This pilot study was a single center RCT conducted at the Montreal Geriatric University Institute in Quebec, Montreal. It employed a two-arm parallel, open-label design with participants assigned to either an intervention or control group. The intervention group engaged in an in-person, music-based VR experience using a VR headset (MetaQuest Pro from Facebook Technologies LLC, equipped with advanced optical technology, featuring a high-resolution display with a combined pixel count of 3664 × 1920 - 1832 × 1920 per eye - LCD panels, a 90Hz refresh rate and light blockers and integrated speakers with spatial audio (Meta Platforms, Inc, 2024). The control group participated in traditional music listening with a standard MP3 headphones (Innioasis, 2024) and Sennheiser HD569 wired supra-aural headphones with sound isolation (Sennheiser, 2024). Participants were randomly assigned to their respective groups using Participants were randomly allocated into intervention and control groups, each participant having a 50% chance of being assigned to either group, thanks to a pre-established randomization list. The randomization list was established using the N’Query randomization software. This method was chosen to ensure equal group sizes and maintain balance in sample distribution over time. Although neither the participants nor the investigators were blinded to the intervention, randomization results were disclosed to participants and other research team members only after the initial assessment (i.e., before the intervention) to minimize bias. This RCT is registered on the ClinicalTrials.gov website (Project Number NCT06296199) and adheres to the CONSORT guidelines for randomized controlled trials (Moher et al., 2012).

The study was conducted at the memory and geriatric outpatient clinics of the Montreal Geriatric University Institute (Quebec, Canada). Inclusion criteria required participants to be 1) community-dwelling, 2) older adults aged 60 and above, 3) who receive care at the memory or geriatric outpatient clinics of the Montreal Geriatric University Institute. Participants with severe visual or hearing impairment or vestibular-origin balance disorders were excluded. Participants unable to provide informed consent or those diagnosed with severe-stage major neurocognitive or psycho-behavioral disorders were excluded. Eligible participants were recruited through open and rolling recruitment within the outpatient clinics. While participants were community-dwelling, this recruitment strategy specifically targeted individuals receiving specialized care in geriatric and memory clinics and was not intended to evaluate the effects on cognitive health in patients with neurocognitive disorders, nor to represent the broader older adult population.

Eligible patients were pre-selected by their attending physician at the memory and geriatric outpatient clinics between August and October 2024. A research team member then contacted 108 pre-selected patients by phone to inform them about the new clinical research opportunity, conduct a final screening, obtain informed consent and schedule an appointment for the intervention at the research center. Of these 108 potential participants, 85 (78.7%) were reached by phone. Of these 85 reached participants, 41 (48.2%) declined to participate. Ultimately, 44 participants (51.8%) agreed to participate in the RCT but 3 (6.8%) of them withdrew their interest. Therefore, 41 participants (93.2%) were enrolled and were randomized into the intervention (n = 21) and control (n = 20) groups. Among them, 1 participant (2.4%) dropped out during the intervention phase (from the intervention group) because of major technical issue. A major technical issue is defined as a technical problem requiring the intervention to be interrupted with the removal of the VR headset and requiring the intervention of a member of the research team. The Figure 1 showed the CONSORT flow diagram detailing the selection of participants.

All participants completed assessments both before and after the intervention. Regardless of group assignment, information was collected on age, sex, body mass index (BMI), neurocognitive disorder diagnoses and preference for classical music. The presence of neurocognitive disorders was determined based on participants’ clinical records and diagnostic assessments made by the referring geriatricians who used the DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition) criteria for major neurocognitive disorders requiring the presence of multiple cognitive deficits, including memory impairment, and at least one other cognitive domain impairment (language, praxis, gnosis, or executive function), with significant functional decline. Additionally, assessments included the CARE scale (Beauchet et al., 2022a) which integrates the baseline clinical and functional parameters (age, ADL, IADL, depression, BMI and frailty). Participants’ prior exposure to VR technology, educational background and prior musical training was not systematically collected, as the study focused primarily on the immediate acceptability and short-term emotional effects of the intervention even if these factors are known to influence emotional processing and engagement with music-based interventions (Hanna-Pladdy and Gajewski, 2012; Schellenberg, 2006).

Before and after the intervention, participants in both the intervention and control groups completed several questionnaires to assess emotional state (with Positive and Negative Affect Schedule scale (Watson et al., 1988), wellbeing (with Warwick-Edinburgh Mental Wellbeing Scale (Bass et al., 2016)) and mood (with Visual Analog Mood Scale (Visual Analogue Mood Scales, 2024).

