Cochlear Implants for Deaf Children With Early Developmental Impairment

A prospective longitudinal cohort study was performed that included all children identified with severe to profound sensorineural hearing loss managed by 2 large pediatric cochlear implant centers: Texas Children’s Hospital (Baylor College of Medicine, Houston, TX) and Lucile Packard Children’s Hospital (Stanford, CA). Institutional review boards at both institutions approved the protocol, and the study was registered with ClinicalTrials.gov (identifier NCT01256229; start date September 2009; completion date September 2017).

A baseline assessment was completed by clinicians in audiology, speech-language pathology, and neuropsychology. We used validated tests to assess cognition (Mullen Scales of Early Learning [MSEL]), adaptive behavior (Vineland Adaptive Behavior Scales [VABS]), language skills (Preschool Language Scale [PLS]), auditory skills (LittlEARS Auditory Questionnaire [LEAQ]), and stress within the parent-child system (Parenting Stress Index [PSI]). Subject recruitment procedures and assessment measures are described in the Supplemental Information.

All children started with hearing aids. When performed, the cochlear implantation surgery was ∼1 month after the baseline assessment. The only exceptions were children younger than the minimum age approved by the US Food and Drug Administration for cochlear implantation (12 months). For these patients, implantation was performed within the month after the children’s first birthday. Follow-up evaluations were scheduled annually using the same assessment protocol.

Developmental impairment was defined on the basis of scores on nonverbal cognitive skills and adaptive functioning. To estimate cognition, we used the visual reception domain from the MSEL. To assess adaptive behavior, we used the daily living skills domain from the VABS. The other domains were excluded because of the possibility that results may not reflect actual capability due to either reduced receptive and expressive language or motor delays. Previously, we validated these 2 measures in a patient population with hearing loss and EDI, and we defined meaningful thresholds for EDI guided by standard definitions of intellectual disability.^21,23,24  Children who scored 2 SDs below the mean for their age in the visual reception domain of the MSEL met the criterion for low nonverbal cognitive skills. Children who scored 1 SD below their age in the daily living skills domain of the VABS met the criterion for low adaptive behavior. Normal nonverbal cognitive skills and normal adaptive behavior were defined as being above these thresholds. Thus, children who met our definition of EDI had severe developmental delays.

For comparing the patient characteristics, which were categorical variables (Table 1), we used the Fisher’s exact test for proportions. For comparing the developmental outcomes, which were continuous variables (Table 2), we used analysis of variance followed by Tukey’s post hoc subgroup comparisons. In both cases, adjusted P values were used to account for multiple comparisons. To perform linear fits for each developmental outcome for each cohort (Fig 2), we used least squares regression. To compare developmental trajectories among the 3 cohorts (Table 3), we used hierarchical linear modeling. Testing age and cohort number were used as fixed-effects predictors, and the subject identifier code was used as a random effect predictor. This statistical approach was chosen for these longitudinal data because the resulting models provide estimates for growth rate, which has less error than each individual data point for each child. The method is also advantageous because it is permissive regarding missing data and can account for random variations in follow-up testing age between subjects. For all analyses, we adjusted for simultaneous inference by term, and the threshold for statistical significance was set at P < .05. Confidence intervals (CIs) were calculated at 95%. Analyses were conducted in R, and figures were produced using the package ggplot2.^25,26 

From the 297 children enrolled in this longitudinal study, 3 subgroups were formed (Fig 1). Cohort 1 included children who received a cochlear implant and had normal cognitive skills and normal adaptive behavior (n = 138). Cohort 2 included children who received a cochlear implant and who met criteria for EDI based on both low cognitive skills and low adaptive behavior scores (n = 37). Cohort 3 included children who continued to be treated with hearing aids and who also met criteria for EDI based on both low cognitive skills and low adaptive behavior scores (n = 29). The remaining 93 children who had low cognitive skills or low adaptive behavior, but not both, were excluded from further analysis. Although these 93 children are worthy of study, the did not meet the criteria for testing the hypothesis proposed in this manuscript.

We found no differences among the cohorts on the basis of study site, sex, household income, race, or mother’s education level (Table 1). As expected, there was a higher rate of preterm births in both cohort 2 and cohort 3 compared with cohort 1. There were also differences among the cohorts in the rates of living in an urban or rural environment, ethnicity, and preferred language, with higher rates of urban Hispanic, Spanish-speaking families in cohort 1. Because we had no hypotheses about the interaction of these demographics with our outcome data, we assumed that the significant P values represented random statistical variability and did not include any demographic data in further analyses.

