach 1994). This association between bacterial colonisation and early‐onset OM has led to speculation about the role of long‐term antibiotic therapy.

Description of the intervention

Prophylactic antibiotics have been recommended to reduce the risk of bacterial infection in selected children. For children at high risk of bacterial OM, oral antibiotics active against S. pneumoniae and H. influenzae can be given once or twice daily. The duration of prophylaxis may vary from weeks to months.

How the intervention might work

Prophylactic antibiotics might either reduce nasopharyngeal colonisation with pathogens or reduce the risk of colonised children developing clinically apparent bacterial infections. Antibiotics should not affect the incidence of viral upper respiratory tract infections.

Why it is important to do this review

While antibiotics appear to reduce rates of subsequent AOM in high‐income countries, important questions remain concerning the relative benefits and costs in populations at high risk of AOM or CSOM. The original systematic reviews (Bonati 1992; Williams 1993) have not been updated with the publication of new studies. In addition, the impact of this intervention in high‐risk populations is unclear.

Objectives

To determine the effectiveness of long‐term antibiotics (equal to or more than six weeks) in the prevention of acute otitis media (AOM) and the complications of AOM (this includes AOM without perforation and AOM with perforation and chronic suppurative otitis media (CSOM)) in children at increased risk of future AOM episodes.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) of long‐term antibiotics (equal to or more than six weeks) versus control (placebo or no treatment) for the prevention of AOM without perforation, AOM with perforation and CSOM. Ideally episodes of AOM should be least two weeks apart. While definitions vary substantially between studies, AOM should be confirmed by otoscopy. All episodes of AOM that were associated with perforation were categorised as AOM with perforation even if the tympanic membrane was intact at the time of examination (that is to say, perforations occurred within two weeks of the clinical assessment and not secondary to trauma). Similarly we attempted to limit the diagnosis of CSOM to episodes of discharge from the middle ear that persisted for longer than six weeks. Ideally, resolution of discharge associated with CSOM was confirmed by a clinician. Cross‐over studies were eligible for inclusion but we only used data from the initial treatment period. There were no language or publication restrictions.

Types of participants

Children aged 0 to 18 years at increased risk of future episodes of AOM (as determined by the study investigators). We excluded studies of children with a diagnosis of AOM, AOM with perforation or CSOM at the time of randomisation. We also excluded studies of children with diseases associated with immunodeficiency, craniofacial abnormalities, undergoing tympanostomy tube insertion or other ear, nose and throat (ENT) surgery at the time of randomisation.

Types of interventions

All randomized comparisons of continuous long‐term systemic antibiotics that are effective against S. pneumoniae or non‐capsular H. influenzae (NCHi) (intervention group) versus placebo or no treatment, or treatment considered to be ineffective (control group). We defined long‐term as six weeks or longer. All doses of antibiotics and all dosing schedules were eligible for inclusion. We included antibiotics given continuously during a period of increased risk (for example, the winter months) and independent of symptoms of upper respiratory tract infections (URTI) if the period of treatment was six weeks or longer. We excluded the use of intermittent antibiotics (for example, at the appearance of signs of URTI). We included studies with additional interventions (but excluding surgery) if the intervention was equally accessible to both intervention and control groups.

Types of outcome measures

Primary outcomes

1. Prevention, any AOM/CSOM during intervention: the proportion of participants who developed any AOM, AOM with perforation, CSOM, complications or withdrawals due to side effects during the intervention period.

2. Prevention, episodes of AOM/CSOM during intervention: the number of episodes per child‐year of any AOM, AOM with perforation or CSOM during therapy.

Secondary outcomes

1. Prevention, recurring AOM/CSOM during intervention: the proportion of participants experiencing recurrent AOM/CSOM (rAOM/CSOM) defined as three or more episodes of AOM within a six‐month period during therapy.

2. Prevention, any AOM/CSOM at the end of intervention: the proportion of participants with any AOM, AOM with perforation, CSOM, complications or withdrawals due to side effects at cessation of therapy.

3. Prevention, any AOM/CSOM following cessation of intervention: the proportion of participants with any AOM, AOM with perforation, CSOM, complications or withdrawals due to side effects following cessation of intervention.

4. Prevention, episodes of illness during intervention: the number of episodes per child‐year of any illness.

5. Side effects, any clinical side effects during intervention: the proportion of participants who experienced significant adverse events such as diarrhea and vomiting or allergic reactions that are sufficient to recommend cessation of intervention.

6. Side effects, any antibiotic resistance during intervention: the proportion of participants colonised (or clinically infected) by S. pneumoniae or H. influenzae with resistance or intermediate resistance to the antibiotic at the first follow‐up visit at least six weeks after starting treatment.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2010, Issue 3) which includes the Acute Respiratory Infections Group's Specialised Register, MEDLINE (January 1966 to July Week 4, 2010), OLD MEDLINE (1950 to 1965), EMBASE (1990 to August 2010) and the references of relevant studies.

We used the following search strategy to search MEDLINE and CENTRAL. The MEDLINE search terms were combined with the Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity‐ and precision‐maximising version (2008 revision); Ovid format (Lefebvre 2009). The search terms were adapted to search EMBASE (see Appendix 1). OLD MEDLINE was accessed via the National Library of Medicine Gateway using the terms (otitis AND antibiotic).

1 exp Otitis Media/
 2 otitis media.tw.
 3 (aom or ome or csom).tw.
 4 or/1‐3
 5 exp Anti‐Bacterial Agents/
 6 antibiotic*.tw.
 7 (amoxicillin or amoxycillin or ampicillin or azithromycin or azithromicin or cefaclor or penicillin or sulphamethoxazole or sulphisoxazole or cotrimoxazole).tw,nm.
 8 or/5‐7
 9 Chemoprevention/
 10 (chemoprevent* or chemoprophyla* or prophyla* or prevent*).tw,nm.
 11 or/9‐10
 12 8 and 11
 13 Antibiotic Prophylaxis/
 14 12 or 13
 15 4 and 14

Searching other resources

We searched reference lists in relevant publications, including RCTs meeting the inclusion criteria, published systematic reviews, Clinical Evidence (Jones 2004), Evidence‐Based Otitis Media (Rosenfeld 2003) and Evidence‐Based Child Health (Moyer 2000). We undertook written communication with the authors of trials included in the review and with major pharmaceutical companies (with offices in Australia) that manufacture antibiotics.

Data collection and analysis

Selection of studies

One review author (AJL) conducted the electronic search to identify potentially relevant articles. Both review authors agreed on the final selection of articles that met the inclusion criteria. Both review authors independently reviewed the published systematic reviews, evidence‐based guidelines, Clinical Evidence and text books.

Data extraction and management

We did not exclude any trials that included a treatment and a control group and that met the inclusion criteria. Both review authors independently reviewed trials that satisfied the inclusion criteria. We recorded the following information on data collection forms: date of study; study setting; source of funding; participant recruitment details (including number of eligible children); inclusion and exclusion criteria; study definitions used; randomisation and allocation concealment methods; numbers of participants randomized; baseline characteristics; blinding (masking) of participants, care providers and outcome assessors; dose and type of antibiotic therapy; compliance; frequency of assessments; other outcome measures used in the study; duration of therapy; duration of follow up post‐therapy; co‐interventions; numbers of participants not followed up; reasons for withdrawals from study protocol (clinical, complications, side effects, refusal and other); details on side effects of therapy; and whether intention‐to‐treat (ITT) analyses were possible. We extracted data on primary and secondary outcomes as described previously. We requested further information from the trial authors where required. We resolved any disagreements by consensus.

Assessment of risk of bias in included studies

Both review authors independently assessed the quality of the studies included in the review. We assessed six components of quality.

1. Method of treatment assignment. Trials were scored as Grade A: randomisation and concealment method correct, or randomisation and concealment stated and group similarity documented; Grade B: randomisation and concealment stated but method not described, or suspect method; Grade C: randomisation claimed but not described, and outcome assessor not blinded; Grade D: randomisation not mentioned (Grade A = high quality).

2. Allocation concealment. Trials were scored as Grade A: adequate concealment, Grade B: unclear, Grade C: clearly inadequate concealment, Grade D: not used (Grade A = high quality).

3. Blinding. Trials were scored as Grade A: participant, care provider and outcome assessor blinded; Grade B: outcome assessor blinded; Grade C: unclear; Grade D: no blinding of outcome assessor (Grade A, B = high quality).

4. Reporting on participants by allocated group. Trials were scored as Grade A: the progress of all randomized children in each group described, Grade B: unclear or no mention of withdrawals or drop‐outs, Grade C: the progress of all randomized children in each group obviously not described (Grade A = high quality).

5. Follow up. Trials were scored as Grade A: outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures), Grade B: outcomes measured in 80% to 90%, Grade C: unclear, Grade D: outcomes measured in less than 80% (Grade A = high quality).

6. Outcome assessments. Grade A: all patients had standardized assessment (scheduled for data collection and attempted to use consistent criteria for diagnosis); Grade B: no standardized assessment, or not mentioned (Grade A = high quality).

While only the allocation concealment quality assessment was displayed in the meta‐analysis figures, we included all assessments in the Characteristics of included studies table. We measured inter‐review author reliability for the identification of high‐quality studies for each component by the Kappa statistic.

Measures of treatment effect

For the dichotomous outcome variables of each individual study, we calculated risk ratio and absolute risk reductions using an ITT analysis. For the incidence rate variables of each individual study, we estimated the log of the incidence rate ratio and the log of the standard error from the raw data using Stata version 7.0 (Stata 2004). In cases where one of the comparison groups had a rate of zero, one case was added to both groups. We carried out an initial qualitative comparison of all the individually analyzed studies to determine whether pooling of results (meta‐analysis) was reasonable. This took into account differences in study populations, inclusion and exclusion criteria, interventions, outcome assessment and estimated effect size.

Unit of analysis issues

We only included studies that randomized individual children. For cross‐over studies, we only included data collected prior to cross‐over. For studies that included non‐antibiotic treatment arms, we only included the antibiotic and control arms in the review.

Dealing with missing data

We applied the principle of an ITT analysis to available outcome data. We did not use imputation. For estimates of rates of infection where the standard deviation was not described, we assumed that events occurred with a Poisson distribution and that the variance was equal to the rate.

Assessment of heterogeneity

Both review authors considered the following sources of clinical heterogeneity: differences in participants, differences in intervention (type, dose and duration of antibiotic treatment) and differences in outcome measurement. We described any statistical heterogeneity between the study results using the I² statistic. We tested statistical significance using a Chi² test (P < 0.1 regarded as significant). We calculated the effect estimate and 95% confidence interval (CI) using a random‐effects model whenever there were concerns about statistical heterogeneity.

Assessment of reporting biases

We assessed the possibility of reporting bias qualitatively using a funnel plot. We did not used statistical tests or imputation.

Data synthesis

We included the results from studies that met the inclusion criteria and that reported any of the outcomes of interest in the subsequent meta‐analyses. We calculated the summary weighted risk ratio and 95% CI by the Mantel‐Haenszel method using a fixed‐effect model (Cochrane statistical package, Review Manager (RevMan 2008). We used the number of episodes of AOM and illness per child‐year to calculate the summary weighted incidence rate ratio by the generic inverse variance method (Cochrane statistical package, RevMan 2008). We calculated the absolute risk reduction and the number needed to treat using the summary odds ratio and the average control event rate described in the relevant studies. This was done by multiplying the odds of the event in the control arm by the odds ratio to determine the odds of the event in the intervention arm. We then converted the estimated odds to probabilities. 

Where individual studies reported outcomes at several different follow‐up times, a summary analysis used data at 'any time greater than three months' and at the time 'closest to nine months' after the intervention period. Since we were unsure if there were any long‐term effects, and if they remain constant over time, we also described the primary outcomes at the following time points after the end of the intervention period:

a. one to two months;
 b. three to five months;
 c. six to 11 months;
 d. 12 to 23 months; and
 e. greater than 23 months.

Subgroup analysis and investigation of heterogeneity

We planned a priori subgroup analyses of the primary outcomes for: 1) children aged less than 12 months, greater than 12 months or not separated at time of randomisation; 2) otitis‐prone children (defined as children having four episodes of AOM within 12 months or three episodes within a six‐month period), not otitis‐prone, or not separated; 3) high‐risk children (defined as children who have experienced an episode of AOM with perforation) at time of randomisation; and 4) children living in high‐risk populations (defined as children from study populations where the CSOM prevalence rate is 4% or more).

Sensitivity analysis

We planned sensitivity analyses of the primary outcomes to assess the impact of potentially important factors on the overall outcomes.

