Touch-Based Therapies in Dementia Care: A Systematic Review and Narrative Synthesis

Background

Dementia is a term that encompasses a number of neurodegenerative conditions including a range of dementia specific conditions such as Alzheimer’s Disease, Frontotemporal Dementia, Young Onset Dementia and other Cerebrovascular diseases. Across these conditions, individuals can experience cognitive impairment that is considered severe enough to impact on their ability to manage everyday activities. Dementia tends to occur in later life but can impact younger populations too. For the purposes of this review, an all-encompassing definition of dementia is adopted with all types and severity levels of dementia being considered. This inclusive adoption of the term is necessary due to the growing concerns of the global increase in dementia prevalence which is estimated to increase from 57.4 million in 2019 to 152.8 million in 2050 (Nichols et al., 2022).

Dementia is commonly associated with behavioural and psychological symptoms which further impact the quality of life and wellbeing of those living with the condition (Nelis et al., 2018) and their loved ones (Gilhooly et al., 2016). Recent research has found that the prevalence of depression, anxiety and apathy remain relatively consistent across people living with mild, moderate and severe levels of dementia (Leung et al., 2021) with prevalence rates ranging from 37-55% across different types of dementia, and up to 97% of individuals experiencing at least one Behavioural and Psychological Symptom of Dementia (BPSD) as the condition progresses (2020). BPSD is a debated term, however, it remains the most commonly used construct in published studies (Soofi, 2025; Warren, 2023). It is used here to maintain authenticity and alignment with the terminology applied in the evidence base, while noting its conceptual limitations.

Pain is increasingly recognised as a contributing factor to BPSD with research evidence that inadequate pain management in people with dementia exacerbates and may even cause behavioural symptoms (Achterberg et al., 2013). For example, research has found that 65.5% of residents living with dementia in aged care facilities in Australia experience pain and out of these, 48.8% experience moderate to severe pain (Atee et al., 2021). When comparing those who were experiencing pain to those who did not experience pain, Atee et al. (2021) found that the pain group were 3.8 times more likely to experience aggressive and agitated behaviours compared to the non-pain group, suggesting that pain may be a key contributor to these types of behaviours.

Touch based therapies (including massage, reflexology, shiatsu and therapeutic touch) have been widely used within acute settings across a range of conditions in the management of pain and symptoms (Adams et al., 2010). However, their effectiveness in the alleviation of pain and management of symptoms such as depression, irritability and anxiety among people with dementia is underexplored and there is a need to develop the evidence base for the efficacy of these intervention(s).

To date, there are a limited number of reviews that have explored touch-based therapies for pain. Some reviews, such as Saragih et al. (2023) and Bao and Landers (2021), have considered touch-based therapies’ under the umbrella term of non-pharmacological interventions. Saragih et al. (2023) found evidence for massage being effective in managing pain. However, questions remain regarding the specific types of massage that are effective, as well as how effectiveness is impacted by patients’ age and frequency of the intervention. Bao and Landers (2021) explored massage and ear acupressure and reported mixed findings suggesting that non-interactive non-pharmacological interventions were not as effective for managing pain compared to interactive types. However, it should be noted that there were only 2 studies related to massage included in the review and these studies had contradictory findings resulting in the lack of confidence in the findings. By focusing more broadly on non-pharmacological interventions rather than touch-based therapies more specifically, it is challenging to establish firm conclusions regarding touch-based therapies effectiveness due to their limited presence within these reviews. Therefore, more work is required to explore developments in this research area and to revisit the possibility of more confident conclusions regarding the management of pain for people with dementia using touch-based therapies.

Greater attention has been focused on exploring the effectiveness of touch-based therapies for behaviours associated with dementia, and anxiety, agitation and aggression (Watt et al., 2019; Wu et al., 2017). Watt et al. (2019) reported massage and touch therapy with and without music were more effective than usual care for managing aggression and agitation, however, the concern within this review was the high risk of bias seen across 46% of the studies. Similar challenges in drawing firm conclusions due to the limitations number of studies, samples sizes and quality of research in the area have also been reported by Wu et al. (2017) who examined the effectiveness of massage and touch. With the potential of touch-based therapies and limitations of previous reviews, it is necessary to consider how the research has developed in this area by conducting a more up to date and comprehensive review of the literature.

