Introduction
Hearing loss is a significant global health issue, affecting an estimated 466 million people worldwide and contributing to impaired communication, social withdrawal and reduced quality of life.Reference Podury, Jiam, Kim, Donnenfield and Dhand1 Unilateral sensorineural hearing loss (USNHL) or single-sided deafness (SSD) refers to clinically unaidable hearing loss in one ear and preserved normal hearing in the contralateral ear.Reference Snapp and Ausili2 It has been more specifically defined as a pure-tone audiometric (PTA) average hearing threshold of ≥70 dB across four frequencies (0.5, 1, 2 and 4 kHz) in the poorer hearing ear and ≤35 dB in the better hearing ear.Reference Van de Heyning, Tavora-Vieira, Mertens, Van Rompaey, Rajan and Muller3 SSD disrupts binaural hearing, impairing spatial localisation, temporal processing and speech perception in noisy environments.Reference Qi, Hui and Nunez4 A recent study by Gürses et al.Reference Gürses, Kılıç and Çıldır5 reported that adults with SSD experienced reduced concentration and motivation and greater subjective fatigue compared with normal-hearing controls. These deficits contribute to social withdrawal, reduced occupational productivity and diminished overall well-being.Reference Snapp and Ausili2, Reference Chiossoine-Kerdel, Baguley, Stoddart and Moffat6
Treatment options comprising both implanted and non-implanted modalities have been developed for SSD. For instance, contralateral routing of signals (CROS) hearing aids provides a non-implanted approach to SSD rehabilitation by rerouting sound from the affected ear to the better hearing ear via wireless transmission, improving audibility but not restoring true binaural hearing.Reference Snapp, Holt, Liu and Rajguru7 Bone-conduction devices (BCDs) are another established solution for rerouting signals.Reference Snapp8 The earliest BCD audio processor transferred sound as vibrations through the intact scalp skin to the underlying skull bone, which is described as the skin drive option.Reference Son and Choo9 In this non-implanted approach, the processor was attached to a headband or the frame of spectacles overlying the scalp to achieve snug fixation.Reference Reinfeldt, Håkansson, Taghavi and Eeg-Olofsson10 Due to discomfort from fixation, a plethora of BCD options have since been developed to implant the transducer surgically (i.e. implanted BCD). Implanted treatments include cochlear implant (CI) devices, which seek to restore binaural hearing in patients affected by SSD.Reference Vermeire and Van de Heyning11
Although patient preference is essential in determining which option is selected, patient opinion is seldom studied or reported in the literature. The SSD literature has traditionally focused on audiological outcomes and surgical candidacy considerations.Reference Snapp, Holt, Liu and Rajguru7 Patient-reported outcome measures (PROMs) are tools that allow the measurement of patients’ perspectives on the impact of treatments on their disease-specific health and well-being or health-related quality of life (HRQoL utility). Instruments for measuring HRQoL utility can be classified as generic or disease specific. These approaches differ in whether they measure the impact on dimensions of health relevant to a wide range of health conditions or on a particular disease. Importantly, generic HRQoL utility measures allow for comparative cost–effectiveness analyses with other non-hearing health conditions and interventions and thus inform healthcare policy decision-making.Reference Borre, Kaalund, Frisco, Zhang, Ayer and Kelly-Hedrick12 Health-related quality of life utility measures have been increasingly adopted in recent evaluations of SSD treatments.Reference Kitterick, Lucas and Smith13
Both implanted and non-implanted hearing-assistive devices have been shown to significantly improve patient satisfaction and reduce hearing handicaps.Reference Cho, Park, Seol, Lim, Hong and Moon14, Reference Ellsperman, Nairn and Stucken15 However, it remains unclear whether implanted interventions yield greater patient satisfaction and HRQoL utility compared with non-implanted interventions and which, if any, patients prefer. An earlier systematic review by Kitterick et al.Reference Kitterick, Lucas and Smith13 assessed the impact of CI, BCDs and CROS hearing aids on HRQoL utility and reported small to moderate improvements across all interventions. However, this review did not directly compare outcomes across treatment modalities, leaving uncertainty about which intervention conferred the most significant subjective benefit. More recent, meta-analyses have reported on single treatment options rather than comparing treatments: Hampton et al.Reference Hampton, Milinis, Whitehall and Sharma16 evaluated QoL outcomes with BCDs, while Daher et al.Reference Daher, Kocharyan, Dillon and Carlson17 and Oh et al.Reference Oh, Mavrommatis, Fan, DiRisio, Villavisanis and Berson18 analysed QoL outcomes with CI but did not provide a comparative patient-centred perspective on the range of available SSD treatments.
Therefore, this systematic review and meta-analysis aimed to compare hearing disability and HRQoL utility outcomes between implanted and non-implanted hearing solutions, as evaluated using hearing-specific and generic PROMs, respectively, in adult patients with SSD.
Materials and methods
This study follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.Reference Page, McKenzie, Bossuyt, Boutron, Hoffmann and Mulrow19 The protocol is registered in the International Prospective Register of Systematic Reviews (PROSPERO) (ID: CRD42021286941) and is publicly available to ensure transparency and methodological rigour.20
Search strategy
The Centre for Agriculture and Bioscience International (CAB) Abstracts, Cumulative Index to Nursing and Allied Health Literature (CINAHL), ClinicalTrials.gov, Cochrane Register of Controlled Trials, Embase, International Standard Randomised Controlled Trial Number (ISRCTN), PubMed and Web of Science databases were chosen to achieve comprehensive coverage of clinical trials and observational studies related to audiology and otolaryngology patient-reported outcomes research. These databases were searched for titles and/or abstracts that included: (1) at least one term relating to SSD (‘unilateral’, ‘single-sided’, ‘hearing loss’, ‘deafness’) or that were assigned MeSH terms (‘hearing loss, unilateral’); and (2) at least one term relating to the interventions (‘implant*’, ‘device*’, ‘prosthes*’, ‘instrument*’, ‘hearing aid*’) or that were assigned MeSH terms (‘bone conduction’, ‘hearing aids’, ‘cochlear implants’). The reference lists of selected articles were searched for additional articles.
Study selection
As the first CROS hearing aid was introduced in January 1965,Reference Harford and Barry21 articles published from January 1965 to 1 October 2025, were evaluated for inclusion. Participants, Interventions, Comparators, Outcomes and Study Design (PICOS) selection criteria for identifying eligible studies for inclusion in the current systematic review and meta-analysis are described. Firstly, Participants: patients aged ≥18 years with profound SSD defined as a hearing threshold of ≥70 dB across four frequencies (0.5, 1, 2 and 4 kHz) in the poorer hearing ear and ≤35 dB in the better hearing ear.Reference Van de Heyning, Tavora-Vieira, Mertens, Van Rompaey, Rajan and Muller3 Secondly, Interventions: any implanted or non-implanted hearing-assistive solution, including, but not limited to, BCDs, CROS devices and CIs. Implanted hearing solutions were defined as any that required surgical implantation and non-implanted hearing solutions were those that did not require a surgical procedure. Thirdly, Comparators: all alternative treatments, including no intervention. Fourthly, Outcomes: primary outcomes were post-intervention changes in hearing disability, as measured by hearing-specific PROMs and HRQoL utility, as measured by generic PROMs. Studies reporting either measure were included. The secondary outcome was the assessment of adverse effects associated with the interventions in studies meeting all other inclusion criteria.
