Published online July 3, 2006
PEDIATRICS Vol. 118 No. 1 July 2006, pp. 1-13 (doi:10.1542/peds.2005-1879)
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow P3Rs: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when P3Rs are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via CrossRef
Right arrow Citing Articles via ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Smith, C. G.
Right arrow Articles by Colborn, D. K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Smith, C. G.
Right arrow Articles by Colborn, D. K.
Related Collections
Right arrow Dentistry & Otolaryngology

Tympanometric Findings and the Probability of Middle-Ear Effusion in 3686 Infants and Young Children

Clyde G. Smith, MSa, Jack L. Paradise, MDb,c, Diane L. Sabo, PhDa, Howard E. Rockette, PhDd, Marcia Kurs-Lasky, MSd, Beverly S. Bernard, RN, BSb and D. Kathleen Colborn, BSb

a Departments of Audiology and Communication Disorders
b Pediatrics, Children's Hospital of Pittsburgh, Pittsburgh, Pennsylvania
c Department of Pediatrics, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
d Department of Biostatistics, University of Pittsburgh Graduate School of Public Health, Pittsburgh, Pennsylvania


   ABSTRACT
TOP
 ABSTRACT
METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
OBJECTIVE. We examined relationships between tympanometric findings and the presence or absence of middle-ear effusion in a population-based sample of children under the age of 3 years.

METHODS. In a study of children’s development in relation to early-life otitis media, we enrolled 6350 infants soon after birth and evaluated them regularly for the presence of middle-ear effusion. In 3686 of the children, we compared tympanometric findings with otoscopic diagnoses. We categorized tympanograms according to varying combinations of tympanometric peak height, peak pressure, and width, and calculated for each resulting category the percentage of the associated ears diagnosed as having effusion. Using these findings we developed algorithms for estimating the probability of middle-ear effusion associated with tympanograms of any configuration.

RESULTS. For tympanograms generally, the lower their height and the greater their width, the greater was the probability of associated middle-ear effusion; the probability also was greater when peak pressure was negative rather than positive. Among children ≥6 months of age, effusion was diagnosed in only 2.7% of ears with tympanometric height ≥0.6 mL, but in 80.2% of ears with flat tympanograms. Relationships among younger infants were similar but less consistent. In both age groups, the tympanographic configurations most commonly encountered were associated with either a relatively low probability (<30%) or a relatively high probability (>70%) of the presence of middle-ear effusion.

The receiver operating characteristic curve we generated using the algorithm we developed for children ≥6 months of age gave an area under the curve of 0.84. The algorithm performed equally well when applied to a separate group of children, suggesting that it is generalizable to other unselected populations.

CONCLUSIONS. The present report offers two alternative methods for estimating the probability of middle-ear effusion in children aged 6 through 35 months, given any combination of tympanometric values.

Key Words: tympanometry • middle-ear effusion • otitis media • otoscopic diagnosis • aural acoustic immittance testing • tympanometric-otoscopic relationships • receiver operating characteristic curve

Abbreviations: MEE—middle-ear effusion • ASHA—American Speech-Language-Hearing Association • ROC—receiver operating characteristic • CI—confidence interval

Otoscopic diagnosis of middle-ear effusion (MEE) in infants and young children is often problematic,1 and clinicians' skills in that regard are notably variable.2,3 Tympanometry, on the other hand, a component of aural acoustic immittance testing, is a simple, noninvasive procedure that gives objective findings reflecting middle-ear status. The procedure is now widely used to assist in determining the presence or absence of MEE, particularly among infants and young children.4 Tympanometry can be of value to the practitioner, educator, or clinical researcher in a variety of ways: in affirming or calling into question clinical diagnoses; in simply and inexpensively objectifying eardrum findings for purposes of clinical follow-up and clinical research; in providing useful information when inspection of the eardrum is not feasible or when expert evaluation of eardrum findings is not available; and in determining, through measurements of external canal volume, the presence of eardrum perforations and the patency of tympanostomy tubes.5

Previous reports in the pediatric literature have described the tympanometric procedure and its mode of operation in detail.6,7 Although the technique of the procedure has remained unchanged over the last half century, systems for classifying tympanometric results have varied. In early research, results were classified according to arbitrary values and graphic patterns6,8 that were not readily comparable across studies. In 1987 the American National Standards Institute issued guidelines (S3.39) that established current standards for immittance equipment, as well as standardized quantitative measurement units for immittance testing,9 and in 1997 the American Speech-Language-Hearing Association (ASHA) issued still-current guidelines (drafted in late 1996 and referred to as 1996 guidelines) for the use of immittance testing in screening for middle-ear disorders in infants and young children.10

The ASHA guidelines regarding MEE concerned children beginning at 6 months of age and were based on data from 3 studies that related tympanometric findings to clinical findings as obtained otoscopically or at myringotomy. The guidelines for infants aged 6 to 12 months were based on data reported by Roush et al,11 and those for children aged 1 to 6 years, on data reported in 2 studies by Nozza et al.12,13 Although the data provided by these 3 studies have been useful, they may have limited applicability to unselected, generally healthy children in the first few years of life and especially to children <1 year of age, for 2 main reasons. First, the numbers of children <1 year of age in the samples studied were relatively small, in the face of the fact that relations between tympanometric variables and middle-ear status may undergo substantial change over the first 2 years of life.6,14 Second, the samples were limited in their demographic diversity, their representativeness, or both, or consisted only of children with histories of chronic or recurrent otitis media.11–13 Apart from those limitations, the studies addressed only binary, pass/fail cutoff values, rather than addressing the more clinically relevant issue of the varying probability of MEE associated with the myriad constellations of tympanometric values that may be found in infants and young children.

