Estimating Hearing Thresholds With Evoked Potentials: ABR, ASSR, MLR and Cortical Responses

In the clinic, hearing is usually measured with behavioral audiometric tests that require the person’s active participation. Yet there are situations where these subjective tests cannot be performed: infants and young children too young for behavioral testing, people with intellectual disability or communication disorders, neurological disease, comatose patients, medicolegal cases and nonorganic hearing loss. In all of these, the primary tool for assessing hearing objectively is the auditory evoked potential.

With the spread of newborn hearing screening programs, most children are referred to a full audiological evaluation on the basis of a screening result, and hearing is assessed in infancy. To estimate the behavioral audiogram, frequency-specific auditory brainstem response (FS-ABR) and auditory steady-state responses (ASSR) are used first and foremost, along with middle latency responses (MLR) and cortical auditory evoked potentials. However, the thresholds obtained with these potentials do not coincide exactly with behavioral thresholds; correction factors are needed to close the gap.

A “threshold” obtained with evoked potentials is not a raw number; to be interpreted correctly it must pass through a correction process that jointly accounts for the stimulus, transducer, age and calibration.

Why are correction factors needed?

Evoked potentials and behavioral thresholds come from different levels. Auditory evoked potentials arise from the brainstem and various regions of the brain, whereas the behavioral threshold is obtained at cortical and consciousness levels that involve a decision to respond to a stimulus. Corrections are therefore needed to bring the responses obtained from potentials closer to behavioral thresholds. A second reason is calibration: evoked-potential systems are calibrated to adult data, so the stimuli used in infants and young children must be transformed to better represent this population. Third, a separate correction must be used for each transducer, since transducers deliver stimuli differently.

Anatomical differences sharpen this picture. Because infants have a smaller ear canal, the sound level delivered to the canal by an insert earphone differs from that in adults; studies report that this difference can reach 5–8 dB and that the level delivered is significantly higher at 500–4000 Hz⁴⁵. For bone conduction, the infant’s smaller head means the bone vibrator drives a smaller mass, so the stimulus is delivered at a higher level; age-dependent corrections are used for this reason. By contrast, no significant adult–infant difference has been found with supra-aural earphones.

There are different stimulus options for recording: click, tone burst/tone pip, frequency-specific tonal stimuli and the increasingly common chirp stimuli. These differ both from one another and from the stimuli of pure-tone audiometry; the stimulus used must therefore be taken into account when interpreting the result.

Threshold estimation with the ABR: from nHL to eHL

Thresholds estimated with the ABR are typically higher than behavioral thresholds; the difference can reach 20–30 dB depending on frequency, which is important for intervention decisions. Frequency-specific correction factors are therefore applied to ABR thresholds to better estimate the behavioral threshold; to distinguish the result from nHL, it is expressed as eHL¹. In some systems the correction is embedded in the calibration; in those without it, corrections must be applied manually before the thresholds are used.

An important reason the thresholds come out higher is temporal summation. The ABR requires short-duration stimuli; clicks are used for screening and diagnosis, and short tonal stimuli for frequency-specific measures. Because the auditory system perceives brief sounds at higher levels, thresholds also rise when these stimuli are measured behaviorally. The ASSR, by contrast, uses continuous, long-duration stimuli.

The main factors affecting the nHL-to-eHL correction are system calibration values, stimulus parameters (tone burst, click, chirp), transducer type (air or bone conduction) and the age at which the ABR is performed. The correction can be applied in two ways: the same fixed correction for all degrees of loss, or a correction dependent on the degree of hearing loss. Hearing loss has been shown to affect the correction factors³.

All of these factors make it unlikely that correction factors are universal; indeed, a recent study showed that behavioral-threshold predictions and the required corrections can vary markedly between different ABR systems¹⁵. Each clinic should therefore verify the corrections specific to its own device and protocol. In addition, the spectrum of infant ABR responses is dominated by low-frequency energy (much of it below 100 Hz); a 30 Hz high-pass filter is recommended during recording.

Threshold estimation with the ASSR

The ASSR builds up electrophysiological thresholds at different frequencies quickly and objectively; because it can assess both ears and several frequencies simultaneously, it shortens test time. Threshold estimation is affected by conditions such as stimulus frequency, degree of loss, age and test duration. The narrow spectrum of continuous sinusoidal stimuli provides at least as good — and often better — frequency specificity than tone-burst ABR; because the modulated tones resemble behavioral warble stimuli, they can be calibrated to pure-tone standards and thresholds obtained directly in dB HL.

However, ASSR thresholds (in dB HL) are markedly higher than behavioral thresholds, especially in young infants, and this relationship is not fully understood; infant ASSR thresholds must therefore be interpreted cautiously as “dB HL,” and regression formulas or correction factors should be used to convert them to an estimated behavioral threshold. In people with hearing loss, the mean difference between ASSR and behavioral thresholds is about 7 (±5) dB⁸⁹. This difference decreases as loss becomes more severe; it is thought to result from recruitment due to outer-hair-cell damage, which abnormally increases physiological response amplitudes at suprathreshold levels. In infants and young children, ASSR thresholds are about 10–15 dB higher than in adults across all frequencies and about 10 dB higher than in older children⁶⁷.