Study design: controlled trials and observational studies
The PICOS criteria were used independently by at least two investigators to select studies from the titles and abstracts yielded by the database search. There were no language restrictions except that the abstracts and titles in the databases searched must be in English. Authors of selected articles were contacted for missing information as required. Studies without PROMs data, non-human studies, reviews, protocols, case reports, letters, books or meeting abstracts were excluded. The senior author resolved any discrepancies in the study selection process.
Data extraction
Two investigators extracted data from the eligible studies using a predefined data extraction form comprising the following variables: year of publication, authors, study design, sample sizes, participants’ age and sex distribution, hearing loss severity, hearing solution modality (implanted or non-implanted), PROM used, pre-intervention and post-intervention global PROM scores, time of PROM administration and adverse treatment effects. The PlotDigitizer online tool (Version 3.1.6)22 was used to extract data when it was presented exclusively in graphical form in the eligible studies. The validity and the reliability of this data extraction method have been previously reported.Reference Aydin and Yassikaya23 Disagreements about numerical data extracted from figures were resolved by averaging the data. When data were not reported in the published article, the study authors were contacted via email and given four weeks to respond.
Study quality assessment
Study quality was assessed independently by two investigators and any disagreements were resolved by consultation with the senior author. The investigators were not blinded to study details.
The Cochrane Risk of Bias tool version 2 (RoB-2) was used to assess the risk of bias in RCTs, as it was the preferred tool for evaluating the risk of bias in randomised trials.Reference Farrah, Young, Tunis and Zhao24 The RoB-2 tool assesses bias across five domains: bias arising from the randomisation process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in the measurement of the outcome and bias in the selection of the reported result.Reference Sterne, Savović, Page, Elbers, Blencowe and Boutron25 A high-quality study was defined as having a low RoB score in all assessed domains.
For the non-randomised studies, the risk of bias was evaluated using the Risk of Bias in Non-Randomised Studies of Interventions, Version 2 (ROBINS-I V2) tool.Reference Sterne, Hernán, Reeves, Savović, Berkman and Viswanathan26 This instrument assesses bias across six domains: confounding, participant selection, classification of interventions, missing data, outcome measurement and selective reporting. The bias in each domain was rated as critical, serious, moderate or low. The overall bias was determined by an algorithm provided in the ROBINS-I V2 tool.
Statistical analysis
For data analysis, multiple PROMs from different studies were pooled if substantial correlationsReference Holder, Reynolds, Sunderhaus and Gifford27 or satisfactory validityReference Hampton, Milinis, Whitehall and Sharma16 had been demonstrated between the instruments. This was done to ensure that the intervention and comparator arms each included at least 50 patients, thereby improving the robustness of the analysis.
Meta-analyses comparing the outcomes of implanted and non-implanted solutions with each other or with no treatment were performed. Pooled effect size estimates of differences in mean post-intervention change in outcome scores, standard deviations, 95 per cent confidence intervals (95 per cent CI) and p-values were calculated using inverse variance and Mantel–Haenszel methods. The summary intervention effect was calculated as a weighted average, with larger studies with lower variance and/or more participants given greater weight. The standard error of the summary intervention effect was used to derive a confidence interval and a p-value. The significance cut-off value was set at p = 0.05.
Inter-study heterogeneity was identified and assessed using the I2 test.Reference Higgins and Thompson28 When the I2 test result was less than 50 per cent, a fixed-effect model was used for meta-analysis. When heterogeneity could not be explored or explained by the study’s characteristics or the I2 test result was 50 per cent or greater, a random-effects model was used for meta-analysis. A sensitivity analysis was planned unless a fixed-effects model was adopted. The sensitivity analysis consisted of conducting a fixed-effects model meta-analysis to determine if and to what extent the results changed.Reference Higgins, Thomas, Chandler, Cumpston, Li and Page29
Publication bias was visually assessed based on the asymmetry of the funnel plot of standard mean difference and standard error of post-intervention change in outcome (PROMs) scores and confirmed statistically using a p-value cut-off of <0.05 on Egger’s test.Reference Egger, Davey Smith, Schneider and Minder30
All meta-analyses comparing the outcomes of implanted or non-implanted solutions were performed using Review Manager software (RevMan version 5.3; The Nordic Cochrane Center, The Cochrane Collaboration, 2014). However, the Comprehensive Meta-Analysis Version 4 software (Biostat, Englewood, New Jersey, USA)Reference Borenstein, Hedges, Higgins and Rothstein31 was used to generate funnel plots and conduct Egger’s test.
Results
Study selection
Of the 7234 records retrieved from the comprehensive database search, 2355 were duplicate records and were thus excluded from the screening process. In total, 4879 records were screened based on title and abstract, of which 4794 were excluded for not meeting our inclusion criteria; the remaining 85 articles progressed to the full-text screening stage. Three of the 85 articles could not be retrieved despite contacting the corresponding author. Therefore, 82 full-text records were successfully retrieved, of which one was a conference abstract and was therefore excluded. Thirteen of the remaining 81 records met the inclusion criteria and were included in the systematic review and meta-analysis.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32–Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 Study screening and selection process adhered to the PRISMA 2020 guidelinesReference Page, McKenzie, Bossuyt, Boutron, Hoffmann and Mulrow19 (Figure 1).
PRISMA flowchart.
Figure 1 Long description
The process of identifying studies through databases and registers is outlined. Initially, 7234 records were identified from databases, with none from registers. Before screening, 2355 duplicate records were removed, leaving 4879 records to be screened. Of these, 4794 were excluded. The remaining 85 reports were sought for retrieval, but 3 were not retrieved. Thus, 82 reports were assessed for eligibility. Reports were excluded for reasons such as inappropriate timing of intervention (12), wrong study design (15), no desired outcomes (14), no global PROMS (10) and wrong population (16). Ultimately, 13 studies were included in the review.
Characteristics of selected studies
The included studies were prospective, pre-intervention and post-intervention observational studies in which patients acted as their own controls (Table 1). Here, 12 of the 13 included studies had a single-arm, non-randomised design. One study was a randomised controlled trial (RCT).Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 The studies were conducted in different settings, including the USA (n = 3),Reference Dillon, Buss, Rooth, King, Deres and Buchman34, Reference Newman, Sandridge and Wodzisz39, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 Germany (n = 2),Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 the Netherlands (n = 2),Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 China (n = 1),Reference Yang, Wang, Huang, Chai, Jia and Wu43 South Korea (n = 1),Reference Ryu, Moon, Byun, Jin, Park and Jang41 Colombia (n = 1),Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35 France (n = 1)Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36 and two were European multi-centred studies.Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38 Collectively, these 13 studies included data from 403 participants. The mean age of the participants reported in individual studies ranged from 41.7 to 57 years.
