The effects of music-based interventions on behavioural and psychological symptoms of people living with dementia: a systematic review and network meta-analysis protocol

Abstract

Objectives

People living with dementia often experience behavioural and psychological symptoms of dementia (BPSD), which severely affect their well-being during the course of the disease. Particularly for BPSD outcomes, there is a high demand for increasing the evidence-based knowledge of non-pharmacological approaches, such as music-based interventions. Although previous reviews emphasize the potential effects of music-based interventions in people with dementia, they cover a wide range of different interventions and outcomes.

Method

Therefore, this systematic review (SR) and network meta-analysis (NMA) aims to not only investigate the efficacy of music-based interventions on BPSD, but also to compare the impact of different types of music-based interventions on outcomes. Preferred reporting items for SR and meta-analysis protocols (PRISMA-P) and the PRISMA NMA extension were followed. Several databases will be searched from inception to the date the search will be performed, for relevant randomized or non-randomized controlled trials comparing a music-based intervention with treatment as usual, active controls, or another music-based intervention. Multivariate pairwise meta-analyses will be conducted for each outcome. NMA based on a frequentist random-effects model will be used to estimate the comparative effects of each type of music-based intervention and related components across outcomes. Heterogeneity will be investigated by meta-regression models.

Conclusion

Based on our knowledge, this may be the first SR and NMA study to compare the efficacy of different types of music-based interventions. In addition, combined with our multivariate analysis approach, it will allow us to identify potential effect modifiers in music-based intervention for treating BPSD.

Keywords:

• Network meta-analysis
• music interventions
• music therapy
• dementia
• behavioural and psychological symptoms

Trial registration number::

• Prospero CRD42023457452

Introduction

Dementia is a clinical syndrome which is characterised by deterioration in cognitive, behavioural, social, daily activity and emotional functions. The World Health Organization (2021) estimated that 55.2 million people worldwide were living with dementia in 2019, a figure that may grow to 78 million in 2030, and 139 million by 2050. Given the rapid increase and that dementia effects many domains of wellbeing (i.e. cognitive, psychological, behavioural, social), the availability of appropriate interventions is critical.

Behavioural and psychological symptoms of dementia (BPSD) are defined as symptoms of disturbed perception, thought content, mood, or behaviour that will affect nearly all people with dementia over the course of their illness (Finkel et al., 1996; Kales et al., 2015). Pharmacological interventions are available but mainly show limited effect on alleviating BPSD and have negative side effects (Harrison et al., 2021; Huybrechts et al., 2011). Given the contraindications of these interventions and their limited effects on symptoms, non-pharmacological interventions, such as music-based interventions, may be an effective alternative to alleviating and treating BPSD.

Music-based interventions

The umbrella term ‘music interventions’ is used to denote all therapeutic interventions where music is the key component in effecting the desired treatment effect (De Witte, 2021). Music-based interventions may include singing, performing, or creating music, moving to music, and listening to music, or any combination of these. Music-based interventions may be implemented by music therapists, in the context of music therapy, but can also be self-administered or delivered by other professionals or family members (American Music Therapy Association, 2019; Bradt et al., 2013; Gold et al., 2011). Studies show that music interventions are strongly associated with a wide range of positive health and well-being outcomes for a various patient population (e.g. Koelsch, 2015).

There are several explanations as to why music-based interventions are an effective and appropriate treatment to alleviate BPSD in people with dementia. Proposed mechanisms encompass the stimulation of neuroplasticity and neurochemical functions, synchronization of auditory and motor systems, entrainment of neural activity, pathways connecting arousal and mood, as well as activation of autobiographical and implicit memory (Brancatisano et al., 2020; Gold et al., 2019; Koelsch, 2014; O’Kelly et al., 2013; Park et al., 2016; Vuilleumier & Trost, 2015). The interplay between the musical components regulates behavioural and psychological change (Koelsch, 2014). Music regulates (1) cardiovascular activity (Watanabe et al., 2017), (2) limbic, paralimbic and cortical brain activity responsible for emotion (Koelsch, 2014), (3) mesolimbic dopaminergic reward pathways, the hypothalamus-pituitary-adrenal axis stress response (Schaefer, 2017), and (4) involuntary movement and motoric expressions (Koelsch, 2014). In addition, shared musical experiences between people with dementia and others results in social bonding and connectedness (Thompson et al., 2023). Collectively, the activation of these mechanisms results in improved memory, reduced anxiety, stress and agitation and improved attention and orientation to space, time and person (Baker, 2001).

