A Systematic Review to Define the Speech and Language Benefit of Early (<12 Months) Pediatric Cochlear Implantation

Through the introduction of universal newborn hearing screening, infants with profound sensorineural hearing loss are recognized earlier after birth. This has led to earlier cochlear implantation (CI) in congenitally deaf children [Lammers et al., 2015]. In 1990, the Food and Drug Administration (FDA) approved pediatric CI (PCI) from the age of 2 years [Clark, 2000], whereas currently the FDA has approved PCI in children from 12 months onwards [Bronstein, 2012].

Neuroplastic and neurolinguistic dynamics are the main reasons to opt for early CI, mainly to gain optimal benefit from implantation during the critical period of cortex neuroplasticity, a sensitive period in which speech and language are developed. In this critical period, auditory experience must occur to organize the neural connections in the brain. Human central auditory pathways are thought to be maximally plastic for a period of 3.5 years [Sharma et al., 2002, 2007]. CI outside this sensitive language period might result in the development of different and delayed patterns of speech and language. Because the period of neurolinguistic development is flexible and varies between children, determining the optimal timing for CI based on these time frames remains difficult and has not yet been strictly defined [Vlastarakos et al., 2010].

The ‘earlier the better' trend [Leigh et al., 2013] in CI originated mainly from results and assumptions from both physiological studies and extrapolation of data of studies with children using hearing aids. Children fitted with hearing aids within the first 2 months of life were found to have significantly better language development than children aided between 3 and 12 months [Mah-rya and Yoshinaga-Itano, 1995], which suggested an indication to start implanting children earlier to accomplish optimal language development and minimize the period of auditory deprivation.

Some surgeons suggest that CI should be performed under the age of 1 year, or even before the age of 6 months [Colletti et al., 2005]. Despite the fact that early implantation is considered to be a predictor of good language and speech development, there is conflicting and incomplete evidence regarding the benefits of implantation under 12 months and particularly under 6 months [Tomblin et al., 2005; Leigh et al., 2013]. An underlying reason for this conflicting evidence could be that early implantation (<9 months) leads to the loss of the ability to discern: the accurate determination of hearing abilities, hearing aid benefit [Tomblin et al., 2005] and coexisting cognitive and behavioral anomalies, which could all affect the performance and outcome following CI.

We aim to identify the existing evidence of the additional benefit on speech and language development and auditory performance of PCI performed within the 1st year of life compared to implantation after 12 months of age.

To systematically identify all relevant studies regarding the influence of age at PCI, we performed a literature search in the following three databases: (1) PubMed, (2) EMBASE and (3) CINAHL (Cumulative Index to Nursing and Allied Health Literature) on April 24, 2014. Because search results can change over time, monthly search updates were checked for additional article inclusions. Up to date, this resulted in no additional article inclusion. We included studies focusing on the assessment of speech and language outcome after PCI by comparing groups of children implanted at different ages. We developed a search strategy by establishing a matrix of synonyms to cover all possible outcome measures of speech perception, speech production and language development following PCI (Appendix). The authors can be contacted to receive the review protocol. The search term ‘age' (or related synonyms as ‘below 1 year') was not included in the search strategy, because age was the principal prognostic factor in the current review.

Two authors (H.B. and I.S.) performed independent systematic title and abstract screening based on predefined selection criteria (fig. 1). Studies that included a comparison between different age groups at the time of PCI were included. Subsequently, the same authors screened the full text of the selected articles for eligibility. As 1999 was the year of FDA approval of implantation before the age of 2 years [Clark, 2000], we included studies published after this date to increase the likelihood to retrieve data for analysis on the study population of interest. Because it could take a significant amount of time before the benefits of early CI can be demonstrated [Beadle et al., 2005], we aimed to include studies with a minimal follow-up of 5 years. If several studies were retrieved that analyzed the same study cohort, the study with the largest sample size was included. Disagreement was solved by discussion between the two authors. No language restrictions were applied.

