Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
J Am Acad Audiol 6 : 150-162 (1995) Preliminary Descriptions of Transient-Evoked and Distortion-Product Otoacoustic Emissions from Graduates of an Intensive Care Nursery Brenda M. Bergman* Michael P. Gorga* Stephen T. Neely* Jan R. Kaminski* Kathryn L. Beauchaine* Jo Peters* Abstract Transient-evoked (TEOAE) and distortion-product otoacoustic emissions (DPOAE) were measured in 51 graduates of an intensive care nursery and compared to data obtained from 80 normal-hearing children and adults . All infants had click-evoked auditory brainstem responses (ABR) at 30 dB nHL or less while the older subjects had pure-tone thresholds of 20 dB HL or less for octave frequencies from 250 to 8000 Hz . OAE data were collected using commercially available devices. All data were analyzed in terms of emission amplitude, emission-to-noise ratio, and response reproducibility as a function of frequency. DPOAEs were measured at three points per octave between f2 frequencies of approximately 500 and 8000 Hz . TEOAEs were elicited by clicks and were analyzed in both octave and'/3-octave bands centered at frequencies from 500 to 4000 Hz, as well as in the broadband condition . In addition, stimulus amplitudes for the clicks used to elicit TEOAEs were analyzed within octave and 1/3-octave bands to determine whether any age-related differences in responses can be accounted for on the basis of stimulus differences . Both emission amplitude and noise amplitude were greater in neonates than adults, although there was variability across frequency. Emission-to-noise ratio and response reproducibility were more similar between groups . For TEOAEs, high-frequency emission-to-noise ratios were larger in neonates compared to older subjects, while the reverse was true in the lower frequencies . Less obvious frequency effects were observed for DPOAEs . These findings are discussed in relation to the potential use of OAEs as screening measures for neonatal hearing loss . Key Words: Distortion-product otoacoustic emissions (DPOAEs), infants, intensive care nursery (ICN), otoacoustic emissions (OAEs), transient-evoked otoacoustic emissions (TEOAEs) here has been considerable interest in otoacoustic emissions (OAEs) since they T were first described by Kemp (1978) . The presence of OAEs provides evidence of the existence of normal nonlinear properties within the cochlea. One consequence of cochlear damage is a reduction or loss of nonlinear behavior, which is reflected by a loss of OAEs . As a con- 'Boys Town National Research Hospital, Omaha, Nebraska Reprint requests : Brenda Bergman, Boys Town National Research Hospital, 555 North 30 Street, Omaha, NE 68131 15 0 sequence, ears with hearing loss are unlikely to produce OAEs . In their application with newborns, OAE measurements may provide basic information regarding the development of cochlear nonlinearities but also could be useful in the early identification of hearing loss . A number of screening programs are currently investigating the use of OAE testing as a method of screening for hearing loss (Kemp and Ryan, 1991 ; White et al, 1993). Data are accumulating that describe OAEs from the perinatal period in humans (e .g ., Johnsen et al, 1983 ; Kemp et al, 1986 ; Norton and Widen, 1990 ; Lafreniere et al, 1991 ; Stevens et al, 1991 ; Uziel and Piron, 1991 ; Kok et al, OAEs from ICN GraduatesBergman et al 1992 ; Smurzynski et al, 1993). Many of these data have shown that it is possible to measure OAEs in virtually all newborns with "normal hearing." Further, there is evidence that OAEs are larger in newborns compared to older children and adults with normal hearing (e .g., Norton and Widen, 1990 ; Kok et al, 1992). It is possible that both stimulus and response amplitudes may be greater in infant ear canals due to their smaller size compared to older subjects . Whether this can account for differences in response amplitudes remains undetermined . Almost all available OAE data from infants are restricted to TEOAEs . Few data exist that describe distortion-product otoacoustic emissions (DPOAEs) (Lasky et al, 1992) or that compare TEOAEs and DPOAEs from the same group of infants (Bonfils et al, 1992 ; Smurzynski et al, 1993). Data reported by Gorga et al (1993b) suggest that TEOAEs and DPOAEs, although differing in their measurements, result in more similar test performance, defined by their ability to separate normal from impaired ears . They did report some differences across frequency that are not yet understood . Finally, detailed comparisons of infant TEOAEs and DPOAEs to the OAEs obtained from older subjects are only beginning to emerge (e .g . the data of Smurzynski and Kim [1992] for adults versus the data of Smurzyinski et al [1993] in neonates). The present paper represents a preliminary description of both TEOAEs and DPOAEs from graduates of an intensive care nursery (ICN) and compares these data to similar measurements from normal-hearing older subjects . OAE amplitudes, noise amplitudes, OAE-to-noise ratios, and response reproducibility are described and compared for these two groups of subjects . In addition, stimulus amplitudes for the clicks used to elicit TEOAEs are analyzed within octave and 1/3-octave bands in order to determine the extent to which any age-related differences in responses can be accounted for on the basis of stimulus differences . We consider these data preliminary in that they were obtained from babies with normal hearing and thus can only be used to estimate the false positive rate . METHOD Subjects Fifty-one graduates from an ICN served as one group of subjects . All babies spending more than 24 hours in our ICN receive hearing