Baseline data were collected at the time of enrollment, with repeated evaluations performed approximately annually over the next 1 to 5 years. The final evaluation was defined as the last data point we collected before the grant ended. During the study, 39 of the children in cohort 1 and 5 of the children in cohort 2 were lost to follow-up, typically because their family moved out of town. Interestingly, none of the children in cohort 3 were lost to follow-up. The rates of attrition were acceptable for a study of this duration. Children in cohorts 1 and 2 had more follow-up evaluations than children in cohort 3 (Table 2). This difference is not surprising because insurers who were willing to cover cochlear implantation were also willing to cover repeated neurocognitive examinations.

At baseline, cohort 1 was the youngest and cohort 3 was the oldest. There were no differences in cognition, adaptive behavior, language, auditory skills, or stress within the parent-child system between cohorts 2 and 3. Because the children were all deaf, cohort 1 had language and auditory skills like cohorts 2 and 3. However, cohort 1 had less stress within the parent-child system relative to the 2 other cohorts. By the time of final testing, average developmental scores were highest for cohort 1 and lowest for cohort 3.

Next, we compared the developmental trajectory for each cohort by calculating a linear fit for all evaluations from birth, where all developmental ages and raw scores would be 0, with the final evaluation (Fig 2). However, to compare the growth trajectories, we did not simply compare these linear fits. Instead, we performed hierarchical linear modeling (Table 3). This analysis permitted us to control for age of testing, differences in the number of follow-up evaluations, and random differences between individuals. The estimated coefficient values reflect the differences in the slope of either cohort 1 or 3 compared with cohort 2. Positive values mean that the slope is steeper than for cohort 2; negative values mean that the slope is shallower.

To analyze growth in cognitive skills and adaptive behavior, we plotted the developmental ages for the MSEL and VABS versus chronological age. A perfectly normal developmental trajectory would have a slope of 1. Cohort 1 performed best, with near-normal trajectories in intelligence and adaptive behavior. In addition, the R² values were high (>0.8), demonstrating that most children in this cohort fit our regression model. In contrast, the developmental trajectories of cohort 2 were shallower and more variable, and cohort 3 underperformed cohort 2.

To analyze language and auditory skills outcomes, we plotted raw PLS and LEAQ scores versus chronological age. Cohort 1 had the best outcomes and high R² values, demonstrating that raw scores and chronological age were strongly correlated. In contrast, the developmental trajectories for cohort 2 were not as steep as those in cohort 1, and cohort 3 underperformed cohort 2.

To assess for changes in stress within the parent-child system, we plotted the PSI score versus chronological age. There were no differences between cohorts 1 and 2, but cohort 3 demonstrated more stress in the parent-child system over time.

To test our hypothesis that cochlear implantation was better than hearing aids for children with EDI, we compared developmental trajectories between cohorts 2 and 3. We found that cohort 3 did worse than cohort 2 in all assessments, including cognition (95% CI, −0.213 to −0.026; P = .01), adaptive behavior (95% CI, −0.365 to −0.073; P = .003), language (95% CI, −0.460 to −0.008; P = .04), and auditory skills (95% CI, −0.363 to −0.124; P < .001). In addition, parental stress was higher in cohort 3 (95% CI, 0.208 to 2.448; P = .02), indicating larger increases in stress within the parent-child system.

To verify that site did not affect these findings, we repeated the hierarchical linear modeling analysis, including site as a random effect. This produced only slight changes in the values given in Table 3, and the statistical significances for every finding did not change.

In this study, we provide the most compelling data available that cochlear implantation in children who are born deaf and who have EDI is associated with better developmental outcomes and faster rates of development in multiple domains than continued use of hearing aids. Our experience in assessing children with EDI at 2 large pediatric hospitals reveals that cochlear implantation is associated with improved scores on measures of nonverbal cognitive skills, adaptive functioning, language, auditory skills, and reduced stress in the parent-child system compared with treatment with hearing aids alone. On the basis of our data, we argue that cochlear implantation should be considered for all children with severe to profound hearing loss, regardless of their cognitive skills and adaptive behavior. Because many studies in children who are deaf with normal development have revealed that cochlear implantation at a younger age is associated with better developmental outcomes^4,27–30  than delaying implantation, we argue that this decision should be made as early as possible before a critical window for rapid development has passed.