500 Hz is the most challenging frequency across all methods: responses to low-frequency stimuli are smaller and more variable owing to neural asynchrony and greater intrinsic delay (jitter), broader basilar-membrane activation and immature frequency-to-place mapping in infants, poor acoustic coupling between the earphone and the ear, higher background noise at low frequencies, and immature processing of fast modulation rates. As a result, 500 Hz thresholds are obtained 5–10 dB worse than at other frequencies and the risk of error rises for mild loss; longer recording times improve accuracy¹⁰.

Two developments in stimulation and analysis stand out in recent years. First, narrow-band CE-Chirp stimuli: by compensating for cochlear delay they produce larger and more synchronous responses; responses about 31% larger at 500 Hz and about 52% larger at 2 kHz than tone pips have been reported, with high agreement with behavioral thresholds including at 500 Hz¹⁶. Second, the presence of a response is now judged not by eye but with statistical tests (F-test, q-sample, mutual-information–based methods); this objectifies decisions, including when to stop recording near threshold. Recent reviews emphasize that the ASSR is reliable in children and that recent work is better able to distinguish mild loss from normal hearing¹⁴.

Middle latency responses (MLR)

The MLR wave series, appearing 12–50 ms after a click, arises primarily from neurons at the thalamic and cortical level — the posteromedial region of Heschl’s gyrus. It has three advantages for threshold estimation: the amplitude of the Pa wave is markedly larger than the ABR’s wave V (better signal-to-noise); it is easily recorded with long-duration frequency-specific tone bursts and gives more reliable responses at low frequencies than the ABR; and it can use the same instrumentation and electrode montage as the ABR.

The MLR was frequently used in pediatrics in the past but lost popularity when the ABR — thanks to being unaffected by sleep and sedation — became widespread in the 1970s; the MLR is markedly affected by age, neurological immaturity and movement. It continued to be used in adults for suspected functional/nonorganic loss, and its use declined further with the advent of the ASSR. The Pa wave is generally used as the response index; the lowest level at which Pa is obtained is about 10 dB above the behavioral threshold at that frequency. When myogenic noise is low, the MLR can be more reliable than the ABR; in patients with disrupted neural synchrony, MLR responses can be recorded even when the ABR is abnormal or absent¹³.

Cortical responses: closest to the gold standard in adults

Cortical auditory evoked potentials (CAEPs) reflect the processing of a stimulus in the auditory cortex. For objective threshold measurement — particularly in adults — auditory late responses (ALR) are recommended as the most suitable method, because they demonstrate audibility at the cortical level without the listener taking an active role. Reasons they agree better with behavioral thresholds include their resistance to myogenic activity, the high frequency specificity of long-duration tonal stimuli, calibration of stimuli to pure-tone standards, and responses arising close to the cortical and consciousness level.

CAEPs have been reported to estimate thresholds with an accuracy of 10–20 dB. Lightfoot and Kennedy found a mean error of 6.5 dB for the N1-P2, independent of frequency, and Ross et al. found 7.5 dB; after correction, 94% of estimates are within 15 dB and 80% within 10 dB of the behavioral threshold¹¹. Cesur and Derinsu compared frequency-specific thresholds from ALR and behavioral audiometry in normal-hearing and sensorineural-hearing-loss groups and reported that the results largely agreed within ±10 dB, with agreement of about 5–7.4 dB with behavioral thresholds¹².

CAEP thresholds can come out worse than the pure-tone average; this may be due to stimulus calibration, recording/analysis methods or comorbidities. A clinician can determine the deviation values in their own clinic and apply corrections, but the report must state at which frequency and by how many dB the correction was made. The main limitation is test time; recent work focuses on rapid cortical audiometry protocols and fully objective response detection to overcome this limit¹⁷. On the other hand, CAEPs are not suitable for threshold estimation in infants; owing to maturation, the P1 recorded at suprathreshold levels does not reliably reflect auditory sensitivity — in newborns a mean P1 threshold of about 25–30 dB nHL has been reported.

Conclusion

Evoked potentials are the primary tool for estimating hearing threshold objectively when behavioral testing is not possible — but the resulting number must pass through corrections that depend on the stimulus, transducer, age and calibration to be interpreted correctly. It is clear in today’s evidence that correction factors are not universal, that they can vary between systems and that each clinic must verify its own values; that chirp stimuli and objective response detection improve accuracy; and that 500 Hz remains the most challenging frequency. eHL is an estimate; not using it as a “measured threshold” is the foundation of sound clinical decisions.

Author

Alperen Akbulut

Audiologist · Audiology PhD(c) · Founder of İşitme Atölyesi

After completing his undergraduate degree in Audiology at Istanbul University, he earned a master’s degree in the Audiology and Speech Disorders program at Marmara University, where he is continuing his doctoral studies. From 2019 to 2021 he worked as a clinical specialist at Advanced Bionics; since 2021 he has been a lecturer in the Department of Audiology at the University of Health Sciences, Hamidiye Faculty of Health Sciences. His research focuses on music perception, music-related quality of life and the music–memory relationship in cochlear implant users; his work has appeared in journals such as Ear and Hearing and European Archives of Oto-Rhino-Laryngology. He is the founder of İşitme Atölyesi, which aims to communicate audiology in an evidence-based and accessible way.