Prospective longitudinal observational study at an academic medical centre of lifestyle and cognition in older adults with a cochlear implant or hearing aid: study protocol

Abstract

Introduction

Dementia is a major global health problem with increasing prevalence. Hearing loss has been identified as the most modifiable risk factor for dementia. The Age-Related Cognition and Hearing (ARCH) study is a 3-year prospective, controlled, observational comparative cohort study comparing cochlear implants (Implants) and hearing aids (HAs) for reducing cognitive decline associated with age-related hearing loss (ARHL), based on patient-reported real-world outcomes of auditory function, cognitive performance, listening environment, social interaction and psychosocial well-being. Upon its completion in 2029, the ARCH study is expected to yield significant evidence regarding the comparative effects of two primary hearing interventions—Implants and HAs, to delay and ameliorate cognitive decline associated with ARHL.

Methods and analysis

210 older adults are divided into six study subgroups (N=35) with: (1) moderate to profound hearing loss or age-typical normal hearing, (2) use of Implants or HAs and (3) mild cognitive impairment (MCI) or normal cognition. Listeners in the HA groups have hearing loss that is consistent with Implant candidacy and qualification through Centers for Medicare & Medicaid Services in the USA. The primary study outcome is a 3-year change in real-time patient-reported outcomes collected while participants are in their natural listening environments using ecological momentary assessment (EMA) methodologies. Secondary outcomes include lab-based audiometric and neuropsychological testing, and patient-reported outcomes of social isolation, loneliness, depression, anxiety and quality of life.

Cross-sectional analyses will use factor analysis to reduce EMA items into domains, followed by regression and mixed-effects models to test group differences and identify specific EMA items driving those effects. Machine-learning approaches will complement these models by predicting outcomes, identifying key variables and uncovering data-driven patterns. Longitudinal mixed-effects models will assess how EMA factor scores and cognition change over time and whether real-world EMA experiences mediate cognitive trajectories. Additional analyses will compare real-time EMA responses with retrospective patient-reported outcome measures and laboratory-based cognitive and auditory assessments, with sample size adjusted for up to 10% attrition.

Ethics and dissemination

All study procedures follow institutional review board requirements and the Declaration of Helsinki at the University of Iowa (IRB# 202403385), with informed consent processes tailored to ensure understanding among participants with MCI. Study findings will be disseminated through a multi-tiered strategy aimed at maximising scientific, clinical and public health impact. Peer-reviewed manuscripts will be submitted to leading journals in audiology, geriatrics, cognitive ageing and public health, with interim and final results presented at national and international conferences and professional society meetings.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The use of real-world ecological momentary assessment outcomes will generate meaningful evidence for personalised recommendations to minimise age-related hearing loss (ARHL)-related cognitive decline.

• Lab-based audiometric tests, comprehensive neuropsychological testing and patient-reported outcomes of social isolation, loneliness, depression, anxiety and quality of life will enhance understanding of the relationship between hearing loss and cognitive decline.

• The study dataset will be leveraged for precision medicine, clinical decision-support and data sharing, with predictive models of outcomes (using cochlear implants (Implants) or hearing aids (HAs)) for individuals and patient subgroups.

• Randomisation and blinding are not feasible as it would be unethical to impose a more invasive implant procedure on patients choosing HAs, and the device type is apparent to study staff and participants.

• Lack of randomisation raises concern that the choice of Implant or HA may be related to cognition.

Introduction

Age-related hearing loss (ARHL) is a result of cumulative, gradual effects of ageing on the auditory system and is present in two-thirds of adults older than 70 years.^{1 2} Hearing aids (HA) are the most common intervention for listeners with mild-to-moderately-severe hearing loss (HL), despite not being covered by Centers for Medicare & Medicaid Services (CMS) in the USA and most private insurance plans. HA adherence is a problem, with less adherence in those with dementia and lower socioeconomic status.³,⁵ In contrast, cochlear implants (Implants) are indicated for those with moderate-to-profound sensorineural HL who receive limited benefit from HAs and are covered by CMS and the most private insurance plans.

