Chapter  15:  Management of Acute Otitis Media: Evidence Report/Technology Assessment Number 15

Objectives

Our objective with this report was to analyze the evidence on the initial management of uncomplicated acute otitis media (AOM) in children. We assessed three questions: What is the natural history of AOM without antibiotic treatment? Are antibiotics effective in preventing clinical failure? and What is the relative effectiveness of specific antibiotic regimens such as: (1) amoxicillin or trimethoprim-sulfamethoxazole vs. other antibiotics, (2) oral fluoroquinolones, (3) >60 mg/kg/day amoxicillin or amoxicillin-clavulanate vs. 40 mg/kg/day, (4) high-dose amoxicillin therapy twice a day vs. three times a day, and (5) short-term vs. long-term antibiotic therapy?

Search Strategy

The technical experts and project staff developed a literature search strategy that used MEDLINE, the Cochrane Library, HealthSTAR, International Pharmaceutical Abstracts, Cumulative Index to Nursing & Allied Health Literature, BIOSIS, and EMBASE. Headings included "om" (otitis media), "mastoiditis," "om w/effusion," "om, suppurative" with the subheading "drug therapy," and "anti-infective agents," including "antibiotics" and "other drugs." Text words included otitis media, antibiotic, antimicrobial, antibacterial, specific antibiotic names and natural history, natural course, untreated, spontaneous, and self-limited for the natural history search. Additional articles were identified by review of reference lists in proceedings, published articles, reports, and guidelines.

Selection Criteria

The selection criteria included human studies addressing AOM in children age 4 weeks to 18 years. Excluded were studies on patients with immunodeficiencies or craniofacial deficiencies, including cleft palate. Randomized clinical trials and cohort studies were included for the question on natural history. Randomized controlled trials were used to address the questions on antibiotic effectiveness.

Data Collection and Analysis

Two physician reviewers independently evaluated articles and abstracted data. They excluded articles that did not address any of the key questions or that had unabstractable data. They assessed the quality of the studies. Project staff developed evidence tables using the abstracted data. Meta-analyses were performed as appropriate. Twenty-four experts and consumers critiqued the draft of this evidence report.

Main Results

Clinical failure was defined generally as the failure to resolve or improve clinical signs and symptoms at specific time intervals. In children with AOM not initially treated with antibiotics, the reported rate of clinical failure at 24-48 hours was 7.7 percent in one study and at 24-72 hours was 26 percent in another study. The pooled estimated failure rate at 1-7 days was 18.9 percent and was 22.2 percent at 4-7 days. Compared with this failure rate, children treated with ampicillin or amoxicillin showed a reduction in clinical failure at 2-7 days of 12.3 percent. This result generally was robust to sensitivity analysis. Eight children with AOM would need to be treated with ampicillin or amoxicillin rather than no antibiotic treatment to avoid a case failure. The pooled comparisons of specific antibiotic regimens showed no difference in clinical failure rates. Adverse effects, primarily gastrointestinal, were more common in children on cefixime than those on ampicillin or amoxicillin and more common in children on amoxicillin-clavulanate (original formulation) than those on azithromycin.

Conclusions

There is a wide range of reported clinical failure rates in the first week in children with AOM who are not treated with antibiotics. The pooled estimates indicate that a majority of these children have clinical resolution within the first week after presentation. These studies, which generally are of small sample size, report few episodes of suppurative complications, regardless of antibiotic treatment. In general, the children in the natural history studies had close followup. Antibiotic treatment with ampicillin or amoxicillin produces a statistically significant improvement in failure rates compared with no antibiotic treatment of AOM. We found no difference in clinical failure rates in the antibiotic regimens assessed. Since we were unable to do subgroup analysis in the meta-analytic comparisons, we cannot generalize those findings to children in specific age groups, such as younger than or older than 2 years of age, or by otitis-prone status. Several studies suggest greater caution be taken with children younger than 2 years old.

Future Research

Randomized controlled studies still are needed to address definitively the initial key questions. Future research should focus on: (1) establishing uniform definitions of AOM and relevant outcomes, (2) establishing uniform diagnostic criteria, (3) strengthening internal and external validity, (4) addressing relevant influencing factors such as age and otitis-prone state, and (5) measuring long-term as well as short-term outcomes. Studies also should be of sufficient power to establish differences in antibiotic efficacy. In addition, future studies should address the bacterial resistance issue.

This document is in the public domain and may be used and reprinted without permission except those copyrighted materials noted for which further reproduction is prohibited without the specific permission of copyright holders.

Suggested Citation

Marcy M, Takata G, Shekelle P, et al. Management of Acute Otitis Media. Evidence Report/Technology Assessment No. 15 (Prepared by the Southern California Evidence-based Practice Center under Contract No. 290-97-0001). AHRQ Publication No. 01-E010. Rockville, MD: Agency for Healthcare Research and Quality. May 2001.

Overview

The objective of this report was to analyze the evidence on the initial management of uncomplicated acute otitis media (AOM) in children. AOM is one of the most common diagnoses in children. Data from the National Ambulatory Medical Care Surveys (NAMCS), which did not differentiate between AOM and otitis media with effusion, indicated that the number of office visits for AOM increased more than two-fold from 1975 to 1990. Gates (1996a) believed that the majority of these cases represented AOM. Based on national data from NAMCS and the National Hospital Ambulatory Medical Care Surveys, and several assumptions detailed in Appendix A, we estimated that 5.18 million episodes of AOM occurred in 1995 at a cost of approximately $2.98 billion, including direct and indirect costs and the costs of sequelae such as otitis media with effusion and chronic middle ear infection. These estimates suggest that increased effectiveness in treating AOM could achieve significant national cost savings.

AOM was defined by the Technical Expert Panel as the presence of middle ear effusion in conjunction with the rapid onset of one or more signs or symptoms of inflammation of the middle ear. Uncomplicated AOM was defined as AOM that is limited to the middle ear cleft. An episode of uncomplicated AOM was considered distinct from a previous episode of AOM and was eligible for initial treatment if the most recent course of antibiotics ended 4 weeks prior to the episode of AOM in question, or if there was documentation by an examiner that a prior episode of AOM had been cleared.

Various medical and surgical treatments have been used for AOM. There is controversy about the need for antibiotics, particularly in children older than 6 months of age with uncomplicated AOM. It is routine to use antibiotics for AOM in the United States, whereas in the Netherlands the standard practice is to observe selected children older than 2 years of age for 48 hours and selected children age 6 months to 2 years for 24 hours before initiating antibiotics. Antibiotics are prescribed if clinical resolution does not occur during the observation period. The presence of alternative treatment paradigms of questioned effectiveness suggests that an evidence-based analysis on the management of AOM is needed to determine the efficacy of antibiotic therapy.

Reporting the Evidence

The Technical Expert Panel limited the scope of this evidence report to three key clinical questions:

• What is the natural history of AOM without antibiotic treatment?

• Are antibiotics effective in preventing clinical failure?

• What is the relative effectiveness of specific antibiotic regimens?

The regimens we analyzed were:

(1) amoxicillin or trimethoprim-sulfamethoxazole vs. other antibiotics,

(2) oral fluoroquinolones,

(3) 60 mg or higher per kg/day of amoxicillin or amoxicillin-clavulanate vs. the standard 40 mg per kg/day,

(4) high-dose amoxicillin therapy twice a day vs. three times a day, and

(5) short-term vs. long-term antibiotic therapy.

The Technical Expert Panel decided that this systematic review would focus on children ages 4 weeks to 18 years, who had uncomplicated AOM and were seeking initial treatment. The major outcomes included presence or absence of signs and symptoms within 48 hours, at 3 to 7 days, 7 to14 days, 14 days to 3 months, and after 3 months; presence or absence of adverse effects from antibiotic treatment; and presence or absence of bacteria and/or resistant bacteria.

Of a total of 41 factors that the Technical Expert Panel felt could influence outcomes apart from antibiotics, the panel ranked age (younger and older than 2 years) and otitis-prone state as the two most important factors. The otitis-prone child was defined as the child who had three or more episodes of AOM in a 6-month period or four or more episodes of AOM in a 12-month period.

Methodology

An 11-member Technical Expert Panel consisting of clinical experts, a consumer, and a representative of a managed care organization convened to:

• advise the project in the ranking of proposed key questions and influencing factors,

• guide the development of the scope and definition of AOM,

• advise in development of the search strategy, and

• review and comment on the analysis plan.

The Technical Expert Panel and project staff developed a literature search strategy. The initial strategy was developed for MEDLINE and was customized for other databases. Project staff searched MEDLINE (1966-March 1999), the Cochrane Library (through march 1999), HealthSTAR (1975-March 1999), International Pharmaceutical Abstracts (1970-March 1999), the Cumulative Index to Nursing & Allied Health Literature (1982-March 1999), BIOSIS (1970-March 1999), and EMBASE (1980-March 1999). Additional articles were identified by review of reference lists in proceedings, published articles, reports, and guidelines.

The initial module of search statements included an explode of "om" (otitis media), which included the headings "om, mastoiditis," "om w/effusion," and "om, suppurative" with the subheading "drug therapy." The next module included the explode of "om," with "om" as a text word. The anti-infectives module used an explode of the mesh heading for anti-infective agents, including antibiotics and other drug groups, and the text words antibiotic, antimicrobial, antibacterial, and specific antibiotic names. Combinations of these modules were used for the literature search. The search was limited to human or undesignated studies and to infant, child, preschool, adolescence, or undesignated subjects. For the natural history search, natural history, natural course, untreated, spontaneous, and self-limited were added as keywords. Two physicians independently screened all titles and/or abstracts for potential inclusion, evaluated the quality of the articles, and abstracted data from full-length articles onto predesigned forms. The selection criteria included human studies that addressed a key question about AOM in children ages 4 weeks to 18 years. Excluded were case reports, editorials, letters, reviews, practice guidelines, and studies on patients with immunodeficiencies or craniofacial deficiencies, including cleft palate. Randomized clinical trials and prospective and retrospective comparative cohort studies were included for the key question on natural history. Only randomized controlled trials were used to address the key questions on the effectiveness of antibiotics. The physician reviewers assessed the quality of controlled trials by using an established tool that focused on randomization, blinding, and reporting of subject withdrawals. They also noted approximation to the diagnostic criteria for AOM established by the Technical Expert Panel.

A physician reviewer also reviewed articles in 13 non-English languages with the assistance of a translator. An initial review of 97 non-English-language articles yielded only two eligible studies, both of which already were known to the project staff. Based on this outcome, the Technical Expert Panel advised the project staff to discontinue further screening of non-English-language citations. Based on peer reviewers' suggestions, the remaining five Dutch and Scandinavian articles were reviewed and no additional studies were found.

Meta-analyses were performed to determine the effectiveness of antibiotic vs. placebo or observational treatment of uncomplicated AOM and to determine the effectiveness of particular antibiotic regimens. Comparisons were established based on the type of antibiotics and the outcome variable under consideration. The technical experts decided to establish comparison groups by individual antibiotic, as this best fit the information required for clinical practice. A meta-analysis was performed for each comparison with three or more randomized controlled studies. Heterogeneity was measured for all meta-analyses, and a random effects model was used to estimate the absolute rate differences. Although subgroup and sensitivity analyses were initially planned based on age of the subject, otitis-prone status of the subject, study-design quality, and approximation to the Technical Expert Panel's diagnostic criteria for AOM, they could not be performed because an insufficient number of studies addressed these factors or because there was an insufficient number of studies within a comparison group.

This evidence report was critiqued by the Technical Expert Panel as well as a 25-member peer-review panel consisting of content experts, consumers, representatives of managed care organizations, an expert in pediatric pharmacology, and methodologists. All comments received from these individuals were reviewed and acted upon as appropriate.

Findings

Future Research

Randomized controlled studies of high internal validity and adequate generalizability are needed to adequately address the key clinical questions asked at the start of this systematic review, including the question of the role of antibiotics in the treatment of uncomplicated AOM. Placebo-controlled trials of adequate power with sufficient patient variation for subgroup analysis are needed. Close monitoring of patients in these studies with a priori plans for appropriate intervention should allay any concerns about suppurative complications and should be a focus of research. Future research should: (1) establish uniform definitions of AOM and relevant clinical, bacteriologic, and societal outcomes; (2) establish uniform diagnostic criteria for AOM; (3) strengthen internal and external validity; (4) address relevant influencing factors, such as age and otitis-prone state, and (5) measure long-term as well as short-term outcomes. Uniform definition and diagnostic criteria are needed to ensure that results across studies can be compared. Appropriate randomization and blinding methods would establish the internal validity of future research. Sufficient variation in future study populations in terms of influencing factors as well as sufficient number of cases may allow results to be applied more readily to specific patients or populations and to be generalized to other populations. Although not addressed in this evidence-based analysis, future research should consider the impact of bacterial resistance on AOM outcomes. Investigators should prioritize AOM outcomes based, in large part, on their importance to patients, their families, and society as a whole because they bear the burden of AOM.

Purpose

The purpose of this evidence-based report is to review the evidence on the natural history of acute otitis media (AOM) and the role of antibiotics in the treatment of uncomplicated AOM in children. The evidence compiled in this report is intended to aid the nominating organizations and others to develop clinical practice guidelines or medical review criteria for AOM. The report also will identify areas for future research.

Scope of Work

The nominating organizations to this topic of the evidence report initially proposed five broad areas of inquiry: (1) factors influencing the course of an episode of AOM, (2) variation in definitions of AOM used in studies, (3) the role of antibiotics, (4) followup, and (5) prevention. Based on degree of importance, including level of controversy and feasibility of answering the question, the Technical Expert Panel limited the scope of this evidence report to the natural history of AOM and the role of antibiotics in treatment of children ages 4 weeks to 18 years. The study also evaluated the influence of specific factors apart from antibiotics on the course of AOM.

Definition

Although readily recognized as a clinical condition, AOM appears to be without a standard definition. In 1985, Paparella, Bluestone, Arnold, et al. (1985), reporting on an international effort to define terms related to otitis media, stated "The panel recognizes the impossibility of recommending terms unanimously acceptable to all." The following four sets of references are often cited as sources for authoritative definitions of AOM and demonstrate the lack of uniform specificity:

• Although not defining acute otitis media, Paparella, Bluestone, Arnold, et al. (1985) defined otitis media as "an inflammation of the middle ear (which may or may not be of infectious origin in contrast to infection, which implies a microbiologic etiology)." An earlier attempt by the same group (Senturia, Paparella, Lowery, et al., 1980) had resulted in 8 categories of otitis media out of 28 that were modified by the term "acute." The group defined "acute process" as the "initial 3 weeks" associated with "onset -- may be short and rapid or signs and symptoms may be subtle and onset slow and insidious."

• In 1992, the Infectious Diseases Society of America (Chow, Hall, Klein, et al., 1992) offered the following definition of AOM: "inflammation of the middle ear as evidenced by the presence of fluid and accompanied by specific signs or symptoms such as ear pain, ear drainage, hearing loss, or nonspecific findings such as fever, lethargy, irritability, anorexia, vomiting, or diarrhea." Diagnostic criteria were embedded in a definition that emphasized middle ear inflammation.

• The Agency for Health Care Policy and Research Otitis Media with Effusion in Young Children. Clinical Practice Guideline (Stool, Berg, Berman, et al., 1994a) mentioned two definitions of AOM within the text of their work: "inflammation of the middle ear with signs or symptoms of middle ear infection" and "fluid in the middle ear accompanied by signs or symptoms of ear infection." One emphasized middle ear inflammation, the other middle ear infection.

• Bluestone and Klein (1996), in a standard text of pediatric otolaryngology, provided two definitions of AOM: "the rapid onset of signs and symptoms of acute infection within the middle ear" and "the rapid and short onset of signs and symptoms of inflammation in the middle ear." The first definition emphasized middle ear infection, the second middle ear inflammation.

In 1985, Paparella, Bluestone, Arnold, et al. (1985) concluded, "It is not the panel's purpose to force usages, but rather to suggest terms to improve clinical and scientific communication. It is appropriate and right and, therefore, incumbent upon every author or speaker to define his/her own terms when communicating with colleagues." For the purpose of systematic review for this evidence-based analysis, a consensus was required among the technical experts on the definition of AOM to evaluate the studies retrieved for analysis of the key questions. A definition would aid in understanding the condition and would describe in codified terms characteristics of the condition that make it unique from other conditions, alluding to etiology and pathophysiology. A well-structured definition also would provide a framework to develop diagnostic criteria to diagnose the presence of AOM in individual patients. This process is detailed in the Methodology section.

Epidemiology

Accurate estimates on the prevalence or incidence of AOM were not found because published population-based estimates are not available on the specific diagnosis of AOM. Data reported from the National Ambulatory Medical Care Survey (NAMCS) provided the best indication of prevalence and incidence of the disease, although nonsuppurative, suppurative, and unspecified otitis media were grouped into the term otitis media, and AOM and OME were not separated in the analyses (Schappert, 1992; Schappert, 1996; Woodwell, 1997a; Woodwell, 1997b; Woodwell and Schappert, 1995). Gates (1996a), commenting on the NAMCS data, stated, "For children, it is probably safe to presume that AOME (AOM with effusion) is the principal disorder noted in these surveys."

Schappert (1992) reported on the 1975-90 NAMCS data. Office visits by patients younger than 15 years of age with the principal diagnosis of otitis media constituted 70.6 percent in 1975; 78.9 percent in 1980; 81.9 in 1985; and 80.5 percent in 1990. From 1975 to 1990, the percentage of office visits with otitis media as the principal diagnosis increased in patients younger than age 15: from 7.3 to 17.4 percent for children younger than age 2; from 10.4 to 18.1 percent for ages 2 to 5; from 6.9 to 10.5 percent for ages 6 to 10; and from 2.6 to 5.2 percent for ages 11 to 14. The number of visits with a principal diagnosis of otitis media per 100 persons per year for the same time period (1975-90) increased from 31.5 to 102.1 for children younger than 2 years of age, 20.8 to 47.8 for those 2 to 5 years of age, 10.2 to 18.2 for those 6 to 10 years of age, and 3.3 to 8.0 for those 11 to 14 years of age.

From 1975 to 1990, the percentage of office visits with a principal diagnosis of otitis media increased from 8.1 to 14.3 percent for pediatricians, from 1.3 to 3.5 percent for general practitioners and family physicians, and 12.8 to 20.2 percent for otolaryngologists. The proportion of office visits for otitis media was higher for pediatricians compared with general practitioners and family physicians or otolaryngologists. In 1990, the number of visits with a principal diagnosis of otitis media per 100 persons per year in children younger than 2 years was 62.9 for pediatricians, 24.0 for general practitioners and family physicians, and 9.1 for otolaryngologists. In 1990, the number of visits with a principal diagnosis of otitis media per 100 persons per year in children 2 to 5 years old was 29.0 for pediatricians, 11.4 for general practitioners and family physicians, and 6.6 for otolaryngologists (Schappert, 1992).

The reports on the NAMCS data for 1993 to 1996 did not stratify by age (Schappert, 1996; Woodwell, 1997a; Woodwell, 1997b; Woodwell and Schappert, 1995). If the 1993-96 data were similar to that in 1975-90, it would be reasonable to conclude that the majority of these patients were younger than 15 years of age. Suppurative and unspecified otitis media was the third most frequently listed principal diagnosis in 1993, the sixth most frequent in 1994, the fourth most frequent in 1995, and the seventh most frequent in 1996 for ambulatory care visits to physician offices, hospital outpatient departments, and emergency departments (Schappert, 1996; Woodwell, 1997a; Woodwell, 1997b; Woodwell and Schappert, 1995). In 1996, visits for a principal diagnosis of otitis media and eustachian tube disorders occurred 82.8 percent of the time in physician offices, 5.3 percent in hospital outpatient departments, and 11.9 percent in emergency departments (Woodwell, 1997a).

Using the published NAMCS data and other published national data such as the National Health Interview Survey, an internal analysis estimated that 5.18 million episodes of AOM and 1.04 million episodes of OME or chronic middle ear infection following AOM occurred in 1995. Each episode of AOM entailed 1.75 visits to the physician, and each episode of OME or chronic middle ear infection entailed 5.1 visits to the physician. The analysis assumed that 20 percent of AOM episodes were followed by OME or chronic middle ear infection. (See Appendix A.)

The NAMCS data also provided information on the duration of office visits for otitis media. The percent of visits for otitis media of 6-15 minutes duration increased between 1975 and 1990 from 64 to 78 percent and was associated with a decrease in visits of less than 6 minutes in duration from 24 to 13 percent (Schappert, 1992). In terms of surgical procedures, the rate of ambulatory surgery visits per 10,000 population for those younger than 15 years of age for otitis media and eustachian tube disorders was 86.9 in 1994 and 83.9 in 1995, based on 498,000 and 484,000 visits respectively (Hall and Lawrence, 1997; Kozak, Hall, Pokras, et al., 1997). In 1995, the number of myringotomy with tympanostomy tube placements reported by NAMCS was 521,000 for a rate of 90.2 procedures per 10,000 children younger than 15 years of age (Hall and Lawrence, 1997).

In general, the NAMCS data demonstrated the importance of otitis media -- and by implication AOM -- based on the prevalence and incidence of the disease and the frequency and duration of visits and surgical interventions.

Risk Factors

Uhari, Mantysaari, and Niemela (1996) conducted a systematic review of studies on the risk factors for development of AOM. This review only searched for studies via MEDLINE and might not have been comprehensive in its scope. Based on the studies included in their study, Uhari, Mantysaari, and Niemela (1996) found the following factors to be associated with increased risk of AOM: family member having AOM, relative risk (RR) 2.63 (95 percent CI, 1.86 and 3.72); day care outside the home, RR 2.45 (95 percent CI, 1.51 and 3.98); family day-care setting vs. care at home, RR 1.59 (95 percent CI, 1.19 and 2.13); parental smoking, RR 1.66 (95 percent CI, 1.33 and 2.06); and pacifier use, RR 1.24 (95 percent CI, 1.06 and 1.46). Having at least one sibling and being in a day-care center also seemed to be associated with increased AOM risk; however, the actual relative risks were not reported. Breast-feeding for at least 3 months was associated with a decreased risk of AOM, RR 0.87 (95 percent CI, 0.79 and 0.95). Breast-feeding in general and breast-feeding for at least 6 months also seemed to be associated with decreased AOM risk; however, the actual relative risks were not reported. Being in a day-care center vs. a family day-care setting did not confer a significant risk for AOM.

Uhari, Mantysaari, and Niemela (1996) also examined the risk factors for recurrent AOM and found the following factors to be associated with an increased risk of recurrent AOM: having at least one sibling, RR 1.92 (95 percent CI, 1.29 and 2.85); child care outside of the home, RR 1.82 (95 percent CI, 1.21 and 2.73); parental smoking, RR 1.76 (95 percent CI, 1.36 and 2.28); and day-care center vs. care at home, RR 1.38 (95 percent CI, 1.19 and 1.61). Breast-feeding in general, RR 0.48 (95 percent CI, 0.32 and 0.72), and breast-feeding for 6 months or longer vs. breast-feeding for less than 6 months, RR 0.69 (95 percent CI, 0.49 and 0.97) were associated with a decreased risk of recurrent AOM. The following factors were not found to be significantly associated with recurrent AOM risk: family day-care setting vs. care at home, day-care center vs. family day-care setting, atopy or allergy, and breast-feeding for 3 months or longer vs. less than 3 months. Rosenfeld (1996) also cited the following factors as mentioned in the literature as risks for recurrence of AOM: young age (younger than 2 years old); multiple previous episodes; bottle feeding; history of ear infections in parents or siblings; and pacifier use.

Risk factors for the outcome of AOM at 2 months were studied by Froom, Culpepper, Bridges-Webb, et al. (1993). Children younger than age 13 months were less likely to be asymptomatic at 2 months than older children. In the study population, several factors had independent risk for poor symptomatic outcome 2 months following diagnosis of AOM. For children age 12 months or younger, the following factors were associated with poor 2-month outcome: history of serous otitis media, odds ratio (OR) 2.3 (95 percent CI, 0.95 and 5.70); episodes of AOM within the past 18 months, OR 1.9 (95 percent CI, 0.82 and 4.51); and male sex, OR 1.7 (95 percent CI, 1.00 and 3.00). For children age 13 months to 15 years, the following factors were associated with poor 2-month outcome: history of ear tubes, OR 2.3 (95 percent CI, 1.46 and 3.58); pus on examination, OR 2.2 (95 percent CI, 1.48 and 3.29); episodes of AOM within the past 18 months, OR 1.8 (95 percent CI, 1.31 and 2.37); and history of serous otitis media, OR 1.4 (95 percent CI, 1.00 and 1.88). Other studies found that bottle feeding, a history of ear infections in parents or siblings, and attendance at a day-care center were factors associated with persistent middle ear effusion after AOM.

Diagnosis

Related to the issue of a uniform definition of AOM is the diagnosis of AOM and the criteria by which the diagnosis is made. Hayden (1980) reviewed 43 full-length articles on AOM that were selected from six journals published from 1955 to 1979. Forty percent of these articles did not describe the criteria by which the diagnosis of AOM was established. In the 26 articles that did describe criteria for the diagnosis of AOM, 18 unique criteria sets were found, and identical criteria sets usually were found in studies performed by the same investigator. These findings, however, are expected considering the lack of a uniform definition.

Rosenfeld (1996) listed the following as possible diagnostic criteria for AOM: acute symptoms either alone, in the presence of middle ear effusion, or without resolution after 24 to 48 hours; purulent otorrhea; bulging tympanic membrane; and middle ear effusion in the presence of acute symptoms. The Otitis Media with Effusion in Young Children. Clinical Practice Guideline (Stool, Berg, Berman, et al., 1994a) stated that pneumatic otoscopy was an appropriate technique to establish the presence of middle ear effusion and that tympanometry could be used as a confirmatory test. Barnett, Klein, Hawkins, et al. (1998) found that acoustic reflectometry was as effective as tympanometry when compared to the gold standard of tympanocentesis.

Froom, Culpepper, Grob, et al. (1990) found that a large proportion of physicians in their study were unsure of the accuracy of their diagnosis of AOM. They reported diagnostic certainty in 58 percent of patients younger than age 12 months, 66 percent of patients age 13 to 30 months, and 73 percent of patients older than age 30 months. Studies from the University of Pittsburgh indicated that practitioners could be trained to diagnose the various forms of otitis media more appropriately (Kaleida and Stool, 1992). The validated otoscopist studies were based on an algorithm which in one study (Mandel, Rockette, Bluestone, et al., 1987) was felt to have potential for systematic bias as judged by the OME Guideline Panel (Culpepper, Long, and Sisk, 1994). As with the issue of the definition, this systematic review required a consensus on the diagnostic criteria for AOM to evaluate the studies retrieved for analysis of the key questions. This process is described in the Methodology section.

Alternatives for Treatment

Various medical and surgical treatments have been used for AOM. For the most part, medical treatment has concentrated on the administration of antibiotics, including penicillins, aminopenicillins, cephalosporins, carbacephems, macrolides, azalides, lincomycins, and combination antibiotics (Rosenfeld, 1996). However, the efficacy of antibiotics in comparison to observation has been in question. The Dutch standard of care is to observe the patient for 48 hours in select children older than age 2 years and for 24 hours in select children age 6 months to 2 years before initiating antibiotics (Froom, Culpepper, Jacobs, et al., 1997). Antibiotics are prescribed if clinical resolution does not occur during the observation period. Other issues pertinent to antibiotics are dose, schedule, and duration of treatment. Antihistamines and decongestants, although shown to be ineffective in the treatment of OME (Cantekin, Mandel, Bluestone, et al., 1983), and analgesics and antipyretics often are used as adjunctive therapies in the treatment of AOM and can be purchased over-the-counter without prescription. One topical analgesic has been shown to be effective in reducing the pain associated with AOM (Hoberman, Paradise, Reynolds, et al., 1997). Alternative medications also have been used and their effectiveness has been subjected to investigation (Barnett, Levatin, Klein, et al., 1999). Myringotomy plus amoxicillin has been shown to be no more effective than antibiotics alone (Kaleida, Casselbrant, Rockette, et al., 1991).