Characteristics of the Studies Included in the Review
Table 1 Long description
Study characteristics are listed for hearing loss thresholds, number treated, mean age with standard deviation, study group, intervention, global patient reported outcome measure scores with standard deviation for APHAB, SSQ, and HUI3, and follow up duration. Bosman, et al. in the Netherlands enrolled 7 patients with an intervention group using CROS and a control group with no hearing assistive solution, reporting APHAB 31.2 for intervention and 38.58 for control at 1 month after the intervention. Newman, et al. in the USA enrolled 8 patients, mean age 55.7 with standard deviation 6.7, comparing BCD with no hearing assistive solution, reporting APHAB 33.7 for intervention and 61.7 for control at 12 months after the intervention. Arndt, et al. in Germany enrolled 11 patients, mean age 42.6 with standard deviation 12.7, reporting SSQ 3.7 for CROS, 3.6 for BCD, 6.43 for CI, and 3.6 for control at 6 months after the intervention. Dillon, et al. in the USA enrolled 20 patients, mean age 50 with standard deviation 11.5, with CI intervention and patient as own control, reporting APHAB 49.8 with standard deviation 18.7 and SSQ 4.0 with standard deviation 1.4 after CI surgery, and APHAB 20.1 with standard deviation 8.5 and SSQ 6.7 with standard deviation 1.3 with no hearing assistive solution at 6 months after CI surgery. Ryu, et al. in South Korea enrolled 21 patients, mean age 41.7 with standard deviation 13.7, with CROS intervention and patient as own control, reporting SSQ 5.8 with standard deviation 1.9 after the CROS hearing aid and SSQ 7.5 with standard deviation 1.2 with no hearing assistive solution at 2 to 4 weeks after the CROS hearing aid. Yang, et al. in China enrolled 15 patients, mean age 42.7 with standard deviation 11.2, with BCD intervention and patient as own control, reporting APHAB 46.5 with standard deviation 18.5 after BCD implant and APHAB 28.5 with standard deviation 13 with no hearing assistive solution at 6 months after BCD implant. Garcia, et al. in Colombia enrolled 10 patients, mean age 57 with standard deviation 9.8, with CI intervention and patient as own control, reporting APHAB 43.8 with standard deviation 24.8 after CI surgery and APHAB 33.4 with standard deviation 20 with no hearing assistive solution at 6 months after CI surgery. Leterme, et al. in France enrolled 18 patients, mean age 50 with standard deviation 11.6, listing intervention groups CROS, BCD Headband, and BCD Invasive with patient as own control, reporting APHAB 44.9 with standard deviation 15.9 for CROS and 36.7 with standard deviation 13.17 for BCD Headband at 3 months after implantation, with other PROM cells left blank. Ramos Macias, et al. 2015 in a European multicentre study enrolled 13 patients with CI intervention and patient as own control, reporting SSQ 5.3 with standard deviation 1.93 for intervention and 4.1 with standard deviation 0.98 for control at 6 months. Ramos Macias, et al.’s 2018 European multicentre study enrolled 16 patients with CI intervention and patient as own control, reporting SSQ 6.3 for intervention and 4.2 for control, and HUI3 0.63 for intervention and 0.45 for control at 12 months after activation. Peters, et al. in the Netherlands enrolled 120 patients, mean age 53.2 with standard deviation 13.1, listing intervention groups CROS, BCD Invasive, and CI plus a control group with no hearing assistive solution, reporting APHAB 45.25 with standard deviation 14.68 for CROS, 42 with standard deviation 6.78 for BCD Invasive, 42.25 with standard deviation 8.41 for CI, and 46.25 with standard deviation 17.86 for control at 3 months after implantation. Plath, Marienfeld in Germany enrolled 100 patients, mean age 55.3 with standard deviation 16.90, with CI intervention and patient as own control, reporting APHAB 46.16 with standard deviation 10.56 for intervention and 46.80 with standard deviation 13.03 for control at 3 months after implantation. Ullah, et al in the USA enrolled 44 patients, mean age 51.6 with standard deviation 14.0, with CI intervention and patient as own control, reporting SSQ 0.83 with standard deviation 0.14 for intervention and 0.56 with standard deviation 0.17 for control, and HUI3 0.83 with range 0.72 to 0.91 for intervention and 0.56 with range 0.48 to 0.71 for control at 3 months after activation.
The definition of deafness in the affected ear varied across the studies. Most studies defined deafness as a PTA threshold of ≥70 dB,Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Dillon, Buss, Rooth, King, Deres and Buchman34, Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ryu, Moon, Byun, Jin, Park and Jang41, Reference Yang, Wang, Huang, Chai, Jia and Wu43, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 while fewer studies used higher thresholds of >80 dBReference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 or ≥90 dB.Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Newman, Sandridge and Wodzisz39 The definition of normal hearing in the contralateral ear also varied with respect to the PTA threshold and the frequencies used to determine it. A threshold of ≤30 dBReference Bosman, Hol, Snik, Mylanus and Cremers33, Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42–Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 was the most commonly cited value, but thresholds of ≤25 dB,Reference Ryu, Moon, Byun, Jin, Park and Jang41 ≤35 dB,Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Dillon, Buss, Rooth, King, Deres and Buchman34, Reference Newman, Sandridge and Wodzisz39 or <50 dBReference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37 were also reported. The interventions in the studies included CI,Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Dillon, Buss, Rooth, King, Deres and Buchman34, Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 BCDs,Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Newman, Sandridge and Wodzisz39, Reference Yang, Wang, Huang, Chai, Jia and Wu43, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 and CROS.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Ryu, Moon, Byun, Jin, Park and Jang41, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44
The minimum follow-up duration varied considerably and depended on the type of intervention. Studies assessing outcomes after receiving implanted hearing solutions, such as implantable BCDs and CIs, followed patients for at least 3 monthsReference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 and up to 12 months.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Dillon, Buss, Rooth, King, Deres and Buchman34–Reference Newman, Sandridge and Wodzisz39, Reference Yang, Wang, Huang, Chai, Jia and Wu43 In contrast, follow-up periods were far shorter for studies assessing non-implanted solutions, at two to four weeks after intervention with CROS hearing aids.Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Ryu, Moon, Byun, Jin, Park and Jang41 This variation reflected the fact that CI and implanted BCD surgical procedures require a period of time for wound healing before the external hearing component can be fitted. For CI, tuning and recipient training to adjust to the non-sound CI stimulus must occur before the optimal effect of treatment can be achieved. In contrast, CROS hearing aids and BCDs worn on headbands utilise a non-implanted approach, requiring no wound-healing period; they present a rerouted, already familiar sound stimulus to the recipient’s better-functioning ear. Therefore, no protracted period of tuning and retraining is required before the treatment’s effect can be assessed.
