Associations among hearing loss, cognition, and mobility

Multitasking in our everyday lives involves successfully integrating and coordinating information from our sensory, cognitive, and motor systems, such as when listening, looking, and thinking while staying balanced. Age-related changes across these systems are common [1–6]. In older adults, clinically significant impairments in one domain, such as hearing, are also often associated with declines or impairments in the other domains, such as cognition and mobility [7]. For example, hearing loss is associated with cognitive decline [1,7–14] and dementia [7,8,10,11,14] with greater hearing loss severity linked to a greater risk of dementia [10,15]. Hearing loss has also been identified as a top potentially modifiable mid-life risk factor for dementia [10]. Separately, other research has demonstrated that hearing loss is associated postural instability [16–20, although see 21 for a critical review], slower walking pace [22], elevated vestibular perceptual thresholds [23], poorer physical functioning [24], and a higher likelihood of falls [22,25–28]. In fact, there is evidence that older adults with hearing loss have a threefold higher risk of falling compared to older adults with normal hearing [28], with a recent study reporting a 17% increased longitudinal risk and 51% increased cross-sectional risk of falls in older adults with hearing loss compared to older adults with normal hearing [29]. These associations among hearing loss, cognitive decline, and mobility problems may reflect interactive mechanisms. Even among older adults with normal hearing and cognitive decline, poorer executive functioning (EF), in particular, is known to be associated with poorer mobility-related outcomes [30–35]. These findings suggest that cognitive mechanisms may also partially underlie the association between hearing loss and poor mobility given that high cognitive load during effortful listening, may a) limit the remaining cognitive resources available to support other concurrently performed tasks, such as when staying balanced [26,36,37], and/or, b) may be associated with changes in brain structure and functioning over time that negatively affect cognitive abilities [37,38]. According to the Framework for Understanding Effortful Listening model, effortful listening may draw on domain-general cognitive resources, particularly under degraded auditory conditions [39] and this could be detrimental to other important tasks that draw on similar resources, such as balancing [40,41].

Dual-task paradigms to evaluate the effects of the competing demands of hearing and cognition on balance and mobility

Dual-task paradigms are often used to evaluate how cognitive resources are distributed when managing hearing and competing mobility task demands such as balancing, or walking [see 42 and 43 for reviews]. In the context of understanding the interactions and the distribution of cognitive resources among hearing, cognition, and motor tasks, dual-task paradigms have introduced auditory and/or cognitive tasks that are performed during standing balance or walking tasks. Specifically, single-task auditory-cognitive performance, such as listening, counting, or remembering, and single-task mobility-related performance, such as standing or walking, are compared to dual-task performance; standing or walking while performing an auditory-cognitive task simultaneously. This approach allows for the determination of task prioritization and allocation of cognitive resources [32,35,41,44,45]. For example, in a dual-task study comparing older and younger adults with normal hearing, Carr et al. [46] used a listening-while-balancing task and found that, compared to younger adults, older adults exhibited poorer balance when listening demands were high on a repeating words in noise task [46]. Nieborowska and colleagues [47] implemented a listening-while-walking dual-task experiment and demonstrated worse listening performance on a multi-talker word identification task and greater postural prioritization in older compared to younger adults [47]. Further, Li et al. [48] used a memorizing-while-walking dual-task and showed that older adults experienced more dual-task costs on memory performance compared to walking performance, suggesting prioritization of walking [48]. Our previous study examining dual-task performance in older adults with and without hearing loss [49] implemented a listening-while-balancing task and found higher dual-task costs in conditions of lower task demands for older adults with hearing loss compared to older adults with normal hearing [49]. Bruce et al. [50] implemented a dual-task paradigm whereby younger and older adults with normal hearing and older adults with hearing loss completed a working memory task during standing balance perturbations [50]. Individuals with hearing loss experienced greater dual-task costs; specifically, worse cognitive performance in dual-task compared to single-task conditions in the presence of noise than younger adults and older adults with normal hearing [50]. Lau et al. [51] used a listening-while walking dual-task to test older adults with and without hearing loss and demonstrated worse performance on a multi-talker word identification task in the single- and dual-task conditions and greater stride time variability in older adults with hearing loss compared to those with normal hearing in the dual-task condition [51]. Taken together, the evidence that older adults, and particularly older adults with hearing loss, experience difficulty managing simultaneously auditory, cognitive, and balance-related tasks, indicates that an important next step is to explore whether targeted interventions could mitigate these challenges and improve their dual-task performance.