Rosenfeld (1996) suggested that AOM may be prevented through modification of environmental risk factors and through immunizations against both bacteria and viruses associated with AOM. Large doses of xylitol seemed to be effective in reducing the recurrence rate of AOM (Uhari, Kontiokari, and Niemela, 1998). Tympanostomy tubes (Casselbrant, Kaleida, Rockette, et al., 1992; Gebhart, 1981; Gonzalez, Arnold, Woody, et al., 1986) and adenoidectomy (Paradise, Bluestone, Rogers, et al., 1990) appeared effective in preventing recurrent otitis media.

The NAMCS data described practitioner followup patterns for management of AOM. In 1990, compared with 1975, return visits were scheduled to a greater extent (73.2 percent vs. 60.2 percent) for children age 0 to 5 years and to a less extent (46.6 percent vs. 58.3 percent) for those age 15 years or older. As noted above, the NAMCS data did not separate AOM from otitis media effusion (Schappert, 1992).

Outcomes

A wide range of short-term and long-term outcomes must be considered when assessing the effectiveness of treatments for AOM. The following are outcomes often measured in studies: total resolution of signs and symptoms of middle ear inflammation, persistent AOM, chronic OME, suppurative complications, chronic complications, hearing deficits, speech and language problems, behavior problems, and cognitive deficits (Rosenfeld, 1996). Suppurative complications include mastoiditis, petrositis, labyrinthitis, meningitis, subdural empyema, focal otitic encephalitis, brain abscess, and lateral sinus thrombosis (Bluestone and Klein, 1996). Sequelae of the disease include cholesteatoma, tympanosclerosis, chronic tympanic membrane perforation, and ossicular discontinuity and fixation (Bluestone and Klein, 1996; Rosenfeld, 1996).

Adverse effects of the treatments may occur. Antibiotics may lead to diarrhea, rash, anaphylaxis, and symptoms of the hematologic, cardiovascular, central nervous, renal, hepatic, and respiratory systems (Stool, Berg, Berman, et al., 1994b). Antimicrobial drug resistance also may increase with increased use of antibiotics (Dowell, Marcy, Phillips, et al., 1998). Antimicrobial drug resistance appears to be less common in the Netherlands (where antibiotics are not used as frequently for AOM) than in countries that use antibiotics routinely for AOM (Culpepper and Froom, 1997). Antihistamines and decongestants may lead to insomnia, drowsiness, behavior changes, changes in blood pressure, and seizures (Stool, Berg, Berman, et al., 1994). Myringotomy with tympanostomy tube insertion has the following possible adverse effects: external canal laceration, persistent otorrhea, granuloma formation, cholesteatoma, permanent perforation of the tympanic membrane, tympanic membrane flaccidity, retraction, tympanosclerosis, and tube intrusion into the middle ear cleft (Stool, Berg, Berman, et al., 1994).

Cost is another important outcome. Gates (1996b) estimated the cost of AOM for children age 0 to 4 years as $3.15 billion for the year 1994. Stool and Field (1989) estimated the cost of otitis media in general for children age 0 to 6 years as $2.2 to $3.4 billion for the year 1987. Using published NAMCS data, other published national data, and estimates from previously published cost analyses, an internal analysis estimated that the cost of AOM in 1995 was $2.98 billion. The estimated cost of AOM includes the cost of its sequelae. Based on findings by Berman, Roark, and Luckey (1994), it was assumed that OME or chronic middle ear infections followed an episode of AOM in 20 percent of cases. Costs for AOM and OME or chronic middle ear infections were estimated from the existent literature and included direct and indirect costs. The cost per episode for AOM was estimated at $214.19 and included 1.75 visits to a physician. The cost per episode for OME or chronic middle ear infection was estimated at $1,811 and included 5.1 visits to a physician. The estimated number of episodes of AOM and OME or chronic middle ear infection following AOM was described earlier. The $2.98 billion estimate is a minimum estimate because most of the assumptions were conservative. Sensitivity analysis indicated that the estimated national cost for 1995 could be as high as $6 billion. In any case, the national cost of AOM, whether accepting previous estimates or the internally derived estimate, is high. (See Appendix A.)

Ultimately, the importance of these outcomes is dependent on the viewpoint of the evaluator (Russell, Siegel, Daniels, et al., 1996). Furthermore, the perspective of the patient, parent, and physician must be taken into consideration. In general, the societal viewpoint is crucial in terms of balancing health-care costs against the nonhealth-care needs of society; but, the individual patient's condition and circumstances within the societal context also must be taken into account (Gold, Patrick, Torrance, et al., 1996). Quality of life and functional status may be more important than the individual medical outcomes listed above. For otitis media, including AOM, various paradigms for looking at quality of life and functional status are being developed (Alsarraf, Jung, Perkins, et al., 1998; Haggard and Smith, 1999a, 1999b; Rosenfeld, Goldsmith, Tetlus, et al., 1997).

Establishment of Conceptual Framework

The project team prepared a conceptual framework (Appendix B) for the management of AOM for the purpose of discussion and future development of the evidence report. It outlined 11 steps for the management of AOM: steps 1 through 6 related to initial evaluation and treatment and steps 7 through 11 related to followup care. Factors influencing outcomes, outcome measures, and decisions were noted as footnotes to the framework. The framework formed the basis on which key questions and the scope of the evidence report were developed. It was understood that not all steps could be addressed in this evidence report.

Nomination of Technical Experts

A total of 11 organizations were identified by the project team from whom nominations of technical experts and peer reviewers were solicited. The 11 organizations included: 3 topic-nominating partner -- the American Academy of Family Physicians, the American Academy of Pediatrics, and the American Academy of Otolaryngology-Head and Neck Surgery Foundation; 2 hearing and speech organizations -- the American Academy of Audiology (AAA) and the American Speech-Language-Hearing Association (ASHA); 1 multidisciplinary organization -- the Society for Ear Nose Throat Advances in Children (SENTAC); 2 managed care organizations -- PacifiCare Health Systems and Prudential HealthCare; 1 nurse practitioner organization -- the National Association of Pediatric Nurse Associates and Practitioners (NAPNAP); and 2 consumer groups -- Family Voices and Foundation for Accountability (FACCT). A letter of invitation to nominate technical experts and peer reviewers was faxed to each organization, together with a description of the tasks and commitments of each panel member.

Table 1. Technical expert panel

Table 1. Technical expert panel

Technical Expert	Area of Expertise	Affiliation/Location
Theodore G. Ganiats, M.D.	Family medicine	University of California San Diego, CA
Martin C. Mahoney, M.D., Ph.D.	Family medicine	State University of New York Buffalo, NY
Margaretha Casselbrant, M.D., Ph.D.	Otolaryngology	Children's Hospital Pittsburgh, PA
Richard M. Rosenfeld, M.D., M.P.H.	Otolaryngology	SUNY Health Science Center Brooklyn, NY
Leonard B. Weiner, M.D.	Pediatrics, infectious diseases	SUNY Health Science Center Syracuse, NY
F. Lane France, M.D.	Pediatrics practice & ambulatory medicine	Tampa Bay Pediatrics Tampa, FL
Judith S. Gravel, Ph.D.	Audiology	Albert Einstein College of Medicine Bronx, NY
Joanne Roberts, Ph.D.	Speech-language and audiology	Frank Porter Graham Child Development Center Chapel Hill, NC
Tony Arboleda, M.D.	Health plan	MedPartners Medical Group Burbank, CA
Janice Goertz, R.N.,CPNP	Nurse practitioner	Portage, MI
Jennifer Postley, B.A.	Consumer	Family Voice Los Angeles, CA

Topic Assessment and Refinement

A draft workplan for the topic assessment and refinement phase was mailed to the technical experts and partners for review and comments. Two conference calls among all technical experts, partner liaisons, project staff, and the Task Order Officer were held to review data on existing literature, refine the causal pathway, specify patient populations and practice settings, and refine the key questions to be addressed in the evidence report. The conference calls were chaired by the Principal Investigator and were attended by all principal project staff, the technical experts, and the AHRQ Task Order Officer. Reference materials -- including a review of the existing literature, causal pathways, conceptual framework, and a list of potential questions -- were mailed to all participants prior to the calls.

The project staff compiled a summary of information on:

• The incidence and prevalence of AOM, treatment and management alternatives, characteristics and size of the affected populations, and the most affected practice settings and providers.

• The burden of illness associated with AOM, including morbidity, mortality, impact on developmental milestones, quality of life, loss of productivity, medical costs to treat the condition, and other economic costs or burdens associated with the condition.

• The extent to which there is variation in practice associated with the prevention, diagnosis, treatment, or management of AOM.

Further, we reviewed and compiled a summary of six meta-analyses on otitis media that already had been conducted. Both summaries were distributed to the technical experts between the first and second conference call. The project staff also examined in detail the OME Guideline Panel's report Managing Otitis Media with Effusion in Young Children. Quick Reference Guide for Clicians (Stool, Berg, Berman, et al., 1994b) for parallel issues in AOM.

Identification of Key Questions

About 2 weeks before the first conference call, the following documents were distributed to the technical experts: (1) the preliminary conceptual framework for management of AOM,(2) the two-page letter nominating AOM as a topic for evidence analysis that was submitted to AHRQ by the three nominating agencies, (3) a list of the key questions included in the Request for Task Order by AHRQ, and (4) the draft workplan for topic assessment and refinement. Each technical expert was asked to submit a list of key questions within a week using the conceptual framework as a guide and to provide comments and any suggestions for revision of the conceptual framework and the draft workplan for topic assessment and refinement. The letter to the technical experts emphasized that not all steps in the framework would be addressed in the evidence report and that only three to five key questions would be selected.

The project staff compiled a master list of proposed key questions that incorporated the key questions submitted by the technical experts. The master list was forwarded to all technical experts and conference call participants and was used as the basis for discussion during the first conference call.

Table 2. Initial seven key questions for ranking by (more...)

Table 2. Initial seven key questions for ranking by technical experts

Question 1: On the Natural History of AOM. What clinical characteristics (history, symptoms, physical findings) of children are associated with rapid improvement and with prolonged duration of morbidity during an episode of (uncomplicated) acute otitis media (AOM) that is initially managed WITHOUT antibiotics in terms of: Resolution of pain and/or fever at 24 hours, 2-3 days, and 4-7 days? Complete clinical resolution, excluding middle ear effusion (MEE), at 7-14 days? Prevalence of asymptomatic MEE at 2 weeks, 1 month, and 3 months? Incidence of acute suppurative complications (e.g., mastoiditis)?

Question 2: On Definition and Physician Diagnosis of AOM. How accurate is physician/health care provider diagnosis of acute otitis media? What is the definition of AOM or the criteria by which AOM is diagnosed? Is middle ear effusion a diagnostic standard for the diagnosis of AOM? Is inflammation of the tympanic membrane a diagnostic standard for the diagnosis of AOM? Is the presence of acute local or systemic symptoms a diagnostic standard for the diagnosis of AOM? What is the gold standard for establishing the diagnosis of AOM? Is it possible to improve the diagnostic accuracy of physicians and other health care providers in identifying AOM? Reduce false-positive diagnoses? What is the value or utility of parental involvement in diagnosis, for example, with home diagnostics such as acoustic reflectometry or otoscopy, compared with diagnosis without such parental involvement?

Question 3: On Antibiotic Treatment of AOM. Are antibiotics (compared to placebo) effective (in terms of statistical significance and magnitude of absolute clinical benefit above and beyond natural history) in treatment of AOM with respect to the following outcome indicators: Resolution of pain and/or fever at 24 hours, 2-3 days, and 4-7 days? Complete clinical resolution, excluding middle ear effusion (MEE), at 7-14 days? Prevalence of asymptomatic MEE at 2 weeks, 1 month, and 3 months? Incidence of acute suppurative complications (e.g., mastoiditis)? To what degree is selective initial use of antibiotics for (uncomplicated) AOM attributable to: Clinical characteristics, including hearing and/or speech delay? Age of the child? Provider degree of diagnostic uncertainty for true AOM? Parent reliability (low vs. high) to report symptom resolution or progression at 48-72 hours if antibiotics are initially withheld? Parent preference concerning antibiotic therapy? Parent education?

Question 4: On Antibiotic Regimen for AOM. Does the specific antibiotic regimen make a difference? (Antibiotic class) Are antibiotics with broader coverage than amoxicillin or trimethoprim sulfa more cost-effective or cost-beneficial in the initial treatment of AOM? (Antibiotic class) Should antibiotic selection be determined on the basis of local or regional bacterial resistance pattern? (Antibiotic class) What is the utility of fluoroquinolones in treatment of AOM in childhood? What are the side effects? (Dose of antibiotic) What is the value of using 60-80 mg/kg/d of amoxicillin or amoxicillin-clavulanate vs. the standard 40 mg/kg/d? (Schedule of antibiotic) Is twice-a-day high-dose amoxicillin therapy as effective as three times a day amoxicillin therapy in the initial treatment of AOM? (Schedule of antibiotic) Is twice-a-day high-dose amoxicillin therapy as effective as three times a day amoxicillin therapy in the initial treatment of AOM? (Length of treatment) What is the comparative efficacy of short- vs. long-term antibiotic therapy in children younger than 2 years of age and those older than 2 years of age? Is the use of a single injection of ceftriaxone or the use of other cephalosporin for treatment of AOM more or less likely than 7-10 days of oral amoxicillin to predispose to creation of a penicillin/cephalosporin-resistant pneumococcal nasopharyngeal flora? Does the risk-benefit justify the use of a single injection of ceftriaxone? How long after an antibiotic is used is it reasonable for it to be used again to treat AOM in terms of reducing the likelihood of developing resistance to antibiotics?

Question 5: On Non-Antibiotic Pharmaceutical Treatment of AOM. What is the effectiveness (or cost-effectiveness) of non-antibiotic treatments of AOM? Do any of the new nonsteroidal anti-inflammatory drugs hasten the resolution of AOM or decrease the incidence or duration of subsequent otitis media with effusion? Do oral and/or topical analgesics reduce the duration of pain in AOM? Do anti-inflammatory nasal sprays, such as cromolyn or beclomethasone, for relief of allergic rhinitis reduce the incidence of AOM?

Question 6: On Followup Strategies. What are the appropriate (and/or cost-effective) clinical followup strategies (including type, schedule, and frequency of strategies) for children with AOM? Is the effectiveness of followup dependent on age group? Is the effectiveness of followup dependent on clinical severity (single episode of AOM, recurrent AOM, persistent AOM)? Is the effectiveness of followup dependent on clinical characteristics (e. g., do children with speech-language delay or developmental delay have better outcomes with more intense followup)? Is the effectiveness of followup dependent on parent characteristics (including compliance)? What are the indications for change in therapy at followup if followup is effective in improving outcomes in children with AOM? Does tympanocentesis in children with AOM who are still symptomatic after 48 or 96 hours of treatment lead to a change in therapy that improves outcomes? Does audiometry during followup lead to a change in therapy that improves outcomes?

Question 7: On Prevention of AOM. What is the role of each of the following preventive strategies in managing sporadic or recurrent AOM? Do vaccines reduce the occurrence of AOM? Do risk-factor modification measures (interruption of secondhand smoke exposure, hygienic measures) reduce the occurrence of AOM? Do prophylactic antibiotics reduce the occurrence of AOM? Do surgical procedures such as tympanostomy tube insertion and adenoidectomy reduce the occurrence of AOM? Do anti-inflammatory nasal sprays, such as cromolyn or beclomethasone, for relief of allergic rhinitis reduce the incidence of AOM?

AOM=acute otitis media; MEE=middle ear effusion.

The new set of seven key questions -- together with a form to record rankings, the summary of information on the incidence, alternatives for treatment, and impact/costs of AOM; the summary of published meta-analyses; and preliminary counts of citations by key questions -- was sent to the technical experts about 1 week before the second conference call. The technical experts were advised to use the background materials as references to rank five of the seven key questions, using "5" as the most important and "1" as the least important. The main criteria to be used in ranking were importance and feasibility. Importance was to be determined by the potential impact on the outcomes associated with the management of AOM. The importance of a question might be diminished if a recent evidence analysis of high quality already existed. Feasibility was determined by the availability of published studies and the time required to complete the evidence analysis for that question. Technical experts were asked to submit the ranking to the Task Order Coordinator before the second conference call.

Table 3. Rankings of initial key questions by technical experts (more...)

Table 3. Rankings of initial key questions by technical experts

Technical Expert	Key Question for Ranking
	Q1 Natural History	Q2 Definition and Diagnosis	Q3 Antibiotic vs. No Antibiotic	Q4 Antibiotic Regimen	Q5 Non-Antibiotic Treatment	Q6 Followup Strategy	Q7 Prevention
T1	5	4	3	0	0	1	2
T2	4	3	2	5	1	0	0
T3	1	0	5	4	0	3	2
T4	5	0	1	2	0	3	4
T5	5	0	1	2	0	3	4
T6	5	4	2	3	0	1	0
T7	3	0	4	1	0	5	2
T8	2	1	5	4	0	3	0
T9	0	5	4	0	1	2	3
T10	3	1	5	2	4	0	0
T11	0	5	2	4	0	3	1
Total Rank	33	23	34	27	6	24	18

Note: The technical expert numbers were assigned randomly and do not correspond to the listing order.

Table 4. Final version of the three key questions

Table 4. Final version of the three key questions

Question 1: On the Observational or No Treatment or Natural History of AOM. During an episode of uncomplicated acute otitis media (AOM) that is initially managed WITHOUT any active intervention (pharmacologic or surgical) other than topical or systemic medications (that do not contain antibiotics) given for symptomatic relief (e.g., analgesics, antipyretics, antihistamines, decongestants, ear or nose drops), what proportion of children have the outcomes delineated in the Scope of the Evidence Report? To what degree are the above outcomes attributable to the influencing factors delineated in the Scope of the Evidence Report?

Question 3: On Antibiotic Treatment of AOM. Are antibiotics effective (in terms of statistical significance and magnitude of absolute clinical benefit above and beyond placebo/observational/no treatment/natural history) in the initial 1 treatment of uncomplicated AOM with respect to the outcomes delineated in the Scope of the Evidence Report? When antibiotics are used in the initial 1 treatment of uncomplicated AOM, which of the influencing factors delineated in the Scope of the Evidence Report are associated with better outcomes compared with placebo/observational/no treatment/natural history?

Question 4: On Antibiotic Regimen for AOM. Does the specific antibiotic regimen make a difference? (Antibiotic class) Is treatment with antibiotics other than amoxicillin or trimethoprim sulfa more effective or result in lower total cost in the initial 1 treatment of uncomplicated AOM? (Antibiotic class) What is the utility of oral fluoroquinolones in the initial 1 treatment of uncomplicated AOM in childhood? What are the side effects? (Dose of antibiotic) What is the value of using > 60 mg/kg/d of amoxicillin or amoxicillin-clavulanate vs. the standard 40 mg/kg/d in the initial 1 treatment of uncomplicated AOM? (Schedule of antibiotic) Is twice a day high-dose amoxicillin therapy as effective as three times a day amoxicillin therapy in the initial 1 treatment of uncomplicated AOM? (Length of treatment) What is the comparative effectiveness of short- vs. long-term antibiotic therapy in children younger than two years of age and those older than two years of age in the initial 1 treatment of uncomplicated AOM?

1

An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial" treatment if the most recent course of antibiotic ended 4 weeks prior to the episode of AOM in question or if there is documentation by an examiner that a prior episode of AOM has been cleared.

Several issues were brought up during the discussion and were addressed during the revision of the key questions and the drafting of the causal pathways:

• One issue related to the difference between "natural history of acute otitis media when nothing is done, including not seeking care from the health care provider" and "observational treatment prescribed by the health care provider." It was decided that the issue brought up an important point, and Question 1 was retitled to reflect this fact.

• Another issue concerned the real objective of Question 3b. The technical experts discussed whether it was more important to address the amount of influence of the listed factors on the "selective initial use of antibiotics" or to address the "effectiveness of the initial use of antibiotics." The former is an analysis of the prescribing pattern (which is a process variable), and the latter is an analysis of the effectiveness of initial use of antibiotics (which is an outcome variable). Question 3b was thus reworded to address the impact of these factors on outcome.

• On the issue of defining uncomplicated AOM, there were several suggestions, including "inflammation of the middle ear cleft" to "one sign and one symptom" to "definitions used by the different studies." The project team decided that a definition and diagnostic criteria had to be established before the study could continue because it would affect the search strategy. The experts were polled to establish the best definition to use for this evidence report. Several of the technical experts felt that we might have to accept whatever definition of AOM was found in any given study because they would be quite disparate; however, the project staff felt it important to establish a standard definition for comparison.

• Subquestions in Question 4 that addressed complicated AOM rather than uncomplicated AOM were discussed: (4b) Would antibiotic selection based on local or regional bacterial resistance pattern improve the outcome of antibiotic treatment of AOM? and (4h) How long after an antibiotic is used is it reasonable for it to be used again to treat AOM in terms of reducing the likelihood of developing resistance to antibiotics? They were not included because this evidence report was limited to uncomplicated otitis media.

• The definition of endpoints for the study was discussed. The group decided to leave them open to definitions used by different studies. However, it was decided that it would be important to define the most appropriate endpoints before the literature search and then compare the definitions with those found and/or used in the literature.

• The next discussion topic was on the age of the child to be included in the evidence report. It was decided to exclude the neonate. Epidemiologically, a neonate is a newborn up to 28 days old. For this evidence report, studies on children between age 28 days and 18 years were included.

• The following factors that might affect outcomes were added: gender, ethnicity, presence of ear infection in sibling, craniofacial problems, and cost. In addition, based on information from review of literature, project staff added race, sibling(s) in day care, pacifier use, presence of sibling(s), and atopy or allergy. In this report, the term "influencing factors" included both risk factors and effect modifiers that might actually or potentially affect the outcomes.

Identification of Causal Pathways, Study Populations, Practice Settings, and Target Audience

Table 5. Causal pathways for key questions

Table 5. Causal pathways for key questions

Table 6. Preliminary literature search on seven initial (more...)

Table 6. Preliminary literature search on seven initial key questions

Key Question	Search Topic	MEDLINE (1966- 1998)
Natural History	AOM and natural history or untreated	7
	AOM and outcome & process assessment	95
	AOM and persistent	110
	AOM and recurrent	284
	AOM and complications	322
Definition & Diagnosis	AOM and blood, microbiology, csf, classification, cytology, parasitology, diagnosis, radiography, radionuclide imaging, urine, virology	525
	AOM and sensitivity/specificity/false-negative/false-positive/predictive/likelihood/randomized controlled trial/blinded methods	468
	AOM and diagnosis/diagnostic techniques and procedures/laboratory techniques	498
	AOM and sensitivity/specificity/predictive value/likelihood	63
	AOM and medical history taking	5
	AOM and physical exam	39
	AOM and tympanogram	29
	AOM and tympanocentesis	93
	AOM and acoustic reflectometry	9
	AOM and diagnostic techniques otological	124
	AOM and otoscopes/otolaryngology (MeSH 1998+)	5
	AOM and decisionmaking/judgment/problem solving/decision support techniques/physician's practice patterns/referral and consultation/evidence-based medicine	45
Antibiotic Therapy	AOM and antibiotics	790
Non-Antibiotic Therapy	AOM and steroids	22
	AOM and analgesics	29
	AOM and antihistamines/decongestants	44
	AOM and tympanostomy/middle ear ventilation/tympanostomy/tube insertion	111
Followup Strategy	AOM and combination of therapies and followup studies	77
Prevention	AOM and antibiotic prophylaxis/preventive medicine/immunoprophylaxis	136
Review Articles
General	AOM and review articles and human and English	172
Cost 1	AOM and cost 1	52
Meta-analysis	AOM and meta-analysis	27
	AOM and information synthesis (meta-analysis, decision analysis, cost effectiveness analysis, cost-benefit analysis)	29
Incidence/Prevalence	AOM and incidence/prevalence	251 (1966- 1998)
		148 (1987- 1998)
Randomized Control Trials (RCTs)
RCT	AOM and RCT as publication type	198 (1966- 1998)
		129 (1987- 1998)
	AOM and RCT (very broad strategy)	617 (1966- 1998)
		441 (1987- 1998)

1 From HealthSTAR, 59 citations from 1966- 1998.

The three key questions, definition of AOM, and the scope of the evidence report have gone through several revisions as a result of conference discussions and polling. The initial and revised versions of these three documents are included in Appendices C, D, and E, respectively.

During the second conference call, the technical experts provided input regarding the patient populations, practice settings, and target audience for the evidence report. The study population included all children between 28 days and 18 years with uncomplicated AOM. The interventions of interest included placebo or observation with no treatment or antibiotic treatment. Influencing factors included demographic, environmental, clinical, pharmaceutical, and cost factors. All health care settings and types of providers were included. The endpoints of the study included four short-term and four long-term indicators. Short-term indicators at 48 hours included disappearance of signs (discharge, inflammation of the tympanic membrane), symptomatic relief of pain and/or fever, resolution of acute suppurative complications, and complete clinical resolution (except middle ear effusion). Long-term indicators at 3 months included absence of asymptomatic middle ear effusion, resolution of suppurative complications (e.g., mastoiditis), absence of speech and/or hearing problems, and absence of recurrence of AOM.

Refinement of Key Questions, Definition of AOM, and Scope

Table 7. Technical expert's responses to four issues (more...)

Table 7. Technical expert's responses to four issues related to definition of acute otitis media January 7, 1999

	Expert Number
	1	2	3	4	5	6	7	8	9	10	11
1. Definition of Rapid Onset:
(a) Within 48 hours	x	x	.	x	x	x	x	.	x	x	x
(b) Within 7 days	.	.	.	.	.	.	.	x	.	.	.
(c) Other (72 hours)	.	.	x	.	.	.	.	.	.	.	.
2. GI symptoms to be included?
(a) Yes	.	.	x	x 2	.	.	.	.	.	.	.
(b) No	x 1	x	.	.	x	x	x	x	x	x	x
3. Hearing loss included?
(a) Yes	x	.	x	.	.	x	.	x	x 5	x 5	x
(b) No	.	x	.	x 3	x	.	x 4	.	x 5	x 5	.
4. The two signs for MEE
(a) Agree	x 6	x	x 7	.	x	x	.	x	x 10	x 10	x
(b) Disagree	.	.	.	x 8	.	.	x 9	.	.	.	.

Notes:

x Answer checked by expert.

1

GI findings can occur, but to have rapid onset of GI findings and an abnormally colored ear doesn't make it for me.

2

GI symptoms should be included on the list because they may be the only manifestation of acute illness in an infant or young child. Anorexia (secondary to nausea) and vomiting may be direct vestibular manifestations of middle ear infection or fluid. Diarrhea is not as easy to justify, and if the panel feels strongly about removing diarrhea from the symptom list I have no objection. Be aware, however, that "traditional" AOM definitions have always included diarrhea as a qualifying symptom.

3

Hearing loss is a manifestation of middle ear effusion, not acute infection. Further, not all middle ear fluid is accompanied by hearing loss, and not all hearing loss caused by middle ear fluid is readily perceptible to parents or clinicians. Traditional AOM definitions have never included hearing loss as a qualifying symptom.