screenings, mainly because risk registries tend to miss as many as 50 percent of hearing-impaired infants (Ellsmann et al, 1987). Thus, most but not all of these babies presented with risk factors, such as prematurity, elevated bilirubin (not necessarily requiring transfusion), and treatment with potentially ototoxic drugs . Twenty-seven subjects were males and 24 were females. Their conceptional ages ranged from 35 to 45 weeks, but the infants were at least 7 days old at the time of the test . Age at the time of the test appears to be an important variable affecting the false positive rate (Kok et al, 1992). ICN babies were tested just prior to hospital discharge on the same day on which ABRs were measured . All babies had click-evoked ABRs at 30 dB nHL or less (0 dB nHL = 35 dB pSPL as measured in a 2-cc coupler) . Data are also included from 80 normal-hearing children and adults . These "older" subjects had audiometric thresholds of 20 dB HL (ANSI, 1989) or less for octave frequencies from 250 to 8000 Hz and no evidence of middle-ear dysfunction, based on audiometric, immittance, and otoscopic examinations . No other subject selection criteria were used . Instrumentation ABR stimuli were delivered through an insert earphone (Etymotic, ER-3A) . Rarefaction clicks were presented at a rate of 17 per second and responses were measured between electrodes placed at the high-forehead and ipsilateral mastoid, with the center of the forehead serving as ground . Responses were amplified (100k), filtered (100-3000 Hz, 12 dB/octave), digitized at a rate of 33k, and averaged (1024 stimuli) using a Bio-Logic Navigator. Each response was replicated once. TEOAEs were measured with the IL088 Otodynamic Analyzer at its default stimulus conditions and in its nonlinear recording mode . Default conditions were used because to date, data do not exist that specify either stimulus or response conditions that result in optimal OAE test performance . Data from large samples of both normal-hearing and hearing-impaired neonates are needed to develop these optimal conditions. In the absence of such well-defined criteria, we chose to use those conditions in routine clinical use. An adult probe unit and impedancetip adapters were used for all measurements with the older group of subjects while the specially designed infant probe and its associated coupling cuffs were used with ICN graduates. Probe tips differed for infants and older Journal of the American Academy of Audiology/Volume 6, Number 2, March 1995 subjects because of differences in ear canal size . As a consequence, it was often more difficult to obtain good seals in newborn ear canals . This affected primarily the low-frequency energy in the ear canal signal . The IL088 provides a fixed voltage to its earphone, which, under default stimulus conditions, produces a level of about 80 dB pSPL in adult ear canals for the highest-level click in the stimulus ensemble (see below for more details) . In an effort to achieve comparable sound pressures in infant ear canals, the neonatal probe incorporates attenuation of 19 .5 dB . The noise rejection level was also set at the default level of 45 dB SPL. If noisy test conditions precluded collection of responses at this level, the rejection level was increased but did not generally exceed 50 dB SPL. Figure 1 provides mean estimates (+ 1 standard deviation) of the amplitude of the clicks used to elicit TEOAEs . These stimuli were analyzed in either octave (top) or 1/3-octave bands (bottom) for both groups of subjects . It appears that greater amplitude was achieved in the ears of older subjects in low and mid frequencies, while slightly greater stimulus amplitudes were seen for infants in the higher frequencies. The largest difference occurred in the 1/3-octave band centered at 1260 Hz, where, on average, the level in infant ears was 6.8 dB lower than that achieved in older subjects . Averaged across all frequency bands, the levels in the ear canals of older subjects exceeded those observed in infants by 2.6 dB . Possible explanations for these differences include differences in ear canal resonances and/or low-frequency leakage due to imperfect seals of the probe in the ear canal of infants. The stimuli appear to be similar between the two groups in the higher frequencies with greater difference in the lower frequencies.' Other features of the IL088 system have been described elsewhere (Kemp et al, 1986 ; Bray and Kemp, 1987), and only a brief description will be provided here . A miniature microphone is used to measure ear-canal sound pressure changes following the presentation of 80usec clicks, presented at an overall rate of 50 per second . These stimuli are presented in blocks of four, with the first three clicks presented at about 70 dB pSPL and one polarity while the fourth click is presented at 80 dB pSPL but opposite polarity. The responses from these four 'Differences in stimulus amplitudes as a function of frequency are less of a concern for DPOAEs, where amplitudes were set on a frequency-by-frequency basis . 