Recent reviews on music-based interventions for people with dementia

An increasing number of empirical studies have shown that music-based interventions are promising and potentially cost-effective non-pharmacological approaches for people living with dementia (Livingston et al., 2014). This is reflected by the growing number of systematic reviews and meta-analyses in this research field (more than 60 since 2010). In a meta-analysis of randomized controlled trials (RCTs) Chang et al. (2015) reported that music therapy significantly improved disruptive behaviours (g = −0.66; 9 RCTs) and decreased anxiety (g = −0.51; 5 RCTs) in people with dementia. The meta-analysis of Zhang et al. (2017) also showed significant but more modest effects of music therapy on disruptive behaviour (SMD −0.42; 26 trials) and anxiety (SMD −0.20; 11 trials). They reported a positive trend for cognitive functioning, depression, and quality of life. However, substantial heterogeneity was observed due to music intervention, study design and measurement instruments used. The meta-analysis by Fusar-Poli et al. (2018) focused only on RCTs that examined the effects of music-based interventions delivered by music therapists on specific cognitive functions, such as attention, memory or language. Analysis of the six studies found no significant effects on all outcomes, however a subgroup analysis found positive effects of specifically active music therapy interventions on global cognition (SMD = 0.29; 3 RCTs). The most recent Cochrane review showed that music-based interventions reduced depressive symptoms (SMD −0.27; 11 RCTs) and overall behavioural problems (SMD −0.23; 14 RCTs) in people with dementia (van der Steen et al., 2018). Positive effects were also found for emotional well-being, quality of life and anxiety reduction, however, included studies were of (very) low quality. The review by Sousa et al. (2020) of music-based interventions for people with dementia in acute settings, found that all the nine included studies reported positive effects, however no overall effects could be calculated due to high methodological diversity, small sample sizes, considerable risk of bias, and lack of homogeneity of music-based interventions used. Finally, the network meta-analysis (NMA) by Dhippayom et al. (2022) of music-based interventions for addressing depression in older adults (not specifically for people with a dementia diagnosis), suggested that music therapist-delivered active music therapy was more effective (SMD −3.00; 15 RCTs), than music listening delivered by other healthcare professionals (SMD −2.06; 15 RCTs). Subgroup analyses showed that none of the selected music-based interventions had influential effects on depression in people with dementia.

Aims of the current review

Previous reviews highlight the potential effects of music interventions for a range of wellbeing outcomes. However, reviews often included a wide range of types of music-based interventions, study designs and/or outcomes, and were unable to reliably determine which types or combination of interventions are the most effective in effecting change across different outcomes. Key outstanding questions remain about the overall effects of music-based interventions on BPSD, especially the comparison between multiple types of interventions within a single analysis. Network meta-analysis (NMA) combines direct and indirect evidence across a network of treatments of a similar class or combinations thereof to compare their efficacy head-to-head or against control conditions. Dhippayom et al. (2022) recently conducted an NMA to compare the effects of different types of music-based interventions, but due to the lack of direct comparisons, it was only feasible to compare the different music-based interventions with Care as Usual (CAU). Further, the analysis was broad, focusing on depression in older adults, not specifically for those with dementia. Thus, to the best of our knowledge, NMA has not been used to compare different types of music-based interventions and their music-related components for people living with dementia. Given the potential of music-based interventions for managing and alleviating BPSD and improving wellbeing of people with dementia and their caregivers, as well as current lack of clarity around what type of music-based intervention or music-related component might be beneficial, this systematic review and network meta-analysis aims to:

1. Investigate the efficacy of music-based interventions on BPSD outcomes, as well as on outcomes related to cognition, quality of life and daily functioning in people with dementia;

2. Evaluate the quality of the available evidence;

3. Investigate potential effect moderators; and

4. Compare the efficacy of the common types of music-based interventions across outcomes.

Methods

This protocol follows the Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P, Shamseer et al., 2015) and the PRISMA extension for network meta-­analyses (Hutton et al., 2015). This protocol has been submitted and pending registration with the International Prospective Register of Systematic Reviews (PROSPERO) database, CRD42023457452.