4

Anyone with AOM will have hearing loss at diagnosis.

5

We are concerned about including hearing loss as a clinical finding. First, it is highly unlikely that a formal audiologic evaluation would be administered when a child presents to the physician with acute OME. Clinical symptoms (fever, irritability, otorrhea, otologia) could prevent a reliable hearing test (audiogram and tympanogram) from being obtained. During the course of the episode when the acute clinical symptoms had subsided, an audiologic evaluation would be more appropriate. Second, without a direct measure of hearing (that is, obtaining an audiogram) there is no way to determine if hearing loss exists. Parent report has been demonstrated to be a highly inaccurate means of detecting hearing loss (Rosenfeld et al.). Further, we are not uncertain if studies examining hearing loss and OM have specifically examined hearing loss during AOM. However, if hearing loss stays in as a clinical finding, please delete "older children." Hearing loss, when present, occurs in both young children as well as older children, and can be reliably assessed in infants and young children. Therefore, hearing loss would be a clinical finding regardless of the age of the child.

6

Need to clarify one or both signs.

7

Change 3(b) to full or bulging with red or white opacification.

8

Middle ear fluid with AOM generally manifests as a full or bulging tympanic membrane, which is opaque and has limited or absent mobility to pneumatic otoscopy. Changes in tympanic membrane color are NOT indicative of effusion. Changes in lucency are also NOT indicative, unless accompanied by reduced mobility.

9

Suggest that OME and inflammation with changes go under part two.

10

We agree with the addition with of the two clinical signs, but suggest that position of the TM (e.g., bulging, retracted) be included as a third clinical sign (Karma Penttila, Sipila, et al., 1989).

More Comments on Definition of Acute Otitis Media

More Comments on Definition of Acute Otitis Media

Comment 1: I continue to be troubled by the AOM definition as it seems to confuse the terms "symptom" & "sign." To clarify: Symptom - change in normal function Sign - physical finding, evidence

Specifically, the current AOM definition (12/23) seems to mix signs & symptoms in items #2 & 3. Symptoms would include ear pulling, irritability, otalgia (subjectively) & hearing loss (subjectively), while signs include otalgia, otorrhea, fever, and hearing loss (objectively). Also, the basic elements which define inflammation (e.g., redness, pain, fever, swelling) are not consistently addressed in the present definition.

Comment 2: Revised proposed AOM definition. I still have some problems with the wording. As noted in a prior e-mail, I recommend the following definition:

"Acute otitis media is defined as: 1) signs of middle-ear effusion (new onset or pre-existing) accompanied by 2) rapid and short onset (within 48 hours) of clinical findings including one or more of the following: otalgia (ear pulling in an infant), otorrhea, fever, irritability, anorexia, vomiting, or diarrhea."

In contrast to your proposed definition, the one above puts emphasis on middle-ear effusion (by placing it first) and eliminates ambiguity about middle-ear effusion duration (your definition implies recent onset, yet AOM can develop on a pre-existing effusion of long duration).

Table 8. Questionnaire to poll technical experts on (more...)

Table 8. Questionnaire to poll technical experts on four unresolved issues related to definition of AOM, key questions and scope

In the definition of AOM, should we add "excluding retracted tympanic membrane" to statement (1)? That is, would you prefer to state AOM is defined as (1) presence of middle ear effusion as indicated by limited or absent mobility of the tympanic membrane, excluding retracted tympanic membrane, with or without....?" Yes___ No___ No preference___

Comments:

2 In the definition of AOM, the majority of you defined rapid onset as "within 48 hours." Now we need to define the beginning and end points of the 48-hour period. Please write down the beginning and end points to define rapid onset. (e.g., Rapid onset is defined as less than or equal to 48 hours from the time the acute signs or symptoms -- otalgia, otorrhea, irritability, or fever -- were first noticed by the parent to the time the parent first contacted the physician's office, whether by phone or in person.) 3 In the key questions, we need to define "initial treatment" or "initially managed." We are proposing the following options: [circle (a), (b), and/or (c)] a) An episode of uncomplicated AOM may be considered distinct from a previous episode of acute otitis media and eligible for "initial treatment" if the most recent course of antibiotic ended 2 weeks prior to the episode of AOM in question. b) An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial treatment" if the most recent course of antibiotic ended 4 weeks prior to the episode of AOM in question. c) An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial treatment" if there is documentation by an examiner that a prior AOM has been cleared.

Comments:

4 One of the influencing factors in the scope is "otitis prone," which is used to replace "prior history of recurrent AOM" and "multiple previous episodes." How would you define "otitis prone"? [circle one] a) > three episodes in 6 months or > four episodes in 12 months b) Other (please specify and write rationale)

Table 9. Technical expert's responses to four unresolved issues (more...)

Table 9. Technical expert's responses to four unresolved issues¹ related to definition, key questions, and scope

Expert	Issue 1	Issue 2	Issue 3	Issue 4
1	Yes	<=48 hours from point signs/symptoms noted by parent and health system contacted 4	b	a
2	Yes	<=48 hours from point of initial signs/symptoms	c	four episodes in 12 months
3	Yes	<=48 hours from point signs/symptoms noted by parent and health system contacted	a and c	a 7
4	No	<=48 hours from point signs/symptoms noted by parent and health system contacted	b 6 and c	a 8
5	No preference 2	<=48 hours from onset of acute signs or symptoms until diagnosis confirmed by practitioner	b and c	a
6	Yes	<=48 hours from point signs/symptoms noted by parent and health system contacted	b and c	a
7	No 3	<=48 hours from point signs/symptoms noted by parent and health system contacted5	b	a
8	No	<=48 hours from point signs/symptoms noted by parent and health system contacted	a	a
9	No preference	<=48 hours from point signs/symptoms noted by parent and health system contacted	c	a
10	Yes	<=48 hours from point signs/symptoms noted by parent and health system contacted	c	a

1

Issues: Issue 1: In the definition of AOM, should we add "excluding retracted tympanic membrane" to statement (1)? Issue 2: In the definition of AOM, the majority of you defined rapid onset as "within 48 hours..." Please write down the beginning and end points to define rapid onset. Issue 3: In the key questions, we need to define "initial treatment" or "initially managed." We are proposing the following options [circle (a), (b), and/or (c)]: (a) An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial treatment" if the most recent course of antibiotic ended 2 weeks prior to the episode of AOM in question. (b) An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial treatment" if the most recent course of antibiotic ended 4 weeks prior to the episode of AOM in question. (c) An episode of uncomplicated AOM may be considered distinct from a previous episode of AOM and eligible for "initial treatment" if there is documentation by an examiner that a prior AOM has been cleared. Issue 4: One of the influencing factors in the scope is "otitis prone," which is used to replace "prior history of recurrent AOM" and "multiple previous episodes." How would you define "otitis prone?" [circle one]: (a) three episodes in 6 months or four episodes in 12 months; (b) other (please specify and write rationale).

2

"I remain troubled by the definition. For example, a child with stable OME who develops febrile rotavirus diarrhea meets the definition (whether they have any of the 1) a, b, or c findings). I'm not sure how that can work."

3

"Unless there is evidence to support its use in diagnosis, remove acoustic "reflectometry" from Definition as well as in Scope of the Evidence Report: Monitoring during episode. (See section on acoustic reflectometry in Clin. Pract. Guideline 12 on OME, Stool, et al., 1994, p. 34-35.)

4

"I always thought 'within 48 hours' referred to the speed of onset, not the duration. If we go with duration, I think your example is fine. If we go with speed, I have a hard time coming up with the right words."

5

"I Agree with above, but change 'physician' to 'examiner' (to include family physician, otolaryngologist, pediatrician, nurse practitioner, PA, etc.)."

6

"AOM within 4 weeks of a prior episode would classify as recurrence."

7

"Patient should also be asymptomatic."

8

"In the context of an 'influencing factor' you may wish to expand this definition to include episodes of OME, not just AOM." "Episodes should also be 'well documented and separate.'"

OME=otitis media effusion.

Based on the responses to the unresolved issues, the three documents were revised further. The next set of revisions of the key questions, definition, and scope (Appendices C.3, D.3, and E.3) were distributed to the panel of experts for review and for discussion at the third conference call and resulted in more changes. The final revisions of the three documents are included in Appendices C.4, D.4, and E.4.

The short version of the final definition of AOM is as follows: Acute otitis media is the presence of middle ear effusion in conjunction with the rapid onset of one or more signs or symptoms of inflammation of the middle ear.

The long version of the final definition is as follows:

• Presence of middle ear effusion as demonstrated by the actual presence of fluid in the middle ear as diagnosed by tympanocentesis or the physical presence of liquid in the external ear canal as a result of tympanic membrane perforation or indicated by limited or absent mobility of the tympanic membrane as diagnosed by pneumatic otoscopy, tympanogram, or acoustic reflectometry with or without the following: (a) opacification, not including erythema, (b) a full or bulging tympanic membrane, or (c) hearing loss, and rapid onset (i.e., up to 48 hours from the onset of acute signs or symptoms first noted by the parent or guardian to the time of contact with the health system) of one or more of the following signs or symptoms, with or without anorexia, nausea, or vomiting: (a) otalgia (or pulling of ear in an infant), (b) otorrhea, (c) irritability in the infant or toddler, or (d) fever.

Literature Search

We conducted the literature search on seven databases: MEDLINE, the Cochrane Library, HealthSTAR, International Pharmaceutical Abstracts (IPA), Cumulative Index to Nursing and Allied Health Literature (CINAHL), BIOSIS, and EMBASE. We also searched for relevant articles in symposiums and reference lists of published articles, reports, and guidelines.

The MEDLINE database is produced by the U.S. National Library of Medicine and is widely recognized as the premier source for bibliographic coverage of biomedical literature. It encompasses information from Index Medicus, Index to Dental Literature, and International Nursing, as well as other sources of coverage in the areas of allied health, biological and physical sciences, humanities and information science as they relate to medicine and health care, communication disorders, population biology, and reproductive biology. The MEDLINE database includes citations and abstracts since 1966.

The Cochrane Library contains several databases: (1) The Cochrane Database of Systematic Reviews containing Cochrane reviews published by the Cochrane Collaboration, an international organization dedicated to applying evidence-based medicine principles to the review of important clinical topics; (2) The Cochrane Controlled Trials Register, which is a bibliographic database of controlled trials; (3) The Database of Abstracts of Reviews of Effectiveness (DARE), which includes structured abstracts of systematic reviews that have been critically appraised by reviewers at the NHS Centre for Reviews and Dissemination in York and by other reviewers (e.g., the American College of Physicians' Journal Club and the journal Evidence-Based Medicine); and (4) The Cochrane Review Methodology Database, which is a bibliography of articles on the science of research synthesis.

The HealthSTAR database contains citations and abstracts (when available) of journal articles, monographs, technical reports, meeting abstracts and papers, book chapters, government documents, and newspaper articles on health services, technology, administration, and research. It focuses on both the clinical and nonclinical aspects of health care delivery. Topics covered include evaluation of patient outcomes; effectiveness of procedures, programs, products, services and processes; and health services research. It is produced cooperatively by the U.S. National Library of Medicine and the American Hospital Association. HealthSTAR replaces the former Health Planning and Administration database. The database includes citations and abstracts since 1975.

The IPA is produced by the American Society of Health System Pharmacists. It provides worldwide coverage of pharmaceutical science and health-related literature. Coverage includes drug therapy, toxicity, and pharmacy practice, as well as legislation, regulation, technology, utilization, biopharmaceutics, information processing, education, economics, and ethics as related to pharmaceutical science and practice. The database currently contains over 300,000 records and includes citations since 1970.

The CINAHL database is produced by CINAHL Information Systems. It provides comprehensive coverage of the English-language journal Literature for Nursing and Allied Health Disciplines. Material from more than 950 journals is included in CINAHL. Also included are health care books, nursing dissertations, selected conference proceedings, standards of professional practice, and educational software. There is selective coverage of journals in biomedicine, the behavioral sciences, management, and education. The database currently contains over 250,000 records. In total, more than 500 journals are regularly indexed; online abstracts are available for more than 150 of these titles. The database includes citations since 1982.

BIOSIS Previews, produced by BIOSIS, is the world's most comprehensive reference database in the life sciences. It covers original research reports and reviews in biological and biomedical areas. Nearly 7,000 serials are monitored for inclusion. In addition, the database covers content summaries, books (including software from 1992 to present), and information from meetings. Content summaries include notes and letters, technical data reports, reviews, U.S. patents from 1986 to 1989, translation journals, meeting reports from 1980 to present, bibliographies, nomenclature rules, and taxonomic keys. The BIOSIS Previews database includes the contents of Biological Abstracts (1969 to present), Biological Abstracts/RRM (1980 to present), and BioResearch Index (1969 to 1979). The BIOSIS database includes citations since 1970.

EMBASE, the Excerpta Medica database produced by Elsevier Science, is a major biomedical and pharmaceutical database indexing more than 3,800 international journals. EMBASE is one of the most widely used biomedical and pharmaceutical databases. The database currently contains more than 6 million records, with more than 400,000 citations and abstracts added yearly. The EMBASE database contains citations since 1980.

The project librarian developed an overall search strategy incorporating the input from the technical experts and following the scope of the project. The initial search strategy was developed for MEDLINE and then customized for the other databases. The MEDLINE search strategy used both controlled vocabulary terms and keywords. The strategy was organized into modules or clusters of search statements. The initial module (called om with sh dt or otitis media with subheading drug therapy) included using what is referred to as an "explode" of om, which includes the controlled vocabulary headings om, mastoiditis, om w/effusion, om, suppurative with the subheading drug therapy. The next module (called om) included the explode of om as well as om as a text word. The anti-infectives module used explode of the mesh heading for anti-infective agents, which includes antibiotics and other drug groups as well as the text words antibiotic, antimicrobial, and antibacterial. The antibiotics module included the names of specific antibiotics previously identified, which were entered as text words. The search involved a combination of the om module with the anti-infectives module and another combination of the om module with the antibiotics module. We also used either the om with dt set or the om with anti-infectives set or the om with antibiotics as a keyword set in the search. From this set editorials, letters, reviews, practice guidelines, consensus development conferences, and case reports were excluded, except when the case report was also a randomized controlled trial or clinical trial. This set was then limited to human or undesignated. In other words, animal studies were excluded. The final set was limited to infant, child, preschool, adolescence, or undesignated. Newborn, adult, middle age, and aged were excluded.

For the search for natural history, natural history, natural course, and untreated were added as keywords. During the third telephone conference with the technical experts, an expert advised adding "spontaneous" and "self-limited" as keywords to the search. Adding these terms, however, uncovered no additional articles.

The first MEDLINE search resulted in 2,284 titles/abstracts. Two subsequent MEDLINE alerts (additions of new articles after the last update) added another 27 titles/abstracts. Additional titles/abstracts included 217 from the Cochrane Library search, 11 from the HealthSTAR search, 88 from the CINAHL search, 248 from the IPA search, 745 from the BIOSIS search, and 2,418 from the EMBASE search. Further screening of and removal of duplicates from BIOSIS and EMBASE resulted in retaining 154 titles/abstracts from BIOSIS and 421 from EMBASE. The search strategy included the following criteria: publication years after 1966, all languages, keywords, human or non-human study, and search hierarchy. The ending date of all searches was March 31, 1999. All databases were searched using the Ovid search system through the World Wide Web with the exception of the Cochrane Library. The files searched included the Cochrane Database of Systematic Reviews, DARE, and The Cochrane Controlled Trials Register.

EndNote software (EndNote Windows Version 3.0, 1st Edition. Niles Software Inc., Berkeley, CA) was used to keep a complete record of all titles/abstracts and to identify duplications. It was able to store, organize, and track references by source (e.g., identified in MEDLINE), search strategy (date of search, index code specifying search criteria used), and a unique identification (UI) code for each article (assigned by source used to find article). Electronic removal of duplicate citations was supplemented by manual cross-checking. In the event an article was identified through an expert panel member or reference checking, the title and author of the reference were entered into MEDLINE through the Ovid search system (Ovid Technologies, Inc. 1998, Version: 7.8 Millennium source ID 1.3932.1.156.1.7, Revision: 1.303.2.8) to determine the UI. If a UI could not be found for the article, an alternate identification code was assigned.

EndNote assigned a record number to each new reference added to the master file. This number would not change once an article was added to the list and was used, in addition to the UI, to sort references for article retrieval and review.

Upon completion of the literature search and duplicate checking, the master list generated from EndNote was exported to a Microsoft Excel spreadsheet for data export for analysis. Codes including status of article retrieval, reviewer, and the results of the review were added.

Review of Retrieved Abstracts Against Screening Criteria

After retrieval of titles and abstracts from the literature search, two physician reviewers reviewed the abstracts against the inclusion/exclusion criteria to determine eligibility for inclusion in the evidence synthesis as defined in the scope and key questions. Titles/abstracts were not masked prior to review. A predesigned screening form (Appendix F) was used to record the reviews. Instructions for screening also were provided (Appendix F). The screening results for each title/abstract were matched between the two reviewers by the Task Order Coordinator. Discrepancies on inclusion or exclusion were resolved in conference among the two reviewers and the coordinator. The data were entered into a Microsoft Excel database from which interrater reliability statistics of agreement and agreement adjusted for chance (kappa statistic) were calculated. Summary reports indicated those abstracts that passed the screening criteria and those that failed, along with the reasons for failure.

Table 10. Preliminary screening results of titles and (more...)

Table 10. Preliminary screening results of titles and abstracts from seven databases

Source	Total Title/Abstract	Before Resolution			After Resolution
		Reject	Accept	Unsure	Reject	Accept
MEDLINE	2,322	1,800 (78%)	259 (11%)	263 (11%)	1,931 (83%)	391 (17%)
Cochrane Library	217	161 (74%)	46 (21%)	10 (5%)	163 (75%)	54 (25%)
HealthSTAR	11	5 (46%)	1 (9%)	5 (46%)	9 (82%)	2 (18%)
CINAHL	88	79 (90%)	1 (1%)	8 (9%)	83 (94%)	5 (6%)
IPA	248	210 (85%)	27 (11%)	11 (4%)	239 (96%)	9 (4%)
BIOSIS	154	41 (27%)	75 (49%)	38 (25%)	53 (34%)	101 (66%)
EMBASE	421	204 (48%)	101 (24%)	116 (28%)	223 (53%)	198 (47%)
All Sources 1	3,461	2,500 (72%)	510 (15%)	451 (13%)	2,701 (78%)	760 (22%)

1Citations from each source are not independent.

CINAHL=Cumulative Index to Nursing and Allied Health Literature; IPA=International Pharmaceutical Abstracts.

Retrieval and Review of Full Articles

The titles/abstracts identified as requiring further review were forwarded to the library for full article retrieval. Libraries at both the Los Angeles County - University of Southern California Medical Center and at the University of Southern California Health Sciences Campus were the primary sources of the articles. Those not found were retrieved through the Inter-Library Loan Program.

Because a large number of titles/abstracts had inadequate information for full evaluation, a second screening with the full articles was conducted. Like the first screening, two physician reviewers independently reviewed each article and filled out a screening form. Articles were not masked prior to review. Discrepancies on inclusion/exclusion were resolved through conference calls.

Table 11. Results of secondary screening of 760 articles (more...)

Table 11. Results of secondary screening of 760 articles

	English	Non-English
Total Number of Articles Number of Articles obtained, not reviewed Number reviewed Number rejected Number accepted	487 0 487 415/487 (85%) 72/487 (15%)	273 176 97 95/97 (98%) 2/97 (2%)
For Rejected Articles: Reason of Rejection R1=Not a study R2=Nonhuman subjects R3=Not on acute otitis media R4=Not in age range of study R5=Study population not in scope R6=Study design not appropriate R7=Not addressing any key questions 1 R8=Duplicate of other studies R9=Data not abstractable, require contacting author	415 56 (13%) 2 (0%) 41 (10%) 1 (0%) 0 (0%) 25 (6%) 228 (55%) 28 (7%) 34 (8%)	95 19 (21%) 2 (2%) 16 (17%) 11 (12%) 0 (0%) 1 (0%) 37 (39%) 9 (9%) 0 (0%)
For Accepted Articles: Key Questions Addressed (n=74) Key Question 1: Natural History of AOM 2 Key Question 3. Use of Antibiotics 3 Key Question 4a: Amoxicillin or TMP-SMZ 4 vs. others Key Question 4b: Oral fluoroquinolones vs others Key Question 4c: High-dose vs. standard-dose amoxicillin Key Question 4d: Bid vs. tid high-dose amoxicillin Key Question 4e: Short- vs. long-term antibiotic therapy	72 10 8 34 0 1 1 35	2 0 0 0 0 0 0 2

1

163 of the 228 addressed potential key questions.

2

Four articles were added to Key Question 1 after review of references of other articles.

3

One article was added to Key Question 3 after review of references of other articles.

4 TMP-SMZ: trimethoprim-sulfamethoxazole

Of the 72 English articles eligible from the electronic search, 10 addressed the natural history question (Key Question 1), 8 addressed the use of antibiotic question (Key Question 3), 34 addressed the comparison of amoxicillin or ampicillin with other antibiotics (Key Question 4a), none addressed the comparison of oral fluoroquinolones with other antibiotics, 1 addressed the comparison of high dose vs. standard dose of amoxicillin, 1 compared twice a day vs. three times a day of amoxicillin, and 35 compared short-term versus long-term antibiotic therapy.

Table 12. Distribution of 273 non-English articles (more...)

Table 12. Distribution of 273 non-English articles by language

Language	First Batch 1 N=2,875	Second Batch 2 N=585	Total
Japanese	22	79	101
French	17	12	29
German	14	10	24
Russian	13	3	16
Italian	6	16	22
Spanish	4	17	21
Danish	7	3	10
Portuguese	1	9	10
Polish	8	0	8
Dutch	4	2	6
Swedish	3	0	3
Norwegian	2	0	2
Romanian	2	0	2
Czech	0	1	1
Finnish	1	0	1
Hebrew	1	0	1
Turkish	0	1	1
Ukrainian	1	0	1
Unknown	14	0	14
Total	120	153	273

1

First batch included MEDLINE, Cochrane, IPA, CINAHL, and HealthSTAR.

2

Second batch included BIOSIS and EMBASE.

Table 13. Results of reviewing 97 non-English articles (more...)

Table 13. Results of reviewing 97 non-English articles

Language	Number Articles Reviewed	Reject	Accept	R1	R2	R3	R4	R5	R6	R7	R8	R9
Japanese	20	20	0	4	1	2	7			6
French	17	15	2	2	1		1			10	1
German	13	13	0	5		4				4
Italian	5	5	0	1		1	1			2
Spanish	4	4	0	1		1				2
Russian	13	13	0	2		5	2			4
Danish	7	7	0	1		1					5
Polish	8	8	0	1		1				6
Dutch	3	3	0	1							2
Swedish	3	3	0	1						1	1
Norwegian	2	2	0	1						1
Romanian	1	1	0			1
Hebrew	1	1	0							1
Total	97	95	2	20	2	16	11			37	9

Reason for Rejection:

R1=Case report/editorial/letter/clinical practice/overview/practice guidelines/consensus statements

R2=Nonhuman subjects

R3=Study condition NOT acute otitis media

R4=Not within age range of study

R5=Study population is on immunodeficiencies or craniofacial deficiencies

R6=Study design not appropriate

R7=Not addressing any key questions

R8=Duplicate of other studies in the database

R9=Data not abstractable from article, required contacting author

The screening and review of the 97 non-English articles resulted in the use of 2 articles for one of the key questions, a yield rate of two percent. The results were presented to the Technical Expert Panel during the last conference call and it was decided unanimously that the yield was too low to warrant further screening and review of the second batch of non-English articles.

Finalization of Articles for Inclusion

The last activity in searching for articles was the identification and review of potential articles from references of studies and/or articles. The tables of contents from the second to sixth Proceedings of the International Symposium on Recent Advances in Otitis Media were screened by the literature reviewer, and a total of seven articles were retrieved for review of potential inclusion. This process yielded the inclusion of one additional article (Ostfeld, Segal, Kaufstein, et al., 1988) for Key Question 1. The screening of references of other articles, also by the literature reviewer, resulted in the retrieval and review of another 23 articles, 5 of which addressed the key questions. Among the five, articles by Townsend (1964), Thalin, Densert, Larsson, et al. (1986), and Bollag and Bollag-Albrecht (1991) were added to the articles for Key Question 1. The Thalin, Densert, Larsson, et al. (1986) article also was added to the articles for Key Question 3. The Howie and Owen (1987) study was added to the list of articles for Key Question 4a, and the article by Stickler, Rubenstein, McBean, et al. (1967) was added to the list for Key Question 4e.

Further review of articles found that the article by Cohen, de La Rocque, Boucherat, et al. (1997) in French reported results of the same study as the article by Cohen, Bingen, Varon, et al. (1997) in English except that different outcomes were reported. Based on the peer reviewers' comments on the draft evidence report, the study by van Buchem, Peeters, and van't Hof (1985) was added to the list of studies addressing the natural history question. Three Dutch and three Danish articles were reviewed with the assistance of translators; no additional articles were found for inclusion. Three of the six articles were duplicates of the English-language articles that already had been reviewed. Two of these were review articles and one was not retrievable.

Combining all the sources, including the electronic databases, searches of reference lists, and peer reviewers' comments, a total of 80 studies in 85 articles were included in answering the key questions. Of the 80 studies, 74 were randomized controlled trials and 6 were observational cohort studies; 15 addressed Key Question 1, 8 addressed Key Question 3, 35 addressed Key Question 4a, none addressed Key Question 4b, 1 each addressed Key Questions 4c and 4d, and 36 addressed Key Question 4e.

The remaining 680 articles that were not included in addressing the key questions in this evidence report are listed in the bibliography.

Review and Assessment of Study Quality

The criteria for the assessment of study quality were established prior to the review of articles. The criteria developed by Jadad, Moore, Carroll, et al. (1996) were used to evaluate the quality of randomized controlled trials. The Jadad score had a range of 0 to 5. For a given study, 1 point was awarded if it was randomized, 1 point if it was double-blind, and 1 point if it described withdrawals and dropouts. An extra point was awarded if the method of randomization and/or double-blinding was appropriate; conversely, one point was subtracted if the method was inappropriate. The criteria used to evaluate the quality of cohort studies and case-control studies were based on the work by the McMaster University Group (Sackett, 1981; Trout, 1981; Tugwell, 1981). The quality of cohort studies was evaluated against eight components, which included presence or absence of a clear definition of the study cohort, an early inception point, a clear pathway of patient entry, complete followup, description of dropouts, objective outcome criteria, "blind" outcome assessment, and adjustment for extraneous factors. An Article Quality Review Form (Appendix G.1) was developed. The quality of definition of AOM was evaluated using the three components of the proposed definition established by the Technical Expert Panel: middle-ear effusion, rapid onset, and signs/symptoms of inflammation. The form collected data on the study design, the key questions and subquestions being addressed, quality of definition of AOM, and quality of study. Instructions also were prepared for reviewing the quality of articles and completing the quality review form (Appendix G.2).

Quality reviews were carried out in the same manner as the screening of articles for inclusion/exclusion. Articles were not masked prior to review. Two physician reviewers independently evaluated the quality of the articles and filled out the quality review forms. The Task Order Coordinator matched the reviews of the two reviewers and resolved minor discrepancies. Conferences were held to resolve discrepancies whenever needed. During these meetings, participants determined the disposition of each article in terms of eligibility for inclusion into analysis and finalized resolutions for all items.