152 OCTAVE BAND 70 65 so CIO 055 55 v .... i .......... .D 50 O - Older Children 6 Adults H - ICN Graduates Q (470 1/3 OCTAVE BAND t E 7 E 65 N 60 55 50 500 1000 2000 Frequency (Hz) 4000 Figure 1 Mean estimates (+ 1 standard deviation) of the amplitude of the TEOAE stimuli. Stimulus amplitude is in dB re : equivalent SPL in a 10-kHz bandwidth. Top panel shows octave-band analysis of the stimuli, and bottom panel shows 1/3-octave band analysis . clicks are added together, thus cancelling linear (stimulus related) components of the ear-canal waveform while leaving a portion of the nonlinear (cochlear) component of the response intact . This process eliminates the stimulus from measurement of the emission . The summed responses from alternate blocks of clicks are then stored in two buffers until each buffer includes 260 responses, at which point the test is terminated . The IL088 software provides three useful descriptions of these ear-canal waveforms. TEOAE amplitude or level is estimated as the cross-power spectrum of the two waveforms while background noise is estimated as the difference between these two waveforms. Response repeatability is estimated from the cross- OAEs from ICN GraduatesBergman et al correlation between the contents of the two buffers and is expressed as a percentage . While the above analyses are based upon broadband responses, we were also interested in knowing response and noise amplitudes, as well as percent reproducibility, in frequency bands associated with those frequencies typically evaluated during pure-tone behavioral audiometry. To this end, the ear-canal waveforms were analyzed in octave and 1/-octave bands. Octave bands were centered at the four octave frequencies from 500 to 4000 Hz, while the 1/3octave bands were centered at frequencies from 400 to 5040 Hz . For each octave and 1/3-octave band, TEOAE level, noise amplitude, and percent reproducibility were estimated using a similar, although not identical, approach to that used by the IL088 software to describe the broadband response . Specifically, response amplitudes were estimated from the spectral magnitude of the sum of the two buffers, as opposed to their cross-power spectrum (Gorga et al, 1993b; Prieve et al, 1993). DPOAEs were measured using the CUBDIS system distributed by Etymotic Research . This system has been described in detail elsewhere (Allen, 1990). Briefly, it includes an Ariel DSP16 signal processing board, an ER-10B low-noise microphone system, a battery powered amplifier providing 40 dB of gain, two ER-2A earphones, a transformer, and software for extracting acoustic distortion products from the ear-canal waveform. Impedance-tip adapters were used to couple the probe unit to the ear canals of older subjects . As with TEOAE measurements, default conditions of the CUBDIS system were used . These included 200 samples of 20 .48-msec duration, resulting in a total time at each frequency of 4 .096 seconds. No other stopping rule was used and no noise reduction was performed. Polyethylene tubing was cut to size and then slipped over the probe unit in order to couple it to the ear canal of neonates . Stimuli consisted of two sinusoids that were independently generated by the two channels ofthe Ariel board and then presented separately to the two ER-2A earphones, the sound-delivery tubes of which exited from the probe unit housing the low-noise microphone system. The two sinusoids (f1 and f2) were mixed acoustically in the ear canal. Stimulus conditions, which were identical to those used in our previous work (Gorga et al, 1993a), were as follows: f2/f1= 1 .2 ; L2 = 50 dB SPL; L1/L2 = 15 dB (L1 = level of lower frequency primary; L2 = level of higher frequency primary) ; three points per octave . These primary levels were chosen based upon the assumption that the interaction is greatest near the f2 place. Additionally, there is some indication that ears with mild cochlear damage perform more normally (i .e ., produce more normal DPOAE amplitudes) for high-level primaries, but the responses decrease rapidly and/or disappear for moderate level stimuli. During the calibration procedures included with CUBDIS, signal amplitude is measured in the ear canal of each subject so that specified primary levels are always achieved, taking into account any effects of earcanal size . Procedures For both groups of subjects, OAE data collection was not initiated until hearing sensitivity had been evaluated . Depending upon subject group, this meant that a pure-tone audiogram or an ABR had been obtained prior to OAE testing and the results were known to the tester. Although middle-ear dysfunction was not present in any of the older subjects, it was not possible to be equally certain about middle-ear status for the newborns . However, the fact that these babies had ABRs at least as low as 30 dB nHL and normal wave I latencies for high-level stimuli reduces the likelihood that significant middle-ear dysfunction was present. Tests were performed on a randomly selected ear in a quiet, although not soundtreated, environment . Older subjects were seated comfortably, while neonates were sleeping in open cribs. While data related to hearing sensitivity were always obtained first, the order of OAE measurements was alternated . For all subjects, check-fitting procedures were performed prior to TEOAE measurements in an effort to obtain a reasonably flat stimulus spectrum in the ear canal. Recall from Figure 1 that this was more easily achieved in older subjects compared to infants. For DPOAE measurements, data collection began only after calibration procedures indicated that the two earphones produced reasonably flat spectra, which were similar to each other. The amount of time devoted to probe fitting for either OAE measurement was not captured . RESULTS Broadband TEOAE Responses Figure 2 shows mean TEOAE amplitude (circles), background noise (squares), TEOAE Journal of the American Academy of Audiology/Volume 6, Number 2, March 1995 Figure 2 Mean TEOAE amplitude (circles), background noise (squares), TEOAE amplitude/noise (triangles), and percent reproducibility (diamonds) as a function of frequency in both older subjects (left column) and in neonates (right column) . All amplitudes are equivalent dB SPL in a bandwidth of 10 kHz . 