Hearing and cognitive interventions to support balance and mobility

Several strategies have been proposed to improve balance in older adults with hearing loss, with specific interventions addressing different hypothesized mechanisms. One such intervention is hearing aids. Hearing aids may amplify important sound signals from the environment that can support balance, and/or mitigate cognitive load by reducing listening demand, thereby preserving more cognitive resources that may be allocated to support balance. However, the literature is mixed with respect to the effectiveness of hearing aids on improving balance and mobility, with some studies demonstrating positive effects [52–57], and other studies showing no or minimal benefits [49,58–62]. As such, hearing aids may not be an adequate solution for improving balance and mitigating falls risk. Other multimodal interventions should be considered to support balance in older adults with hearing loss. An alternative intervention is to more directly address the cognitive load hypothesis through cognitive training to increase the availability of cognitive resources for other tasks. Understanding how the effects of sound amplification from hearing aids can be further augmented through cognitive training to improve balance has not been investigated previously.

Cognitive training refers to standardized activities involving structured practice on tasks targeting various cognitive abilities, such as working memory, EF, attention, processing speed, with the aim of improving overall cognitive performance [63–65]. The literature on the benefits of cognitive training interventions is mixed, with strong evidence showing improved performance on the specific tasks that are trained, weaker evidence for the transfer of training to similar tasks, and very limited evidence that training benefits broader cognitive abilities or everyday functioning [66; see 67 and 68 for reviews]. In the context of balance and mobility, training EFs may be more promising due to functional overlap with gait/balance. Specifically, training-related improvements in EF following training, particularly those involving attentional control and inhibitory processing, can lead to improvements in gait and balance performance in older adults [32], including single-support balance, sit-to-stand, dynamic posturography, and a 40-foot walk test [69]. Downey et al. [70] and Pothier et al. [71] used the EF training program developed by Lussier et al. [72] that specifically targets different aspects of EF (i.e., inhibition, task switching, divided attention, and working memory) and have shown improvements in cognitive and mobility task performance, including more accurate 2-back-while-walking performance [70] and faster spontaneous walking speed [71]. It has been proposed that EF training may improve the ability to engage in compensatory scaffolding; that is, the recruitment of additional neural resources to compensate for age-related structural and functional brain decline [32,73], such that the likelihood of transfer to untrained tasks increases when there is shared neural circuitry between the trained functions and the targeted outcomes [74]. Using EF training to increase the cognitive resources available for multitasking [75] could support better management of simultaneous task demands in older adults with and without hearing loss.

For older adults with hearing loss, increased availability of cognitive resources gained from EF training could reduce competition amongst cognitive and neural resources to better manage greater listening demands associated with hearing loss. Very few studies, however, have evaluated the benefits of EF training on balance and mobility-related outcomes in older adults with hearing loss specifically. Bruce et al. [76] implemented exercise (i.e., cycling) and cognitive training (i.e., computerized, visual discrimination dual-task) interventions simultaneously or sequentially in older adults with and without hearing loss. They found that both groups improved in single- and dual-task cognitive performance, especially during low task demands and sit-to-stand performance under dual-task conditions. However, they did not find improvements in balance on measures from computerized dynamic posturography, including double support, single-support, and visual sway referenced [76]. However, given the need to carefully allocate cognitive resources as a function of multiple demands, it is likely that any effects of EF training would be most pronounced during balance and mobility-related tasks that are complex and multisensory involving looking, listening, remembering, and moving at the same time. Further, these types of complex scenarios are reflective of everyday life, and it is important to understand whether EF training-related benefits generalize to these types of ecologically valid contexts [77,78]. Previous studies have strategically introduced these challenges using virtual reality and simulation technologies, which enable the study of complex, realistic situations under safe and controlled conditions [46,47,49,51]. However, no previous virtual reality-based studies have evaluated the effects of EF training on balance-related outcomes in older adults with hearing loss.

When implementing an EF training program as a strategy to mitigate age-related declines in cognitive processing, an additional important consideration is the point at which it is the optimal time to intervene, recognizing that training-related effectiveness is likely influenced by age, stage of cognitive changes/decline, and hearing loss trajectory. In fact, declines in domains of cognitive functions, such as working memory and speed of processing, can begin as early as mid-life [79–83]. Declines in hearing can also begin in mid-life, with nearly 20% of middle-aged adults between 45–65 years old having elevated high-frequency hearing thresholds, and 5% of middle-aged adults having mild hearing loss [84–86]. While very few studies have specifically investigated the associations between hearing and balance-related outcomes in middle-aged adults, a recent study by Wang et al. [87] found that poorer hearing abilities (defined using audiometric pure-tone average (PTA)) was associated with worse postural control in adults aged 40–69 years. Importantly, even with clinically normal audiometric hearing thresholds, supra-threshold declines in hearing [88], such as difficulties understanding speech in noise, often starts in middle age [89,90]. These supra-threshold declines could draw cognitive resources in more subtle ways, as it may result in more effortful listening, thereby introducing greater processing demands compared to when auditory processing is normal [90]. Therefore, comparing the effects of cognitive training on balance-related outcomes in middle-aged adults could, provide unique insights into the effects of EF interventions across the adult lifespan.