Many HA users receive insufficient benefit due to HL severity or aetiology. HAs fail to address the damaged cochlea, while implants directly stimulate the auditory nerve, improving audibility in moderate-to-profound HL. Fewer than 10% of implant-qualifying HL patients get Implants.^{6 7} Implant uptake is limited due to limited understanding of recent Implant candidacy criteria, among healthcare providers who have limited experience with Implant patients (ie, HA audiologists, primary care providers), as well as the requirement for surgery, the complexity of care delivery and an insufficient number of implant clinics.^{6 7}

Older adults are at increased risk for cognitive decline, and growing evidence links HL with cognitive impairment.^{2 5 8 9} The 2017 Lancet Commission for Dementia Prevention, Intervention, and Care meta-analysis reported HL has the greatest relative dementia risk (1.9) of nine modifiable risk variables.⁹ The 2020 Lancet Commission reported that eliminating HL risk reduced dementia prevalence by 8% among 12 modifiable risk variables.⁵ In a prospective trial of 639 adults with normal cognition, 58 developed dementia over 12 years (64% Alzheimer’s). Baseline HL severity highly predicted dementia risk (HR 1.9–4.9).¹⁰ A longitudinal ageing study (N=1515) showed baseline HL increased the probability of mild cognitive impairment (MCI) or dementia (HR 2.3).¹¹

The results of studies of HA effectiveness to prevent cognitive decline are mixed.¹²,¹⁹ Longitudinal studies show HAs alleviated ARHL-related cognitive decline, dementia and cognitive dysfunction.¹²,¹⁹ In a Danish population-based study (N=573 088) with 23 023 dementia cases over 8.6 (SD 4.3) years,²⁰ HL increased dementia risk (HR 1.07), severe HL showed greater risk (HR 1.13–1.20) and dementia risk was greater with HL without HAs (HR 1.20). In contrast, an 11-year observational study demonstrated no difference in cognitive deterioration between HAs and non-HAs.¹⁵ The ACHIEVE study (Aging and Cognitive Health Evaluation in Elders) examined whether HAs reduced cognitive decline in seniors with untreated HL and no cognitive impairment.^{21 22} When HAs were compared with a health education control, 3-year cognitive change was not statistically different between the groups. A prespecified sensitivity analysis showed that HAs may reduce cognitive decline in older adults at increased risk of cognitive decline, with lower baseline cognitive scores and greater cardiovascular risk factors.²³ ARHL is also linked to frailty, falls, social isolation, late-onset depression and functional disability^{24 25} and an ACHIEVE secondary analysis found that HAs reduced loneliness and social isolation after 3 years.²⁶

Few studies have examined Implant effects on cognition. Small longitudinal studies demonstrate Implants prevent dementia and improve language, memory, attention and communication.²⁷ A recent systematic review and meta-analysis described eight studies (N=126 903) of the effects of HAs and Implants on cognitive decline and dementia.²⁸ In those using Implants or HAs, long-term cognitive decline risk dropped 19% and cognition improved 3%. Subgroup analysis by hearing restorative device showed these findings remained significant with Implants but not HAs. HAs and Implants may improve ARHL-related cognitive impairment by reducing listening effort and stimulating auditory neural circuitry.²⁹

The Age-Related Cognition and Hearing (ARCH) study is a 3-year prospective observational comparative cohort study comparing Implants and HAs to reduce ARHL-related cognitive decline, based on meaningful real-time outcomes of: (1) listening environment context (eg, indoor/outdoor, one-on-one, group interaction); (2) auditory function (eg, listening effort, speech perception); (3) cognitive function (eg, attention, processing speed, multi-tasking) and (4) hearing-related psychosocial well-being (eg, social isolation, loneliness, depression).

Study objectives

The ARCH trial compares the effects of Implants and HAs on cognitive impairment using real-world outcomes. The specific aims are to: (1) compare the cross-sectional associations of Implants versus HAs on real-world outcomes in ARHL with normal cognition, (2) compare the longitudinal effects of Implants versus HAs with and without baseline cognitive impairment on cognitive function and (3) deploy study data in our visual analytics platform for predictive modelling, clinical decision-support and data sharing. Implant use is expected to show greater improvement than HAs on: (1) cross-sectionally, ecological momentary assessment (EMA) outcomes of auditory function, cognitive performance, social interaction and psychological well-being and (2) longitudinally, cognitive performance and less cognitive decline as measured by EMA cognitive items and neuropsychological testing in those with baseline MCI.