Eligibility criteria

Eligible studies investigated the effects of music-based interventions on BPSDs, cognition, quality of life or everyday functioning against at least one other treatment in people with dementia. We will include randomized or non-randomized controlled trials. Observational and single-arm studies will be excluded. Cross-over trials will be included if results of the first phase are reported. There are no limitations on publication type, date, or language. For articles in any other language than English, a translator will be sought from the global research network of the authors involved. Unpublished data will be considered if available from authors.

Population

Eligible studies included adults living with dementia at baseline (of any age, stage, or aetiology), established through a clinical diagnosis or when baseline values are indicative of high prevalence of dementia in the study sample based on reported functional assessment scores. Studies that did not specifically target people with dementia or including a mixed population (e.g. a mix of dementia and mild cognitive impairment) will be included if at least 50% of the sample has a diagnosis of dementia or if outcomes for the dementia group can be obtained separately. Studies will be excluded if focused specifically on people with sensory impairment or psychiatric conditions (e.g. schizophrenia).

Interventions

Eligible studies include at least one group that received a music-based intervention targeting a specific a therapeutic outcome provided by (1) a music therapist in the context of music therapy, or (2) another healthcare professional who may need to use the skills of both musicians and therapists to appropriately select and apply the music, tailored to an individual’s needs and abilities. This is consistent with the most recent Cochrane Review on music-based interventions for people living with dementia (van der Steen et al., 2018).

Music-based interventions generally fall into two broad categories: receptive interventions (listening to pre-recorded or live performed music) and active interventions (singing and instrument playing). These interventions often combine multiple approaches, such as listening to music in parallel to singing. Studies where the music intervention was combined with other activities (e.g. dancing, exercise, art making) will be included only when music was the main therapeutic component of the intervention. Studies where music was used for non-therapeutic purposes, such as playing music while undertaking exercises during physical therapy or using music during daily activities such as meals or bathing, are excluded from this review.

Controls and comparators

Studies are included if they compared a music-based intervention with treatment/care as usual (TAU/CAU), active control (e.g. sham or social, physical or cognitive stimulation activities) or another music-based intervention condition. In multi-arm studies, all eligible arms will be included.

Outcomes

Eligible studies should report change from baseline to at least one follow-up on validated measures of global BPSD (e.g. the Neuropsychiatric Inventory), or specific symptoms (e.g. agitation, depression, apathy). Studies may also report on changes in global cognition, specific cognitive domains (e.g. learning, short-term memory), function (e.g. IADL, mobility), and quality of life and engagement. These additional secondary outcomes will be collected to examine the extent to which the putative efficacy of music interventions on BPSD has a flow-on effect on cognitive and everyday functioning, and whether specific interventions and their effects are related to engagement. Eligible outcome measures may include self-, proxy-, carer- or informant-report tools, as well as clinical, subjective, or objective assessments which are validated for the use in people with dementia.

Information sources and search strategy

We will search MEDLINE, EMBASE, PsycINFO, COCHRANE CENTRAL and CINAHL. The search strategy for MEDLINE via Ovid is provided in the Appendix. No database limits will be applied and articles in languages other than English will be translated. The database search will be complemented by manually searching the reference lists of tables of previous systematic reviews in the field. Authors will be contacted if published information is unavailable or insufficient to determine eligibility or extract outcome data.

Study selection

Search results will be combined and deduplicated, and then uploaded into Covidence for the study selection process. Screening of results by title and abstracts will be done based on single vote by one member of the review team and checked by a senior reviewer using a random selection procedure. The full-text versions of potentially eligible articles will be reviewed in duplicates by two independent reviewers, who will also contact authors when eligibility is unclear. Disagreements between reviewers will be resolved by a third reviewer. In cases where it is difficult to determine whether ‘music’ is the key component of the intervention or whether it would be ‘dance’ or ‘drawing’, for example, a third reviewer will be involved, and the final decision will be made through consensus. The reason for exclusion of papers will be recorded.

