 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
REVIEW ARTICLE |
,
* Division of General Pediatrics, Childrens Hospital Los Angeles, Los Angeles, California
Division of Biostatistics and Outcomes Assessment, Los Angeles County+University of Southern California Medical Center, Los Angeles, California
Center for Pediatric Health Outcomes Research, Department of Pediatrics, University of Southern California, Los Angeles, California
|| Department of Pediatrics, University of California, Los Angeles, Los Angeles, California
¶ Southern California Evidence-Based Practice Center, Rand, Santa Monica, California
# Statistics Group, Rand, Santa Monica, California
** Health Services Research and Development Service, Greater Los Angeles Veterans Affair Healthcare System, Los Angeles, California
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Methods. We searched Medline (1966–January 2000), the Cochrane Library (through January 2000), and Embase (1980–January 2000) and identified additional articles from reference lists in proceedings, published articles, reports, and guidelines. Excluded were nonhuman studies; case reports; editorials; letters; reviews; practice guidelines; non–English-language publications; and studies on patients with immunodeficiencies, craniofacial anomalies (including cleft palate), primary mucosal disorders, or genetic conditions. From each eligible study, we calculated the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and prevalence of OME in the cohort. We determined the number of studies for each comparison of a diagnostic method and a reference standard listed within the scope of our assessment. For comparisons with 3 or more studies, we derived random effects estimates of sensitivity, specificity, and prevalence rate. Using the pooled estimates, we plotted the performance of each diagnostic test in terms of sensitivity and (1 – specificity) and identified the best performer among the tests included in the comparison.
Results. Among 8 diagnostic methods, pneumatic otoscopy had the best apparent performance with a sensitivity of 94% (95% confidence interval: 92%–96%) and a specificity of 80% (95% confidence interval: 75%–86%). However, examiner qualifications were reported inconsistently, and training was not specified.
Conclusions. The finding that pneumatic otoscopy can do as well as or better than tympanometry and acoustic reflectometry has significant practical implications. For the typical clinician, pneumatic otoscopy should be easier to use than other diagnostic methods. The important question may be what degree of training will be needed for the clinician to be as effective with pneumatic otoscopy as were the examiners in the studies reviewed in this report.
Key Words: diagnostic methods • otitis media with effusion
Abbreviations: OME, otitis media with effusion • AHCPR, Agency for Health Care Policy and Research • CI, confidence interval
Various methods have been proposed for the diagnosis of otitis media with effusion (OME). The 1994 Agency for Health Care Policy and Research (AHCPR) OME guideline panel drew several conclusions regarding diagnosis of OME.1 The panel recommended the use of pneumatic otoscopy as the primary diagnostic method with tympanometry as a confirmatory diagnostic method. These recommendations were based on limited scientific evidence and strong panel consensus and on limited scientific evidence and expert opinion, respectively. The OME guideline panel found no evidence linking the outcome of algorithms that combine the results of pneumatic otoscopy and tympanometry to the presence of middle ear effusion. In addition, the panel believed that the evidence was insufficient to make any recommendation regarding the use of acoustic reflectometry in the diagnosis of OME. Finally, the panel decided not to make a recommendation on the use of tuning fork tests in the diagnosis of OME because of the lack of adequate studies. The OME guideline panel did not present any meta-analyses on diagnostic methods.
In an attempt to improve the evidence base for such decisions, we conducted an evidence assessment for various methods of diagnosing OME as 1 of 4 key objectives in our evidence assessment of the diagnosis, natural history, and late effects of OME sponsored by the Agency for Healthcare Research and Quality.2 The findings are reported in this article.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Evidence Assessment Process
We searched Medline (1966–January 2000), the Cochrane Library (through January 2000), and Embase (1980–January 2000) and identified additional articles by review of reference lists in proceedings, published articles, reports, and guidelines. The search used diagnosis and diagnostic techniques and procedures, as well as the text words "audiometry," "diagnosis," "diagnostic," "otoscopy," and "tympanometry" as key words. Two physicians or 1 physician and 1 health services researcher 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 OME in children. Excluded were case reports, editorials, letters, reviews, practice guidelines, non–English-language publications, and studies on patients with immunodeficiency disorders or craniofacial anomalies, including cleft palate.