Methods and analysis

Design and participants

ARCH is a non-randomised, controlled, single-centre, repeated-measure, prospective observational comparative cohort study of 210 older adults. There are six study subgroups (N=35; table 1) with: (1) moderate to profound HL or age-typical normal hearing, (2) use of implants or HAs and (c) MCI or normal cognition. Groups 1 and 3 are people who have an Implant in one or both ears. Groups 2 and 4 target people who are implant candidates (meeting audiological candidacy for implantation by CMS) with inadequate benefit from best fitted HAs who naturally chose to retain HAs. Groups 5 and 6 have normal age-typical hearing.³⁰ Each hearing group has either normal cognition (groups 1, 2 and 5) or MCI (groups 3, 4, and 6).

Table 1. Study groups.

Group	Description
1	Implant/NC: implant and normal cognition
2	HA/NC: HA retainers and normal cognition
3	Implant/MCI: implant and MCI
4	HA/MCI: HA retainers and MCI
5	NH/NC: NH and normal cognition (control group)
6	NH/MCI: NH and MCI (control group)

HA, hearing aid with inadequate benefit-meets Implant candidacy; Implant, cochlear implant; MCI, mild cognitive impairment; NC, normal cognition; NH, normal hearing.

All groups will be followed longitudinally for 3 years, with study assessments (table 2) performed at baseline and annually, for a total of four sessions. Each session consists of two visits, separated by 1 week (figure 1).

Table 2. Summary of study assessments, including measures, assessment time required, approach and administration site.

Domain and test measure by category	Time	Approach	Site
Demographics and clinical characteristics
Demographics (age, gender, race, geographic location, etc)	5	PROM	Off
Hearing history, medical comorbidities, medication use	5	PROM	Off
Cumulative Illness Rating Scale-Geriatric46	5–10	PROM	On
Auditory environment
Noise/Quiet, inside/outside, familiar/unfamiliar talker, visual cues, 1:1 vs multiple people, etc	5/day	EMA	Off
Auditory function
Audition: Audiogram	15	LAB	On
Speech perception: CNC words47 and AzBio Sentences in Noise48	30	LAB	On
Speech perception, localisation, sound quality: Speech, Spatial and Qualities of Hearing Scale-12 (SSQ-12)49 50	5	PROM	Off
Listening strategies: Communication Profile for the Hearing Impaired (CPHI)51	10	PROM	Off
Speech perception, listening effort, efficiency, etc	5/day	EMA	Off
Cognitive function
Premorbid function: Wide Range Achievement Test 5 Reading (WRAT5)	5	LAB	On
Language: RBANS-HI52,54 Picture Naming, Semantic Fluency; WRAT Sentence Comprehension; Controlled Oral Word Association (COWAT)55 (verbal fluency)	15	LAB	On
Memory: Brief Visual Memory Test-Revised (BVMT-R)56 (immediate and delayed); RBANS-HI52,54 List Learning (immediate), Story Memory (immediate), List Recall (delayed), Story Recall (delayed), Figure Recall (delayed)	15	LAB	On
Executive function: WAIS-IV Matrix Reasoning (visual reasoning),57 Trails B58 (set-shifting, cognitive flexibility), Stroop (inhibition)	25	LAB	On
Attention/working memory: RBANS-HI52,54 Digit Span, Coding	10	LAB	On
Processing speed: Trails A43 (visual scanning), Stroop-Word/Color Naming59 (reading/colour naming)	5	LAB	On
Visuospatial function: RBANS-HI52,54 Figure Copy, Line Orientation	10	LAB	On
Dementia screen: Montreal Cognitive Assessment-Hearing Adapted (MoCA-HI)60	10–15	LAB	On
Participant perceived daily function: PROMIS Cognitive Function-Abilities61	5	PROM	On
Participant perceived functional abilities: Form B7 NACC Functional Assessment Scale (FAS)	5	PROM	On
Attention, processing speed, executive function, learning and memory, language skills	5/day	EMA	Off
Psychosocial Well-Being
Hearing-specific quality of life: Cochlear Implant Quality of Life (CIQOL)-Short62	10	PROM	Off
Social isolation: PROMIS-Social Isolation-Short Form	5	PROM	Off
Social integration: Social Network Inventory	10	PROM	Off
Loneliness: UCLA Loneliness Scale63 64	10	PROM	Off
Anxiety and depression: General Anxiety Disorder-7 (GAD-7); Patient Health Questionnaire-8 (PHQ-8)	10	PROM	Off
Social interaction: characteristics and number of social interactions	5/day	EMA	Off
Depression, anxiety, social isolation	5/day	EMA	Off

Test measures adapted for hearing impairment. Time is in minutes. EMA measures are daily for 1 week (1× per year); Approach: PROM; LAB; real-time EMA. Site: measured onsite (on) or off-site (off) in subject’s natural surroundings.

