Hearing in Schoolchildren After Neonatal Exposure to a High-Dose Gentamicin Regimen

Children included in this study had been admitted to the NICU at the University Hospital of North Norway and received gentamicin therapy between 2004 and 2012. This NICU is the only unit offering care for infants born before 32 weeks’ gestation and all other newborn infants (≥32 weeks’ gestation) in need of mechanical ventilation or intensive care in the 2 northern-most counties in Norway. We previously validated our extended-interval, high-dose (6 mg/kg) gentamicin dosing regimen in 440 neonates who were exposed to at least 3 doses of gentamicin between 2004 and 2012.²⁰  The vast majority of TPCs (94%) were within the normal range, there was a low rate of prescription error, and we found no evidence of early-onset ototoxicity using a transient evoked otoacoustic emissions screening test before hospital discharge.²⁰ 

For the current study (Fig 1), 357 children from the original cohort were invited for a detailed hearing assessment at age 6 to 14 years. We also, from public primary schools, recruited a control group of 33 healthy children with no history of previous use of aminoglycosides and no previous hearing problems or tympanostomy tubes. Parents of all children filled out a questionnaire including any history of middle-ear infections, treatment with tympanostomy tubes, or use of intravenous antibiotics after the neonatal period.

For the gentamicin-exposed cohort, we collected data on birth weight, gestational age (GA), Apgar scores, neurologic abnormalities, mechanical ventilation, and any phototherapy for jaundice. Preterm neonates are more susceptible to bilirubin-induced neurologic damage, suffer adverse effects at lower total serum bilirubin (TSB) levels, and receive more phototherapy than term infants.^21,22  We recorded the peak TSB level within the first 2 weeks of life and divided this value by GA in weeks, creating an age-adjusted variable of possible bilirubin toxicity instead of using crude peak TSB levels. To assess level of gentamicin exposure during hospitalization, we recorded 2 variables: the highest measured gentamicin TPC (mg/L) and the cumulative gentamicin dose (mg/kg). For the healthy-control group, we collected data on birth weight, admission to a NICU for reasons other than infection, and any phototherapy for jaundice.

Participants attended 1 study visit between September 2017 and September 2018. We did otoscopy and tympanometry at 226 Hz (Zodiac; Otometrics, Taastrup, Denmark) before pure tone audiometry. Tympanogram results were classified as type A (normal), B (flat), and C (negative pressure). We collected a urine sample for analysis of the mitochondrial 1555A>G gene mutation in all gentamicin-exposed children. DNA was extracted by using the Quick-DNA Urine Kit (Zymo Research, Irvine, CA). The m.1555A>G gene mutation was analyzed by using polymerase chain reaction amplification and melting curve analysis (LightCycler 480; Roche, Basel, Switzerland).

Pure tone audiometry thresholds were measured with the Equinox 2.0 clinical audiometer by using Equinox Suite 2.9.0 software (Interacoustics A/S, Middelfart, Denmark). The audiometer was calibrated according to the manufacturer’s specifications and in accordance with International Organization for Standardization references.^23,24  We used the DD45 supra-aural earphones (RadioEar, Middelfart, Denmark) for the conventional frequencies (0.125–8 kHz) and Sennheiser HDA200 closed circumaural earphones (Sennheiser, Wedemark, Germany) for the EHFs (9–16 kHz). Testing was done first in the conventional frequency range before the EHF range. We used the ascending method to acquire thresholds.²⁵  Special care was taken for each child to avoid fatigue and loss of concentration. The first ear tested (left or right) was randomly assigned by the survey management software (Research Electronic Data Capture). Audiometry testing was done by a trained audiologist or an audiology-trained ear-nose-throat physician. The hearing thresholds are expressed as dB hearing levels (HLs).

The main audiological outcomes were average hearing thresholds in the conventional frequencies and the EHF range. We calculated the established pure tone average (PTA), representing the mean of the conventional midfrequencies (0.5, 1, 2, and 4 kHz), according to an established reference method.²⁶  There is no established equivalent to PTA in the EHF range. We chose to use the average of all 6 EHFs (9, 10, 11.2, 12.5, 14, and 16 kHz), hereafter termed the extended high-frequency average (EHFA). Middle-ear problems can be unilateral, but ototoxic hearing losses are most often bilateral. Thus, we used the PTA and EHFA for the best ear in the final analysis. A relevant clinical hearing loss was defined as PTA and/or EHFA threshold >20 dB in the best ear. We report tympanogram results corresponding to the best-ear result.

