Share this article on:

Sound Advice for ABR Analysis: No Squinting or Guessing

Smith, Joanna T. MS; Wolfe, Jace PhD

doi: 10.1097/01.HJ.0000451365.01338.5b
Tot 10

Ms. Smith, left, is cofounder and executive director of Hearts for Hearing in Oklahoma City.

Dr. Wolfe is director of audiology at Hearts for Hearing and an adjunct assistant professor at the University of Oklahoma Health Sciences Center and Salus University.

Figure 1

Figure 1

The stakes are high. You miss, and auditory brain development may be forever hampered. On the other hand, erroneously diagnosing a hearing loss leads to treatment that may be detrimental to a child's auditory function.



Without a doubt, accurate auditory brainstem response (ABR) evaluation in infants necessitates a clinician who is savvy in the science and the art of the assessment.

Last month's installment of the Tot Ten discussed techniques for acquisition of the ABR in infants ( HJ May 2014, p. 36 This month, we aim to provide sound advice for ABR interpretation.



Again, we dedicate these columns to experts who serve as our mentors in auditory electrophysiological assessment, including, but not limited to: David Stapells, Jay Hall, Guy Lightfoot, Michael Dennis, Richard Talbott, Yvonne Sininger, Linda Hood, Chuck Berlin, Michael Gorga, and Martin Hyde.

The Hearts for Hearing ABR protocol has been shaped by the excellent guidelines developed in Canada (British Columbia and Ontario, the United Kingdom, and Australia

Back to Top | Article Outline


Figure 2

Figure 2

ABR waveforms can be categorized as clear, absent, or inconclusive responses. Figures 1 and 2 provide examples of clear and absent responses, respectively.

Many clinicians only use these first two descriptors to categorize ABR waveforms, but the third option is very important.

Figure 3

Figure 3

Figure 3 shows waveforms that are best described as inconclusive. As indicated by the red arrow, there is activity that might be a response, but the waveforms do not have repeatable peaks that are significantly greater in amplitude than the averaged EEG activity.

Additionally, we may encounter cases in which the residual noise is high enough to potentially mask a threshold response.

In both these cases, the most prudent conclusion is to state that we are not certain whether a response exists.

Back to Top | Article Outline


The expected latencies of infant ABR waveforms are frequency specific. We are most concerned with having a latency range for which we can expect to see wave V, since this component is often the only one that's identifiable down to threshold.

Wave V possesses a similar latency for a 4,000-Hz tone burst as for a click-evoked ABR. For an infant with normal hearing, the response is often obtained between 6.5 and 10 ms at a presentation level of 25 dB nHL.

At moderate stimulus levels (e.g., 65-75 dB nHL), wave V typically falls between 6.0 and 9.0 ms.

The expected latency ranges for 500-Hz, 1,000-Hz, and 2,000-Hz tone-burst frequencies are 12.0 to 16.0 ms, 10.5 to 14.5 ms, and 8.5 to 11.5 ms, respectively, at a presentation level of 25 dB nHL.

Back to Top | Article Outline


Infant tone-burst ABR morphology is both frequency- and level-specific. High-frequency tone-burst (e.g., 2,000 and 4,000 Hz) ABR waveforms have a similar morphology to that of a click-evoked ABR, with relatively robust waves I, III, and V.

Near threshold, infant ABR wave V is best described as a downward-going trough, negative slope, or displacement from the baseline of the response.

Figure 4

Figure 4

Figure 4 provides an example of ABR waveforms obtained with a 500-Hz tone burst presented near threshold to an infant with normal auditory function.

At low and moderate presentation levels, low-frequency ABR responses typically only possess a rounded, relatively indistinct, negative slope within the expected latency range.

In the interpretation of low-frequency responses, the use of an expected shape for a clear response must be reinforced by the repeatability of the waveform and the presence of the response within the expected latency range.

At high presentation levels (e.g., greater than 60 dB nHL), the 500-Hz tone-burst response may possess a more prominent wave V peak, with multiple other peaks appearing in the waveform.

Back to Top | Article Outline


When ABR responses are robust, such as at presentation levels well above threshold, it is not entirely critical for noise levels to be low.

However, in the evaluation of threshold responses, the noise level must be quite low in order to confidently identify the response and exceptionally low in order to conclude that no response is present.

We attempt to average until the noise level is below 25 nV before concluding that a response is absent, as recommended in the UK guidance and the following reference: Stapells DR. Frequency-specific threshold assessment in young infants using the transient ABR and the brainstem ASSR. In: Seewald R, Tharpe AM, eds. Comprehensive Handbook of Pediatric Audiology. San Diego, CA: Plural Publishing; 2011: 409-448.

Estimated residual noise levels are calculated differently across ABR systems from different manufacturers. The UK guidelines include manufacturer-specific information.