One cognitive aging framework that has been proposed to explain why cognitive training effects may vary across individuals is the compensation mechanism. This framework predicts that individuals with lower baseline sensory and/or cognitive performance can benefit most from cognitive training with poorer baseline performance predicting larger training gains [91] and will be used to formulate predictions for the current study.

Current research

EF training has the potential to improve cognitive processing, thereby improving balance-related outcomes. The most pronounced balance-related effects of EF training are likely to be observed during complex, ecologically valid contexts that mirror the demands of everyday life. These training-related benefits may start in middle age, increase with older age, and may be more pronounced in older adults with hearing loss, who may be at risk for declines in EF, balance problems and falls. It is also unknown whether EF training can further augment any benefits already provided to hearing aid users through amplification. Therefore, this randomized control study evaluated the effects of a 12-week EF training program on balance outcomes (COP AP). Age and hearing status were evaluated by testing middle-aged and older adults with normal hearing (MA and OA, respectively) and older adults with hearing loss who use hearing aids (OAHL). Pre- and post-training effects (T1 vs. T2) were evaluated using a dual-task paradigm (auditory-cognitive + balancing tasks) in an immersive, realistic, virtual reality environment. Secondary analyses also examined whether sensory, cognitive, and mobility functions, as evaluated by baseline standardized assessments, predicted training-related outcomes. Based on a compensation hypothesis [91] it would be predicted that individuals with lower baseline cognitive or hearing performance would benefit most from training [76,92].

Participants

The recommended sample size based on a G*Power analysis (power = 0.95, α = 0.05, and a small-to-moderate effect size) was 18 participants per group. To allow for approximately 10% attrition and to maintain a balanced design, we aimed to recruit 20 participants per group. We tested 19 MA (M_age = 52, SD_age = 5; 11 EF training, 8 controls; 14 F, 5 M), 23 OA (M_age = 70.87, SD_age = 3.83; 13 EF training, 10 controls; 13 F, 10 M), and 23 OAHL (M_age = 76.48, SD_age = 6.48; 14 EF training, 9 controls; 9 F, 14 M; see Fig 1 for consort diagram). Recruitment strategies included both advertisements posted across Toronto and digital outreach via social media. Participants were included in the MA group if they were 45–60 years old and had normal audiometric hearing thresholds, the OA group if they were 65 + years old and had normal audiometric hearing thresholds, and the OAHL group if they were 65 + years old, had bilateral hearing loss and had used hearing aids for a minimum of 6 months. While this ensured that participants were comfortable with and well-adapted to their hearing aids, there was variability in the length of time participants used their current hearing aids (M = 2.14 years, range = 1 week – 5 years), frequency of use (M = 13.32 h/day, SD = 5.10, range = 3–24), fit (hearing had been assessed by an audiologist within the previous 2 years on average), and model type (n = 14 behind-the-ear; n = 3 in-the-canal; n = 2 in-the-ear). Hearing thresholds were measured using SHOEBOX^TM [94]. A 4-frequency PTA was calculated for each ear using thresholds at 0.5, 1, 2, and 4 kHz. Participants with a better-ear PTA of ≤ 25 dB HL were classified as having normal hearing and those with a PTA ≥ 26 dB HL were classified as having hearing loss [95]. While the updated World Health Organization [96] standard cut-off is < 20 dB HL and has been used in current studies [e.g., 97], we used the previously recommended cut-off of ≤ 25dB HL [95] to be comparable and consistent with previous research in this field. All participants had symmetrical hearing, defined as a right-left ear threshold difference of no more than 15 dB at two or more adjacent frequencies. Participants also had to speak and have learned English fluently before the age of five and have normal or corrected-to-normal visual acuity with their habitual correction on the Early Treatment Diabetic Retinopathy Study chart (ETDRS; logMAR of −0.2–0.5 in the better eye) [98,99]. All participants were required to have normal cognition as determined through screening using the Montreal Cognitive Assessment (MoCA) [100], with a ≥ 23 cutoff off [101]. See S1 for detailed descriptions of all standardized assessment measures. Participants had to ambulate ≥ 10 meters independently, have a home computer or tablet with Internet connection. Participants were excluded if they had any major diseases affecting the ear, damage to the ear (e.g., occupational noise) or onset of hearing loss prior to adulthood. Further, participants were excluded if they had a self-reported history of any major neurological or psychiatric illnesses (e.g., depression), mild cognitive impairment or dementia, Parkinson’s disease or other neurological disorder or sequelae, recent stroke, seizure disorder or epilepsy, claustrophobia, susceptibility to motion sickness, arthritis, or a serious injury affecting the hand or arm, as the responses during the experiment were made using a game controller. Additional exclusion criteria included mobility-related impairments, such as the use of mobility aids, diagnosed vestibular disorders, or major musculoskeletal conditions that interfere with daily activities and current participation in any other cognitive training or exercise programs. This study was approved by the University Health Network’s research ethics board (Study ID: 19–5857). Data collection began June 12th, 2023, and was completed by April 2nd, 2024. Participants were compensated a minimum of $220 which comprised of $30 for the four in-person sessions and $100 for the 12-week at-home training period. See published protocol [102] and registered clinical trial (https://clinicaltrials.gov/ct2/show/NCT05418998). The authors confirm that all ongoing and related trials for this intervention are registered.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Consort flowchart of participant recruitment, eligibility, and assignment [93].

