Treatment of Acute Otitis Media With a Shortened Course of Antibiotics:  A Meta-analysis FREE

ACUTE OTITIS MEDIA (AOM) is one of the most common reasons for a child to be brought to a source of health care. Before age 7 years, 65% to 95% of children will have experienced 1 or more episodes of AOM.¹ Acute otitis media is the most common indication for antibiotic prescribing in the United States; over 90% of children with AOM will be treated with a 10-day course of antibiotics.^2- 3 It has been estimated that more than $3.5 billion is spent annually on managing AOM in the United States.⁴ The decline in suppurative complications of otitis media (OM) in North America and Europe during the 1940s and 1950s has been attributed to antibiotic therapy.⁵ More recent evidence suggests that long-term outcomes are similar in antibiotic-treated and untreated children with AOM living in developed countries.^2,6- 8 Practitioners in North America, however, may be hesitant to discontinue prescribing antibiotics altogether. In fact, 2 recent meta-analyses of children with AOM reported that compared with placebo, antibiotic treatment did hasten the resolution of acute symptoms.^9- 10 In some of the trials, symptoms abated by 2 to 3 days of antibiotic treatment, suggesting that a course of antibiotics of less than 10 days of therapy may be an effective alternative.^6- 8

The emergence of drug resistance in community-acquired infections, such as that reported in children receiving prophylaxis for recurrent OM, has increased concern over the indiscriminate prescribing of antibiotics.^11- 12 This concern is evident in recent expert opinion which recommends that the duration of antibiotic treatment in uncomplicated AOM be decreased to 5 days; this recommendation has been further qualified to exclude young children.^13- 14 The quality of scientific evidence to support a policy for shorter courses of antibiotic treatment has been assessed,¹⁵ but a systematic, quantitative evaluation of this evidence is lacking. Therefore, a meta-analysis of the available evidence on the antibiotic treatment of AOM was undertaken, with the primary objective of determining whether treatment with a short course of antibiotics was comparable to treatment with a longer course.

Randomized controlled trials comparing different durations of antibiotic treatment in children with a diagnosis of AOM were identified from MEDLINE (January 1966 to July 1997), EMBASE (January 1974 to July 1997), Current Contents (January to July 1997), and Science Citation Index searches. There were no language restrictions.¹⁶ In MEDLINE, search terms employed were otitis media in medical subject headings, modified by acute in the title or abstract. These terms were combined with the terms of randomized controlled trials, random allocation, double-blind method, single-blind method, or with the term randomized controlled trial in publication type. This strategy, which was approximated for the EMBASE search, was an adaptation of one deemed optimally sensitive for identifying randomized controlled trials in MEDLINE.¹⁷ Reference lists of relevant publications were reviewed to identify further trials. Where abstracts were referenced, attempts were made to obtain the published version.

Clinical trials of the antibiotic treatment of AOM were independently assessed for inclusion by the 7 authors, according to the following criteria: (1) patients between the ages of 4 weeks and 18 years, (2) clinical diagnosis of AOM and no antimicrobial therapy at time of diagnosis, (3) random assignment to antibiotic treatment of fewer than 7 days vs at least 7 days, and (4) an assessment of clinical resolution of AOM.¹⁸ Trials were excluded if there was planned surgical cointervention, but not if myringotomy subsequent to treatment failure or tympanocentesis was performed. Concurrent use of a second antibiotic was not a criterion for trial exclusion, but children given additional antibiotics were counted as treatment failures.

The internal validity of included trials was assessed by the same authors using the Jadad scale.¹⁹ The scale assigned scores from 0 to 5 (best-quality trial), based on the following criteria: (1) study participants were randomly allocated to treatment using an appropriate method such as a random numbers table, (2) the intervention was double-blinded using an identical looking and tasting placebo, and (3) an accounting and description of study withdrawals were done.²⁰ Concealment of treatment allocation was also evaluated for adequacy.^21- 22 Rater assessment was blinded in the majority of trials by removing all identifying features of the published trial, such as author names, affiliations, and sources of financial support and by random ordering of trials.²⁰ Interrater agreement on trial inclusion and the quality score was assessed by the κ statistic using PC Agree software (Stephen Walter, McMaster University, Hamilton, Ontario). Agreement was categorized using Landis and Koch guidelines.²³