Instruments used to evaluate hearing disability
Among the included studies, seven studies assessed hearing disability exclusively with the Abbreviated Profile of Hearing Aid Benefit (APHAB),Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Newman, Sandridge and Wodzisz39, Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40, Reference Yang, Wang, Huang, Chai, Jia and Wu43, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 five with the Speech, Spatial and Qualities of Hearing scale (SSQ),Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ryu, Moon, Byun, Jin, Park and Jang41, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 and one used both APHAB and SSQ scores.Reference Dillon, Buss, Rooth, King, Deres and Buchman34 The APHAB is a 24-item questionnaire comprising four subscales: Ease of Communication (EC), Reverberation (RV), Background Noise (BN) and Aversiveness (AV), each rated on a 0–100 per cent scale; the total score is the average of all subscale scores.Reference Cox and Alexander45, Reference Johnson, Cox and Alexander46 Lower scores indicated less difficulty and greater benefit from hearing aids. The SSQ is a 50-item questionnaire grouped into three subscales: speech, spatial and other qualities of hearing, such as clarity or listening effort.Reference Gatehouse and Noble47 Each item is rated on a 0–10 scale and scores are averaged across the three subscales to derive the SSQ global score, with higher scores indicating better hearing performance.
As previous validation studies have demonstrated a high correlation (r = 0.6084; p < 0.0001) between SSQ and APHAB global scores,Reference Holder, Reynolds, Sunderhaus and Gifford27 global SSQ scores were standardised to the APHAB scale for data synthesis. The global SSQ score was reversed by subtracting 10 from the score, then scaled to the APHAB range by multiplying by 10. Reported SDs were scaled by multiplying them directly by 10 without reversing.
The included studies also reported other hearing-related PROMs that commonly focused on other symptoms, such as the Tinnitus Handicap Inventory (THI; n = 5).Reference Dillon, Buss, Rooth, King, Deres and Buchman34, Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 Others symptoms were supplementary measures of hearing disability in a small number of studies that included APHAB or SSQ scores, for example International Outcome Inventory for Hearing Aids (IOI-HA; n = 2),Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ryu, Moon, Byun, Jin, Park and Jang41 and the Single-Sided Deafness Questionnaire (SSDQ; n = 2),Reference Newman, Sandridge and Wodzisz39, Reference Yang, Wang, Huang, Chai, Jia and Wu43 Nijmegen Cochlear Implant Questionnaire (NCIQ; n = 1),Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 Glasgow Hearing Aid Benefit Profile (GHABP; n = 1),Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36 Hearing Handicap Inventory for the Elderly (HHIE; n = 1),Reference Ryu, Moon, Byun, Jin, Park and Jang41 Hearing Handicap Inventory for Adults (HHIA; n = 1),Reference Newman, Sandridge and Wodzisz39 Spatial Hearing Questionnaire (SHQ; n = 1)Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38 and Hearing Participation Scale (HPS; n = 1).Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 Scores from these and other rarely used hearing disability PROMs were not assessed in this review because they were redundant for hearing disability change analysis purposes or lacked evidence of comparability to the APHAB scale.
Instruments used to evaluate health-related quality of life utility
The included studies reported scores from three different accepted generic HRQoL utility questionnaires: Health Utilities Index (HUI; n = 3),Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 Glasgow Benefit Inventory (GBI; n = 2)Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 and Short Form 36 (SF-36; n = 1).Reference Newman, Sandridge and Wodzisz39 We used HUI as the measure of HRQoL utility as it was the most widely used generic HRQoL utility PROM in the studies selected for this study and it is also widely used in the QoL literature.Reference Kaplan and Hays48 It covers multiple health attributes, such as vision, hearing, speech, mobility/dexterity, emotion, cognition and pain, thus providing a comprehensive and standardised framework for comparing different health conditions.Reference Horsman, Furlong, Feeny and Torrance49
When using PROMs, statistically significant changes in scores may not have correlated with meaningful improvements in patient experience.Reference Guyatt, Walter and Norman50 For the APHAB, a reduction of 10 per cent between scores suggested reasonable certainty that the changes represented a minimal clinically important difference (MCID) between conditions.Reference Cox51 For the SSQ, a change of 1 unit on the SSQ score indicated MCID.Reference Anderson, Rallapalli, Schoof, Souza and Arehart52 The MCID of the generic HUI3 was estimated at 0.03.Reference Horsman, Furlong, Feeny and Torrance49
Meta-analysis of hearing disability
Data from the 13 studies included in the systematic review allowed for three separate meta-analyses comparing implanted hearing solutions versus no hearing solution, CROS hearing solutions versus no hearing solution and CROS hearing solutions versus BCD non-implanted hearing solutions. Meta-analyses for these comparisons are presented separately in this review.
Implanted hearing solution vs no hearing solution
Five studiesReference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Newman, Sandridge and Wodzisz39, Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40, Reference Yang, Wang, Huang, Chai, Jia and Wu43, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 evaluated changes in participants’ hearing disabilities after implanted hearing solutions using the APHAB, while two studiesReference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37 used the SSQ and one studyReference Dillon, Buss, Rooth, King, Deres and Buchman34 used both. The global score of the SSQ, as reported by Arndt et al.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32 and Ramos Macias et al.Reference Ramos Macías, Falcón González, Manrique, Morera, García-Ibáñez and Cenjor37 was converted to the APHAB scoring scale. Only the global APHAB score from Dillon et al.Reference Dillon, Buss, Rooth, King, Deres and Buchman34 was selected for meta-analysis to reduce sampling bias.
Pooled analysis favoured implanted hearing solutions [mean difference (MD) = −14.71 (95%CI: −23.48, −5.93), p < 0.00001] indicating that a significantly greater number of patients reported hearing disability improvement over that of no intervention (Figure 2). Five of the eight studies individually reported a statistically significant advantage of implanted hearing solutions over no hearing solution, the three exceptions being studies by Plath et al.,Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 and Garcia et al.Reference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35 Importantly, the pooled MD of 14.7% exceeded the APHAB’s MCID of 10%. However, considerable heterogeneity was noted among the included studies (I2 = 88%).
Forest plot comparing the mean post-intervention change in hearing disability score (APHAB) with implanted hearing solutions and no hearing solutions.
