Tone-burst elicited auditory brainstem responses in full-term and pre-term infants

Abstract

Objective

To investigate the reliability of tone-burst auditory brainstem response (TB-ABR) latencies and thresholds in pre-term and full-term infants.

Design

TB-ABRs to 500 Hz and 4000 Hz tone-burst stimuli were recorded at two-week intervals over the first six weeks of life in a group of full-term infants (40–46 weeks gestational age [GA]) and over ten weeks (36–46 weeks gestation) in a group of preterm infants. Linear mixed model analyses evaluated within-subject changes and the effects of the subject group, age at assessment, and stimulus frequency on ABR latency and threshold.

Study sample

Twenty-four infants participated. Nine were full-term (GA: 39–41 weeks) and fifteen were healthy pre-term (GA: 28–34 weeks) at birth.

Results

TB-ABR wave V latencies at 70 dBnHL decreased throughout the study (p < 0.001) in pre-term babies for both test frequencies by approximately 0.5 ms. There were, however, no group or GA (at birth) effects indicating that response latency normalized in these children by the due date. Similarly, TB-ABR threshold levels in the premature group were elevated (p = 0.001) by approximately 5 dB in pre-term evaluation but were equivalent to those of full-term participants in the post-term assessment period.

Conclusions

In healthy, pre-term infants, tone-burst ABR testing is reliable from 36 weeks gestation.

Keywords:

• Auditory brainstem response
• hearing
• premature
• pre-term
• maturation

Introduction

Early detection of infant hearing impairment has direct implications for long-term speech and language outcomes (Kennedy et al. 2006; Moeller 2000; Yoshinaga-Itano et al. 1998). Even “minimal” deficits including mild degree- and unilateral hearing loss have been associated with significant psychosocial abnormality, poorer health-related quality of life, and compromised academic performance (Bess et al. 1998; Borton et al. 2010; Porter et al. 2013). Identifying significant hearing loss early relies on an efficient newborn screening program and thorough diagnostic follow-up.

Measurement of evoked responses from the auditory pathway is a commonly used assessment approach for infant subjects. In particular, techniques based on auditory brainstem response (ABR) which is reliably recordable in newborns, have achieved widespread clinical application in the last 40 years. The ABR is a series of scalp-recorded electrical potentials occurring in 10–20 ms immediately following the presentation of an abrupt auditory stimulus such as an acoustic click or tone burst. Identification of the response within the electroencephalogram requires synchronous neural firing in the auditory pathway which is best generated by stimuli with rapid rise times. Acoustic clicks, by their near-instantaneous onset, generate the clearest ABR responses, but contain energy across a broad frequency range and only reflect hearing in the 1000–4000 Hz region (Stapells 1994). Similarly, broad-band “CHIRP” stimuli, which are designed to compensate for basilar-membrane-related time delays, may elicit large amplitude brainstem responses but represent the combined response from different frequency regions (Bargen 2015). As successful management (and intervention if warranted) requires test stimuli that can provide frequency-specific estimates of sound detection across the audiometric range, acoustic clicks, and broad-band CHIRPs have limited diagnostic value. Higher frequency specificity is offered by tone-burst stimuli (Stapells et al. 1990, 1995) which show good correspondence between ABR threshold, the lowest presentation level that can elicit a repeatable response, and behavioral hearing levels for a range of test frequencies in both normal and hearing-impaired subjects (Gorga et al. 1993; Johnson and Brown 2005; Purdy and Abbas 2002; Stapells et al. 1990, 1995).

Tone-burst-elicited ABRs can be recorded to low-level stimuli in healthy, full-term babies. Several studies have shown minimum response levels in normally hearing neonates that are comparable with those seen in older subjects, with mean TB-ABR thresholds typically around 5–15 dBnHL for frequencies across the audiometric range (500 Hz–4000 Hz) (Klein 1984; Rance et al. 2006; Sininger et al. 1997). In contrast, there is currently a lack of research describing TB-ABR result patterns for premature babies tested before their due date.

Hearing assessment in the pre-term period (Spong 2013) is desirable for several reasons. Firstly, babies are typically most accessible (i.e. hospital in-patients) during this period and may be discharged before their due date if healthy and stable. Secondly, medical interventions - such as the provision of antiviral medicines for hearing-impaired infants testing positive for cytomegalovirus may be most effective in minimizing hearing loss within this window (Rawlinson et al. 2017, 2018). Thirdly, parental anxiety about the potential (hearing) consequences of prematurity may be alleviated and finally, early intervention may be initiated if a hearing deficit is identified.