It is unfortunate and inequitable that, in our society, children with EDI may be excluded from receiving potentially beneficial interventions because their developmental outcomes are below that of children who function in the normal range in terms of cognition and adaptive behaviors. Our study reveals that this problem is seen for children who need cochlear implantation. Besides the obvious problem that EDI has been used as a reason to deny insurance coverage, a more complex hurdle to overcome is that providers may also fail to refer these patients for cochlear implant evaluation on the basis of the notion that their cognitive delays render them ineligible for insurance coverage for the procedure. For example, we found that children with EDI who continued with hearing aid use were older and had lower auditory skills at initial evaluation than children who went on to receive a cochlear implant. This finding might be explained by differences in pediatricians’ approaches to referring these children. Although the types of developmental testing paradigms in use today are wide ranging and highly valid, it remains difficult to accurately predict the ultimate cognitive outcome of any individual child from tests performed when they are infants and toddlers, like those included in this study. To give all children the best opportunity to fully develop their maximum potential, cochlear implantation should be carefully considered regardless of the presence of severe developmental delays.

We recognize the several limitations of this study. Most importantly, a randomized controlled trial was not feasible given the current state of clinical practice. When we initially proposed this study, we planned a prospective, randomized controlled trial of children with EDI in which some children would receive a cochlear implant and some would continue with hearing aid use. In the early 2000s, pediatric cochlear implantation in children with normal cognition was still considered a novel procedure, and it was rarely performed in any child with EDI. However, by the time we were awarded National Institutes of Health funding for this trial, nearly all children with normal cognition and most children with EDI were undergoing cochlear implantation. This change occurred because the evidence for the benefit of cochlear implantation in children with normal cognition had become clear, and parents and medical professionals considered it clinically unwise and unethical to withhold the “best” treatment of the child with EDI. However, equipoise had not been lost, as the question of whether cochlear implantation is beneficial to children who may not have the cognitive ability to use the auditory information provided by the cochlear implant remained unclear.

We attempted to overcome our inability to perform a controlled clinical trial by capitalizing on differences in insurance coverage to create randomization. However, we could not determine with certainty the underlying reasons for the decisions made by families of children with deafness and EDI in opting for cochlear implants or hearing aids. For most of the families in cohort 3, the decision regarding cochlear implantation was taken out of their hands by the reimbursement agency. Nevertheless, it was impossible to quantify how differing eligibility policies affected the implantation rate because neither pediatric cochlear implant team determined the care plan for any patient solely on the basis of insurance coverage. Instead, according to best practice, the teams described the risks, benefits, and typical outcomes for hearing aids and for cochlear implantation. The parents then decided how they wished to proceed. Interestingly, none of the families whose child was denied by insurance was adamant about receiving a cochlear implant. Over time and repeated visits, most parents ultimately determined for themselves that cochlear implantation would not be best for their child. In contrast, parents who did not have this eligibility issue tended to make the decision to proceed with implantation.

Another caveat is that these children were only followed for an average of 2 years after their initial evaluation. We know that in children with normal intelligence, cochlear implantation produces better long-term outcomes than hearing aids.³¹  Although we found that children with EDI treated with hearing aids have a slower developmental trajectory than those treated with cochlear implantation, it is possible that both cohorts will ultimately end up with similar developmental outcomes.

To conclude, our data are supportive of the value of cochlear implantation in children with multiple disabilities/developmental delays and provide an argument against insurance plans requiring that certain developmental milestones be met prior to authorization. Developmental impairments will adversely affect a child’s outcome after cochlear implantation because the combination of hearing loss and EDI synergistically affect other developmental outcomes, with the loss of auditory input creating a cascade of higher-order neurocognitive deficits.³²  This outcome should be discussed carefully with the parents to ensure that their expectations are appropriate. However, EDI should not be the deciding factor alone. Every child is an individual and has unique health and developmental limitations, together with unique family, home, and school environments. All these factors should be considered during the process of cochlear implantation candidacy.

We thank E. O’Brian Smith, Lane Strathearn, Ann Stark, Jody Haymond, Sue Caudle, Dian Baker, Ross Tonini, Homer Abaya, Jody Winzelberg, Barbara Bentley, Ryan Lucash, Laurie Eisenberg, and Karen Johnson for their contributions. Deidentified individual subject data (including data dictionaries) will be made available, in addition to study protocols, the statistical analysis plan, and the informed consent form. The data will be made available upon publication to researchers who provide a methodologically sound proposal for use in achieving the goals of the approved proposal. Proposals should be submitted to [email protected].

CI

confidence interval

EDI

early developmental impairment

LEAQ

LittlEARS Auditory Questionnaire

MSEL

Mullen Scales of Early Learning

PLS

Preschool Language Scale

PSI

Parenting Stress Index

VABS

Vineland Adaptive Behavior Scales