Consort diagram of participant recruitment and the final eligible sample analyzed after exclusions and dropouts. Specifically, one hundred participants were screened, sixty of whom were eligible. Nineteen participants were ineligible, and twenty-one individuals dropped out (n = 40) of the study due to various reasons (i.e., sickness including COVID-19, could not commit to study time commitment, no longer interested in participating, travel abroad). Of the sixty eligible participants, thirty-one were randomly assigned to the executive function (EF) condition, while twenty-nine were assigned to the control condition. Eight participants who were originally assigned to the control condition came back to complete the 12-week training and were added to the EF training condition, bringing this sample size of the EF training group to thirty-nine. Three participants, one from the EF condition and two from the control condition, were excluded due to technical errors that resulted in missing data. The final sample analyzed included thirty-eight participants in EF training and twenty-seven controls across the three age groups (blue outlined boxes on the last line). MA = Middle-Aged Adults with normal hearing, OA = Older Adults with normal hearing, OAHL = Older Adults with Hearing Loss. See full reference in the references section below.

https://doi.org/10.1371/journal.pone.0331276.g001

Study design and procedure

First, a screening was conducted by telephone to determine a participant’s initial eligibility to participate in the study. The screening included demographic and self-reported responses regarding eligibility criteria, as well as questions about the participant’s level of comfort in using a computer or tablet at home to complete the computerized training. Eligible participants were then asked to come to KITE-Toronto Rehabilitation Institute, University Health Network to provide written consent and complete a battery of sensory, mobility, and cognitive assessments (secondary measures of interest at T1 and T2, see below). If participants met the screening criteria; that is, the pre-determined cut-offs for the ETDRS and MoCA measures, and the age and hearing criteria for inclusion in one of the MA, OA, or OAHL groups, they were invited to complete the experimental session.

The experimental session involved completing an auditory-cognitive task (i.e., 2-back task) and a standing balance task (each task alone and simultaneously) in an immersive virtual reality simulator (see procedural details below). Participants were then randomized into either the EF training condition or the control condition for 12 weeks, with an attempt to balance the number of males and females within each group. After 12 weeks, participants returned to the lab and completed the same set of secondary measures (see Downey et al. [102] and below for details) and experimental tasks again at T2. Upon completing the study (T1 and T2), participants in the control condition were then offered access to the at-home EF training. A subset of the participants in the control condition (n = 8; see Fig 1) chose to complete the 12-week EF training intervention and following training they were brought back to the lab to complete the full post-training protocol. This waitlist-control subset included one MA, three OA, and four OAHL participants. These eight participants did not show any differences in performance compared to the participants in the original training condition. Therefore their initial data were included in analyses for the control condition and their post EF training data were included in the analyses for the training condition (see Fig 1).

Characterizing the participant sample and secondary outcome measures

To address the objectives of determining whether sensory, cognitive, and mobility-related abilities predict training-related outcomes and to further characterize the participant sample beyond eligibility criteria; that is, aside from the PTA, MoCA and ETDRS measures, additional assessments were conducted in-person or online via Qualtrics (an online survey software; Qualtrics, Provo, Utah).

Hearing. The Canadian Digit Triplet Test [103], Hearing Handicap Inventory for the Elderly [104] and Listening Self-Efficacy Questionnaire [105] were administered. Cognition. The Rey Auditory Verbal Learning Test [106], Digit Span Forward & Backward [107], Letter Number Sequencing [107], Digit Symbol-Coding [107], Stroop Color-Word Interference [108], Trail Making Test [109] and Frequency of Forgetting Questionnaire [110] were administered. Vision. The Pelli-Robson Contrast Sensitivity Test were administered [111]. Mobility. Mini-BESTest [112] and Activities-Specific Balance Confidence Scale [113] were administered. Other. The Health History Questionnaire (see Table 1 for a summary of demographics and S1 for assessment descriptions) was administered which asked questions related to health conditions and demographics, such as sex and education level. All assessments were administered again at T2 except for PTA, Hearing Handicap Inventory for the Elderly, ETDRS and Health History Questionnaire.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographics table of baseline sensory, cognitive, and mobility function.

https://doi.org/10.1371/journal.pone.0331276.t001

Experimental session and primary outcome measures

StreetLab virtual reality simulator.