While hearing screening (using automated ABR techniques) is feasible from 34 weeks gestation (Coenraad et al. 2011; Jacobson et al. 1990), diagnostic assessment in the pre-term period is hampered by the lack of studies supporting feasibility, reliability, and impact of maturation on the response. As a result, hearing evaluation capable of forming a basis for intervention, may be delayed until the infant reaches term or is outside the neonatal period (Coenraad et al. 2011; Eggermont and Salamy 1988; Lary et al. 1985; Picton et al. 1994; Salamy 1984).

Changes in ABR thresholds with maturation in premature infants are not well understood. The response can be recorded in babies as early as 25 weeks gestation (Starr et al. 1977) and there have been numerous studies describing the course of ABR latency and amplitude development through the pre-term period (Jiang 2015; Martins-Ribeiro and Carvallo 2008; Pasman et al. 1992; Seethapathy et al. 2018, 2019; Stipdonk et al. 2016), but there is relatively little data describing response threshold changes in premature babies tested before their due date. Starr et al. (1977) investigated click-ABR threshold levels for premature infants and found a decrease in minimum response level from around 45 dBnHL for babies tested at 27 weeks post-conceptual age to approximately 20 dBnHL for babies at term. Similarly, Lary et al. (1985) in their cross-sectional sample of premature infants tested in the first week of life, found a progressive decrease in click-ABR thresholds obtained from 27 weeks GA to term with mean thresholds for infants assessed at 28–34 weeks gestation approximately 20 dB higher than those of full-term babies. Likewise, Ping and Jiang (2013) found a linear decrease in click-ABR thresholds with increasing gestational age and a mean threshold approximately 10 dB higher for babies tested at 30 weeks compared to those assessed at term.

Findings for frequency-specific ABR thresholds in the pre-term period are particularly scarce. Martins-Ribeiro and Carvallo (2008), did report that ABRs could be recorded at low presentation levels (20 dBnHL) to a range of tone-burst stimuli in babies tested at 28–35 weeks gestation, but these were not threshold data and potential developmental changes through the pre-term period were not considered.

In this longitudinal study, we sought to track tone-burst ABR latency and threshold changes in individual premature babies through the pre-term and immediate post-term periods. Our study hypotheses were as follows:

1. TB-ABR latencies in premature babies would be longer and more variable in the pre-term period and would be equivalent to those of the full-term participants by the due date.

2. TB-ABR thresholds in premature babies would be elevated and more variable in the pre-term period and would be equivalent to those of the full-term participants by the due date.

Materials and methods

Ethics

This study carried approval from the Ethics Committee of the Royal Victorian Eye and Ear Hospital (07/757H). All testing conformed to the tenets of the Declaration of Helsinki (2013) and informed consent was obtained from all participant families after an explanation of the nature, purpose, and expected outcomes of the study.

Participants

A total of 24 babies participated including nine full-term and fifteen pre-term infants. Children were recruited from the special care and well-baby nurseries of the Epworth Freemasons Hospital and Box Hill Hospital in Melbourne, Australia.

Participants were divided into two groups based on gestational age at birth. The full-term infants (four males) were born between 39 and 41 weeks GA (mean: 40 weeks). None of these required special care or had any risk factors for hearing loss. The pre-term participants were born at gestational ages ranging from 28 to 34 weeks (mean: 32 weeks). Each individual in this group (nine males) spent time in the special care nursery, where they received care for relatively mild neonatal health issues. The mean length of stay was 8 days. None suffered a critical illness and, as such, did not require treatment through a neonatal intensive care unit. Children were excluded from the study if they had any risk factors for hearing loss, including hyperbilirubinemia requiring blood transfusion, birth asphyxia requiring resuscitation, congenital abnormalities of the head or neck, infections requiring aminoglycoside administration, any neurological condition, or a family history of childhood hearing loss.

Apparatus and procedures

Infants were assessed at two-week intervals until they were six weeks post-due date. In the full-term group, the first of these tests was carried out 3–6 days after birth with subsequent tests every two weeks (42-, 44-, and 46 weeks gestation). Pre-term infants were also assessed at two-week intervals but over a longer period with assessments at 36-, 38-, 40-, 42, 44- and 46 weeks.