Nearly 3 decades ago, we reported probabilities of MEE in relation to tympanometric findings in a sample of 280 children aged 10 days to 5 years.6 The study was limited, however, by the fact that the sample was not unselected, in that many of the children had been scheduled for myringotomy and tube insertion; only 140 of the children were <3 years of age, the age group in whom the prevalence of MEE is greatest and the diagnosis of MEE most problematic; and the probabilities of MEE were reported in relation to arbitrary, instrument-specific measurement units rather than the quantitative units in current standard use.

In the present study, we report relations between tympanometric findings and otoscopic diagnoses of the presence or absence of MEE in 3686 children during their first 3 years of life. The children were part of a population-based, sociodemographically diverse sample of otherwise healthy children who were being followed prospectively as participants in a larger study of possible effects of early life otitis media on speech, language, cognitive, and psychosocial development.15–17 The present findings add precision to findings we reported earlier6 and, to our knowledge, constitute the largest and most representative data set yet reported concerning tympanometric-otoscopic relationships in infants and young children. For the clinician, the present findings provide a basis for estimates of the probability of MEE for the entire spectrum of possible tympanogram types.


   METHODS
TOP
 ABSTRACT
 METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
General Procedures
Details concerning study procedures have been described previously.15–17 In brief, between June 1991 and December 1995, we enrolled 6350 healthy infants from 2 to 61 days of age at 8 sites in the greater Pittsburgh area: Children's Hospital of Pittsburgh, Mercy Hospital of Pittsburgh, and 2 small-town and rural and 4 suburban private pediatric group practices. The study was approved by the institutional review boards of the 2 hospitals, and written informed consent was obtained from 1 or both parents of each enrolled infant. We used pneumatic otoscopy to evaluate children's middle-ear status at least monthly until they were 3 years old, and we regularly monitored the validity of study clinicians' otoscopic observations.15 Until December 31, 1994, the study protocol called for each participant to also undergo tympanometric testing either immediately before or immediately after each otoscopic examination at all routine monthly and illness-related visits. On occasion, when tympanometric and otoscopic findings were seemingly discordant, repeat otoscopic examination was undertaken, and, in some of those instances, the otoscopic diagnosis was revised.

Of the 6350 children enrolled, 5049 were enrolled through December 31, 1994. The present analyses are based on data from 3686 of those children who underwent, at least once before reaching the age of 3 years, both immittance testing yielding an interpretable instrument-generated tympanogram and, at the same visit, otoscopic examination in 1 or both ears. Of the 1363 children not included in the present analysis, 454 (33.3%) had been observed for ≤4 months, and 602 (44.2%) had had ≤3 visits during the period covered by the analysis.

Classification of Children
For each child included in the present analysis, we used data from only 1 visit. We divided the 3686 children into 2 groups (Fig 1). Group 1 consists of 259 children who underwent 421 joint tympanometric-otoscopic assessments of individual ears in which the tympanometric and otoscopic assessments were known to have been conducted independently, that is, the otoscopic assessments were made by examiners who were unaware of the tympanometric findings. These assessments were undertaken at the various study sites as part of the ongoing appraisal of the study-team otoscopists' interobserver reliability, and all of the otoscopic examinations used for the group 1 analysis were performed by one or the other of 2 otoscopists (J.L.P. and B.S.B.), whose diagnoses of the presence or absence of MEE had been confirmed by findings at myringotomy in clinical studies,6,18 formal validation exercises,2 or both, involving children both younger and older than 6 months of age. To maximize the number of assessments for analysis, children in whom assessments were available for only 1 ear were included in group 1.


Figure 1
View larger version (21K):
[in this window]
[in a new window]
  		FIGURE 1 Flow diagram showing selection of study participants for the present analysis.

 
Group 2 consists of the remaining 3427 children, for whom records had not been kept at any visit as to whether the joint tympanometric-otoscopic assessment had been conducted independently. For each of these children, 1 visit was selected randomly, using a computer-generated table of random numbers. If fully analyzable data (ie, otoscopic diagnoses and interpretable tympanometric results) had been obtained for both ears on that date, those data were used for the present analysis. If fully analyzable data had not been obtained, analyzable data obtained on the nearest randomly selected visit before or after the originally selected visit were used instead. Thus, the 3427 children provided 6854 joint assessments of individual ears.

Tympanometric Procedures
Tympanometry was performed by trained study-team assistants. Two experienced audiologists (C.G.S. and D.L.S.) supervised testing throughout the study. A version 1 or version 2 GSI-33 Middle Ear Analyzer (Grason-Stadler, Inc, Milford, NH) was used at each study site. Each instrument was calibrated annually by an independent contractor in accordance with American National Standards Institute guidelines S3.39.9 Tympanograms were recorded using a 226-Hz probe tone and a positive (+300 daPa) to negative (–600 daPa) air-pressure sweep at either fast (600/200 daPa/second [ie, 200 daPa/second at the peak and 600 daPa/second before and after the tympanometric peak]) or slow (50 daPa/second) test speeds. During approximately the first half of the 43-month study period, slow speed was used mainly and during the second half because of evolving audiologist consensus, fast speed was used mainly. Of the tympanograms used for the present analysis, 71.6% were conducted at fast test speed.

Tympanometric Measurements and Values
The GSI-33 displays printed tympanometric results in 2 forms: as a graph (ie, tympanogram) plotted on calibrated x- and y-axes (Fig 2), and as values for 3 distinct immittance measures derived from the graph. These measures are: (1) peak compensated static admittance, often referred to as tympanometric peak height, and measured in millimhos (mmho), a unit of electric conductance equal to 1/1000 of a mho (the reciprocal of ohm), or in mL, which are numerically equivalent to millimhos; (2) tympanometric peak pressure, determined by the location of the peak deflection on the x-axis in relation to atmospheric pressure, and measured in decapascals; and (3) 1 of 2 measures of the sharpness of the tympanometric peak, either gradient, measured as the ratio a:b, where "b" equals overall peak height and "a" equals the vertical distance from the peak to a horizontal line intersecting the tympanogram so that its width between the points of intersection is 100 daPa,6,19 or tympanometric width, defined as the pressure interval in daPa where a horizontal line intersects the tympanogram at 50% of the peak height. When the tympanogram is flat, the instrument prints "NP," indicating "no peak," rather than providing any numerical values. In the interest of simplicity, we will hereafter refer to the 3 immittance measures as height, pressure, and width, respectively.