Two authors (H.B. and F.Z.) independently assessed the methodological quality of included studies. This assessment was performed using a constructed critical appraisal tool assessing both directness of evidence (DoE) and risk of bias (RoB) at both study level and outcome level, represented in 11 and 5 domains, respectively (table 1). Both main domains were graded per complete domain and rated as having a low, moderate or high DoE or RoB. Due to the fact that both long-term follow-up is important and speech/language development should be assessed on various language domains, we selected studies with a high DoE. Consensus on quality assessment was reached by discussion between the authors (H.B. and F.Z.). Publication bias can mark results of PCI. To prevent selective reporting of identified evidence, we aimed to present results on all speech/language outcome assessments that were performed in each independent study.

The same authors (H.B. and F.Z.) collected the following information from studies: authors, publication year, study design and sample size (table 1). The first reviewer (H.B.) independently collected additional information of the included studies on: age group comparisons, the applied speech and language outcome measures, the follow-up visits and the summary of the principal outcome measures. Original results on postoperative speech and language performance from the papers are presented in the tables and classified according to age at implantation. Data were extracted from original articles by magnifying the figure size to 500%. This paper was written according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [Moher et al., 2009]. In the case of homogeneity of: age group comparisons, applied outcome measures, the type of applied statistical analysis and elected follow-up visits, we aimed to combine results of studies in a meta-analysis. When consistency measure I² was below 50%, we performed statistical pooling of the data using Review Manager (RevMan, version 5.3; Department of Informatics and Knowledge Management, The Nordic Cochrane Center, Copenhagen, Denmark).

We retrieved 3,360 articles by performing a search on April 24, 2014. After title and abstract screening, 203 articles were assessed for eligibility in full text by two authors (H.B. and I.S.). Sixty-three articles were selected for inclusion (fig. 1). We screened reference lists of selected articles, which did not result in inclusion of additional articles. Figure 1 shows that 15 articles were excluded after full text assessment. Reasons for exclusion were: the use of similar patient cohorts (6 studies), no direct comparison of various CI age groups (8 studies) and 1 study assessed child's health using visual analogue scale scores instead of assessing speech and language outcomes (fig. 1). Three studies were not written in English: 2 German articles and 1 Turkish article. Both reviewers screened the German articles and both a Turkish ear, nose and throat surgeon (V.T.) and a Turkish student (F.Z.) from our department reviewed the Turkish article.

Ten articles showed to contain a high DoE and were, therefore, included in the current review. RoB of included studies ranged from low to high (table 1). Half of the included studies were retrospective and the other 5 studies were prospective case series (table 1). All included studies represented level 2 evidence (grade B). Twenty-six selected studies showed medium DoE. In addition to the aforementioned articles, we decided to discuss the studies with a moderate DoE (> score above 5) and a medium RoB. Due to this decision, 4 studies were included in addition (italicized in table 1). Only 11 included studies defined the etiology of the deafness of their patients. One study defined whether the cause of deafness was (un)known, but did not define exact etiologies [Manrique et al., 2004]. Although 5 studies did not define whether patients were unilaterally or bilaterally implanted, 9 studies did mention how many implants were used in each individual patient (table 1). Twelve of the 14 studies clearly defined baseline characteristics. The exact age at CI of all implanted individuals was provided in 5 studies. Eight studies provided the mean age at CI and the age range of PCI patients in the study groups. The remaining study provided the mean age at CI in months, but did not define the exact age range of each age group. Six of the included studies tested children on 3 speech and language outcomes: 4 studies on speech production, speech perception and receptive language and 2 studies on auditory performance, speech production and receptive language. Although we aimed to include studies with a minimal follow-up of 5 years, only 3 of the selected studies had a follow-up that was longer than 5 years. Eight studies had a follow-up between 2 and 5 years and 3 studies shorter than 2 years. In 4 studies, the loss to follow-up was less than 20% and in 6 studies above 20%; 3 studies did not report patients lost to follow-up. The applied surgical procedure was described in 4 studies. The type of cochlear device used was mentioned in 12 of the 14 studies. In only 4 studies, outcome measurements were performed according to a defined protocol. Four studies defined the method of handling of missing data. Five studies mentioned the amount of missing data, but did not describe how these missing values were accounted for in their analysis.