Initial assessment was performed at the hospital either in the special care nursery or in the private hospital room. Subsequent testing was undertaken at the family home. Background noise level recordings (from within the child’s ear canal) were taken at each test occasion to monitor potential environmental effects in the different settings. Measurements were made using a probe tube placed in the ear canal and attached to an external sound level meter as per Rance et al. 2006. The mean noise level for the hospital recordings was 51.8 dBSPL and the mean levels for the home assessments ranged from 45.4 to 49.1 dBSPL. Analysis of variance showed no significant difference between test settings.

Otoscopy was completed on each test occasion and normal middle-ear function was confirmed by the presence of transient evoked otoacoustic emission (TEOAE) responses to clicks at 80 dB SPL at each of the data collection points (Owens et al. 2008). An emission was considered present when a response signal-to-noise ratio of >6 dB and a waveform reproducibility of 50% was obtained for at least 3 octave bands.

All of the participants showed evidence of normal hearing at the 4000 Hz test frequency presenting with a repeatable ABR to a 4000 Hz tone-burst presented at ≤20 dBnHL at some point in the recording period. (Tharpe and Seewald 2016).

Tone-burst auditory brainstem response

Evoked response measurement was conducted while the infant was in natural sleep, either in a crib or being held by a parent. One ear was randomly selected at the initial assessment and the same ear was subsequently assessed on each occasion. In the full-term group, five babies had their left ear tested and in the pre-term group, nine had their left ear evaluated.

Responses were recorded using the GSI AUDERA evoked response system to 500 Hz and 4000 Hz tone-burst stimuli. Positive polarity Blackman (2:1:2 cycle) tones were presented via ER-3A insert earphones at a rate of 39.09 Hz. Differential recordings were made between electrodes placed on the high forehead (Fz) or vertex (Cz) and the ipsilateral mastoid. A third electrode, placed on either the low forehead or contralateral mastoid, served as a ground. The electroencephalogram in a post-stimulus window of 0–20 ms was recorded and bandpass filtered from 30 Hz to 1.5 kHz. Artifact rejection (set at 25 µV) was employed to avoid response contamination from high EEG levels and myogenic activity.

At each test frequency, a 70 dBnHL stimulus was presented to obtain a clear reference waveform. Response latencies were based on this recording. The stimulus presentation level was then reduced in 20 dB steps until a response was no longer present (determined by the testing audiologist). The approximate threshold was then determined using step sizes of 5 dB around that provisional point. At each presentation level, two runs consisting of averaged responses from 2000 samples were obtained. Post-hoc analysis was carried out by two independent judges experienced in clinical infant diagnostic testing who determined the latency of ABR Wave V and the TB-ABR threshold level. The threshold was defined as the minimum stimulus presentation level required to elicit a repeatable waveform. If there was a discrepancy between the judge-determined thresholds, (in the present study this occurred in <3% of occasions), an average of the two estimates was used in the data analysis.

Statistical analyses

Linear mixed-model analyses of variance were used to analyze the relationship between ABR measures (response latency [Wave V at 70 dBnHL] and threshold) with a random effect for a participant and fixed effects for age at assessment (timepoint), subject group (pre-term versus full-term) and test stimulus frequency (500 Hz tone-burst versus 4000 Hz tone-burst). The model was also adjusted for gestational age at birth as a covariate.

ABR result variability was evaluated using the test for equal variances (Levene’s).

Statistical significance was defined as p < 0.05.

Results

Auditory brainstem response: Wave V latency

Repeatable ABRs were obtained from all participants to acoustic stimuli at 70 dBnHL at each of the data collection points. Averaged response waveforms for a typical participant from the premature cohort are shown in Figure 1.

Figure 1. Auditory brainstem response waveforms were obtained over the 10-week data collection period for a female participant born at 32 weeks gestation. There are two tracings for every timepoint, each reflecting the averaged waveform from 2000 post-stimulus EEG samples. Waveform morphology and latencies are representative of those seen for the “pre-term” participant group showing improved response definition and decreased response latency throughout the study. Responses were obtained to 500 Hz tone bursts (panel A) and 4000 Hz tone bursts (panel B) presented to the test ear at 70 dBnHL. The arrow markers represent the Wave V peak in the response complex.