Data extraction

Outcome data

One reviewer will extract study information, characteristics, and outcome data independently into an Excel data extraction form. Results will be checked by a second reviewer and discrepancies between reviewers will be resolved by discussion. Authors of primary studies will be contacted for missing data.

The primary outcome will be change in overall BPSD (a composite of all BPSD domains) from baseline to post-intervention. Secondary outcomes will include specific BPSD domains, cognition, functioning, quality of life and engagement. BPSD is divided to 12 sub-domains by the Neuropsychiatric Inventory (NPI) including delusion, hallucinations, agitation or aggression, dysphoria or depression, anxiety, euphoria or elation, apathy or indifference, disinhibition, irritability or lability, aberrant motor activity, sleep and night-time behavioural disturbances, and appetite or eating abnormalities. When studies report more than one outcome measure per domain, all eligible outcome measures will be extracted.

Results of individual outcome measures will be extracted as mean and standard deviation (SD) for each group at each time point, mean change and SD, or other measures of between-group difference (e.g. mean difference, F for within-between difference). Outcomes will be extracted for all eligible intervention and control groups and coded into broad and specific domains. The time frame for evaluation of outcomes should be baseline and immediately after intervention. Outcomes in cross-over studies will be included if outcome data are provided before and after first phase of their study before switching, and for both groups. Intention-to-treat data will be preferred if reported.

Coding of study characteristics and potential effect moderators

In addition to outcome data, various factors with a potential moderating effect on the relation between music-based interventions and BPSD will be investigated. These factors will be coded and divided into study-, outcome-, participant-, intervention- and comparator characteristics, comparable to the methods of previous meta-analyses in the field of music-based interventions (de Witte et al., 2020; 2022).

• Study characteristics: first author, publication year, country, trial registration number, study design (RCT or Controlled Clinical Trial [CCT]), study setting (community, aged care, or clinical settings)

• Outcome characteristics: type of outcome (illness reduction outcome or positive health outcome), type of measure (observational/proxy, self-report, physiological), outcome subdomain (agitation, anxiety, depression, neuropsychiatric symptoms).

• Participant characteristics: weighted mean age, percentage of women, dementia severity and type (if reported), residential status, sample composition (dementia only vs majority with dementia)

• Intervention characteristics: type of music-based intervention/combination (receptive, active, or combination of both), specific intervention components (music listening, vocalization, instrument playing, movement to music), delivery format (group, individual or remote), facilitator (trained music therapist or not), intervention following a structured protocol (yes/no), session length (minutes), number of sessions, number of weeks, frequency of sessions (per week), total duration (in hours)

• Comparator: type of comparator; for comparator other than TAU: delivery format (group, individual or remote), session length (minute), number of sessions, number of weeks, frequency of sessions (per week), total duration (in hours)

Defining and coding the nodes for the network meta-analysis

Regarding the type of music-based intervention, we will code whether it was a receptive, active, or a combined music intervention. In receptive music interventions, people with dementia are not actively engaging in music-making, but rather respond to music provided by the facilitator/therapist, such as listening to live or pre-recorded music (Wheeler, 2015). Active music interventions involve the client doing something with the music during the sessions, such as instrument playing, composing music or songs, movement to music, or singing or vocalizing (Wigram, 2004). However, music-based interventions often involve a combination of both active and receptive methods (de Witte et al., 2022).

In addition to coding music interventions into these general intervention categories, we will also code several intervention components. This means that at least one or more components can be coded per study, which will eventually result in a new set of intervention nodes. Vocalization will be coded as an intervention component if voice is used within the intervention by singing, chanting, humming, making vocal sounds or improvisation. Instrument playing will be coded if the participants played on musical instruments, either jointly with others or alone, which could vary from small percussion such as a handshaker, to playing a piano or djembe. The movement to music component will be coded when participants were encouraged to move along with the music, such as hand clapping, rubbing, foot tapping, or dancing/moving to the beat of the music. Music listening will be coded when the participants are purely listening to music, individually or in a group, regardless of whether it is pre-recorded music or music played live by the therapist/facilitator.