The experimental session was completed in StreetLab, a fully immersive, state-of-the-art, multisensory, projection-based, virtual reality simulator used to simulate realistic conditions (see Fig 2). StreetLab has a 240° horizontal by +15° to −90° vertical field-of-view curved projection screen extending from floor to ceiling. The virtual environment used in this study depicted a large, urban, 6-lane intersection street-crossing in Toronto (developed using OpenScene Graph www.openscenegraph.org). StreetLab is outfitted with an AMTI (Advanced Mechanical Technology, Inc., Watertown, MA) BP12001200–2000 strain gage force plate that measures ground reaction forces. A harness was used to ensure participants could not fall to the ground surface but did not provide body-weight assistance during the study while standing. There is a spatialized surround sound system with seven speakers (Meyersound MP-4XP; Meyersound Laboratories, Inc., Berkeley, CA) distributed behind the projection screen, at approximately the height of a participant’s head when they are standing on the force plate positioned at 0° azimuth across a horizontal plane at ± 28° (right front and left front), ± 90° (right side and left side), and ± 127° (right rear and left rear). One subwoofer (Meyersound MP-10XP) is located in the floor under the centre speaker. All loudspeakers are at a distance of 2.14 m from the participant when standing on the force plate. A sound level meter recorded ambient noise in the laboratory at 51 dB SPL during quiet standing conditions (full acoustic details of the StreetLab environment are available in Campos et al. [114]).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. StreetLab.

(a) Exterior view of StreetLab. (b) Interior view of StreetLab depicting a participant standing on the force plate, oriented toward the projection screen displaying a simulated city street-crossing.

https://doi.org/10.1371/journal.pone.0331276.g002

Statistical analyses

All analyses were completed using R-studio (Version 2024.12.0 + 467). Histograms were used to visualize the distribution of all dependent measures (i.e., RTW, COP APSD) and boxplots were used to visualize outliers. Assumption checks confirmed acceptable model fit (i.e., homoscedasticity and linearity were adequately met). Multicollinearity was assessed using variance inflation factor diagnostics on the final linear mixed-effect models and all lower-order predictors fell within acceptable ranges, indicating that collinearity did not compromise model stability. If data points were above or below three SDs from the mean, instead of excluding them altogether, winsorization was used to replace these data points with the next most non-extreme value (cutoff: z <−3 or z > +3), affecting 3.46% of RTW values and 12.55% of COP APSD values. Outcome measures were normally distributed or near-normally distributed (skewness of −.31 for RTW and.48 for COP APSD) and platykurtic to slightly platykurtic (kurtosis of 2.27 for RTW and 2.26 for COP APSD). However, these skewness and kurtosis ranges were deemed to be within acceptable ranges. To preserve the original structure and interpretability of the dataset, no transformations were applied. Continuous variables were group-mean centered to facilitate interpretation of main effects and interactions.

For the postural data a sampling frequency of 1000 Hz was used for the force plate. MATLAB was used to process the data using a Butterworth filter with a filter cut-off of 6 Hz. Only COP path length variability in the AP direction is reported in the manuscript, as participants performed the balance task in a standardized shoulder-width stance, thereby stabilizing mediolateral sway. While mediolateral COP parameters were also analyzed, as expected, no significant effects were observed, therefore, this parameter is not discussed further.

Welch’s two-tailed, between-subjects t-tests were conducted to confirm that the participants assigned to the EF training condition were not statistically different from those assigned to the control condition at T1 with respect to primary outcomes and secondary measures of sensory, cognitive, and motor functioning. Visual inspection of individual participant data and mixed factorial ANOVAs were conducted to confirm whether the EF training condition improved over the course of the 12 weeks in the evaluation modules for Week 1 versus the evaluation modules at Week 12 for the visual n-back, Stroop, and visual discrimination dual-task.