The primary outcome of interest was treatment failure, which included lack of clinical resolution or relapse or recurrence of AOM, during a 31-day period following the initiation of therapy. Clinical resolution meant that the presenting signs or symptoms of AOM had improved or resolved. A secondary outcome was the cumulative number of treatment failures, relapses, and recurrences reported from time of diagnosis until a final evaluation point between 1 to 3 months. Middle ear effusion was not classified as a treatment failure because of its documented persistence during the course of the disease, regardless of treatment.⁹

Two authors independently abstracted primary and secondary outcome data from each blinded trial, using a standardized data extraction form. Where possible, outcome data were extracted for children both older and younger than 2 years, and for children with perforated and nonperforated tympanic membranes.^8,24- 26 Descriptive data were also collected on the treatment site, patient baseline characteristics, cointerventions, and inclusion, exclusion, and outcome criteria. The number of withdrawals in each treatment arm, including those lost to follow-up, those with adverse drug effects, and those not complying with the study protocol, was recorded.

Meta-analyses were performed on trials grouped by the pharmacokinetic behavior of the antibiotic used in the short-treatment arm, as follows: (1) short-acting oral antibiotics, eg, penicillin V potassium, amoxicillin, cefaclor, cefuroxime; (2) oral azithromycin; or (3) intramuscular ceftriaxone. Additional meta-analyses were conducted in the short-acting antibiotic group for treatment duration of 48 hours or less and greater than 48 hours.

The meta-analyses were carried out using the Cochrane Collaboration's Revman 3.0 program (Cochrane Collaboration, Oxford, England. The risk of treatment failure with a short course of antibiotics, in comparison to a longer course, was expressed as an odds ratio (OR). An OR greater than 1 indicated a greater number of failures with the short course of antibiotics and superiority of the long course of antibiotics. The ORs were calculated for individual trial outcomes and, using the Peto fixed-effects model, a summary OR was determined for trials pooled by antibiotic type. Statistical heterogeneity among trials was assessed,^27- 28 and trials with outlying ORs were further investigated.²⁹ In addition, summary ORs were calculated using the DerSimonian and Laird random-effects model.³⁰ To provide additional clinical meaning, results were also expressed as a summary risk difference in the failure rate and the number needed to treat to experience an additional failure in the short treatment arm.²⁸

Sensitivity analyses were conducted to assess the robustness of the meta-analysis by comparing summary ORs among groups redefined by (1) excluding trials of a lower methodological quality (quality ≤2), (2) excluding trials that included children with recurrent or chronic OM, and (3) excluding trials of comparisons between different antibiotics.²⁷ A sensitivity analysis of outcome criteria was also conducted by redefining clinical resolution to include cure, but not improved symptoms. A funnel graph of the sample size vs the OR was plotted to determine whether publication bias existed.²⁷

The MEDLINE search yielded 209 citations of randomized controlled trials in AOM, of which 36 were assessments of the antibiotic treatment duration of AOM in children. An additional 4 trials were found among 236 clinical trials of AOM identified by the EMBASE search. Four trials not identified by MEDLINE or EMBASE were retrieved from the Current Contents index search and reference listings. Three clinical trials of antibiotics no longer used today were excluded,^31- 33 leaving 41 citations for further assessment according to inclusion criteria. Substantial interrater agreement for trial inclusion into the meta-analysis for the remaining trials (κ score, 0.89; 95% CI, 0.86-0.92) was attained. Twelve of these trials were excluded from the meta-analysis for the following reasons: duration of antibiotic therapy was equivalent in both treatment arms^34- 39; duration of antibiotic therapy in both treatment arms was either fewer than 7 days^40- 41 or7 days or greater⁴²; there was no comparison of different durations of antibiotic therapy⁴³; trial participants were predominantly adults⁴⁴; or insufficient data were reported in abstract.⁴⁵ This left 32 trials, contained in 29 published reports, for inclusion in the meta-analysis. The 32 trials were distributed in antibiotic groups as follows: 17 were short-acting antibiotics,^26,46- 58 4 were ceftriaxone sodium,^59- 62 and 11 were azithromycin.^63- 73 Two of the trials were in a language other than English.^53,57

