Study eligibility criteria were defined according to the PICO acronym (population, interventions, comparators, and outcomes) (McKenzie et al., 2019). The review included studies in children (boys and/or girls), aged 0–12 years, with prelingual hearing loss, CI and/or HA users, submitted to cognitive, and auditory and/or language assessments. Studies involving populations with additional disabilities and auditory cognitive training were excluded.

Only primary studies were considered for inclusion, except for case reports. Secondary research, book chapters, annals, dissertations, theses, and animal models were excluded. Only studies published in Portuguese and English were reviewed, and no restrictions were applied regarding the year of publication.

The review protocol was registered in the International Prospective Register of Systematic Review (PROSPERO) database under the registration number CRD42020203974.

Preliminary research was conducted to test, refine, and validate the strategy and terms adopted. The final strategy was reviewed by the authors.

The following databases were searched for potential studies: Virtual Health Library (VHL), Medical Literature Analysis and Retrieval System Online (MEDLINE/PubMed), Cumulative Index to Nursing and Allied Health Literature (CINAHL/EBSCOhost), Scopus, Web of Science (WOS), and EMBASE (Elsevier). CINAHL, WOS, and EMBASE were accessed via the CAPES/MEC Journal Portal.

The search terms were previously identified in Portuguese in the Health Science Descriptors (DeCS), and the corresponding vocabulary in English was identified in the Medical Subject Headings (MeSH). The controlled vocabulary in EMBASE (EMTREE) and CIHNAL (CINAHL Headings) was also consulted.

The strategies were modeled and adapted, when necessary, to ensure a highly sensitive search. All possible combinations were used between the following index terms, synonyms, and free terms: child preschool, child, hearing loss, deafness, cochlear implants, hearing aids, cognition, neuropsychological tests, executive function, memory, attention, auditory perception, speech perception, language, language development. During the research, the language filter was applied to retrieve only studies published in Portuguese and English. There was no limitation regarding the publication period, and retrieved studies had been published by August 2020.

All records retrieved from the databases were imported into the reference manager EndNote Web to remove duplicates. The resulting records were exported to the Rayyan (web version) systematic review manager (Ouzzani et al., 2016) for screening.

Two reviewers (JV and NE) blindly and independently screened study titles and abstracts. Studies whose abstracts were unavailable were removed. Each record was labeled as “included,” “maybe,” or “excluded.” The reasons for excluding the latter were identified.

After such screening, the reviewers’ decisions were unblinded. When they diverged, the conflicting decisions were solved by a third independent reviewer (CF). Then, the selected records were imported to the Zotero reference manager for full-text reading. The CAPES/MEC Journal Portal was used as a primary alternative to retrieve full texts. If they were unavailable in this virtual library, a second attempt was made in ResearchGate. If this also failed, full texts were directly requested from the authors via ResearchGate.

After the retrieval, two reviewers (JV and CF) independently screened the full texts. If they diverged, it was solved through consensus and, if necessary, a third evaluator was consulted (ACR).

Two independent researchers (JV and CF) extracted data into a standardized Microsoft Excel spreadsheet to avoid measurement bias. Extracted data were compared, and any discrepancies were solved by discussion. The following data were extracted from each record: title, authors, year of publication, country of the study, study objective, sample size, participant exclusion and inclusion criteria, sex, chronological age, hearing deprivation time, age at the time of intervention, the device used, length of use, study design, and control group (if any). If the authors did not explain the research design, features of the design were noted.

Results were described regarding subjective and objective research instruments used for the audiological evaluation; speech, language, and cognitive functions; description of the cognitive domains evaluated; evaluation results; associations between tests; and study limitations.

Studies were evaluated according to the Oxford Center for Evidence-Based Medicine: Levels of Evidence (OCEBM Levels of Evidence Working Group, 2011). The levels of evidence were assessed by two independent judges; if they disagreed, a third judge solved the conflict.

Due to outcome measure heterogeneity, it was not possible to conduct a meta-analysis. Therefore, the results were descriptively analyzed, and the summarized data are presented in tables.