Figure 2 Long description
Study of Subgroup. Columns labeled, Implanted hearing solutions with subcolumns Mean, Standard deviation, Total; No hearing solutions with subcolumns Mean, Standard deviation, Total; Weight; Mean Difference with subcolumns, I V, Random, 95 percent confidence interval. The horizontal scale is labeled Mean Difference, I V, Random, 95 percent confidence interval, with tick labels minus 50, minus 25, 0, 25, 50. Text under the scale reads, Implanted hearing solutions and No hearing solutions. A vertical reference line is drawn at 0. Rows listed as: Anders et al. 2010. Dillon et al. 2018. Garcia et al. 2020. Ramos Macias et al. 2015. Newman et al. 2018. Peters et al. 2021. Plath et al. 2021. Yang et al. 2018. Numeric entries per row: Anders et al. 2010: Implanted hearing solutions mean 4.2, standard deviation 19, total 3; No hearing solutions mean 0, standard deviation 32, total 3; Weight 1.2 percent; Mean difference 4.20, 95 percent confidence interval minus 27.47 to 35.87. Dillon et al. 2018: Implanted hearing solutions mean 35.1, standard deviation 7.55, total 21; No hearing solutions mean 48.1, standard deviation 17.1, total 21; Weight 13.3 percent; Mean difference minus 13.00, 95 percent confidence interval minus 20.62 to minus 5.38. Garcia et al. 2020: Implanted hearing solutions mean 28.1, standard deviation 7.8, total 10; No hearing solutions mean 34.8, standard deviation 9.6, total 10; Weight 3.7 percent; Mean difference minus 6.70, 95 percent confidence interval minus 14.44 to 1.04. Ramos Macias et al. 2015: Implanted hearing solutions mean 4.7, standard deviation 19.3, total 13; No hearing solutions mean 59, standard deviation 9.8, total 13; Weight 12.0 percent; Mean difference minus 54.30, 95 percent confidence interval minus 62.37 to minus 46.23. Newman et al. 2018: Implanted hearing solutions mean 43.26, standard deviation 15.1, total 8; No hearing solutions mean 61.7, standard deviation 7.21, total 8; Weight 8.9 percent; Mean difference minus 18.44, 95 percent confidence interval minus 28.10 to minus 8.78. Peters et al. 2021: Implanted hearing solutions mean 32.59, standard deviation 8.41, total 24; No hearing solutions mean 46.85, standard deviation 12.86, total 24; Weight 13.4 percent; Mean difference minus 14.26, 95 percent confidence interval minus 20.01 to minus 8.51. Plath et al. 2021: Implanted hearing solutions mean 46.16, standard deviation 10.56, total 100; No hearing solutions mean 46.8, standard deviation 13.83, total 100; Weight 50.6 percent; Mean difference minus 0.64, 95 percent confidence interval minus 4.39 to 3.10. Yang et al. 2018: Implanted hearing solutions mean 28.5, standard deviation 13, total 15; No hearing solutions mean 46.5, standard deviation 18.5, total 15; Weight 10.5 percent; Mean difference minus 18.00, 95 percent confidence interval minus 29.44 to minus 6.56. Total row: Total 201 for Implanted hearing solutions and 201 for No hearing solutions; Weight 100.0 percent; Mean difference minus 14.71, 95 percent confidence interval minus 23.48 to minus 5.93. Text below: Heterogeneity, Tau squared equals 134.02, Chi squared equals 59.87, degrees of freedom equals 7, P less than 0.00001, I squared equals 88 percent. Test for overall effect, Z equals 3.29, P equals 0.001.
Contralateral routing of signals hearing solutions vs no hearing solution
Four studiesReference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Ryu, Moon, Byun, Jin, Park and Jang41, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 provided data for comparing changes in hearing disability with CROS hearing solutions versus no hearing solution. The meta-analysis yielded a non-significant MD of −8.75 (95%CI: −21.94, 4.45; p = 0.19) (Figure 3). The inter-study heterogeneity among the included studies remained considerable (I2 = 88%) and was likely to be due to variability in device technology and study design.
Forest plot comparing the mean post-intervention change in hearing disability score (APHAB) with CROS hearing solutions and no hearing solutions.
Figure 3 Long description
Study of Subgroup. A forest plot with a table and a mean difference axis. The table columns are, Study or Subgroup; CROS Hearing Solution with subcolumns Mean, Standard Deviation, Total; No Hearing Solution with subcolumns Mean, Standard Deviation, Total; Weight; Mean Difference with subcolumns Inverse Variance, Random, 95 percent Confidence Interval. Four study rows are listed: Arndt et al, 2011; Bosman et al, 2003; Peters et al, 2021; and Ryu et al, 2015. Arndt et al, 2011 shows CROS Hearing Solution mean 0.27, standard deviation 17.47, total 11; No Hearing Solution mean 0.64, standard deviation 16.71, total 11; weight 27.9 percent; mean difference minus 0.37 with 95 percent confidence interval minus 15.93 to 15.19. Bosman et al, 2003 shows CROS Hearing Solution mean 53.17, standard deviation 14.8, total 17; No Hearing Solution mean 40.67, standard deviation 23.15, total 11; weight 17.9 percent; mean difference 12.50 with 95 percent confidence interval minus 3.53 to 28.53. Peters et al, 2021 shows CROS Hearing Solution mean 45.25, standard deviation 14.68, total 32; No Hearing Solution mean 46.25, standard deviation 17.86, total 24; weight 27.6 percent; mean difference minus 1.00 with 95 percent confidence interval minus 10.01 to 8.01. Ryu et al, 2015 shows CROS Hearing Solution mean 42.5, standard deviation 16.04, total 21; No Hearing Solution mean 42.5, standard deviation 16.04, total 21; weight 26.5 percent; mean difference minus 20.00 with 95 percent confidence interval minus 33.33 to minus 6.67. A total row shows total 81 for CROS Hearing Solution and total 67 for No Hearing Solution, with weight 100.0 percent and pooled mean difference minus 8.75 with 95 percent confidence interval minus 21.94 to 4.45. Below the table, heterogeneity text reads, Tau squared equals 143.82; Chi squared equals 25.38, degrees of freedom equals 3, P equals 0.00001; I squared equals 88 percent. Test for overall effect reads, Z equals 1.30, P equals 0.19. The right side plot shows one square marker per study with a horizontal confidence interval line and a diamond for the pooled estimate. The horizontal axis label reads, Mean Difference, Inverse Variance, Random, 95 percent Confidence Interval. The scale shows tick labels at minus 50, 0 and 50. The left side under the axis reads, CROS Hearing Solution and the right side reads, No Hearing Solution.
Contralateral routing of signals hearing solutions vs bone-conduction devices non-implanted hearing solutions
Four studiesReference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Bosman, Hol, Snik, Mylanus and Cremers33, Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 reported changes in hearing disability with CROS and BCD non-implanted hearing solutions. A meta-analysis of the four studies revealed that patients using the BCD headband had a significantly greater APHAB score change than those using CROS [MD = 4.70 (95%CI: 0.27, 9.14); p = 0.04], with no inter-study heterogeneity (I2 = 0%) (Figure 4). However, an MD of 4.7% did not cross the 10% APHAB MCID threshold, suggesting no clinically meaningful difference in APHAB scores.
Forest plot comparing mean post-intervention change in hearing disability score (APHAB) with CROS hearing solutions and BCD headbands.