1. Study quality.

2. Study size.

3. Variation in the inclusion criteria.

4. Differences in the medications used in the intervention and comparison groups.

5. Differences in outcome measurement.

6. Analysis by 'treatment received' rather than ITT.

Results

Description of studies

Results of the search

We identified 17 studies that met our inclusion criteria. However one study reported acute otitis media (AOM) combined with otitis media with effusion (OME) as their outcome measure and was not able to contribute any data to the meta‐analyses (Schwartz 1982a). Sixteen studies are included in the meta‐analysis. Fifteen of the studies were identified through searching the databases described above. All of these were listed on MEDLINE. We were notified about one additional study that had been published as an extended abstract (Gray 1981). Four of the studies used a cross‐over design (Liston 1983a; Perrin 1974a; Schwartz 1982a; Varsano 1985a). For these studies we only used data from the first treatment phase. Three studies also randomized children to non‐antibiotic treatment arms: tympanostomy tubes (Casselbrant 1992a; Gonzalez 1986a) and pneumococcal vaccine (Schuller 1983a). For these studies we have only used data from the antibiotic treatment or control treatment arms.

Included studies

All the 16 studies that contributed data to the meta‐analyses randomized young children who were regarded as being at increased risk of AOM. In seven studies the children met the accepted criteria for being prone to otitis (three episodes in the previous six months or four episodes in the previous 12 months). At least half the studies excluded children who had congenital abnormalities associated with increased rates of OM (for example, cleft palate and immunodeficiency). Most studies were conducted in the USA and were completed prior to 1990. The most commonly used long‐term antibiotic treatments were sulphisoxazole and amoxicillin. Nearly all of the studies were placebo‐controlled. The long‐term antibiotics were generally given for three to six months. Antibiotics were given once or twice daily. Most studies used a combination of active surveillance (with monthly reviews) and visits to a healthcare provider because of illness to document episodes of AOM. The diagnostic criteria used for AOM were variable but generally children had to have symptoms and signs consistent with the illness. Relatively few studies provided data on the long‐term outcomes following cessation of intervention.

The age eligibility criteria for the 16 studies that contributed data to the meta‐analyses varied. Three studies specifically targeted children less than 18 months of age (Gray 1981; Leach 2008; Teele 2000a). Two additional studies did not enrol children older then 30 to 35 months of age (Casselbrant 1992a; Persico 1985a). The other 11 studies did not exclude older children. Most children in the studies were younger than five years of age.

All the studies randomized children at increased risk of AOM. This was consistent with the criteria for recurrent AOM (three episodes in six months or four episodes in 12 months) in seven studies (Casselbrant 1992a; Liston 1983a; Persico 1985a; Principi 1989a; Roark 1997a; Schuller 1983a; Varsano 1985a). In five studies children had two episodes of AOM within six months or three episodes within 18 months (Gaskins 1982a; Gonzalez 1986a; Gray 1981; Perrin 1974a; Sih 1993a). One study enrolled children who had experienced an episode of AOM in the first six months of life or two episodes in the first 12 months (Teele 2000a). One study enrolled children with frequent middle ear effusion (MEE) but infrequent AOM (Mandel 1996a). Two studies enrolled children living in populations known to have high rates of severe OM (Leach 2008; Maynard 1972a).

None of the 11 studies that described ear status at entry to the study included children with AOM at the start of treatment (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Gray 1981; Leach 2008; Liston 1983a; Mandel 1996a; Principi 1989a; Schuller 1983a; Sih 1993a; Varsano 1985a). Four of these studies only included children who were free of middle ear infection (MEE) at the start of treatment (Casselbrant 1992a; Mandel 1996a; Schuller 1983a; Sih 1993a).

Of the 11 studies that described exclusion criteria, nine excluded children with congenital abnormalities associated with OM (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Perrin 1974a; Principi 1989a; Roark 1997a). Seven studies did not describe exclusion criteria.

Most studies addressing our research question were small. Ten studies randomized fewer than 100 children (Gaskins 1982a; Gonzalez 1986a; Gray 1981; Liston 1983a; Perrin 1974a; Principi 1989a; Schuller 1983a; Schwartz 1982a; Sih 1993a; Varsano 1985a). Six studies randomized from 100 to 200 children (Casselbrant 1992a; Leach 2008; Mandel 1996a; Persico 1985a; Roark 1997a; Teele 2000a), and one study randomized more than 200 children (Maynard 1972a).

Eleven of the studies were conducted in the USA (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Gray 1981; Liston 1983a; Mandel 1996a; Maynard 1972a; Perrin 1974a; Roark 1997a; Schuller 1983a; Teele 2000a). The other studies were conducted in Israel (Persico 1985a; Varsano 1985a), Italy (Principi 1989a), Brazil (Sih 1993a) and Australia (Leach 2008).

All the studies were completed before 1995 with the possible exception of one in which the date of the study was not reported (Teele 2000a) and one completed in 2001 (Leach 2008).

A large proportion of the 16 studies that contributed data to the meta‐analyses had more than two treatment arms. Three studies randomized children to either one of two antibiotic groups (amoxicillin or trimethoprim sulphamethoxazole (or sulfisoxazole)) or a control group (Principi 1989a; Sih 1993a; Teele 2000a). One study randomized children to once‐daily or twice‐daily amoxicillin (Roark 1997a). Ten studies included sulphur drugs as their long‐term antibiotic intervention: sulfisoxazole (Gonzalez 1986a; Liston 1983a; Perrin 1974a; Schuller 1983a; Teele 2000a; Varsano 1985a); or trimethoprim sulphamethoxazole (Gaskins 1982a; Gray 1981; Principi 1989a; Sih 1993a). Nine studies included penicillins as their long‐term antibiotic intervention: amoxicillin (Casselbrant 1992a; Leach 2008; Mandel 1996a; Principi 1989a; Roark 1997a; Sih 1993a; Teele 2000a); ampicillin (Maynard 1972a); or penicillin V (Persico 1985a). Only three studies did not use a placebo treatment in their control group (Gaskins 1982a; Persico 1985a; Schuller 1983a). Two studies used a no‐treatment control group (Gaskins 1982a; Persico 1985a). One study had both a no‐treatment control group and another control group where symptomatic use of antihistamines for a blocked nose was recommended (Schuller 1983a).

In nine studies the antibiotics were given once daily (Casselbrant 1992a; Gray 1981; Mandel 1996a; Maynard 1972a; Persico 1985a; Principi 1989a; Roark 1997a; Sih 1993a; Teele 2000a). Eight studies used a twice‐daily dose (Gaskins 1982a; Gonzalez 1986a; Leach 2008; Liston 1983a; Perrin 1974a; Roark 1997a; Schuller 1983a; Varsano 1985a). One study included a once‐daily treatment arm and a twice‐daily treatment arm (Roark 1997a).

Three studies continued treatment for less than three months (Roark 1997a; Schwartz 1982a; Varsano 1985a). Three studies continued treatment for three months (Liston 1983a; Perrin 1974a; Sih 1993a). Six studies continued treatment for three to six months (Gaskins 1982a; Gonzalez 1986a; Leach 2008; Persico 1985a; Principi 1989a; Teele 2000a). Four studies continued treatment for more than six to 12 months (Gray 1981; Mandel 1996a; Maynard 1972a; Schuller 1983a), and one had a two‐year intervention period (Casselbrant 1992a).

All studies used active surveillance and the interval between clinical reviews ranged from: monthly (Casselbrant 1992a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Maynard 1972a; Roark 1997a; Sih 1993a; Teele 2000a); greater than one, to three‐monthly (Gaskins 1982a; Perrin 1974a; Principi 1989a; Varsano 1985a); three monthly (Persico 1985a; Schuller 1983a; Schwartz 1982a); or was unclear (one to three‐monthly) (Gray 1981). No studies relied only on presentation to a health provider to document episodes of illness. All but two studies, in which it was unclear (Gray 1981; Perrin 1974a), also included episodes detected during presentation.

Compliance with antibiotics was reported by most studies. Only two studies (Gray 1981; Maynard 1972a) compared outcomes for high or not‐high compliance by group.

Upper respiratory tract cultures were collected in seven studies (Casselbrant 1992a; Gray 1981; Leach 2008; Mandel 1996a; Perrin 1974a; Persico 1985a; Teele 2000a) and middle ear fluid specimens were collected in five studies (Casselbrant 1992a; Gaskins 1982a; Liston 1983a; Mandel 1996a; Persico 1985a). One study reported the proportion of swabs (but not children) with carriage and resistance of both H. influenzae and S. pneumoniae by group (Leach 2008). Two studies reported the proportion of throat or nasopharyngeal cultures positive for beta‐lactamase producing H. influenzae or M. catarrhalis, or both, by group (Casselbrant 1992a; Mandel 1996a). One study reported proportion of throat cultures positive for pneumococci (but not resistant strains) by group (Mandel 1996a). One study reported middle ear fluid culture results for S. pneumoniae (but not resistant strains) and H. influenzae (including beta‐lactamase producing strains) by group (Casselbrant 1992a). Others failed to report microbiological findings by group (Gray 1981; Perrin 1974a; Persico 1985a), combined pre‐ and post cross‐over data (Liston 1983a), or performed microbiological studies on immunocompromised patients alone (Gaskins 1982a).

Excluded studies

One study was excluded because children also received tympanostomy tubes at the time of randomisation (Koivunen 2004). None of the other excluded studies assessed the impact of antibiotics on AOM with a valid control group (see Characteristics of excluded studies table).

Risk of bias in included studies

The overall quality was high with nine studies providing a description of the random assignment or stating random assignment and providing group similarity details, and nine studies providing adequate information on allocation concealment. Fourteen studies had at least the outcome assessor blinded. Nine studies reported outcomes for all randomized children in each group; 10 studies measured outcomes in over 90% of participants and evaluated withdrawals as treatment failures. Standardised assessments were used in 14 studies. Agreement was highest for blinding (agreement 94%, kappa 0.76) and reporting by allocated group (agreement 88%, kappa 0.75); and lowest for allocation concealment (agreement 69%, kappa 0.35) and outcome assessment (agreement 81%, kappa ‐0.1). Consensus was readily achieved by re‐reviewing original publications. A third author was not required to achieve consensus (Table 3).

1. Quality of studies included in the review.

Quality component	High quality	Agreement	Kappa
Randomisation	9/17	75%	0.48
Allocation concealment	9/17	69%	0.35
Blinding	14/17	94%	0.76
Reporting by allocated group	9/17	88%	0.75
Follow up	10/17	81%	0.63
Outcome assessments	14/17	81%	‐0.1

Summary number of studies meeting high quality for each quality measure

1. Randomisation: nine studies (Casselbrant 1992a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Principi 1989a; Roark 1997a; Schwartz 1982a; Teele 2000a).

2. Allocation concealment: 10 studies (Casselbrant 1992a; Gonzalez 1986a; Gray 1981; Leach 2008; Liston 1983a; Mandel 1996a; Perrin 1974a; Roark 1997a; Schwartz 1982a; Teele 2000a).

3. Blinding: 14 studies (Casselbrant 1992a; Gonzalez 1986a; Gray 1981; Leach 2008; Liston 1983a; Mandel 1996a; Maynard 1972a; Perrin 1974a; Persico 1985a; Principi 1989a; Roark 1997a; Schwartz 1982a;Teele 2000a; Varsano 1985a).

4. Reporting by allocated group: eight studies (Casselbrant 1992a; Gaskins 1982a; Leach 2008; Mandel 1996a; Maynard 1972a; Principi 1989a; Roark 1997a; Teele 2000a).

5. Follow up: 10 studies (Casselbrant 1992a; Gaskins 1982a; Gray 1981; Leach 2008; Mandel 1996a; Maynard 1972a; Perrin 1974a; Persico 1985a; Principi 1989a; Teele 2000a).

6. Outcome assessment: 14 studies (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Persico 1985a; Principi 1989a; Roark 1997a; Schuller 1983a; Schwartz 1982a; Sih 1993a; Teele 2000a; Varsano 1985a).

We categorised four studies as high quality for all six components of quality that we assessed (Casselbrant 1992a; Leach 2008; Mandel 1996a; Teele 2000a). We categorised eight studies as high quality for randomisation and allocation concealment (Casselbrant 1992a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Perrin 1974a; Roark 1997a; Teele 2000a). We categorised 12 studies as high quality for blinding of outcome assessment (Casselbrant 1992a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Maynard 1972a; Perrin 1974a; Persico 1985a; Principi 1989a; Roark 1997a; Teele 2000a; Varsano 1985a). We decided that the studies were sufficiently clinically homogeneous for meta‐analysis.

Allocation

Overall, allocation generation and concealment was recorded and was appropriate in 9/17 studies (53%) ‐ see additional Table 3. Any associated bias may exaggerate the pooled estimate of beneficial effect of the intervention.