The number of dropouts, linked or not to major technical incidents, was recorded to assess the acceptability of the VR intervention. Additionally, participants in the intervention group also completed two additional questionnaires (Acceptability Questionnaire, internally defined on a Likert scale model, and Simulator Sickness Questionnaire [SSQ] (Kennedy et al., 2019)) post-VR experience. Emotional state was assessed using the Positive and Negative Affect Schedule (PANAS) (Watson et al., 1988), a scale designed to evaluate both the valence and intensity of an individual’s emotional state defined as momentary expressions within that broader affective experience. PANAS consists of two subscales: positive and negative emotional states, each containing 10 items, for a total of 20 items. The scale categorizes individuals into four emotional profiles, reflecting different interactions between positive and negative emotions. Scores for both subscales range from 10 to 50. A low negative emotional state score indicates minimal negative emotions, reflecting calm and serenity, whereas a high score suggests psychological distress with various unpleasant emotions. Conversely, a high positive emotional score reflects a strong presence of positive emotions, while a low score indicates reduced positive emotional state.

The self-administered Warwick-Edinburgh Mental Wellbeing Scale (WEMWBS) (Bass et al., 2016) was used to assess overall wellbeing. This scale consists of 14 positively worded items, with scores ranging from 14 (indicating “none of the time”) to 70 (indicating “all of the time”), providing a comprehensive measure of mental wellbeing. The Visual Analog Mood Scale (VAMS) (Visual Analogue Mood Scales, 2024), was used to evaluate mood. This scale assesses eight mood states using ideogrammatic icons, including two positive mood states (happy and energetic) and six negative mood states (scared, confused, sad, angry, tired, tense). Each mood state is represented by a neutral face on a continuous line, where participants mark the point corresponding to their mood intensity. The score is determined by measuring the distance (in millimeters) from the neutral face to the marked point, greater distances indicate stronger mood. Higher scores for positive mood reflect a better mood, whereas higher scores for negative emotions indicate a worse mood.

The intervention consisted of a 15-min experience in which participants in the intervention group listened to and watched classical musicians perform via a VR headset, while participants in the control group only listened to the same musical suite, defined here as a curated and thematically organized sequence of musical pieces, through an audio headset.

The 15-min single-session format was chosen for its suitability in geriatric care, where brief interventions can enhance feasibility, engagement, and emotional responsiveness, and are supported by theoretical models emphasizing the benefits of short, targeted affective experiences (Bannon et al., 2025) and minimum side-effects (Cinalioglu et al., 2023; Sekhon et al., 2024). All participants, regardless of group allocation, experienced the same pre-recorded musical suite performed by professional musicians with extensive experience in chamber music, ensuring a high level of artistry. A carefully curated selection of classical and popular music, chosen by the concerned professional artists in accordance with directives of the research team, was recorded and presented by a trio ensemble consisting of flute, violin, and cello. The repertoire featured both instrumental works and lyrical excerpts, offering a diversified listening experience. The performance was filmed in 360° during a live rehearsal session, emphasizing the authenticity and intimacy of in-the-moment performance while maintaining a high standard of musical interpretation.

The selected musical works included in the musical suite were, by chronological order, 1) Romance for Violin and Orchestra No. 2 in F major–Ludwig van Beethoven (1798), a gentle and lyrical piece, 2) Suite No. 2 in B Minor, BWV 1067 – Johann Sebastian Bach (1738), an uplifting and structured Baroque composition, 3) Lascia ch’io pianga (Almira) – George Frideric Handel (1705), an aria expressing longing and introspection, 4) La donna è mobile (Rigoletto) – Giuseppe Verdi (1851), in its instrumental version, 5) Por una cabeza–Carlos Gardel (1935), in a classical arrangement of this well-known tango, 6) El Choclo–Ángel Villoldo (1903), another iconic tango, also in a classical arrangement, 7) Quién quiero no me quiere–Lucho Barrios (1964), Peruvian waltz, arranged for classical trio, and 8) Petradaki Patradaki–Nikos Xanthopoulos (2015), a contemporary Greek melody with strong folkloric undertones.

The selection of musical works was designed to span a wide range of musical eras and cultural contexts with the aim of stimulating affective responses, autobiographical memory and a sense of comfort in older adults. The combination of classical repertoire and popular melodies, reinterpreted in a chamber music format and enhanced by lyrical elements, enriched the sensory experience and intentionally reflected the diverse cultural backgrounds and musical preferences of older adult populations.