Rationale

The National Institute for Health and Care Excellence (NICE), and Scottish Intercollegiate Guidelines Network (SIGN) have increasingly emphasised the importance of non-pharmacological interventions in the management and support of people with dementia, particularly in the earlier stages of the disease (National Institute for Health and Care Excellence, 2018; Scottish intercollegiate guidelines Network, 2023Scottish Intercollegiate Guidelines Network, 2023). While pharmacological treatments are recommended for use to manage symptoms, NICE recommends that non-drug approaches be considered as a first-line treatment or alongside medication. NICE guidelines suggest a balanced approach incorporating pharmacological and non-pharmacological methods in the treatment of pain and behavioural symptoms in people with severe dementia (National Institute for Health and Care Excellence, 2018). Despite these recommendation, the evidence base for non-pharmacological interventions is relatively poor. This review aims to address this gap and synthesise the available evidence on the ongoing use of touch-based therapies within dementia care to strengthen the evidence base and clarify their role in in dementia care.

Review Objectives

The primary aim of this review is to assess the individual and overall efficacy of Touch Based Therapies in Dementia Care in the alleviation of Dementia Symptomology and Pain.

For dementia symptomatology, we summarised changes in symptom severity using validated measures such as the CMAI, NPI/NPI-NH, and Behave-AD. For pain, we summarised both the proportion of participants experiencing pain and any changes in pain severity over the course of the intervention.

A secondary aim of the review included providing a narrative synthesis on specific aspects of the intervention delivery to provide a blueprint for future research such as; Length, Frequency, and Duration of intervention; Dementia Stages (Early, Mid and Late); Product Type/Combined Therapies; Expertise/Level of Training of Practitioner. The authors initially planned a mixed-methods synthesis including meta-analysis, however, marked heterogeneity in interventions and outcomes and incomplete/non-comparable reporting precluded pooling, so the authors undertook a structured narrative synthesis.

As touch-based therapies are generally considered as low-risk interventions, there are no known typical areas of adverse event. However, to achieve a balanced perspective, the review adopted an exploratory approach when looking at adverse effects, aiming to review all recorded adverse events within returned articles and categorising them in relation to measured outcomes. i.e. increase in pain, increase in dementia symptomatology.

Eligibility Criteria

The authors applied a single set of inclusion/exclusion criteria covering population (people with dementia), intervention (touch-based therapies), comparators, outcomes, study design, and timeframe. For each outcome synthesis (agitation, pain), only studies reporting the relevant validated measures were summarised. A summary table of the eligibility criteria (Population, Intervention, Comparator, Outcomes, Study Design, Setting, Time frame, Language) is provided in Supplemental Table 1.

Population/Participant/s

Individuals were described as having a diagnosis of dementia. This was determined to be of any type and severity where the following sufficed as formal diagnosis of dementia; The international Statistical Classification of Diseases and Related Health Problems (ICD-10) (World Health Organisation, 1992); the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) (Lewis, 1994). Cognitive screening and professional clinical diagnosis of well validated assessment scales for cognitive function were also included such as; The Mini-Mental State Examination (MMSE) (Larner, 2012), and the Alzheimer’s Disease Assessment Scale-Cognitive Subscale (ADAS-Cog) (Kueper et al., 2018); Clock Drawing Test (Brodaty & Moore, 1997); Six-Item Cognitive Impairment Test (6-CIT) (Jefferies, 2012); general Practitioner Assessment of Cognition (GPCOG) (Brodaty et al., 2002); Test Your Memory (TYM) (Hancock & Larner, 2011); Montreal Cognitive Assessment (MoCA) (Nasreddine et al., 2005); Addenbrooke’s Cognitive Assessment- Revised (ACE-R) (Davies & Larner, 2012); Memory Impairment Screen (MIS) (Buschke et al., 1999).

All care settings were included both from within the community and long-term care facilities including care homes, nursing homes, residential homes, dementia specialist units, inclusive of international language.

Intervention Inclusion

The interventions that were included are listed below. A full breakdown of how each of these interventions is defined can be located in Appendix 1. All forms of massage (‘Swedish’, ‘Esalen’ ‘Lymphatic Drainage’ ‘Seated Massage’ ‘Sports Massage’ ‘Aromatherapy Massage’ ‘Indian Head Massage’ ‘Cranial Sacral Massage’ etc); Reflexology; Shiatsu; Accupressure; Bowen Therapy; Touch (Healing/Therapeutic).