Blinding

Overall blinding was reported and appropriate in 14/17 studies (82%). This was achieved through the use of placebo medications. Any associated bias is unlikely to have much impact on the pooled estimate of beneficial effect of the intervention.

Incomplete outcome data

Overall, outcome data were reported and complete for more than 90% of participants in 10/17 studies (59%). Any associated bias may alter the pooled estimate of effect of the intervention. The direction of any associated bias is not clear.

Selective reporting

Most of the trials were conducted prior to requirement for trial registration. However, most studies reported the same primary outcome (making selective reporting unlikely). Selective reporting of secondary outcomes is possible.

Other potential sources of bias

There were no other important sources of bias related to the conduct of the included studies identified. Publication or small study bias is possible (see Figure 1).

1.

Funnel plot of comparison: 1 Antibiotic versus control ‐ primary outcomes, outcome: 1.1 Prevention ‐ any AOM or CSOM during intervention.

Effects of interventions

See: Table 1; Table 2

Seventeen studies involving 1586 children met our inclusion criteria. All studies enrolled children at increased risk of AOM and in seven studies the children met the accepted criteria for being prone to otitis (three episodes in the previous six months or four episodes in the previous 12 months). Most eligible studies reported data on our two primary outcomes (14 studies reported on any AOM and 13 studies reported on rates of AOM). One study combined AOM and otitis media with effusion (OME) episodes and thus contributed no data to this analysis (Schwartz 1982a). Long‐term antibiotics reduced any episode of AOM and the number of episodes of AOM. The benefits were modest (absolute risk reduction of 20%) but highly statistically significant. The results of individual studies were statistically heterogeneous but this was not sufficient to prevent pooling of studies. No study described the proportion of children with recurrent AOM or CSOM at end of treatment. Only one study provided data on long‐term outcomes. Clinical side effects of treatment were reported in 12 studies. In these studies, the use of long‐term antibiotics was not associated with a significant increase in adverse events.

For the following prespecified analyses there were no data available and no forest plots produced: i) Prevention ‐ any AOM or CSOM following cessation of intervention at one to two months, three to five months, 12 to 23 months, and more than 23 months; ii) Prevention ‐ any AOM or CSOM during intervention in high‐risk children; and iii) Prevention ‐ episodes of AOM or CSOM during intervention in high‐risk children.

01 Primary outcomes

1. Prevention ‐ any AOM/CSOM during intervention

Fourteen studies involving 1461 children assessed this outcome (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Maynard 1972a; Perrin 1974a; Persico 1985a; Principi 1989a; Roark 1997a; Sih 1993a; Teele 2000a; Varsano 1985a). The size of the studies varied from 21 to 364 participants. The quality of the included studies was variable. Overall, 293 of 800 children (37%) in the antibiotic group and 368 of 661 children (56%) in the control group experienced at least one episode of AOM during the intervention period. The pooled risk ratio was 0.65 (95% CI 0.53 to 0.79; random‐effects model, I² = 68%). We estimated that this is equivalent to an absolute risk reduction of 21%. In other words, you would need to treat five children (95% CI 4 to 6) with long‐term antibiotics to prevent one child experiencing an episode of AOM while on treatment.

2. Prevention ‐ episodes of AOM/CSOM during intervention

Thirteen studies involving 1327 children assessed this outcome (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Gray 1981; Leach 2008; Liston 1983a; Mandel 1996a; Maynard 1972a; Principi 1989a; Roark 1997a; Schuller 1983a; Sih 1993a; Varsano 1985a). The number of events in each study varied from 10 to 259 AOM episodes. The rate of AOM per child year varied from less than one to more than seven (see Table 4). We categorised three studies as high quality for all six components of quality that we assessed (Casselbrant 1992a; Leach 2008; Mandel 1996a), and seven as high quality for randomisation and allocation concealment (Casselbrant 1992a; Gonzalez 1986a; Leach 2008; Liston 1983a; Mandel 1996a; Principi 1989a; Roark 1997a). Overall, 464 events occurred in the antibiotic group and 863 in the control group. The pooled incidence rate ratio was 0.51 (95% CI 0.39 to 0.66; random‐effects model). The studies were statistically heterogeneous (P = 0.0001, I² = 73%). The average rate of AOM in children in the control group was around three episodes per child‐year. Assuming the effect of antibiotics persists over any treatment period, a child (typical of those included in these trials) will have their rate of AOM reduced to 1.5 episodes per year. That is to say, on average, antibiotics will prevent 1.5 episodes of AOM for every 12 months of treatment per child.

2. AOM episodes per child year.

Study	Antibiotic	Control
Casselbrant	0.6 (n = 86)	1.08 (n = 80)
Gaskins	0 (n = 10)	2.0 (n = 11)
Gonzalez	2.8 (n = 21)	4 (n = 80)
Gray	1.2 (n = 26)	1.6 (n = 24)
Leach	4.25 (n = 52)	5.0 (n = 51)
Liston	3.6 (n = 19)	6.7 (n = 16)
Mandel	0.28 (n = 55)	1.04 (n = 51)
Maynard	0.42 (n = 173)	0.74 (n = 191)
Principi	0.66 (n = 66)	1.68 (n = 30)
Roark	2.6(n = 99)	2.95 (n = 59)
Schuller	1.15 (n = 18)	4.9 (n = 20)
Sih	0.8 (n = 40)	2.6 (n = 20)
Varsano	2.08 (n = 15)	7.3 (n = 17)

02 Secondary outcomes ‐ first primary outcome (any AOM/CSOM during intervention)

1. Prevention ‐ any rAOM/CSOM during intervention

Five studies involving 329 children assessed this outcome (Gaskins 1982a; Liston 1983a; Principi 1989a; Sih 1993a; Teele 2000a). The size of the studies varied from 21 to 117 participants. Overall, nine of 211 (4.2%) children in the antibiotic group and 12 of 118 (10.2%) in the control group experienced recurrent AOM during the intervention period. The pooled risk ratio was 0.45 (95% CI 0.20 to 1.01, fixed‐effect model; I² = 5%). No studies reported data on CSOM.

2. Prevention ‐ any AOM/CSOM at end of intervention

One study involving 103 children assessed this outcome (Leach 2008). Eighteen of 52 versus 24 of 51 children had AOM or CSOM at the end of therapy (RR 0.74, 95% CI 0.46 to 1.18).

3. Prevention ‐ any AOM/CSOM following cessation of intervention

Only one study reported findings after the intervention period (Teele 2000a). At six to 11 months, 51 of 76 (67%) children in the antibiotic group and 32 of 41 (78%) in the control group experienced AOM (RR 0.86, 95% CI 0.69 to 1.08). None of these children were reported to have CSOM. There were no data reported on AOM following cessation of intervention for any of the other proposed follow‐up times.

4. Prevention ‐ episodes of illness during intervention

None of the studies reported directly on rates of illness during the intervention period. One study reported rates of penicillin injections for illnesses other than OM (Maynard 1972a). Fewer injections were given to children in the amoxicillin group (318 injections to 165 children or 1.93 per child‐year) than the placebo group (412 injections to 179 children or 2.30 per child‐year) (incidence rate ratio (IRR) 0.84, 95% CI 0.72 to 0.97).

5. Side effects ‐ any clinical side effects during intervention

Twelve studies involving 817 children reported side effects (Casselbrant 1992a; Gaskins 1982a; Gonzalez 1986a; Gray 1981; Leach 2008; Liston 1983a; Perrin 1974a; Principi 1989a; Schuller 1983a; Sih 1993a; Teele 2000a; Varsano 1985a). Ten of 457 children in the antibiotic group and three of 360 in the control group experienced side effects (RR 1.99, 95% CI 0.25 to 15.9; random‐effects model, I² = 53%).

6. Side effects ‐ any antibiotic resistance during intervention

Two studies reported carriage of resistant respiratory bacterial pathogens (Casselbrant 1992a; Mandel 1996a). Overall, 31 of 101 (30.7%) of the antibiotic group and 18 of 80 (22.5%) of the control group carried beta‐lactamase resistant H. influenzae or M. catarrhalis, or both, at three to four months post‐study entry (RR 1.37, 95% CI 0.83 to 2.26); fixed‐effect model, I² = 0%).

03 Subgroup analyses ‐ first and second primary outcomes (any AOM/CSOM and episodes of AOM/CSOM during intervention)

1. Children less than 12 months or age, older than 12 months or not separated

Two studies reported any AOM/CSOM during intervention in children less than 12 months of age (Leach 2008; Teele 2000a). Seventy‐one of 128 (55%) children in the antibiotic group and 65 of 92 (71%) in the control group had AOM/CSOM (RR 0.81, 95% CI 0.45 to 1.44). Two studies reported episodes of AOM/CSOM during intervention in children aged less than 12 months (Leach 2008; Roark 1997a); 127 episodes were reported in the antibiotic group and 121 in the control group (IRR of 1.00, 95% CI 0.6 to 1.7). Only one study reported any AOM/CSOM in children older than 12 months (RR 0.06, 95% CI 0.0 to 0.99) (Gaskins 1982a). Two studies reported episodes of AOM/CSOM in children older than 12 months (IRR 0.56, 95% CI 0.09 to 3.67) (Gaskins 1982a; Roark 1997a). Studies that did not separate age had a pooled risk ratio for prevention of AOM/CSOM (RR 0.7, 95% CI 0.59 to 0.84; random‐effects model, I² = 45%) that was consistent with the overall estimate.

2. Otitis‐prone children, non‐otitis prone, or not separated

Seven studies involving 636 children reported any AOM/CSOM during the intervention for otitis‐prone children (Casselbrant 1992a; Gonzalez 1986a; Liston 1983a; Persico 1985a; Principi 1989a; Roark 1997a; Varsano 1985a). There were 156 of 366 (42.6%) children in the antibiotic group and 167 of the 270 (61.9%) children in the control group with AOM/CSOM (RR 0.73, 95% CI 0.61 to 0.88; random‐effects model, I² = 33%). Eight studies reported episodes of AOM/CSOM in otitis‐prone children (Casselbrant 1992a; Gonzalez 1986a; Gray 1981; Liston 1983a; Principi 1989a; Roark 1997a; Schuller 1983a; Varsano 1985a). Overall, there were 264 episodes in the antibiotic groups and 532 in the control group (IRR 0.52, 95% CI 0.37 to 0.73; random‐effects model, I² = 73%). One study (Teele 2000a) reported slightly greater benefits in prevention of any AOM/CSOM in non‐otitis prone children (RR 0.6, 95% CI 0.42 to 0.84). Studies that did not separate otitis‐prone and non‐otitis prone had a greater benefit of antibiotics in the prevention of any AOM/CSOM (RR 0.51, 95% CI 0.30 to 0.87; random‐effects model, I² = 83%). Similarly for prevention of episodes of AOM/CSOM, benefits were greater in studies that did not separate otitis‐prone from non‐otitis prone children (RR 0.47, 95% CI 0.29 to 0.76; random‐effects model, I² = 73%).

3. High‐risk children or not high‐risk children

No studies reported outcomes for high‐risk children (defined as children who had experienced an episode of AOM with perforation) at time of randomisation. One study (Maynard 1972a) appeared to include a large proportion of children who had previously experienced AOM with perforation. The outcomes for this subgroup of children were not described.

4. High‐risk populations or not high‐risk populations

Two studies (Leach 2008; Maynard 1972a) reported outcomes from a high‐risk population (defined as a population with a CSOM prevalence rate of 4% or more). Overall, 82 of 255 children in the antibiotic group and 113 of 242 children in the control group experienced any AOM/CSOM (RR 0.81, 95% CI 0.44 to 1.5). There were 177 episodes of AOM or CSOM in the antibiotic group and 262 episodes in the control group (IRR 0.67, 95% CI 0.42 to 1.08). Removing these studies from the pooled analysis for any AOM/CSOM or episodes of AOM/CSOM did not substantially change the pooled estimates (RR 0.62, 95% CI 0.51 to 0.76; random‐effects model, I² = 56%) (IRR 0.51, 95% CI 0.3 to 0.87; random‐effects model, I² = 90%).

04 Sensitivity analyses ‐ first and second primary outcomes (any AOM/CSOM and episodes of AOM/CSOM during intervention)

We conducted a range of sensitivity analyses. In all of these meta‐analyses (excluding compliance) the pooled effect size remained consistent with the overall pooled effect size for both primary outcomes. The studies included in this meta‐analysis were statistically heterogeneous (P = 0.02, I² = 52%). This statistical heterogeneity was less marked (but still significant) if the odds ratio rather than risk ratio was used (P = 0.06, I² = 42%). We used the sensitivity analyses to explore the potential causes of this heterogeneity. All of the planned sensitivity analyses provided pooled estimates of effect that were consistent with the overall estimate (see below). Visual assessment of the funnel plot asymmetry identified four smaller studies that described the largest effect of antibiotics (RR 0.33, 95% CI 0.2 to 0.54; random‐effects model, I² = 0%). Removal of these studies from the pooled analysis did not change the pooled estimate of effect but did reduce the heterogeneity between studies (RR 0.7, 95% CI 0.6 to 0.82; I² = 34%).