Figure 2
View larger version (19K):
[in this window]
[in a new window]
  		FIGURE 2 Typical tympanograms showing, for each, the estimated probability of MEE. Tympanometric peak height is measured in milliliters (numerically equivalent to millimhos). Tympanometric peak pressure is determined by the location of the peak deflection on the x-axis in relation to atmospheric pressure and is measured in decapascals. Tympanometric width is defined as the pressure interval in decapascals where a horizontal line intersects the tympanogram at 50% of the peak height. For graphs A and B, the probability values shown were derived by using the algorithm described in this article. Note that for graph A, the value closely matches the percentage in the corresponding cell in Table 5 (1.2%; n = 247) as derived from the empirical data. For graph B, the estimated probability differs somewhat from the percentage in the corresponding cell in Table 5 (36.7%; n = 49) but falls well within the 95% CI of that value. For graph C, because the graph is flat, an algorithm is not available, and the estimated probability shown was derived from the empirical data summarized in Table 5 (n = 217).

 
Selection of Tympanograms for Analysis
To be included in the present analysis, a tympanogram had to have been accompanied by a successfully completed otoscopic examination, had to be indicative of an intact tympanic membrane (ie, showing external auditory canal volume within normal limits), and had to be interpretable by virtue of showing good agreement between the printed numeric results and the graphic representation. (Patient movement or crying may produce artifacts, causing the instrument to record spurious results.) Tympanograms were rated as "good," "fair," or "poor" regarding their interpretability by 1 of 2 audiologists (C.G.S. and D.L.S.), and those rated as fair or poor were excluded from additional consideration. (The 2 audiologists' degree of interobserver agreement in this regard had been assessed by comparing their independent ratings in a separate sample of 346 randomly selected individual tympanograms. Ratings agreed for 286 [82.7%] of these tympanograms [205 good, 63 fair, and 18 poor] and disagreed between fair and poor for 21 [6.1%], between good and fair for 36 [10.4%], and between good and poor for 3 [0.9%] [weighted {kappa}20 = 0.70].)

Conversion of Values for Tympanometric Gradient From Ratios to Widths
Because some of our GSI-33 instruments recorded gradients rather than tympanometric widths, we performed a computerized image-analysis measurement of width for each tympanogram in which gradient had been originally recorded as a ratio value. One graduate student in audiology performed all of these measurements. To assess her accuracy, we compared her independent width measurements with machine-recorded widths on a series of tympanograms selected randomly from among those not included in the present analysis. The exercise consisted of 2 sets of 100 comparisons each before beginning the measurements for the present analysis, and a third set of 100 comparisons after having performed approximately one half of the measurements required for the analysis. Linear regression analyses of the 3 sets of comparisons between her measurements and machine-recorded widths yielded R2 values of 0.98, 0.91, and 0.96, respectively.

Otoscopic Assessments
At each monthly and interim visit, a study-team clinician examined each ear using a pneumatic otoscope with an air-tight lens assembly (model 20200, Welch Allyn Inc, Skaneateles Falls, NY). Cerumen was removed as necessary, and the clinician made a forced-choice decision for each ear concerning the presence or absence of MEE. For the present analysis, we did not distinguish between MEE associated with acute otitis media and MEE associated with otitis media with effusion. The study-team clinicians consisted of pediatricians, pediatric nurse practitioners, a physician assistant, and specially trained nurse otoscopists. We monitored the validity of the clinicians' otoscopic diagnoses on an ongoing basis, using concurrent independent diagnosis by 1 of the 2 aforementioned validated otoscopists as the criterion standard. Previously we reported that, at Children's Hospital, in 489 comparisons of independent diagnoses concerning the presence or absence of MEE in individual ears, agreement was found in 92.6% ({kappa} = 0.85). At the other 7 study sites, in 2099 such comparisons, the overall degree of agreement was 91.0% ({kappa} = 0.76), and the range at the respective sites was 85.0% ({kappa} = 0.58) to 96.5% ({kappa} = 0.89).15 Further details concerning otoscopic diagnosis and the management of study patients have been reported previously.15 For the purpose of the present report, the respective otoscopic diagnoses made by study-team clinicians served as the determinants of the presence or absence of MEE.

Statistical Analysis
Because of earlier observations that correlations between tympanometric variables and otoscopic findings were less consistent in infants ≤6 months of age than in older children,6 and in the light of ongoing clinical experience confirming those observations, we analyzed data from children in the first 6 months of life and from children aged 6 to 35 months separately. On the other hand, because correlations involving tympanometric values obtained at fast and at slow speeds, respectively, were similar, we combined the data from the 2 speeds for the purpose of further analysis. In the case of ears with flat tympanograms, for which the tympanometer recorded NP (no peak), we considered height to be 0, but we assigned no value to pressure or, with the exception noted in the footnote to Table 4, to width.


View this table:
[in this window]
[in a new window]
  		TABLE 4 Mean Tympanometric Width Values in Relation to Study Group, Age in Months, and Absence or Presence of MEE

 
We used 2-tailed tests for all of the analyses and we set statistical significance at P ≤ .05. We used {chi}2 tests to evaluate differences between proportions. To evaluate differences between mean values, we assumed that distributions were normal and used the method of generalized estimating equations. This procedure takes into consideration the correlation between the data for the 2 ears in a given child.21 We used logistic regression to evaluate empirical relationships between tympanometric values and otoscopic diagnoses (MEE versus no MEE) after adjusting for potentially confounding variables, and also to test for interactions.