The reason for including only these particular interventions is predominantly to do with governing bodies. The majority of the provided list with the exception of touch (healing/therapeutic), are considered as voluntary regulated since they do not have any statutory regulation. Other reasons include the need to physically place your hands upon the other person, and that the touch involved is part of a therapeutic intervention, and not that of in-the-moment care. In-the-moment care refers to spontaneous, informal physical contact that occurs naturally during routine interactions. For example, assisting with dressing or offering comfort, rather than touch that is delivered as a structured therapeutic intervention. Multi-component interventions were also included when touch formed a central therapeutic element, reflecting typical practice in dementia care (e.g., massage commonly delivered with aromatherapy). However, because these designs limit the ability to isolate the specific effect of touch, we interpreted their findings cautiously and reported when outcomes may reflect combined rather than touch-specific effects.

Database

With the support of a skilled academic librarian, the research team conducted a comprehensive database search and managed the retrieved articles using Zotero reference management software (Nikam, 2015).

CENTRAL (Cochrane Register of Controlled Trials); MEDLINE; Cummulative Index for Nursing and Allied Health Literature (CINAHL); PsycINFO; BioMed Central; Biosis Previews/Web of Science; British Medical Journal (BMJ Journals); Cambridge Core; Emerald; Health Source: Nursing Academic Edition; Taylor and Francis Online; Karger Journals; OVID; Oxford Academics; Sage Journals Online; Scopus; Wiley Online.

Some of these databases were grouped together based on the search engine, such as EBSCO Host. The choice to include these databases within the review was to ensure comprehensive coverage of the literature from diverse sources, enhancing the robustness of the findings (Bramer et al., 2017). Utilising both specialised and multidisciplinary databases, the review can minimise the risk of missing important evidence and provide a more complete and unbiased synthesis of available research on touch-based therapies in dementia care.

Search Strategy

The structure of the search strategy was fairly similar for each search where the authors wanted to capture as much literature as possible. To do this the authors followed a PIO format of Problem/Population, Intervention, and Outcome using varying words to both broaden and filter appropriate articles (Table 1).

OPEN IN VIEWER

Table 1.

Example PIO Search Strategy

	Problem/Population	Dementia ‘OR’ dem*	(inc. associated variation words such as Alzheimer’s, Lewy body, Cognitive Impairment etc)
‘AND’	Intervention	Massage ‘OR’ massage therapy	(inc. inclusion interventions and associated variations words such as reflexology/accupressure)
‘AND’	Outcome	Behavioural or psychological symptoms of dementia ‘OR’ BPSD	(inc. additional inclusion outcomes and variation words)

A full breakdown of the search strategy that was followed for each database search from our skilled academic librarian can be seen in Appendix 2.

Data Extraction and Management

A data extraction template was developed to extract all relevant information from each included study after screening. The extracted data included: Reviewer; Author(s); Objective(s); Sample characteristics; Method(s); Outcome(s)/Measure(s); Type of Therapy; Location of Intervention; Duration of Intervention; Country of Study; Summary of Results; and Critical Appraisal Skills Programme (CASP) Concerns. The CASP tool was used to support the data-extraction process, as it provided a structured and reliable method for assessing the quality and rigour of included studies and was attached to each included article within the Zotero software (Long et al., 2020). As there were three reviewers, the CASP checklists supported efforts to reduce bias, improve the accuracy of the findings, and enhance the overall reliability of the reviewers’ conclusions.

To begin the screening process, all records were exported into Zotero and underwent initial title/abstract screening by all three reviewers. Each record was independently double-screened using a traffic-light system (green = include, red = exclude, yellow = uncertain). The reviewing team met iteratively to discuss all yellow and discordant ratings, ensuring a shared interpretation of the eligibility criteria. During full-text screening, studies were divided among the three reviewers, and to maintain consistency, a second reviewer checked a proportion of each reviewer’s full-text decisions. Any disagreements were resolved through group discussion.

Data extraction was conducted using a standardised template covering study characteristics, intervention details, outcomes, and CASP concerns. CASP item-level assessments were then synthesised into an overall risk-of-bias rating: studies with all or nearly all criteria met and no major methodological concerns were classified as low risk; studies with some unclear or partially met criteria were classified as moderate risk; and studies with several unmet criteria or substantial methodological concerns were classified as high risk.