The intervention was conducted in a dedicated room under researcher supervision in a quiet clinical setting at the Research Centre of Montreal Geriatric University Institute (Quebec, Canada), where participants were seated comfortably in an armchair. Each session was conducted individually in a dedicated quiet room. Participants in the intervention group experienced a single 15-min VR session using the MetaQuest Pro headset. The control group listened to the same musical program through stereo headphones while seated comfortably. Before the intervention, participants completed the pre-intervention assessments and received an explanation of the standardized procedure and equipment and device adjustment. Volume levels were individually adjusted prior to each session to accommodate hearing differences among participants. During the intervention, the environment remained silent. No visual stimuli were presented in the control condition. Participants were not given explicit instructions regarding whether to keep their eyes open or closed, nor were they guided to imagine content. In both conditions, they were asked to remain silent during the session to avoid external influence on emotional experience. If any technical issues arose, participants could signal the research team member, who stayed in the room during the entire intervention.

Since it was not possible to determine the required sample size based on acceptability (assessed through descriptive criteria in the intervention group only) which is the primary outcome, the calculation was instead based on the change in the positive and negative affect schedule (PANAS) positive score before and after the intervention, using the formula: ((score M ⁿ⁺¹ – score M ⁿ)/ ((score M ⁿ⁺¹ + score M ⁿ)/2))x100. PANAS reflects immediate emotional state and is sensitive to short-term interventions (Watson et al., 1988). An expected intergroup difference of 8 points was chosen based on prior research reporting smaller but significant changes following brief VR or music interventions (Chan et al., 2011), combined with clinical judgment that a larger effect could be observed in an immersive VR context. A free online sample size calculator (http://www.sample-size.net/sample-size-means/) was used to determine the required sample size. To detect a significant absolute difference between the two groups (intervention and control) with an alpha level of 5% and a power of 90%, a minimum of 20 participants per group was required.

The primary outcome corresponded to the acceptability of the VR intervention. Considering that acceptability remains poorly defined and lacks a unified framework (Bucyibaruta et al., 2022), we decided to use three specific criteria to better define it (M. Sekhon et al., 2017). It included adoption (i.e., willingness to complete the entire study), satisfaction (i.e., user satisfaction, perceived effectiveness) and tolerance (i.e., cybersickness) (Sekhon et al., 2017). First, adoption was defined as the retention rate which was determined by calculating the proportion of participants who completed the entire study without dropout or withdrawal due to major technical incidents or for other reasons. Retention rate was expressed as a percentage, with a retention rate of ≥80% deemed acceptable (Walters et al., 2017). Second, satisfaction was determined by calculating the proportion of participants willing to repeat the VR experience (user satisfaction) and the proportion of participants who believed in its potential mental health benefits (perceived effectiveness). Both variables were assessed using 7-point Likert scales, where higher scores indicated stronger agreement (1 = strong disagreement, 7 = strong agreement). A score of ≥6 was considered indicative of very good agreement for the willingness to repeat the VR experience and score of ≥5 was considered indicative of very good agreement for the perceived effectiveness on health. A high level of perceived interest was defined as at least 80% of participants who completed the study scoring ≥6 on the will to reuse the intervention measure and ≥5 on belief in a potential mental health benefit (Anastasiadou et al., 2024). Lastly, tolerance was assessed by using the Simulator Sickness Questionnaire (SSQ) (Kennedy et al., 2019). This questionnaire consists of 16 items evaluating symptoms such as fatigue, headache and nausea, each rated from 0 (no symptoms) to 3 (severe symptoms). The total SSQ score ranges from 0 (excellent tolerance) to 48 (poor tolerance), with ≤9 indicating very good tolerance (Lin et al., 2023). A high level of tolerance was defined as at least 80% of participants in the intervention group scoring ≤9 on the SSQ. The primary outcome was assessed based on participants randomized to the intervention group. For the adoption criterion, all randomized participants in intervention group were considered (n = 21), whereas satisfaction and tolerance were evaluated only among those who completed the full intervention (n = 20).

The second category of outcomes corresponded to the intervention’s mental health effects. For this purpose, we looked after the mean value of scales assessing emotional state, wellbeing and mood; each scale being performed before and after intervention in the intervention group and the control group. In addition, we used the score variations for these scales between pre- (T0) and post- (T1) intervention for control group and the intervention group using the formula: [(score T1) – (score T0)/n (score T1 + score T0)/2] X 100).