1. Study quality

Pooling the results of studies with high‐quality randomisation and allocation concealment resulted in a reduced (but still statistically significant) effect of the intervention (RR 0.75, 95% CI 0.59 to 0.95; random‐effects model, I² = 67%) (IRR 0.63, 95% CI 0.45 to 0.88; random‐effects model, I² = 70%). The studies without high‐quality randomisation and allocation concealment estimated a larger effect size (RR 0.53, 95% CI 0.40 to 0.69; I² = 39%) (IRR 0.43, 95% CI 0.31 to 0.59; I² = 39%). Although most studies had at least one high‐quality feature, only three studies that reported any AOM/CSOM (283 children) were categorised as high quality for all six criteria (Casselbrant 1992a; Mandel 1996a; Teele 2000a). The pooled risk ratio for prevention of any AOM/CSOM from these three studies (RR 0.61, 95% CI 0.50 to 0.76; I² = 2.5%) removed heterogeneity but remained consistent with the overall estimate.

2. Study size

Smaller studies (fewer than 100 participants) tended to describe a larger effect size (RR 0.49, 95% CI 0.3 to 0.8; random‐effects model, I² = 74%) than larger studies (more than 100 children) (RR 0.72, 95% CI 0.57 to 0.89; I² = 70%) (Casselbrant 1992a; Leach 2008; Mandel 1996a; Maynard 1972a; Persico 1985a; Roark 1997a; Teele 2000a). Similarly for studies that reported prevention of episodes of AOM/CSOM, smaller studies indicated a larger benefit (IRR 0.42, 95% CI 0.3 to 0.61; random‐effects model, I² = 57%) than larger studies (IRR 0.62, 95% CI 0.44 to 0.87; random‐effects model, I² = 78%). This apparent study size effect may be due to publication bias (see Figure 1).

3. Variation in inclusion criteria

Sensitivity analyses examining a range of differences in entry criteria did not identify any important effect measure modification. Potentially the most important inclusion criterion is the presence or absence of ear disease at the time of randomisation. Three studies (332 children) randomized children free of middle ear effusion (MEE) at the start of treatment (Casselbrant 1992a; Mandel 1996a; Sih 1993a). The pooled risk ratio for prevention of any AOM/CSOM (RR 0.53, 95% CI 0.35 to 0.80; I² = 48%) was consistent with the overall estimate.

4. Differences in medications used (class or regimen)

A range of antibiotics, doses and regimens were used in the studies. In eight studies (1131 children) antibiotics were given once daily (Casselbrant 1992a; Mandel 1996a; Maynard 1972a; Persico 1985a; Principi 1989a; Roark 1997a; Sih 1993a; Teele 2000a). The pooled risk ratio for prevention of any AOM/CSOM was consistent with the overall estimate (RR 0.61, 95% CI 0.52 to 0.72; I² = 30%). Similarly, we were not able to identify any important differences in outcome according to the major classes of antibiotics used or the duration of the intervention period.

5. Differences in outcome measurement

In nine studies (1100 children) monthly active surveillance was applied (Casselbrant 1992a; Gonzalez 1986a; Liston 1983a; Leach 2008; Mandel 1996a; Maynard 1972a; Roark 1997a; Sih 1993a; Teele 2000a). The pooled risk ratio for prevention of any AOM/CSOM (RR 0.71, 95% CI 0.58 to 0.86; random‐effects model, I² = 67%) was consistent with the overall estimate. Less frequent surveillance (four to six‐weekly or three‐monthly) in five studies estimated a greater benefit of antibiotics for prevention of any AOM/CSOM (RR 0.45, 95% CI 0.28 to 0.72; random‐effects model, I² = 58%) and for prevention of episodes of AOM/CSOM (IRR 0.32, 95% CI 0.19 to 0.56; random‐effects model, I² = 74%).

6. Differences in the date of study

The pooled estimates of effect were consistent between studies conducted in the 1970s, 1980s and 1990s. The only study conducted since 2000 indicated no benefit of antibiotics over placebo for any AOM/CSOM (RR 1.06, 95% CI 0.85, 1.33; random‐effects model) or episodes of AOM or CSOM (IRR 0.85, 95% CI 0.65 to 1.12; random‐effects model). However, a reduction in rates of perforation was observed (an outcome not assessed in this review). At this point in time, there is a lack of evidence from randomized trials of antibiotics to prevent AOM in populations conducted in the era of relatively high rates of penicillin‐resistant pneumococcal infection.

7. Analysis by compliance

Where reported, compliance with recommended treatment was generally high and there was little difference in compliance between antibiotic and control groups. Two studies (Gray 1981; Maynard 1972a) formally compared outcomes between 'compliers' and 'non‐compliers'. One study (Maynard 1972a) reported any AOM/CSOM; a greater benefit of antibiotics over placebo was seen in the high compliance group (RR 0.47, 95% CI 0.29 to 0.76) compared to a non‐significant benefit in the not high compliance group (RR 0.77, 95% CI 0.5 to 1.19). A highly significant benefit of antibiotics over placebo in preventing episodes of AOM/CSOM was noted for high compliance (IRR 0.36, 95% CI 0.24 to 0.54; I² = 0) compared to a non‐significant benefit for not high compliance (IRR 0.87, 95% CI 0.62 to 1.22; I² = 0%). The pooled incidence rate ratio for prevention of episodes of AOM/CSOM in these studies (IRR 0.61, 95% CI 0.35 to 1.07; random‐effects model, I² = 75%) was associated with significant heterogeneity and was substantially less than the overall estimate (IRR 0.48, 95% CI 0.37 to 0.62; I² = 65%).

Discussion

Strength of evidence

The evidence contained within this review is consistent with antibiotics preventing any AOM in children at risk of future infections. This conclusion is based on the results of 14 RCTs involving 1461 children. The estimated effect size was a 35% reduction in risk of experiencing any episode of AOM (RR 0.65, 95% CI 0.53 to 0.79 for any episode of AOM). In the 13 RCTs involving 1215 children, a similar reduction in episodes of AOM/CSOM during the intervention period was described (IRR 0.51, 95% CI 0.39 to 0.66 for rates of AOM). These results were not substantially affected by a range of sensitivity analyses. The clinical importance of this beneficial effect was limited by the spontaneous reduction in rates of AOM over time in children not receiving treatment. If these studies are valid, and there is no publication bias, the probability that antibiotics appeared effective by chance alone is remote. More than half of the studies used appropriate methods of randomisation and blinding. The effect size tended to be slightly larger in the studies that were of poorer quality or were small. However, exclusion of these studies had little impact on the pooled effect size.

Overall, the observed beneficial effect of antibiotics cannot be explained by bias or co‐intervention. The pooled effect size was also clinically consistent with the results from individual studies. However, there was some statistical heterogeneity that could not be easily explained. There may be other factors not measured in the individual studies that influence outcome. The conclusion that antibiotics are effective in the prevention of AOM is supported by the knowledge that AOM is often a bacterial infection, which antibiotics act to resolve (Rosenfeld 2003). While this review provides strong evidence that antibiotics are effective in preventing AOM during the intervention period, the long‐term effects of the intervention remain uncertain. Long‐term use of antibiotics should take into account the possibility of developing antibiotic resistance and side effects, which are discussed in the 'Trade‐offs' section.

Applicability

Nearly all studies were conducted in developed countries and only two studies involved a population of children at high risk of long‐term complications such as CSOM (Maynard 1972a). Most studies enrolled children with a history of recurrent AOM. While there was statistical heterogeneity in the results this could not be explained by biologic or cultural variation, variation in compliance, or variation in baseline risk.

Although antibiotics will prevent AOM, the overall beneficial effect is not large. This is because rates of AOM generally decreased even in children randomized to placebo or no treatment. The control event rate for any AOM during the intervention period varied from 40% to 85%. Overall, 56% of children enrolled in these trials and randomized to the control group experienced one or more episodes of AOM; they experienced around three episodes of AOM per child‐year. This review did provide some evidence that the relative benefit of antibiotics was reasonably consistent across a range of risk groups. Therefore, the potential benefit is most readily determined by the baseline risk. Since the risk of disease also tends to resolve spontaneously over time, children who have either persistently frequent or severe episodes of AOM are most likely to benefit. For children living in temperate climates the baseline risk is considerably higher during winter (Rosenfeld 2003).

Other relevant information

The results of this systematic review are consistent with previous systematic reviews and evidence summaries of the use of antibiotics to prevent AOM. They are also consistent with the size of the beneficial effect described in the systematic review of pneumococcal vaccines. Antibiotics have also been shown to be similarly effective in treating AOM and other mucosal respiratory infections such as sore throat, persistent rhinosinusitis and bronchitis.

There is some evidence that the frequency of AOM diagnosed by clinicians has increased but the severity of episodes (as determined by tympanic membrane perforation) has decreased. If we assume that the total number of respiratory infections experienced by young children has not changed substantially, it may be that clinicians are identifying AOM in children previously thought to have a non‐specific viral upper respiratory infection. Whether the reduction in tympanic membrane perforations is due to changes in the severity or aetiology of the infection or earlier treatment is not clear. An appreciation of the size of the beneficial effect of antibiotics, the low risk of long‐term complications like CSOM, and the association between antibiotic use and antibiotic resistance has meant that clinicians in high‐income countries are encouraged to withhold antibiotics wherever feasible (Dowell 1999). While this approach is reasonable, clinicians should be aware that this decision is based on a trade‐off between the benefits and risks of treatment

For clinicians using antibiotics to prevent AOM in children, local knowledge of the resistance patterns of the infecting organisms may be helpful. Rates of infection with penicillin‐resistant and multi‐resistant pneumococci have increased dramatically since the mid‐1990s (Dowell 1999). Since no studies have been conducted since this time, this review was unable to assess the impact of these changes. More recently, the introduction of the conjugate pneumococcal vaccine has meant that rates of penicillin‐resistant infection are likely to decrease again in countries with a universal childhood immunisation programme.

Trade‐offs

It would appear that the benefit of antibiotics must be weighed up against the cost and inconvenience of therapy and the risk of occasional side effects. Although the increase in side effects was not statistically significant in this review, it is generally accepted that antibiotics are associated with a small increased risk of gastrointestinal side effects and allergic reactions. How often these effects lead to withdrawal of the intervention is unclear. In the studies included within this review only 10 of 405 children in the antibiotic group and three of 309 children in the control group experienced significant side effects. Not all studies in this review reported on side effects or antibiotic resistance, hence there are limitations to the power of this meta‐analysis to detect potentially important adverse events or antibiotic resistance. Antibiotics may also influence the prevalence of antibiotic resistance within the community. This was only reported by two of the studies included within the review (Casselbrant 1992a; Mandel 1996a). Although there was an increase in carriage of antibiotic‐resistant pneumococci or haemophilus in both studies, the pooled estimate was not statistically significant. While this remains an important issue, the hypothesis that antibiotic use leads to increased antibiotic resistance has still not been tested in appropriate large randomized controlled trials. Increasing antibiotic resistance may also limit the potential benefits of this therapy. Local knowledge of epidemiology, aetiology and resistance patterns is likely to be helpful. In addition to these potential side effects, the cost of purchasing the medication and the inconvenience of giving it to young children as instructed must also be taken into account.

Summary of main results

Prophylactic antibiotics reduce the risk of episodic AOM while on treatment by 20% to 50% (see Table 1). This is equivalent to a reduction of one to two episodes per year of treatment. In the studies included in this review, prophylactic antibiotics were not associated with a significant increase in side effects (see Table 2).

Overall completeness and applicability of evidence

There is a reasonable amount of evidence about the impact of prophylactic antibiotics on episodic AOM without perforation in children living in high‐income countries. There is still uncertainty about the impact of prophylactic antibiotics on episodic AOM with perforation and CSOM, and the impact on all forms of AOM in children living in low‐income countries. The results are consistent across the studies. The pooled estimates of effect are likely to be applicable to similar populations.

Quality of the evidence

Overall, 17 RCTs (1586 children) have addressed this research question. The quality of evidence included in this review is generally high (see Table 3). The results across the studies are consistent. The pooled estimates of effect are likely to be internally valid and qualitatively reliable. However, sensitivity analyses of high‐quality studies (see Data and analyses), suggest the true effect may be more modest than that described.