The age range of the children in the included studies varied widely (tables 2, 3, 4, 5). We identified 6 studies with a high DoE that compared children implanted under the age of 1 year with children implanted between 12 and 24 months. Inclusion of these studies resulted in the identification of 125 children implanted under the age of 1 year (tables 2, 3, 4, 5). The timing of assessments of children included ranged from preoperative measurements (5 studies) to 9 years of implant experience (IE; tables 2, 3, 4, 5).

The study results are presented in tables 2, 3, 4, 5, subdivided by the four categories of speech and language development: receptive language, speech perception, speech production and auditory performance (tables 2, 3, 4, 5). Studies are enumerated by similarly applied outcome measures to quantify speech and language development. Due to the heterogeneity of groups regarding age at CI, outcome measures, applied statistical analysis and elected follow-up visits, we did not perform statistical pooling of the data (tables 2, 3, 4, 5).

Speech perception of included studies was assessed using consonant-nucleus-consonant (CNC), phonetically balanced kindergarten (PB-K) and Glendonald auditory screening procedure (GASP) scores (table 2).

Leigh et al. [2013] did not find significant differences when comparing CNC word or phoneme scores between groups implanted between 6-12 and 13-24 months at 2 years of IE. However, Dunn et al. [2014] showed that younger implanted children (<24 months) performed superior on combined CNC and PB-K-scores at 5 years of age (p < 0.001) compared to older implanted children. Scores of the 2 groups remained significantly different (p < 0.05) at 8, 9, 10 and 12 years of age. Uziel et al. [2007] tested children on PB-K scores at 5- and 10-year follow-up visits and showed that speech perception skills continued to grow after 5 years of IE: no plateau was reached in the children. The authors detected a positive effect for implantation under the age of 4 years (p < 0.00001). Uziel et al. [2007] showed higher speech perception scores compared to the modeled values at 60 months of Dunn et al. [2014]: 55% at 7 years of age (<2 years) and 48% at 8 years of age (2-4 years). Both studies testing children on GASP scores [Zwolan et al., 2004; Lesinski-Schiedat et al., 2005] found older implanted children (implanted above 12 and 36 months, respectively) to perform better at the 12-month follow-up. However, after 24 months of follow-up, the youngest implanted children (<12 months [Lesinski-Schiedat et al., 2005] and between 1 and 3 years [Zwolan et al., 2004]) outperformed the older implanted children (nonsignificant) in both studies (table 2).

Speech production was assessed using the diagnostic evaluation of articulation and phonology (DEAP), speech intelligibility rate (SIR) and Infant-Toddler Meaningful Auditory Integration Scale (IT-MAIS) in the included studies (table 3). Children are currently implanted in their prelexical period; therefore, a major landmark in their development becomes the onset of babbling [Schauwers et al., 2004]. Normally, babbling occurs between 6-10 months [Koopmans-van Beinum and van der Stelt, 1986; Oller, 1986]. Schauwers et al. [2004] tested the onset of the babbling spurt postoperatively and results showed that the earlier the children were implanted, the more consistent their results were with the development of their normal hearing (NH) peers.

In line with this finding, Leigh et al. [2013] showed a significant effect (p < 0.05) for early implanted children (<12 months) compared to later implanted children at 2-year IE on DEAP scores. Both tested groups performed poorer than their hearing peers (p = 0.002 and p = 0.001, respectively) (table 3).

Two included studies [Uziel et al., 2007; Artières et al., 2009] compared groups implanted before and after 48 months on SIR scores; however, follow-up moments were different (table 3). Artières et al. [2009] assessed young children (<2 years) who reached ceiling scores from 6 years onwards and could not confirm significant differences between groups. Uziel et al. [2007] showed that after 10 years of IE, the mean SIR score of the group of children implanted before 4 years was significantly higher than in children implanted after 4 years (p < 0.0005).