To arrive at algorithms that relate the presence or absence of MEE as diagnosed in individual ears to combined values for the 3 associated tympanometric variables (ie, height, pressure, and width), we used the data set from group 2 because of its large size. Where values were available for each of the variables for both left and right ears (ie, other than in the case of flat tympanograms, from which numeric values for pressure and width cannot be calculated), we applied a logistic regression model, the general form of which is given by the equation:

Formula 1(1)
where ln is the natural logarithm (ie, logarithm to base e [whose value is ~2.71828]), p is an estimate of the probability that an ear contains effusion, ß0 is an intercept, and the other ß values are coefficients estimated from the data, using maximum likelihood estimation.22 The value of p can be estimated using the equation:

Formula 2(2)
where

Formula 3(3)
We performed analyses separately for infants in the first 6 months of life and for older children and separately for left ears and right ears. Because in each age group the models obtained for left and right ears, respectively, were similar, we averaged the coefficients of the tympanometric variables for left and right ears to arrive at a single model for each age group. We assessed the predictive ability of the resulting models by using the variable p from the logistic regression model to plot the receiver operating characteristic (ROC) curves for the given data sets and by then estimating the areas under the curves from the combined data for left and right ears, taking into consideration the correlation between the data for the 2 ears in a given child.23 (The area under an ROC curve for a test summarizes its overall diagnostic accuracy. If the area under the curve approaches 1.0, the test has excellent diagnostic accuracy; if the area approaches 0.5, the test is unable to differentiate between those with and those without the condition. One might consider the diagnostic accuracy of a test with an area under the curve of 0.60–0.70 to be poor; of 0.70–0.80 to be fair, of 0.80–0.90 to be good, and of 0.90–1 to be excellent.24) Finally, we validated our models by applying the algorithms developed from the data for group 2 to the independently obtained data for group 1.


   RESULTS
TOP
 ABSTRACT
 METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Subjects
Table 1 shows selected demographic characteristics of the children in groups 1 and 2. Demographics in both group 1 and group 2 approximated those in the study as a whole. However, groups 1 and 2 differed from each other in certain respects: the joint tympanometric-otoscopic assessments were, on the whole, conducted at earlier ages in group 1 than in group 2 (P = .002); and a larger percentage of the ears of children in group 1 than in group 2 had MEE diagnosed otoscopically at the time of evaluation (P < .001).


View this table:
[in this window]
[in a new window]
  		TABLE 1 Characteristics of Study Participants

 
Immittance Values Available for Analysis
Tympanometric results were recorded from 421 ears of the 259 children in group 1 and from each ear of the 3427 children in group 2. Of the 421 group 1 ears, numeric results were available for height in all and for pressure and width in 397 (94.3%). Of the 6854 group 2 ears, numeric results were available for height in all and for pressure and width in 6602 (96.3%). Missing values all concerned pressure and width from ears with flat tympanograms. Of the 397 calculable width values for group 1, 374 (94.2%) were determined using computerized image analysis, as discussed earlier in the "Methods" section, and 23 (5.8%) were recorded directly by the GSI 33. Of the 6602 calculable width values for group 2, the corresponding values were 5391 (81.7%) and 1211 (18.3%), respectively.

Individual Immittance Measurements in Relation to Age and the Presence or Absence of MEE
We examined the distributions of individual ears in group 1 and in group 2 according to each of the 3 immittance parameters in relation to children's age category at the time of assessment (ie, first 6 months of life versus ages 6–35 months) and the presence or absence of MEE as diagnosed otoscopically. The distributions in group 1 differed little from corresponding distributions in group 2; detailed data are available from the authors. In each age category, height values <0.3 mL, pressure values more negative than –100 daPa, and width values >200 daPa tended to be more prominently represented among ears with effusion than among ears without effusion; the differences were more pronounced among children in the older age group. Those differences notwithstanding, substantial overlap in values existed for each of the 3 parameters between ears with and ears without effusion.

Tables 2 through 4 show mean values in group 1 and group 2 for tympanometric height, pressure, and width, respectively, in relation to more narrowly categorized children's ages at the time of assessment and to the presence or absence of MEE. In group 2, no statistically significant differences in mean values were found between ears with and ears without effusion during children's first 3 months of life, but thereafter, significant differences became apparent and remained so. The pattern of values in group 1 was similar to that in group 2, but within most individual age categories the numbers were insufficient to achieve statistical significance. Corresponding comparisons of median values between ears with and ears without effusion gave the same pattern of findings. More detailed data concerning the comparisons are available from the authors. In both group 1 and group 2, for each of the tympanometric parameters, the interactions between age and MEE status were statistically significant (P ≤ .001).


View this table:
[in this window]
[in a new window]
  		TABLE 2 Mean Tympanometric Height Values in Relation to Study Group, Age in Months, and Absence or Presence of MEE

 

View this table:
[in this window]
[in a new window]
  		TABLE 3 Mean Tympanometric Peak Pressure Values in Relation to Study Group, Age in Months, and Absence or Presence of MEE

 
Combined Tympanometric Findings in Relation to the Presence or Absence of MEE
To determine relationships in individual ears between the presence or absence of MEE and combined findings for tympanometric height, pressure, and width, we first addressed the data set from group 2. We subdivided the data set into categories according to height intervals of 0.1 mL, pressure intervals of 50 daPa, and width intervals of 100 daPa. We then collapsed categories so as to show apparent similarities and differences in the likelihood of MEE with no more detail than seemed necessary. The resulting depictions of relationships are shown in Table 5 for the children 6 to 35 months of age and in Table 6 for the infants <6 months of age.