Potential biases in the review process

The most important limitation of this review is that publication or small study bias cannot be excluded (see Figure 1). The methods used in this review are unlikely to have introduced bias. Both review authors were investigators of one included study.

Agreements and disagreements with other studies or reviews

The findings of this systematic review are similar to the findings of other systematic reviews and evidence‐based guidelines. Although the evidence for individual studies appears consistent, the expert opinion about the role of prophylactic antibiotics has changed over time. Currently, many experts propose that the risk of antibiotic resistance outweighs the benefit of reduced infection.

Authors' conclusions

Implications for practice.

Parents should be advised that, for children at risk of future episodes of AOM, the available evidence suggests that antibiotics given once or twice daily will reduce the probability of AOM while the child is on treatment. The effects of antibiotics beyond the treatment period are unclear. The benefits indicate that around five children must be treated in order for one child to avoid experiencing one or more episodes of AOM. A typical otitis‐prone child who has experienced at least three episodes of AOM in the previous six months (or at least four episodes in the previous 12 months) will have an average rate of AOM of around three episodes in the next year even if they receive no treatment. Antibiotics will reduce the rate of AOM to 1.5 episodes per year (that is to say, antibiotics will prevent 1.5 episodes of AOM for every 12 months of treatment per child). Larger absolute benefits are likely in children who have more frequent or more severe AOM. These conclusions are based on a large number of randomized controlled trials and are not affected by sensitivity analyses. The overall pooled estimate of effect is unlikely to change dramatically as additional data become available.

Antibiotics are not without risk. In these studies around one additional child experienced diarrhea or an allergic reaction for every 100 treated. This difference was not statistically significant. Unnecessary antibiotic prescribing may also contribute to increasing rates of antibiotic resistance. Parents or children who place greatest value on avoidance of any further episodes of AOM during the treatment period and who are willing to risk the possible side effects associated with this intervention are most likely to choose antibiotic therapy.

Implications for research.

Despite increasing concern about inappropriate antibiotic prescribing for the prevention of AOM, this intervention is associated with some benefits. The challenge is to now identify the children for whom the absolute benefits outweigh the potential risks of long‐term antibiotic treatment. Consequently, large simple trials comparing the effects of antibiotics with placebo in children at greatest risk of complications of AOM, children at high risk (children with previous perforation), and children living in high‐risk populations would provide useful information. Children should be stratified according to the frequency and severity of previous infections. Outcomes should be assessed at the completion of the intervention and again at least three to six months later. Additional research is needed to examine the impact of antibiotics on the bacteria infecting the nasopharynx. Any subsequent increases in antibiotic resistance should be described.

What's new

Date	Event	Description
29 July 2013	Amended	We have corrected in the Footnotes of the Characteristics of included study table for Roark 1997a. The intervention described “Amoxicillin 20 mg/kg/d QD or BD” with QD defined in the Footnotes as “four times daily”. The original study states once daily versus twice daily, and the text of the review also reports this correctly.

History

Protocol first published: Issue 3, 2003
 Review first published: Issue 4, 2006

Date	Event	Description
6 August 2010	New search has been performed	Searches conducted. One new study identified (Leach 2008). Conclusions of review unchanged. New conflict of interest as the review authors are investigators in the new included study.
20 June 2008	Amended	Converted to new review format.
28 March 2006	New search has been performed	Searches conducted.

Notes

Protocol amendment: the outcome colonised with resistant Streptococcus pneumoniae or Haemophilus influenzae was limited to the first follow‐up visit following treatment, and at least six weeks after starting treatment. Although this meant that some of the data from some studies that measured antibiotic resistance at multiple time points was not used, we decided that the estimates of antibiotic resistance would become more comparable across a range of studies (27 August 2003).

Appendices

Appendix 1. Embase.com search strategy

18. #14 AND #17 
 17. #15 OR #16 
 16. random*:ab,ti OR placebo*:ab,ti OR factorial*:ab,ti OR crossover*:ab,ti OR 'cross‐over':ab,ti OR 'cross over':ab,ti OR volunteer*:ab,ti OR assign*:ab,ti OR allocat*:ab,ti OR ((singl* OR doubl*) NEAR/2 (blind* OR mask*)):ab,ti 
 15. 'randomized controlled trial'/exp OR 'single blind procedure'/exp OR 'double blind procedure'/exp OR 'crossover procedure'/exp 
 14. #3 AND #13 
 13. #11 OR #12 
 12. 'antibiotic prophylaxis'/de 
 11. #7 AND #10 
 10. #8 OR #9 
 9. chemoprevent*:ab,ti OR chemoprophyla*:ab,ti OR prophyla*:ab,ti OR prevent*:ab,ti 
 8. 'chemoprophylaxis'/de 
 7. #4 OR #5 OR #6 
 6. amoxicillin:ab,ti OR amoxycillin:ab,ti OR ampicillin:ab,ti OR azithromycin:ab,ti OR azithromicin:ab,ti OR cefaclor:ab,ti OR penicillin:ab,ti OR sulphamethoxazole:ab,ti OR sulphisoxazole:ab,ti OR cotrimoxazole:ab,ti 
 5. antibiotic*:ab,ti 
 4. 'antibiotic agent'/exp 
 3. #1 OR #2 
 2. 'otitis media':ab,ti OR aom:ab,ti OR ome:ab,ti OR csom:ab,ti 
 1. 'otitis media'/exp

Data and analyses

Comparison 1. Antibiotic versus control ‐ primary outcomes.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Prevention ‐ any AOM or CSOM during intervention	14	1461	Risk Ratio (M‐H, Random, 95% CI)	0.65 [0.53, 0.79]
2 Prevention ‐ episodes of AOM or CSOM during intervention	13	1327	Incidence Rate Ratio (Random, 95% CI)	0.51 [0.39, 0.66]

1.1. Analysis.

Comparison 1 Antibiotic versus control ‐ primary outcomes, Outcome 1 Prevention ‐ any AOM or CSOM during intervention.

1.2. Analysis.

Comparison 1 Antibiotic versus control ‐ primary outcomes, Outcome 2 Prevention ‐ episodes of AOM or CSOM during intervention.

Comparison 2. Antibiotic versus control ‐ secondary outcomes.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Prevention ‐ any rAOM or CSOM during intervention	5	329	Risk Ratio (M‐H, Fixed, 95% CI)	0.45 [0.20, 1.01]
2 Prevention ‐ any AOM or CSOM at end of intervention	1	103	Risk Ratio (M‐H, Fixed, 95% CI)	0.74 [0.46, 1.18]
3 Prevention ‐ any AOM or CSOM following cessation of intervention	1		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
3.1 More than three months	1	117	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.69, 1.08]
3.2 Six to 11 months	1	117	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.69, 1.08]
4 Prevention ‐ episodes of illness during intervention	1	224	incidence rate ratio (Fixed, 95% CI)	0.84 [0.72, 0.97]
5 Side effects ‐ any clinical side effects during intervention	12	817	Risk Ratio (M‐H, Random, 95% CI)	1.99 [0.25, 15.89]
6 Side effects ‐ any antibiotic resistance during intervention	2	181	Risk Ratio (M‐H, Fixed, 95% CI)	1.37 [0.83, 2.26]

2.1. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 1 Prevention ‐ any rAOM or CSOM during intervention.

2.2. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 2 Prevention ‐ any AOM or CSOM at end of intervention.

2.3. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 3 Prevention ‐ any AOM or CSOM following cessation of intervention.

2.4. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 4 Prevention ‐ episodes of illness during intervention.

2.5. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 5 Side effects ‐ any clinical side effects during intervention.

2.6. Analysis.

Comparison 2 Antibiotic versus control ‐ secondary outcomes, Outcome 6 Side effects ‐ any antibiotic resistance during intervention.

Comparison 3. Antibiotic versus control ‐ subgroup analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Prevention ‐ any AOM or CSOM during intervention: children < 12 months, more than 12 months, or not separated	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.1 Less than 12 months	2	220	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.45, 1.44]
1.2 More than 12 months	1	21	Risk Ratio (M‐H, Random, 95% CI)	0.06 [0.00, 0.99]
1.3 Not separated	11	1190	Risk Ratio (M‐H, Random, 95% CI)	0.70 [0.59, 0.84]
2 Prevention ‐ episodes of AOM or CSOM during intervention in children less than 12 months	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
2.1 Less than 12 months	2	248	Incidence Rate Ratio (Random, 95% CI)	1.00 [0.60, 1.67]
2.2 More than 12 months	2	33	Incidence Rate Ratio (Random, 95% CI)	0.56 [0.09, 3.67]
2.3 Not separated	10	1046	Incidence Rate Ratio (Random, 95% CI)	0.45 [0.35, 0.57]
3 Prevention ‐ any AOM or CSOM during intervention: otitis prone, non‐otitis‐prone, or not separated	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
3.1 Otitis prone	7	636	Risk Ratio (M‐H, Random, 95% CI)	0.73 [0.61, 0.88]
3.2 Non‐otitis prone	1	117	Risk Ratio (M‐H, Random, 95% CI)	0.60 [0.42, 0.84]
3.3 Not separated	6	706	Risk Ratio (M‐H, Random, 95% CI)	0.51 [0.30, 0.87]
4 Prevention ‐ episodes of AOM or CSOM during intervention in OP, non‐OP or separated	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
4.1 Otitis‐prone	8	796	Incidence Rate Ratio (Random, 95% CI)	0.52 [0.37, 0.73]
4.2 Non‐otitis prone	0	0	Incidence Rate Ratio (Random, 95% CI)	0.0 [0.0, 0.0]
4.3 Not separated	5	531	Incidence Rate Ratio (Random, 95% CI)	0.47 [0.29, 0.76]
5 Prevention ‐ any AOM or CSOM during intervention in high‐risk or not high‐risk populations	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
5.1 High‐risk populations	2	467	Risk Ratio (M‐H, Random, 95% CI)	0.81 [0.44, 1.50]
5.2 Not high‐risk populations	12	994	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.51, 0.76]
6 Prevention ‐ episodes of AOM or CSOM during intervention in high‐risk or not high‐risk population	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
6.1 High‐risk populations	2	439	Incidence Rate Ratio (Random, 95% CI)	0.67 [0.42, 1.08]
6.2 Not high‐risk populations	11	888	Incidence Rate Ratio (Random, 95% CI)	0.51 [0.30, 0.87]

3.1. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 1 Prevention ‐ any AOM or CSOM during intervention: children < 12 months, more than 12 months, or not separated.

3.2. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 2 Prevention ‐ episodes of AOM or CSOM during intervention in children less than 12 months.

3.3. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 3 Prevention ‐ any AOM or CSOM during intervention: otitis prone, non‐otitis‐prone, or not separated.

3.4. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 4 Prevention ‐ episodes of AOM or CSOM during intervention in OP, non‐OP or separated.

3.5. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 5 Prevention ‐ any AOM or CSOM during intervention in high‐risk or not high‐risk populations.

3.6. Analysis.

Comparison 3 Antibiotic versus control ‐ subgroup analyses, Outcome 6 Prevention ‐ episodes of AOM or CSOM during intervention in high‐risk or not high‐risk population.