Two linear mixed-effect models were conducted for each primary outcome measure: auditory 2-back RTW (ms) and COP APSD (mm). Following the approach outlined by [120], a backward elimination strategy was used to derive the most simplified model. Within-subject differences across cohorts were modeled using nested random intercepts. Specifically, random intercepts were specified for participant ID nested within testing cohort to account for repeated measures and potential procedural variability across testing waves. Fixed effects included the following and their interactions: Group (MA, OA, OAHL), Intervention (EF training, control), Time (pre-training/T1, post-training/T2), Task (single-task, dual-task), Eyes, (eyes-open, eyes-closed), PTA, MoCA, Mini-BESTest and covariates such as Age (continuous variable), Education (years) and Sex (male, female). For continuous significant fixed effects (i.e., PTA, MoCA, Age) there were a range of abilities. For example, for PTA participants across the spectrum of hearing abilities were recruited (normal hearing to hearing loss). Therefore, when reporting significant results of PTA below, the different levels of hearing loss were binned into tertiles across the entire sample. Specifically, “lower PTAs” were 9 dB HL or lower, “average PTAs” ranged from 10 dB HL to 45 dB HL (average = 28 dB HL), and “higher PTAs” were 46 dB HL or higher. The total range of PTAs in the entire sample was 3 dB HL to 71 dB HL. When reporting significant results of MoCA below, the different levels of MoCA were binned into tertiles across the entire sample. Specifically, “lower MoCA scores” ranged from 23 to 25 out of 30, “average MoCA scores” ranged from 26 to 28 out of 30 (average = 27), and “higher MoCA scores” were 29 or 30 out of 30. The range of MoCA scores across the total sample was 23–30 out of 30. When reporting significant results of Age, different Age levels were binned into tertiles across the entire sample. Specifically, “lower Ages” were 56 years or younger; “average Ages” were 57–78 years (average = 67); “higher Ages” were 79 years or older. The range of ages across the entire sample was 45–90 years. Note that for the fixed effects included in the models (Age, PTA, MoCA) the tertile-binned analyses were only done for post-hoc analyses to interpret the main effects of interactions which showed significant effects on the continuous variables. Post-hoc tests were conducted with Bonferroni correction. Linear mixed-effect models were refined based on likelihood ratio tests evaluating model fit. Non-contributory fixed effects and interactions were excluded unless they represented key confounding variables, such as age, education, sex or were integral to testing the primary or secondary objectives. Note that only significant main and interaction factors of Time or Time x Intervention are reported below, as these significant results allow us to directly evaluate our primary and secondary research questions of interest, which are focused on training-related effects. The threshold for significance was set to p < 0.05.

Comparability of participants in the control and training conditions

To ensure participants in the EF training and control conditions had similar sensory, cognitive, and mobility abilities at baseline, we conducted Welch’s two-tailed independent-samples t-tests across all three groups between the participants assigned to EF training condition and those assigned to the control condition across all standardized measures and found no significant differences with two exceptions. Specifically, there was a significant difference in age between those in the OA group who were assigned to the EF training condition compared to those assigned to the control condition (p = 0.024), with those in EF training condition being younger (M = 69.54) than those in the control condition (M = 72.60) and Mini-BESTest performance (p = 0.041) of those in the EF training condition being significantly worse (M = 23.08) than those in the control condition (M = 25.10; see Table 1 for demographics across all three groups). To ensure participants assigned to the EF training and control conditions had similar primary outcomes at baseline, we conducted Welch’s two-tailed independent-samples t-tests across all three groups between the participants assigned to the EF training condition and those assigned to the control condition across RTW and APSD and found no significant differences.

Demonstrated effects of the 12-week training program

To determine whether participants in the EF training condition demonstrated improvements from Week 1 of training to Week 12 of training, the evaluation modules for the visual n-back, Stroop, and visual discrimination dual-task RTs were compared. Overall, visual inspection of the data revealed that over the 12 weeks, every individual improved on the visual n-back task, apart from 1 OAHL and the Stroop task, apart from 1 MA. However, the results for the visual discrimination dual-task were mixed, with some participants improving, including 3 MA, 4 OA, 5 OAHL and others not improving, including 7 MA, 9 OA and 9 OAHL. To evaluate these training-related effects statistically, separate mixed-factorial ANOVAs, including the within-subject factor Time (Week 1, Week 12) and between-subjects factor Group (MA, OA, OAHL) were conducted for each of the three training tasks. For the Stroop task and visual n-back task, a main effect of Time was observed indicating that Week 12 RT performance was better/faster than Week 1 RT performance; (F(1, 34) = 47.29, p < 0.001 and F(1, 32) = 61.82, p < 0.001, respectively). However, for the visual discrimination dual-task there were no significant improvements. While a main effect of Time was also observed for the visual discrimination dual-task (F(1, 34) = 7.23, p = 0.011), the effects were in the opposite direction, such that RT performance was worse at Week 12 compared to Week 1. There were no main effects of Group or interaction effects for any of the three training tasks.

Effects of EF training on balance and auditory-cognitive performance in StreetLab

Standing balance performance.

To address our primary objective of investigating the effects of EF training across Time and Groups on standing balance performance, a linear mixed-effect model was conducted with the primary balance outcome of interest, COP APSD (mm), as the dependent variable, where higher values were interpreted in this context as indicating worse performance. First, a significant four-way Intervention*Time*Age*PTA interaction was demonstrated on COP APSD performance, F(1,425.11) = 5.0022, p = 0.026, η_p² = .01, with post-hoc tests demonstrating that for those in the EF training conditions, but not controls, T2 balance was significantly better (smaller COP APSD) than T1 balance for participants of older ages and lower dB HL levels (i.e., better hearing; t = 2.78, p = 0.006, Fig 3). Further, for those in the EF training condition, there were no improvements in balance at T2 versus T1 in those with the average to higher levels of hearing loss across all age tertiles (younger, average, older; performance was worse; see S2 for full statistics). To summarize, those in the EF training condition who were of older ages and had lower levels of hearing loss/better hearing demonstrated significant improvements in balance between T1 and T2; however, those in the EF training condition with average to higher levels of hearing loss across all ages showed no improvements after training.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Linear mixed-effect model of intervention x time x age x PTA interaction effect on COP APSD.