Two studies compared children on IT-MAIS scores at the same follow-up visits: 3, 6 and 12 months [Lesinski-Schiedat et al., 2005; Holman et al., 2013]. Studies compared similar age groups, but 1 study [Holman et al., 2013] combined IT-MAIS with LittlEARS questionnaire scores, whereas only MAIS scores were applied in the study of Lesinski-Schiedat et al. [2005]. Holman et al. [2013] found over 60% of correct scores of their youngest implanted children at the 12-month follow-up. Similarly, Lesinski-Schiedat et al. [2005] found over 70% correct answers at several subsets of the MAIS scores on their youngest group of implanted patients (<12 months) at the 1-year follow-up. In both studies, at 21 months of age [Holman et al., 2013] and 24 months of follow-up [Lesinski-Schiedat et al., 2005], scores of both age groups varied between 80 and 100% correct scores, without differences with respect to age at CI being identified between groups.

Receptive language was measured on oral and written language skills (OWLS), Clinical Evaluation of Language Fundamentals (CELF), Preschool Language Scale (PLS), Peabody Picture Vocabulary Test (PPVT) and Reynell Developmental Language Scale (RDLS) scores (table 4).

Holman et al. [2013] assessed PLS-4 and OWLS scores and concluded that children implanted before 12 months reached speech and language skills by 24 months of age compared to 41 months for the group implanted between 12 and 24 months of age (p < 0.05). Niparko et al. [2010] confirmed significantly higher rates of both comprehension and expression of language (RDLS scores) in children implanted before 18 months compared with children undergoing implantation between 18 and 36 months of age, and especially compared to children implanted after 36 months of age. Manrique et al. [2004] applied similar RDLS scores and showed that children implanted before 36 months had a delay of 2 years compared to NH peers, whereas older (>36 months) implanted children deviated more than 4 years from their NH peers. A comparison on CELF-3 scores [Dunn et al., 2014] indicated that at 7 years of age, the younger-implanted group (<2 years) achieved scores that were on average 12 points higher than the later implanted group (p = 0.01).

Colletti et al. [2012] used the revised PPVT (PPVT-R) to assess receptive language measures, and their results showed that the youngest group (implanted between 2 and 6 months) significantly outperformed children implanted above 1 year (p < 0.001), an effect that remained significant till 48 months of follow-up. A comparison [Colletti et al., 2012] between children implanted before 6 months (n = 12) and at 6-12 months (n = 9) failed to show a significant difference at any follow-up visit. Artières et al. [2009] compared PPVT-R scores of children implanted before and after the age of 2 years, and found a statistically significant difference between groups followed up at 1 year (n = 32) and 2 years (n = 15; p < 0.05) and at 4 and 5 years (table 4). Uziel et al. [2007] showed children implanted before the age of 4 years outperformed children implanted after 4 years at the 10-year follow-up on PPVT-R scores (p < 0.05).

The PPVT-3 and PPVT-4 are highly correlated [Dunn and Dunn, 2007]. Leigh et al. [2013] compared data of groups with different age at CI (<1 vs. >1 year) at the 36-month follow-up, tested on both PPVT scores. The authors found younger children to achieve higher receptive vocabulary scores compared to older implanted children (p = 0.033).

Suh et al. [2009] showed patients implanted before 24 months to catch up with the NH population after 2-3 years (Korean PPVT scores). In addition, after 2 and even after 3 years of follow-up, the children implanted after 3 years caught up with only the 20th percentile of NH children. Therefore, Suh et al. [2009] concluded that 2 years of age seems to be the critical CI time point for children.

In the included studies, auditory performance was assayed using Categories of Auditory Performance (CAP) scores only (table 5).

At the 4-year follow-up, Colletti et al. [2012] found their youngest CI group (implanted between 2 and 6 months) to outperform their peers with later CI on the second version of the CAP (CAP-II) score (p < 0.001). Their results indicated that using the CAP-II score is needed to show differences between children implanted before 2 years of age. No significant difference was seen when their youngest CI group was compared to NH peers. Schauwers et al. [2004] compared children at the 12-month follow-up and found that 80% of the children operated between 6 and 18 months achieved a CAP score between 5 and 6. The authors concluded that children with CI at approximately 18 months of age lag a bit behind their NH peers, whereas those receiving their implant in their 1st year of life follow the normal line (nonsignificant). Suh et al. [2009] found their youngest CI cohort to show a more rapid CAP improvement than their older CI peers. However, this CAP improvement was not significantly different between groups regarding age at CI.