30 1 1 -o O - TEOAE Level o - No" _,o 30 A - TEOAE/Noise--- ---- -I - 1s -w loo w ii i i 0 - ti Relw o&xibility Older Children 6 Adults soo 1000 2000 4000 ICN Graduates Be soo 1000 2000 4000 so Frequency (Hz) amplitude/noise (triangles), and percent reproducibility (diamonds) as a function of frequency in both older subjects (left column) and in neonates (right column). Error bars represent ± 1 standard deviation. The results of broadband response analyses, provided by the IL088, are shown at the far right of each panel. Although the broadband TEOAEs were larger in neonates, estimates of background noise also were larger by a comparable amount (top row), resulting in TEOAE/noise that were more similar across subject groups (middle row) . Estimates of response reproducibility also were comparable across the two groups, which is an expected outcome, given the close relationship between TEOAE/noise and percent reproducibility (Gorga et al, 1993b; Prieve et al, 1993) . The observation of larger noise amplitudes in neonates may be due to their greater breathing noise and their more rapid breathing pattern . Higher noise levels also may be a consequence of poorer probe fit in 154 neonatal ear canals, allowing more ambient room noise to be picked up by the microphone . Octave Band Analyses of TEOAE Waveforms Asomewhat different pattern emerged when broadband responses were analyzed in octave bands, which is also shown in Figure 2 . TEOAE amplitudes in newborns were larger for all bands but especially for the two octave bands centered at 2000 and 4000 Hz . While ICN graduates showed greater background noise for all bands, noise levels and variability were much greater in the lower frequency bands (top row) . As a consequence, within the group of older subjects, TEOAE/noise were larger for the two octave bands centered at 500 and 1000 Hz . In contrast, TEOAE/noise were larger for newborns in the octave bands centered at 2000 and 4000 Hz (middle row) . As expected, a similar pattern was observed for response reproducibility (bottom OAEs from ICN Graduates/Berg-man et al O - I Figure 3 Mean 1/3-octave band TEOAE amplitude (circles), background noise (squares), TEOAE amplitude/noise (triangles), and percent reproducibility (diamonds) as a function of frequency in both older subjects (left column) and in neonates (right column). All amplitudes are equivalent dB SPL in a bandwidth of 10 kHz. TEOAE Level s 0 f -30 a - 30 TEOAE/wee J IS . i.a- . .L 0 -1s -30 too 60 so .0 O <r 20 0 0 - K -20 I Rpreducibility 0 -40 -60 Older Children & Adults S00 1000 2000 ICN Graduates 4000 S00 1000 2000 4000 Frequency (Hz) row), although the high-frequency effects are not as clear due to ceiling effects on this response measure. That is, differences in response reproducibility decrease as the signal-to-noise ratio increases beyond about 10 dB . The variability of the OAE/noise measures tended to be less in the low frequencies and greater in the high frequencies because the low-frequency measures were dominated by the noise. '/s-Octave Band Analyses of TEOAE Waveforms As might be expected, a virtually identical pattern was observed when the broadband responses were analyzed in Y3-octave bands, which is shown in Figure 3. TEOAE amplitude and background noise overlap in newborns for low-frequency regions, whereas some separation exists in the data from older subjects (top row) . As a result, positive TEOAE/noise are observed for lower frequencies in older subjects but not in infants (middle row) . However, greater separation in neonates clearly exists between TEOAE amplitude and the background noise for higher frequencies, resulting in larger high-frequency TEOAE/noise for this group. These patterns are reflected in estimates of response reproducibility (bottom row), although perhaps less clearly due to the previously described ceiling effects. Journal of the American Academy of Audiology/Volume 6, Number 2, March 1995 Figure 4 Means (+ 1 standard deviation) for DPOAE amplitude, noise amplitude (top row), DPOAE/noise (middle row), and derived correlation (bottom row) as a function of f2 frequency. Data for older subjects are shown in the left column while data from neonates are shown in OLDER CHILDREN & ADULTS 0 -00 I {{1iI°I °'` e o - rb» . o - OPO~E the right column . 1.1111111 f a- OPDX/N .i.. -30 too I so a 11 .. .I f0 K p L m 1 I p 0 0 -20 -~0 0 - D.riv .d r (2) -i0 Soo tow x000 4000 11000 7w t000 .Ow .ON Im Frequency (Hz) DPOAE Measurements Figure 4 shows means (±- 1 standard deviation) for DPOAE and noise amplitudes (top row), DPOAE/noise (middle row), and derived correlation (bottom row) as a function of f2 frequency. Data for older subjects are shown in the left column while data from neonates are shown in the right column . Using the present paradigm for measuring DPOAEs, it is not possible to obtain a direct cross-correlation comparable to that used with TEOAEs . However, it is possible to derive a correlation from signal-tonoise ratios using the following formula: r = (S - 1)/(S + 1) where S is defined as the ratio of DPOAE energy to noise energy (Mills, 1924). The mathematical relation between signal-to-noise ratios and con 156 relations (Mills, 1924) and its application to OAE data (Gorga et al, 1993b) has been described elsewhere. This equation was used to estimate a derived correlation for DPOAE measurements following one additional consideration. The present paradigm estimates the amplitude of the DPOAE at the 2f1 f2 frequency but estimates noise from the average amplitude of six points adjacent to 2f,42 (three above and three below), each spaced by 48 .8 Hz . If the amplitude at any of the six points is greater than the amplitude at 2f1-f2, negative DPOAE/noise may be observed . Obviously, there is no response under these conditions . However, if these values are used to calculate a derived correlation, large negative correlations will be observed . To avoid this problem, values of 0 dB were assigned to all conditions when the DPOAE/noise was less than 0 dB, resulting in a derived correlation of 0 for these conditions . This rule was used both for ICN graduates and for older subjects . Mean derived OAEs from ICN GraduatesBergman et al Octave TEOAE E. 0 0 0 ..e...'