View this table:
[in this window]
[in a new window]
  		TABLE 5 Percentages of Ears of Group 2 Children 6 to 35 Months of Age With MEE According to Combined Findings of Tympanometric Height, Width, and Pressure

 

View this table:
[in this window]
[in a new window]
  		TABLE 6 Percentages of Ears of Group 2 Children <6 Months of Age With MEE According to Combined Findings of Tympanometric Height, Width, and Pressure

 
In the 2 age groups, the overall prevalences of MEE in individual ears were 13.5% and 12.2%, respectively. In general, in the older age group, the lower the height and the greater the width, the greater was the likelihood of MEE; the likelihood of MEE also tended to be greater when pressure was negative rather than positive, although the degree of negativity seemed to have little or no influence. Thus, at any given width-pressure combination, the proportion of ears with MEE tended to increase as height decreased; at any given height-pressure combination, the proportion of ears with MEE tended to increase as width increased; and at any given height-width combination, the proportion of ears with MEE tended to increase as peak pressure decreased from positive to negative. Among the infants <6 months of age, analogous relationships were obtained regarding height and width, but the relationships were less consistent overall than those in the older children, and pressure seemed to be negligible as a factor.

In children 6 to 35 months of age, tympanometric height ≥0.6 mL was infrequently associated with the presence of MEE (Table 5), but in the younger infants, this association was found with some frequency (Table 6). At the opposite extreme, MEE was diagnosed in 174 (80.2%) of the 217 ears of group 2 children aged 6 to 35 months with flat tympanograms (106 of 132 [80.3%] in infants aged 6–11 months and 68 of 85 [80%] in children aged 12–35 months) but in only 20 (57%) of the 35 ears of such children <6 months of age (3 of 4 [75%] in infants aged 0–2 months and 17 of 31 [55%] in infants aged 3–5 months).

Estimating Probabilities of MEE Algorithmically
For tympanograms that are not flat, use of the equations

Formula 4(3)
and

Formula 5(2)
as described earlier in "Methods" affords an alternative, mathematically based method of estimating the probability of MEE associated with a given combination of tympanometric values. Logistic regression analysis of the group 2 data, where values for each of the variables were available for both left and right ears, resulted in the intercepts and coefficients described in Table 7 to be used in arriving at values for A. These values are averages of values for left ears and right ears; data for the left and right ears separately are available on request. Application of these equations to the group 2 data set results in ROC curves with areas under the curve of 0.84 (95% confidence interval [CI]: 0.81–0.86) in children from 6 to 35 months of age (Fig 3A) and of 0.65 (95% CI: 0.59–0.70) in infants in the first 6 months of life (data not shown). Application of the equations derived from the children in group 2 to the data from children in group 1, in whom all otoscopic and tympanometric findings were obtained independently, results in areas under the ROC curve of 0.85 (95% CI: 0.78–0.92) for the children 6 to 35 months of age (Fig 3B) and of 0.45 (95% CI: 0.31–0.59) for the infants in the first 6 months of life (data not shown). Models including interaction terms involving the 3 tympanometric variables did not give significantly larger areas under the ROC curves. Figure 2 shows 3 representative tympanograms from our group 2 data set and, for each of the 2 tympanograms that are not flat, the calculated probability estimates of MEE using the above equation.


View this table:
[in this window]
[in a new window]
  		TABLE 7 Intercepts and Coefficients for Use in the Algorithm to Estimate the Probability of MEE

 

Figure 3
View larger version (9K):
[in this window]
[in a new window]
  		FIGURE 3 ROC curves summarizing tympanometric findings in relation to otoscopic diagnoses in study children 6 to 35 months of age. A, Curve generated by applying the algorithm developed from data for children in group 2 with values for both right and left ears for height, pressure, and width (ie, excluding children with a flat tympanogram in 1 or both ears) to all ears in the group 2 data set with values for height, pressure, and width (4761 ears). In group 2 not all of the joint tympanometric-otoscopic assessments were independent. The area under the curve is 0.84 of the plot area. The lettered rectangles A through F represent the respective plots obtained by applying the various 1996 ASHA-recommended pass/fail cutoff values shown in Table 8 to the data set. B, Curve generated by applying the algorithm developed from data for children in group 2 with values for both right and left ears for height, pressure, and width (ie, excluding children with a flat tympanogram in 1 or both ears) to all ears in the group 1 data set with values for height, pressure, and width (245 ears). In group 1, all of the joint tympanometric-otoscopic assessments were independent. The area under the curve is 0.85 of the plot area.

 
Correspondence Between Empirical and Algorithmic Estimates of MEE Probability
We examined the degree of correspondence between empirical estimates of MEE probability as shown in Tables 5 and 6 and estimates derived from the algorithms by comparing the results in samples composed of every fifth tympanogram associated with a diagnosis of MEE (n = 141) and every 75th tympanogram associated with a diagnosis of no MEE (n = 78). Empirical estimates and algorithmic estimates seemed generally similar, and differences in most cases seemed clinically insignificant. In only 15 (6.8%) of the 219 comparisons did empirical and algorithmic estimates of probability fall on opposite sides of 50%. Further details are available from the authors.

Applying the 1996 ASHA-Recommended Pass/Fail Cutoff Values to the Present Group 2 Data Set
Table 8 shows the test characteristics (sensitivity, specificity, and positive and negative predictive values) that would result from applying, individually and in combination, the 1996 ASHA-recommended pass/fail cutoff values10 to the present group 2 data set. The data illustrate the inherent trade-off between sensitivity and specificity. Both in the infants 7 to 12 months of age and in the older children, using tympanometric width as the sole determinant gives higher values for sensitivity and negative predictive value than using tympanometric height as the sole determinant but lower values for specificity and positive predictive value. Using a combination of width and height as the determinant, which would result in a larger proportion of ears being counted as test-positive than using either alone, accordingly gives the highest values for sensitivity and negative predictive value and the lowest values for specificity and positive predictive value. Figure 3A shows the points in the group 2 ROC curve that would result from applying each of the sensitivity-specificity combinations shown in Table 8 to the group 2 data set.