Our review shows consistent evidence for the benefit of early PCI, but the literature remains indistinct on defining the additional speech and language benefits of CI before the age of 12 months. The number of available studies was substantial. Ten of the 14 discussed studies showed to contain high DoE, but RoB ranged from low to moderate.

The best available evidence is based on independent subjective outcome measures and indicates that implantation before the age of 2 is beneficial when considering speech perception (on combined PB-K and CNC but not on GASP scores) [Leigh et al., 2013; Dunn et al., 2014; Lesinski-Schiedat et al., 2005]. Implantation before 12 months resulted in better speech production (DEAP and IT-MAIS) [Ching et al., 2013; Leigh et al., 2013], auditory performance (CAP-II score) [Colletti et al., 2012] and two out of the five receptive language scores (combined PLS-4 and OWLS and PPVT scores) [Colletti et al., 2012; Leigh et al., 2013; Holman et al., 2013]. One study showed that implantation before 6 months resulted in superior 4-year auditory performance (CAP-II) [Colletti et al., 2012]. Although the latter study of Colletti et al. [2012] showed safe and effective results, the majority of ENT surgeons will refrain from this elective surgery before the age of 12 months [Bronstein, 2012]. The exception for performing CI in early infancy remains the occurrence of deafness following meningitis [Bronstein, 2012]. However, in Europe, a trend to implanting children before their 1st year of life is emerging.

Due to concerns regarding unreliable preoperative auditory assessment, underdeveloped anatomy, lack of FDA approval [Bronstein, 2012] and a possibly increased risk for anesthetic complications, CI has not been performed widely in the population under 1 year [Holman et al., 2013]. However, increasing evidence shows that CI can be performed without increased risk of anesthetic and surgical complications in this population. Four selected studies reported on complication rates in children [Colletti et al., 2012; Holman et al., 2013; Lesinski-Schiedat et al., 2005; Manrique et al., 2004] of which 3 studies [Colletti et al., 2012; Holman et al., 2013; Lesinski-Schiedat et al., 2005] included children operated before the age of 12 months. All 3 studies did not report any significant difference in anesthetic or surgical complications between early (<12 months) and later (>12 months) implanted patients. Colletti et al. [2012] did report that young children (2-6 months) experienced a significantly (p < 0.05) higher heart rate; however, this reflected an age-appropiate heart rate for these young children.

We found variation in the recommended age children should be implanted to be able to close gaps in speech and language delays compared to NH peers. The recommended age for implantation varied from 6 to 24 months [Niparko et al., 2010; Boons et al., 2012]. Coene et al. [2011] stated that implantation before the age of 16 months will prevent speech and language delays; however, this small study sample lacked a comparison of age-at-CI groups. Svirsky et al. [2004] used developmental trajectory analysis and showed that implantation before the age of 2 years resulted in significant speech and language advantages. The identified variation in recommended implantation age might be the consequence of the inconsistent, incomplete and conflicting evidence that was identified from the current literature.

To monitor the initial positive ‘age-at-CI effect', adequate longitudinal analysis is essential to account for assessment of confounding effects. For example, Dunn et al. [2014] showed initial significant differences at the 7-year follow-up among age-at-CI groups, but no differences were found at the 10- and 11-year follow-up (table 4). The initial speech and language growth rate might be higher in early implanted children due to auditory stimulation during the sensitive developmental period [Tajudeen et al., 2010]. Alternatively, this group has the advantage of earlier diagnosis, earlier hearing aid intervention, more time to learn to listen with the implant (‘starting early') and earlier education intervention. Therefore, the lower performance level in older implanted children might be a consequence of their lower level of device experience [Tomblin et al., 2005; Dettman et al., 2007]. Tajudeen et al. [2010] underlined this in their analysis by showing that younger implanted children outperformed older implanted children; however, when implanted children were compared at the same follow-up moment postoperatively, there was almost a complete overlap in scores. Therefore, studies that compare children at the same time postoperatively are essential. The CDaCI (Childhood Development after Cochlear Implantation) study [Niparko et al., 2010] is one of the few retrieved prospective, longitudinal studies of auditory and language benefits obviating these limitations.