~ n Qan . ' 00 Figure 5 DPOAE 1/3 octave TEOAE .. . too " ". 0200. .0-0o:&seo :o 0 - ICN craWora. 0 - oldw CNdrw A A". i ".. .___ ta. " s. . . a ._._ . sr "f. Q`law. CPua °au. .Wa QO. .a ." ` bt] D : Q ~ . . :e " - Oldw CNdrm & Ad&@ b so a . . .. : : ;;... a' 1s 0 :: : .. . e a: n a° 6 . .. e.a.ea e The means from Figures 2, 3, and 4 are superimposed without error bars . Data from older subjects are shown as open symbols while data from ICN graduates are displayed with corresponding filled symbols. TEOAEs, analyzed into octave bands, are shown in the left column, along with the results of the broadband response analyses . These same data, analyzed into 1/3-octave bands, are shown in the middle column. For TEOAEs, all amplitudes are equivalent dB SPL in a bandwidth of 10 kHz . The DPOAE data are shown in the right column. a - ICN crosxeta. e - Ok1w artrw A Adu1M 00 0 00 Gradvotes "0" -- ICN o1a.r CMdrw s Adult. s00 1000 :000 .000 0 """ es s00 0.9 . " 1000 2000 .000 Froqu.ncy (1s) 0, a~ " 0 s00 correlations, using the above rules, are given in the bottom row of Figure 4. As with TEOAE measurements, both DPOAE amplitude and background noise tend to be larger in newborns compared to older normalhearing subjects, but the DPOAE/noise are much more comparable . Also, the DPOAE/noise measures show less variability in the low frequencies. The effect of frequency on mean responses, although present, is less obvious in these measurements compared to TEOAE measurements . Comparisons across OAE Measures All of the above observations are summarized in Figure 5, where the means from Figures 2, 3, and 4 are superimposed without error bars . Data from older subjects are shown as open symbols while data from ICN graduates are displayed with corresponding filled symbols. TEOAEs, analyzed into octave bands, are shown in the left column, along with the results of the broadband response analyses . These same data, 1000 2000 .000 x000 analyzed into 1/3-octave bands, are shown in the middle column . The DPOAE data are shown in the right column . First, let us consider the TEOAE data . Both TEOAE amplitude (top row) and the background noise (second row) are greater in newborns . However, the separation between these two measures is more similar across groups, resulting in virtually identical TEOAE/noise for broadband conditions ("BB", third row, left column). In contrast, both octave and 1/3-octave band analyses showed poorer signal-to-noise ratios for neonates in lower frequencies and larger ratios for higher frequencies (third row, left and middle columns). Broadband analyses tended to obscure the fact that neonates have larger TEOAE/noise in higher frequencies but lower TEOAE/noise for low-frequency conditions . Some of these effects are undoubtedly a consequence of the higher noise levels in neonates, which has its greatest energy in the lower frequencies . It is possible that including a greater number of samples in each averaged response 157 Journal of the American Academy of Audiology/ Volume 6, Number 2, March 1995 would reduce the noise floor sufficiently to result in more clearly observed low-frequency TEOAEs . One other possible explanation for these frequency effects is the fact that stimulus amplitudes were greater in the ears of older subjects in the low and mid frequencies compared to the amplitude achieved in infant ear canals (see Fig. 1) . These greater amplitudes might have elicited stronger responses in the ears of older subjects . However, in no circumstance was TEOAE amplitude greater in older subjects than in infants . Differences in these responses were more a function of differences in the level of background noise . Additionally, a relation between stimulus amplitude and response should be observed if stimulus amplitude is the source of these differences . However, correlations between TEOAE and stimulus amplitude within each of four octave bands for infants were not apparent . The strongest correlation between response and stimulus amplitude (r = .25) was observed at 500 Hz, which is the frequency at which a particularly poor signal-tonoise ratio was observed . At the three higher octave frequencies, the correlations were less and sometimes even negative . Although the range of stimulus amplitudes was not large, these data suggest that the present response differences are not the result of differences in stimulus amplitude . There is little doubt, however, that stimulus and response amplitudes are related (Stover and Norton, 1992). It is possible that the stimulus conditions used in the present study were on the asymptotic, saturated portion of the input/output function relating OAE amplitude to stimulus amplitude. Had a wider (lower) range of stimulus amplitudes been used, the influence of stimulus amplitude would have been observed . The above arguments are based on the premise that the response spectrum, measured in the ear canal, reflects the stimulus spectrum that reaches the cochlea. It is important to remember, however, that the measured OAE is influenced by both forward and reverse transmission characteristics. In a sense, the stimulus spectrum in the ear canal can be viewed as an estimate ofthe energy provided to the cochlea. Although direct estimates of reverse transmission (i .e ., from the cochlea to the external ear) are not available, modelling data have been used to suggest that the spectral shaping of broadband TEOAEs may be a consequence of the reverse transfer of energy (Kemp, 1980 ; Kemp et al, 1986). Thus, it is possible that both the energy provided to the cochlea (the forward 158 transmission as estimated by the stimulus spectrum in the ear canal) and the reverse transmission serve to attenuate low-frequency energy in neonatal ears . A slightly different picture emerged when DPOAE measurements from these two subject groups were overlapped, which is shown in the right column of Figure 5. DPOAE/noise data from normal-hearing older subjects and from presumably normal-hearing newborns are similar across a wide range of frequencies (third row), whereas TEOAE measurements revealed more obvious frequency-dependent effects. Cumulative Distributions Cumulative distributions may have some advantage over summary statistics, such as means and standard deviations . While means and standard deviations can be used to estimate any percentile, this holds only when the data on which these estimates are based are normally distributed. This is unlikely to be the case for OAE measurements, where both ceiling and floor effects potentially will skew the distributions for both normal-hearing and hearingimpaired ears . No such assumptions are necessary when cumulative distributions are used. Examples of this approach for TEOAE data are shown in Figure 6. The left and right columns represent data from older children/adults and neonates, respectively. The top four rows represent TEOAE data analyzed into octave bands centered at 500, 1000, 2000, and 4000 Hz, respectively, while the bottom row represents data analyzed as broadband responses. Within each panel, solid lines represent cumulative distributions of response amplitude while dashed lines represent similar data for background noise. At 500 Hz, there is only slight separation between response and noise cumulative distributions in older subjects and virtually no separation for the infant responses . At 1000 Hz, there is greater separation in older subjects, with less separation between response and noise cumulative distributions in infants. This trend reverses at 2000 Hz and especially at 4000 Hz, where the separation is much greater in the infant responses. These effects are a direct consequence of the interaction between frequency and the known behaviors of OAEs in infants compared to older subjects . Infant responses tend to be larger but so does the background noise. However, background noise is dominated by low-frequency energy in both groups of subjects . When the analysis is restricted to a OAEs from ICN GraduatesBergman et al important frequency effects. Thus, a narrowband analysis should allow for a better choice of pass/fail criteria that accurately discriminate impaired from normal ears . INFANTS OLDER CHILDREN I 500 Hz DISCUSSION he following observations can be made from T the present study: 1. >K d t v E U -10 o to 10 30 .o -to o Amplitude (dB SPL) to m ao w Figure 6 Cumulative distributions of response (solid lines) and noise (dotted lines) amplitudes for older subjects (left column) and neonates (right column). Top row shows 500-Hz data, with 1000-, 2000-, and 4000-Hz data in the second, third, and fourth rows, respectively. Broadband response analyses are shown in the bottom row. frequency region where the noise is low in amplitude for both groups (i .e ., 4000 Hz), then the larger responses of neonates result in larger signal-to-noise ratios . The previously described stimulus differences also may be factors contributing to these effects. Although not identical, the differences between infants and older subjects are not as obvious for the broadband condition . This occurs because both overall response and noise amplitudes are greater in infants, effectively offsetting each other. As noted above, however, these analyses obscure potentially TEOAEs and DPOAEs can be measured in graduates of the ICN with click-evoked' ABR thresholds of 30 dB nHL or better and with no evidence of middle-ear dysfunction, as well as in older children and adults with normal hearing sensitivity and normal middle-ear function . 2. Both OAE amplitudes and noise amplitudes are larger in newborns compared to older normal-hearing subjects, resulting in more similar OAE/noise for both DPOAE and TEOAE measures . Reproducibility was also similar between groups . 3. Subtle frequency differences were noted when TEOAEs were measured that were not as obvious in DPOAE measurements . Whereas TEOAE/noise measures were less in the low frequencies for newborns, TEOAE/noise were greater in the high frequencies . These low-frequency decrements and high-frequency increments were reduced or absent when DPOAE/noise from infants were compared to those seen in older subjects . 