Non-music comparisons will be coded as treatment-us-usual (TAU, i.e. no additional intervention is introduced), physical activity (guidance to engage in motion such as walking, yoga, stretching, or exercising), cognitive activity (stationary activities based on cognitive or learning engagement such as lectures, reading or practicing cognitive or behavioural strategies), or social activity (non-specific social interactions such regular chats session or cooking).

Assessment of risk of bias within individual studies

Two reviewers will independently evaluate risk of bias within studies using the Revised Cochrane Risk of Bias (RoB 2, Sterne et al., 2019) tool for randomized trials which covers six domains of bias: selection bias, performance bias, detection bias, attrition bias, reporting bias, and other bias. For non-randomized trials the Non-randomized Studies of Interventions (ROBINS-I, Sterne et al., 2016) tool will be used covering seven domains of bias. The first two domains address possible bias before the start of the intervention: bias due to confounding and in the participant selection. The third domain addresses possible bias in the classification of the interventions. The other four domains address post-intervention risk of bias due to deviations from intended interventions, missing data, measurement of outcomes, and selection of the reported result. Masking of participants will be assessed based on evidence of masking participants to the study hypothesis. Masking of therapists will not be assessed as this is not feasible in music intervention trials. Disagreements will be resolved with a senior reviewer (MTV or AL).

Statistical analyses

Pairwise meta-analysis

Data from all eligible outcomes will be converted to standardised mean difference (SMD), calculated as Hedges’ g with 95% confidence interval (CI). A positive SMD will indicate greater clinical benefit between intervention and comparison groups, regardless of the scale of the original measure. For example, a greater reduction in symptoms from baseline to post-training in the music intervention group over control will be coded as a positive SMD.

Multivariate pairwise meta-analyses will be conducted for each broad (e.g. overall BPSD) and narrow (e.g. BPSD sub-­domains) outcome. When studies report results for more than one eligible outcome measure per analysis, all eligible outcome measures will be included in the multivariate model. When studies report results for three or more arms, all eligible comparisons of music interventions and control will be included into a single unit of analysis except for comparisons of two music interventions. The multivariate pairwise analyses will account for dependency of effect sizes due to inclusion of multiple outcomes or comparisons.

Between-study heterogeneity will be measured using τ². The proportion of heterogeneity out of total observed variance will be expressed as I². Multivariate meta-regressions will be used to examine associations between study-level moderating factors and effect size as potential sources of heterogeneity. Analyses will be done using the packages metafor (Viechtbauer & Cheung, 2010), robumeta (Fisher et al., 2023) and clubSandwich (Pustejovsky, 2023) for R.

A forest plot will be generated following the recommendations by Fernández-Castilla et al. (2020) for specific use in multivariate meta-analyses to give a pictorial overview of all study results. This modified forest plot contains additional confidence intervals based on the sample variance of both individual observed effect sizes as well as the total number of effect sizes within studies, and thus provides information on the variability in effect sizes between studies and the relative contribution to the overall effect size estimate. Precision of effect estimates will be expressed using 95% CI. The range of predicted effects will be expressed using 95% prediction interval.

Network meta-analysis

A network meta-analysis based on a frequentist random-effects model will be used to estimate the comparative effects of each music-based intervention or combined interventions across outcomes. Analyses will be conducted for overall BPSD as well as individual domains for which sufficient evidence is available. TAU will be used as reference for estimations of effect sizes and P-scores (i.e. the relative probability of each intervention being the best option) for the main analyses. Relative SMDs and 95% CI for each pair of arms will be presented using a leagues table. All analyses will be conducted using the netmeta package for R (Balduzzi et al., 2023).

Subgroup analysis and investigation of heterogeneity

Where warranted, heterogeneity within NMA effect estimates will be investigated by meta-regression models using the study-level variables described previously. Similar to the planned meta-regressions of the pairwise comparisons, the aim of these analyses is to estimate whether design factors such as intervention settings, delivery and risk of bias may moderate the effect estimates.

Assessment of transitivity and inconsistency

Transitivity will be assessed by examining whether effect modifiers (design characteristics and risk of bias) differ systematically across different treatment comparisons. Distribution of modifiers will represent how various treatments are comparable across different trials. Inconsistency between direct and indirect estimates will be examined using node-splitting and formally tested using the back-calculation method.