View this table:
[in this window]
[in a new window]
  		TABLE 8 Test Characteristics Resulting From Application of 1996 ASHA-Recommended Pass/Fail Cutoff Values10 to the Group 2 Data Set

 

   DISCUSSION
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
The Present Findings
The pass/fail cutoff values recommended in the 1996 ASHA guidelines for use in tympanometric screening, as shown in Table 8, were based, for infants aged 6 to 12 months, on data reported by Roush et al,11 and for children aged 1 to 6 years, on data reported by Nozza et al12,13 Key details of those studies are summarized in Table 9. The present report provides data concerning tympanometric-otoscopic relationships in 7275 ears from a demographically diverse, population-based sample of 3686 children ranging in age from <1 to 35 months. This is a more representative and far larger sample than those on which the 1996 ASHA guidelines were based, and it embodies a broad spectrum of differing combinations of values for the 3 tympanometric variables. In keeping with earlier findings,12 we found that tympanometric peak height and tympanometric width, each independently, as well as in combination, strongly influenced the probability of MEE, whereas tympanometric peak pressure had only minor influence.


View this table:
[in this window]
[in a new window]
  		TABLE 9 Studies on Which the 1996 ASHA Guidelines10 Were Based

 
As with diagnostic tests in general, and as we6 and others25 have noted previously, test results at the extremes of the testing spectrum gave the highest probability of being associated with the absence or the presence, respectively, of disease. As also found previously,6,14 relationships were relatively consistent in children ≥6 months of age but considerably less so in younger infants. The poorer performance of the test in younger infants seems attributable to the fact that the acoustic response properties of the external- and middle-ear systems in young infants differ from those in older infants and children. The difference is probably attributable, in turn, to physical changes that occur in the external and middle ear in the first few months after birth, mainly an increase in bone density and a loss of mesenchyme, with a resulting decrease in overall mass.26,27 Even among the children ≥6 months of age, however, most of the various combinations of tympanometric results obtained were represented in substantial proportion both among ears with effusion and among ears without effusion. Accordingly, in children ≥6 months of age, one should expect the clinical result to be derived from tympanometry to consist of an estimate of the probability of MEE, rather than a determination of either its presence or its absence.

That said, as shown in Table 5, most tympanograms are associated with either a relatively low probability (<30%) or a relatively high probability (>70%) of MEE and, accordingly, can be helpful in either reassuring the clinician that MEE probably is not present or raising the clinician's index of suspicion that MEE is present. Thus, the types of tympanograms found most commonly among healthy, asymptomatic children >6 months of age, namely, tympanograms with height ≥0.3 mL and width ≤200 daPa, have such a low probability of associated MEE (Table 5) that they might serve to obviate the necessity of otoscopic examination in evaluating such children clinically. That would especially be the case if the ears in question had previously been examined otoscopically and found to be normal. On the other hand, a flat tympanogram in a child initially diagnosed otoscopically as having no MEE might well prompt a second look by the clinician.

Earlier, Nozza et al12 noted that various anatomic, physiologic, and pathologic variables can influence tympanometric data and that such data, therefore, cannot be expected "to discriminate, without error, ears with MEE from ears without MEE." In that regard, it is of interest that if one were to apply the most stringent 1996 ASHA-recommended pass/fail cutoff values to our data set, most of the ears failing the screening test on the basis of those values would actually not have MEE and, further, would have shown relatively low probabilities of MEE as determined using the methods described in the present report.

The main offering of the present study differs from that of most previous studies of tympanometry in relation to MEE, which have focused on test characteristics (ie, sensitivity, specificity, etc) and pass/fail cutoff values for screening purposes, rather than on clinical diagnosis. Instead, the present findings in our large, normative population afford the clinician or investigator a choice of 2 methods for estimating the probability of MEE in the ear of any otherwise well infant or child <3 years of age, given any particular tympanographic configuration. First, reference may be made to Tables 5 and 6, which show percentages of ears with MEE by combined findings of tympanometric height, width, and pressure. Probabilities in any unselected population of the same age group should approximate those values. Alternatively, in ears other than those with flat tympanograms, values found for the 3 tympanometric measures may be entered into the algorithm described earlier to provide a calculated estimate of probability.

For most clinicians, it seems likely that reference to the age-appropriate table will ordinarily be the more practical of the 2 approaches. The investigator, however, may find the algorithm useful when comparing outcomes of competing treatments for otitis media at various stages of follow-up. Furthermore, the ROC curve generated using the algorithm may serve as a standard of comparison when evaluating the accuracy of other diagnostic tests for MEE (eg, spectral gradient acoustic reflectometry).

That the present findings are valid is suggested by 3 sets of circumstances. First, the tympanometric-otoscopic relationships found in group 2 of the present study (much the larger of the 2 groups), in which not all of the tympanometric and otoscopic observations were independent of each other, seemed generally similar to the relationships found in group 1, in which all of the observations were known to have been independent. Second, the probability algorithm derived from group 2 functioned comparably when applied to group 1. And third, a large majority of the algorithmic estimates of the probability of MEE were reasonably similar to the empirical estimates, with similar clinical implications, and few of the paired estimates were clearly discordant.