Data Abstraction

Of the articles eligible for inclusion in the Evidence Report, data abstraction was carried out by two physician reviewers. Data abstracted included parameters necessary to define study groups, inclusion/exclusion criteria, influencing factors, and outcome measures to be used in analysis. A sequential resolution strategy was used to match and resolve the screening and review results of the two reviewers. Telephone conferences were held among the two reviewers and the Task Order Coordinator, who was a health services researcher, to resolve discrepancies and refine instructions. The data abstraction form used is included in Appendix H.

Procedures to Reduce Bias, Enhance Consistency, and Check Accuracy

The following procedures were used to minimize biases:

• Two physician reviewers screened and reviewed titles/abstracts and full articles in every stage of the selection process to reduce selection bias. Their percentage of agreement on inclusion/exclusion was 90 percent.

• Completeness of the retrieved articles was assessed by cross-checking with studies used in other meta-analyses and references listed in review articles. EndNote was used to check batches of articles added to the master list for duplication. The software checked for duplicate references by comparing author, year, title, and reference type. Following the importation of the first literature search, subsequent references were examined for duplication using this EndNote feature prior to their addition to the master list. After the master list was completed, a second check was performed to ensure there were no duplicate entries. This was done manually by scanning the master list after sorting by author and title.

• Multiple sources and unpublished material identified by the expert panel and internal content experts were searched. Funnel plots -- scatter plots of sample size vs. the estimated effect size from each study -- were studied to assess the extent of publication bias. When publication bias existed, a "bite" would be taken out of the funnel plot, typically at the null effect level.

• To reduce bias in the assessment of study quality, explicit preset criteria by Jadad, Moore, Carroll, et al. (1996) were used for study quality. Inter-rater reliability was used to assess the extent of bias in assessing study quality. Because of time and resource constraints, original authors were not contacted for additional information regarding study quality and reporting.

The mechanisms used to enhance consistency include the use of predesigned forms with explicit instructions and continuous and prompt resolution of discrepancies. After data collection, each of the variables collected on the data abstraction forms was entered into a Microsoft Excel spreadsheet. Cross-checking of variables for individual studies abstracted by each data collector identified discrepancies, which were then resolved by rechecking the article or consensus. The accuracy of study reviews was checked by using two data extractors who independently collected data from each article, cross-checked data entry by random sampling, and resolved any inconsistencies.

Figure 1. Percent of titles/abstracts requiring (more...)

   Figure 1. Percent of titles/abstracts requiring resolution over time

Figure 1

Review of Results of Literature Review

Project staff provided the Technical Expert Panel with a summary report from the literature review that included the number of abstracts reviewed, articles retrieved, eligible studies identified, and reasons for exclusion. Their comments on the search strategy and analysis, which were discussed in the third conference call, aided in planning supplemental analyses.

Ranking of Influencing Factors

Table 14. Ranking of 41 influencing factors for analysis (more...)

Table 14. Ranking of 41 influencing factors for analysis by technical experts

Factor ID#	Influencing Factor	Total Rank
1	Age of child	79.0
23	Otitis prone	73.0
5	Attendance at day care center	54.5
9	Tobacco smoke exposure	48.0
25	Prior antibiotic use and when used	36.0
30	Presence of tube	35.5
7	Feeding mode -- bottle vs. breast	25.0
22	Middle ear effusion	24.0
21	Purulent otorrhea	21.0
12	Otalgia and severity	19.5
41	Cost of treatment	15.0
36	Setting (public, private, PPO, HMO, etc.)	15.0
24	Underlying viral infection	14.0
11	Season of the year	14.0
29	Atopy or allergy	13.0
19	Tympanic membrane inflammation	12.0
10	Ear infections in parents or siblings	12.0
18	Hearing loss	11.0
20	Retracted TM	10.0
14	Pulling of ear in an infant	9.0
16	Irritability	9.0
32	Parent/caretaker preference	9.0
17	Fever	8.5
3	Ethnicity/race	8.0
26	Concurrent use of non-antibiotics	6.0
15	Otorrhea	5.0
2	Gender	4.0
4	Presence of sibling(s)	4.0
8	Pacifier use	3.0
6	Sibling(s) in day care center	2.0
40	Type of method to monitor episode	1.0
35	Skill to diagnose	1.0
38	Frequency of monitoring of episode	1.0
39	Primary person monitoring episode	0.0
37	When monitoring is done during episode	0.0
13	Hearing deficit and severity	0.0
33	Parent/caretaker education	0.0
27	Prior hearing deficit	0.0
34	Type of examiner	0.0
28	Inability to express symptoms	0.0
31	Parent/caretaker availability	0.0

HMO=health maintenance organization; PPO=preferred provider organization; TM=tympanic membrane.

Preparation of Evidence Tables

An evidence table was prepared for each key question. Each evidence table provides a comprehensive tabular display of data abstracted from the literature, including the name of the first author, year of publication, language of study (if other than English), study design and its quality score, number of components addressed in the definition of AOM (according to the three preestablished criteria by the technical experts), when and where the study took place, inclusion and exclusion criteria, interventions compared, important influencing factors, sample sizes, outcome measures and their definitions, and study results. A total of five evidence tables were prepared and are included in the Evidence Tables section.

Supplemental Analysis

To address Key Question 1 on natural history, project staff analyzed data from the placebo or observational arm of randomized clinical trials, as well as data from nontrials (including prospective and retrospective single or comparative cohort studies). Incidence rates of each outcome indicator at each time point were estimated by pooling data from studies addressing the same patient population. For Key Question 3 on antibiotic use, only randomized controlled trials were used in meta-analysis.

Table 15. Comparisons for key question 3, antibiotics (more...)

Table 15. Comparisons for key question 3, antibiotics vs. no antibiotics

Question 3:The general principle agreed upon was to separate amoxicillin-clavulanate, penicillin G, penicillin V, erythromycin estolate, triple sulfonamide, and erythromycin estolate-triple sulfonamide from ampicillin/amoxicillin and each other. Penicillin G is oxidized in the stomach and is not well absorbed. Penicillin V does not cover Haemophilus Influenzaewell. Erythromycin estolate is quite different from the other antibiotics. Triple sulfonamide is no longer in common usage.

Comparison

Number of articles

ampicillin or amoxicillin vs. placebo amoxicillin-clavulanate vs. placebo penicillin G plus sulfisoxazole vs. placebo penicillin V vs. placebo erythromycin estolate vs. placebo triple sulfonamide vs. placebo erythromycin estolate-triple sulfonamide vs. placebo

6 1 1 2 1 1 1

The next major factor for consideration in the meta-analyses for each comparison was the outcome variables. A meta-analysis was performed on outcome measures that were considered clinically homogeneous, such as failure rate of antibiotics at 2-7 days. The assumption was that the 2-day failure rate reported in one study was clinically the same as the 7-day failure rate reported in another study.

Further subdivision of the articles within each comparison into subgroups for meta-analysis according to major influencing factors was considered but was not feasible due to the small number of articles. A sensitivity analysis was planned to evaluate the influence of age, otitis proneness, language, time, place, study-design quality, and appropriateness of definition of AOM. This was also not feasible due to the small number of articles in each comparison group.

Table 16. Comparisons for key question 4a: amoxicillin (more...)

Table 16. Comparisons for key question 4a: amoxicillin or trimethoprim-sulfamethoxazole vs. other anitibiotics

Question 4: The general principle agreed upon was to compare by individual antibiotic rather than by antibiotic class, spectrum, or pharmacokinetics.
penicillin amoxicillin-clavulanate cephalexin cephradine cefuroxime axetil cefaclor loracarbef cefixime ceftriaxone erythromycin estolate clarithromycin clindamycin penicillin V and sulfisoxazole triple sulfonamide penicillin G plus triple sulfonamide erythromycin ethylsuccinate-sulfisoxazole erythromycin ethylsuccinate-acetyl sulfafurazole oxytetracycline and procaine penicillin plus benzathine penicillin G injection plus sulfisoxazole	3 0 2 1 2 5 1 5 3 2 2 1 2 1 1 1 1 1
Trimethoprim-sulfamethoxazole vs.	Number of Articles
amoxicillin-clavulanate cephalexin cephradine cefuroxime axetil cefaclor loracarbef cefixime ceftriaxone erythromycin estolate erythromycin ethylsuccinate clarithromycin clindamycin penicillin-sulfisoxazole triple sulfonamide erythromycin ethylsuccinate-sulfisoxazole erythromycin ethylsuccinate-acetyl sulfafurazole	1 0 0 0 3 0 1 0 0 0 0 0 0 0 0 0

Table 17. Comparisons for key question 4e: short-term (more...)

Table 17. Comparisons for key question 4e: short-term vs. long-term antibiotic therapy

Question 4e:The general principle agreed upon was to compare by individual antibiotic stratified by therapy duration < 5 days vs. 5 days.

amoxicillin (<5d) vs. amoxicillin (7-10d) penicillin V (<5d) vs. penicillin V (7-10d) penicillin V (5d, either 25mg/kg/d or 50 mg/kg/d) vs. penicillin V (7-10d) benthazine penicillin G/procaine penicillin G/potassium penicillin G (Bicillin) (1 dose) vs. tetracycline (7-10d) benthazine penicillin G/procaine penicillin G/potassium penicillin G (Bicillin) (1 dose) vs. benthazine penicillin/procaine penicillin G/potassium penicillin G (Bicillin) (1 dose) plus triple sulonamide (7d) amoxicillin-clavulanate (5d, either 45 mg/kg/d or 80 mg/kg/d) vs. amoxicillin-clavulanate (7-10d, either 40 mg/kg/d or 45 mg/kg/d or 80 mg/kg/d) cefaclor (<5d) vs. cefaclor (7010d) cefaclor (5d) vs. amoxicillin (7-10d) cefaclor (5d) vs. cefaclor (7-10d) cefuroxime axetil (5d) vs. amoxicillin-clavulanate (7-10d) cefuroxime axetil (5d) vs. cefixime (7-10d) cefpodoxime proxetil (5d) vs. amoxicillin-clavulanate (7-10d) cefpodoxime proxetil (5d) vs. cefaclor (7-10d) cefpodoxime proxetil (5d) vs. cefixime (7-10d) cefprozil (5d) vs. cefprozil (7-10d) ceftibuten (5d) vs. ceftibuten (10d) ceftriaxone (1 dose) vs. amoxicillin (7-10d) ceftriaxone (1 dose) vs. amoxicillin-clavulanate (7-10d) ceftriaxone (1 dose) vs. cefaclor (7-10d) ceftriaxone (1 dose) vs. cefuroxime axetil (7-10d) ceftriaxone (1 dose) vs. trimethoprim-sulfamethoxazole (7-10d) azithromycin (<5d) vs. amoxicillin-clavulanate (7-10d, either 40 mg/kg/d or 45 mg/kg/d) azithromycin (<5d) vs. cefaclor (7-10d) azithromycin (<5d) vs. clarithromycin (7-10d) azithromycin (5d) vs. amoxicillin-clavulanate (7-10d)

3 1 1 with two 5-day arms 1 2 2 1 1 1 1 1 2 1 1 1 1 3 2 1 1 1 5 2 1 3

The DerSimonian and Laird random effects model (DerSimonian and Laird, 1986) was used to pool effect sizes across studies. This method produces a summary measure that is a weighted mean. It weights each study's measure by the inverse of the sum of the within-study variance and the between-study variance. This approach allows both sampling variation and between-study heterogeneity to affect the pooled estimate. Among the three effect measures -- rate difference, relative risk, and odds ratio -- the Technical Expert Panel and the project staff chose as most suitable the rate difference and its 95 percent confidence limits, which are presented both numerically and graphically. It should be noted that the absolute rate difference was used rather than the relative rate difference to measure the effect size throughout the report.

In addition to the pooled estimate, we report the Q statistic and p-value for the Chi-squared test of heterogeneity, which tests the null hypothesis that the individual study results are homogeneous (Laird and Mosteller, 1990). Because the test is known to lack power to detect heterogeneity, we protected against spurious conclusions resulting from combining clinically heterogeneous patients or treatments in two ways, regardless of the outcome of the Chi-squared test of heterogeneity. First was the use of random effects estimates that incorporate some between-study variance even if the Chi-squared test does not reject. Second, subgroup analyses and sensitivity analyses were planned to assess the impact of possible heterogeneity on the conclusions, but these were not feasible because of the small number of studies in each comparison group.

When necessary, we had planned to conduct meta-regression analysis for comparisons where multiple arm studies where used more than once. However, we did not have any multiple-arm studies that were used more than once in our comparisons using the same outcome indicator.

We also had planned to conduct a power analysis for those comparisons where the rate difference was 10 percent or higher but did not reach statistical significance. However, none of the comparisons that did not reach statistical significance was 10 percent or higher; thus, a power analysis was not performed.

To prepare for a meta-analysis for each comparison, data were abstracted from the evidence table (one meta-analysis for each outcome measure). The following data elements were entered in an SAS program to be converted into an SAS data set: the study ID number, author, year of publication, number of adverse outcomes in the experimental group, total number of patients in the experimental group, number of adverse outcomes in the control group, and total number of patients in the control group. An SAS macro software program that was developed by RAND statistical staff was used to perform all meta-analyses; the beta-test version of the software package "MetaGraphs" (1998, Belmont Research, Inc., 84 Sherman Street, Cambridge, MA 02140) was used for the graphing.

The following statistics were generated from the SAS macro program: (a) study-level statistics (incidence rate, relative risk, risk difference, number needed to treat [NNT], odds ratio, and their 95 percent CI); (b) crude estimates and their 95 percent CI for all studies combined; (c) fixed effects estimates and their 95 percent CI for all studies combined; (d) random effects estimates and their 95 percent CI based on the DerSimonian and Laird method for pooling study results and the Chi-squared test of homogeneity; and (e) weight for each study for both fixed effects model and random effects model in calculation of risk difference and relative risk.

To use MetaGraphs for graphing, the data were entered into ASCII files using the UltraEdit-32 software. Funnel plots were produced for the purpose of screening possible publication bias and the shrinkage plots were generated to display the effect size of each study and compare it against the overall model estimate, together with their 95 percent confidence limits.

Summary of Characteristics of Articles in Evidence Report

Of the 80 studies used in the evidence report, 40 (50 percent) were published in the 1990s, 25 (31 percent) in the 1980s, 8 (10 percent) in the 1970s, and 7 (9 percent) in the 1960s. The earliest article was published in 1964 and the latest in 1998. Forty-one studies (51 percent) were conducted in the United States. Six studies did not specify either a lower or upper age limit, although they all stated that the subjects were children (Bollag and Bollag-Albrecht, 1991; Froom, Culpepper, Grob, et al., 1990; Howie and Ploussard, 1972; Laxdal, Merida, and Jones, 1970; Tilyard, Dovey, and Walker, 1997; and Townsend, 1964). Of the six studies, four were observational cohort studies and two were randomized controlled trials. All addressed Key Questions 1 and 3. The otitis-prone status of the study population, the other influencing factor considered important by the Technical Expert Panel, was not reported in 64 percent (51/80) of the studies.

Of the 80 studies, 74 were randomized controlled trials. Their study-design quality was evaluated using the Jadad score (Jadad, Moore, Carroll, et al., 1996). Of the 74 studies, 6 (8 percent) had the highest score (5 on the Jadad scale); 12 (16 percent) scored 4; 21 (28 percent) scored 3; 26 (35 percent) scored 2; 8 (11 percent) scored 1; and 1 (1 percent) scored 0. Project staff evaluated the study-design quality of the six observational cohort studies by the presence or absence of eight components of study quality. Of the six studies, two studies addressed four of the eight components, one study addressed three, two studies addressed two, and one study addressed one. Of the 74 randomized controlled studies, 30 (41 percent) mentioned double-blinding, 52 (70 percent) described the characteristics of the dropouts, 28 (38 percent) used appropriate randomization methods, and 17 (23 percent) used appropriate blinding strategies.

For definition of AOM, none of the 80 studies used all three components (middle-ear effusion, rapid onset, and signs/symptoms of inflammation) of the proposed definition established by the Technical Expert Panel; 18 (22.5 percent) used two components; 34 (42.5 percent) used only one component; and 28 (35 percent) used none. Of the 80 studies, 42 (52.5 percent) included the middle ear effusion component, 2 (2.5 percent) included the rapid onset component, and 26 (32.5 percent) included the signs/symptoms of inflammation component.

With respect to the Technical Expert Panel's criterion that an "initial" episode of AOM be separated by at least 4 weeks from the end of an antibiotic course of treatment for the last episode of AOM, 21 of 74 (28 percent) randomized controlled trials and 6 of 6 (100 percent) cohort studies did not address this issue. Many of the other studies also did not address this issue: 9 of the 15 (60 percent) studies on natural history; 3 of the 9 (33 percent) studies on the effects of antibiotics on AOM; 12 of the 34 (35 percent) studies on ampicillin or amoxicillin or trimethoprim-sulfamethoxazole vs. other antibiotic, the single study on high-dose vs. standard-dose amoxicillin; and 5 of the 36 (14 percent) studies on short-duration vs. long-duration. Of the studies that addressed duration of the initial AOM episode, 17 met the criterion of 4 weeks or longer. None required that the last episode of AOM be documented as having cleared.

Identification of Peer Reviewers

In the letter to the 11 organizations requesting nominations for technical experts, nominations for peer reviewers also was requested. As of November 3, 1998, a total of 21 nominations for the Peer Review Panel were received. On November 16, 1998, a letter was mailed to the Ambulatory Pediatric Association (APA) and to the American Association of Health Plans (AAHP) requesting nomination of peer reviewers. Nine additional experts expressed interest to serve on the Peer Review Panel and submitted their curriculum vitae.

Experts in systematic reviews and meta-analysis were selected from a pool of experts maintained by the Southern California Evidence-based Practice Center who were not involved with this project.

Table 18. Peer Review Panel

Table 18. Peer Review Panel

Technical Expert	Area of Expertise	Affiliation/Location
Alfred O. Berg, M.D., M.P.H.	Family Medicine	University of Washington, Seattle, WA
Larry Culpepper, M.D.,M.P.H.	Family Medicine	Boston Medical Center, Boston, MA
Robert Ruben, M.D.	Otolaryngology	Montefiore Medical Center, Bronx, NY
Stanford T. Shulman, M.D.	Pediatrics, Infectious Disease	Children's Memorial Hospital, Chicago
Elizabeth Susan Hodgson, M.D.	Pediatrics, Ambulatory	St. Peter's Pediatric Faculty Group PracticeNew Brunswick, NJ
Jerome Klein, M.D.	Pediatrics, Infectious Disease	Boston Medical Center, Boston, MA
Lisa L. Hunter Ph.D.	Audiology	University of Minnesota, Minneapolis, MN
Terese Finitzo, Ph.D.	Audiology	Dallas, TX
Carol Rudy, M.S.N.,C.P.N.P.	Nurse Practitioner	Spokane, WA
Tracy Lieu, M.D.	Health Plan	Harvard Pilgrim Health Care, Boston, MA
Michael Siegel, M.D.	Health Plan	Prudential Health Care Plan, CA
Fran Goldfarb, M.A.	Consumer	Family Voices, Los Angeles, CA
Mark P. Haggard, Ph.D.	Epidemiology and Outcomes Methods	Institute for Hearing Research, UK
Anne G.M. Schilder, M.D., Ph.D.	Otolaryngology	University Medical Center Utrecht, TheNetherlands
Hanan S. Bell, Ph.D.	Methodology reviewer	Seattle, WA
Vic Hasselblad, Ph.D.	Meta-analysis reviewer	Duke University, Durham, NC
Katherine Harris, Ph.D.	Cost-analysis reviewer	RAND, Santa Monica, CA
Lynne Haverkos, M.D.	Government reviewer	NICHD
Robin Yurk, M.D., M.P.H.	Consumer	The Foundation for Accountability, Portland, OR
Peer Reviewer #20	Otolaryngologist
Peer Reviewer #21	Nurse practitioner
Peer Reviewer #22	Pediatric pharmacologist
Peer Reviewer #23	Pediatrician
Peer Reviewer #23	Methodology reviewer

Peer Review Process

Table 19. Instructions for reviewing draft evidence (more...)

Table 19. Instructions for reviewing draft evidence report

Enclosed is a draft evidence report on the management of acute otitis media. You may make your comments either directly on the draft evidence report, or on a separate sheet of paper. If you choose to record your comments on a separate piece of paper, please use the page and paragraph number to identify to which part of the report your comments pertain. We ask that you consider the following questions while you read this report. We realized that some of the questions may not pertain to your area of expertise. Please feel free to comment only on those that you feel most suited to answer.

1. Overall evaluation Is it clear what we did? You may agree or disagree with our methods, findings, or conclusions, but you should be able to understand what we did in order to produce this report.

2. Methodology Are the methods we used appropriate: (a) for identifying the key questions of interest from the panel of technical experts? (b) for searching and reviewing the identified literature? (c) for synthesizing the literature?

3. Evidence (a) Did we miss any crucial pieces of information in our literature search? (b) Does the evidence support the conclusions?

4. Utility Would you find this information to be useful if you had to develop clinical practice guidelines or medical review criteria for management of acute otitis media in children?

Natural History

To understand the true effect of an intervention on a clinical condition such as AOM, the natural history of that condition without intervention must be known. The timing of processes and intermediate outcomes, as well as the ultimate outcomes of acute otitis media (AOM) without intervention, must be known. Also important are influencing factors, either genetic, familial, cultural, or environmental, that may be difficult to control and that may affect the outcomes of AOM apart from, in addition to, or in interaction with the interventions of interest. The Technical Expert Panel decided to concentrate on age and otitis-prone state as such influencing factors to analyze. To properly estimate the marginal effects of an intervention on the outcomes of AOM, the natural history of AOM without intervention should be clearly understood.

The question we asked was: During an episode of uncomplicated AOM that is initially managed without any active intervention (pharmacologic or surgical) other than topical or systemic medications (that do not contain antibiotics) given for symptomatic relief (e. g., analgesics, antipyretics, antihistamines, decongestants, ear or nose drops), what proportion of children have the outcomes delineated in the Scope of the Evidence Report (Appendix E.4). To what degree are the above outcomes attributable to the influencing factors delineated in the Scope of the Evidence Report (Appendix E.4).

Previous Information Syntheses

Rosenfeld (1999a) assessed the natural history question in a previous information synthesis. As in the present analysis, Rosenfeld (1999a) evaluated children from epidemiologic studies, untreated control groups in randomized clinical trials, and cohort studies with an untreated group. Rosenfeld (1999a) assessed the same set of studies as in the present evidence-based analysis except for the inclusion of Fry (1958) and the absence of Tilyard, Dovey, and Walker (1997). We excluded Fry (1958) from our analysis because we could not extract the data on children from the data on adults in that study. Data were then pooled to calculate outcome estimates. Rosenfeld (1999a) estimated that 59 percent (3 studies; 315 children; 95 percent confidence limits, 53 percent and 65 percent) of children had relief of pain and fever within 24 hours of diagnosis. We note that one study used in this estimate (Burke, Bain, Robinson, et al., 1991) reported on resolution of pain and fever separately, and another study (Thalin, Densert, Larsson, et al., 1986) reported on pain and ear discharge, but not fever. To include the data in the estimate, an assumption must be made that those without pain did not have fever. Rosenfeld (1999a) estimated that 87 percent (5 studies; 808 children; 95 percent confidence limits, 84 percent and 89 percent) had relief of pain and fever within 2-3 days. Data from the Thalin, Densert, Larsson, et al. (1986) study was used in this estimate. Relief of pain and fever at 4-7 days was estimated to be 88 percent (5 studies; 503 children; 95 percent confidence limits, 85 percent and 91 percent). Again, data from the estimates in two studies -- Burke, Bain, Robinson, et al. (1991) and Thalin, Densert, Larsson, et al. (1986) -- were included in this estimate.

Finally, clinical resolution within 7-14 days of diagnosis was estimated at 73 percent (4 studies; 270 children; 95 percent confidence limits, 65 percent and 78 percent) based on the absence of all presenting signs and symptoms except middle ear effusion. The numerator used from one study (Halsted, Lepow, Balassanian, et al., 1968) appears to include success as judged by clinical or bacteriologic failure rather than clinical failure alone. In another study (Mygind, Meistrup-Larsen, Thomsen, et al., 1981) used in this estimate, the denominator is not clear from the report, and the initial denominator was used. Finally, Rosenfeld (1999a) found two cases of acute mastoiditis in 1,802 patients managed without antibiotics out of 2,368 patients, predominantly children, in six observational studies. (Although we mention some concerns about the data in these estimates, we also acknowledge that the direction and general magnitude of the estimates would not change with exclusion of that data. These same concerns apply to the use of these data in Rosenfeld (1999b), which is mentioned in the previous meta-analyses section Antibiotics vs. No Antibiotics in this chapter.)

Observation or Placebo

As proxies for natural history, we used the observational arm of cohort studies and the observational or placebo arm of randomized controlled trials of AOM. The treatment arm of such trials could be of any modality, antibiotic, myringotomy, etc., as long as the control arm consisted solely of observation or placebo.

Outcomes

The outcomes reported for these nine randomized controlled trials and six cohort studies share some commonalities but also many differences (Evidence Table 1). Outcome descriptors often relate to success or failure, pain or otalgia, fever, otoscopic findings, middle ear exudates or otorrhea, middle ear effusion, tympanometry results, relapse, recurrence, need for further intervention such as placement of pressure equalizing tubes, and suppurative complications such as mastoiditis.

Table 20. Key question 1: selected outcomes

Table 20. Key question 1: selected outcomes

Author/Year	Outcome(s) Measured 1	Time(s) Measured
Randomized Controlled Trials
Halsted/1968	early improvement	24-72 hours
	resolved	14-18 days
	recurrence	1 year
Laxdal/1970	excellent	7 days
	good	14 days
	fair	21 days
	poor	21 days
	failure	7 days
Howie/1972	improvement/failure	2-7 days
	fever	2-7 days
	nonsterile exduate	2-7 days
Mygind/1981	satisfactory course of acute phase	7 days
	symptom free	2 days
	otorrhea	>5 days
	middle ear effusion	1 week, 1 month, 3 months
	contralateral otitis media	1 week
	relapse	1 week, 1-3 months
van Buchem/1981	pain	>24 hours, >7 days
	abnormal otoscopy	>7 days, >14 days
	otorrhea	>24 hours, >7 days
	relapse	6 months
Thalin/1986	complete resolution	30 days
	failure	30 days
	relapse	30 days
Appelman/1991	irregular course	>3 days
	tympanometry	1 month
Burke/1991	failure	7 days
	pain	2 days, 5-7 days
	fever	2 days, 5-7 days
	abnormal otoscopy	>7 days
	middle ear effusion	1 month, 3 months
	recurrence	1 year
Kaleida/1991	treatment failure	24-48 hours, 1 year
	middle ear effusion	2 weeks, 6 weeks
	recurrence	2-6 weeks
Cohort Studies
Townsend/1964	complication	1 year
Ostfeld/1988	otorrhea	24 months
	chronic middle ear effusion	24 months
	recurrence	24 months
	pressure equalizing tube placement	24 months
	mastoiditis	24 months
Froom/1990	recovery	2 months
Bollag/1991	mastoiditis	2 months
	hearing deficit	2 weeks
Tilyard/1997	failure	30 days
van Buchem/1985	cure	14 days
	severe course	3-4 days
	persistent discharge	14 days

1 See Table 21 for definitions for outcome measures related to success or failure of treatment

Table 21. Key question 1. Terms and definitions: success (more...)