The study was approved by the Committee for Human Medical Research Ethics for Northern Norway. All parents signed a written informed consent form, and all participating children received age-appropriate written information about the study. The study was registered with www.clinicaltrials.gov (number NCT03253614) in August 2017.

On the basis of previous studies,^27,28  we estimated that the mean EHFA threshold would be ∼5 to 10 dB in the healthy-control group. We realistically hoped to include 60% to 70% of the 357 invited gentamicin-exposed children. We considered that a 10 dB difference in the EHFA hearing threshold would represent a clinically relevant difference between healthy controls and the gentamicin-exposed group. By including ∼30 healthy controls and ∼250 gentamicin-exposed children, we would have 80% power with a 2-sided 5% level of significance to detect a difference of 4 to 5 dB between the groups. Moreover, within the group of gentamicin-exposed children, we knew that approximately half of them had a gentamicin TPC ≥1.0 mg/L, and the rest had a TPC <1.0 mg/L. With 125 children in each group, we would have 80% power with a 2-sided 5% level of significance to detect a difference of 3 to 4 dB between the groups.

All clinical data were first entered into Research Electronic Data Capture, a secure, Web-based software platform designed to support data capture for research studies (Vanderbilt University, Nashville, TN). Clinical and audiometry data were analyzed by using IBM SPSS Statistics version 23 (IBM SPSS Statistics, IBM Corporation). Descriptive results are expressed as medians and interquartile ranges (IQRs). We used a univariable linear regression model to analyze level of gentamicin exposure and other predictors that may affect hearing thresholds.²⁹  We then plotted all predictors in a directed acyclic graph, and on the basis of clinical and biological knowledge, we identified birth weight as the central confounder of both the outcome and other predictor variables. Finally, we adjusted each predictor separately for birth weight. Results from univariable and adjusted analyses are presented as regression coefficients with 95% confidence intervals (CIs). We defined P <.05 as significant.

After parental consent, 226 of 357 (63%) gentamicin-exposed children were included. Eight children had relevant hearing loss (Table 1). Five of these had known etiology (3 with ongoing middle-ear disease and 2 with developmental delay and genetic hearing loss) and were therefore not included in the main audiological analysis. The 3 remaining children had hearing loss of uncertain etiology and were included in the main audiological analysis. Two more children were excluded from the main audiological analysis because of obvious lack of concentration during testing with uncertain validity of audiometry results.

High-quality audiometry results were obtained for 219 children exposed to gentamicin in the neonatal period and for 33 healthy controls (Table 2). In the gentamicin cohort, 39 (17%) had very low birth weight (VLBW) (<1500 g birth weight), and 46 (20%) had been treated with mechanical ventilation. One child was diagnosed with a m.1555A>G gene mutation. This child had culture-confirmed group B streptococcal early-onset sepsis and received gentamicin for 12 days but had normal audiometry results (best-ear thresholds: PTA 6 dB and EHFA 8 dB). Three term children who underwent therapeutic hypothermia because of severe perinatal asphyxia also received gentamicin; all 3 later had normal psychomotor development and no hearing loss.

Overall, the gentamicin-exposed cohort and the control group had normal hearing thresholds for the whole frequency range (Table 2, Fig 2). Unadjusted statistical analysis showed a 2.5 dB absolute difference in median EHF hearing thresholds between the gentamicin-exposed children and healthy controls, which is not of clinical significance. After adjusting for birth weight, the statistical difference was lost (Table 2). No International Organization for Standardization references exist for the EHF range in children. We compared our results with data from the hitherto largest published reference study, which included 90 healthy children and adolescents aged 5 to 19 years.³⁰  EHF hearing thresholds between groups from the current study and the reference study were comparable (Fig 2).

Tables 3 and 4 display the linear regression analysis of predictors for hearing thresholds in the conventional midfrequencies and EHFs. In the conventional midfrequencies, we found that birth weight, mechanical ventilation, and tympanometry results were all significant predictors in the unadjusted analysis. After adjusting each predictor for birth weight, only birth weight and tympanometry result remained significant predictors. In the EHFs, we found that cumulative gentamicin dose, birth weight, phototherapy, being small for GA, mechanical ventilation, and tympanostomy tubes were significant predictors in the unadjusted analysis. After adjusting each predictor for birth weight, only birth weight and tympanostomy tubes remained significant predictors.