Wave V latencies decreased throughout the study. A mixed-effects analysis of variance revealed a significant effect of timepoint with earlier recordings showing longer latencies (F_{(5, 185)} = 7.48, p < 0.001). Mean wave V latency values for each data collection point are shown in Figure 2 and listed in Table 1. There was also a stimulus frequency effect with ABRs elicited by 500 Hz tone bursts significantly longer than those elicited by 4000 Hz stimuli (F_{(1, 185)} = 1822.35, p < 0.001). There were no main effects for subject group (pre-term versus full-term) (F_{(1, 185)} = 0.5, p = 0.829) or gestational age at birth (F_{(1, 185)} = 0.60, p = 0.448) and no interaction effects between timepoint and test frequency (F_{(5, 185)} = 1.07, p = 0.380) or subject group and test frequency (F_{(1, 185)} = 0.38, p = 0.537).

Figure 2. Mean ABR wave V latency values for pre-term and full-term groups show latency decreases over the recording period. Panel A shows findings for a 500 Hz tone-burst stimulus presented to the test ear at 70 dBnHL and panel B shows results for a 4000 Hz stimulus presented to the test ear at 70 dBnHL. “Timepoint” is GA in weeks. Error bars represent the mean ± 1 SD range.

Table 1. Auditory brainstem response Wave V latency means, standard deviations, and standard deviation confidence intervals for each subject group at each data collection point.

Timepoint (weeks)	Group	Frequency (Hz)	Mean (ms)	SD (ms)	SD 95% CI
36	Pre-term	500	12.8	0.8	(0.5, 1.6)
		4000	7.7	0.3	(0.1, 5.5)
	Full-term	500	*	*	*
		4000	*	*	*
38	Pre-term	500	12.5	0.9	(0.5, 2.2)
		4000	7.6	0.3	(0.2, 0.7)
	Full-term	500	*	*	*
		4000	*	*	*
40	Pre-term	500	12.1	1.0	(0.5, 2.7)
		4000	7.4	0.3	(0.1, 1.4)
	Full-term	500	12.6	1.0	(0.1, 22.6)
		4000	7.2	0.2	(0.1, 3.6
42	Pre-term	500	12.2	0.6	(0.3, 1.9)
		4000	7.3	0.4	(0.2, 1.0)
	Full-term	500	11.6	0.9	(0.1, 56.0)
		4000	7.1	0.2	(0.0, 14.4)
44	Pre-term	500	11.8	1.0	(0.6, 2.4)
		4000	7.3	0.3	(0.2, 0.6)
	Full-term	500	11.3	0.9	(0.2, 8.1)
		4000	7.1	0.3	(0.0, 3.7)
46	Pre-term	500	11.8	0.9	(0.5, 2.0)
		4000	7.1	0.4	(0.2, 0.8)
	Full-term	500	11.3	0.7	(0.1, 16.8)
		4000	7.1	0.3	(0.0, 5.5)

Results obtained to tone-burst stimuli presented at 70 dBnHL.

The variability of Wave V latency was consistent throughout the study. Tests for equal variance showed equivalent values across study time points for both stimulus frequencies (500 Hz: p = 0.731; 4000 Hz: p = 0.916). Furthermore, there was no latency variance difference between subject groups across time points (p = 0.922) (Table 1). In contrast, there was a significant frequency difference with wave V latency estimates more variable for 500 Hz, than 4000 Hz stimuli (500 Hz SD: 0.95 ms; 4000 Hz SD: 0.37 ms, P < 0.001).

Auditory brainstem response threshold

Tone-burst ABR thresholds decreased throughout the study. Analysis of variance revealed a significant timepoint effect with earlier assessments showing higher threshold levels (F_{(5, 222)} = 4.51, p = 0.001). Mean response threshold levels for each test frequency are represented in Figure 3 and listed in Table 2. There was also a stimulus frequency effect with ABR thresholds elicited by 500 Hz tone bursts significantly higher than those elicited by 4000 Hz stimuli (F_{(1, 222)} = 637.92, p < 0.001). No main effects were observed for subject group (pre-term versus full-term) (F_{(1, 222)} = 0.17, p = 0.683) or gestational age at birth (F_{(1, 222)} = 0.01, p = 0.923) and no interaction effects between timepoint and test frequency (F_{(5, 222)} = 1.36, p = 0.242) or subject group and test frequency (F_{(1, 222)} = 1.51, p = 0.187).