The mean quality score for all of the trials was 2.66 (SD, 0.97) of a maximum score of 5 (Table 1). Ceftriaxone trials scored highest (mean, 3.50; SD, 0.58), followed by short-acting antibiotic trials (mean, 2.71; SD, 1.07) and azithromycin trials (mean, 2.27; SD, 0.79). Substantial interrater agreement was reached (weighted κ, 0.82; 95% CI, 0.79-0.85). Only 9 trials in 8 publications reported adequate concealment of treatment allocation.^{48,53- 54,57,60,69- 70,73} Cointervention was permitted in 38% of trials, usually with analgesics and decongestants.^{46- 48,50- 51,53,57,59,64- 66} In 2 trials, myringotomy had been conducted in early treatment failures.^48,59 Concurrent use of a second antibiotic was seen in 1 trial.⁶¹ A check for compliance with antibiotic administration was carried out in half of the trials^{26,46,48,52- 54,57- 59,62,67,70,72- 73}; substantial noncompliance was noted in 3 trials in 2 publications, contributing to the withdrawal of patients.^54,67 One trial had an exceptionally high withdrawal rate due to protocol violation, which could not be explained by noncompliance, adverse effects, or loss to follow-up.⁵⁸ With the exception of 3 trials in 2 publications,^54,72 withdrawal rates were similar in both treatment arms.

Two short-acting antibiotic trials were excluded because outcomes were not reported as treatment failures.^50- 51 Trials in which short-acting antibiotics were administered for at least 48 hours in the short arm were conducted in the United States or Europe. Although the majority of trials excluded comorbid conditions, only 31% excluded children with chronic or recurrent OM.^{46,53,55- 56} In all trials, diagnosis of AOM was made according to the presence of signs and symptoms. Tympanometry⁵⁸ and tympanocentesis^26,49,54 were not commonly performed. Among trials which reported data on the distribution of age (representing 57% of all children in trials), 45% of children were younger than 2 years.^{26,47,49,53,57} In 3 trials,^54- 56 representing 43% of children, the mean age was 3.7 years. Fifty-two percent of children were male in trials that reported gender information.^{26,47,49,53- 58} Ten trials in 8 publications defined a satisfactory outcome as the resolution or improvement of signs and symptoms, including middle ear effusion.^{26,46,49,54- 58} Three trials in 2 publications regarded persistent effusion as a treatment failure.^47,53 One thousand children were evaluated at 8 to 19 days after treatment was begun,^53,58 and 462 children were evaluated at both 8 to 19 days and 20 to 30 days after entry into trial.^26,57 In the remainder of children (n=1653), primary outcomes were assessed only at the 20- to 30-day time period. Secondary outcomes were assessed at 30 to 40 days and 3 months following trial entry.^26,46,53,58

The ceftriaxone trials were conducted in the United States or Israel. A diagnosis of AOM was made using clinical and tympanometric criteria. Comorbid conditions were a criterion of exclusion, and half of the trials excluded children with chronic or recurrent OM as well.^60- 61 The mean age of children in each treatment arm was approximately 2 years and 51.2% of children were male. Treatment outcome criteria included the resolution of signs and symptoms, with or without persistent effusion, in 3 trials; in the fourth trial a satisfactory outcome was defined as only the resolution of symptoms.⁶⁰ Primary outcomes were evaluated at 1 month following the initial assessment, and secondary outcome assessment ranged from 1 to 3 months.

The azithromycin trials were predominantly multicenter trials, conducted in the United States, Europe, or developing countries. Comorbid conditions were frequently excluded; 4 trials excluded children with chronic or recurrent OM.^65,68,70,72 Acute otitis media was diagnosed by the presence of signs and symptoms. Tympanometry^{63,68- 70,72- 73} and tympanocentesis^{63- 65,67- 68,70,72- 73} were frequently carried out. The mean age of each treatment arm was approximately 4 years and children were predominantly male (56%). Treatment outcomes were defined in terms of the resolution and improvement of signs and symptoms. Five trials specified that persistent middle ear effusion was an acceptable outcome.^{63,69- 70,72- 73} All children were evaluated at 10 to 14 days following entry into the trial. Six trials (n=1254) also included a 1-month evaluation.^{63,68- 70,72- 73}

Secondary and not primary outcomes were evaluated in 1 of the trials,⁴⁶ leaving 14 trials in the meta-analysis of primary outcomes.