Table 21. Key question 1. Terms and definitions: success and failure

Author/Year	Term(s)	Definition(s)	Time Measured
SUCCESS
Randomized Controlled Trials
Halsted/1968	early improvement	decreased symptoms	24-72 hours
	resolved	asymptomatic and normal tympanic membrane appearance	14-18 days
Laxdal/1970	excellent	no evidence of middle ear inflammation	7 days
	good	signs of otitis media resolved	14 days
	gair	signs of otitis media resolved	21 days
Howie/1972	improvement	absence of exudate	2-7 days
Mygind/1981	satisfactory course of acute phase	(1) not crying from pain after 1day (2) no analgesics after 1day (3) no otitis symptoms after 2days (4) otorrhea <6 days (5) no contralateral otitis first week	1 week
Thalin/1986	complete resolution	not clear if refers to "satisfactory" defined as normal otomicroscopy and normal audiogram and/or tympanogram	30 days
Cohort Studies
Townsend/1964	recovery	(1) tympanic membrane normal (2) pneumatic otoscopy normal (3) audiogram and hearing per parents and teacher normal	1 year
Van Buchem/1985	cure	not severe course and without persistent discharge	14 days
Froom/1990	recovery	based on pain, ear drainage, hearing problem, or other	2 months
FAILURE
Randomized Controlled Trials
Laxdal/1970	poor	persistent signs of minimal infection	21 days
	failure	no improvement or deterioration	7 days
Howie/1972	failure	presence of xudates	2-7 days
Thalin/1986	failure	Remaining nonnegligible symptoms (pain, fever, etc.) or insufficient resolution of infectious signs during the medical treatment period	7 days
Appelman/1991	irregular course	otalgia or > 38 degrees Centigrade	>3 days
Burke/1991	failure	second-line antibiotic required, presumably due to nonresolution or recurrence of symptoms	1 week
Kaleida/1991	initial treatment failure	"severe" criterion present >24 hours or >38 degrees Centigrade oral or 38.5 degrees Centigrade rectal or >6 otalgia score	24-48 hours
	ultimate treatment failure	cumulative >180 days middle ear effusion in same ear or > four severe AOM episodes or > five any AOM or new otitis media with effusion within 6 months or fourth myringotomy in 6 months or fifth in 12 months or suppurative complication or cholesteatoma or an allergic reaction to penicillin	1 year
Cohort Studies
Townsend/1964	complication	purulent progression of disease or serous otitis media	1 year
Van Buchem/1985	Severe course	Persistent high temperature or severe pain	3-4 days
	persistent discharge	Persistent discharge	14 days
Froom/1990	no or uncertain recovery	based on pain, ear drainage, hearing problem, or other	2 months
Tilyard/1997	failure	return <10 days with same problem or change antibiotic within 30 days	30 days

The differences in the actual outcomes measured were due both to the outcome selection, measurement description, and the time of measurement relative to the start of the study (Table 20). Forty-eight outcomes are listed in Table 20. In those cases where outcomes seemed to be shared, the description of those outcomes was not always the same. In addition, in those cases where the outcomes seemed to have the descriptors of the same meaning such as success or failure but not the same time of measurement, a reading of the actual definition of those descriptors revealed differences (Table 21). Terms used for the meaning success included "early improvement," "resolved," "excellent" or "good," "improvement," "satisfactory course of acute phase," "recovery," and "cure." In some studies, the reviewer had to assume that the opposite of failure was success. Terms used for failure included "failure," "poor," "irregular course," "complication," and "no" or "uncertain recovery." Some definitions of success or failure were quite detailed (Kaleida, Casselbrant, Rockette, et al., 1991; Mygind, Meistrup-Larsen, Thomsen, et al., 1981). For the most part, the definitions of success or failure were nonspecific and relied solely on: symptoms; signs; signs and symptoms; signs, symptoms, and tympanic membrane appearance, specifically; or signs, symptoms, tympanic membrane appearance, and absence of middle ear effusion, specifically. Of the definitions of success and failure listed in Table 21, each is unique to the study in which it is described.

Results

This report concentrates on failure of clinical resolution, pain, fever, and middle ear effusion in the natural history of AOM, mastoiditis, and other suppurative complications. The estimates for the other outcomes noted in Table 20 may be found in Evidence Table 1.

Table 22. Key question 1: failure1 rate in the placebo (more...)

Table 22. Key question 1: failure1 rate in the placebo or observational group

Author/Year	Time Measured/Influencing Factor	Failure Rate
Randomized Controlled Trials
Halsted/1968	24-72 hours	7/27 (26%)
	14-18 days	0/21 (0%)
Laxdal/1970	7 days	18/48 (38%)
	0-3 years old	46% (denominator not reported)
	3-6 years old	38% denominator not reported)
	6-9 years old	25% (denominator not reported)
	9-14 years old	0% (denominator not reported)
Howie/1972	2-7 days	92/116 (79%)
Mygind/1981	7 days	31% (denominator not reported)
Thalin/1986	7 days	12 (count only, denominator unknown)
Appelman/1991	>3 days	10/54 (18%)
	<2 years old	7/12 (58%)
	>2 years old	3/42 (7%)
Burke/1991	7 days	17/118 (14%)
	3-5 years old	10/66 (15%)
	6-9 years old	7/52 (13%)
	<2 prior AOM 2	6/55 (11%)
	>2 prior AOM	7/48 (15%)
Kaleida/1991	24-48 hours	38/492 (7.7%)
	<2 years age	25/254 (9.8%)
	>2 years age	13/238 (5.5%)
	1 year	46/170 (27.1%)
	<2 years age	30/82 (36.6%)
	>2 years age	16/88 (18.2%)
Cohort Studies
Townsend/1964	1 year	11/189 (5.8%)
Van Buchem/1985	14 days	20/465 (4.3%)
Froom/1990	2 months	40/4 19 (9.5%)
	0-12 months old	5/39 (12.8%)
	13 30 months old	7/69 (10.1%)
	>31 months old	28/311 (9.0%)
Tilyard/1997	1 month	8/74 (10.8%)
	<2 years old	4/29 (14%)
	2 5 years old	4/22 (18%)
	6-15 years old	0/12 (0%)
	>15 years old	0/11 (0%)

1

See definition of success and failure in Table 21

2

AOM=acute otitis media

For those randomized controlled trials that evaluate some descriptor of success or failure in a placebo or observational group, the following statements may be made while cognizant of the differences in definition of success or failure as noted above and in Table 21. (See Table 22 and Evidence Table 1.) Early failure in a placebo group monitored at 24-72 hours was reported as 7.7 percent by Kaleida, Casselbrant, Rockette, et al. (1991) and 26 percent by Halsted, Lepow,

Balassanian, et al. (1968). Appelman, Claessen, Touw-Otten, et al. (1991) reported a failure rate of 18 percent measured after 3 days. Failure monitored at 7 days was reported as 8 percent by Thalin, Densert, Larsson, et al. (1986), 14 percent by Burke, Bain, Robinson, et al. (1991), and 38 percent by Laxdal, Merida, and Jones (1970). An unsatisfactory clinical course of the acute phase (7 days) was reported for 31 percent of the placebo group by Mygind, Meistrup-Larsen, Thomsen, et al. (1981). Thalin, Densert, Larsson, et al. (1986) reported 12 subjects in their placebo group failing therapy at 7 days but did not specify a denominator. According to the definition of failure by Howie and Ploussard (1972) as the presence of middle ear exudate at 2-7 days, that group reported a rate of failure of 79 percent in the placebo group and 57 percent in the antibiotic group. Failure in the placebo group at 14-18 days was 0 percent (Halsted, Lepow, Balassanian, et al. 1968) and failure for the placebo group at 1 year was 27.1 percent (Kaleida, Casselbrant, Rockette, et al. 1991). The cohort study (Tilyard, Dovey, and Walker 1997) reported a failure rate of 10.8 percent in their group not treated with antibiotics at 1-month measurement, (Froom, Culpepper, Grob, et al. 1990), a failure rate of 9.5 percent at 2-month measurement, and a failure rate of 5.8 percent at 1-year measurement (Townsend 1964).

Pooling the failure data at 1-7 days from the randomized controlled studies in cases where denominators were certain, we calculated a failure rate of 18.9 percent (five studies; 739 children; 95 percent CI, 9.9 percent and 28.0 percent) for those not initially receiving antibiotics (Appelman, Claessen, Touw-Otten, et al., 1991; Burke, Bain, Robinson, et al., 1991; Halsted, Lepow, Balassanian, et al., 1968; Kaleida, Casselbrant, Rockette, et al., 1991; Laxdal, Merida, and Jones, 1970). In other words, 81.1 percent of these children with AOM not initially treated with antibiotics would have clinical resolution within 1-7 days of presentation. The pooled estimate for the failure rate at 3-7 days was 22.2 percent (three studies; 220 children; 95 percent CI, 10.1 percent and 34.3 percent) (Appelman, Claessen, Touw-Otten, et al., 1991; Burke, Bain, Robinson, et al., 1991; Laxdal, Merida, and Jones, 1970). During this time, 77.8 percent of children would have clinical resolution. (See Evidence Table 1 for data on individual studies.) We did not pool the data at 1-3 days because only two studies were available (Halsted, Lepow, Balassanian, et al., 1968; Kaleida, Casselbrant, Rockette, et al., 1991).

Table 23. Key question 1: presence of pain and/or fever (more...)

Table 23. Key question 1: presence of pain and/or fever in the placebo or observational group

Author/Year	Time Measured/Influencing Factor	Number of patients with Symptom
Randomized Controlled Trials
Pain
van Buchem/1981	>24 hours	11/49 (28%)
	>7 days	4/38 (10%)
Burke/1991	2 days	56/117 (48%)
	5-7 days	29/114 (25%)
Fever
Howie/1972	2-7 days	0/116 (0%)
Burke/1991	2 days	19/93 (20%)
	5-7 days	8/70 (11%)
Pain or Fever
Mygind/1981	2 days	38% (denominator not reported)
van Buchem/1981	>24 hours	11/40 (28%)
	>7 days	4/38 (10%)
Thalin/1986	7 days	12 (count only, denominator unknown)
Appelman/1991	>3 days	10/54 (18%)
	<2 years old	7/12 (58%)
	>2 years old	3/42 (7%)
Kaleida/1991	24-48 hours	38/492 (7.7%)
	<2 years old	25/254 (9.8%)
	>2 years old	13/238 (5.5%)

The presence of pain or otalgia was a common outcome reported in the randomized controlled studies (Table 23). Twenty-eight percent of the placebo group had pain at 24 hours and 10 percent had pain at 7 days (Van Buchem, Dunk, van't Hof, et al. 1981). 48 percent had pain at 2 days and 25 percent had pain at 5-7 days (Burke, Bain, Robinson, et al. 1991). Initial treatment failure was based on presence of fever or pain (Kaleida, Casselbrant, Rockette, et al. 1991). This study found 7.7 percent of children with nonsevere AOM in their placebo group had fever or pain at 24-48 hours. Howie and Ploussard (1972) reported that none of the placebo group had fever at 2-7 days; Burke, Bain, Robinson, et al. (1991) reported that 20 percent of the placebo group had fever at 2 days and 11 percent had fever at 5-7 days. Mygind, Meistrup-Larsen, Thomsen, et al. (1981) reported that 38 percent of the placebo group had pain and fever at 2 days. Twenty-eight percent of the placebo group had pain and fever at 24 hours, and 10 percent had pain and fever at 7 days (van Buchem, Dunk and van't Hof 1981); 18 percent had pain and fever at 3 days (Appelman, Claessen, Touw-Otten, et al. 1991). Despite the variability of the studies, it appears that children with AOM not treated with antibiotics experience a significant decrease in pain and fever during the first week of the clinical condition.

Table 24. Key question 1: presence of middle ear effusion (more...)

Table 24. Key question 1: presence of middle ear effusion in the placebo or observational group

Author/Year	Time Measured/Influencing Factor	Number of Patientst with Symptom
Randomized Controlled Trials
Howie/1972	2-7 days	92/116 (79%)
Mygind/1981	1 week	50% (denominator not reported)
	1 month	32% (denominator not reported)
	3 months	24% (denominator not reported)
Thalin/1985	30 days	41% (denominator not reported)
Burke/1991	1 month	41/116 (35%)
	3 months	31/111 (28%)
Kaleida/1991	2 weeks	255/408 (63%)
	<2 years old	143/209 (68%)
	>2 years old	112/1 19 (56%)
	6 weeks	169/328 (52%)
	<2 years old	99/175 (57%)
	>2 years old	70/153 (46%)
Cohort Study
Ostfeld/1987	24 months	42% (denominator not reported)

The presence of middle ear effusion also was reported in many of the studies (Table 24). Asymptomatic middle ear effusion is an expected part of the clinical course of AOM (Rosenfeld, 1996). Persistence of middle ear effusion beyond 3 months, however, is not considered normal. Fifty percent of children in the placebo group had middle ear effusion at 1 week in the study of Mygind, Meistrup-Larsen, Thomsen, et al. 1981); at 1 month and 3 months, 32 percent and 24 percent of patients, respectively, had middle ear effusion. In Burke, Bain, Robinson, et al. (1991), middle ear effusion occurred in 35 percent of the placebo group at 1 month, and 28 percent of these patients at 3 months. In Kaleida, Casselbrant, Rockette, et al. (1991), 63 percent had a middle ear effusion at 2 weeks, and 52 percent at 6 weeks. A cohort study found 42 percent with chronic middle ear effusion at 24-month followup (Ostfeld, Segal, Kaufstein, et al. 1988). In general, these studies indicate that middle ear effusion is quite common after AOM.

Several of the studies reported estimates of these outcome measures by age group or otitis-prone state (Evidence Table 1, Tables 22, 23, and 24). Laxdal, Merida, and Jones (1970) reported that 46 percent of children 0-3 years of age failed to improve by 7 days in the observational group, while none of the children 9-14 years of age in the observational group failed to improve. In Burke, Bain, Robinson, et al. (1991), 15 percent of children 3-5 years of age had no improvement of symptoms by 1 week, while 13 percent of those 6-9 years of age had no symptomatic improvement. Kaleida, Casselbrant, Rockette, et al. (1991) reported that in patients in the placebo group, the initial failure rate was 9.8 percent in children younger than 2 years of age compared with 5.5 percent of children 2 years of age or older. Kaleida, Casselbrant, Rockette, et al. (1991) also found that at 1-year followup, 37 percent of those younger than 2 years of age in the placebo group and 18 percent of those 2 years of age or older met their criteria for failure. Burke, Bain, Robinson, et al. (1991) also found that children in the placebo group with one or two previous episodes of AOM had an 11 percent rate of symptomatic failure compared with 15 percent of children with more than two previous episodes of AOM. In the prospective questionnaire study, 12.8 percent of those aged 0-12 months who were not on antibiotics failed to recover by 2 months, whereas 10.1 percent of those 13-30 months of age and 9.0 percent of those older than 30 months failed to recover (Froom, Culpepper, Grob, et al. 1990). In the retrospective cohort study (Tilyard, Dovey, and Walker 1997), 14 percent of those younger than 2 years of age without antibiotic treatment failed to recover, as did 18 percent of those 2-5 years of age, and 0% of those 6 years of age and older (Table 22). A significant difference was seen for the resolution of pain and fever as reported by Appelman, Claessen, Touw-Otten, et al. (1991): 58 percent with pain or fever after 3 days in those younger than 2 years of age and 7 percent in those 2 years of age or older (Table 23). As shown in Table 24, 57 percent of children younger than 2 years of age in the placebo group of Kaleida, Casselbrant, Rockette, et al. (1991) had a middle ear effusion at 6 weeks, whereas that was true of 46 percent of those 2 years of age or older.

Table 25. Mastoiditis/suppurative complications and (more...)

Table 25. Mastoiditis/suppurative complications and AOM: no antibiotic and associated antibiotic arms from cohort and randomized controlled studies¹

Article/Year	Time of Study; Locale	Antibiotic	Age	Otitis-Prone Status	Denominator	Mastoiditis	Other Suppurative Complication
Cohort Studies 2
Bollag/1991	2/86-2/88; Switzerland	none	mean 51.0 m 3 (SEM 35.9 m)	not addressed	153	0	0 (meningitis)
	3/88-2/89; Switzerland	none	mean 68.6 m 3 (SEM 38.9 m)	not addressed	56	0	0(meningitis)
Ostfeld/1988	3/81-2/83; Locale not noted	none	<6 m=110 4 pts 7-12 m=166pts 13-24m=75pts >24 m=46pts	not addressed	397	2	0 (meningitis)
		antibiotics "for associated medical conditions"	<6 m=87 4 pts 7-12 m=116pts 13-24 m=67pts >24 m=26pts	not addressed	196	3	0 (meningitis)
Randomized Controlled Studies
Burke/1991	10/86-4/87, 10/87-4/88, 10/88-4/89; Southampton, Bristol, and Portsmouth, England	placebo	3-5 y=66pts 6-9 y=52pts	previous AOM: 0-2 55 >2 48	111 5	0	0 (meningitis)
	10/86-4/87, 10/87-4/88, 10/88-4/89; Southampton, Bristol, and Portsmouth, England	amoxicillin	>3 y < 10 y	not specified for this group 6	110 5	0	0 (meningitis)
Kaleida/1991	5/81-8/85; Pittsburgh, U.S.A.	placebo	<2 y=136pts 2-5 y=107pts 6-12 y=30pts	previous AOM in last year: 0/98 1-2/120 >2/55	170 from the nonsevere group	0 7	0 7
	5/81-8/85; Pittsburgh, U.S.A.	amoxicillin	<2 y=133pts 2-5 y=108pts 6-12 y=22pts	previous AOM in last year: 0 105 1-2 100 >2 58	169 from the nonsevere group	1 7	0 8
Thalin/1986	7/84-6/85; Halstad, Sweden	placebo	2-15 y 9	not specified for this group 10	159 11	0	0
	7/84-6/85; Halstad, Sweden	penicillin VK	2-15 y 9	not specified for this roup 10	158 11	0	0
van Buchem/1981	1/79-3/79; Tilburg, The Netherlands	placebo	2-12 y 12	not addressed	40 13	0	0
	1/79-3/79; Tilburg, The Netherlands	amoxicillin	2-12 y 12	not addressed	47 13	0	0
Mygind/1981	11/77-4/78; Copenhagen, Denmark	placebo	mean 4.1 y	not addressed	77	0	0
	11/77-4/78; Copenhagen, Denmark	penicillin VK	mean 3.7 y	not addressed	72 14	1 14	0
Laxdal/1970	1/66-9/68; Saskatchewan, Canada	none	<14 y 15	not addressed	48	0 16	0 16
	1/66-9/68; Saskatchewan, Canada	penicillin G or ampicillin	<14 y 15	not addressed	94	0 16	0 16

1

Mixed-treatment arms are not reported in this table.

2

In addition to these two cohort studies, van Buchem, Peeters, van't Hof (1985) mention two cases of mastoiditis in a 17-month study period for which the authors estimate a denominator of 4,860 children with AOM based on a 3-month convenience sample.

3

Lower age limit was not specified.

4

Lower and upper age limits were not specified. Counts are estimated from Figure 1 in Ostfeld, Segal, Kaufstein, et al. (1987).

5

This is the denominator for the 3-month follow up.

6

Of the total group (placebo plus amoxicillin), 53 percent of patients had 0-2 previous episodes of AOM, and 47 percent had >2 previous episodes of AOM.

7

These counts are by implication because no episodes of mastoiditis were reported in the "Complications" section of "RESULTS" nor any other suppurative complications in the placebo group.

8

It is not clear from the article if this child with unilateral paresis of the marginal mandibular nerve was in the nonsevere or severe group. She was on antimicrobial treatment.

9

Of the total group (placebo and penicillin VK), 51 episodes of AOM involved children 2 years old, and 266 episodes of AOM involved older children.

10

Of the total group (placebo and penicillin VK), 23 percent of the children had their first episode of AOM and as many as 27 percent had suffered from AOM more than 10 times.

11

The article does not indicate the denominators for serious complications. These were the denominators at the start of the study.

12

Of the total group (placebo and amoxicillin), 18 were 2 years old, and 153 were older.

13

Reference to "Complications" is reported for an observation period of 2 years, but a 2-year denominator is not reported. These were the denominators at the start of the study.

14

Mastoiditis occurred sometime after the "immediate period after treatment", so the choice of denominator was unclear. The denominator "72" was at the start of study. Denominators at 1 month and 3 months were not reported by study group.

Lower-age limit was not specified. Of the total group (no antibiotic and antibiotic), 50 percent were < 3 years old, 25 percent were 3-6 years old, 18 percent were 6-9 years old, and 7 percent were > 9 years old.

16

These counts are by implication because authors report "no serious complications."

The following articles of no antibiotic treatment of AOM do not report mastoiditis or suppurative complications as outcomes:

1. Cohort studies: Froom, Culpepper, Grob, et al. (1990); Tilyard, Dovey, and Walker (1997); Townsend (1964) [The author mentions 1 case of mastoiditis in a child in the symptomatic treatment plus myringotomy group that included 2 cases but does not comment on the presence or absence of suppurative complications in the other treatment groups, including the 218 cases that were treated symptomatically.]

2. Randomized controlled studies: Appelman, Claessen, Touw-Otten, et al. (1991); Halsted, Lepow, Balassanian, et al. (1968); Howie and Ploussard (1972)

Mastoiditis and other suppurative complications are important outcomes to consider. Clinical failure based on signs and symptoms -- including pain, fever, and middle ear effusion -- are important, but serious complications would have greater weight in the treatment decision of AOM. Table 25 summarizes the information on mastoiditis and other suppurative complications found in the literature retrieved on natural history. Of the six cohort studies, two report specifically on the occurrence of suppurative complications in their study populations. Bollag and Bollag-Albrecht (1991) report no episodes of mastoiditis or meningitis in the two small cohorts that they followed. Ostfeld, Segal, Kaufstein, et al. (1988) noted two cases of mastoiditis in the 397 patients not treated with antibiotics and three cases of mastoiditis in the 296 patients treated with antibiotics; none of their patients had meningitis. In addition to these two cohort studies, van Buchem, Peeters, and van't Hof (1985) mention two cases of mastoiditis in a 17-month period for which they estimated a denominator of 4,860 children with AOM based on a 3-month convenience sample. Townsend (1964) mentions one case of mastoiditis in the symptomatic treatment plus myringotomy group of his study, which included two patients, but he does not comment explicitly on suppurative complications in the other treatment groups, including the 218 who were treated symptomatically.

The six randomized controlled studies listed in Table 25 have smaller sample sizes. No cases of mastoiditis or other suppurative complications were reported in the placebo or antibiotic arms of these studies. Of interest, two of these studies (Kaleida, Casselbrant, Rockette, et al., 1991; Ostfeld, Segal, Kaufstein, et al., 1988) included large proportions of children younger than 2 years of age. The footnotes to Table 25 detail some of the problems with these studies in addition to their small sample sizes and nonpopulation-based sample selection.

Despite these cautions, it appears that placebo or no antibiotic treatment in the setting of studies with close followup of patients is not necessarily associated with a high risk of mastoiditis or suppurative complications.

The natural history results must be considered in light of the role of antibiotic intervention in the "no antibiotic" treatment groups. Although the placebo or observational groups in all the studies were not treated initially with antibiotics, all nine of the randomized controlled trials and two of the six cohort studies mention clinical circumstances that would allow the investigator to administer antibiotics to children in the placebo or observational groups. Persistent symptoms or complications were common reasons for giving antibiotics to children in the placebo or observational groups. The majority of these studies had close followup of children, which allowed for timely clinical assessment. Most of the studies did not state explicitly how many children received antibiotics from these groups; however, in view of the criteria given for administering antibiotics, the numbers were probably small. (The only exception would be the Howie and Ploussard (1972) study where the investigators decided arbitrarily to give a majority of children in the placebo group amoxicillin starting at 2-5 days.) In most cases, the children in the placebo or observational group who received antibiotics were not removed from the study; in any case, the administration of antibiotics would not have had an effect on the early outcomes. This discussion also applies to the results on antibiotics vs. no antibiotics.

Antibiotics vs. No Antibiotics

We asked the following: Are antibiotics effective (in terms of statistical significance and magnitude of absolute clinical benefit above and beyond placebo/observational/no treatment/natural history) in the initial treatment of uncomplicated AOM with respect to the outcomes delineated in the Scope of the Evidence Report (Appendix E.4)? When antibiotics are used in the initial treatment of uncomplicated AOM, which of the influencing factors delineated in the Scope of the Evidence Report (Appendix E.4) are associated with better outcomes when compared with placebo/observational/no treatment/natural history?

Before asking which antibiotics are effective in treating AOM, we first ask if antibiotics have any marginal benefit in the treatment of AOM compared with natural history. In the scientific setting, the randomized controlled trial is the ideal method to answer this question by comparing the effect of antibiotics to that of placebo or observational treatment as proxies for natural history.

Previous Meta-Analyses

Del Mar, Glasziou, and Hayem (1997); Rosenfeld, Vertrees, Carr, et al. (1994); and Rosenfeld (1999b) have conducted quantitative syntheses assessing the marginal benefit of antibiotics in the treatment of AOM compared with placebo or observational therapy. Del Mar, Glasziou, and Hayem (1997) and Rosenfeld, Vertrees, Carr, et al. (1994) meet the validity criteria for systematic reviews of Oxman, Cook, and Guyatt (1994). Rosenfeld's (1999b) methodology is not described in detail, but it follows the guidelines of Oxman, Cook, and Guyatt (1994). A quantitative synthesis by Damoiseaux, van Balen, Hoes, et al. (1998) attempted to evaluate the effectiveness of antibiotic treatment of AOM in children younger than 2 years old, but examination of the report reveals that data from one of the four studies (Halsted, Lepow, Balassanian, et al., 1968) included children older than 2 years; in another study (Englehard, Cohen, Strauss, et al., 1989), the placebo group is treated with myringotomy as well. For these reasons, the results of the Damoiseaux, van Balen, Hoes, et al. (1998) quantitative synthesis will not be reported.

Table 26. Summary of Del Mar, Glasziou, and Hayem (1997) meta-analysis (more...)