The initial search identified 7,309 manuscripts. Duplicates were removed, totaling 4,065 records. Another 162 duplicates had not been identified at first and were manually removed. A total of 3,082 records were screened by title and abstract. Of these, 3,006 were excluded for not meeting the inclusion criteria. The full texts of 74 out of the 76 potential records were retrieved and screened for eligibility – 53 of them were excluded because they did not meet the inclusion criteria. Thus, 21 articles were included in the review. The study selection process with all PRISMA 2020 stages (Page et al., 2021a,b) is shown in Figure 1.

The general study characterization is provided in Table 1 and Supplementary Digital Content 1 (see Table, Supplementary Digital Content 1, which summarizes the characteristics regarding study design, objective, main outcomes, and level of evidence of the 21 studies included in the review). All 21 studies included in the review were published in English between 2002 and 2019. They were conducted in the following countries: the United States (Pisoni and Cleary, 2003; Conway et al., 2011a,b; Stiles et al., 2012; Beer et al., 2014; Bharadwaj et al., 2015; Meinzen-Derr et al., 2018; Davidson et al., 2019; McCreery et al., 2019), the United Kingdom (Figueras et al., 2008; Edwards and Anderson, 2014; Botting et al., 2017), Italy (Bubbico et al., 2007; Colletti et al., 2011), Australia (Dawson et al., 2002), Saudi Arabia (Hassan et al., 2014), South Korea (Lee et al., 2012), Denmark (Udholm et al., 2017), the Netherlands (de Hoog et al., 2016), Iran (Soleymani et al., 2016), and Taiwan (Lo and Chen, 2017).

The sample size of the studies ranged from 10 to 176 participants, totaling 1,098 hearing-impaired children who used CI and/or HA. Approximately 53% of the total population were males. Sex data were not available in four studies (Pisoni and Cleary, 2003; Figueras et al., 2008; Colletti et al., 2011; Conway et al., 2011b). Their mean chronological age ranged from 0.16 to 12.6 years, and the length of CI and/or HA use (hearing age) ranged from 0.04 to 8.7 years (see Table, Supplementary Digital Content 2, which presents the population characteristics). Ten out of the 21 articles did not mention the time of device use in their study populations.

Three studies included patients with mild to profound hearing loss (Botting et al., 2017; Lo and Chen, 2017; Meinzen-Derr et al., 2018), two included patients with mild to severe hearing loss (Stiles et al., 2012; McCreery et al., 2019), two included moderate to profound hearing loss (Bubbico et al., 2007; Figueras et al., 2008), three included severe to profound hearing loss (Hassan et al., 2014; Bharadwaj et al., 2015; Udholm et al., 2017), and 11 studies included patients with profound hearing loss (Dawson et al., 2002; Pisoni and Cleary, 2003; Colletti et al., 2011; Conway et al., 2011a,b; Lee et al., 2012; Beer et al., 2014; Edwards and Anderson, 2014; de Hoog et al., 2016; Soleymani et al., 2016; Davidson et al., 2019).

Regarding device use, 17 studies included participants who received unimodal auditory stimulation. Among them, four studies (Bubbico et al., 2007; Stiles et al., 2012; Lo and Chen, 2017; McCreery et al., 2019) specifically focused on HA users, all of them bilaterally. Eleven studies exclusively approached CI users (Dawson et al., 2002; Pisoni and Cleary, 2003; Colletti et al., 2011; Conway et al., 2011a,b; Lee et al., 2012; Beer et al., 2014; Edwards and Anderson, 2014; Bharadwaj et al., 2015; Soleymani et al., 2016; Udholm et al., 2017) and two studies included both groups (Figueras et al., 2008; Hassan et al., 2014). Additionally, three studies included participants with bilateral auditory stimulation (de Hoog et al., 2016; Botting et al., 2017; Davidson et al., 2019).