4. The slight differences in stimulus amplitudes between newborns and older subjects alone may not fully account for differences in response amplitudes . Our data are most comparable to those reported by Smurzynski et al (1993) . For example, they evaluated both DPOAEs and TEOAEs in neonates . They analyzed TEOAEs in frequency bands, much like the present study. The shape of their response amplitudes as a function of frequency also was similar to those observed presently. Direct comparisons of absolute TEOAE amplitudes, however, are not feasible, due to possible differences in the reference for dB SPL. Similarly, direct comparisons of noise amplitudes cannot be made . In contrast, direct comparisons between the two sets of DPOAE data are possible . Our estimates of DPOAE amplitude are similar to theirs, including a slight decrease in amplitude around 3000 Hz (although they observed a greater reduction in amplitude than Journal of the American Academy of Audiology/ Volume 6, Number 2, March 1995 was presently observed). The present estimates of background noise during DPOAE measurements, however, were greater than those observed by Smurzynski et al . In fact, their estimates of background noise are more similar to those we observed in older normal-hearing subjects . The reasons for some of these observations may be related to differences between studies in the criteria that were used for including data from individual subjects in the analyses . Smurzynski et al did not include data that were characterized by high levels of background noise, whereas we included data from all subjects regardless of noise levels . Our reasons for including the data from every baby who presumably had normal hearing (i .e ., passed an ABR test) reflects our interest in understanding how OAEs perform as a clinical test. Whether an infant fails an OAE test because of cochlear hearing loss, background noise, or even middle-ear dysfunction is perhaps irrelevant in terms of what happens next . That infant must either receive a repeat OAE test or move to the next level of care . One must include data from all babies (regardless of the reasons for a failure) if the goals of the program are to develop a screening paradigm that is universally applicable . It is important to consider some of the frequency effects observed in the present study in relation to the clinical application of OAEs . Depending upon the stimulus paradigm, both TEOAEs and DPOAEs can be measured very rapidly and thus may be well suited as efficient screening measures for perinatal hearing loss . The vast majority of educationally significant sensorineural hearing loss involves at least the higher frequencies . Low-frequency sensorineural hearing loss in the presence of normal high-frequency hearing is much less common . Rehabilitative options, including the use of personal amplification, are greater for hearing losses involving higher frequencies. Indeed, it is less likely that any intervention will occur in the perinatal period, even if low-frequency hearing loss was identified and quantified, as long as normal hearing was present in the mid to high frequencies. Given these facts, the observation that OAEs in neonates are more easily measured in the higher frequencies but difficult to measure at lower frequencies may have little negative clinical consequence . Whatever the reasons for poorer OAE/noise in the low-frequency responses of neonates (i .e ., reduced stimulus amplitudes due to leakage or ear canal resonances, greater ambient noise due to poor fit or greater breath160 ing noise), the present data suggest that OAEs may provide useful information for the frequency region over which such information may be of the greatest clinical importance . Regardless of the frequency region(s) for which predictions about auditory status are to be made, cumulative distributions of response properties perhaps provide the most straightforward approach to selecting pass/fail criteria . Distributions of response properties from normal ears would allow selection of response criteria to achieve any desired false-alarm (false positive) rate . In a patient population where the incidence of hearing loss is very low (i .e ., the wellbaby nursery), the cost of the test may be driven by the false-alarm rate . Under these circumstances, it might be acceptable, at least to a first approximation, to develop test criteria based solely on data from normal-hearing infants. Cumulative distributions from impaired ears (not presently available) would allow the selection of response criteria resulting in any hit (true positive) rate . Selection of test criteria probably should be driven more by this information if the patient population was more likely to include patients with hearing loss, such as those being evaluated through hearing clinics and/or otolaryngology departments. In the absence of data from hearing-impaired infants, we have not specified optimal response criteria ; however, the data presented here suggest that best OAE test performance will occur for mid to high frequencies, much like ABR test performance . Thus, cumulative distributions of response properties for normal and impaired ears allow one to select criteria resulting in any combination of both hit and false-alarm rates over the range of levels of performances for a particular test. This approach has been applied to data from older subjects (see, for example, Fig. 5 from Gorga et al, 1993a and Fig. 6 from Prieve et al, 1993). Unfortunately, these previous data may be of limited applicability with neonates because their response properties apparently differ from those of older subjects . Thus, the present data can be used to tell only one part of the story. Similar cumulative distributions from infants with hearing loss are needed in order to fully describe OAE test performance in a neonatal population. It is important to recognize some additional limitations of the present set of data . First, no special efforts were undertaken to improve signal-to-noise ratios, such as increasing the number of averages or restricting the analyses to more favorable frequency regions. For DPOAEs, OAEs from ICN GraduatesBergman et al data were collected without the use of an artifact rejection system . Improved signal-to-noise ratios most likely would have been observed had such a system been used . A system that incorporates measurement-based stopping rules might also result in more efficient data collection and perhaps extend the range of frequencies over which reliable