Figure 4 Long description
Study or Subgroup. A forest plot with a table on the left and a horizontal mean difference axis on the right. Table columns are labeled, Study or Subgroup; CROS with subcolumns Mean, Standard Deviation, Total; BCD Headband with subcolumns Mean, Standard Deviation, Total; Weight; Mean Difference with subcolumns Inverse Variance, Random, 95 percent Confidence Interval. Row 1 label, Peters et al., 2021. CROS Mean 27.7, Standard Deviation 17.47, Total 11. BCD Headband Mean 61.7, Standard Deviation 20.06, Total 11. Weight 10.6 percent. Mean Difference 34.00, 95 percent Confidence Interval 15.08 to 52.92. Row 2 label, Lettermeier et al., 2016. CROS Mean 30.17, Standard Deviation 14.7, Total 9. BCD Headband Mean 27.8, Standard Deviation 10.59, Total 9. Weight 17.0 percent. Mean Difference minus 2.37, 95 percent Confidence Interval minus 14.05 to 9.31. Row 3 label, Bosman et al., 2003. CROS Mean 44.9, Standard Deviation 15.9, Total 18. BCD Headband Mean 36.7, Standard Deviation 13.17, Total 18. Weight 21.6 percent. Mean Difference minus 8.20, 95 percent Confidence Interval minus 18.34 to 1.74. Row 4 label, Arndt et al., 2010. CROS Mean 45.25, Standard Deviation 14.68, Total 32. BCD Headband Mean 42.76, Standard Deviation 20.14, Total 32. Weight 50.8 percent. Mean Difference minus 2.49, 95 percent Confidence Interval minus 10.34 to 5.36. A summary row labeled Total with CROS Total 70 and BCD Headband Total 70. Weight 100.0 percent. Mean Difference 4.70, 95 percent Confidence Interval 0.27 to 9.14. Additional text below the table reads, Total, 95 percent Confidence Interval 4.70, 0.27, 9.14. Heterogeneity, Tau squared equals 0.00; Chi squared equals 1.12, degrees of freedom equals 3, P equals 0.77; I squared equals 0 percent. Test for overall effect, Z equals 2.08, P equals 0.04. The right side shows square markers with horizontal confidence interval lines for each study and a diamond for the total. The horizontal axis is labeled Mean Difference, Inverse Variance, Random, 95 percent Confidence Interval, with tick labels minus 50, minus 25, 0, 25, 50. Text under the axis reads, Favours CROS on the left and Favours BCD Headband on the right. A vertical reference line is drawn at 0.
Meta-analysis of generic quality of life
Four studies reported on generic HRQoL utility measures; three used the HUI3Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 and Newman et al.Reference Newman, Sandridge and Wodzisz39 used the SF-36. However, Newman et al.Reference Newman, Sandridge and Wodzisz39 did not report global SF-36 scores in patients undergoing BCD surgical implantation and these were therefore not included in our meta-analysis. This allowed only a comparison of generic QoL among patients using CI versus those with no hearing solution. The PlotDigitizer online tool (Version 3.1.6)22 was used to extract the standard deviation in pre and post-Cl activation from one study.Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38 Pooled analysis of data from three studiesReference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 with 123 participants indicated that CI significantly improved HUI3 scores compared with no hearing solution [MD 0.21 (95%CI: 0.10, 0.31); p = 0.0001], albeit with high heterogeneity (I2 = 66%) (Figure 5). Importantly, the MD of 0.21 exceeds the HUI3’s MCID threshold of 0.03.
Forest plot comparing mean post-intervention change in generic health-related quality of life measure (HUI3) among patients using cochlear implants vs no hearing solution.
Figure 5 Long description
Study of Subgroup, Cochlear Implant, No Hearing Solution, Mean Difference and Mean Difference. A forest plot with a table on the left and a horizontal effect scale on the right. The table columns are labeled, Study of Subgroup; Cochlear Implant with subcolumns Mean, Standard Deviation, Total; No Hearing Solution with subcolumns Mean, Standard Deviation, Total; Weight; Mean Difference, inverse variance, Random, 95 percent confidence interval. A second header on the right reads, Mean Difference, inverse variance, Random, 95 percent confidence interval. Three study rows are listed. Arndt et al. 2010 shows Cochlear Implant mean 0.80, standard deviation 0.12, total 11; No Hearing Solution mean 0.56, standard deviation 0.27, total 11; weight 21.7 percent; mean difference 0.24, 95 percent confidence interval 0.07 to 0.41. Macias et al. 2018 shows Cochlear Implant mean 0.57, standard deviation 0.09, total 16; No Hearing Solution mean 0.44, standard deviation 0.16, total 18; weight 36.6 percent; mean difference 0.12, 95 percent confidence interval 0.02 to 0.22. Ullah et al. 2023 shows Cochlear Implant mean 0.83, standard deviation 0.14, total 25; No Hearing Solution mean 0.56, standard deviation 0.17, total 44; weight 41.6 percent; mean difference 0.27, 95 percent confidence interval 0.20 to 0.34. A total row reads, Total, 95 percent confidence interval, with Cochlear Implant total 55, No Hearing Solution total 73, weight 100.0 percent and mean difference 0.21, 95 percent confidence interval 0.10 to 0.32. Below the table, heterogeneity lines read, Heterogeneity: Tau squared equals 0.01; Chi squared equals 5.93, degrees of freedom equals 2, P equals 0.05; I squared equals 66 percent. A final line reads, Test for overall effect: Z equals 3.89, P equals 0.0001. On the right, the horizontal scale is labeled with tick labels minus 0.5, minus 0.25, 0, 0.25 and 0.5. The left side under the scale reads, No Hearing Solution and the right side reads, Cochlear Implant. Each study is shown as a square with a horizontal confidence interval line and the pooled result is shown as a diamond. The vertical reference line is at 0. The three study squares and their confidence intervals are positioned to the right of 0, matching the listed mean differences and 95 percent confidence intervals. The pooled diamond spans 0.10 to 0.32 with center at 0.21.
Adverse events
Two studies reported BCD-related adverse events: Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 reported 8% (n = 2) serious adverse events, both of which were implant extrusions, among 25 patients who opted for BCD implantation. Local pain in 36% (n = 5) was the most commonly reported adverse effect among the 14 patients who had undergone BCD implantation described in the study by Leterme et al.,Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36 followed by hematoma 14% (n = 2), physical discomfort 7% (n = 1) and local abscess 7% (n = 1). No device-related adverse events were noted in the study by Yang et al.Reference Yang, Wang, Huang, Chai, Jia and Wu43 among 15 patients who underwent BCD implantation.
Ramos Macias et al.Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38 reported no major post-surgical complications in 16 patients, after CI implantation. Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 reported no device-related complications in their 34 CROS group patients, while Leterme et al.Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36 reported physical discomfort in 38.9% (n = 7) and headache in 5.6% (n = 1) of 18 patients in the CROS group.
In summary, the incidence of adverse events was 20.3% (n = 11) among the 54 patients who underwent BCD surgical implantation, 0% among the 16 patients who received CI and 19% (n = 8) among the 42 patients who received CROS. However, adverse-event reporting following CROS and BCD implantation appeared to be comprehensive; for instance, discomfort or pain accounted for all the CROS adverse events. CI adverse-event reporting, conversely, was restricted to ‘major surgical’. It is very unlikely that none of the patients undergoing CI experienced some postoperative discomfort.
Risk of bias assessment
The risk of bias assessment for the RCT was conducted using the Cochrane Collaboration’s RoB2, per our PROSPERO-registered study protocol.20 However, as 12 of the 13 included studies were non-randomised, the risk of bias was assessed using the ROBINS-I V2 tool.Reference Farrah, Young, Tunis and Zhao24
The one RCT by Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 was assessed as having a low risk of bias using the RoB2 tool (Table S1), thereby strengthening the validity of their findings. The risk of bias assessment of the 12 observational studies using the ROBINS-I V2 tool identified three as having a serious risk of bias and nine as having a moderate risk of bias. The three studies with serious risk of bias ratings had high attrition rates (Plath, 2021), incomplete follow-up (Ramos Macias, 2015) and were retrospective studies with small samples (≤10) and no adjustment for the reported results (Garcia et al., 2019). For the other studies, a moderate risk of bias was assigned for some domains. The main limitations of these studies included limited adjustment for baseline confounding factors, inadequate description of screening and exclusion processes and small sample sizes (Table S1).