The UK guidelines also note that noise levels may be inferred from the average gap between a pair of superimposed waveforms recorded at the same presentation level.

Figure 5

Figure 5

When such waveforms possess a small gap, as in figure 5, noise levels are sufficiently low. In contrast, when there is a high degree of dissimilarity, as in figure 6, then it may be concluded that noise levels are high.

Figure 6

Figure 6

Examination of figure 6 reveals that certain portions of the waveforms are overlapping and quite similar, while a large amount of separation exists within other segments of the waveforms.

A good estimate of the noise level may be obtained by finding the place where there is the maximum amount of separation and multiplying the magnitude of that separation by one-third (personal correspondence with Dr. Lightfoot).

When appropriate steps have been taken to reduce noise, EEG noise levels can be averaged to below 25 nV with fewer than 1,000 sweeps. For noisier patients, we may have to average to 2,000 sweeps or beyond (e.g., 4,000-6,000 sweeps).

Back to Top | Article Outline


While there is no unanimous rule stating how much larger the ABR wave V should be relative to the EEG noise, most experts suggest that the response should be 2.5 to 4 times greater.

At Hearts for Hearing, if the response is at least three times greater than the noise, then it is not necessary to continue averaging until the residual noise reaches 25 nV.

Back to Top | Article Outline


Auditory brainstem responses that are recorded at the same presentation level should be highly replicable.

If we are uncertain whether a response exists after collecting two waveforms, then we run a third waveform. If there is a true physiologic response, then the third waveform should be almost identical to the first two. Any differences among the waveforms are noise.

In addition, we may also use digital analysis to add the waveforms together, yielding a collective response that, if a response does exist, should possess a wave V. This result corroborates the presence of the wave V in individual waveforms.

Finally, for threshold-level response, we must be certain that the display scale allows for visualization of low-level responses.

Again turning to the UK guidelines, we set the display scale so that 0.1 µV on the vertical axis (i.e., ABR amplitude) corresponds to the same distance as 1 ms on the horizontal axis (e.g., time scale).

If we suspect that a low-level response exists, then we “zoom in” so that.05 µV corresponds to the same distance as 1 ms.

Back to Top | Article Outline


To minimize the margin for potential error, we suggest using a bracketing approach in auditory brainstem response similar to that used in behavioral audiometry.

We begin the ABR assessment with a 2,000-Hz tone burst presented at 60 dB nHL and then seek to obtain a response at 30 dB nHL in order to rule out a significant hearing loss.

If no response is present at 30 dB nHL—that is, a probable hearing loss exists—or if there is a questionable response, then we increase the presentation level 10 dB above the suspected threshold to confirm the presence of a clear response.

For infants with hearing loss, we collect two waveforms at a presentation level that is 10 dB below the response threshold.

Some experts recommend the completion of two “response-absent” waveforms. However, if the sufficiently averaged EEG response clearly shows no response with an objective residual noise level less than 25 nV, then it is highly unlikely that a response would be identified by an additional waveform.

We do not obtain waveforms 10 dB below the response threshold level if we have collected clear responses at levels ruling out a significant hearing loss greater than 25 dB nHL.

Back to Top | Article Outline


For ABR testing, there is potential for crossover to the opposite ear when bone-conduction testing is used or when there is a large difference between air-conduction thresholds obtained for the two ears.

To limit the likelihood of crossover, the clinician may routinely present 50-55 dB nHL of broadband noise to the opposite ear, which is unlikely to affect the response obtained on the test ear and should sufficiently mask the opposite ear.

If a more refined approach to determining masking necessity and levels is desired, the reader is referred to Guy Lightfoot's excellent summary called “Masking the ABR,d.b2k.” Dr. Lightfoot's masking calculator spreadsheet is described in the article.

Two-channel recordings should be made during bone-conduction testing so that the latencies of the responses may be compared between the two channels. The latency of wave V should be earlier for the channel that corresponds to the responding ear.

Back to Top | Article Outline


Since we set up the parameters within our ABR protocols to be similar to those used by David Stapells and colleagues in generating correction factors for the estimation of behavioral threshold from tone-burst ABR thresholds, we feel comfortable using the same correction factors: -15, -10, -5, and 0 dB at 500, 1,000, 2,000, and 4,000 Hz, respectively.

For children who have no response present at equipment limits, we subtract the frequency-specific correction from the highest level we assessed at each tone-burst frequency.

Back to Top | Article Outline


To read more about auditory brainstem response interpretation, we recommend the following publications:

The clinician who is new to infant ABR assessment should seek mentorship from an audiologist seasoned in the art and science of ABR testing in young children. Don't fly solo until you are certain you can navigate turbulent skies.

Even better, establish a local peer review network in which valuable constructive feedback is obtained on your practice.

Copyright © 2014 Wolters Kluwer Health, Inc. All rights reserved.