Study Selection: (PRISMA Diagram)

Study Characteristics

Concluding the screening process, 33 studies met the inclusion criteria and were incorporated into this review. Of these, 21 studies were focussed on the intervention of massage, 8 Acupressure, 3 touch modalities (i.e healing touch), 1 Reflexology, and 1 Shiatsu. Data is discussed as per these intervention listings as a way to compare across interventions (Table 2).

OPEN IN VIEWER

Table 2.

Literature Search Prisma

Summary of Evidence

Across 33 studies, touch-based approaches delivered as brief, repeated sessions were consistently associated with short-term calming, most notably reductions in agitation observed immediately post-session or across brief treatment courses. The clearest and most reproducible positive signals came from massage and acupressure, while findings for therapeutic/healing touch and reflexology/shiatsu were fewer and more variable. Effects on broader neuropsychiatric symptoms (e.g., NPI/NPI-NH) were mixed, and evidence for pain relief was limited but suggestive in a subset of protocols. Formal harms were rarely reported where monitored, no serious adverse events were identified. Minor/transient issues were infrequent, and acceptability was generally high, especially when sessions were short, delivered in a quiet space, and embedded in routine care by trained staff.

Methodological heterogeneity (intervention content and dose, comparators, outcome timing/measures) and mixed study quality (about half low risk of bias, a substantial minority moderate, and a small number high) precluded meta-analysis and limit precision. Nonetheless, converging patterns across low-/moderate-risk studies suggest that dose matters. Short sessions (5–20 min), delivered several times per week over 2–6 weeks, in a low-stimulus environment and with simple, standardised protocols, were the conditions most often linked to positive agitation outcomes.

Taken together, the preliminary evidence suggests pragmatic use as an adjunct to person-centred care while underscoring key uncertainties, durability beyond 4–8 weeks, optimal dose, and which subgroups benefit most. These signals directly inform the blueprint the authors propose. Clear, testable dose parameters (minutes per session, sessions per week, course length), standard comparators, core outcomes (e.g., CMAI/NPI; PAINAD where appropriate), blinded assessment where feasible, and planned follow-up to evaluate maintenance of effect.

Strengths and Limitations

This review synthesised 33 studies across massage, acupressure, therapeutic/healing touch, and reflexology/shiatsu, most conducted in real-world long-term care or inpatient settings, which enhances practical relevance. The author team also recognise that as the search concluded in February 2023, more recent studies may not be captured. Beyond direction of effect, the authors extracted and compared dose and delivery parameters such as minutes per session, sessions per week, course length, setting, who delivered, and basic fidelity cues, so the discussion could offer a trial-ready blueprint rather than a purely descriptive summary. Study quality was appraised with CASP, concerns were logged study-by-study, and our narrative weighting highlighted high-risk evidence while emphasising areas of convergence across low- and moderate-risk studies. The authors also collated the safety signal across modalities; where adverse events were explicitly monitored, no serious harms were reported and acceptability was generally high.

At the same time, substantial heterogeneity in intervention content and dose, comparators (usual care, attention/quiet presence, sham/light-touch), settings, and outcome timing/measures precluded meta-analysis and limits precision. Many studies were small, follow-up was short, and assessor blinding and fidelity reporting were inconsistent. These factors can inflate apparent benefits or obscure null effects. Agitation was frequently assessed (often with CMAI), but broader BPSD (e.g., NPI/NPI-NH domains), pain, and sleep/mood outcomes were inconsistently collected, reducing cross-study comparability and hampering any assessment of dose–response. Few studies examined durability beyond 4–8 weeks, so maintenance of effect remains uncertain. The preliminary evidence is concentrated in care homes/inpatient units, making generalisability to community-dwelling populations or specific dementia aetiologies/severities less clear. Furthermore, because a narrative “direction-of-effect” synthesis relies on reported statistics without pooling, the authors could not formally assess small-study or publication bias.

Across outcomes, the overall certainty of evidence was low to very low. For agitation, the certainty was judged to be low: several studies demonstrated short-term calming effects, but heterogeneity in protocols, small samples, and variable comparators reduced confidence in the consistency and precision of these findings. For broader BPSD outcomes, certainty was very low due to mixed findings, inconsistent outcome measures, and underpowered analyses. For pain, certainty was also very low, as pain was infrequently and inconsistently measured, and studies often lacked adequate controls, resulting in imprecise and highly uncertain estimates.