Limitations
One limitation of this study is its reliance on pneumatic otoscopic diagnosis by study-team clinicians, rather than on findings at tympanocentesis or myringotomy, as the criterion standard in determining the presence or absence of MEE. This is especially the case with very young infants. To the extent that there may have been misdiagnosis, it would have detracted from the accuracy of the percentages shown in Tables 5 and 6 and from the precision of the algorithms, but because the values at issue are probabilities of MEE rather than determinations of its presence or absence, small variations would have little or no practical import. It is also the case that reliance on otoscopic diagnosis is unavoidable in any ethical study involving a normative population, because one would not subject children without apparent middle-ear disease to tympanocentesis or myringotomy. Finally, as noted earlier, the diagnoses made by study-team clinicians consistently showed good agreement with those of validated otoscopists.15

A second limitation of the study is the absence or near absence from the study population of children from certain ethnic groups, notably Asians, Native Americans, and Inuits. A third limitation is the fact that not all of the otoscopic and tympanometric assessments in group 2 children were independent of each other (although, as noted earlier, findings in group 2 children were similar to those in group 1 children, in whom all of the otoscopic assessments were made independently of tympanometric assessments and by validated otoscopists). A fourth limitation lies in the fact that currently available immittance instruments give values for peak height in whole tenths of a milliliter, with intermediate values rounded either up or down, thereby detracting somewhat from precision in generating algorithmic results. A fifth limitation lies in the fact that, in categorizing the tympanometric data shown in Tables 5 and 6, the mathematical demarcations we used are arbitrary and, relatedly, that the estimates assigned to those categories are but estimates, not determinations. Finally, one must note that the present population-based findings in children <3 years of age cannot properly be extrapolated to children in that age group who are at high risk for middle-ear abnormalities, such as those referred for specialty otolaryngological or audiological services, nor is it certain how similar the findings would be in otherwise healthy older children.

Remaining Questions
Recent studies in newborns28,29 raise the possibility that the use of a higher-frequency probe tone, for example, 1000 Hz, rather than the currently standard 226-Hz tone, may offer better discrimination between ears with and without MEE in very young infants. Particularly because of the difficulty of otoscopic diagnosis in such infants, further study of this issue is called for.


   CONCLUSIONS
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
We have analyzed tympanometric data in relation to the presence or absence of MEE in a large, diverse, population-based sample of healthy infants and children <3 years of age. Tympanometric-otoscopic relationships were relatively consistent in children ≥6 months of age but not very consistent in younger infants. The findings can serve as a basis for determining, in children 6 to 35 months of age, the probability of MEE in association with any individual variant in the entire spectrum of possible tympanographic configurations. On the other hand, the findings also underscore the fact that no single tympanometric value or set of values can be used to differentiate with certainty between ears with and ears without MEE. Fuller understanding of both the potential and the limitations of tympanometry, as documented in the present analysis, should hopefully help clarify the place of the procedure in the armamentariums of the clinician, the clinical educator, and the clinical researcher.


   ACKNOWLEDGMENTS
 
This work was supported by grant HD26026 from the National Institute of Child Health and Human Development and the Agency for Healthcare Research and Quality and by gifts from GlaxoSmithKline and Pfizer Inc.

We are indebted to the following pediatricians who made the decisions, participated in the planning, and assisted in the efforts to integrate this study into their practices and who, at no small inconvenience and cost, provided unflagging support for study activities: at Beaver, David J. Cahill, MD, James Scibilia, MD, and Julius A. Vogel Jr, MD; at Brentwood, Mark Diamond, MD, and Thomas D. Skelly, MD; at Gibsonia, Amelia V. Agustin, MD, and Eva A. Vogeley, MD; at Kittanning, Harold A. Altman, MD, James K. Greenbaum, MD, Kenneth R. Keppel, MD, and Donald J. Vigliotti, MD; at Mt Lebanon, Scott L. Tyson, MD, and Celeste J. Welkon, MD; at Pleasant Hills, K. Gopalkrishna Pai, MD, and Harvey M. Rubin, MD; and at Mercy Hospital of Pittsburgh, Bradley J. Bradford, MD. We also thank J. Douglas Swarts, PhD, and Juliane M. Banks, BS, for assistance in obtaining computerized image-analysis measurements of tympanometric width and Bethaney Tessitore, MA, who performed all of the measurements. Regrettably, space limitations preclude a complete listing of the many other persons, most acknowledged previously,17 who served as study team clinicians or assisted in other clinical, administrative, or analytic capacities.


   FOOTNOTES
 
Accepted Feb 13, 2006.

Address correspondence to Jack L. Paradise, MD, Department of Pediatrics, Children's Hospital of Pittsburgh, 3705 Fifth Ave, Pittsburgh, PA 15213. E-mail: jpar{at}pitt.edu

The authors have indicated they have no financial relationships relevant to this article to disclose.