Comparison 4. Antibiotic versus control ‐ sensitivity analyses.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Prevention ‐ any AOM or CSOM during intervention by study quality	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.1 High‐quality randomisation and allocation concealment	8	780	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.59, 0.95]
1.2 Not high‐quality randomisation and allocation concealment	6	681	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.40, 0.69]
1.3 High‐quality for blinding of outcome assessment	12	1380	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.56, 0.82]
1.4 High‐quality for all six criteria	4	492	Risk Ratio (M‐H, Random, 95% CI)	0.69 [0.48, 1.01]
2 Prevention ‐ episodes of AOM or CSOM during intervention by study quality	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
2.1 High quality for randomisation and allocation concealment	6	667	Incidence Rate Ratio (Random, 95% CI)	0.63 [0.45, 0.88]
2.2 Not high quality for randomisation and allocation concealment	7	660	Incidence Rate Ratio (Random, 95% CI)	0.43 [0.31, 0.59]
2.3 High quality for blinding of outcome assessment	10	1036	Incidence Rate Ratio (Random, 95% CI)	0.58 [0.46, 0.74]
2.4 High quality for all six criteria	3	499	Incidence Rate Ratio (Random, 95% CI)	0.54 [0.33, 0.90]
3 Prevention ‐ any AOM or CSOM during intervention by study size	14	1461	Risk Ratio (M‐H, Random, 95% CI)	0.65 [0.53, 0.79]
3.1 Fewer than 100 participants	7	339	Risk Ratio (M‐H, Random, 95% CI)	0.49 [0.30, 0.81]
3.2 Greater than 100 participants	7	1122	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.57, 0.89]
4 Prevention ‐ episodes of AOM or CSOM during intervention by study size	13	1327	Incidence Rate Ratio (Random, 95% CI)	0.51 [0.39, 0.66]
4.1 Fewer than 100 participants	8	548	Incidence Rate Ratio (Random, 95% CI)	0.42 [0.30, 0.61]
4.2 Greater than 100 participants	5	779	Incidence Rate Ratio (Random, 95% CI)	0.62 [0.44, 0.87]
5 Prevention ‐ any AOM or CSOM during intervention by inclusion criteria	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
5.1 Excluded age greater than 36 months	4	494	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.57, 0.98]
5.2 Included age greater than 36 months	10	967	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.43, 0.77]
5.3 Otitis‐prone at entry (2/6 or 3/12)	6	595	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.54, 0.85]
5.4 Less otitis‐prone at entry (2/6 or 3/18; 1 before 6 months of age or 2/12)	5	293	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.29, 0.88]
5.5 Free of MEE at entry	3	332	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.35, 0.80]
5.6 Free of MEE at entry not required	5	225	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.33, 0.99]
5.7 Excluded congenital abnormalities	9	798	Risk Ratio (M‐H, Random, 95% CI)	0.65 [0.46, 0.91]
5.8 Exclusion criteria not described	5	681	Risk Ratio (M‐H, Random, 95% CI)	0.61 [0.51, 0.71]
5.9 Conducted in USA	8	1021	Risk Ratio (M‐H, Random, 95% CI)	0.65 [0.52, 0.81]
5.10 Not conducted in USA	5	399	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.37, 0.90]
6 Prevention ‐ episodes of AOM or CSOM during intervention by inclusion criteria	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
6.1 Excluded age greater than 36 months	3	509	Incidence Rate Ratio (Random, 95% CI)	0.70 [0.52, 0.94]
6.2 Included age greater than 36 months	10	818	Incidence Rate Ratio (Random, 95% CI)	0.44 [0.31, 0.61]
6.3 Otitis‐prone at entry (2/6 or 3/12)	6	657	Incidence Rate Ratio (Random, 95% CI)	0.47 [0.30, 0.72]
6.4 Less otitis‐prone at entry (2/6 or 3/18; 1 before 6 months of age or 2/12)	4	171	Incidence Rate Ratio (Random, 95% CI)	0.59 [0.38, 0.91]
6.5 Free of MEE at entry	3	306	Incidence Rate Ratio (Random, 95% CI)	0.39 [0.22, 0.67]
6.6 Free of MEE at entry not required	7	526	Incidence Rate Ratio (Random, 95% CI)	0.44 [0.30, 0.65]
6.7 Excluded congenital abnormalities	8	722	Incidence Rate Ratio (Random, 95% CI)	0.58 [0.42, 0.80]
6.8 Exclusion criteria not described	5	605	Incidence Rate Ratio (Random, 95% CI)	0.41 [0.27, 0.64]
6.9 Conducted in USA	8	946	Incidence Rate Ratio (Random, 95% CI)	0.50 [0.36, 0.70]
6.10 Not conducted in USA	4	312	Incidence Rate Ratio (Random, 95% CI)	0.44 [0.23, 0.85]
7 Prevention ‐ any AOM or CSOM during intervention by medication	14		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
7.1 Amoxicillin, penicillin	9	1189	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.50, 0.82]
7.2 Sulfisoxazole, trimethoprim sulfamethoxazole	8	363	Risk Ratio (M‐H, Random, 95% CI)	0.58 [0.40, 0.84]
7.3 Not placebo‐controlled	2	129	Risk Ratio (M‐H, Random, 95% CI)	0.28 [0.02, 3.69]
7.4 Once daily	8	1131	Risk Ratio (M‐H, Random, 95% CI)	0.61 [0.52, 0.72]
7.5 Twice daily	7	389	Risk Ratio (M‐H, Random, 95% CI)	0.78 [0.56, 1.09]
7.6 Less than 3 months therapy	2	190	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.26, 1.72]
7.7 Three to six months therapy	9	635	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.48, 0.86]
7.8 More than six months therapy	3	636	Risk Ratio (M‐H, Random, 95% CI)	0.62 [0.50, 0.76]
8 Prevention ‐ episodes of AOM or CSOM during intervention by medication	13		Incidence Rate Ratio (Random, 95% CI)	Subtotals only
8.1 Amoxicillin, penicillin	7	846	Incidence Rate Ratio (Random, 95% CI)	0.55 [0.40, 0.76]
8.2 Sulfisoxazole, trimethoprim sulfamethoxazole	7	503	Incidence Rate Ratio (Random, 95% CI)	0.52 [0.28, 0.96]
8.3 Not placebo‐controlled	2	269	Incidence Rate Ratio (Random, 95% CI)	0.27 [0.19, 0.39]
8.4 Once daily	7	701	Incidence Rate Ratio (Random, 95% CI)	0.53 [0.39, 0.71]
8.5 Twice daily	7	682	Incidence Rate Ratio (Random, 95% CI)	0.54 [0.33, 0.90]
8.6 Less than three months therapy	2	86	Incidence Rate Ratio (Random, 95% CI)	0.62 [0.15, 2.53]
8.7 Three to six months therapy	6	404	Incidence Rate Ratio (Random, 95% CI)	0.54 [0.36, 0.81]
8.8 More than six months therapy	5	837	Incidence Rate Ratio (Random, 95% CI)	0.45 [0.32, 0.63]
9 Prevention ‐ any AOM or CSOM during intervention by outcome measure	14	1411	Risk Ratio (M‐H, Random, 95% CI)	0.64 [0.53, 0.77]
9.1 Monthly active surveillance	9	1100	Risk Ratio (M‐H, Random, 95% CI)	0.71 [0.58, 0.86]
9.2 Greater than one to three‐monthly active surveillance	5	311	Risk Ratio (M‐H, Random, 95% CI)	0.45 [0.28, 0.72]
10 Prevention ‐ episodes of AOM or CSOM during intervention by outcome measure	13	1327	Incidence Rate Ratio (Random, 95% CI)	0.47 [0.36, 0.63]
10.1 Monthly active surveillance	8	913	Incidence Rate Ratio (Random, 95% CI)	0.59 [0.45, 0.77]
10.2 Four to six‐weekly, or three‐monthly active surveillance	5	414	Incidence Rate Ratio (Random, 95% CI)	0.32 [0.19, 0.56]
11 Prevention ‐ any AOM or CSOM during intervention by date of study	12	1312	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.53, 0.83]
11.1 Completed in 1970s	4	474	Risk Ratio (M‐H, Random, 95% CI)	0.56 [0.34, 0.94]
11.2 Completed in 1980s	4	411	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.53, 0.87]
11.3 Completed in 1990s	3	324	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.29, 1.11]
11.4 Completed in 2000s	1	103	Risk Ratio (M‐H, Random, 95% CI)	1.06 [0.85, 1.33]
12 Prevention ‐ episodes of AOM or CSOM during intervention by date of study	8	734	Incidence Rate Ratio (Random, 95% CI)	0.56 [0.41, 0.77]
12.1 Completed in 1970s	1	43	Incidence Rate Ratio (Random, 95% CI)	0.50 [0.26, 0.98]
12.2 Completed in 1980s	3	338	Incidence Rate Ratio (Random, 95% CI)	0.55 [0.42, 0.71]
12.3 Completed in 1990s	3	138	Incidence Rate Ratio (Random, 95% CI)	0.47 [0.16, 1.35]
12.4 Completed in 2000s	1	215	Incidence Rate Ratio (Random, 95% CI)	0.85 [0.65, 1.12]
13 Prevention ‐ any AOM or CSOM during intervention by compliance	1	364	Risk Ratio (M‐H, Fixed, 95% CI)	0.61 [0.44, 0.84]
13.1 High compliance	1	171	Risk Ratio (M‐H, Fixed, 95% CI)	0.47 [0.29, 0.76]
13.2 Not high compliance	1	193	Risk Ratio (M‐H, Fixed, 95% CI)	0.77 [0.50, 1.19]
14 Prevention ‐ episodes of AOM or CSOM during intervention by compliance	2	293	Incidence Rate Ratio (Random, 95% CI)	0.61 [0.35, 1.07]
14.1 High compliance	2	140	Incidence Rate Ratio (Random, 95% CI)	0.36 [0.24, 0.54]
14.2 Not high compliance	2	153	Incidence Rate Ratio (Random, 95% CI)	0.87 [0.62, 1.22]

4.1. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 1 Prevention ‐ any AOM or CSOM during intervention by study quality.

4.2. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 2 Prevention ‐ episodes of AOM or CSOM during intervention by study quality.

4.3. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 3 Prevention ‐ any AOM or CSOM during intervention by study size.

4.4. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 4 Prevention ‐ episodes of AOM or CSOM during intervention by study size.

4.5. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 5 Prevention ‐ any AOM or CSOM during intervention by inclusion criteria.

4.6. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 6 Prevention ‐ episodes of AOM or CSOM during intervention by inclusion criteria.

4.7. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 7 Prevention ‐ any AOM or CSOM during intervention by medication.

4.8. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 8 Prevention ‐ episodes of AOM or CSOM during intervention by medication.

4.9. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 9 Prevention ‐ any AOM or CSOM during intervention by outcome measure.

4.10. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 10 Prevention ‐ episodes of AOM or CSOM during intervention by outcome measure.

4.11. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 11 Prevention ‐ any AOM or CSOM during intervention by date of study.

4.12. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 12 Prevention ‐ episodes of AOM or CSOM during intervention by date of study.

4.13. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 13 Prevention ‐ any AOM or CSOM during intervention by compliance.

4.14. Analysis.

Comparison 4 Antibiotic versus control ‐ sensitivity analyses, Outcome 14 Prevention ‐ episodes of AOM or CSOM during intervention by compliance.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Casselbrant 1992a.

Methods	Double‐blind Random assignment (method not stated) Stratified by seasonal occurrence of previous AOM and by age group Review monthly plus ENT illnesses Group similarity reported (T1)
Participants	Pittsburgh, USA March 1981 to January 1988 Eligibility ‐ children 7 to 35 months (mo) of age with rAOM (more than 2 AOMs in 6 mo or more than 3 in 12 mo if mainly in previous 6 mo). Exclusion ‐ potentially complicating or confounding conditions Number eligible not given. 264 enrolled, allocation: 90 amoxicillin; 88 placebo; 86 tympanostomy (not used in this analysis) Baseline characteristics age, recent OM history, sex, race, site, parent occupation and season comparable for all 264
Interventions	2 years Amoxicillin (20 mg/kg/d, daily (OD)) Placebo
Outcomes	Measured in more than 90% Format of reporting ‐ rate (episodes of new AOM per child‐year) (95% CI), median time to first AOM, mean proportion of time with any OM (95% CI) Diagnostic criteria ‐ otoscopic signs, OR more than 1 symptom in the presence of MEE of both. Signs ‐ erythema or white opacity, fullness or bulging and decreased mobility. Symptoms ‐ fever, otalgia and irritability Subgroups ‐ age, seasonality of OM history, sex, race, season at entry Microbiology ‐ 3‐monthly NP and OP ‐ only report carriage of beta‐lactamase positive Hi or M. cat by allocated group (AOM and OME combined). MEF ‐ Spn, Hi and M. cat and beta‐lactamase positive Hi and M. cat by allocated group and diagnosis
Notes	Q: AAAAAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Low risk	No other potential source of bias identified

Gaskins 1982a.

Methods	Not blinded Random assignment, method of randomisation not clear Review 6 weekly plus at presentations Group similarity reported (T1)
Participants	San Juan, Puerto Rico Late fall to early Spring. Prior to May 1982 Eligibility ‐ children 1 year (y) to 14 y of age. 'High risk' more than 3 AOMs in 18 mo, or 5 AOMs. Exclusion ‐ cleft palate, drug sensitivity, tubes Number eligible not given. 21 enrolled. Allocation: 10 TMP/SMX, 11 not treated Baseline characteristics for all; age, race, risk, recent OM history comparable by group
Interventions	6 months TMP/SMX/ 5 to 8/25 to 40 mg/kg/d BD. Mean dose 6.8/34 No treatment controls
Outcomes	Measured in more than 90% Format of reporting ‐ episodes of AOM per child during therapy Diagnostic criteria ‐ general symptoms plus any 2 signs; red TM, bulging TM, no landmarks, abnormal pneumatic otoscopy, tympanocentesis +ve (only 4 done) Subgroups ‐ none Microbiology ‐ none
Notes	Q:BBCAAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	B ‐ randomisation stated but method not described, or suspect method
Allocation concealment (selection bias)	Unclear risk	B ‐ unclear
Blinding (performance bias and detection bias) All outcomes	Unclear risk	C ‐ unclear
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Gonzalez 1986a.