This risk of bias determination affects the reliability of the conclusions to be drawn from each of the individual meta-analyses depending on the level of the risk of bias determined for the included studies.
Publication bias
Changes in hearing disability
Two funnel plots were generated to separately assess publication bias for the primary outcome (change in APHAB scores) in implanted and non-implanted surgical hearing solutions. A further funnel plot was generated to separately assess publication bias in studies that compared changes in APHAB scores between CROS and BCD headband hearing solutions. Studies reporting changes in hearing disability outcomes with implanted solutions (n = 8) appear to exhibit significant asymmetry, with a skew towards smaller sample sizes and the left side of the funnel plot (Figure S1). The trim-and-fill method imputes missing studies on the right, suggesting underreporting of non-significant results. The Egger’s regression test identified a significant asymmetric funnel plot (intercept = −3.86 (95% CI: −6.59 to −1.14); p = 0.015), suggesting potential publication bias.
The four included studies that reported changes in hearing disability outcomes with non-implanted solutions had an asymmetric distribution on the funnel plot, with the study by Ryu et al.Reference Ryu, Moon, Byun, Jin, Park and Jang41 positioned far to the left of the plot, suggesting potential publication bias (Figure S2). However, Egger’s test did not identify significant asymmetry [intercept = −3.32 (95% CI: −16.02 to −0.32); p = 0.08].
The four studies that compared changes in hearing disability outcomes with different non-implanted surgical hearing solutions (CROS versus BCD headband) had a symmetrical distribution on the funnel plot. The lack of significant publication bias was confirmed by Egger’s regression test [intercept = 0.26 (95% CI: −7.98 to 8.52); p = 0.90] (Figure S3).
Changes in generic quality of life
The three studies that reported generic QoL did not show any asymmetry in their funnel plots. Egger’s regression test confirmed that there was no significant asymmetry for the three studies [intercept = −3.83 (95% CI: −57.94 to 50.27); p = 0.53] (Figure S4).
Reliability of Egger’s test assessments
Egger test results for all four analyses presented here should be interpreted with caution, as Egger’s regression is unreliable when fewer than 10 studies are included.Reference Higgins, Thomas, Chandler, Cumpston, Li and Page29, Reference Sterne and Egger53 While the Egger test failed to identify statistically significant publication bias with the very small study samples analysed in Figures S2–S4, publication bias became significant when the number of studies reached eight (Figure S1). Thus, it is safest to conclude that publication bias exists in this field and any summary conclusions drawn should be interpreted cautiously.
Discussion
Hearing disability
It is well known that patients’ perceptions of hearing disability change do not match audiometric measures of hearing change in response to interventions such as ossiculoplasty and stapedectomy.Reference Smyth and Patterson54 Therefore, there is a growing recognition of the importance of assessing PROMs to discern the effect of hearing-related interventions. Despite the availability of several robust hearing disability hearing-related and generic QoL PROMs, the literature remains deficient in studies that use these tools to determine which of the multiple SSD treatment modalities patients report as most effective.
Our meta-analyses identified that implanted surgical hearing solutions, including CIs and implanted BCDs, yielded greater improvements in hearing disability than non-implanted surgical hearing solutions, that is BCDs on a headband and CROS hearing aids (Figure 2). Subgroup analyses did not reveal a difference in hearing disability improvement with CROS hearing aids compared with no intervention (Figure 3) or non-implanted BCDs (Figure 4). Two studiesReference García, Urquijo, Puerta, Mosquera, Hernández and Aparicio35, Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 included in the meta-analysis comparing implanted and non-implanted hearing solutions are at high risk of bias. Each of these studies on their own did not detect a difference between implanted and non-implanted hearing solutions and together the findings accounted for almost 29% of the weight of the conclusion in our meta-analysis. Despite this, implanted hearing solutions were found to deliver a 14.7% greater improvement in APHAB scores than non-implanted solutions. A difference that exceeds the 10% APHAB MCID. It is reasonable to infer that implanted solutions have an even greater impact than non-implanted solutions than was found here. There was no high risk of bias study in the meta-analyses comparing the effect of CROS hearing aids with either no treatment (Figure 3) or with BCDs on a headband (Figure 4). Therefore, the finding of no discernible effect of non-implanted hearing solutions on PROMs of hearing disability is reliable.
To our knowledge, this systematic review and meta-analysis is the first comparing implanted and non-implanted hearing solutions effects on hearing disability and QoL PROMs. Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 systematic review revealed mixed hearing disability outcomes with BCD and CROS, based on five studies, one of whichReference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32 is included in the current review. However, Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 did not conduct a pooled analysis due to expressed concerns about heterogeneity and did not provide a comparative synthesis of outcomes using implanted and non-implanted solutions. While the meta-analysis by Kitterick et al.Reference Kitterick, Lucas and Smith13 assessed pooled post-intervention individual domain scores across various PROMs from 23 studies, the original article describing the development of the APHAB by Cox and AlexanderReference Cox and Alexander45 underscored the limitation of using individual APHAB domain scores to differentiate between hearing aids.
We identified only one study that had conducted a head-to-head comparison of implanted BCD and CI,Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 and therefore this was insufficient for a subgroup meta-analysis comparing these two modalities. That particular study was an RCT by Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 that reported only APHAB domain scores but not global APHAB scores. Other earlier systematic reviews and meta-analyses assessing the impact of implanted interventions were limited to evaluating a single intervention. For instance, the systematic reviews by Cabral Junior et al.Reference Cabral Junior, Pinna, Alves, Malerbi and Bento55 and Oh et al.20 exclusively assessed CI use in SSD patients. Therefore, it was not possible to determine which, if any, of these implanted solutions resulted in superior patient-reported hearing disability improvement.
In summary, the meta-analysis of global PROMs for hearing disability change in the current study provides comprehensive evidence supporting the advantage of implanted over non-implanted hearing interventions. Our findings support the consideration of implanted BCDs and CIs, in SSD patients, including those who had experienced minimal improvement in hearing disability with non-implanted treatments.
Health-related quality of life utility
Three studies in our review reported on HRQoL utility using the HUI3.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 Pooled data from these studies demonstrated a statistically significant MD of 0.21 (p = 0.0001) in generic HRQoL utility in favour of CI compared with no hearing solution (Figure 5). This is a clinically relevant improvement as it exceeds the HUI3 MCID of 0.03. The study by Arndt et al.Reference Arndt, Aschendorff, Laszig, Beck, Schild and Kroeger32 utilised the HUI3 to study the HRQoL utility effect of CROS hearing aids, BCDs worn on a headband and CIs. Ullah et al.Reference Ullah, Cevallos, Shen, Carver, Dunham and Marsiglia42 and Ramos Macias et al.Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38 administered the HUI3 before and after CI surgical procedures but did not study any other hearing solutions. The risk of bias in these three studies was judged to be moderate (Table S2). Therefore, as no high risk of bias studies were included in this meta-analysis and all three studies consistently identified a positive effect of CI, it is reasonable to conclude that SSD patients who had undergone CI derived improved HRQoL utility.