Figure 3. Mean ABR threshold values for pre-term and full-term groups show minimum response level decreases across the recording period. Panel A shows findings for a 500 Hz tone-burst stimulus and panel B shows results for a 4000 Hz stimulus. “Timepoint” is GA in weeks. Error bars represent the mean ± 1 SD range.

Table 2. Auditory brainstem response threshold means, standard deviations and standard deviation confidence intervals for each subject group at each data collection point.

Timepoint (weeks)	Group	Frequency (Hz)	Mean (dBnHL)	SD (dB)	SD 95% CI
36	Pre-term	500	45.7	7.8	(4.3, 17.6)
		4000	21.0	6.3	(3.2, 15.7)
	Full-term	500	*	*	*
		4000	*	*	*
38	Pre-term	500	44.2	9.7	(4.5, 28.3)
		4000	17.1	4.5	(2.4, 11.3)
	Full-term	500	*	*	*
		4000	*	*	*
40	Pre-term	500	44.2	9.3	(5.6, 20.2)
		4000	18.1	6.9	(3.0, 20.6)
	Full-term	500	40.6	9.2	(3.9, 32.2)
		4000	21.7	6.6	(2.9, 22.8)
42	Pre-term	500	38.5	7.1	(3.1, 23.4)
		4000	15.0	4.1	(2.4, 9.9)
	Full-term	500	35.0	6.1	(3.0, 18.9)
		4000	20.6	8.1	(3.0, 33.3)
44	Pre-term	500	38.6	8.4	(4.4, 19.7)
		4000	16.5	8.0	(4.4, 18.8)
	Full-term	500	33.3	3.5	(1.5, 12.9)
		4000	14.4	5.3	(2.2, 18.6)
46	Pre-term	500	42.5	9.9	(5.5, 24.0)
		4000	18.2	8.5	(3.3, 27.5)
	Full-term	500	33.9	4.9	(2.0, 17.4)
		4000	13.3	6.1	(3.2, 17.4)

Tone-burst ABR threshold variability was consistent throughout the study. Tests for equal variance showed equivalent values across study time points for each stimulus frequency (500 Hz: p = 0.731; 4000 Hz: p = 0.916). Furthermore, there was no threshold variance difference between pre-term and full-term subject groups across timepoints (p = 0.219) (Table 2). In contrast, response thresholds were more variable for 500 Hz, than for 4000 Hz stimuli (500 Hz SD: 8.8 dB; 4000 Hz SD: 6.9 dB, p = 0.002).

Discussion

The longitudinal results presented in this study indicate developmental changes affecting both ABR latency and threshold in pre-term assessment. Response variability, was, however, equivalent across the recording period suggesting that clinical testing for healthy premature babies will be as reliable in the late pre-term period (36–40 weeks gestational age) as evaluation undertaken post-due date.

Response latency

Auditory brainstem response latency showed a consistent frequency effect with ABRs elicited by the 500 Hz tone-burst showing significantly later Wave V peaks than those evoked by the 4000 Hz stimulus. This finding, which has been reported previously (Martins-Ribeiro & Carvallo 2008; Ponton et al. 1992; Sininger et al. 1997), is likely due (at least in part) to cochlear mechanics. The tonotopic frequency arrangement of the basilar membrane means that low-frequency acoustic inputs are processed more apically than high-frequency and, as such, are relatively delayed due to increased traveling wave times (von Békésy 1960).

Wave V latency decreases were observed for both low- and high-frequency stimuli throughout the study. This was particularly evident for recordings obtained from pre-term babies assessed before their due date where mean values were approximately 0.5 ms longer than those obtained at term (Table 1). Similar differences have been reported by other authors (Jiang 2015; Martins-Ribeiro & Carvallo 2008; Pasman et al. 1992; Seethapathy et al. 2018, 2019) who have demonstrated (primarily in cross-sectional data) increases in both absolute Wave V latency and Wave I–V interpeak latency with earlier assessment. Importantly, there was no group difference in our results and no effect of gestational age at birth indicating that peripheral auditory function and/or the efficiency of neural conduction in the auditory nerve/brainstem had normalized by the due date (40 weeks). This result is similar to those reported by Ponton et al. (1992) who found no difference in neural conduction time (ABR I–V interpeak latency) for groups of premature and full-term babies tested at the same gestational age. Our findings are, however, inconsistent with those reported by Stipdonk et al. (2016) whose meta-analysis of published click-ABR latencies suggested neural conduction delays, particularly between peaks III and V, in premature babies tested at term. Methodological differences make direct comparison difficult, but the conflicting results may reflect differences in the study populations. Some of the studies described in the meta-analysis, for example, included children born at 24–28 weeks gestation, where each of the participants in the current study was born at 28+ weeks. Furthermore, the meta-analysis included children who were intensive care unit graduates and found that “neonatal complications” were a significant factor in the rate of auditory neural development. In contrast, none of the infants in our group required treatment in the intensive-care unit.