An earlier review marked the limited and lower quality of evidence for age-at-CI effects on PCI performance [Vlastarakos et al., 2010]. We confirm the lack of level 1 evidence but provide additional evidence from more recent studies [Colletti et al., 2012; Leigh et al., 2013; Holman et al., 2013; Dunn et al., 2014] comparing children implanted before the age of 1 year with children implanted between 12 and 24 months on longer follow-up (>48 months). In addition to the review of Vlastarakos et al. [2010], we assessed receptive language and auditory performance outcome measures.

Due to the recent trend of earlier pediatric implantation, the majority of the children have not yet reached an age in which objective measures can be applied. In addition, elected subjective measures might be too grammatically complex (e.g. GASP scores) for these young children [Zwolan et al., 2004; Lesinski-Schiedat et al., 2005]. Another consequence of the aforementioned trend is that a limited number of children are implanted early and current study samples might be too small to show significant differences between different age-at-CI groups [Lesinski-Schiedat et al., 2005; Tomblin et al., 2005]. Therefore, there is a need for age-normed test standards for both meaningful comparisons of these young implanted children with NH peers and comparisons of study results with respect to age at CI. A recent survey [Uhler and Gifford, 2014] proved the lack of consistency in the preoperative and postoperative selection of speech perception measures across pediatric CI centers. The need for uniform protocols to assess children preoperatively and the development of a working group to establish a standard pediatric postoperative test battery (similar to the adult Minimum Speech Test Battery) was underlined [Uhler and Gifford, 2014]. In addition, variability in CI fitting protocols exist: in a worldwide survey [Vaerenberg et al., 2014], the large variability in all aspects of the CI fitting process and the small role that objective measures play in this process were marked.

Some limitations of this review should be mentioned. First, we refrained from including noncomparative studies of patients implanted before the age of 1 year, because the number of comparative studies provided sufficient direct evidence to address our review query. The fact that positive study outcomes might be more likely to be reported (reporting bias) could have influenced our conclusions. In addition, various confounders are known to influence the post-CI performance [Cheng et al., 1999; Zwolan et al., 2004], such as the communication mode (speech only or speech/sign combined) and the intelligence and participation and support of the child's family during the rehabilitation [Suh et al., 2009]. Therefore, there is a need for additional multivariate analysis in studies to accurately assess the effect of age at CI. The majority of studies consisted of retrospective designs with inconsistent or incomplete language measures and lacked multivariate analysis. Therefore, we applied critical appraisal to select the literature that most adequately corrected for these confounders and lack of transparency in data collection. Third, since language is complex behavior consisting of multiple sensitive periods of various speech and language skills [Houston and Miyamoto, 2010], it is difficult to assess language as one exact outcome measure. By assessing multiple speech and language outcome measures (tables 2, 3, 4, 5), we aimed to assess language development as complete and accurately as possible.

In conclusion, our systematic review provides consistent evidence for early PCI; however, the literature remains indistinct about the additional benefit of implantation before the age of 12 months. The current best evidence showed that early implanted children (<12 months) score better on speech production (DEAP and IT-MAIS-scores), auditory performance (CAP-II score) and on two out of the five receptive language scores (combined PLS-4 and OWLS and PPVT scores) compared to their later implanted peers (>12 months). This evidence consists of cohort studies with a moderate to high RoB; therefore, protocols for standardized preoperative and postoperative evaluations and CI fitting procedures should be developed to allow consistent comparisons of speech and language outcomes between various age-at-CI groups and to gather additional high-level evidence for timely implantation for deaf-born children.

We thank B. Kramer, PhD (information specialist in health and medical sciences at Utrecht University Library), for her help in establishing the search terms for this systematic review.

In addition, we would like to thank Prof. Dr. Govaerts, The EarGroup, Deurne, Belgium, for his feedback and suggestions for improvement in this paper.

W.G. receives unrestricted grants from Cochlear, MED-EL and Advanced Bionics.

None of the authors reports any conflict of interest with any other institution or company.