This study is conducted in accordance with the ethical standards set forth in the Helsinki Declaration (1983). Participants were included after giving written, informed consent for research. Data were anonymized and the study received was approved by the CIUSSS Centre-Sud-de-l’Île-de-Montréal (Quebec, Canada) Research Ethics Committee Vieillissement et neuroimagerie (# 2024-2064 – CÉR VN 23-24-40). This RCT was registered on ClinicalTrials.gov (project number NCT06296199) and adheres to the CONSORT guidelines for RCTs (Moher et al., 2012).

Analyses were conducted on an intention-to-treat basis. Missing data, which constituted less than 10% for each variable, were addressed using multiple imputation techniques. This method involved replacing missing values with the mean of the observed values for the respective variable within the corresponding allocated group (Mohamed et al., 2018). The imputed data represented 0.4% of the complete dataset.

Descriptive statistics (means, standard deviations [SD], frequencies, and percentages) were used to summarize participants’ characteristics. Between-group comparisons of change scores (post–pre) were conducted using Mann-Whitney U tests, while within-group differences were analyzed using Wilcoxon signed-rank tests. Multiple linear regressions examined the association of the intervention (used as independent variable) with change (between after and before intervention) in mean values of scales assessing emotional state, wellbeing and mood (used as the dependent variable, separated model for each variable). In order to limit the risk of overfitting in this sample (n = 40) while still accounting for key baseline differences, the models were adjusted for four covariates selected a priori based on theoretical relevance and clinical comprehensiveness: music preference, sex, neurocognitive disorder status and the CARE score. Bonferroni corrections were applied where relevant: for baseline comparisons (n = 11), the adjusted threshold was p < 0.0045; for outcome comparisons across groups and time (n = 20), p < 0.0025 was used. For regression analyses, each model tested a separate hypothesis, and a conventional threshold of p < 0.05 was applied. All statistics were performed using SPSS (version 29.0.2.0; SPSS, Inc., Chicago, IL).

As shown in the Table 1, participant ages ranged from 66 to 95 years (M = 79.8, SD = 6.5), with a BMI range of 16.3–44.3 kg/m² (M = 26.6, SD = 6.2). Participants were predominantly female (n = 24 [60.0%]), all were of Caucasian ethnicity (n = 41 [100%]), 23 [57.5%] had a neurocognitive disorder (NCD).

Most participants reported liking classical music (n = 38 [95.0%]) and 21 [52.5%] were not depressed. Additionally, 26 [65.0%] participants were independent in activities of daily living (ADL) and 28 [70.0%] in instrumental activities of daily living (IADL). Half of the participants were equally categorized as robust or frail (n = 20 [50.0%]).

No significant differences were observed between the intervention and control groups for demographic variables, except for sex. There were fewer females in the intervention group compared to the control group (n = 24, p = 0.003).

A descriptive analysis showed a high retention rate in the intervention group (Figure 1; Table 2, n = 20, 95.2%, 95% CI [−0.1; 0.2]) across 20 full VR sessions. It was also showed a high satisfaction with 85.0% of participants indicating a strong willingness to repeat the VR experience (Table 2, n = 17, 95% CI [0.7, 1.0]) and 85.0% agreeing that it could benefit their mental health (Table 2, n = 17, 95% CI [0.7, 1.0]). Tolerance to the intervention was also high with 95.0% (Table 2, n = 19, 95% CI [-0.5; 0.2]) of participants scoring below 9 on the SSQ.

Intra-group comparisons (pre-post) using Wilcoxon signed-rank tests revealed no significant changes in any outcome for either the intervention or control group (Table 3). Inter-group comparisons of change scores (post-pre), performed using Mann-Whitney U tests, also showed that there was no statistically significant difference (Table 3). However, trends favoring the intervention group were observed in the PANAS positive score (U = 136.5, p = 0.072) and the WEMWBS wellbeing score (U = 133.5, p = 0.086) (Figure 2).

Multiple linear regression analyses further indicated a significant increase in the PANAS positive score in the intervention group (β = 15.9, 95% CI [6.9, 25.1], p = 0.001; adjusted R ² = 0.32), after adjusting for baseline characteristics. No other outcome was significantly associated with the intervention, and the adjusted R ² values for these models ranged from −0.02 to 0.13, indicating minimal variance explained. Cohen’s d values were also computed to complement regression estimates and assess the magnitude of between-group differences. A moderate effect size was observed for the PANAS positive score (d = −0.66), supporting the clinical relevance of the observed improvement. Other outcomes showed small or negligible effects (see Table 4), consistent with the non-significant findings.