Auditory brainstem response latency estimates to 500 Hz tone bursts were more variable than those evoked by 4000 Hz tones. This is likely related to the fact that the peaks in the response waveform were less well defined to low-frequency stimuli—see Figure 1. This phenomenon, which suggests a lesser degree of neural synchrony in the auditory nerve/brainstem, may be stimulus-related. An acoustic click, which has a near-instantaneous onset, maximizes the consistency of neural activity across the stimulus series to produce an averaged ABR. Tonal stimuli, in contrast, require a rise period, which is typically related to the number of cycles in the signal. In this study, we used Blackman tone bursts with a two-cycle rise, so the onset time for the 500 Hz stimulus was 4 ms whereas for the 4000 Hz tone, it was only 0.5 ms. This onset difference results in relatively poorer temporal precision for the low-frequency burst as the point at which the neural activity is triggered is potentially more variable. Alternatively, the relative lack of synchrony to the 500 Hz burst could have been related to cochlear processing, as the traveling wave is slower in the apical turns than the basal which results in a greater temporal spread across the basilar membrane regions contributing to the low-frequency response (Johnston and Boyle 1967; von Békésy 1960).

In contrast to the frequency results, there were no timepoint, group, or age-at-birth effects on latency variability. This suggests that clinicians can be confident of where to look for a potential ABR waveform within the post-stimulus recording window, regardless of a baby’s degree of prematurity or gestational age—at least for testing carried out at 36 weeks GA or later.

Response threshold

Auditory brainstem response thresholds to 500 Hz tones were significantly higher than those elicited by the 4000 Hz stimuli across the recording period. This consistently reported phenomenon (Hawes and Greenberg, 1981; Rance et al. 2006; Stapells et al. 1995) is also likely a reflection of the relative lack of temporal precision in the neural response to low-frequency sounds, which makes the resulting ABR waveform less distinct and harder to identify within the background EEG at levels close to the psychophysical hearing threshold.

Response thresholds, for both test stimuli, decreased over the recording period with mean thresholds obtained at the 36-week data point approximately 5 dB higher than those observed from the same babies at or post-due date (Table 2). There were, however, no group or gestational age-at-birth effects indicating that ABR threshold levels, even in the most premature participants, had normalized by the due date.

There are several possible causes underlying ABR sensitivity changes in the pre-term period. Differences in ear canal size and resonance are one possibility, although had the ear canals been smaller in the pre-term period (which is likely), the equivalent dBSPL levels at the tympanic membrane would have been higher and the resultant ABR thresholds relatively lower than later recordings (Rance et al., 2006; Sininger et al., 1997).

Cochlear immaturity in the pre-term period is another conceivable explanation. The human cochlea is anatomically mature at birth (Bredberg 1968; Lavigne-Rebillard and Pujol 1987) and studies of otoacoustic emission tuning curves (Abdala et al. 1996) and basilar membrane traveling wave times (Eggermont et al. 1991) have indicated maturity of the cochlear mechanical mechanisms at term. Less is known about cochlear physiology in the pre-term period, but Gkoritsa et al. (2007) have shown normal transient OAE amplitudes in babies tested at 30–36 weeks post-conceptual age suggesting that a cochlear contribution to the elevated ABR thresholds in our pre-term data is improbable.

A more likely explanation for ABR threshold elevation in the pre-term period is the immaturity of the auditory neural system. Unlike the cochlea, the brainstem pathways are not mature at term. This can be most clearly seen in response latency. ABR wave I latencies are adult-like soon after birth in full-term infants (Eggermont and Salamy 1988) but absolute wave V and wave I–V interpeak latencies decrease over the first 2–3 years of life (Salamy 1984). Furthermore, as discussed previously, latency findings in the pre-term period suggest significant changes in the efficiency of neural conduction in the auditory nerve/brainstem.