   REFERENCES
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 

  1. Paradise JL. Testing for otitis media: diagnosis ex machina. N Engl J Med. 1977;296 :445 –448[ISI][Medline]
  2. Kaleida PH, Stool SE. Assessment of otoscopists' accuracy regarding middle-ear effusion. Am J Dis Child. 1992;146 :433 –435[Abstract]
  3. Pichichero ME, Poole MD. Comparison of performance by otolaryngologists, pediatricians, and general practioners on an otoendoscopic diagnostic video examination. Int J Pediatr Otorhinolaryngol. 2005;69 :361 –366[CrossRef][ISI][Medline]
  4. American Academy of Family Physicians; American Academy of Otolaryngology-Head and Neck Surgery; American Academy of Pediatrics, Subcommittee on Otitis Media With Effusion. Otitis media with effusion. Pediatrics. 2004;113 :1412 –1429[Abstract/Free Full Text]
  5. Paradise JL. Otitis media. In: Behrman RE, Kliegman RM, Jenson HB, eds. Nelson Textbook of Pediatrics 17th ed. Philadelphia, PA: Saunders; 2004:2138 –2149
  6. Paradise JL, Smith CG, Bluestone CD. Tympanometric detection of middle ear effusion in infants and young children. Pediatrics. 1976;58 :198 –210[Abstract/Free Full Text]
  7. Brookhouser PE. Use of tympanometry in office practice for diagnosis of otitis media. Pediatric Infect Dis J. 1998;17 :544 –551[CrossRef][ISI][Medline]
  8. Jerger J. Clinical experience with impedance audiometry. Arch Otolaryngol. 1970;92 :311 –324[Medline]
  9. American National Standards Institute. Specifications for Instruments to Measure Aural Acoustic Impedance and Admittance (Aural Acoustic Immittance . New York, NY: American National Standards Institute; 1987
  10. ASHA Audiologic Assessment Panel 1996 (1997). Guidelines for screening infants and children for outer and middle ear disorders, birth through 18 years. In: ASHA 2002 Desk Reference Vol 4. Rockville, MD: American Speech-Language-Hearing Association; 2002:342–349. Available at: www.asha.org/NR/rdonlyres/A13D0ECC-684B-4E33-BA0E-CAF6681CEDEB/0/v4GLAudScreening.pdf. Accessed June 15, 2005
  11. Roush J, Bryant K, Mundy M, Zeisel S, Roberts J. Developmental changes in static admittance and tympanometric width in infants and toddlers. J Am Acad Audiol. 1995;6 :334 –338[Medline]
  12. Nozza RJ, Bluestone CD, Kardatzke D, Bachman R. Towards the validation of aural acoustic immittance measures for diagnosis of middle ear effusion in children. Ear Hear. 1992;13 :442 –453[ISI][Medline]
  13. Nozza RJ, Bluestone CD, Kardatzke D, Bachman R. Identification of middle ear effusion by aural acoustic admittance and otoscopy. Ear Hear. 1994;15 :310 –323[ISI][Medline]
  14. Paradise JL, Smith CG, Sabo DL, et al. Tympanometric detection of middle-ear effusion (MEE) in the first vs the second year of life [abstract]. Arch Pediatr Adolesc Med. 1994;148 :31
  15. Paradise JL, Rockette HE, Colborn DK, et al. Otitis media in 2253 Pittsburgh-area infants: prevalence and risk factors during the first two years of life. Pediatrics. 1997;99 :318 –333[Abstract/Free Full Text]
  16. Paradise JL, Feldman HM, Campbell TF, et al. Effect of early or delayed insertion of tympanostomy tubes for persistent otitis media on developmental outcomes at the age of 3 years. N Engl J Med. 2001;344 :1179 –1187[Abstract/Free Full Text]
  17. Paradise JL, Dollaghan CA, Campbell TF, et al. Otitis media and tympanostomy tube insertion during the first three years of life: developmental outcomes at the age of four years. Pediatrics. 2003;112 :265 –277[Abstract/Free Full Text]
  18. Paradise JL, Bluestone CD, Felder H. The universality of otitis media in 50 infants with cleft palate. Pediatrics. 1969;44 :35 –42[Abstract/Free Full Text]
  19. Brooks DN. The use of the eletroacoustic impedance bridge in the assessment of middle ear function. Int Audiol. 1969;8 :563 –569
  20. Fleiss JL. Statistical Methods for Rates and Proportions 2nd ed. New York, NY: Wiley; 1981
  21. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73 :13 –22[Abstract/Free Full Text]
  22. SAS Institute Inc. SAS/STAT User's Guide Version 8. Cary, NC: SAS Institute Inc; 1999
  23. Obuchowski NA. Nonparametric analysis of clustered ROC curve data. Biometrics. 1997;53 :567 –578[CrossRef][ISI][Medline]
  24. Tape TG. Interpreting diagnostic tests. Available at: http://gim.unmc.edu/dxtests/Default.htm. Accessed June 20, 2005
  25. Palmu AA, Syrjanen R. Diagnostic value of tympanometry using subject-specific normative values. Int J Pediatr Otorhinolaryngol. 2005;69 :965 –971[ISI][Medline]
  26. Keefe DH, Bulen JC, Arehart KH, Burns EM. Ear-canal impedance and reflection coefficient in human infants and adults. J Acoust Soc Am. 1993;94 :2617 –2638[CrossRef][ISI][Medline]
  27. Keefe DH, Levi, E. Maturation of the middle and external ears: acoustic power-based responses and reflectance tympanometry. Ear Hear. 1996;17 :361 –373[ISI][Medline]
  28. Kei J, Allison-Levick J, Dockray J, et al. High-frequency (1000 Hz) tympanometry in normal neonates. J Am Acad Audiol. 2003;14 :20 –28[Medline]
  29. Margolis RH, Bass-Ringdahl S, Hanks WD, Holte L, Zapala DA. Tympanometry in newborn infants–1 KHz norms. J Am Acad Audiol. 2003;14 :383 –392[Medline]
PEDIATRICS (ISSN 1098-4275). ©2006 by the American Academy of Pediatrics



This article has been cited by other articles:


Home page
 J. Clin. Microbiol.Home page
J. Jacques, H. Moret, D. Minette, N. Leveque, N. Jovenin, G. Deslee, F. Lebargy, J. Motte, and L. Andreoletti
Epidemiological, Molecular, and Clinical Features of Enterovirus Respiratory Infections in French Children between 1999 and 2005
J. Clin. Microbiol., January 1, 2008; 46(1): 206 - 213.
[Abstract] [Full Text] [PDF]

Home page
 Arch Pediatr Adolesc MedHome page
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]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow P3Rs: Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when P3Rs are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow E-mail this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My File Cabinet
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via CrossRef
Right arrow Citing Articles via ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Smith, C. G.
Right arrow Articles by Colborn, D. K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Smith, C. G.
Right arrow Articles by Colborn, D. K.
Related Collections
Right arrow Dentistry & Otolaryngology