It is important to clarify that the studies by Conway et al. (2011a,b) included participants with bimodal stimulation and bilateral CI; however, they were tested with only one CI activated (the first CI). In addition, only one study (Meinzen-Derr et al., 2018) did not make it clear whether auditory stimulation was unimodal or bimodal and whether participants were unilateral or bilateral device users. Data on communication mode, therapeutic approach, and/or setting enrollment were absent in three (Bubbico et al., 2007; Hassan et al., 2014; Udholm et al., 2017) of the 21 studies. Hassan et al. (2014) reported that the children were attending regular language therapy sessions. However, the authors did not specify any details of the communication mode or therapeutic approach. Most of the study population (78.7%) used oral language as the main mode of communication, while 15.6% of participants used total communication (oral language associated with sign language). Only a small portion (5.3%) communicated exclusively through sign language.

The studies by Conway et al. (2011a,b) highlight that, although several of the children had been exposed to sign language, none of them relied exclusively on signs or gestures, and all children were tested using oral-only procedures. In addition, Botting et al. (2017) reported that part of their study population used Sign Supported English (an adapted sign system using English grammar) as their main communication mode. These participants were included in the group of exclusive sign language communication.

The 21 studies used heterogeneous neuropsychological tests/subtests, and some of them used more than one tool (see Table, Supplementary Digital Content 3, which outlines the cognitive tools and subtests used to assess the population and what domains were assessed). Altogether, 46 different cognitive tests/subtests were used to assess the following cognitive domains: STM (auditory and visual), WM (auditory, visual, and visuospatial), nonverbal cognition, reasoning, attention (auditory and visual), executive functions, language, perceptual-motor function, visuoconstructive ability, processing speed, and phonological processing/phonological memory. The Wechsler Intelligence Scale for Children (WISC) was the most frequently used test (47.6%), followed by NEPSY (28.6%) and Leiter International Performance Scale-Revised (19%).

Study findings by Dawson et al. (2002) suggest that a period of auditory deprivation in early years did not cause a specific deficit in auditory STM. Likewise, findings by Pisoni and Cleary (2003) did not show a correlation between auditory digit span and hearing data, such as age at onset of deafness, duration of deafness, age at implantation, and duration of implant use. Furthermore, Bubbico et al. (2007) did not find a statistical difference between the early and late age of enrollment relating to nonverbal IQ. However, the lack of auditory input was associated with lower scores on the cognitive test.

In contrast, Conway et al. (2011b) established a positive correlation between sequence learning and hearing age (duration of implant use). In other words, the longer the children had auditory experiences through the implant, the higher were their scores. Beer et al. (2014) also established a relationship between the Planning/Organization BRIEF subscale and hearing age – the longer the duration of CI use, the fewer the problems with executive functions.

Regarding memory and speech perception tests, Pisoni and Cleary (2003) found a positive correlation between the performance in digit span and scores in spoken word recognition tests with forward digit span, explaining almost 7% of the variance in word recognition scores. Deficits in verbal WM seem to persist even in children with good audibility (Davidson et al., 2019).

According to Dawson et al. (2002), STM performance accounts for significant variance in receptive language scores in children using CI, being spatial memory the strongest predictor. Bharadwaj et al. (2015) also reported a positive correlation between visual WM, visual STM, auditory STM, and reading measures. Opposite findings were stated by Davidson et al. (2019), only simple verbal WM tasks were significantly correlated with vocabulary scores. Thus, complex verbal WM and simple and complex visuospatial WM are not as closely related to language outcomes.

Similar results were reported by Hassan et al. (2014), who suggest that decreased auditory STM in children using CI and HA may be due to associated language impairments. Stiles et al. (2012) also found that children with poorer WM had a smaller vocabulary.

Pisoni and Cleary (2003) showed that early sensory and linguistic experiences immediately after cochlear implantation may have effects on the digit span. Children who were exposed to auditory/oral environments had longer forward digit spans than those in total communication environments; also, digit span was correlated with articulation rate measures. Conversely, another study did not find a significant correlation between digit span and articulation rate measures (Stiles et al., 2012).

Analogous results were obtained by de Hoog et al. (2016) concerning language input. Children enrolled in auditory/oral education performed better on lexical measures than those on total communication.