A final limitation relates to the use of the term BPSD, which we acknowledge is contested; although it remains the predominant terminology in the research literature, its use may inadvertently reinforce biomedical framings that overlook the communicative or contextual nature of behaviours (Soofi, 2025; Warren, 2023).

These strengths and constraints directly shape our blueprint. It is recommended that future trials should pre-specify dose (minutes per session, frequency, course length), standardise core outcomes (e.g., CMAI/NPI for agitation/BPSD; PAINAD where appropriate), use concealed allocation and blinded outcome assessment where feasible, include routine fidelity checks, extend follow-up to test durability, and incorporate economic evaluation. Taken together, these steps will help the field move from suggestive signals to robust, comparable evidence that can be implemented and scaled with confidence.

Comparison with other Reviews

Our findings are consistent with earlier syntheses of non-pharmacological approaches for agitation/BPSD in dementia. Touch-based modalities show immediate, short-term calming, but overall certainty is limited by small samples, heterogeneous protocols, and inconsistent follow-up. In particular, a network meta-analysis found massage/touch (and music) more effective than usual care for aggression/agitation, while noting a high risk of bias in nearly half the included studies (Watt et al., 2019). This pattern, a signal of benefit with methodological caveats, is echoed in other reviews focused on touch modalities (Bao & Landers, 2021; Saragih et al., 2023).

Where this review extends prior work is by (i) separating modalities (massage, acupressure, therapeutic/healing touch, reflexology/shiatsu) while explicitly extracting dose parameters (minutes/session, sessions/week, course length); (ii) foregrounding implementation details (who delivers, setting, fidelity/acceptability) often mentioned only briefly elsewhere; and (iii) translating these into a recommended trial blueprint with standard comparators and core outcomes (e.g., CMAI/NPI; PAINAD where appropriate). This adds a practical scaffold that earlier broad reviews could not provide because touch-based interventions were typically pooled with other sensory or activity-based approaches (Bao & Landers, 2021; Saragih et al., 2023; Watt et al., 2019).

For pain, broader non-pharmacological overviews report mixed findings for touch-based therapies. Massage sometimes appears beneficial, but evidence is sparse and heterogeneous; conclusions are cautious given few eligible trials and contradictory results across small samples (Bao & Landers, 2021; Saragih et al., 2023). These conclusions align with our interpretation that dose clarity, outcome standardisation, and longer follow-up are prerequisites for firmer inferences.

Finally, current guidelines endorse non-pharmacological approaches as first-line or co-interventions for behavioural symptoms in dementia, but highlight the need for better-designed trials, a gap which our blueprint aims to directly target (National Institute for Health and Care Excellence, 2018; Scottish Intercollegiate Guidelines Network, 2023).

It should be noted that with wider contextual issues such as COVID-19, recruitment and delivery of close-contact, touch-based interventions were substantially constrained during 2020–2021, particularly in care homes, due to infection-control restrictions and redeployment of research staff. In the UK, the NIHR paused set-up/recruitment for non-COVID studies in March 2020 and only later issued restart frameworks, while care homes limited visitor access and staff movement to reduce transmission (Department of Health and Social Care, 2020; National Institute for Health and Care Research, 2020a, National Institute for Health and Care Research, 2020b). This context likely contributed to the relative paucity of post-2019 trials in our corpus and may have biased the evidence base toward pre-pandemic, small, short-duration designs. Interpretation should consider that feasibility, acceptability, and implementation logistics observed pre-pandemic may differ under current practices.

Minimum Viable Package (Deliverable by Trained Staff)

The following session structures are derived from patterns observed across the included studies and are intended as pragmatic parameters to guide future feasibility work, rather than evidence-based practice recommendations.

• Hand/foot massage: 10–20 min, 3–5 times/week for 4–6 weeks; neutral oil; seated, low-stimulus environment.

Interventions should be delivered by trained care staff in massage/acupressure after a short skills session and return-demonstration. Use specialist practitioners where required for reflexology/shiatsu/therapeutic touch.

It is suggested that practitioners should use a quiet space, log minutes delivered, a brief pre/post agitation rating (e.g., CMAI-SF or NPI), and any refusals or adverse events. Re-evaluate at 4 weeks, continue, modify, or stop based on response and tolerance. Safety, patch-test oils, screen for fragile skin, wounds, or acute illness and pause if agitation escalates.