Risk of bias across studies

Small-study effect (‘publication bias’) will be assessed for each outcome in the pairwise and the NMA models using funnel plots of SMD vs standard error. A multivariate adaptation of the Egger’s regression test for funnel plot asymmetry (Egger et al., 1997) will be performed when at least 10 studies are available for analysis. If evidence for asymmetry is detected in the pairwise analyses, we will use the trim-and-fill procedure (Duval & Tweedie, 2000) to estimate the extent to which small-study effect biases the effect estimates.

Assessment of the confidence in the evidence from network meta-analysis

CINeMA framework (Confidence in network meta-analysis) will evaluate the certainty of evidence due to network meta-analysis in six different domains, including within-study bias, reporting bias, indirectness, imprecision, heterogeneity and incoherence (Nikolakopoulou et al., 2020).

Discussion

Although an increasing number of empirical studies show that music-based interventions are promising non-pharmacological approaches for alleviating and treating BPSD, there are many different types of methods and interventions used. In addition, several music-related components can be distinguished in music-based interventions that, either by themselves or in combination with others, can influence the overall effect. Based on our knowledge, this may be the first SR and NMA study in the field of dementia care to compare the efficacy of key types of music-based interventions, as well as music-related components.

We are aware that there are more possible ways to categorise music-based interventions. However, the framework of the different types of music interventions and its components as created in our study protocol is based on both evidence-based literature and the extensive clinical expertise of involved experts. This, we believe, will overcome some of the limitations of prior studies that conflated music interventions with different characteristics. In addition, combined with our multivariate analysis approach, it will allow us to investigate possible factors as sources of heterogeneity and potential effect modifiers. As such, we aim to examine not only whether music-based interventions are efficacious in alleviating and treating BPSD, but also which key intervention types and intervention components, appear to be most promising in future trials and clinical practice.

The results of this NMA will not only sharpen future effectiveness studies but may also identify specific intervention factors that account for the ways in which therapeutic change occurs. This aligns with emerging change process research in the field of music-based interventions (de Witte et al., 2021; Hakvoort & Tönjes, 2023). This will enable health policymakers to make better-informed decisions about the type of music intervention needed within their own specific clinical context, resulting in a more effective and targeted use of music-based interventions. As a relatively low-cost, non-pharmacological approach without side effects, these interventions can effectively treat and alleviate BPSD in people with dementia.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by a McKenzie Postdoctoral Fellowship from the University of Melbourne (MdW), an International Postgraduate Scholarship from the University of Melbourne (LN), and a grant from the Australian Medical Research Future Fund (ID 2007411) (AW, NL, FAB, and AL).

Appendix

Search strategy (MEDLINE via Ovid)

1. dementia.mp. or exp Dementia/

2. exp Alzheimer Disease/or exp Amyloid beta-Peptides/or alzheimer$.mp.

3. exp Delirium/or deliri$.mp.

4. exp Cerebrovascular Disorders/

5. (cerebr$ adj2 (impair$ or declin$ or deficit$ or degenerat$ or deteriorat$ or los$ or disease$ ordisorder$ orcomplain$ or disturb$)).mp.

6. (mental adj2 (impair$ or declin$ or deficit$ or degenerat$ or deteriorat$ or los$ or complain$ ordisturb$)).mp.

7. ((cognit$ or neuro?cognit$) adj2 (impair$ or declin$ or deficit$ or degenerat$ or deteriorat$ orlos$ or disorder$ ordisease$ or complain$ or disturb$)).mp.

8. exp Parkinson Disease/or exp Parkinsonian Disorders/

9. (‘parkinson$ disease dementia’ or PDD or ‘parkinson$ dementia’).mp.

10. Homes for the Aged/or residential aged care.mp. or Health Services for the Aged/

11. nursing home$.mp. or exp Home Nursing/

12. institutionali$.mp.

13. Acute Care Setting.mp.

14. or/1-13

15. music$.mp.

16. Singing.mp.

17. singing/

18. choir.mp.

19. Music therapy/

20. choir/

21. Music/

22. Sensory Art Therapies/mt

23. Sensory Art Therapies/

24. Art Therapy/mt

25. Art Therapy/

26. Music/px

27. Improvis*.mp.

28. or/15-27

29. 15 and 28