Bold indicates test domain per each study assessment category.

EMA, ecological momentary assessment; LAB, laboratory performance; NACC, National Alzheimer’s Coordinating Center; PROM, patient-reported outcome measure.

Figure 1. Description of ARCH study session. Each session includes two on-site visits (visit 1 and visit 2) and off-site real-time EMA surveys and PROMs. At study end, participants complete a total of four sessions (baseline and annually for 3 years). ARCH, Age-Related Cognition and Hearing Study; EMA, ecological momentary assessment; NACC-FAS, National Alzheimer’s Coordinating Center Functional Assessment Scale; PROMs, patient-reported outcome measures.

Recruitment and eligibility criteria

Participants are community-dwelling, older adults recruited from the University of Iowa Cochlear Implant Hearing Registry, Otolaryngology Clinic, Brain Boosters Cognitive Enhancement Programme, Seniors Together in Aging Research Registry, HA retailers, and retirement and senior centres. Participants with normal hearing are recruited from advertisements, email groups or Implant/HA participants’ families or friends. ARCH recruitment began in May 2025 with completion anticipated by July 2029.

Eligibility criteria are: (1) ≥65 years of age at enrolment, (2) adult onset HL, (3) English as primary language, (4) access to Android or iPhone mobile phone, (5) willingness to travel to University of Iowa, (6) functional independence (score ≤8 on National Alzheimer’s Coordinating Center Functional Assessment Scale (NACC-FAS)), (7) absence of psychiatric/behavioural issues that could affect study assessments and (8) use of HAs for at least 6 months or Implant(s) for ≥1 year and <10 years.

Groups 1 and 3 (Implant groups) have unilateral or bilateral implants. For those with unilateral implants, the non-implanted ear needs Hearing Number (HN)³¹ ≥50 dB HL and minimally at least one pure-tone air conduction threshold ≥70 dB HL. HN is pure-tone air conduction threshold average at 500, 1000, 2000, 4000 Hz. To align the HA groups, groups 2 and 4 meet CMS guidelines for coverage for cochlear implants. These participants need bilateral moderate-to-profound HL in both ears with one ear’s HN≥60 dB HL and contralateral ear’s HN≥50 dB HL, at least one pure-tone air conduction threshold ≥70 dB HL and word scores ≤60% in the poorer ear. Groups 4 and 5 need age-typical hearing in both ears.³⁰ Table 3 provides a summary of the inclusion criteria.

Table 3. Study inclusion criteria.

Group	English primary language	Smartphone access	Functional independence	Adult onset HL	Audiometry ear 1	Audiometry ear 2	Speech perception
1	Y	Y	Y	Y	Implant	Implant or HN≥50 with a threshold ≥70 dB HL
2	Y	Y	Y	Y	HN≥60 with a threshold ≥70 dB HL	HN≥50 with a threshold ≥70 dB HL	Ear 1: CNC Word score ≤60%
3	Y	Y	Y	Y	Implant	Implant or HN≥50 with a threshold ≥70 dB HL
4	Y	Y	Y	Y	HN≥60 with a threshold ≥70 dB HL	HN≥50 with a threshold ≥70 dB HL	Ear 1: CNC Word score ≤60%
5	Y	Y	Y	N/A	Female HN: ≤28; male HN: ≤33	Female HN: ≤28; male HN: ≤34	N/A
6	Y	Y	Y	N/A	Female HN: ≤28; male HN: ≤33	Female HN: ≤28; male HN: ≤34	N/A

Y indicates required. N/A indicates not applicable.

HL, hearing loss; HN, hearing number; SNR, signal-to-noise ratio.

This study is approved by the Universities of Iowa and Maryland Institutional Review Boards. Participants must provide informed consent as defined by the Evaluation to Sign Consent (ESC) to ensure understanding of research activities. Participants who fail the ESC are excluded. Participants are compensated with US$200 for session completion with a maximum of US$800 at completion of four sessions (figure 1). Participants who do not meet inclusion criteria at initial screening are compensated US$10.

Determination of MCI

MCI is defined as: (1) cognitive change reported by patient, informant or clinician, (2) objective evidence of impairment in ≥1 cognitive domain and (3) general preservation of function.

