Animal-assisted treatment in spinal cord injury rehabilitation: a scoping review

Abstract

Study design:

Scoping review.

Objectives

Animal-assisted treatment (AATx) has found application in the rehabilitation of individuals with neurophysiological conditions, such as spinal cord injury (SCI), to improve factors such as gait, respiratory function, and well-being. However, no overview of the scientific literature in this area currently exists. Therefore, this scoping review aimed to assemble all empirical studies concerning the use of AATx in the rehabilitation of individuals with SCI.

Methods

Seven databases (Embase, Medline, PsycINFO, Web of Science, Scopus, CENTRAL, and CINAHL) were systematically searched for relevant studies. Via double-screening, eligible studies were required to be peer-reviewed, report original data, include adult participants ( ≥ 18 years) living with SCI, and evaluate an AATx intervention in a rehabilitation context. Data were double-extracted from all included studies.

Results

From 257 unique records screened, 10 studies were included per our eligibility criteria. Results were categorized according to intervention type. Five studies evaluated canine-assisted treatment interventions, finding reductions in such metrics as muscle activity, pushrim kinetics, and cardiovascular outcomes, although pain severity and interference were not significantly improved. Five studies evaluated equine-assisted treatment interventions, finding improvements in such outcomes as spasticity, weight distribution, and ankle range-of-motion, though long-term effects were contested.

Conclusions

Preliminary findings suggest positive effects across a range of physiological and psychological outcomes, although some outcomes were not significantly improved or sustained. There is a need for more high-quality, longitudinal research and examples of AATx integrated into standard treatment in order to make recommendations for the use of AATx in SCI clinical practice.

Introduction

Despite advances in neurorehabilitation, many individuals continue to experience persistent functional limitations and secondary complications. Identifying innovative and motivating rehabilitation approaches therefore remains a clinical priority. One such emerging approach is animal-assisted treatment (AATx). In the current work, AATx is operationally defined as a goal-oriented, planned, and structured therapeutic intervention directed and/or delivered by health, education, or human service professionals, such as psychologists, physiotherapists, or occupational therapists [1]. This operationalization is guided by the 2025 position paper from the International Association of Human–Animal Interaction Organizations (IAHAIO), which associates AATx with explicit therapeutic goals, professional responsibility for treatment planning and delivery, and the structured integration of the animal to support these goals [2]. As such, AATx is often embedded into treatment programmes with a clear experimental set-up and measurable objectives [1, 3]. This is distinguished from animal-assisted support programmes (AASP), where emphasis is placed on interactions with animals purely to enhance well-being [1, 2, 4]. A recent review in the general adult population suggests that AATx has the potential to achieve multimodal benefits, improving psychological outcomes such as depression and anxiety, and physiological outcomes such as balance and gait [5]. When examining clinical subgroups, AATx has been associated with greater postural control in people with autism [6], improved mobility for individuals with cognitive impairment [7], and pain relief in pediatric populations [8]. On account of this versatility, AATx interventions have been used in a number of therapeutic contexts and for a wide range of clinical populations, including assisting individuals with neurological conditions [9]. For instance, structured, goal-directed horseback riding has been shown to stimulate human gait and support motor recovery in individuals post-stroke [10] as well as improve balance, fatigue, and quality of life for individuals with multiple sclerosis [11]. Equally, studies using dogs as therapeutic partners in AATx interventions have demonstrated improved pulmonary function post-stroke [12] and reduced stress-related hormones in persons with dementia [13]. Individuals with spinal cord injury (SCI) face similarly complex and multifaceted challenges during rehabilitation, impacting physical [14], psychological [15], and social domains [16]. Despite this, the use of AATx in SCI rehabilitation remains only an emerging field [17], and there is currently no systematic overview of the existing empirical studies to guide future developments within this field. In response to this knowledge gap, the present review aimed to compile, synthesize, and present the existing scientific literature on the use of AATx in SCI rehabilitation.

Methods

This scoping review is written in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) [18] Checklist and Explanation. The study protocol, reported in adherence to PRISMA-Protocols (PRISMA-P) [19], was pre-registered in Open Science Framework as of 11th November 2024 (DOI: 10.17605/OSF.IO/VWX3S). The search strategy, described below, is also reported according to PRISMA-S [20] guidelines.

Database searches

The database search strategy was structured around two blocks of search terms. The first block contained search terms associated with the population, i.e., individuals with spinal cord injury, while the second block contained search terms associated with the intervention, i.e., animal-assisted treatment. The search terms were developed in a continuous process involving consultation with clinical experts, preliminary literature searches, orientation in existing research, and examining indexed terminology within the selected databases: Embase (Ovid, 1947-present), Medline (Ovid, 1946-present), PsycInfo (Ovid, 1806- present), Web of Science (1900-present), Scopus (1823-present), CINAHL (1982-present), and Cochrane Central (1940-present). Database searches were conducted by SDW using a combination of free-text search terms and, where possible, subject headings. Proximity operators and truncation symbols (*) were used to capture multiple search term variants and derivations. Searches were configured to identify terms in either the title or abstract fields. No filters were applied to restrict language, design, population characteristics, or source type. The within-block searches were run using the Boolean Operator “OR”, and the across-block searches were run using the Boolean Operator “AND”. Pilot searches were conducted by SDW to refine and finalize the search terms and subject headings with feedback from and discussions with AOA and SLR. The final search was performed on 11th November 2024, and the final search string complete with individual database figures is provided as Supplement 1 in the supplementary materials.