The summary OR for failures at 1 month or less in trials that compared 48 hours or less of antibiotic treatment with at least 7 days was 2.99 (95% CI, 1.04-8.54). Only 2 trials, evaluating a total of 118 children, were included in this meta-analysis.^48,52 One of the trials had been terminated prematurely.⁵²

Antibiotics were administered for 5 days in the short-treatment arm and 8 to 10 days in the long-treatment arm in 12 trials that reported outcomes at 1 month or less. The summary OR for treatment failure in 1549 children treated for 5 days, in comparison to 1569 children treated for 8 to 10 days, was 1.38 (95% CI, 1.15-1.66). One third of these children were evaluated close to the end of the longer course of antibiotics, while the other two thirds were evaluated around 30 days. Primary outcomes were further subdivided by timing of evaluation, subsequent to a concern that evaluations close to the end of treatment would not give a true picture of relapse and reinfection. The odds of treatment failure at an earlier evaluation point (8 to 19 days) in the 5-day treatment arm was increased to 1.52 (95% CI, 1.17-1.98) (Figure 1). The weighted mean failure rate was 19.0% (SE, 7.6%) with 5 days of treatment and 13.7% (SE, 6.4%) in the long-treatment arm. The weighted summary risk difference was 7.8% (95% CI, 4%-11.6%).

However, among 1031 children treated with an antibiotic for 5 days and 1084 children treated for 8 to 10 days, primary outcomes at 20 to 30 days were not significantly different between the 2 groups (OR, 1.22; 95% CI, 0.98-1.54). The weighted mean failure rate at 20 to 30 days was 15.7% (SE, 13.3%) with 5 days of treatment and 12.5% (SE, 12.4%) in the long-treatment arm. The weighted summary risk difference was 2.3% (95% CI,−0.2% to 4.9%). There were no differences in secondary outcomes (failures, relapses, and recurrences during a 3-month period) between 539 children in the short-treatment and 515 children in the long-treatment arms (OR, 1.16; 95% CI, 0.90-1.50).^26,46,53,58 Similar summary ORs were obtained in children evaluated at 30 to 40 days (OR, 1.16; 95% CI, 0.87-1.55) or at 90 days (OR, 1.16; 95% CI, 0.65-2.06) (Figure 1).

The summary OR for antibiotic failure at 30 days or less in children younger than 2 years (n=118) and those who were at least 2 years old (n=235) was 0.71 (95% CI, 0.3-1.64) and 1.01 (95% CI, 0.53-1.94), respectively.^47,49 Significant differences in outcome between 5-day and 10-day treatment among children younger than 2 years old, but not in older children, were reported in an additional trial, but no extractable subgroup data were presented.⁵⁸ In the single trial that reported outcome data for perforated and nonperforated eardrums,²⁶ the OR for treatment failure in children with perforated eardrums (n=27) was 3.62 (95% CI, 0.81-16.06) and 1.06 (95% CI, 0.40-2.75) in children with nonperforated eardrums (n=101).

Meta-analyses were conducted on 3 trials for primary (n=671)^59- 60,62 and 2 trials for secondary outcomes (n=312).^60- 61 The outcomes of treatment with ceftriaxone for 1 month or less and for 3 months or less were not significantly different from those of treatment with a longer course of oral antibiotics, with summary ORs of 1.25 (95% CI, 0.90-1.72) and of 0.91 (95% CI, 0.57-1.47), respectively.