Estimated marginal means of Centre of Pressure Path Length Anterior-Posterior Standard Deviation (COP APSD in mm; higher scores represent worse performance) from T1 to T2 for those in both the EF cognitive training (T) and control (C) conditions. Age and hearing loss (HL) tertiles are represented by the columns and rows respectively. All significant findings are stated in-text. Asterisks indicate statistically significant training-related improvements. Errors bars represent the 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0331276.g003

Second, we found a significant four-way Intervention*Time*Group (MA, OA, OAHL)*MoCA interaction on COP APSD performance, F(2,426.06) = 8.8496, p < .001, η_p² = .04, but for which no post-hoc tests indicated improvements in balance related to training. Instead, there was significantly worse balance performance at T2 compared to T1 in the OAHL in the control condition with lower MoCA levels and the OA in the EF training condition at all MoCA levels (low, average, and high; see S2 for full statistics). To summarize, these interactions did not provide evidence of any unique training-related improvements to balance that were predicted by interactions among group-related factors (age and hearing status) and cognitive abilities.

Auditory 2-back performance.

First, to address our primary objective of investigating the effects of EF training across Time and Groups on auditory-cognitive performance, a linear mixed-effect model was conducted with the auditory 2-back RTW (ms) as the dependent variable, for which higher values indicate worse performance. Results demonstrated a significant four-way Intervention*Time*Age*PTA interaction on 2-back RTW performance, F(1,418.43) = 4.700, p = .031, η_p² = .01. Post-hoc tests revealed that for those in the EF training condition, but not the control condition, RTW at T2 was significantly better/faster than at T1 for participants of average ages and lower PTAs/better hearing (t = 3.576, p < .001, see Fig 4). Further, for those in the EF training condition, but not the control condition, RTW at T2 was significantly better than at T1 for participants of older ages and lower PTAs/better hearing (t = 3.646, p < .001). Those in the control condition did not improve from T1 to T2; in fact, their performance generally worsened in terms of auditory 2-back performance (see S3 for full statistics). To summarize, participants in the EF training condition of average-to-older ages and lower levels of hearing loss demonstrated significant improvements in auditory 2-back RTW, with the control condition showing no improvements.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Linear mixed-effect model of intervention x time x age x PTA interaction effect on auditory 2-back RTW.

Estimated marginal means of auditory 2-back reaction time weighted (RTW in ms, higher scores represent worse performance) from T1 to T2 for both the EF training (T) and control (C) conditions. Age and hearing loss (HL) tertiles are represented by the columns and rows respectively. All significant findings are stated in-text. Asterisks indicate statistically significant training-related improvements. Errors bars represent the 95% confidence intervals around model-estimated marginal means, which may extend beyond observed data ranges (i.e., max of 2000 ms) due to statistical estimation.

https://doi.org/10.1371/journal.pone.0331276.g004

We also found a significant four-way Intervention*Time*Group (MA, OA, OAHL)*MoCA interaction on 2-back RTW performance, F(2,425.46) = 8.7397, p < .001, η_p² = .04. Post-hoc tests demonstrated that for the EF training condition, but not the control condition, RTW was better at T2 than T1 for OAHL with lower MoCA scores (t = 2.145, p = 0.033). In contrast, OA at lower-average MoCA levels in the EF training condition had poorer RTW at T2 compared to T1 (see S3 for full statistics). The control condition did not improve from T1 to T2; in fact, they generally worsened in terms of auditory 2-back performance (see S3 for full statistics). To summarize, participants in the EF training condition with the worst hearing abilities (i.e., OAHL) and poorest cognitive abilities demonstrated significant improvements in auditory 2-back RTW, whereas the participants with normal hearing and lower-to-average cognition did not improve. Again, the control condition showed no improvements.