data can be measured without unacceptable increases in test time . Another possible limitation of these data is related to the fact that middle-ear dysfunction cannot easily be objectively measured in neonates . Although a sizable conductive hearing loss would be ruled out by ABR thresholds of 30 dB nHL and normal wave I latencies for highlevel stimuli, a slight conductive component could potentially be present. Trine et al (1993) demonstrated that negative middle-ear pressure affected the amplitude, reproducibility, and spectral content of TEOAEs . Indeed, since the amplitude of OAEs is typically 10 dB SPL or less, an air-bone gap of 15 to 20 dB is likely to provide sufficient attenuation to reduce or eliminate the OAE. Most importantly, data are needed from hearing-impaired infants. As stated above, a description of OAE behaviors in infants with normal hearing provides only a portion of the information that is needed in order to determine the ability of these measures to identify hearing loss . Data are needed from infants with hearing impairment in order to describe the distributions of responses from both normal and impaired ears and to determine the extent to which data from an individual baby can be correctly assigned to one of these groups . Hopefully, ongoing research efforts will provide these data, thus providing a more quantitative appraisal of OAE test performance as a screening measure for perinatal hearing loss . Acknowledgments. This work was supported in part by NIH grants DC00982 and DC01958. We would like to thank Linda Mace for her help in the preparation of the manuscript and its accompanying figures. We also thank Drs. Lisa Stover and Edward Walsh for their comments on an earlier version of this paper as part of our internal review process. REFERENCES Bonfils P, Avan P, Francois M, Trotoux J, Narcy P. (1992) . Distortion-product otoacoustic emissions in neonates : normative data . Acta Otolaryngol (Stockh) 112 :739-744 . Bray P, Kemp DT. (1987) . An advanced cochlear echo technique suitable for infant screening . Br J Audiol 21 :191-204 . Ellsmann SF, Matkin ND, Sabo MP. (1987) . Early identification of congenital sensorineural hearing impairment . Hear J 9 :13-17 . Gorga MP, Neely ST, Bergman BM, Beauchaine KL, Kaminski JK, Peters J, Jesteadt W. (1993a). Otoacoustic emissions from normal-hearing and hearing-impaired subjects : distortion product responses . JAcoust Soc Am 93 :2050-2060 . Gorga MP, Neely ST, Bergman BM, Beauchaine KL, Kaminski JK, Peters J, Schulte L, Jesteadt W (1993b). Acomparison of transient-evoked and distortion-product otoacoustic emissions in normal-hearing and hearingimpaired subjects . JAcoust Soc Am 94 :2639-2648 . Johnsen NJ, Bagi P, Elberling C . (1983) . Evoked acoustic emissions from the human ear. III . Final results in 100 neonates . Scand Audiol 12 :17-24 . Kemp DT . (1978) . Stimulated acoustic emissions from within the human auditory system . J Acoust Soc Am 64 :1386-1391 . Kemp DT. (1980) . Towards a model for the origin of cochlear echoes . Hear Res 2 :533-548. Kemp DT, Bray P, Alexander L, Brown AM . (1986) . Acoustic emission cochleography - practical aspects. Scand Audiol Suppl 25 :71-83 . Kemp DT, Ryan S. (1991). Otoacoustic emission tests in neonatal screening programmes. Acta Otolaryngol Suppl (Stockh) 482 :73-84 . Kok MR, van Zanten GA, Brocaar MP. (1992) . Growth of evoked otoacoustic emissions during the first days postpartum - a preliminary report . Audiology 31 :140-149 . Lafreniere D, Jung MD, Smurzynski J, Leonard G, Kim DO, Sasek J. (1991) . Distortion-product and click-evoked otoacoustic emissions in healthy newborns . Arch Otol Head Neck Surg 117:1382-1389 . Lasky R, Perlman J, Hecox K. (1992) . Distortion-product otoacoustic emissions in human newborns and adults . Ear Hear 13 :430-441 . Mills FC . (1924) . Statistical Methods. New York : Henry Holt . Norton SJ, Widen JE . (1990) . Evoked otoacoustic emissions in normal-hearing infants and children : emerging data and issues . Ear Hear 11 :121-127 . American National Standards Institute. (1989) . Specifications forAudiometers . (ANSI S3 .6-1969, R 1973) New York : ANSI. Prieve BA, Gorga MP, Schmidt AR, Neely ST, Peters J, Schulte L, Jesteadt W. (1993) . Analysis of transientevoked otoacoustic emissions in normal-hearing and hearing-impaired ears . JAcoust Soc Am 93 :3308-3319 . Allen JB . (1990). User Manual for the CUBDIS Distortion Product Measurement System . Unpublished manual . Smurzynski J, Jung MD, Lafreniere D, Kim DO, Kamath MV, Rowe JC, Holman MC, Leonard G. (1993). Distortion- 161 Journal of the American Academy of Audiology/ Volume 6, Number 2, March 1995 product and click-evoked otoacoustic emissions of preterm and full-term infants. Ear Hear 14 :258-274. Smurzynski J, Kim DO . (1992) . Distortion-product and click-evoked otoacoustic emissions of normally-hearing adults . Hear Res 58 :227-240 . Stevens JC, Webb HD, Hutchinson J, Connell J, Smith MF, Buffin JT. (1991) . Evaluation of click-evoked otoacoustic emission in the newborn . Br JAudiol 25 :11-14. Stover LJ, Norton SJ . (1992) . Comparisons among different otoacoustic emission types . Abstracts, Ass Res Otolaryngol 15 :153. Trine MB, Hirsch JE, Margolis RH . (1993) . The effect of middle ear pressure on transient evoked otoacoustic emissions . Ear Hear 14 :401-407 . Uziel A, Piron J. (1991) . Evoked otoacoustic emissions from normal newborns and babies admitted to an intensive care baby unit . Acta Otolaryngol Suppl (Stockh) 482:85-91 . White KR, Vohr BR, Behrens TR . (1993) . Universal newborn hearing screening using transient evoked otoacoustic emissions: results of the Rhode Island hearing assessment project. Semin Hear 14 :18-29 .