The summary OR for primary outcomes following 3 or 5 days of treatment with azithromycin (n=1347) in comparison to 10 days of treatment with another antibiotic (n=1246) was 1.09 (95% CI, 0.86-1.38). The summary ORs did not change when children were treated for 3 days only (OR, 1.17; 95% CI, 0.71-1.92), when children were evaluated at 10 to 14 days (OR, 1.11; 95% CI, 0.82-1.51), or when children were evaluated at 30 days (OR, 1.02; 95% CI, 0.78-1.34). The odds of treatment failure with azithromycin was 1.92 (95% CI, 0.73-5.04) in children younger than 2 years (n=138; 17.4%) and 1.34 (95% CI, 0.61-2.94) in older children (n=656).^66- 67

Sensitivity analyses were conducted for all primary outcomes up until 30 days following trial entry, and separately for outcomes at a 20- to 30-day evaluation in the 5-day short-acting antibiotic group. Treatment failures at 30 days or less were not significantly more likely in the short- than long-treatment arm among short-acting antibiotic trials grouped as (1) high-quality trials, (2) trials with adequate treatment allocation, and (3) trials that excluded children with recurrent or chronic OM (Table 2). The majority of summary ORs for primary outcomes at 20 to 30 days did not change during any of the sensitivity analyses.

The symmetrical, inverted funnel-shaped plot of the ORs vs sample size, as shown by the wide scattering of ORs from small studies narrowing to a peak among larger trials, suggested no evidence of a publication bias in the meta-analysis trials (Figure 2). Using a random-effects model, which accounts for trial heterogeneity, only 1 of the summary ORs was changed. The random-effects model widened the 95% CI for the summary OR in the short-acting antibiotic group at 8 to 19 days, so that a 5-day course of antibiotics was no longer statistically significantly associated with poorer treatment outcomes than a longer course.

The summary OR for gastrointestinal side effects in the short-acting antibiotics trials was 0.54 (95% CI, 0.43-0.66). An OR less than 1 was observed only in trials in which the short course of antibiotics was compared with a 10-day course of the standard formulation of amoxicillin-clavulanate.^54,57- 58 Once these were excluded, there was no difference in the likelihood of gastrointestinal side effects following a short or long course of antibiotics (OR, 1.13; 95% CI, 0.81-1.57). Children treated with azithromycin were also less likely to experience gastrointestinal side effects than children treated with a long course of antibiotics (OR, 0.26; 95% CI, 0.19-0.37), which was most often amoxicillin-clavulanate.

Our meta-analysis results indicate that at an early evaluation point (8 to 19 days) a reduction in treatment from 10 to 5 days of short-acting antibiotics may slightly increase the risk of a child experiencing signs and symptoms or relapse or reinfection (OR, 1.52; 95% CI, 1.17-1.98). However, by 30 days following initiation of therapy, a longer course of short-acting antibiotics was comparable to a 5-day course in terms of these outcomes (OR, 1.22; 95% CI, 0.98-1.54). We do note that the latter finding is based on a summary OR for which the lower limit of the CI is close to 1. At the 8- to 19-day evaluation, the difference in risk of failure between treatment with 5 days vs 10 days of a short-acting antibiotic was 7.8%. The clinical significance of this risk difference, expressed as the number needed to treat to experience an additional failure in the short-treatment arm (or 1 divided by the risk difference), indicates that 13 children would require treatment with a long course of antibiotics to prevent 1 excess failure following treatment with a shorter course. However, a recent study suggests that a 6-week evaluation period,⁷⁴ rather than the traditional 2-week evaluation,⁷⁵ decreases the necessity of retreating AOM without increasing complications. In our study, the risk difference at 1 month following treatment dropped to 2.3%, which translates into 44 children requiring treatment with a longer course of antibiotics to prevent 1 failure following shorter treatment. This is not likely a clinically important difference because the 44th child with persistent symptoms will generally return to a source of health care. Moreover, a shortened course of antibiotics may protect the child from developing resistant microorganisms,^{11- 12,76- 79} which in our opinion outweighs the minimal risk of increased treatment failure, if such a risk exists at all.