Methods	Double‐blind Random number series. Randomised at Pharmacy Dept. Review monthly plus when AOM suspected Group similarity reported (T1)
Participants	Aurora CO and Tacoma WA, USA 1982 to 1985 Eligibility ‐ children 6 mo to 10 y of age with rAOM (more than 2 AOMs in 6 mo or more than 4 AOMs in 18 mo). With or without OME at entry. Exclusion: cleft palate, Down's syndrome, tubes, sensitivity Number eligible, enrolled or allocated unclear. 65 completed 6‐month follow up, 63 analysed; 21 sulfisoxazole, 20 placebo, 22 tubes Baseline characteristics for 63; sex, mean age (CI not provided) comparable by group
Interventions	6 months Sulfisoxazole (500 mg for less than 5 y; 1 g for more than 5 y) Placebo
Outcomes	Measured in less than 90% Format of reporting ‐ % patients with 0, 1, 2, 3 or 4 episodes of AOM. Attack rate (AOMs per child during therapy), AOM‐free, more than 0 AOM, more than 1 AOM in 3 mo (treatment failure) Diagnostic criteria ‐ rapid and short onset of signs and symptoms of inflammation (otalgia (ear tugging in infant), fever, bulging or erythema, decreased mobility, loss of TM landmarks, otorrhoea) Subgroups ‐ diagnosis at entry Microbiology ‐ none
Notes	Q: AAACCA Excluded children more than 4 y (unclear if randomized)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	C ‐ the progress of all randomized children in each group obviously not described C ‐ unclear
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Gray 1981.

Methods	Double‐blind Randomly assigned. Dispensed by pharmacy Review schedule not provided. Possibly presentations with illness only Group similarity reported by allocated group and compliance category within the SXT group (text)
Participants	Alabama, USA Eligibility ‐ infants with more than or equal to 2 OM beginning in the first 12 mo of life. At least 1 episode documented by culture. No exclusion criteria listed Number eligible not provided. 48 enrolled; 26 SXT, 24 placebo (2 infants re‐enrolled) Baseline characteristics ‐ OM history and age comparable by group
Interventions	12 months Sulfamethoxazole‐trimethoprim (4 mg TMP per 20 mg SX per kg/d). Daily ‐ possibly once daily Placebo
Outcomes	Measured in more than 90% Format of reporting ‐ episodes of OM per group. Mean +/‐SD episodes per patient Diagnostic criteria ‐ tympanocentesis and culture Subgroup ‐ compliance Microbiology ‐ periodic cultures. Reported number of types carried and proportion with resistance by allocated group
Notes	Q: CAABAB Two infants had 1 y placebo, 1 y SXT
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	C ‐ randomisation claimed but not described, and outcome assessor not blinded
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	B ‐ unclear or no mention of withdrawals or drop‐outs A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Leach 2008.

Methods	Double‐blind, placebo‐controlled Random sequence computer‐generated with allocation concealed. Stratified by age Review 4 weekly Group similarity reported (T2)
Participants	Australia. 3 remote Aboriginal communities 70 km from Darwin in tropical northern Australia 1996 to 2001 (prior to conjugate vaccine introduction) Eligibility ‐ Aboriginal children less than 12 months of age with unilateral or bilateral OME. Exclusion ‐ less than 34 weeks gestation, chronic infection requiring prophylactic antibiotic therapy, craniofacial abnormalities or immune deficiency syndromes Number eligible ‐ 188 births, 126 assessed for eligibility, 103 randomized, allocation: 52 amoxicillin; 51 placebo Baseline characteristics age, gestation, birth weight, sex, recent OM history comparable for all 103 except for birth weight (placebo group heavier)
Interventions	6 months Amoxicillin (50 mg/kg/d, daily (BD)) Placebo
Outcomes	Measured and analysed in all children Format of reporting ‐ ear state (primary outcome) and nasopharyngeal carriage at end of therapy (T3), ear state and nasopharyngeal carriage during therapy (additional file 3), rate (95% CI) ‐ episodes of new AOMwoP and AOMwiP per child year, nasopharyngeal carriage per OM pathogen (additional file 4) Diagnostic criteria (T1) ‐ type B tympanogram plus otoscopic signs of moderate or marked bulging of the tympanic membrane. Ear examinations were categorised as AOM without perforation (AOMwoP) even if the child was asymptomatic Subgroups ‐ none reported Microbiology ‐ 4 weekly NP ‐ report carriage of Spn, Hi and M. cat and penicillin non‐susceptible Spn and beta‐lactamase positive Hi by allocated group
Notes	Q: AAAAAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Low risk	No other potential source of bias identified

Liston 1983a.

Methods	Double‐blind Random number table. Hospital pharmacist Review monthly plus illnesses Group similarity reported by group (T2)
Participants	San Antonio, Texas, USA Jan 1978 to June 1979 Eligibility ‐ children 6 mo to 5 y of age. Available for 6 mo. No current AOM. 3 OMs, one every 2 mo (2 diagnosed by author). Exclusion ‐ cleft palate, tubes or perforations, known immunodeficiency Number eligible not provided, 42 enrolled, 43 allocated; 22 sulfisoxazole (one twice), 21 placebo. 35 analysed Baseline characteristics age, sex, OM history, allergy, season not reported by group
Interventions	3 months Sulfisoxazole 75 mg/kg/d BD Placebo
Outcomes	Measured in less than 90% Format of reporting ‐ no. children with 0, 1, 2, 3, 4 or 5 episodes of AOM, by GP, before and after cross‐over. Rate ‐ episodes per patient month by GP Diagnostic criteria ‐ bulging, opaque or red plus loss of landmarks plus either 'B' tymp or discharge through natural perforation Subgroups ‐ age, sex, OM and allergy histories, child care, season Microbiology ‐ MEF cultures only. Spn, Hi. Oxacillin screen, beta‐lactamase. Cross‐over periods combined
Notes	Q: AAACBA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	C ‐ the progress of all randomized children in each group obviously not described B ‐ outcomes measured in 80% to 90%
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Mandel 1996a.

Methods	Double‐blind Randomly assigned in blocks of 4 within 12 stratified groups Stratified by criterion for entry (OM history) and age Review monthly and at ENT illness Group similarity reported (T1)
Participants	Pittsburgh, USA April 1988 to October 1991 Eligibility ‐ children 7 mo to 12 years of age and effusion‐free, 4 months OME or >3 episodes OME in previous 12 months and 1 recent episode, and <= 2 episodes AOM in previous 12 months. Entry criteria revised to 1) if TT in previous 12 mo, TM intact and MEE for >= 2 weeks in previous 3 months; 2) if no TT in previous 12 months, >= 3 episodes or 3 months MEE in previous 12 months, or both. All subjects required documentation of MEE by ENT or research centre clinic at least once or by tympanogram in previous 3 months. Exclusion ‐ potentially complicating or confounding conditions Number enrolled 111, 79 completed (25 had less than 1‐year follow up), allocation: 57 amoxicillin, 54 placebo Baseline characteristics: recent OM history, age, race, sex, allergy, socio‐economic status and hearing comparable by group
Interventions	12 months Amoxicillin (20 mg/kg/d OD) Placebo
Outcomes	Measured in more than 90% Format of reporting ‐ Rates: no. AOM episodes per person year by allocated GP; no. children with 0, 1, 2, more than or equal to 3 new episodes of AOM; proportion of children with rAOM, rOME, rMEE; % time with MEE Diagnostic criteria ‐ otoscopically more than or equal to 1 sign of MEE AND more than or equal 1 symptom of acute infection. Signs ‐ erythema, white opacity, fullness or bulging, fluid level, acute perforation. Symptoms ‐ fever, otalgia, irritability Subgroups ‐ none Microbiology ‐ 4‐monthly throat swabs ‐ report combined carriage of beta‐lactamase positive Hi or M. cat, and Spn by allocated group
Notes	Q: AAAAAA Results exclude subjects with no follow up (2 amox, 3 placebo)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Maynard 1972a.

Methods	Double‐blind Randomisation not described Longitudinal clinical trial. Matched by family size, age, sex, OM history Review monthly plus presentation, standard care for AOM episodes Group similarity (age) reported (T1)
Participants	West Alaskan villages, USA Published Jan 1972. Probably conducted between 1969 and 1971 Eligibility ‐ children 0 to 7 y of age previously studied. If consented. Exclusion criteria not defined Number eligible not given. 364 enrolled. 173 ampicillin and 191 placebo. All allocated Baseline characteristics by group: age 49% and 47% less than 3 y, clinical for 153 and 164 (15% and 12% normal ear drums; 92% and 92% intact tonsils), incidence of OM in preceding year for 164 and 175 "comparable"
Interventions	12 months Ampicillin (125/5 ml; age 2.5 years 5 ml/d; age more than 2.5 years 10 ml/d) Placebo Single dose
Outcomes	Measured in more than 90% Format of reporting ‐ episodes of AOM. New episode after 2 weeks without drainage Diagnostic criteria ‐ AOM = spontaneous occurrence of purulent drainage from 1 or both ears Subgroups ‐ none Microbiology ‐ none
Notes	Q: DBAAAB
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	D ‐ randomisation not mentioned
Allocation concealment (selection bias)	Unclear risk	B ‐ unclear
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Perrin 1974a.

Methods	Double‐blind Cross‐over Randomisation and allocation concealment stated but method not described Review 6 weekly Group similarity not reported
Participants	Rochester, NY, USA Dec 1972 to May 1973 Eligibility ‐ children aged 11 mo to 8 y with rAOM (more than or equal to 3 AOMs in 18 m or more than or equal to 5 anytime). Two enrolled did not meet clinical criteria. Exclusion ‐ anatomic defects, tubes Number eligible 70, 9 refusals. 61 enrolled. 54 completed the study, allocation: 28 sulfonamide, 26 placebo Baseline characteristics by group not provided
Interventions	3 months Sulfisoxazole 500 mg/5 ml. One teaspoon twice daily Placebo
Outcomes	Measured in more than 90% Format of reporting ‐ episodes of AOM. Proportion of children with more than 1 AOM in 3 mo. Proportion of children with MEE at 1 mo, 3 mo, 12 mo post‐therapy Diagnostic criteria ‐ clinician diagnosed Subgroups ‐ none Microbiology ‐ NP swabs during final 3 visits
Notes	Q: BAACAB Definition of post‐therapy unclear ‐ after cross‐over?
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	B ‐ randomisation stated but method not described, or suspect method
Allocation concealment (selection bias)	Low risk	A ‐ Adequate
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	C ‐ the progress of all randomized children in each group obviously not described A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Persico 1985a.

Methods	Outcome assessor blinded Randomly divided Review 3‐monthly and follow up 6 to 15 mo Group similarity not reported (matched for age ? pre‐ or post‐allocation)
Participants	Israel, Haifa Submitted May 1984 Eligibility ‐ children aged 3 to 30 mo with more than one AOM or AOM with discharge per mo during a 3 mo period Exclusion ‐ numbers eligible not provided, 111 enrolled, 108 with data on primary outcome; 60 active, 48 no treatment Baseline characteristics: age distribution given by group but the control group adds to 111%
Interventions	At least 3 months Potassium phenoxymethyl penicillin V 25 mg/kg/d No treatment
Outcomes	Measured in more than 90% Format of reporting ‐ episodes of AOM. Subacute AOM, SOM Diagnostic criteria ‐ bulging, red, immobile TM +/‐ URT symptoms Subgroups ‐ none Microbiology ‐ table missing from publication
Notes	Q: CBBBAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	C ‐ randomisation claimed but not described, and outcome assessor not blinded
Allocation concealment (selection bias)	Unclear risk	B ‐ Unclear
Blinding (performance bias and detection bias) All outcomes	Low risk	B ‐ outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	B ‐ unclear or no mention of withdrawals or drop‐outs A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Principi 1989a.

Methods	Double‐blind Assigned randomly. Stated but method not described Review 4 to 6 weekly and when AOM or URTI suspected Group similarity described (T1)
Participants	Milan, Italy Oct 1986 to Sept 1988 Children aged 9 mo to 5 y with more than 2 AOMs in preceding 6 mo. No AOM at entry. Exclusion ‐ cleft palate, Down's syndrome, drug allergy, immunodeficiency, tubes Numbers eligible not provided, 100 enrolled. Allocation: 34 amoxicillin, 33 SXT, 33 placebo. 96 reported Baseline characteristics; sex, age, diagnosis, time to most recent AOM, season, day care, atopy by group
Interventions	6 months Amox : 20 mg/kg/d. Sulfamethoxazole and trimethoprim (SXT) : 12 mg/kg/d. OD Placebo
Outcomes	Measured in more than 90% Format of reporting ‐ number of children with number of episodes of AOM during therapy. Mean no. AOMs per patient, mean no AOMs per patient‐month Diagnostic criteria ‐ AOM = any combination of otalgia and irritation and hyperaemia or opacity plus fullness, bulging or type B Subgroups ‐ none Microbiology ‐ none
Notes	Q: ABAAAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Unclear risk	B ‐ unclear
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Roark 1997a.