We compared data from the population-based original study cohort, including all gentamicin-exposed neonates during the 8-year study period (n = 440), with data from the follow-up cohort (n = 226) to assess the representativeness of the follow-up cohort. There were no differences in birth weight, the proportion of VLBW infants, cumulative gentamicin doses, the highest median gentamicin TPCs, and the proportion of children with gentamicin TPC >2.0 mg/L between the 2 cohorts (Supplemental Table 5).

Our main objective in this study was to perform a detailed hearing assessment of schoolchildren exposed to a high-dose gentamicin regimen in the neonatal period to assess potential clinical or subclinical signs of ototoxic hearing loss as markers of long-term harm or safety. We tested hearing in both the conventional frequencies and the EHFs, adjusted findings for other potential peri- and postnatal risk factors for hearing loss, and compared audiological data with those of a healthy-control cohort. We found no association between level of gentamicin exposure in the neonatal period and hearing thresholds after 9 years median follow-up time.

Previous studies and reviews indicate a low risk of gentamicin-induced ototoxicity in the newborn period regardless of dosing regimen. However, there is a paucity of long-term, detailed follow-up studies. One recent case-control study compared level of gentamicin exposure in 25 VLBW infants who presented with hearing loss during the first 5 years of life and a matched control group without hearing loss and found no differences in gentamicin exposure between groups.³¹  One study from the 1970s reported 4-year follow-up hearing results after newborn aminoglycoside therapy using play audiometry (0.5–4 kHz). Only 25% of their original cohort were assessed at 4 years, but the authors did not identify any substantial aminoglycoside-attributable hearing loss.³²  Our study is the first long-term follow-up study performing high-quality pure tone audiometry, including the EHFs, of children exposed to gentamicin in the newborn period. A delay between exposure and hearing loss is well known from platinum-induced hearing loss in children.^33,34  This has also been suggested in sporadic cases after neonatal treatment with gentamicin.^35,36  We found no indication of late-onset gentamicin-induced ototoxicity in our study.

The mechanisms behind gentamicin-induced ototoxicity are not fully understood.³⁷  Currently, there is stronger evidence for aminoglycoside ototoxicity in older children than in neonates.^6,18,19  A possible explanation is that older children (eg, with cystic fibrosis or cancer) receive larger cumulative doses than those commonly administered in neonates.^6,18,19  Alternatively, the newborn inner ear is less vulnerable to ototoxicity or gentamicin-induced ototoxicity, so these effects may be partly reversible. Indeed, reversible ototoxic effects from aminoglycosides have been demonstrated in animal models.³⁸  Moreover, transient hearing loss in neonates is reported and could be explained by a transient cochlear dysfunction due to inflammation³⁹  or a delayed maturation of the auditory system.⁴⁰  However, in our study cohort, there were no signs of ototoxicity either at NICU discharge or at follow-up in children exposed to gentamicin.

Hearing loss in infants admitted to NICUs has a prevalence of ∼2% to 4% compared with 0.1% to 0.3% in the general newborn population.^7,29,41,42  Low GA, VLBW, mechanical ventilation, perinatal infections, hyperbilirubinemia, and severe asphyxia are all identified as risk factors for hearing loss.^8,11,13  In line with other studies, we found a strong association between decreasing birth weight and increasing hearing thresholds.¹⁴  Some authors argue that low birth weight itself does not cause hearing loss⁴³  but is rather associated with other perinatal factors that more directly affect hearing. We evaluated other possible predictors for hearing, such as Apgar scores, hyperbilirubinemia and/or phototherapy, and mechanical ventilation, but none of these were associated with increasing hearing thresholds after adjusting for birth weight.