The scope of the assessment included diagnostic studies that addressed children through 12 years of age, had the diagnostic procedure of interest performed within 24 hours of the reference standard, were not an algorithm or combination of multiple diagnostic procedures, and had abstractable data. The criteria established to assess study quality consisted of six components8: 1) Was the reference standard appropriate? 2) Were the test results and the reference standard assessed independent of each other? 3) Were the readers of the results of the diagnostic test or the reference standard blinded? 4) Did the patient sample include an appropriate spectrum of mild and severe and treated and untreated patients to whom the diagnostic tests were applied in clinical practice? 5) Were the reproducibility of the test result (precision) and its interpretation (observer variation) determined? 6) Were the methods for performing the test described in sufficient detail to permit replication?
As a recommendation from the peer reviewers of the evidence report,2 we added "the test performer" as a study-specific quality component. Thus, we abstracted the personnel who performed the diagnostic test for each study.
Data Synthesis
From each eligible study, we calculated the sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and prevalence of OME in the cohort. We then determined the number of studies for each comparison of a diagnostic method and a reference standard listed within the scope of our assessment. For comparisons with 3 or more studies, we derived pooled estimates of sensitivity, specificity, and prevalence rate. We used the DerSimonian and Laird random effects model4 to derive random effects estimates and 95% confidence intervals (CIs). This method produces a summary measure that is a weighted mean and allows both sampling variation and between-study heterogeneity (differences in the characteristics of the study population) to affect the pooled estimate. We pooled the prevalence rates to determine the heterogeneity of the study populations. Using the pooled estimates, we plotted the performance of each diagnostic test in terms of true-positive and false-positive rates and identified the best performer among the tests included in the comparison. We then derived the positive and negative predictive values for the best diagnostic test for various prevalence levels.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Findings
Table 1 presents the initial number of comparisons for each diagnostic method and reference standard pair after preliminary screening of the articles. These articles were subjected to a tertiary review to determine further their eligibility. Figure 1 provides the number of articles and the reasons for their exclusion at each stage of review. Comparisons with 3 or more studies were subjected to meta-analysis, from which we derived pooled random effect estimates; 95% CIs; and measures of heterogeneity for sensitivity, specificity, positive predictive value, negative predictive value, accuracy, and prevalence.
|
|
|
|
When we consider both the true-positive and false-positive rates in the display in Fig 2, pneumatic otoscopy is closest to the optimal point. The pooled sensitivity was 94% (95% CI: 91%–96%), and the pooled specificity was 80% (95% CI: 75%–86%). These findings were based on 2694 children from 7 studies that reported a pooled prevalence of OME of 63% (95% CI: 58%–67%). The estimated prevalence rates ranged from 56% to 71%, and the between-study heterogeneity was statistically significant (P < .001). This indicates that the prevalence rate of OME differed significantly among the 7 studies included in this comparison. To understand further the relationship of positive and negative predictive values with prevalence rate, we used the pooled sensitivity and specificity for pneumatic otoscopy and derived the positive and negative predictive values for various prevalence levels. Figure 3 provides such a plot.
|
4 out of a maximum of 6 possible points. The majority of the studies did not fulfill criterion 4 on the representativeness of patient sample in clinical practice or criterion 5 on determination of reproducibility of test results. Among the 7 studies used to derive the pooled estimates of sensitivity and specificity for pneumatic otoscopy, 1 study scored the minimum of 1 point, 1 scored 2 points, 1 scored 3 points, and 4 scored 4 points.
|
Summary of Findings
The meta-analyses revealed that pneumatic otoscopy and professional tympanometry had the highest sensitivity compared with myringotomy. Although the diagnostic test with the highest specificity was professional tympanometry (using static compensated acoustic admittance at 0.1), pneumatic otoscopy optimized both sensitivity and specificity. However, the poor quality of many of the studies included in the analysis must be considered. Moreover, most studies failed to provide enough information to assess the qualifications of testers.