MCI is determined based on a standardised battery of cognitive tests encompassing multiple domains (table 2). MCI criteria are: (1) the score on ≥2 tasks in ≥1 cognitive domain is >1.5 SD below estimated premorbid intellectual function, as assessed by the Wide Range Achievement Test (WRAT-5) and (2) NACC-FAS shows independent function (score ≤8).

Study assessments

Study assessments are in four categories: (1) auditory environment; (2) auditory function; (3) cognitive function; (4) psychosocial well-being. There are three types of study outcomes data: (1) retrospective patient-reported outcome measures (PROMs): recollection of listening experiences, cognitive ability, psychological well-being, social engagement and physical function; (2) real-time PROMs: reporting from the participant’s natural environment (EMA surveys); and (3) laboratory performance: (a) audiometry and speech perception and (b) neuropsychological assessments (table 2).

Ecological momentary assessment

Our experience with EMA methodologies for hearing impairment includes publication of a comprehensive statistical guide to EMA data³² and documenting feasibility in Implant users, including those with MCI.³³ We also developed AudioSense+^{34 35} to facilitate EMA administration on a smartphone in the participant’s natural environment. For 1 week during each of four study sessions, participants complete ≥5 surveys per day, triggered by pinged notifications or self-initiated reports (figure 1). AudioSense+ adaptively delivers questions, with each subsequent question contingent on the previous response. Each survey requires 2–3 min. Users respond to questions to report their current or very recent experience (online supplemental table).

Neuropsychological testing

Neuropsychological testing is performed by a trained technician under the supervision of two board-certified neuropsychologists. To limit practice effects, participants receive alternate test forms annually, where possible. Cognitive assessments (table 2) include: (1) objective assessment of cognitive domains, including traditional tests with standardised verbal instructions and supplemental auditory stimuli with visual cues adapted for hearing impairment and (2) subjective assessment adapted from the PROMIS Cognitive Function-Abilities Short Form, administered with real-time EMA.

The Montreal Cognitive Assessment-Hearing Impairment (MoCA-HI) characterises global cognitive status for comparison to prior literature. In MoCA-HI, the examiner shows flashcards with written instructions and instructs the participant to read the instructions out loud before completion. The WRAT-5 Reading Standard Score is used to estimate premorbid intellectual function. The Repeatable Battery for the Assessment of Neuropsychological Status for individuals with hearing impairment was modified by providing PowerPoint prompts to enhance comprehension (RBANS-H). In our further modification, a neuropsychologist provides verbal instructions with closed-captioning in PowerPoint (RBANS-HI).

Primary outcome

The primary outcome measure is EMA data, assessing real-time, real-world patient-reported outcomes. Diverse EMA outcomes characterise rich ecological profiles of auditory function, cognitive performance, listening environment, social interaction and psychosocial well-being, which will be used to study how interactions between hearing and lifestyle impact cognition and how interactions between hearing and cognition impact lifestyle.

Secondary outcomes

Secondary outcomes include cognitive domain scores (memory, executive function, language, attention/working memory, processing speed, visuospatial function), audiometry, speech perception and retrospective PROMs (social isolation, loneliness, anxiety, depression and hearing-specific quality of life).

Deploying data in our visual analytics platform

Study data will be deployed in our user-friendly visual analytics platform, POD-Vis (Probing Outcomes Data with Visual Analytics; figure 2).³⁶ POD-Vis performs traditional data analyses, hypothesis-generating visualisations and predictive modelling for clinical decision support. POD-Vis is a web application built using modern JavaScript and Python libraries (VueJS, NumPy and Pandas). POD-Vis functions include: (1) selecting specific predictors and outcome variables to create data queries; (2) filtering study populations to create customised cohorts; (3) using data visualisations to understand distributions and compare effects on different outcome variables; (4) performing instantaneous analyses for iterative data exploration and (5) predicting outcomes for individual patients based on pre-built models. Machine learning (ML) analyses will augment traditional analyses to leverage the value of the complex real-world and conventional data collected in this study and will be democratised by embedding them within the easy-to-use POD-Vis interface.