The m.1555A>G mutation is associated with hearing loss, in particular after exposure to aminoglycoside antibiotics.⁴⁴  In our cohort, only 1 patient (0.44%) had this mitochondrial mutation, and this patient had normal hearing despite a cumulative gentamicin dose of 72 mg/kg. In another cohort of infants treated with gentamicin, 4 of 436 (0.9%) had a mitochondrial 12sRNA mutation, but only 1 showed evidence of possible hearing loss.⁴⁵  Some authors suggest testing for mitochondrial mutations before neonatal aminoglycoside treatment.⁴⁶  A clinical study is planning to assess rapid pharmacogenetic testing of the m.1555A>G mutation to avoid aminoglycoside therapy in at-risk neonates.⁴⁷  However, given the low and variable prevalence of this mutation in different ethnic populations combined with a variable penetrance, this approach may not be justified or cost-effective in all settings.^44,48,49 

Middle-ear disease in childhood may cause mechanical hearing loss because of permanent inflammatory damage and/or sensorineural hearing loss secondary to toxic effects on the inner ear.^50,51  Isolated sensorineural hearing loss in the EHFs after otitis media is also reported in children.⁵²  We found a significant association between previous tympanostomy tubes, which is a marker for more severe middle-ear disease, and EHF hearing thresholds. We also found increased hearing thresholds in the conventional midfrequencies in children with negative middle-ear pressure. The latter may reflect a subtle mechanical hearing loss caused by ongoing middle-ear pathology.

The strength of our study is the unique long-term audiological data that are sensitive enough to detect subtle and subclinical hearing loss. We also present data on different levels of gentamicin exposure, with cumulative dose being the most important proxy for exposure, but found only a weak correlation between cumulative dose and EHF-thresholds, which was not significant after adjusting for birth weight (Table 3, Supplemental Fig 3). It is a paradox that most neonatal gentamicin dosing regimens recommend lower gentamicin doses (4–5 mg/kg) than those of older children (7 mg/kg) despite a proportionally higher distribution volume in neonates.⁴  Since 2004, we have used a dosing regimen with a fixed gentamicin dose (6 mg/kg) for all neonates and a variable dosing interval (24–48 h) depending on GA and postnatal age.²⁰  This dosing regimen has a low risk of prescription errors.²⁰  Our study also has limitations. Children from the original cohort with the most severe comorbidities were not included in our follow-up because of clinical conditions that made them unable to complete audiometric testing. Some of these children may have hearing problems in addition to other disabilities. However, we are only aware of 1 child from the original cohort, who was diagnosed with a congenital cytomegalovirus infection, who has a cochlea implant. Since 2009, our unit has avoided routine use of gentamicin in children with severe asphyxia who undergo therapeutic hypothermia. Therefore, only 3 children with this condition were included in the follow-up cohort, all 3 of whom had normal hearing. There are conflicting results on a possible association between gentamicin exposure and hearing loss in children with severe perinatal asphyxia who have undergone therapeutic hypothermia.^53,54  Only 10% of the children in our study received >10 doses (>60 mg/kg) of gentamicin, and we cannot exclude that long courses of gentamicin have a greater ototoxic potential, also in the neonatal period. Finally, a response rate of 63% adds potential selection bias. Still, the gentamicin exposure data and the proportion of VLBW infants were similar in the original and follow-up cohorts.

In schoolchildren with no severe disabilities and who were therefore able to complete a detailed hearing assessment, we found no association between neonatal exposure to a high-dose, extended-interval gentamicin regimen and increased risk of hearing loss in the conventional midfrequencies and EHFs. Increasing hearing thresholds were associated with lower birth weight and middle-ear disease in childhood, but the vast majority of children had normal hearing. Potential damage to hearing early in life is of great concern because childhood hearing loss, and prelingual hearing loss in particular, may affect both language and general development.⁵⁵  It is therefore important to provide high-quality, long-term follow-up data on hearing after gentamicin exposure in neonates because this drug is widely used in neonates, and safety is therefore paramount.

We greatly appreciate the professional work of the staff at the Clinical Research Department at the University Hospital of North Norway in Tromsø. We are also grateful to Marthe Larsen (Clinical Research Department, University Hospital of North Norway, Tromsø, Norway) for statistical advice and Bo Engdahl (Norwegian Institute of Public Health, Oslo, Norway) for advice on audiological methods and analyses. Finally, we thank all the children and parents for participating in this study; without their voluntarily contribution, this study would not have been possible.

CI

confidence interval

EHF

extended high frequency

EHFA

extended high-frequency average

GA

gestational age

HL

hearing level

IQR

interquartile range

PTA

pure tone average

TPC

trough plasma concentration

TSB

total serum bilirubin

VLBW

very low birth weight