Table 26. Summary of Del Mar, Glasziou, and Hayem (1997) meta-analysis

Title	Literature Sources	Paper Trail	Questions	Results	Author Conclusions/Reviewer Comments
Title	Literature Sources	Paper Trail	Questions	Results	Author Conclusions/Reviewer Comments
Are antibiotics indicated as initial treatment for children with acute otitis media? A meta-analysis.	Index Medicus manual search, 1958- 1965; MEDLINE, 1966-8/94; Current Contents, 1966-8/94; References of all retrieved articles; Not noted if non-English language articles were included.	Does not explain in detail the process of article selection and does not provide much descriptive information about the articles	1). Do antibiotics in the treatment of AOM affect the resolution of pain: (a) at 24 hours? (b) at 2-7 days? 2). Do antibiotics in the treatment of AOM affect the incidence of tympanic membrane perforation? 3). Do antibiotics in the treatment of AOM affect the incidence of vomiting, diarrhea, or rash? 4). Do antibiotics in the treatment of AOM affect the resolution of deafness: (a) at 1 month? (b) at 3 months? 5). Do antibiotics in the treatment of AOM affect the incidence of contralteral AOM? 6). Do antibiotics in the treatment of AOM affect the incidence of recurrent AOM?	1. (a) no, 124/318 vs. 125/315 1. (b) yes, less pain, 90/929 vs. 131/914, RD 41% of who still had pain at 24 hours (14%-60%) 2) possibly fewer perforations but not statistically significant, 7/190 vs. 14/191 3) yes, more vomiting, diarrhea, or rash, 57/345 vs. 38/353, OR 1.97(1. 19-3.25) 4) a) no, 64/183 vs. 66/193 4) b) possibly fewer deaf but not statistically significant, 38/182 vs. 49/188 5) yes, less contralateral AOM, 35/329 vs. 56/337, RD 43% (9%-64%) 6) no, 187/864 vs. 175/804	"Many doctors and their patients may be disinclined to use antibiotics at first presentation of otitis media for so little benefit. Others may regard any potential benefit as worth the inconvenience of purchasing and administering the drugs and the risk of their (usually) minor complications." "...17 children must be treated at first presentation to prevent one child experiencing pain after 2-7 days..." (Pain resolves spontaneously within 24 hours in 60% and within 2-7 days in 86% without antibiotics.)
				(Results are graphically displayed as ORs and in tabular counts with associated statistics, but the actual Ors and confidence intervals are not all reported. Two results are reported as RDs.)	The authors acknowledge the results may not be generalizable to "Third World communities". The authors suggest research to identify subgroups of children who would benefit from antibiotics. Very little discussion of the study's weaknesses by the authors. The authors do not calculate the "file drawer" effect estimate.
				RD-rate difference	Published criticisms: 1) incorrect use of OR to calculate number needed to treat; using RD, the NNT is 23 (15-56) 2) diagnostic criteria that were not reported varied widely among the studies
				OR-odds ratio	Possible RCTs Published Since Analysis: 11?

Inclusion Criteria	Quality Control
RCTs, antimicrobial drugs vs. placebo control	Yes (per Chalmers, Adams, Dickersin, et al., 1990) scores of 11, 10,10, 9, 8, 5, 5, and 2 (0-11 possible)
Exclusion Criteria	Articles/Patients
none noted	6/1833 (The number of articles used for each analysis ranged from two to six)

Table 27. Summary of Rosenfeld, Vertrees, Carr, et (more...)

Table 27. Summary of Rosenfeld, Vertrees, Carr, et al. (1994) meta-analysis

Title	Literature Sources	Paper Trail	Questions	Results	Author(s)'s Conclusions/Reviewer(s)'s Comments
Title	Literature Sources	Paper Trail	Questions	Results	Author(s)'s Conclusions/Reviewer(s)'s Comments
Clinical efficacy of antimicrobial drugs for acute otitis media: Meta-analysis of 5400 children from thirty-three randomized trials1	MEDLINE, 1/66-6/92 inclusive (English and foreign); Current Contents/Life Sciences, 3 months through 6/29/92; bibliographies of textbooks, review articles, source articles, symposium publications; three trials included in the analysis were from non-English language reports	Includes details on numbers of articles at each stage of selection and some descriptive information on included articles	1) Are antimicrobial drugs more efficacious than placebo or no drug for the clinical resolution of AOM? 2) What is the efficacy of various antimicrobials vs. amoxicillin or ampicillin for the same end point? 3) How does study design affect observed rates of AOM control? 4) How do patient characteristics affect observed rates of AOM control?	1) Comparisons with placebo/no drug: vs. PCN RD 15.7(4.7-26.7) vs. aminoPCN RD 12.9(6.8-19.0) vs. any ABX RD 13.7(8.2-19.2) 2) a) Comparisons with aminopenicillins: AMP vs. PCN RD −6.8(−15.2-1.5) AMP vs. PCN/SSX RD 0.9(−7.6-9.4) AminoPCN vs. ERY RD 3.1(−3.9-10.2) AminoPCN vs. TMP-SMX 0.2(−8.8-9.2) AMX vs. CFC RD 6.4(−10.2-22.9) AMX vs. CFX RD −3.9 (−10.4-2.6) 2) b) Comparisons with cefaclor: vs. ERY/SSX RD 7.0(−6.5-20.4) vs. AMX/CLV RD 2.8(−1.3-6.8) vs. CFX RD 1.2(−2.4-4.7) 3) RD was unrelated to year, publication year, outcome assessment day, requirement of MEE resolution for primary control, duration of treatment, cointerventions, other antimicrobial drugs, blinding protocol, compliance check, or the quality score. Diagnostic certainty as an independent variable had no significant impact on the regression analysis of RD as the dependent variable. 4) RD was unrelated to bilateral AOM and days since prior antibiotics.	1) "Should antibiotics be part of the initial empiric therapy for AOM in children? Our meta-analysis suggests that the answer is a qualified yes....Six of every seven children with AOM either do not need antibiotics for primary control or will not respond to antibiotic therapy..." (spontaneous 81% resolution of symptoms without antibiotics) 2) "We did not detect any significant differences in comparative clinical efficacy for standard- vs. extended-spectrum antibiotics..."
			primary end point: clinical response to antimicrobial therapy defined by absence or presence of presenting signs and symptoms 7-14 days after initiation of therapy including improved TM appearance if reported (somewhat vague description) secondary end point: presence or absence of MEE after resolution of acute infection as close to 30 days after initiation of therapy	(ORs also reported for the primary end point.) (None of the RDs for the secondary end point were significant.)	The authors note that these results apply to children older than 4 weeks of age for initial treatment of uncomplicated AOM and not to other situations.
					Good discussion of study weaknesses by the authors. The authors do not calculate the "file drawer" effect estimate.
					Possible RCTs Published Since Analysis: 25?

Inclusion Criteria	Quality Control
RCT; assessment of antimicrobial drugs; initial treatment; simple AOM	Yes (per Marchant and Shurin, 1982) score of 0.62+/-0.20 (0-1 possible)
Exclusion Criteria	Articles/Patients
studies of specific pathogens; myringotomy; OM not described; unable to extract data for ages 4 weeks to 18 years; most patients had treatment failure; most patients otitis prone	33/5400 (each meta-analysis utilized at most 5 studies) (47% did not describe randomization procedure; control group could be another antibiotic; only 4 study arms out of 69 were placebo/no drug; double-blind not required; primarily industry funded (61%); included studies with length of treatment 2 days.)

RD=rate difference, OR=odds ratio,ABX=antimicrobial, AMP=ampicillin, AMX= amoxicillin, CFC=cefaclor, CFX=cefixime,CLV=clavulanate,ERY=erythromycin, PCN=penicillin,TM=tympanic membrance, MEE=middle ear effusion,TMP=trimethoprim,SMX=sulfamethoxazole, SSX=sulfisoxazole

Table 28. Summary of Rosenfeld (1999b) meta-analysis (more...)

Table 28. Summary of Rosenfeld (1999b) meta-analysis

Title	Literature Sources	Paper Trail	Questions	Results	Author Conclusions/Reviewer Comments
Title	Literature Sources	Paper Trail	Questions	Results	Author Conclusions/Reviewer Comments
What to expect from medical therapy. (Only the acute otitis media meta-analyses will be summarized here.)	MEDLINE, 1996 through July 1998; manual search of symposium proceedings, published otitis media meta-analyses, book chapters, bibliographies of retrieved articles	Does not describe the number of articles initially retrieved from the literature search and the reasons for article rejection.	1) Are antimicrobials efficacious in the symptomatic relief of AOM pain and fever at 24 hours? 2) Are antimicrobials efficacious in the symptomatic relief of AOM pain and fever at 2-3 days? 3) Are antimicrobials efficacious in the symptomatic relief of AOM pain and fever at 4-7 days? 4) Are antimicrobials efficacious in the complete clinical resolution of AOM within 7-14 days? 5) Are antimicrobials efficacious in the resolution of OME 4-6 weeks after AOM treatment? 6) Are antimicrobials efficacious in the resolution of OME 3 months after AOM treatment?	Antibiotics vs. placebo/no antibiotic RD 0 (−7-8) (three studies; 633 subjects) 2) Antibiotics vs. placebo/no antibiotic RD 4 (2-7) (five studies; 1,241 subjects) 3) Antibiotics vs. placebo/no antibiotic RD 5 (−3-14) (four studies; 785 subjects) 4) Antibiotics vs. placebo/no antibiotic RD 13 (8- 19) (four studies; 612 subjects) 5) Antibiotics vs placebo/no antibiotic RD 3(−2-8) (5 studies, 1,447 subjects) 6) Antibiotics vs placebo/no antibiotic RD 5(−6-16) (2 studies, 371 subjects)	The author mentions that the results might not be generalizable to children with immune deficiencies, cleft palate, craniofacial anomalies, preexisting OME, complicated AOM, and concurrent bacterial infections, and that the 24 hour findings may not be generalizable to very young children. The author also mentions that the subjects in these studies may have less severe disease.
					Because the results are presented in a textbook, details of the methodology are not presented (e.g., as the number of articles retrieved from the initial literature search and actual quality scores). Also, no measure of heterogeneity is presented; the author assumes that the random effects model will adjust for any heterogeneity. The chapter does have a good discussion on the importance, validity, and generalizability of each result.

Inclusion Criteria	Quality Control
Randomized controlled trial	Quality scores are not reported. A statement is made that all of the studies in these meta-analyses were randomized and all but one were blinded.
Exclusion Criteria	Articles/Patients
Not mentioned	See "ANSWER(S)"

RD=rate difference

OME=otitis media effusion

Comparisons

Individual antibiotics have unique characteristics whose marginal effects on treatment of AOM may be lost by grouping. These marginal effects may be due to antibacterial spectrum, pharmacokinetics, bioavailability, and minimum inhibitory concentrations to specific organisms. In addition, previous meta-analyses (described above) grouped all antibiotics together without regard to antibacterial spectrum or pharmacokinetic properties (Del Mar, Glasziou, and Hayem, 1997; Rosenfeld, 1999b; Rosenfeld, Vertrees, Carr, et al., 1994). The project staff decided that specific comparisons would not need to be repeated unless new studies had been subsequently published that may have had significant effect on the estimate.

Ampicillin or Amoxicillin vs. Placebo or Observation

Study populations

The randomized controlled trials comparing ampicillin or amoxicillin with placebo or observational treatment include Halsted, Lepow, Balassanian, et al. (1967); Laxdal, Merida, and Jones (1970); Howie and Ploussard (1972); van Buchem, Dunk, and van't Hof (1981); Burke, Bain, Robinson, et al. (1991); and Kaleida, Casselbrant, Rockette, et al. (1991). The age and otitis-prone status of the study populations in these investigations are described in the Natural History section of this chapter.

Outcomes

The outcomes assessed in these studies are as described in the Natural History section. Rosenfeld (1999b) examined the issue of pain and fever resolution, middle ear effusion, and clinical resolution of symptoms at 7-14 days. An outcome common to three or more of the studies that has not been evaluated is clinical success or failure at 2-7 days. The difficulties with using success or failure as an outcome are detailed in the Natural History section. In particular, Howie and Ploussard (1972) defined failure as the presence of exudate at 2-7 days -- a definition which was unlike those of the other studies -- that included some measure of sign or symptom resolution. van Buchem, Dunk, and van't Hof (1981) did not measure clinical success or failure at 2-7 days, and it was not included in this meta-analysis. Another difficulty was that the data in Kaleida, Casselbrant, Rockette, et al., (1991) used the number of episodes as a denominator rather than the number of unique patients. Episodes are not independent of each other because a patient could have more than one episode of AOM within the study period. The failure of a patient to respond during one episode of AOM is not independent from that patient's response during a subsequent episode of AOM.

Results

Table 29. Comparison 1: Meta-Analysis 1.1 -- Key question (more...)

Table 29. Comparison 1: Meta-Analysis 1.1 -- Key question 3: ampicillin/amoxicillin vs. placebo; outcome indicator: failure rate at 2-7 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2 yr	Risk Factor 2 Otitis Prone (Prior episodes)	Amoxicillin/Ampicillin Sample Size	Placebo Sample Size	Amoxicillin Failure Rate (%)	Placebo Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
Halsted	1967	75% <2 yr	Not addressed	30	27	33.3	25.9	7.4	(−16.2, 31.0)
Laxdal	1970	49% <3 yr	Not addressed	49	48	10.2	37.5	−27.3	(−43.4, −11.2)
Howie	1972	100% <2.5 yr	Not addressed	36	116	52.8	79.3	−26.5	(−44.4, −8.6)
Burke	1991	0% <2 yr	47%>2 episodes	114	118	1.8	14.4	−12.7	(−19.4, −5.9)
Kaleida	1991	50% <2 yr	Not addressed	488	492	3.9	7.7	−3.8	(−6.7, −0.9)
Random effects estimates				717	801	13.6	32.9	−12.3	(−21.8, −2.8)
Test of heterogeneity Chi-square test value						51.37	341.00	18.83
Test of heterogeneity Chi-square test p value						<0.001	<0.001	0.002
NNT=−8 (−36, −5)

CI=confidence interval

NNT=number needed to treat

Figure 2. Shrinkage plot for outcome=failure (more...)

   Figure 2. Shrinkage plot for outcome=failure rate at 2-7 days of treatment Comparison 1: Meta-analysis 1.1=ampicillin/amoxicillin vs. placebo

Figure 3. Funnel plot for outcome=failure (more...)

   Figure 3. Funnel plot for outcome=failure rate at 2-7 days of treatment Comparison 1: Meta-analysis 1.1=ampicillin/amoxicillin vs. placebo

Table 30. Comparison 1: Meta-Analysis 1.2 -- Key question (more...)

Table 30. Comparison 1: Meta-Analysis 1.2 -- Key question 3: ampicillin/amoxicillin vs. placebo; outcome indicator: failure rate at 2-7 days of treatment (excluded Howie studies)

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Amoxicillin/Ampicillin Sample Size	Placebo Sample Size	Amoxicillin Failure Rate (%)	Placebo Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Halsted	1967	75% <2yr	Not addressed	30	27	33.3	25.9	7.4	(−16.2, 31.0)
Laxdal	1970	49% <3yr	Not addressed	49	48	10.2	37.5	−27.3	(−43.4, −11.2)
Burke	1991	0% <2yr	47%>2 epi	114	118	1.8	14.4	−12.7	(−19.4, −5.9)
Kaleida	1991	50% <2yr	Not addressed	488	492	3.9	7.7	−3.8	(−6.7, −0.9)
Random effects estimates				681	685	6.1	19.3	−9.7	(−19.2, −0.2)
Test of heterogeneity Chi-square test value						16.63	24.33	13.71
Test of heterogeneity Chi-square test p value						0.001	<0.001	0.003	NNT=−10 (−437, −5)

CI=confidence interval

NNT=number needed to treat

Table 31. Comparison 1: Meta-Analysis 1.3 -- Key question (more...)

Table 31. Comparison 1: Meta-Analysis 1.3 -- Key question 3: ampicillin/amoxicillin vs. placebo; outcome indicator: failure rate at 2-7 days of treatment (excluded Howie, and Kaleida studies)

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Amoxicillin/Ampicillin Sample Size	Placebo Sample Size	Amoxicillin Failure Rate (%)	Placebo Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Halsted	1967	75%<2yr	Not addressed	30	27	33.3	25.9	7.4	(−16.2, 31.0)
Laxdal	1970	49%<3yr	Not addressed	49	48	10.2	37.5	−27.3	(−43.4, −11.2)
Burke	1991	0%<2yr	47%>2 epi	114	118	1.8	14.4	−12.7	(−19.4, −5.9)
Random effects estimates				193	193	12.5	25.0	−12.9	(−27.5, 1.7)
Test of heterogeneity Chi-square test value						16.22	9.69	5.89
Test of heterogeneity Chi-square test p value						<0.001	0.008	0.053

CI=confidence interval

Figure 4. Shrinkage plot for outcome=failure (more...)

   Figure 4. Shrinkage plot for outcome=failure rate at 2-7 days of treatment Comparison 1: Meta-analysis 1.2=ampicillin/amoxicillin vs. placebo (excluding Howie study)

Figure 5. Shrinkage plot for outcome=failure (more...)

   Figure 5. Shrinkage plot for outcome=failure rate at 2-7 days of treatment Comparison 1: Meta-analysis 1.3=ampicillin/amoxicillin vs. placebo (excluding Howie and Kaleida studies)

A meta-analysis was performed to synthesize the data from Halsted, Lepow, Balassanian, et al. (1967); Laxdal, Merida, and Jones (1970); Howie and Ploussard (1972); Burke, Bain, Robinson, et al. (1991); and Kaleida, Casselbrant, Rockette, et al. (1991) with respect to failure at 2-7 days comparing placebo or observational treatment with ampicillin or amoxicillin (Table 29, and Figures 2 and 3). This collection of studies is statistically heterogeneous, suggesting the presence of differences in the populations studied and/or research methods employed. Therefore, caution is advised in interpreting overall summary measures. The random effects model calculates a rate difference of −12.3 (95 percent CI, 21.8 and −2.8). In other words, eight children with AOM would have to be treated with ampicillin or amoxicillin to avoid a case of clinical failure. Sensitivity analyses removing various studies did not materially effect the point estimate of −12.3. Subgroup analysis by age or otitis-prone status was not possible (Tables 30 and 31, Figures 4 and 5).

Antibiotic vs. Antibiotic

Having established that antibiotics have some marginal effect on treatment of AOM, it is valid to ask if certain antibiotic regimens have greater marginal effect than others. We asked: Does the specific antibiotic regimen make a difference in outcome? The possibilities for antibiotic comparisons are limitless. These possibilities depend, in part, on differences of antibiotic, dosage, schedule, duration, and other factors. The Technical Expert Panel chose five antibiotic treatments to evaluate.

Previous Meta-Analyses

Comparisons

Antibiotics Other than Amoxicillin or Trimethoprim-Sulfamethoxazole

Traditionally, the first-line antibiotics for treatment of AOM have been amoxicillin and trimethoprim-sulfamethoxazole (Rosenfeld, 1996). A question arose as to whether antibiotics other than these added any marginal benefit to the treatment of AOM. Many of the other antibiotics in question had theoretical benefits in terms of bacterial spectrum, pharmacokinetics, bioavailability, and minimum inhibitory concentrations to specific organisms; however, they also had greater costs and more possible side effects.

Penicillin vs. Ampicillin or Amoxicillin

Study Populations

Three randomized controlled trials were evaluated that compared penicillin with ampicillin or amoxicillin (Bass, Cashman, Frostad, et al. 1973; Laxdal, Merida, and Jones, 1970; and Nilson, Poland, Thompson, et al. 1969). In one study (Milson, Poland, Thompson, et al., 1969), 79 percent of children were younger than 2 years of age; 49 percent of those children in the second study (Laxdal, Merida, and Jones, 1970) were younger than 3 years of age; and subjects in the third study (Bass, Cashman, Frostad, et al., 1973) ranged in age from 2 months to 12 years, but numbers for specific age strata were not reported. The otitis-prone status of the study patients was not addressed in any of these studies.

Outcomes

Table 32. Key Question 4a: meta-analysis comparisons (more...)

Table 32. Key Question 4a: meta-analysis comparisons (terms and definitions; success and failure)

Author/Year	Term(s)	Definition(s)	Time Measured
Penicillin vs. Ampicillin or Amoxicillin
SUCCESS
Laxdal/1970	excellent	no evidence of middle ear inflammation	7 days
	good	signs of otitis media resolved	14 days
	fair	signs of otitis media resolved	21 days
Bass/1973	therapeutic effectiveness	no therapeutic failure, untoward reaction, or relapse	4 weeks
FAILURE
Nilson/1969	unsatisfactory	(1) marked injection >1/3 of TM or (2) marked decrease TM mobility on pneumatic otoscopy or (3) fluid level or bubbles or (4) otorrhea or (5) perforation or (6) marked distortion or obliteration of bony landmarks or (7) any degree of TM distension	10-12 days
Laxdal/1970	poor	persistent signs of minimal infection	21 days
	failure	no improvement or deterioration	7 days
Bass/1973	therapeutic failure	no improvement after 48 hours or exacerbation after initial improvement during 14 days of observation with repeat of therapy	<14 days
Cefaclor vs. Ampicillin or Amoxicillin
SUCCESS
Berman/1983	satisfactory	total resolution of effusion with normal mobility or residual serous effusion	>5 days
Jacobson/1979	cured	no fever, TM near normal appearance, no ear pain on exam	1 week
McLinn/1980	cure	disappearance or improvement of signs and symptoms during therapy and elimination or reduction to insignificant numbers of pathogen	10-21 days
Giebink/1984	satisfactory	absence of fever, irritability, otalgia, and TM erythema	3 days after ending antibiotic
Ploussard/1984	satisfactory	signs and symptoms had improved or disappeared	at end of therapy
FAILURE
Jacobson/1979	improvement without cure	no fever, TM slowly resolving but some inflammation remaining	1 week
	no improvement	fever, pain, TM unaltered or worse	1 week
McLinn/1980	failure	no improvement of signs and symptoms of infection and persistence of pathogen on culture	10-21 days
Berman/1983	failure	otoscopic findings of persistent bacterial infections (e.g. yellow or red immobile, bulging TM) with symptoms such as irritability and fever	>5 days
Ploussard/1984	unsatisfactory	no improvement of signs and symptoms at end of therapy	>3 days to end of therapy
Cefixime vs. Ampicillin or Amoxicillin
SUCCESS
McLinn/1987	favorable=cure or improvement	absence of fever, irritability, otalgia, and TM erythema	10 days
Leigh/1989	cure	all symptoms resolved	10-14 days
	improvement	significant improvement in symptomatology but without complete resolution	10-14 days
Johnson/1991	success	absence of fever, otorrhea, earache, and irritability	3-5 days
Principi/1991	cure, early outcome	normalization of clinical, otoscopic, and tympanometric findings	midtreatment and 15 days
	improvement, early outcome	relief of acute signs and symptoms with ear effusion by otoscopy and tympanometry	midtreatment and 15 days
	cure, late outcome	not failure at early outcome and resolution of otitis media with effusion per otoscopic tympanometric findings	30, 60, and 90 days
Owen/1993	improved	TM with dull appearance and decreased mobility or a dry perforation	4-6 days, 10-13 days, 31-38 days
FAILURE
McLinn/1987	failure	bacteriologic failure in those with repeat tympanocentesis or clinical failure including recurrence or inability to resolve fever and/or significant otalgia during therapy	10 days
Leigh/1989	failure	no response to therapy	10-14 days
Johnson/1991	failure	persistent fever, pain, irritability, or otorrhea	3-5 days
	bacteriologic failure	pathogen not eradicated	3-5 days
Principi/1991	failure	persistence of signs and symptoms of AOM and/or need to discontinue treatment due to adverse effects	midtreatment and 15 days
Owen/1993	bacterial failure	presence of bacterial pathogen in culture obtained while on or <24 hours after discontinuation of study medication	4-6 days if without dry perforation, 10 days at discretion of investigator
	clinical failure	continued or recurrent symptoms with red or yellow, bulging TM or purulent drainage with persistent perforation	4-6 days, 10-13 days, 31-38 days
Cefaclor vs. Trimethoprim-sulfamethoxazole
SUCCESS
Blumer/1984	satisfactory	absence of fever and otalgia and improved otologic exam	9-10 days
Marchant/1984	bacterial eradication	eradication of pathogens from the middle ear	3-6 days
	symptomatic improvement	decreased irritability per parent or guardian or absence of fever	3-6 days
FAILURE
Howie/1985	unsatisfactory	bulging eardrum or middle-ear drainage	14 days

As seen in Evidence Table 3, the outcomes measured in these three studies were not uniform. The only outcome common to all three studies was failure at 7-14 days of treatment. The definitions of failure were not uniform across the studies (Table 32).

Results

Table 33. Comparison 2: Meta-Analysis 2.1 -- Key question (more...)

Table 33. Comparison 2: Meta-Analysis 2.1 -- Key question 4a: penicillin vs. ampicillin/amoxicillin; outcome indicator: failure rate at 7-14 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Penicillin Sample Size	Amoxicillin or Ampicillin Sample Size	Penicillin Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
Nilson	1969	79%<2yr	Not addressed	96	101	28	25	3	(−9, 16)
Laxdal	1970	49%<3yr	Not addressed	45	49	24	10	14	(−1, 29)
Bass	1973	2mo-12yr	Not addressed	100	100	11	9	2	(−6, 10)
Random effects estimates				241	250	20.6	14.4	4.5	(−1.8, 10.7)
Test of heterogeneity Chi-square test value						10.85	9.86	1.97
Test of heterogeneity Chi-square test p value						0.004	0.007	0.374

CI=confidence interval

Figure 6. Shrinkage plot for outcome=failure (more...)

   Figure 6. Shrinkage plot for outcome=failure rate at 7-14 days of treatment Comparison 2: Meta-analysis 2.1=penicillin vs. ampicillin/amoxicillin

A meta-analysis was performed that synthesized the data on failure at 7-14 days from the three studies. The statistical test for heterogeneity was insignificant. The calculated rate difference was 4.5 percent (95 percent CI, −1.8 percent and 10.7 percent), which indicated no significant difference (Table 33 and Figure 6).

The study by Laxdal, Merida, and Jones (1970) indicated that for children younger than age 3 years and those age 3-6 years, ampicillin might be more effective than penicillin in treating AOM. The authors reported a 15 percent failure rate for ampicillin vs. a 32 percent for penicillin in children younger than age 3, and 0 percent failure vs. 30 percent for children age 3-6 years, respectively. However, denominators were not provided for these estimates (Laxdal, Merida, and Jones, 1970) (Evidence Table 3). Nilson, Poland, Thompson, et al. (1969) provided estimates stratified by bacterial isolate, which suggested that ampicillin might be more effective than penicillin V in treating patients whose cultures yielded Haemophilus influenzae (29 percent failure vs. 67 percent, respectively) (Evidence Table 3). Due to lack of information reported in the studies, meta-analysis by subgroups was not performed.

Cefaclor vs. Ampicillin or Amoxicillin

Study Populations

Five randomized controlled trials compared ampicillin or amoxicillin with cefaclor (Berman and Lauer, 1983; Giebink, Batalden, Russ, et al., 1984; Jacobson, Metcalf, Parkin, et al., 1979; McLinn, 1980; and Ploussard, 1984). The study by Berman and Lauer (1983) focused exclusively on infants age 1-3 months and in the study by Ploussard (1984), 84 percent of the subjects were younger than 2 years of age. McLinn (1980) reported that 45 percent of the subjects were younger than 2 years of age, and Giebink, Batalden, Russ, et al. (1984) reported 44 percent. Jacobson, Metcalf, Parkin, et al. (1979) listed an age range of 1-12 years and a mean age of 3.9 years. Giebink, Batalden, Russ, et al. (1984) reported that 41 percent of their study subjects had more than five previous episodes of AOM. The other studies did not address the otitis-prone state of their subjects.

Outcomes

As seen in Evidence Table 3, the outcomes had some commonalities and several differences. Two outcome indicators that were common to more than two studies in this group were failure rate at 3-7 days and failure rate at 5-21 days. As seen in Table 33, the definition of failure or its equivalent term was somewhat different among studies.

Results

Table 34. Comparison 3: Meta-Analysis 3.1. Key question (more...)

Table 34. Comparison 3: Meta-Analysis 3.1. Key question 4a: cefaclor vs. ampicillin/amoxicillin; outcome indicator: failure rate at 3-7 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2 yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefaclor Sample Size	Amoxicillin or Ampicillin Sample Size	Cefaclor Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Jacobson	1979	1-12 yr	Not addressed	13	15	15	13	2	(−24, 28)
Berman	1983	1-3 mo	Not addressed	19	21	26	29	−2	(−30, 25)
Giebink	1984	44% <2 yr	41%>5 episodes	31	30	23	17	6	(−14, 26)
Ploussard	1984	84% <2 yr	Not addressed	27	29	0	14	−14	(−26, −1)
Random effects estimates				90	95	14.6	16.8	−5.4	(−15.2, 4.4)
Test of heterogeneity Chi-square test value						14.85	1.79	3.05
Test of heterogeneity Chi-square test p value						0.002	0.616	0.383

CI=confidence intervals

Figure 7. Shrinkage plot for outcome=failure (more...)