Figure 2. POD-Vis. A tool for visualising clinical variables and building queries based on predictor and outcome variables. It includes five main screens: (1) home page, (2) create new study dataset, (3) use a previously saved dataset, (4) study group selector—to filter predictor and outcome variables to create comparative study groups (eg, female/male, more disability/less disability) and (5) data analytics—to analyse the selected study groups with box plots, longitudinal graphs and regression analyses. The figure is a summary view of selected analyses showing that males and older subjects have worse disease severity and cognitive function over time in a Parkinson’s disease cohort. POD-Vis, Probing Outcomes Data with Visual Analytics.

Covariates

Participant sex, age at enrolment, race, ethnicity, education, premorbid intellectual function, HL aetiology, HL duration/onset, amplification/implant history and device datalogging (if applicable) are documented. Medical comorbidities and life changes (eg, change of home, retirement, death of family member) will be documented. These variables will be adjusted for analyses and used as predictor variables in predictive models.

Statistical analysis

Aim 1: cross-sectional analyses

We want to know if group differences, unrelated to cognitive function, explain EMA variance. EMA survey responses are the key outcome metrics in these cross-sectional analyses. Since EMA items are correlated with each other to various degrees, EMA questions will be combined into a smaller number of EMA domains (online supplemental table). An exploratory factor analysis of Psychosocial Wellbeing and Environment items from prior EMA research showed seven factors explained 52% of EMA variance (unpublished data). Factor analysis of the study data, including EMA Cognitive items, will create participant factor scores. Each factor score will be the dependent variable in multiple linear regression models. Group will be the key independent variable in each model because the goal is to compare study groups. Sex is a biological variable for differentiation. We will adjust for medical comorbidities and cognitive function in the models to determine if group differences explain variation not explained by cognitive function.

When substantial group differences in an EMA factor are found, we will determine which EMA item is responsible. Thus, the regression model’s particular EMA item (eg, Auditory Function, Psychosocial, Cognition) becomes the outcome variable at this step. Group, cognitive function and sex fixed effects will be included in the model. A linear mixed model (LMM) with a random intercept for subject is used to account for multiple EMA item responses per participant. Pairwise comparisons between the six groups will be performed using a Tukey adjustment to adjust for multiple comparisons. Secondary analyses will use multiple regression models with cognitive function as the dependent variable. Group, with sex adjustments, is the key independent variable.

ML analyses will complement the statistical analyses. These analyses will largely focus on supervised learning for predicting outcomes such as EMA response. ML modelling will enable: (1) prediction of Implant versus HA outcomes for individuals or subgroups, (b) quantification of variable importance for selected outcomes and (c) unsupervised learning to identify common trajectories and patterns.

Aim 2: longitudinal analyses

Analysis #1. LMMs will evaluate how EMA factor scores change over time. Within-subject fixed effects will be time, and between-subject fixed effects will be cognitive function and hearing/device (six groups). A random intercept for participant and a random slope for time are expected; however, other random effects and correlation matrices will be evaluated to find the best model fit. We will use hierarchical evaluation to evaluate if group change in EMA factor scores over time differs by group. Models will account for medical comorbidities, education and sex. Group comparisons will be performed at each time point using model estimates to determine when group moderates change in EMA.

Subsequent analysis will determine which EMA questions within the identified factor have group differences based on EMA factor scores. Similar to aim 1, LMMs will be created using raw EMA data, accounting for multiple responses per item at each time point and integrating the same fixed factors.

Analysis #2. Hierarchical LMMs will study how real-world factors affect cognition and ARHL. Cognition will be the outcome variable, with time, group membership, and a random intercept for participants as fixed effects. Hierarchically using EMA, we examine how real-world factors mediate cognitive changes and explain MCI variance.

Analysis #3. PROMs will be compared to lab data. The agreement between retrospective PROMs and real-time EMA questions (dependent variable) to laboratory-based assessments (independent variable) will be examined, comparing EMA and retrospective PROMs with (1) in-lab performance on objective cognitive tests and (2) laboratory auditory function tests.

This study collects longitudinal data and some participants in each group may drop-out. While the coordinator and investigator will encourage participation in final data collection, dropouts will result in missing data. Based on prior research, the attrition rate is expected to be below 5%; we account for 10% to ensure a sufficient sample size.