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
5 or >5 RU; pneumatic otoscopy; portable tympanometry; professional tympanometry using static compensated acoustic admittance at 0.1, 0.2, and 0.3; professional tympanometry using B curve as abnormal; and professional tympanometry using B or C2 curves as abnormal. All comparisons used myringotomy as the reference standard. Although other reference standards were considered, none of them had an adequate number of studies for information synthesis. A search of literature cited in Medline on OME diagnostic methods since the conclusion of this study revealed no additional studies that would have met the inclusion criteria for this evidence assessment. Our findings provide evidence to support recommendations of the 1994 AHCPR OME guideline that had previously been based mainly on expert opinion. The 1994 AHCPR OME guideline stated that "the diagnostic evaluation of suspected otitis media with effusion should include pneumatic otoscopy. Otoscopy alone (without the use of the pneumatic otoscope to test tympanic membrane mobility) is not recommended."1
Our analysis considered several references not cited by the 1994 OME guideline; some were accepted, and some were rejected. The results of our meta-analyses confirm that pneumatic otoscopy had the best operating characteristics among the 8 alternatives examined. Our findings also confirm that certain but not all categories of tympanometry also perform well in identifying middle ear effusion in OME as well as in distinguishing it from other entities. Although the 1994 AHCPR OME guideline did not make a recommendation regarding acoustic reflectometry, our findings suggest that acoustic reflectometry does not perform as well as pneumatic otoscopy and certain types of tympanometry. The use of the spectral gradient angle, as the unit of measurement, may improve the sensitivity of acoustic reflectometry compared with the use of reflectivity, but this observation is based on a single study that found a sensitivity of 95% using a threshold of 95 degrees7 compared with the pooled sensitivity of 64% for a threshold reflectivity of 5 found in this study. However, the specificity was 32% when a spectral gradient angle of 95 degrees was used as the threshold, compared with the pooled specificity of 80% with a threshold reflectivity of 5 in this study. Unlike the 1994 AHCPR OME guideline, which commented on the combination of tympanometry and pneumatic otoscopy, we did not assess combinational diagnostic methods or algorithms.
Several factors should be considered in assessing our findings on pneumatic otoscopy. As noted in Table 3, 5 of the 7 studies are of adequate to high quality, ie, a quality score of 3 or above, and contributed 2263 (84%) of 2694 units of analysis and tended to have higher sensitivities and lower specificities than the 2 studies of lower quality. The study with the largest number of units of analysis (1902 [70.6%] of 2694 units of analysis reported by Karma et al28) was of adequate quality and looked at children who were 6 to 31 months of age, whereas the other studies looked predominantly at older children, and their results are slightly higher in sensitivity than for the 7 studies combined as detailed below in future research considerations.28 Only 1 study44 documented that pneumatic otoscopy was performed by a validated otoscopist41; thus, one might speculate that the performance in this study might represent the higher limits of pneumatic otoscopy characteristics, and that seems to be true in terms of sensitivity although not necessarily for specificity, which was highest in the study with test performers who seemed to be the least qualified among the 7 studies.34 The statistical heterogeneity for the pneumatic otoscopy meta-analyses was significant and may be a reflection of these differences or others not documented by the investigators as well as differences in OME prevalence.
The finding that pneumatic otoscopy can do as well as or better than tympanometry and acoustic reflectometry has significant practical implications. For the typical clinician, pneumatic otoscopy should be easier to use than other diagnostic methods. The important question may be what degree of training will be needed for the clinician to be as effective with pneumatic otoscopy as were the examiners in the studies reviewed in this report. Also, although we did not do a cost-effectiveness analysis, the cost of pneumatic otoscopy, in terms of direct and indirect costs, seem to be less than that for tympanometry or acoustic reflectometry.
Because of inadequate evidence, we could not conduct evaluations of clinical signs and/or symptoms, air and/or bone threshold audiometry, binocular microtympanoscopy, and nonpneumatic otoscopy in the diagnosis of OME. In addition, diagnostic methods that use algorithms or aggregated scorings are important but were not included in the scope of this evidence assessment. The sources of variation of such combination methods are difficult to detect in published articles. In addition, we must emphasize that we assessed the diagnosis of OME at single points in time rather than the diagnosis of persistent or recurrent OME over time. Furthermore, we must point out that different models of instruments were included within the broad categories of diagnostic methods. Our search strategy limiting to the literature in the English language is expected have little or no impact on the random effects estimates. A recent study by Moher et al65 provided no evidence that English-language–restricted meta-analyses lead to biased estimates of intervention effectiveness.