The NCIQ was used in one study in this review to assess HRQoL utility.Reference Plath, Marienfeld, Sand, van de Weyer, Praetorius and Plinkert40 However, since the NCIQ is a CI-specific instrument that does not allow comparison with other treatments, this study was excluded from meta-analysis. Two studies assessed HRQoL utility using the GBI scale.Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 Peters et al.Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 reported only GBI domain-specific subscale scores rather than overall GBI scores that excluded their results from the meta-analysis described in this review. Leterme et al.Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36 assessed overall GBI scores to compare CROS aids and implanted BCDs. A subgroup meta-analysis to compare the effect of CI and implanted BCDs on HRQoL utility was not possible, as no study reported relevant HRQoL utility data. Thus, a robust comparison of the effects of CIs and implanted BCDs on HRQoL utility awaits the reporting of future trials.
We identified only one previous meta-analysis that had evaluated post-interventional changes in QoL among patients with SSD using CIs.Reference Oh, Mavrommatis, Fan, DiRisio, Villavisanis and Berson18 Although the study showed improved QoL after CI surgical procedure, there were significant methodological differences compared with the current study. Oh et al.Reference Oh, Mavrommatis, Fan, DiRisio, Villavisanis and Berson18 exclusively assessed CI implantation and combined data from seven studies that reported measures of hearing disability (SSQ and NCIQ) with that of three studies that reported generic HRQoL utility measures (SF-36 and HUI3) to derive a standardised MD of ‘QoL’ of 0.61. Importantly the lower end of the 95 per cent confidence interval of the standardised difference in two of the three studies in their review that reported accepted measures of HRQoL utility was negativeReference Arts, George, Janssen, Griessner, Zierhofer and Stokroos56, Reference Häußler, Knopke, Dudka, Gräbel, Ketterer and Battmer57 and the one that crossed over the null point of the Forest plot had a lower 95 per cent confidence of only 0.01Reference Ketterer, Knopke, Häußler, Hildenbrand, Becker and Gräbel58. The ‘QoL’ effect size estimate reported in Oh et al.’s review was positively impacted by the SSQ-based studies that together demonstrated the largest positive standardised subgroup MD of 0.93. In contrast, the size of the effect of CIs on SSD participants’ HRQoL utility determined in the current review, based on three studies that all utilised the widely accepted HUI3 HRQoL utility tool, was 0.2. This result is within the effect size range of 0.17 (95 per cent CI: 0.11–0.23) identified after unilateral CIs in adults with bilateral severe to profound hearing loss. It is reasonable to accept that CIs achieved a statistically significant small and clinically important improvement in the HRQoL utility of individuals with SSD as identified here.
Adverse events and safety
Adverse-event data were reported inconsistently across studies. The studies that explicitly reported adverse events,Reference Leterme, Bernardeschi, Bensemman, Coudert, Portal and Ferrary36, Reference Ramos Macías, Falcón-González, Manrique Rodríguez, Morera Pérez, García-Ibáñez and Cenjor Español38, Reference Yang, Wang, Huang, Chai, Jia and Wu43, Reference Peters, van Heteren, Wendrich, van Zanten, Grolman and Stokroos44 when summarised, yielded an adverse-event incidence of 20.3 per cent (n = 11) among 54 BCD-implanted patients, 0 per cent in 26 CI patients and 19 per cent (n = 8) among 42 CROS patients. It is unlikely that all 281 patients in the other studies that did not report adverse events were adverse-event free. While the CI adverse-event rate here was 0%, this was unrealistic when discomfort or pain accounted for all the CROS adverse events reported. Future studies should consistently report the adverse effects of different hearing solutions. This intervention will allow patients’ concerns related to implanted solutions, specifically adverse effects, which remain major barriers to CIs and implanted BCDs, to be better addressed.
Limitations
Despite the promising findings, this review has limitations. The paucity of prospective controlled trials and evidence of publication bias affected the observed findings. Additionally, variation in follow-up duration and small sample sizes reduced the strength of the findings. The absence of blinding in the included studies is inevitable in surgical studies, although it results in higher performance and detection bias scores. In some studies, it was unclear whether the unaided condition was evaluated de novo or after prior use of a hearing solution. Previous use of hearing devices may have influenced patients’ perceptions of their unaided hearing disability, potentially introducing bias into PROM assessments. Future studies should account for prior hearing aid use when analysing treatment effects. Inter-study heterogeneity is a further limitation. However, adopting a random-effects model for meta-analysis helped to mitigate this variability.
Suggestions for future research
Our review highlights a pressing need for more prospective studies that utilise generic PROMs to assess changes in hearing disability and HRQoL utility and to directly compare the various hearing solutions for SSD. Further research should also investigate adverse events and the long-term effectiveness of these hearing interventions to better determine the optimal solution for individuals with profound unilateral hearing loss.
Summary
What is already known on the subject
• Unilateral sensorineural hearing loss or single-sided deafness (SSD) is a severe or profound hearing loss in one ear with normal hearing in the other.
• Several treatment options are available for SSD, including contralateral routing of signal hearing aids, bone-conduction devices headband-mounted or implanted and cochlear implants. These can be broadly categorised into implanted and non-implanted hearing solutions.
• Previous systematic reviews have mostly assessed individual interventions instead of direct comparisons, or have focused solely on audiological hearing outcomes.
• Comparative evidence of treatment efficacy using patient-reported outcome measures remains limited. It is still unclear whether implanted interventions provide greater patient-reported hearing disability or health-related quality of life utility improvement compared with non-implanted ones.
What this paper adds to our understanding
• This systematic review compares implanted and non-implanted solutions for SSD using PROMs and demonstrates that SSD patients report greater improvement in hearing disability with implanted options.
• There was no clinically significant difference in the effect of contralateral routing of signals hearing aids and headband-worn bone-conduction devices on SSD patients’ reported hearing disability.
• SSD patients report small but clinically important health-related quality of life utility improvement after cochlear implantation.
Conclusion
This review synthesises current evidence on the impact of hearing solutions for patients with SSD. It identifies significant improvements in hearing disability with implanted solutions, including CIs and BCDs. Non-implanted options, namely CROS aids and non-implanted BCDs, did not achieve a clinically significant improvement in hearing disability over no treatment. CIs resulted in a significant and clinically relevant albeit small improvement in HRQoL utility over that achieved with no treatment. There was insufficient available data to robustly compare the effect of CIs with BCDs and CROS hearing solutions on HRQoL utility of individuals with SSD. Future prospective comparative studies using generic PROMs to assess changes in HRQoL utility after implanted BCD and CI treatment for SSD are required.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0022215126105891.
Authors’ contributions
ML, Arshbir Singh Aulakh (ASA) and DAN conceptualised the study. ML drafted the protocol. ML, AA and ASA undertook database searches. ML, AA, ASA, AM and DAN participated in article screening, data extraction, analyses and manuscript drafting. RA participated in data extraction and analyses and reviewed the final manuscript.
Funding statement
No funding was received.
Competing interests
There are no conflicts of interest to disclose.
Open access