Sample size

Aim 1: the primary question is a two-group pairwise comparison between independent groups. We evaluate power using a two-independent samples t-test, although the regression analysis that controls for additional variation should have more power than the t-test. The outcome variable is the factor score from one of the EMA factors. Factor scores have an SD of 1 and we wish to detect half a SD effect between the two groups. We have 80% power to detect an effect of Cohen’s d=0.71 with 32 participants per group at the 5% significance level. Our prior work shows that the same sample size will yield approximately 88% power for the LMM analysis approach based on 1000 simulated datasets.³²

Aim 2: we anticipate that EMA responses in the Implant/MCI group, HA/MCI group and NH/MCI group will show changes over time at different rates. We will assess whether a difference of d=0.5 exists at the final follow-up period, using N=32 for each of the six groups. Our analysis indicates approximately 95% power with 1000 simulated datasets to detect a within-group change of d=0.5, using sample size of N=32 at a 5% significance level, and an approximate 84% power to detect a between-group change of d=0.7 with a sample size of N=32 at a 5% significance level. Recruitment target for each group has been increased to N=35 to guarantee an adequate sample size.

Data management and sharing plan

ARCH study data will be preserved and shared to allow other researchers to reproduce the data and obtain additional findings to advance the field. A unique feature of this study is the use of the visual analytics tool, POD-Vis, to enable large-scale data-sharing, including predictive models showing the relative effectiveness of HAs or Implants on cognitive function based on individual profiles (precision medicine). The research community will have access to ARCH data uploaded in POD-Vis for clinical care and research. Data sharing will include: (1) subscale and global scores of PROMs, (2) raw and standardised scoring measurements from neuropsychological tests and (3) deidentified demographic data. Audiological data (unaided thresholds) will be provided in their original form.

Data preservation and sharing will be facilitated in the well-established repository, the Inter-University Consortium for Political and Social Research.

Discussion

Dementia is a major global health problem with increasing prevalence and societal costs associated with the rising elderly population.³⁷,⁴⁰ The Lancet Commission report identifies HL as the most modifiable risk factor for dementia.⁵ In response, ARCH responds to the necessity for research on hearing interventions, including HAs and Implants, to reduce risk and address cognitive impairment associated with ARHL. ARCH is a large prospective longitudinal comparative trial investigating AHRL-related cognitive impairment in participants using either HAs or Implants.

The randomised controlled trial, ACHIEVE, examined the efficacy of HAs in reducing cognitive decline among participants with age-related mild to moderate HL and without significant cognitive impairment. The study did not demonstrate a significant difference between HAs and a health education control. However, a prespecified sensitivity analysis indicated that HAs may reduce cognitive decline in the subgroup characterised by greater age and lower baseline cognitive scores.²¹,²³ In contrast, ARCH targets a population with moderate to profound baseline HL who currently have an implant or who are a candidate for an implant. Participant groups are divided into those with and without MCI. These factors (more severe HL⁴¹ and MCI⁴²,⁴⁵) are associated with a greater risk of more rapid cognitive decline over the 3-year study period. Older adults with age-typical hearing, both with and without MCI, function as control subjects. ARCH’s primary outcome measures are the EMA surveys, assessing real-time cognitive and auditory outcomes, along with the traditional cognitive and audiology measures.

On its completion in 2029, ARCH is expected to yield significant evidence regarding the comparative effects of two primary hearing interventions, Implants and HAs, to delay and ameliorate cognitive decline associated with ARHL.

Ethics and dissemination

All study procedures adhere to ethical standards set by institutional review boards and the Declaration of Helsinki through the University of Iowa Institutional Review Board (IRB# 202403385). Participants provide informed consent prior to enrolment, with additional procedures in place to ensure decisional capacity and comprehension among individuals with MCI. Data confidentiality is ensured through secure, encrypted data capture, privacy-preserving EMA protocols and de-identification of all shared datasets. Continuous risk–benefit monitoring is conducted throughout the 3-year study, including oversight of withdrawal procedures, adverse psychological responses and device-related concerns. Ethical safeguards ensure participant autonomy, protection of vulnerable adults and responsible stewardship of sensitive real-world hearing and cognitive data. Patients and members of the public were not involved in the design, conduct, reporting or dissemination plans of this research.

Study findings will be disseminated through a multi-tiered strategy aimed at maximising scientific, clinical and public health impact. Peer-reviewed manuscripts will be submitted to leading journals in audiology, geriatrics, cognitive ageing and public health, with interim and final results presented at international conferences and professional society meetings. Summaries tailored for clinicians, industry partners and policymakers will be developed to support translation of evidence into hearing-health interventions and dementia-risk reduction strategies.