   Figure 7. Shrinkage plot for outcome=failure rate at 3-7 days of treatment Comparison 3: Meta-analysis 3.1=cefaclor vs. ampicillin/amoxicillin

The meta-analysis synthesized data from four studies for the outcome on failure at 3-7 days of treatment (Table 34 and Figure 7). Although the four studies appeared to be quite different with regard to the age of the study populations, the test for heterogeneity was insignificant. The rate difference was −5.4 percent for cefaclor vs. ampicillin or amoxicillin (95 percent CI, −15.2 percent and −4.4 percent), which indicates lack of statistical significance.

Table 35. Comparison 3: Meta-Analysis 3.2. Key question (more...)

Table 35. Comparison 3: Meta-Analysis 3.2. Key question 4a: cefaclor vs. ampicillin/amoxicillin; outcome indicator: failure rate at 5-21 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefaclor Sample Size	Amoxicillin or Ampicillin Sample Size	Cefaclor Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
Jacobson	1969	1-12yr	Not addressed	13	15	15	13	2	(−24, 28)
McLinn	1980	45% <2yr	Not addressed	64	66	5	9	4	(−13, 4)
Berman	1983	1-3mo	Not addressed	19	21	26	29	−2	(−30, 25)
Giebink	1984	44% <2yr	41%>5 episodes	31	30	23	17	6	(−14, 26)
Ploussard	1984	84% <2yr	Not addressed	27	29	0	14	−14	(−26, −1)
Random effects estimates				154	161	15.8	13.0	0.5	(−5.7, 6.8)
Test of heterogeneity Chi-square test value						15.56	19.70	3.16
Test of heterogeneity Chi-square test p value						0.004	0.001	0.531

CI=confidence intervals

Figure 8. Shrinkage plot for outcome=failure (more...)

   Figure 8. Shrinkage plot for outcome=failure rate at 5-21 days of treatment Comparison 3: Meta-analysis 3.2=cefaclor vs. ampicillin/amoxicillin

Figure 9. Funnel plot for outcome=failure (more...)

   Figure 9. Funnel plot for outcome=failure rate at 5-21 days of treatment Comparison 3: Meta-analysis 3.2=cefaclor vs. ampicillin/amoxicillin

Five studies were eligible for the meta-analysis that evaluated failure at 5-21 days (Table 35 and Figures 8 and 9). The studies were not statistically heterogeneous. The rate difference was 0.5 percent (95 percent CI; −5.7 percent and −6.8 percent), indicating no significant differences.

The Berman and Lauer (1983) study, which evaluated infants 1-3 months of age, also showed no difference in terms of treatment failure after 5 days of treatment (Evidence Table 3). Subgroup analysis by quantitative synthesis could not be performed by age or otitis-prone status due to lack of information reported in this group of studies.

Cefixime vs. Ampicillin or Amoxicillin

Study populations

Five randomized controlled trials were evaluated for the comparison of cefixime with ampicillin or amoxicillin (Johnson, Carlin, Super, et al., 1991; Leigh, Robinson, and Millar, 1989; McLinn, 1987; Owen, Anwar, Nguyen, et al., 1993; and Principi and Marchisio, 1991). The study by McLinn (1987) included 61 percent of subjects younger than 2 years of age, and the Principi and Marchisio (1991) study included 40 percent in that age group. The subjects in the Johnson, Carlin, Super, et al. (1991) study had a mean age of 1.5 years and in the Owen, Anwar, Nguyen, et al. (1993) study, a mean age of 1.8 years. The mean age of subjects in the Leigh, Robinson, and Millar (1989) study was 4.9 years. Twenty-seven percent of children in the Johnson, Carlin, Super, et al. (1991) study had more than three prior episodes of AOM, and 26 percent in the Owen, Anwar, Nguyen, et al. (1993) study had more than two episodes. In the Principi and Marchisio (1991) study, 38 percent had one or more episodes of AOM; in the McLinn (1987) study, that number was 62.5 percent. Leigh, Robinson, and Millar (1989) did not address the otitis-prone status of their patients.

Outcomes

As seen in Evidence Table 3, a wide range of outcomes was measured in these five studies, and they were not always overlapping. Common outcomes included failure at 10-15 days, recurrence at 3-5 weeks, diarrhea, vomiting, and rash. Table 33 shows the differences in definition of failure.

Results

Table 36. Comparison 4: Meta-Analysis 4.1. Key question (more...)

Table 36. Comparison 4: Meta-Analysis 4.1. Key question 4a: cefixime vs. ampicillin/amoxicillin; outcome indicator: failure rate at 10-15 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefixime Sample Size	Amoxicillin or Ampicillin Sample Size	Cefixime Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
McLinn	1987	61%<2yr	62.5%>1 episodes	30	34	3	3	0	(−8, 9)
Leigh	1989	6m-16yr	Not addressed	150	150	5	5	−1	(−6, 4)
Principi	1991	40%<2yr	38%>1 episodes	20	20	10	10	0	(−19, 19)
Owen	1993	2m-6yr	26%>3 episodes	74	61	26	23	3	(−12, 17)
Random effects estimates				274	265	10.0	8.9	−0.1	(−4.2, 3.9)
Test of heterogeneity Chi-square test value						16.58	11.49	0.21
Test of heterogeneity Chi-square test p-value						0.001	0.009	0.976

CI=confidence interval

Figure 10. Shrinkage plot for outcome=failure (more...)

   Figure 10. Shrinkage plot for outcome=failure rate at 10-15 days of treatment Comparison 4: Meta-analysis 4.1=cefixime vs. ampicillin/amoxicillin Comparison 4: Meta-analysis 4.1=cefixime vs. ampicillin/amoxicillin

Four studies were used for the meta-analysis on failure at 10-15 days of treatment (Table 36 and Figure 10). The studies were statistically homogeneous. The calculated rate difference was −0.1 percent (95 percent CI, −4.2 percent and 3.9 percent), indicating that the rate difference was not statistically different from 0.

Table 37. Comparison 4: Meta-Analysis 4.2. Key question (more...)

Table 37. Comparison 4: Meta-Analysis 4.2. Key question 4a: cefixime vs. ampicillin/amoxicillin; outcome indicator: recurrence at 3-5 weeks of treatment

Study (First Author)	Year	Risk Factor 1 Age <2 yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefixime Sample Size	Amoxicillin or Ampicillin Sample Size	Cefixime Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
McLinn	1987	61% <2 yr	62.5% >1 episodes	30	34	3	3	0	(−8, 9)
Johnson	1991	2m-13 yr	27% >3 episodes	20	20	15	15	0	(−22, 22)
Principi	1991	40% <2 yr	38% >1 episodes	20	20	5	0	5	(−5, 15)
Random effects estimates				70	74	5.0	3.1	1.6	(−5.1, 8.4)
Test of heterogeneity Chi-square test value						1.83	3.04	0.37
Test of heterogeneity Chi-square test p-value						0.401	0.291	0.832

CI=confidence intervals

Figure 11. Shrinkage plot for outcome=recurrence (more...)

   Figure 11. Shrinkage plot for outcome=recurrence at 3-5 weeks of treatment Comparison 4: Meta-analysis 4.2=cefixime vs. ampicillin/amoxicillin

Three studies were synthesized for the analysis on recurrence at 3-5 weeks (Table 37 and Figure 11). The studies were not heterogeneous. The rate difference (1.6 percent) was insignificant (95 percent CI, −5.1 percent and 8.4 percent).

Table 38. Comparison 4: Meta-Analysis 4.3. Key question (more...)

Table 38. Comparison 4: Meta-Analysis 4.3. Key question 4a: cefixime vs. ampicillin/amoxicillin; outcome indicator: diarrhea as an adverse event

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefixime Sample Size	Amoxicillin or Ampicillin Sample Size	Cefixime Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
McLinn	1987	61% <2yr	62.5% >1 episodes	60	60	17	15	2	(−11,15)
Leigh	1989	6m-16yr	Not addressed	150	150	13	3	9	(3, 15)
Johnson	1991	2m-13yr	27% >3 episodes	59	51	34	29	4	(−13, 22)
Principi	1991	40% <2yr	38% >1 episodes	20	20	15	15	0	(−22, 22)
Owen	1993	2m-6yr	26% >3 episodes	91	93	29	14	15	(3, 26)
Random effects estimates				380	374	21.0	14.3	8.4	(3.8, 13.1)
Test of heterogeneity Chi-square test value						15.70	26.36	2.95
Test of heterogeneity Chi-square test p value						0.003	<0.001	0.567
NNT=12 (8, 27)

CI=confidence intervals

NNT=number needed to treat

Table 39. Comparison 4: Meta-Analysis 4.4. Key question (more...)

Table 39. Comparison 4: Meta-Analysis 4.4. Key question 4a: cefixime vs. ampicillin/amoxicillin; outcome indicator: vomiting as an adverse event

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefixime Sample Size	Amoxicillin or Ampicillin Sample Size	Cefixime Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
McLinn	1987	61% <2yr	62.5% >1 episodes	60	60	12	2	10	(1, 19)
Leigh	1989	6m-16yr	Not addressed	150	150	1	0	1	(−1, 3)
Johnson	1991	2m-13yr	27% >3 episodes	59	51	8	6	3	(−7, 12)
Principi	1991	40% <2yr	38% >1 episodes	20	20	5	0	5	(−5, 15)
Owen	1993	2m-6yr	26% >3 episodes	91	93	7	4	2	(−4, 9)
Random effects estimates				380	374	5.8	1.6	2	(0, 4)
Test of heterogeneity Chi-square test value						11.26	7.54	3.79
Test of heterogeneity Chi-square test p value						0.024	0.110	0.435

CI=confidence interval

Table 40. Comparison 4: Meta-Analysis 4.5. Key question (more...)

Table 40. Comparison 4: Meta-Analysis 4.5. Key question 4a: cefixime vs. ampicillin/amoxicillin; outcome indicator: rash as an adverse event

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Cefixime Sample Size	Amoxicillin or Ampicillin Sample Size	Cefixime Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
McLinn	1987	61% <2 yr	62.5% >1 episodes	60	60	15	2	13	(4, 23)
Leigh	1989	6m-16 yr	Not addressed	150	150	1	2	−1	(−4, 1)
Johnson	1991	2m-13 yr	27% >3 episodes	59	51	22	10	12	(−1, 26)
Owen	1993	2m-6 yr	26% >3 episodes	91	93	14	10	4	(−6, 13)
Random effects estimates				360	354	12.4	4.7	5.8	(−2.4, 13.9)
Test of heterogeneity Chi-square test value						36.71	9.90	12.11
Test of heterogeneity Chi-square test p value						<0.001	0.019	0.007

CI=confidence limits

Figure 12. Shrinkage plot for outcome=diarrhea (more...)

   Figure 12. Shrinkage plot for outcome=diarrhea as an adverse effect Comparison 4: Meta-analysis 4.3=cefixime vs. ampicillin/amoxicillin

Figure 13. Funnel plot for Outcome=diarrhea (more...)

   Figure 13. Funnel plot for Outcome=diarrhea as an adverse effect Comparison 4: Meta-analysis 4.3=cefixime vs. ampicillin/amoxicillin

Figure 14. Shrinkage plot for outcome=vomiting (more...)

   Figure 14. Shrinkage plot for outcome=vomiting as an adverse effect Comparison 4: Meta-analysis 4.4=cefixime vs. ampicillin/amoxicillin

Figure 15. Shrinkage plot for outcome=rash (more...)

   Figure 15. Shrinkage plot for outcome=rash as an adverse event Comparison 4: Meta-analysis 4.5=cefixime vs. ampicillin/amoxicillin

The five studies in this comparison group were included in three meta-analyses that looked at adverse effects presumably secondary to the antibiotics (Tables 38-40 and Figures 12-15). For diarrhea as the outcome, there was no heterogeneity in the rate difference; however, the heterogeneity of the incidence rates within antibiotic groups was significant. The rate difference was 8.4 percent (95 percent CI, 3.8 percent and 13.1 percent), favoring ampicillin or amoxicillin when compared with cefixime. Twelve children would need to be treated with ampicillin or amoxicillin rather than cefixime to avoid one case of diarrhea. For vomiting as the outcome, there was no heterogeneity statistically. The rate difference between cefixime and ampicillin or amoxicillin was 2 percent (95 percent CI, 0 percent and 4 percent), indicating no significant differences. For rash as the outcome, statistical heterogeneity existed in the incidence rates and rate difference. However, the rate difference of 5.8 percent (95 percent CI, −2.4 percent and 13.9 percent) was not significant.

The persistence of Haemophilus influenzae was reported as 0 of 10 in the cefixime group and 5 of 13 in the amoxicillin group in the Johnson, Carlin, Super, et al. (1991) study and 3 of 34 and 9 of 35, respectively, in the Owen, Anwar, Nguyen, et al. (1993) study. Subgroup analysis based on age or otitis-prone status was not reported in any of the studies and was not performed.

Other Antibiotics vs. Ampicillin or Amoxicillin

Other treatment regimens we compared with ampicillin or amoxicillin included: cephalexin, cephradine, cefuroxime-axetil, ceftriaxone, loracarbef, erythromycin-estolate, erythromycin-ethylsuccinate, clarithromycin, clindamycin, triple sulfonamide, penicillin G plus sulfisoxazole, penicillin V plus sulfisoxazole, penicillin V plus triple sulfonamide, procaine-penicillin-benthazine-penicillin-G plus sulfisoxazole, erythromycin-ethylsuccinate-sulfisoxazole, erythromycin-ethylsuccinate-acetyl-sulfafurazole, and erythromycin-sulfisoxazole.

Study populations

The studies varied in age of subjects. Four studies were very clear in what proportion of subjects were younger than 2 years of age (Brodie, Griggs, and Cunningham, 1990; Halsted, Lepow, Balassanian, et al., 1967; Nilson, Poland, Thompson, et al., 1969; and Scholz and Noack, 1998). The majority of the remaining studies reported the mean age or age range.

Outcomes

Various outcomes were measured in these studies. The definitions for success and failure were not uniform (Evidence Table 3).

Results

Table 41. Key Question 4a. Studies not in meta-analytic comparisons: (more...)

Table 41. Key Question 4a. Studies not in meta-analytic comparisons: synopsis

Author/Year	Comparator Antibiotic	Age	Otitis Prone	Result
	Ampicillin or Amoxicillin vs.:
Bass/1967	Procaine PCN+Bicillin+sulfisoxazole PCN V+sulfisoxazole Oxytetracycline	<5 years (70%); mean 3.6 years	Not addressed	No difference for failure <14 days for any comparator antibiotic
Halsted/1967, 1968	PCN G+sulfisoxazole	<2 years (75%)	Not addressed	No difference for failure at 24-72 hours
Nilson/1969	PCN V+sulfonamide (also PCN V)	<6 months (15%); 6-12 months (27%); 1-<2 years (36%); 2-3 years (21%)	Not addressed	No difference in unsatisfactory outcome at 10-12 days
Bass/1973	PCN V+sulfisoxazole Erythromycin estolate (also PCN V)	<3 years ("half")	Not addressed	No difference in failure at <14 days or relapses
Feigin/1973	Clindamycin	Mean 1.6 years	Not addressed	No difference in response at 3-4 days
Nassar/1974	Cephalexin	Mean 4.9 years	Otitis prone (52%)	No difference in failure at 14 days
Stechenberg/1976	Cephalexin	Mean 1.5 years	Not addressed	No difference in clinical response by 10 days
McLinn/1979	Cephradine	Mean cephradine 1.9 years and amoxicillin 2.4 years	Not addressed	No difference in initial improvement or in overall outcome at 10 days or recurrence through 12 months
Rodriguez/1985	Erythromycin sulfisoxazole	Mean 3.2 years	Not addresssed	No difference in failure at 10-14 days, recurrence at 30 days, or persistent middle ear effusion at 10-14 or 28 days

Table 41. Key Question 4a. Studies not in meta-analytic comparisons: (more...)

Table 41. Key Question 4a. Studies not in meta-analytic comparisons: age and otitis prone status

Brodie/1990	Cefuroxime axetil	<2 years (36%)	Not addressed	No difference for failure 1-4 days; no difference stratified by age
Foshee/1992	Loracarbef	Mean loracarbef 3.36 years and amoxicillin 3.59 years	Not addressed	No difference in failure at 7-10 days
Coles/1993	Clarithromycin	Mean clarithromycin 5.8 years and amoxicillin 5.3 years	Not reported	No difference in failure at 6-9 days
Pukander/1993	Clarithromycin	Range 1-12 years	Not addressed	No difference in failure at 48 hours
Kara/1998	Cefuroxime-axetil (also ceftriaxone)	Range 6 months-6 years	Not addressed	No difference in failure at 5 days
Lenoski/1968	EES; Triple sulfonamide (also EES+triple sulfonamide)	<3 years (58%)	Not addressed	No difference in failure at 14 days but numbers small; no difference stratified by age but numbers small
Scholz/1998	Erythromycin estolate	<2 years (14%)	Not addressed	No difference in failure at 9-11 days nor recurrence up to 41 days
Feldman/1990	Amoxicillin-clavulanate	Mean 60 months	Not addressed	Failure at 12-16 days was 7 of 101 in the TMP-SMX and 18 of 101 for amoxicillin-clavulanate, p=0.03

PCN=penicillin

EES=erythromycin ethylsuccinate

TMP-SMX=trimethoprim-sulfamethoxazole

All comparisons in this group showed no difference in various failure rates measured at various points in time. A few looked at age-stratified outcomes. No difference was found between cefuroxime-axetil and ampicillin when stratified by age (Brodie, Griggs, and Cunningham, 1990). Lenoski, Wingert, and Wehrle (1968), who compared erythromycin-ethylsuccinate with ampicillin, presented the failure at day 14, stratified by age. No difference was noted for subjects younger than 3 years of age, but the number of patients in the study was small. For those 3 years or older, none of the children on erythromycin-ethylsuccinate failed therapy, while 36 percent of those on ampicillin failed; again, the number of patients in the study was small. None of these studies stratified the findings by otitis-prone status (Evidence Table 3 and Table 41).

High Dose vs. Standard Dose of Amoxicillin-Clavulanate

As the incidence of bacterial resistance to antibiotics increases, modifications in treatment of AOM must be considered. These modifications may include use of different or new antibiotics or the use of familiar antibiotics in new ways. One proposal has been to increase the dosage of amoxicillin or amoxicillin-clavulanate. The question we posed was: What is the value of using > 60 mg/kg/day of amoxicillin or amoxicillin-clavulanate vs. the standard 40 mg/kg/day in the initial treatment of uncomplicated AOM?

Amoxicillin Therapy Twice a Day vs. Three Times a Day

Compliance with antibiotic treatment of AOM has been an issue. One proposal to increase compliance is to decrease the daily dosages of antibiotics from three to two.

Short-Term Vs. Long-Term Antibiotic Therapy

As already noted, compliance with antibiotic treatment of AOM has been an important issue to the practitioner and patient. Another proposal to increase compliance is to decrease the duration of therapy. The question we posed was: What is the comparative effectiveness of short-term vs. long-term antibiotic therapy in children younger than 2 years of age and those older than 2 years of age in the initial treatment of uncomplicated AOM?

Previous Meta-Analyses

Table 42. Summary of Kozyrskyj, Hildes-Ripstein, Longstaffe, (more...)

Table 42. Summary of Kozyrskyj, Hildes-Ripstein, Longstaffe, et al. (1998) meta-analysis

Title	Literature Sources	Paper Trail	Questions	Results	Author Conclusions/Reviewer Comments	Source	Inclusion Criteria	Quality Control	Authors	Exclusion Criteria	Articles/Patients
Treatment of Acute Otitis Media With a Shortened Course of Antibiotics: A Meta-analysis.	MEDLINE, 1/66-7/97; EMBASE, 1/74-7/97; Current Contents, 1/97-7/97; Science Citation Index searches; reference lists of relevant publications; no language restrictions and two trials included in the analysis were from non-English language reports	includes details on numbers of articles at each stage of selection and detailed descriptive information on included articles.	1) Are short-acting antibiotics given for 48 hours or less as effective as antibiotics given for at least 7 days in treatment of AOM? 2) Are short-acting antibiotics given for more than 48 hours but < 5 days as effective as antibiotics given for at least 7 days of treatment of AOM? 3) Is ceftriaxone as effective as antibiotics given for 10 days treatment of AOM? 4) Is 3-5 days of azithromycin as effective as 10 days of another antibiotic in treatment of AOM? stratifications: older and younger than 2 years; perforated and nonperforated tympanic membranes primary outcome: treatment failure noted by lack of clinical resolution or relapse or recurrence of AOM during a 31-day period following therapy initiation as defined by improvement or resolution of presenting signs and symptoms secondary outcome: cumulative number of treatment failures, relapses, and recurrences from diagnosis until final evaluation at 1-3 months (MEE was not counted a treatment failure.)	1) OR for treatment failure at less than one month 2.99(1.04-8.54) 2) a) OR for treatment failure at 8-10 days 1.38(1.15-1.66) 2) b) OR for treatment failure at 20-30 days 1.22(0.98-1.54) (Results 2) a) did not change with quality of study, adequacy of concealment of treatment allocation, exclusion of children with chronic or recurrent OM. 2) b) results for the most part were also not affected.) 3) OR for treatment failure at one month or less and at 3 months or less 1.25 (0.90-1.72) and 0.91 (0.57-1.47) respectively 4) OR for treatment failure 1.09 (0.86-1.38) OR for gastrointestinal side effects for short course versus long course short-acting antibiotics 0.54 (0.43-0.66) when amoxicillin-clavulanate was included in the analysis. Children on azithromyicin also had less likelihood of gastrointestinal side effects OR 0.26 (0.19-0.37)	1) Short-acting antibiotics for < 48 hours is not as effective. 2) "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...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...13 children would require treatment with a long course of antibiotics to prevent 1 excess failure following treatment with a shorter course....the risk difference at 1 month following treatment dropped to 2.3%...44 children requiring treatment with a longer course of antibiotics to prevent 1 failure following treatment." 3) Ceftriaxone was comparable. 4) Azithromycin was comparable. The authors state that subgroup samples were too small to assess the proposed stratifications. Good discussion of plausibility of results and weaknesses of the study by the authors. Possible RCTs published since analysis: 4?	JAMA	randomized controlled trials; treatment of AOM; comparing different durations of antibiotic treatment, fewer than 7 days vs at least 7 days; patients 4 weeks to 18 years age; no antimicrobial therapy at time of diagnosis; an assessment of clinical resolution of AOM	Yes (per Moher, Jadad, Nichol, et al., 1995) average score 2.66(SD,0.97) (0-5 possible) Also rated for concealment of treatment allocation. 9 trials in 8 articles reported adequate concealment	Kozyrskyj, Hildes-Ripstein, Longstaffe, et al., 1998	planned surgical cointervention except myringotomy subsequent to treatment failure or tympanocentesis (use of a second antibiotic was deemed a treatment failure and not excluded)	29 with 32 trials/9073 (each meta-analysis utilized at most 5 studies) (only 31% of studies on short-acting antiiotics excluded children with chronic or recurrent OM)

AOM=acute otitis media

MEE=middle ear effusion

OM=otitis media

OR=odds ration

RCT=randomized controlled trial

A meta-analysis was conducted that compared short-duration with long-duration antibiotic therapy for the treatment of AOM (Kozyrskyj, Hildes-Ripstein, Longstaffe, et al. 1998) (Table 42). The meta-analysis met the validity requirements listed in Oxman, Cook, and Guyatt (1994). The primary outcome evaluated was treatment failure, which was defined as the 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. The meta-analysis also looked at "the cumulative number of treatment failures, relapses, and recurrences reported from time of diagnosis until a final evaluation point between 1 to 3 months." Antibiotics in the short-duration arms of the comparison were grouped by "pharmacokinetic behavior" into short-acting oral antibiotics (i.e., penicillin V potassium, amoxicillin, cefaclor, cefuroxime), oral azithromycin, and intramuscular ceftriaxone. Duration of the short-acting oral antibiotics also was stratified into 48 hours or less and greater than 48 hours. Subgroup analyses that were based on two age groups (younger than and older than age 2 years) could not be conducted because the number of subjects was too small (Kozyrskyj, Hildes-Ripstein, Longstaffe, et al., 1998).

Synthesizing two studies on short-acting antibiotics given for 48 hours or less vs. long-duration therapy, the odds ratio was 2.99 (95 percent CI, 1.04 percent and 8.54 percent), which indicates a significant difference in risk when looking at treatment failure at less than 1 month followup; however, only two studies were quantitatively synthesized in this comparison. Comparing short-acting antibiotics given for more than 48 hours and less than 5 days vs. therapy of at least 7 days with treatment failure at 8-10 days as the outcome, the odds ratio was 1.38 (95 percent CI, 1.15 percent and 1.66 percent), which indicates a significant difference in risk. In the previous comparison, if the outcome was treatment failure at 20-30 days, the odds ratio was reduced to 1.22 (95 percent CI, 0.98 percent and 1.54 percent), indicating no significant difference in risk. These last two results were not affected by study quality, adequacy of treatment allocation concealment, or exclusion of subjects with chronic or recurrent otitis media.

The comparison of ceftriaxone to antibiotic duration of 10 days yielded an odds ratio for treatment failure at less than 1 month followup of 1.25 (95 percent CI, 0.90 percent and 1.72 percent), indicating no significant difference in risk. For treatment failure at less than 3 months followup, the odds ratio was 0.91 (95 percent CI, 0.57 percent and 1.47 percent), indicating no significant difference in risk. The meta-analysis on treatment for 3-5 days with azithromycin vs. 10-day therapy with another antibiotic yielded an odds ratio of 1.09 (95 percent CI, 0.86 percent and 1.38 percent), indicating no significant difference in risk.

It was concluded that treatment with short-acting antibiotics given for less than 48 hours was not as effective as antibiotic therapy given for at least 7 days (Kozyrskyj, Hildes-Ripstein, Longstaffe, et al., 1998), and that decreasing the duration of therapy from 10 days to 5 days might slightly increase the risk of treatment failure at 8-10 days (although by 30 days that difference would no longer exist). The latter conclusion is invalid because the difference in risk between 5-day and 10-day therapies at 20-30 days was not significant. In other terms, for the previous comparison, 13 children would need to be treated with the longer-duration therapy to prevent one excess failure that would occur with the 5-day treatment relative to 8-10 day treatment failure. Ceftriaxone and 3- to 5-day azithromycin seemed comparable to 10-day antibiotic treatment.

Ceftriaxone (1 dose) vs. Amoxicillin (7-10 days)

Study populations

Table 43. Comparison 9: Meta-Analysis 9.1. Key Question (more...)

Table 43. Comparison 9: Meta-Analysis 9.1. Key Question 4e: Ceftriaxone (1 dose) with amoxicillin (7-10d) outcome indicator: failure rate at 5-10 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Ceftriaxone Sample Size	Amoxicillin Sample Size	Ceftriaxone Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
Varsano	1988	6 mo-8 yr	58% >2 epidoes	22	22	18	14	5	(−17, 26)
Green	1993	5 mo-5 yr	18% >2 episodes	105	107	6	3	3	(−3, 8)
Kara	1998	6 mo-6 yr	Not addressed	25	25	16	8	8	(−10, 26)
Random effects estimates				152	154	10.5	5.1	3.4	(−1.6, 8.5)
Test of heterogeneity Chi-square test value						3.65	2.79	0.30
Test of heterogeneity Chi-square test p value						0.161	0.248	0.862

CI=confidence intervals

Figure 16. Shrinkage plot for outcome=failure (more...)