Findings by Lo and Chen (2017) suggest that language outcomes in Mandarin-speaking children using HA partially depend on their working memory capacity. Therefore, children with higher memory capacity performed as well as their hearing peers on the receptive and expressive language test.

An association between sequence learning abilities and standardized language measures was also reported by Conway et al. (2011b). Children with CI who performed better on sequence learning tasks also had better language outcomes. In another study, Conway et al. (2011a) found that motor sequence skills are closely associated with language outcomes in children using CI.

Language is a complex cognitive function. Lee et al. (2012) designed a study to investigate if metalinguistic skills could be a predictor of receptive vocabulary in children using CI. Results showed that phonological awareness was a significant predictor of vocabulary. Another study reported a relationship between phonological awareness and language skills in CI children, but not in NH children. In other words, language skills clearly predicted phonological awareness outcomes in children with CI (Soleymani et al., 2016).

Some studies sought a relationship between executive functions domain and language. Figueras et al. (2008) found a positive significant association between language ability and executive functions in children using CI and HA and in NH children. CI and HA children performed worse, especially when the test required language skills, suggesting that language and executive functions are interdependent but also dissociable. Therefore, the authors argued that executive function deficits are more likely to be a result of language delay caused by a lack of sensory auditory experiences rather than a period of deafness itself. Botting et al. (2017) investigated whether language mediates executive function differences in children using CI and HA and vice versa. The study findings showed that language not only influences executive functions but also plays a role in mediating their performance. Nevertheless, the reverse association may not happen; hence, poorer executive functions do not necessarily result in poorer language.

Beer et al. (2014) found a significant correlation with executive functions parent reports but not with their objective measures when data were controlled for language.

Colletti et al. (2011) measured the auditory, speech, and nonverbal cognitive function of children using CI implanted under 3 years old and later, in two distinct moments: at 5 years and 10 years post-implantation. Results showed that children implanted at earlier ages had superior performance in auditory, speech, language, and cognitive outcomes even after 10 years of CI use. On the other hand, results by de Hoog et al. (2016) showed that age at the onset of deafness and the first implantation did not correlate with measures of memory, language, and speech perception tests. Moreover, no correlation between memory and outcome performance on language measures was observed. However auditory perception tests contributed to a significant amount of variance in lexical and morphosyntactic language skills.

In the analysis by Edwards and Anderson (2014), age at implantation was a robust predictor of language, speech, and cognitive function performances of children using CI, and accounted for 7–15% of the variance across the outcomes. They also found that auditory memory accounted for 73% of the variability in measures of speech, phonological processing, vocabulary knowledge, and reading, and visual sequential reasoning and visual memory accounted for 16–25%.

No significant correlation was established between audiological data (including age at first HA, age at CI, and unaided pure-tone audiometry) and outcome measures such as verbal WM, visuospatial WM, perceptual fluid reasoning abilities, and receptive vocabulary (Davidson et al., 2019). On the other hand, Meinzen-Derr et al. (2018) created a language performance ratio that reflected language skills in relation to cognitive abilities. The study findings indicate that the degree of hearing loss and audibility (aided SRT) were factors associated with language underperformance. However, they did not find a significant association with the age at which children received the devices. Higher nonverbal IQ was significantly associated with higher language standard scores.

Udholm et al. (2017) investigated the relationship between cognitive skills, auditory capacity, speech perception, and intelligibility in CI children. They found that children whose cognitive performance was equivalent for their age achieved better results on auditory, speech, and language outcomes. Regardless of the years since implantation, CAP and SIR seem to reach a ceiling effect; hence, they are not the best tools to monitor this relationship between skills. The study concluded that vocabulary measured with PPVT-4 seems to reflect the effects of cognitive skills better than CAP and SIR.

McCreery et al. (2019) examined factors that could contribute to speech recognition in adverse listening conditions in children using HA. Their results showed that speech recognition was partially predicted by language, working memory, and auditory attention. In other words, children with better vocabulary and cognitive outcomes have better speech recognition in challenging conditions. Moreover, better aided speech audibility was a positive predictor of language.