Axonal and/or synaptic immaturity may explain both latency and threshold differences in the pre-term ABR findings. Myelination of auditory axons is present throughout the pathway by 29 weeks gestation, but the density of myelin sheath continues to increase throughout the 1st year of life (Moore et al., 1995). Myelin affects the speed of neural conduction and its developmental course is therefore likely to be a key factor in the maturation of ABR latencies. Similarly, post-natal synaptogenesis is thought to be a factor in the development of ABR peak latencies (Ponton et al. 1996). In addition to affecting the speed of neural conduction, both myelination and synaptic efficiency can impact the synchrony of neural firing in the auditory nerve/brainstem (for a review Rance and Starr 2015). In extreme (pathological) cases, this can mean that the ABR is unrecordable (i.e. auditory neuropathy presentation), but normal developmental changes could result in synchrony improvements over time which could, in turn, affect the precision of the ABR waveform and its detectability at low amplitudes when close to the EEG noise floor. Development of the accuracy of neural firing patterns in the pre-term ABR could therefore result in elevated threshold levels in pre-term testing.

The variability of ABR thresholds was consistent throughout the study. This finding is important for clinical assessment as the reliability of hearing level prediction from the ABR is determined by the consistency of the evoked response/hearing threshold relationship (Picton et al. 2005; Rance et al. 1995). That is, in clinical testing, the average ABR sensation level can be subtracted to determine the most likely hearing threshold, but how confident the clinician can be with that prediction is determined by the range of behavioral hearing levels that can correspond to a particular ABR threshold. In this study, standard deviations for our group of (presumed) normally hearing participants ranged from 5–10 dB (500 Hz stimulus) to 4–8 dB (4000 Hz) stimulus which is similar to those reported in previous studies involving newborns tested post-term (Rance et al. 2006).

Study limitations

A particular limitation of this study is that the participants were not followed up with a behavioral hearing assessment. All had normal TEOAE responses on each test occasion, so there was no evidence of middle ear abnormality, and normal 4000 Hz TB-ABR thresholds (≤20 dBnHL) on at least one test occasion, so it was presumed that they all had hearing within the normal range—but exact hearing levels were not established. As such, we do not know the precise ABR sensation levels in each case. Furthermore, while there was no evidence to suggest hearing level fluctuation, it was assumed that each child’s hearing was stable throughout the study and that any changes to ABR latency and threshold were the result of developmental processes.

Another limitation involves the study assessment range (36–46 weeks GA). Reliable ABR testing may be possible in premature babies before 36 weeks GA, but we were unable to access study participants before this date due to Special Care Unit policies at the participating hospitals.

A final possible limitation is the sample size. Our longitudinal study design limited the number of babies who could participate. Although each participant was evaluated on several occasions and the project included data from 126 separate test sessions, the fact that we had only 24 subjects may limit the generalization of the findings.

Conclusions

The results of this study suggest that diagnostic hearing assessment based on tone-burst ABRs is feasible in healthy, premature babies from at least 36 weeks gestation. While response threshold levels were marginally higher in pre-term assessment, variability across participants was equivalent at each of the pre- and post-term data collection points meaning that hearing level estimation would be equally reliable at any point from 36- to 46 weeks gestational age. The range of observed ABR thresholds for the 4000 Hz stimulus was only 10–25 dB for each data point suggesting that normal hearing and mild hearing loss could be differentiated at this frequency. Results at 500 Hz were more variable, with threshold ranges of 15–35 dB, suggesting that ABR testing at this frequency would struggle to differentiate between normal hearing and mild degree loss, but this low-frequency shortcoming, which has been reported previously for full-term neonates (Rance et al. 2006; Sininger et al. 1997), was a consequence of test technique limitations rather than a reflection of participant immaturity. As such, the findings of this study suggest that frequency-specific quantification of hearing is possible in the pre-term period and that intervention options (amplification, etc.) may be considered with the same degree of confidence as for children tested beyond their due date.

Author contributions

GR: conception and design of the study; acquisition and analysis of data; drafting a significant portion of the manuscript/figures. DT: acquisition and analysis of data; drafting a significant portion of the manuscript/figures. KG: acquisition and analysis of data; drafting a significant portion of the manuscript/figures.

Acknowledgments

The authors would like to thank the families who participated and the staff from the maternity units at the Freemasons and Box Hill Hospitals, Melbourne, Australia.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This project received funding support from the William Buckland Foundation and the Deafness Foundation. GR was supported by the Graeme Clark Chair in Audiology and Speech Sciences.