   Figure 16. Shrinkage plot for outcome=failure rate at 5-10 days of treatment Comparison 9: Meta-analysis 9.1=ceftriaxone (1 Dose) vs. ampicillin/amoxicillin (7-10 days)

Three randomized controlled trials compared ceftriaxone single-injection therapy to 7- to 10-days of amoxicillin therapy (Green and Rothrock, 1993; Kara, Ozuer, Kilic, et al. 1998; and Varsano, Frydman, Amir, et al., 1988) (Table 43 and Figure 16). In the Varsano, Frydman, Amir, et al. (1988) study, the mean age was 2 years. The mean age of patients in the Green and Rothrock (1993) study was 1.8 years. The patients in the Kara, Ozuer, Kilic, et al. (1998) study ranged in age from 6 months to 6 years, but the mean age or age distribution was not reported. Fifty-eight percent of patients in the Varsano, Frydman, Amir, et al. (1988) study had more than two previous episodes of AOM, compared to 18 percent in the Green and Rothrock (1993) study. The Kara, Ozuer, Kilic, et al. (1998) study did not address the issue of the otitis-prone state.

Outcomes

Table 45. Key Question 4e. Meta-analysis comparisons: (more...)

Table 45. Key Question 4e. Meta-analysis comparisons: terms and definitions; success and failure

Author/Year	Term(s)	Definition(s)	Time Measured
Ceftriaxone (1 injection) vs. Amoxicillin (7-10 days)
SUCCESS
Varsano/1988	satisfactory	resolution of middle ear fluid and of signs and symptoms of AOM or persistence of some degree of effusion but without erythema or bulging and with no fever or pain	7 days, 30 days
Green/1993	Successful	resolution of AOM symptoms and no return of symptoms within 10 days of emergency department visit	10 days, measured at 90 days
FAILURE
Varsano/1988	failure, early	persistence or recurrence of fever and/or pain with otoscopic signs of acute ear infection or spontaneous otorrhea during the first 10 days	7 days
Green/1993	unsuccessful	not successful	10 days, measured at 90 days
	failure	persistence or recurrence of symptoms within 10 days of initiating treatment	10 days, measured at 90 days
Kara/1998	no improvement or treatment failure	based on "resolution of symptoms and clinical and tympanometric appearance of the tympanic membrane"	5 days
Azithromycin (<5 days) ves. Amoxicillin-Clavulanate (7-10 days)
SUCCESS
Pestalozza/1992	cured	normalization of clinical, otoscopic, and tympanometric findings	3-5 days, 12-14 days, 30 days
	improved	relief of acute signs and symptoms with persistent middle ear effusion	3-5 days, 12-14 days, 30 days
Daniel/1993	cured	resolution of signs and symptoms of primary infection	3-5 days, 10-12 days
	improvement	signs and symptoms subsided but with incomplete resolution	3-5 days, 10-12 days
Schaad/1993	cure	disappearance of baseline symptoms of infection	7-20 days
Principi/1995	bacterial eradication	baseline pathogen not isolated upon repeat sampling or no culturable material obtainable because of clinical cure	10-14 days, aspirate only if no improvement
	cured	all pretreatment signs and symptoms disappeared	10-14 days
	Improved	Improvement or partial disappearance of pretreatment signs and symptoms	10-14 days
Arguesdas/1996	satisfactory	complete resolution of initial symptoms regardless of middle ear fluid	10-11 days
FAILURE
Pestalozza/1992	failed	no change or worsening of pretreatment signs	3-5 days, 12-14 days, 30 days
Daniel/1993	failure	no apparent clinical response	3-5 days, 10-12 days
Schaad/1993	failure	no change in or worsening of symptoms from baseline	7-20 days
Principi/1995	bacterial persistence	baseline pathogen present in post therapy sample	10-14 days, aspirate only if no improvement
	failed	no change or worsening of pretreatment signs and symptoms	10-14 days
Arguesdas/1996	failure	bacteriologic or clinical failure	10-11 days
	bacteriologic failure	inability to sterilize the middle ear fluid in patients with persistent ear drainage or who had repeated tympanocentesis	at discretion of investigator

The studies looked at different types of outcomes (Evidence Table 5). The outcome common to all three studies was failure rate at 5-10 days, but the definitions of failure were different. In particular, one study (Green and Rothrock, 1993) defined two terms, one failure and the other unsuccessful. Unsuccessful included both failed cases and cases with complications (Table 45).

Results

Table 44. Comparison 9: Meta-analysis 9.2. Key Question (more...)

Table 44. Comparison 9: Meta-analysis 9.2. Key Question 4e: Ceftriaxone (1 dose) vs. amoxicillin (7-10d) outcome indicator: failure rate at 5-10 days of treatment (criterion of failure relaxed to include complications in Green study)

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Ceftriaxone Sample Size	Amoxicillin Sample Size	Ceftriaxone Failure Rate (%)	Amoxicillin Failure Rate (%)	Rate Difference (%)	95% CI of Rate Difference (%)
Varsano	1988	6 mo-8 yr	58% >2 episodes	22	22	18	14	5	(−17, 26)
Green	1993	5 mo-5 yr	18% >2 episodes	105	107	10	9	1	(−7, 9)
Kara	1998	6 mo-6 yr	Not addressed	25	25	16	8	8	(−10, 26)
Random effects estimates				152	154	12.0	9.5	2.5	(−4.4, 9.5)
Test of heterogeneity Chi-square test value					1.12	0.40	0.51
Test of heterogeneity Chi-square test p value					0.570	0.819	0.775

CI=confidence intervals

Figure 17. Shrinkage plot for outcome=failure (more...)

   Figure 17. Shrinkage plot for outcome=failure rate at 5-10 days of treatment Comparison 9: Meta-analysis 9.2=ceftriaxone (1 Dose) vs. ampicillin/amoxicillin (7-10 days) (criterion of failure relaxed to include complications in Green study)

The meta-analysis on failure at 5-10 days synthesized data from three of the eligible studies. The studies were statistically homogeneous. Using "failure" rather than "unsuccessful" as the outcome in one study (Green and Rothrock, 1993), the estimated combined rate difference was 3.4 percent (95 percent CI, −1.6 percent and 8.5 percent) (Table 43 and Figure 16). Also, the results were not sensitive to the choice of outcome in that study (Green and Rothrock, 1993). Subgroup analysis by age or otitis-prone status was not possible (Table 44 and Figure 17).

Azithromycin (<5 days) vs. Amoxicillin-Clavulanate (7-10 days)

Study populations

Five randomized controlled trials were eligible for the comparison of less than 5-day azithromycin therapy vs. 7- to 10-day amoxicillin-clavulanate therapy (Arguedas, Loaiza, Herrera, et al., 1996; Daniel, 1993; Pestalozza, Cioce, and Facchini, 1992; Principi, 1995; and Schaad, 1993). In the Pestalozza, Cioce, and Facchini (1992) study, patients had a mean age of 3 years. The proportion of patients younger than 2 years of age was 1 percent in the Daniel (1993) study, 7.5 percent in the Schaad (1993) study, and 25 percent in the Principi (1995) study. Arguedas, Loaiza, Herrera, et al. (1996) reported an age range of 6 months to 12 years. In the Principi (1995) study, 10 percent of the subjects had recurrent otitis media. The other studies did not report on the otitis-prone status of their subjects.

Outcomes

The outcomes measured in these studies are somewhat similar in terms of the global clinical outcome (Evidence Table 5). The common outcome in this comparison was failure at 10-14 days. (The definitions are found in Table 45).

Results

Table 46. Comparison 10: Meta-analysis 10.1. Key question (more...)

Table 46. Comparison 10: Meta-analysis 10.1. Key question 4e: azithromycin (<5d) vs. amoxicillin-clavulanate (7-10d) outcome indicator: failure rate at 10-14 days of treatment

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Azithromycin Sample Size	Amox-Clav Sample Size	Azithromycin Failure Rate (%)	Amox-Clav Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Pestalozza	1992	11 mo-9 yr	Not addressed	15	15	0	0	0	(−, −)
Daniel	1993	1% <2 yr	Not addressed	103	54	6	0	6	(1, 10)
Schaad	1993	7.5% <2 yr	Not addressed	192	189	3	1	3	(−0, 5)
Principi	1995	25% <2 yr	10% recurrent	203	182	8	6	2	(−3, 7)
Arguedas	1996	6 mo-12 yr	Not addressed	47	45	0	4	−4	(−10, 2)
Random effects estimates				560	485	3.6	1.9	2.1	(−0.6, 4.8)
Test of heterogeneity Chi-square test value						12.66	10.64	5.34
Test of heterogeneity Chi-square test p value						0.013	0.031	0.254

Amox-Clav=amoxicillin-clavulanate

Cl=confidence interval

Figure 18. Shrinkage plot for outcome=failure (more...)

   Figure 18. Shrinkage plot for outcome=failure rate at 10-14 days of treatment Comparison 10: Meta-analysis 10.1=Azithromycin (<5d) vs. amoxicillin-clavulanate (7-10d)

Synthesizing data on failure at 10-14 days for all five of these studies using a random effects model, the pooled rate difference is 2.1 percent (95 percent CI, 0.6 percent and 4.8 percent), indicating no significant differences (Table 46 and Figures 18 and 19). Although the failure rates were heterogeneous among studies, no statistical heterogeneity existed for rate differences.

Principi (1995) stratified some of their results by age (Evidence Table 4). For those younger than 2 years of age, the 10- to 14-day failure rate was 18 percent (11 of 61 subjects) for the short-duration group and 10 percent (5 of 49 subjects) for the long-duration group. For children 2 years and older, the respective findings were 3 percent (5 of 154) and 5 percent (7 of 149). The differences were not statistically significant. Similar calculations were done for "not cured" at 10-14 days and for adverse effects. The numbers were small in those subjects younger than 2 years of age. Schaad (1993) showed that age stratification did not reveal any significant differences between age groups or between therapeutic groups. Here again, the numbers of patients younger than 2 years of age were small.

Azithromycin (5 days) vs. Amoxicillin-Clavulanate (7-10 days)

Study populations

Three randomized controlled trials compared 5-day azithromycin therapy to 7-10 day amoxicillin-clavulanate therapy (Aronovitz, 1996; Khurana, 1996; McLinn, 1996). The Aronovitz (1996) study population was 2 years of age or older. The mean age of the patients in the Khurana (1996) study was 5.6 years and 53 percent were 5 years old or younger. The age range for the McLinn (1996) study population was 1-15 years. Fifteen percent of the Khurana (1996) population had more than two prior episodes of AOM. The other two studies did not address the otitis-prone status.

Outcomes

Some of the outcomes found in these three studies are listed in Evidence Table 5. Common outcomes included any type of adverse effects and gastrointestinal adverse effects.

Results

Table 47. Comparison 11: Meta-Analysis 11.1. Key question (more...)

Table 47. Comparison 11: Meta-Analysis 11.1. Key question 4e: azithromycin (5d) vs. amoxicillin-clavulanate (7-10d) outcome indicator: any mention of adverse events

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Azithromycin Sample Size	Amox-Clav Sample Size	Azithromycin Failure Rate (%)	Amox-Clav Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Aronovitz	1996	0% <2 yr	Not addressed	85	84	4	31	−27	(−38, −17)
Khurana	1996	6 mo-12 yr	15% >2 episodes	263	260	7	17	−10	(−16, −5)
McLinn	1996	1-15 yr	Not addressed	340	334	9	31	−22	(−28, −16)
Random effects estimates				688	678	6.8	26.0	−19.2	(−29.2, −9.2)
Test of heterogeneity Chi-square test value						4.43	17.32	12.49
Test of heterogeneity Chi-square test p-value						0.109	<0.001	0.002
NNT=−5 (−11, −3)

NNT=number needed to treat

CI=confidence interval

Amox-Clav=amoxicillin-clavulanate

Figure 20. Shrinkage plot for outcome=any (more...)

   Figure 20. Shrinkage plot for outcome=any mention of adverse events Comparison 11: Meta-analysis 11.1=Azithromycin (5 d) vs. amoxicillin-clavulanate (7-10 d)

In terms of adverse effects, the meta-analysis calculated a pooled rate difference of −19.2 percent (95 percent CI, −29.2 percent and −9.2 percent). Five children with AOM would need to be treated with azithromycin rather than amoxicillin-clavulanate to avoid an adverse event. However, the outcomes reported in the studies in this comparison were heterogeneous statistically for the rate differences and for incidence rates within the amoxicillin-clavulanate group (Table 47 and Figure 20).

Table 48. Comparison 11: Meta-Analysis 11.2. Key question (more...)

Table 48. Comparison 11: Meta-Analysis 11.2. Key question 4e: azithromycin (5d) vs. amoxicillin-clavulanate (7-10d) outcome indicator: GI-related adverse events

Study (First Author)	Year	Risk Factor 1 Age <2yr	Risk Factor 2 Otitis Prone (Prior episodes)	Azithromycin Sample Size	Amox-Clav Sample Size	Azithromycin Failure Rate (%)	Amox-Clav Failure Rate (%)	Rate Difference (%)	95%CI of Rate Difference (%)
Aronovitz	1996	0% <2 yr	Not addressed	85	84	4	30	−26	(−37, −16)
Khurana	1996	6 mo-12 yr	15% >2 epi	263	260	6	15	−9	(−14, −4)
McLinn	1996	1-15 yr	Not addressed	340	334	8	29	−21	(−26, −15)
Random effects estimates				688	678	6.1	18.7	−18.0	(−28.0, −8.0)
Test of heterogeneity Chi-square test value						3.19	19.83	13.48
Test of heterogeneity Chi-square test p-value						0.203	<0.001	0.001
NNT=−6 (−13, −4)

NNT=number needed to treat

CI=confidence interval

Amox-Clav=amoxicillin-clavulanate

Figure 21 Shrinkage plot for outcome=gastrointestinal (more...)

   Figure 21 Shrinkage plot for outcome=gastrointestinal (GI) related adverse events Comparison 11: Meta-analysis 11.2=Azithromycin (5 d) vs. amoxicillin-clavulanate (7-10 d)

The next comparison examined the gastrointestinal adverse effects, which also showed highly heterogeneous results for both the rate differences and within the amoxicillin-clavulanate group. The random effects estimate of the pooled rate difference was −18.0 percent (95 percent CI, −28.0 percent and −8.0 percent). In other words, six children would need to be treated with azithromycin rather than amoxicillin-clavulanate to avoid a gastrointestinal adverse event. (Although not reported in the studies, the clavulanate concentration was most likely 31.25 mg per 125 mg of amoxicillin.) Subgroup analysis by age and otitis-prone status was not possible. (Table 48 and Figure 21).

In addition, individual studies (Aronovitz, 1996; McLinn, 1996) showed no differences in 11-day and 30-day failure rates between azithromycin (5 days) and amoxicillin-clavulanate (10 days).

Other Short-Duration vs. Long-Duration Comparisons

Table 49. Key question 4e. Studies not in meta-analytic comparisons: (more...)

Table 49. Key question 4e. Studies not in meta-analytic comparisons: synopsis

Author	Age	Otitis Prone	Result 1
Amoxicillin (<5 days) vs. Amoxicillin (7-10 days)
Chaput de Saintonge/1982	>2 years (100%)	>1 episode in previous excluded from study	No difference in sign and symptom resolution days 3 and 15, failure at 15 days, recurrences, complications
Bain/1985	>2 years (100%)	History of otitis media 75%	No difference in earache at 1 day, of 5 days, or 10 days; resolution of eardrum findings at 7 days or recurrence at 1 year
Puczynski/1987	>2 years (100%)	Not addressed	Significantly greater failure in the single dose group 3 of 7 vs. 0 of 10
Penicillin V (<5 day) vs. Penicillin V (7-10 day)
Meistrup-Larsen/1983	Mean 4.5 years	Not addresssed	No difference in unsatisfactory course at 14 days, but small numbers
Penicillin V (5 day) vs. Penicillin V (7-10 day)
Ingvarsson/1982	<2 years (24%)	Not addressed	No difference in not healed, which includes serous otitis media or relapses or therapeutic failure at 28-30 days nor in the individual components; higher percentage not healed if <2 years
Benthazine Penicillin G (1 dose) vs. tetracycline (7-10 day) or Benthazine Penicillin (1 dose) plus Triple-sulfonamide (7 day)
Rubenstein/1965	<2 years (48%)	Not addressed	Failure 4 of 79 (5%) vs. 10 of 78 (13%); no difference between bicillin and bicillin+triple-sulfonamide
Benthazine Pencillin G (1 dose) vs. Benthazine Pencillin (1 dose) plus Triple-sulfonamide (7 day)
Stickler/1967	<2 years (45%)	Not addressed	No difference in 2 week clinical failure
Amoxicillin-clavulanate (5 day) vs. Amoxicillin-clavulanate (7-10 day, either twice-a-day or three-times-a-day)
Hoberman/1997	<2 years (39%)	Not addressed	No difference in failure at 12-14 days or 32-38 days
Cohen/1998	<2.5 years (100%) mean 1.1 years	>3 episodes (9%), 2 episodes (11%), 1 episode (23%)	Greater failure at 12-14 days in shorter duration group (23% vs. 12 %)
Cefaclor (<5 day) vs. cefaclor (7-10 day)
Jones/1986	>2 years (100%) mean 5.7 years	History of otitis media (77%)	No difference in earache > 4 days or resolution of ear signs or recurrence within 6 weeks
Cefaclor (5 day) vs. Amoxicillin (7-10 day)
Ploussard/1984	<2 years (84%)	Not addressed	No difference in unsatisfactory outcome or recurrence but numbers small; all unsatisfactory outcomes and recurrences occurred in <2 year old patients
Cefaclor (5 day) vs. Cefaclor (7-10 day)
Hendrickse/1988	<2 years (53%)	Otitis prone (20% 2 )	With intact TM no difference in failure at <5 days or 6-10 days; reinfection at 14-30 days, or 31-60 days, or 61-90 days; persistent middle ear effusion rate at 10-15 days 30 days, 60 days, 90 days; with perforated TM no difference in reinfection or middle ear effusion rate but 46% to 8% difference in 6-10 day failure rate
Cefuroxime-axetil (5 day) vs. Cefuroxime-axetil (10 day) vs. Amoxicillin-clavulanate (7-10 day)
Gooch/1996	Mean 3.5 years	Prior episode (84%) Prior episode within 1 year (84%)	No difference in clinical failure after 3 days, recurrence 14-18 days after treatment end, or bacteriologic failure
Cefpodoxime-proxetil (5 day) vs. Amoxicillin-clavulanate (7-10 day)
Cohen/1997 (article 426 and 2325)	Mean 1.5 years <2 years (75%)	Not addressed Recurrent acute otitis media (25%)	Difference in nasopharyngeal pneumococcus persistence of (59% vs. 36%, p=0.01); no difference in nasopharyngeal H. influenzae persistence; both 2-6 days after treatment end No difference in failure 4 days after treatment end and at 17 days nor in serous otitis media at 1 month after treatment 24% vs. 18% with new acute otitis media at 1 month after treatment end
Cefpodoxime-proxetil (5 day) vs. Cefaclor (7-10 day)
Adams/1995	Mean 3.5 years	Not addresed	No difference in failure at 11-13 days or relapse at 3 weeks
Cefpodoxime-proxetil (5 day) vs. Cefixime (7-10 day)
Boulesteix/1995	Range 6 months-6 years	Otitis prone excluded	No difference in unsatisfactory course at 8-10 days and 30-40 days or relapse at 30-40 days
Cefprozil (5 day) vs. Cefprozil (7-10 day)
Kafetzis/1997	Mean 4.1 years	>2 episodes (3%) 2 episodes (6%) 1 episode (24%)	No difference in failure at 28-32 days or relapse after end of treatment
Ceftibuten (5 day) vs. Ceftibuten (10 day)
Simon/1997	Mean 3.7 years	Not addressed	More failure with the short-duration group (22% vs. 2%) and more recurrence at 14 days (14% versus 8%)
Ceftriaxone (1 dose) vs. Amoxicillin-clavulanate (7-10 day)
Bauchner/1996	<1.5 years (34%)	>2 prior episodes (11%) 1 prior episode (25%)	Unimproved or worse in 50 of 267 (19%) vs. 28 of 271 (10%); parents dissatisfacton with antibiotic significant with 8% vs. 11%
Varsano/1997	Mean 2.7 years	Otitis prone 2 (13%)	No difference failure at 11 days or relapses at 12-30 days; higher recurrence at 31-90 days in amoxicillin-clavulanate group; more diarrhea and vomiting with amoxicillin-clavulanate
Ceftriaxone (1 dose) vs. Cefaclor (7-10 day)
Chamberlain/1994	>1.5 years (100%) mean 3 years	<4 AOM per year (100% by exclusion of others) mean 1.4 episodes/year	No difference in failure at 14 days, recurrence at <14 days or 90 days, or persistent effusion 90 days
Ceftriaxone (1 dose) vs. Cefuroxime-axetil (7-10 day)
Kara/1988 (also vs. amoxicillin)	Range 6 months-6 years	Not addressed	No difference in failure at day 5 but very small numbers
Ceftriaxone (1 dose) vs. Trimethoprim-sulfamethoxazole (7-10 day)
Barnett/1997	Mean ceftriaxone 17.3 months and TMP-SMX 18 months	>3 prior episodes (41%)	Equivalent noncured rates at 14 days and 28 days but not at 3 days; otitis-prone status did not have an effect; those who failed appeared to be of younger mean age
Azithromycin (<5 day) vs. Cefaclor (7-10 day)
Rodriguez/1996	Mean 4 years	Not addressed	No difference in failure at 10-14 days; less failure in azithromycin at 25-30 days
Azithromycin (<5 day) vs. Clarithromycin (7-10 day)
Arguedas/1997	Mean 4.2 years	Not addressed	No difference in clinical failure by end of therapy
Azithromycin (5 day) vs. Amoxicillin-clavulanate (7-10 day)
Aronovitz/1996	<2 years (0%)	Not addressed	No difference in failure rate at 11 days or 30 days. Relapse assessed at 15-35 days was higher in the amoxicillin-clavulanate group, (21 %), compared to the azithromycin group, (5 %).
Khurana/1996	Range 6 months-12 years	>2 episodes (15%)	No difference in failure rate at 3 days or relapse before 14 days, recurrence after 14 days, or total unsatisfactory response.
McLinn/1996	Range 1-15 years	Not addressed	No difference in failure rate at 11 days or 30 days.

1

comparative results report short-duration result then long-duration result

2

meets technical expert panel of otitis prone

The treatment regimens we compared were the following: amoxicillin (<5 days) vs. amoxicillin (7-10 days), penicillin V (<5 days) vs. penicillin V (7-10 days), penicillin V (5 days) vs. penicillin V (7-10 days), benthazine penicillin G (1 dose) vs. tetracycline (7-10 days), benthazine penicillin G (1 dose) vs benthazine penicillin G (1 dose) plus triple-sulfonamide (7 days), amoxicillin-clavulanate (5 days) vs. amoxicillin-clavulanate (7-10 days), cefaclor (<5 days) vs. cefaclor (7-10 days), cefaclor (5 days) vs. amoxicillin (7-10 days), cefaclor (5 days) vs. cefaclor (7-10 days), cefuroxime-axetil (5 days) vs. amoxicillin-clavulanate (7-10 days), cefpodoxime-proxetil (5 days) vs. amoxicillin-clavulanate (7-10 days), cefpodoxime-proxetil (5 days) vs. cefaclor (7-10 days), cefpodoxime-proxetil (5 days) vs. cefixime (7-10 days), cefprozil (5 days) vs. cefprozil (7-10 days), ceftibuten (5 days) vs. ceftibuten (10 days), ceftriaxone (1 dose) vs. amoxicillin-clavulanate (7-10 days), ceftriaxone (1 dose) vs. cefaclor (7-10 days), ceftriaxone (1 dose) vs. cefuroxime-axetil (7-10 days), azithromycin (<5 days) vs. cefaclor (7-10 days), and azithromycin (<5 days) vs. clarithromycin (7-10 days) (Table 49).

Study populations

Fourteen of these studies reviewed indicated that the proportion of subjects were younger or older 2 years of age. Nine studies reported a mean age alone. Two studies only reported an age range. In nine studies, at least 20 percent of the subjects appeared to have been younger than 2 years of age (Bauchner, Adams, Barnett, et al., 1996; Cohen, de La Rocque, Boucherat, et al., 1997; Cohen, Levy, Boucherat, et al., 1998; Hendrickse, Kusmiesz, Shelton, et al., 1988; Hoberman, Paradise, Burch, et al., 1997; Ingvarsson and Lundgren, 1982; Ploussard, 1984; Rubenstein, McBean, Hedgecock, et al., 1965; Stickler, Rubenstein, McBean, et al., 1967). Thirteen studies indicated the proportion of subjects who had previous episodes of AOM, and three studies excluded patients with varying degrees of previous episodes (Bain, Murphy, and Ross, 1985; Boulesteix, Durbreuil, Moutot, et al., 1995; Chamberlain, Boenning, Waisman, et al., 1994). Twelve studies did not address the issue of otitis-prone status.

Outcomes

Many different outcomes were measured in these studies (Evidence Table 5).

Results

Nineteen of these 24 studies did not demonstrate any difference in effectiveness between the short-duration and long-duration therapy in the outcomes measured. Puczynski, Stankiewicz, and O'Keefe (1987) found, however, a significantly greater failure rate in the single-dose amoxicillin group compared with the 10-day amoxicillin group. Rubenstein, McBean, Hedgecock, et al. (1965) found a higher failure rate in those treated with long-duration tetracycline compared with a single dose of benthazine penicillin G. Cohen, Levy, Boucherat, et al. (1998) found a greater failure rate at 12-14 days in the 5-day duration of the amoxicillin-clavulanate group vs. the long-duration group. Henrickse, Kusmiesz, Shelton, et al. (1988) found a higher failure rate among those with a perforated tympanic membrane who were on short-duration cefaclor therapy at 6-10 days. Simon (1997) found a higher failure rate with short-duration ceftibuten therapy compared with long-duration ceftibuten therapy.

Only three studies looked specifically at the effect of age or otitis-prone status on any of the outcomes. Paradise (1997) reports in an editorial on age-stratified results from the Hoberman, Paradise, Burch, et al. (1997) study that, compared with 5-day treatment, 10-day treatment had significantly improved outcomes at days 12-14 and 32-38 for children younger than 2 years and at days 12-14 for children 2-5 years old. Paradise (1997) only reports on the twice-a-day amoxicillin-clavulanate regimen for 10 days and not the three-times-a-day amoxicillin-clavulanate regimen for 10 days. Ingvarsson and Lundgren (1982) stratified the "not healed" results by age. For treatment with the 5-day penicillin V, 25 percent of the children younger than 2 years were not healed and 20 percent of those 2 years and older were not healed. For treatment with the 10-day penicillin V, the respective findings were 33 percent for children younger than 2 years and 20 percent for those 2 years or older. In the Barnett, Teele, Klein, et al. (1997) study, those identified as otitis prone had a failure rate of 24 percent in the ceftriaxone group and a 23 percent failure rate in the trimethoprim-sulfamethoxazole group. The respective numbers for the patients who were not otitis prone were 26 percent and 19 percent. The data suggests that the otitis-prone status of the child did not have an effect on the outcome of failure at 14 days (Barnett, Teele, Klein, et al., 1997).