Future research on the diagnosis of OME will need to start with the definition of OME. The difficulty in reaching a consensus on the definition of OME was seen in our discussion of this issue with our technical expert panel. The technical experts agreed that OME was defined as "fluid in the middle ear without signs or symptoms of ear infection," as proposed by the 1994 AHCPR OME guideline. However, they could not agree on which signs or symptoms should be absent, ie, which signs or symptoms differentiated OME from acute otitis media. Without such agreement, we believe that little progress can be made in improving the diagnosis of OME.
In terms of child characteristics, the performance of these diagnostic tests in young children is of particular interest to clinicians. Because few studies looked solely at young children or stratified analysis by age, we were unable to analyze by age in the present evidence assessment. Among the studies that met the inclusion criteria of this evidence assessment, 3 specifically addressed the issue of age. Karma et al28 studied 2 groups of children 6 to 31 months of age. The first group was evaluated by an otolaryngologist, and the second group was evaluated by a pediatrician, both in academic settings. Comparing pneumatic otoscopy with myringotomy, they found a sensitivity of 98.8% and a specificity of 90.3% for the first group and a sensitivity of 93.6% and a specificity of 71.4% for the second group. Paradise et al44 looked at a subgroup of their patients who were younger than 7 months and found a sensitivity of 34.8% and a specificity of 85.7% comparing tympanogram curves with myringotomy or otoscopy. The authors concluded that " "normal’ tympanograms are of no diagnostic value in this age group."44 Sassen et al49 stratified their analysis by age and reported a tympanogram sensitivity of 90% and specificity of 67% for 5-month to 2-year-olds and a sensitivity of 81% and specificity of 63% for 2- to 12-year-olds compared with myringotomy. Although the sensitivity seemed to be lower in 5-month to 1-year-olds compared with those older than 1 year, the numbers in these subgroups were too small to arrive at any generalizable conclusions.49 Future studies must provide additional details on the characteristics of the children studied. Also, the study settings must be evaluated in greater detail so that the generalizability of the findings can be assessed in primary care as well as in tertiary care settings, where many of these studies are conducted.
Pneumatic otoscopy might seem to be less costly and more easily used by the typical clinician than other diagnostic options such as tympanometry and acoustic reflectometry. Nevertheless, future studies on the diagnostic assessment of OME should consider cost-effectiveness analysis, which can take into account the variable proficiency of clinicians in performing pneumatic otoscopy as well as the consequences of testing and patient preferences.66 Cost-effectiveness analysis will enable more informed decisions on the best diagnostic method for OME. The assessment of more complex diagnostic methods such as combination tests or algorithms would also benefit from cost-effectiveness analysis. Such analysis should be undertaken in the future.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Address correspondence to Glenn S. Takata, MD, Childrens Hospital Los Angeles, 4650 Sunset Blvd, MS 76, Los Angeles, CA 90027. E-mail: gtakata{at}chla.usc.edu
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  J. Chianese, A. Hoberman, J. L. Paradise, D. K. Colborn, D. Kearney, H. E. Rockette, and M. Kurs-Lasky Spectral Gradient Acoustic Reflectometry Compared With Tympanometry in Diagnosing Middle Ear Effusion in Children Aged 6 to 24 Months Arch Pediatr Adolesc Med, September 1, 2007; 161(9): 884 - 888. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. A. Damoiseaux, M. M Rovers, F. A. Van Balen, A. W Hoes, and R. A de Melker Long-term prognosis of acute otitis media in infancy: determinants of recurrent acute otitis media and persistent middle ear effusion Fam. Pract., February 1, 2006; 23(1): 40 - 45. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Other articles noted: 06 Feb 2004 to 16 Apr 2004 Evid. Based Nurs., July 1, 2004; 7(3): e3 - e3. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2003 American Academy of Pediatrics. All rights reserved. |
||
|
||