Methods	Double‐blind Randomly assigned (method not described) Review monthly plus during URT plus onset of pain, fever or severe crying Group similarity reported (T1)
Participants	Denver, CO, USA Dec 1991 to Sept 1994 Eligibility ‐ children 3 mo to 6 y of age with rAOM. Diagnosis at randomisation not clear. Exclusion ‐ tubes, anatomic defects, immunodeficiency, allergy to penicillin Numbers eligible not provided. 194 enrolled. Allocation; 70 amoxycillin (OD), 58 amoxycillin (BD), 66 placebo. Analysis; 55, 44 and 59 Baseline characteristics sex, age, smoke, family OM history, child care, snoring, LRI, feeding position, mouth breathing, RAD, prematurity, age (mean +/‐ SD of OM onset, all comparable by group
Interventions	Up to 90 days or until 2 x AOM (failure). An average of 6.5 weeks (7243 days for 158 children) on study medication (excludes days treated for AOM) Amoxicillin 20 mg/kg/d OD or BD Placebo
Outcomes	Measured in less than 90% Format of reporting ‐ incidence density of AOM Diagnostic criteria ‐ decreased mobility plus red or yellow TM Subgroups ‐ age, season Microbiology ‐ none
Notes	Q: AAAABA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) B ‐ outcomes measured in 80% to 90%
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Schuller 1983a.

Methods	Not blinded Randomly divided on basis of age, sex and allergy. Method of randomisation and allocation concealment not stated Review 3‐monthly plus illness Group similarity not described
Participants	Danville, PA, USA Presented in Oct 1982, published July 1983 Eligibility ‐ children 2 to 6 y of age with asthma within 6 mo and at least 4 AOMs in 12 mo. Exclusion criteria not stated but no child had enlarged adenoids or tonsils, none had tubes Numbers eligible not provided. 80 enrolled (2 moved, 6 non‐compliant), 72 completed 2 year study. Allocation of 72: 12 control, 12 symptomatic antihistamine, 12 prophylactic sulfisoxazole 2 years, 6 prophylactic sulfisoxazole 1 year, then pneumococcal vaccine, 12 pneumococcal vaccine, 6 pneumococcal vaccine then sulfisoxazole in 2nd year, 12 pneumococcal vaccine plus sulfisoxazole 2 years. Baseline characteristics mean age comparable by group
Interventions	2 years (second year combined therapies used) Sulfisoxazole 500 mg BD (Group 3) No treatment (1) or antihistamines when nose blocked (2) or Pneumovax (4) or Pneumovax and sulfisoxazole (5)
Outcomes	Measured in 72/80 (90%) over all intervention groups Format of reporting ‐ frequency of AOM episodes per y Diagnostic criteria ‐ fever more than 38 C, bulging immobile TM Subgroups ‐ none Microbiology ‐ none
Notes	Q: CCDCBA Included data from first year only (combined therapies used in year 2)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	C ‐ randomisation claimed but not described, and outcome assessor not blinded
Allocation concealment (selection bias)	High risk	C ‐ Inadequate
Blinding (performance bias and detection bias) All outcomes	High risk	D ‐ no blinding of outcome assessor
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	C ‐ the progress of all randomized children in each group obviously not described B ‐ outcomes measured in 80% to 90%
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Schwartz 1982a.

Methods	Double‐blind, cross‐over Method of randomisation unclear (pharmacy prepared, assigned at random by blinded nurse) Review 2 ‐ 4 weekly after randomisation, then monthly for 3 months (combined cross‐over periods) Group similarity not reported
Participants	Washington and Virginia, USA Enrolled between Dec 1979 and January 1980. Published 1982 Eligibility ‐ children 4 to 72 mo of age. Mean 31 mo with more than 3 AOMs in preceding 12 mo and recent AOM. Exclusion ‐ none described Numbers eligible not provided. 43 enrolled, 43 allocated; 23 SMX, 20 placebo (pre‐cross‐over). 33 completed; 17 SMX, 16 placebo. 10 drop‐outs; 6 no follow up, 3 refused SMX, 1 not accounted Baseline characteristics not reported by group
Interventions	2 months Sulfisoxazole 25 mg/kg OD Placebo
Outcomes	Measured in 33/43 (77%) Format of reporting ‐ number of episodes in 2‐month period OM = 1 bulge of opaque TM. 2. Dec mobility (pneumatic otosc + type B tymp. 3. Yellow or fiery red TM Subgroups ‐ tympanometry at baseline Microbiology ‐ none
Notes	Q: AAAADA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Sih 1993a.

Methods	Randomised, placebo‐controlled Methods not stated, 'blinding' not mentioned Review monthly plus during illness Group similarity reported
Participants	Sao Paulo, Brazil Enrolled between March 1989 and March 1990. Published 1993 Eligibility ‐ children aged 9 to 108 mo with rAOM (3 AOMs in preceding 12 mo). Enrolled with AOM, treated then randomized if type A or C tympanogram and no symptoms of AOM. Exclusion ‐ type B tympanogram on day 11 of AOM Rx Number enrolled and randomized 80, 60 completed (20 lost to follow up all enrolled through public hospital), allocation: 20 amoxicillin, 20TMP‐SMX, 20 placebo Baseline characteristics: age, site, recent OM history, atopy comparable by group
Interventions	3 months TMP‐SMX (12 mg/kg/d) AMX (20 mg/kg/d) OD. Placebo
Outcomes	Measured in 75% Method of reporting ‐ proportion of children with any AOM. Number of episodes for each child. Number of episodes by month of therapy for placebo and combined antibiotic groups. Diagnostic criteria ‐ AOM = acute symptom (pain, irritability, fever, respiratory symptoms) and pneumo‐otoscopic findings (red, bulging or reduce mobility) Subgroups ‐ none Microbiology ‐ none
Notes	Q: BBCCDA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	B ‐ randomisation stated but method not described, or suspect method
Allocation concealment (selection bias)	Unclear risk	B ‐ unclear
Blinding (performance bias and detection bias) All outcomes	Unclear risk	C ‐ unclear
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	C ‐ the progress of all randomized children in each group obviously not described D ‐ outcomes measured in less than 80%
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Teele 2000a.

Methods	Double‐blind Randomised computer‐generated. Stratified by age and sibling history of rAOM Review at 0, 2, 6, 10, 14, 18, 22 and 26 weeks, plus presentation Similarity not reported by group
Participants	Boston, MA, USA Dates not provided. Published in 2001 Birth cohort Eligibility ‐ single AOM prior to 6 months of age or 2 discrete episodes before 12 months. Exclusion ‐ none described Numbers eligible not provided. 117 enrolled and followed 1 y. Allocation: 40 amoxicillin, 36 sulfisoxazole, 41 placebo Baseline characteristics provide, not by group
Interventions	6 months Sulfisoxazole (50 mg/kg/d) Amoxicillin (20 mg/kg/d). Frequency not stated. Assumed once daily Placebo
Outcomes	Measured in 100% Format of reporting ‐ proportion of children with none, more than or equal to 1, more than or equal to 2, more than or equal to 3 AOM episodes during 6 months therapy and during therapy and an additional 6 months of post‐therapy follow up Diagnostic criteria ‐ AOM = effusion plus more than 1 sign of acute illness (fever, irritability, lethargy, otalgia, anorexia, vomit) New episode 21 days after previous AOM Subgroups ‐ none Microbiology ‐ throat cultures at entry and monthly. Spn, Hi and M. cat tested for antimicrobial susceptibility. Results not reported
Notes	Q: AAAAAA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	A ‐ randomisation method correct, or randomisation stated and group similarity documented
Allocation concealment (selection bias)	Low risk	A ‐ concealment method correct, or concealment stated and group similarity documented
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Low risk	A ‐ the progress of all randomized children in each group described and outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures) A ‐ outcomes measured in more than 90% (where withdrawals due to complications and side effects were categorised as treatment failures)
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

Varsano 1985a.

Methods	Double‐blind, cross‐over Randomised (method not described) Review 5 weekly and when symptoms of URT or OM suspected Group similarity not reported
Participants	Israel 1 October to 31 January. Year of study not provided. Presented in 1984, published 1985 Children aged 6 mo to 5 y with rAOM. Free of AOM at randomisation. Exclusion ‐ not eligible for TTs Number entered 40, 32 completed (8 not co‐operative or non‐compliant). 17 placebo, 15 sulfisoxazole
Interventions	10 weeks Sulfisoxazole (250 mg tablet if less than 2 y; 500 mg if 2 to 5 y BD) Placebo
Outcomes	Measured in 32/40 (80%) Format of reporting ‐ proportion of children with AOM, total number of episodes, and rate of AOM per patient per 10‐week period Diagnostic criteria ‐ AOM = erythema or white‐yellow opacification plus bulging and poor mobility, or acute perforation with pus in canal Subgroups ‐ age (less than 2 versus 2 to 5 yrs), OME at entry Microbiology ‐ none
Notes	Q: BBACBA
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	B ‐ randomisation stated but method not described, or suspect method
Allocation concealment (selection bias)	Unclear risk	B ‐ Unclear
Blinding (performance bias and detection bias) All outcomes	Low risk	A ‐ participant, care provider and outcome assessor blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	C ‐ the progress of all randomized children in each group obviously not described B ‐ outcomes measured in 80% to 90%
Selective reporting (reporting bias)	Unclear risk	Study conducted prior to clinical trial registration
Other bias	Unclear risk	No other potential source of bias identified

METHODS: Blinding. Randomisation method. Stratified. Review schedule. Group similarity reported.
 PARTICIPANTS: Country, city. Dates of study. Eligibility (age, inclusion). Numbers eligible, enrolled, allocated by group. Baseline characteristics by group.
 INTERVENTIONS: Duration of therapy, Active and Control group dose and schedule.
 OUTCOMES: Proportion with at least one follow‐up assessment. Format of reporting. Diagnostic criteria. Subgroups. Microbiology (carriage and resistance outcomes)
 NOTES: Q: Quality scores for 1) Randomisation (A: Randomisation & Blinding clear, or R&B stated AND group similarity reported); 2) Allocation concealment (A: adequate); 3) Blinding (A: double‐blind or outcome assessor blinded); 4) Participant reporting by allocated group (A: progress of all randomized children reported); 5) Follow up (A: > 90%); 6) Standardised outcome assessments (A: described and used for all participants)

AOM = acute otitis media
 BD = twice‐daily
 CI = confidence interval
 d = day
 ENT = ear, nose and throat
 Hi = Haemophilus influenzae 
 LRI = lower respiratory infection
 M. cat = Moraxella catarrhalis 
 MEE = middle ear effusion
 MEF = middle ear fluid
 mo = months
 NP = nasopharynx
 OD = daily
 OM = otitis media
 OME = otitis media with effusion
 OP = oropharynx
 rAOM = recurrent acute otitis media
 Rx = treatment
 SD = standard deviation
 SOM = suppurative otitis media
 Spn = Streptococcus pneumoniae 
 TM = tympanic membrane
 TTs = tympanostomy tubes
 URTI = upper respiratory tract infection
 y = year
 TMP = trimethoprim
 SMX = sulfamethoxasole
 SX = sulfasoxasole
 RAD = reactive airway disease
 (AOM)woP = acute otitis media without perforation
 (AOM)wiP = acute otitis media with perforation

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Berman 1992	Cases of AOM not reported. No placebo (treatment of URTI symptoms)
De Diego 2001	No placebo group
Donaldson 1990	Children with AOM withdrawn
Harrison 1999	No placebo group. Intervention not antibiotics or therapy thought to be effective (homeopathic)
Koivunen 2004	All children received tympanostomy tubes at time of randomisation

AOM = acute otitis media
 URTI = upper respiratory tract infection

Contributions of authors

Amanda Leach (AL) was responsible for designing the study, searching the literature, identifying the relevant studies, extracting the data, analysing the results and writing the review. Peter Morris (PM) extracted data independently and assisted with the design, analysis and writing of the review.

Sources of support

Internal sources

• No sources of support supplied

External sources

• National Health and Medical Research Council, Australia.

  Both authors received individual and project grant research support

Declarations of interest

Both review authors were investigators on a randomized controlled trial of antibiotics to prevent otitis media in remote Australian Aboriginal children. This study was funded by the Australian National Health and Medical Research Council.

Edited (no change to conclusions)