Differences in outcome between the evaluations at 8 to 19 days and 20 to 30 days likely reflect bias in the timing of evaluation. Children treated with a long course of antibiotics had fewer days to experience an outcome than those treated with 5 days when the time to evaluation was 8 to 19 days, as opposed to 30 days.¹⁵ Comparisons of treatment outcome over a 3-month period, although based on a smaller number of children, revealed no significant differences between the 2 treatment regimens. The long-term comparability between a short and long course of antibiotics is biologically plausible, on the basis of (1) spontaneous resolution of untreated AOM,^9- 10 (2) early eradication of pathogens after 3 to 5 days of treatment,⁸⁰ (3) poorer penetration of the antibiotic into the ear with continued administration as inflammation decreases,⁸¹ and (4) treatment of children without AOM because of diagnostic uncertainty.²

Potential weaknesses of meta-analysis techniques are that they incorporate existing biases and introduce new biases, some of which have predicted discordance of results between meta-analyses and single, large randomized controlled trials.^82- 83 To minimize bias during study selection, we used predetermined inclusion criteria, and most trials were assessed in a blinded fashion, although recent evidence suggests that blinded evaluations are not necessary.⁸⁴ Publication bias was assessed in our funnel plot of sample size vs the ORs and none was evident.²⁷ Using the validated Jadad scale to assess for biases within individual trials,^19- 20 the majority of short-acting antibiotic trials scored less than 3 because they were not double-blinded. Lack of double-blinding in trials has yielded larger estimates of treatment effects.²¹ In AOM trials, the absence of blinding may bias outcomes in favor of longer treatment with antibiotics because parents and physicians may be more vigilant in detecting failures in a child receiving a shorter course.⁴⁶ An increased risk of treatment failure with a 5-day course of antibiotics was indeed documented among poor-quality trials in our sensitivity analyses. Less favorable outcomes in the shorter-treatment arm were also limited to trials with inadequate concealment of treatment allocation, consistent with evidence that poor concealment of treatment allocation results in biased treatment effects.^21- 22 The issue of trial heterogeneity was addressed in our grouping of antibiotics according to pharmacokinetic profile.⁸⁵ In addition, the fixed-effects model, which assumes trial homogeneity, and the random-effects model, which incorporates between-study variation,⁸² produced similar ORs in most of the antibiotic subgroups. Specific inclusion criteria for the diagnosis of AOM were not used, but diagnostic criteria were comparable across trials. However, outcome criteria did vary among trials and changes to the definition of a successful outcome altered the 20- to 30-day summary OR.

Most of the short-acting antibiotics in this study are commonly prescribed in primary practice and our sensitivity analyses indicate that comparisons between different short-acting antibiotics did not alter treatment outcomes. Similar outcomes were observed between a 3- or 5-day course of azithromycin and a 10-day course of other antibiotics. Comparability was also demonstrated between ceftriaxone and a longer course of antibiotics, although the sample size of the consolidated trials was small. It is difficult to support the use of azithromycin or ceftriaxone based on their cost and concern over indiscriminate use of broad-spectrum antibiotics.⁸⁶ Ceftriaxone's intramuscular mode of administration may further limit its role.⁸⁷ While antibiotic use in children with AOM may increase the incidence of diarrhea and vomiting,¹⁰ we did not find that a shortened course of antibiotics decreased the likelihood of gastrointestinal effects, except when compared with a longer course of amoxicillin-clavulanate.

Our findings can be safely applied to children who present to ambulatory care settings with uncomplicated AOM, with some noteworthy exceptions. These findings do not apply to children with underlying disease and potentially children with recurrent or chronic OM. Treatment failure at less than 1 month was not more likely with a 5-day course of antibiotics once children with chronic or recurrent OM were excluded. Subgroup sample sizes were too small to provide a reliable estimate of the risk of treatment failure with a shortened course of antibiotics in children younger than 2 years or children with perforated eardrums. In addition, while our meta-analysis included data on children from various populations, it remains unproven whether our findings can be extrapolated to other high-risk children.^5,88

The meta-analysis results support the use of 5 days of a short-acting antibiotic in uncomplicated AOM in the event that clinicians and parents decide to use antibiotics. Treatment with a shortened course of antibiotics has the potential to greatly reduce antibiotic use in regions where 10 days of treatment is considered the standard, with anticipated cost savings, improved compliance, and decreased antibiotic resistance.¹⁵