Cognitive aging frameworks to understand dual-task auditory-cognitive and balance performance

Several theoretical models of cognitive aging may help explain the factors that influence responsiveness to cognitive training. Broadly, these models emphasize two primary mechanisms: compensation and magnification [91]. Compensation includes the following principles: 1) individuals with lower baseline performance can benefit the most from cognitive training through compensation and 2) there is a negative correlation between baseline performance/cognitive abilities and cognitive training gains [91]. There is some evidence supporting compensation in the current study for both balance-related outcomes and auditory-cognitive related outcomes. For example, training-related improvements to balance were observed only for participants who were in the older age range and who therefore might have more to gain from training compared to participants who were younger. Similarly, training-related improvements to auditory-cognitive task performance were mostly observed for older adults with hearing loss who had poorer cognition who may have had more to gain from cognitive training than those with normal hearing and better cognition. In contrast, magnification refers to a pattern in which individuals with higher cognitive processing gain more from training due to an increased ability to acquire and refine cognitive strategies and includes the following principles: 1) individuals with higher baseline performance derive greater benefits from training than those with lower baseline performance and 2) there is a positive correlation between baseline performance/cognitive abilities and cognitive training gains [91]. There is some evidence supporting magnification in the current study as the training-related improvements to both primary outcomes (RTW, APSD) were found in the individuals with better hearing, but not in those with poorer hearing.

For the balance outcomes, older adults with better hearing benefited the most from training compared to poorer hearing at any age. For auditory-cognitive task performance, average to older adults with better hearing benefited from training, while those with the poorest cognition and poorer hearing benefited the most from training. Overall, this pattern of results suggests that there may be some processing limit for hearing that was reached, such that beyond a certain degree of hearing loss, cognitive training may no longer be effective, or more cognitive training may be required to improve balance during realistic dual-tasks with auditory-cognitive demands. It is also possible that structural changes to the brain related to auditory deprivation over time may alter how cognitive resources are managed and allocated, influencing training-related effects. However, as demands increase, eventually a resource-limited ceiling is reached, leading to an inability to compensate, leading to associated behavioural performance declines and perhaps associated declines in training-related benefits. In the current study, it is possible that once an upper threshold for hearing loss is reached, EF training may no longer be sufficient to compensate for the high demands required to perform an auditory-cognitive task while maintaining standing balance.

Limitations and future directions

An important limitation of the current study is the nature of the participant samples. We recruited generally healthy, high functioning adults with no major health conditions or other comorbidities that are common with older age and other health conditions that increase risk for cognitive decline and hearing loss (e.g., diabetes). For some measures, including the MoCA, we had a very small range of abilities that were limited to the better end of the ability spectrum. While this strategy of only recruiting physically and cognitively healthy participants was intentional to limit potential confounds in the interpretation of the data, it also limits the generalizability of these findings to the broader older adult population, which is highly heterogeneous. Consequently, the current results likely provide a conservative estimate and might not adequately reflect the potential magnitude of EF training benefits on cognitive-balance performance in the general population or in real-world settings. Future research would benefit from broadening eligibility criteria to include adults with a wider range of sensory, cognitive, and balance abilities to better evaluate the extent to which these abilities predict training-related outcomes.

Efforts were made to achieve a balanced representation of sex across study groups; however, dividing participants by intervention conditions (EF training vs. control) yielded subgroup sizes too small to permit meaningful analyses of sex-related differences. Nonetheless, sex and gender are important factors to consider, as both may influence the efficacy of EF training on cognitive and sensorimotor performance. Moreover, hearing loss tends to be more prevalent in men than in women, a disparity attributed to both biological and environmental influences [121–123]. For example, estrogen is thought to exert a protective effect on auditory function in premenopausal women (i.e., a sex-related factor), while greater exposure to occupational noise in male-dominated industries (e.g., construction, manufacturing) may contribute to elevated risk among men (i.e., a gender-related factor; [121–123]). Future studies should recruit larger, stratified samples to examine how these sex and gender differences may interact with auditory, cognitive, and balance outcomes [124–126].

Another potential limitation of the current study is the absence of long-term follow-up assessments to evaluate the sustainability of the observed EF training effects on auditory-cognitive and balance performance. While short-term post-training improvements were observed in both auditory-cognitive and postural measures in the current study, it remains unclear whether these gains could be sustained over time. Without longitudinal data, we cannot determine the extent to which the observed benefits reflect temporary performance boosts versus lasting functional improvements. Of note are two studies [127,128] that examined the long-term effects and durability of EF training on cognitive abilities and everyday functions in independently living older adults. Both studies trained memory, reasoning, and speed of processing separately and demonstrated training resulted in fewer declines to independent activities of daily living compared to the control condition [127], fewer declines to independent activities of daily living after reasoning training specifically [128] these gains were sustained for at least 5 years [128]. However, these longitudinal studies did not consider the effects of hearing loss on observed benefits to independent activities of daily living. Future studies should incorporate follow-up assessments to evaluate the durability of cognitive training effects on balance and cognitive outcomes over time and in individuals with hearing loss. Important future considerations to determine the efficacy of at-home executive function training can include the length of time spent training, the time and intensity of training (e.g., time per day/days per week), and types of cognitive domains trained. Moreover, longitudinal studies are also required to determine whether any training-related improvements in balance and/or auditory-cognition also result in real world outcomes such as reduced fall risk or improved mobility. Future research could also explore whether lifestyle factors, such as physical activity levels, nutrition, and social engagement moderate training-related gains [93].