Additional search strategy

To identify potential additional materials not identified by the database searches, we undertook two additional steps. First, additional searches were conducted in Google Scholar [21] by SDW, and in the Physiotherapy Evidence Database (PEDro) [22] by AOE, on 21st November 2024. Five combinations of search terms from both blocks (e.g., spinal cord injur* + animal assist*) were searched in Google Scholar, and the first 50 results sorted by relevance were screened according to the eligibility criteria. The search term “spinal cord injur*” was entered into PEDro to assess the relevance of all retrieved clinical trials. The additional searches were performed following the database searches and in a conservative, overinclusive manner. Potentially relevant studies found via additional searches were imported into a separate Covidence workspace [23] for independent double-screening at the full-text level, as described below. Secondly, all included studies in addition to content-relevant studies excluded on account of their format underwent forward and backward snowballing by SDW [24]. Forward snowballing, which is designed to identify materials referencing a given article, was performed using Google Scholar. Backward snowballing, which is designed to identify materials based on all cited materials within a given article, was performed manually by reviewing reference lists.

Eligibility criteria

To gain inclusion, studies needed to be written in English, Danish, Swedish, or Norwegian, corresponding to the languages that the author group could reliably interpret, and to report original data from a peer-reviewed academic journal as a full-text research article. The study sample had to consist of ≥50% of participants with SCI, or SCI data had to be reported separately. Participants were required to be aged 18 years or older, and the study needed to evaluate an AATx intervention in a rehabilitation context. To avoid definitional ambiguity, an AATx intervention was required to meet the characteristics described in the most recent IAHAIO Position Paper, including (i) a clearly articulated rehabilitative goal; (ii) a structured intervention performed within a rehabilitation context; and (iii) professional responsibility for intervention delivery or oversight by a health, education, or human service professional [1, 2]. Participants were allowed to be familiar with the animal(s) so long as they co-operated to achieve a goal-oriented, planned, and structured therapeutic intervention.

Screening procedure

All records found via database and additional searches were imported into separate Covidence [23] workspaces for screening. Duplicates were first automatically detected and removed during the import process and later checked by AOE. Additional duplicates observed during screening were removed manually. Screening was performed in two stages. First, the titles and abstracts were independently double screened by SDW, AOA, and SLR. Secondly, all studies forwarded from the title-and-abstract stage received independent double-screening of their associated full-text by the same authors. All disagreements pertaining to individual studies were resolved within the screening group. At the full-text screening stage, reasons for exclusion were documented. Information missing from the full text, which was considered essential to the decision-making process, was sought from corresponding authors.

Data extraction

Relevant data from all included studies were independently extracted by AOA and AOE into two standardized tables. SDW and AOA identified discrepancies and thoroughly rechecked data. As at the screening levels, all disagreements pertaining to individual studies were resolved within the screening group. The first table detailed the study’s authors, country of origin, sample characteristics, injury characteristics, setting, design, and relevant outcomes. The second table detailed the study’s intervention and key findings. Both tables are provided in the supplementary materials as Supplements 2 and 3. The Template for Intervention Description and Replication (TIDieR) checklist was used to systematically assess the completeness of intervention reporting for each included study [25]. Preceding data extraction, Google searches were performed to ensure that no errata, addendums, or other published information connected to the included studies would alter the interpretation of findings.

Synthesis methods

To summarize study flow, SDW inputted data from Covidence’s native PRISMA diagram into an external PRISMA-compliant web application [26]. This application visualizes the database findings, the incremental exclusion of records at both screening levels, and documents reasons for exclusion at the full-text level. Findings were synthesized by SDW using a structured, narrative approach informed directly by the data extraction tables (Supplements 2 and 3), containing information on sample characteristics, injury characteristics, setting, design, relevant outcomes, intervention characteristics, and key findings. Data reporting was limited to outcomes and statistical information available in the original studies. No retrospective calculation of standardized effect sizes was undertaken.

Results protocol violations

While the exclusion criteria dictate that only studies containing participants aged 18 years old or older may be considered for inclusion, Lechner et al. [27] reported data from a 16-year-old participant. However, per the JBI Manual for Evidence Synthesis [28], protocol violations are acceptable when clearly justified. Since the 16-year-old participant represented 1/34 of the published data, and since the individual’s data was in line with the group-level data, we did not believe that their inclusion meaningfully biased the results. As such, we decided to include Lechner et al. [27], as it otherwise met all pre-specified criteria. Therefore, the decision serves as a protocol violation.

Study flow

A total of 257 unique records were screened at the title-and-abstract level, of which 37 were advanced to full-text screening. At this level, records were excluded for their format, e.g., reviews, book chapters, conference abstracts (n = 16), relevancy of intervention (n = 5), language (n = 4), and relevancy of sample (n = 2). No further studies were identified via additional searches, leaving 10 relevant studies included for analysis. Figure 1 illustrates the study flow.

Fig. 1. PRISMA-compliant flow diagram of study selection.

The figure illustrates the identification, screening, and inclusion of studies in the scoping review, complete with reasons for exclusion at the full-text screening level.

Study characteristics

In terms of study design, three case studies [29–31], three quasi-experimental designs [32–34], two non-controlled pre-post studies [27, 35], one cross-over trial [36], and one randomized controlled trial (RCT) [37] featured. All 10 studies were published from 2003–2020. The number of participants per study ranged from 1-32 participants, with ages ranging from 16 to 72 years across the 8 studies reporting individual participant data. The remaining two studies reported mean ages of 40.4 [32] and 35.4 [37]. Fewer women (n = 23) than men (n = 84) participated in intervention groups. Regarding injury characteristics, several studies targeted the cervical (C3–C7) [27, 29, 31–37] and thoracic (T7–T12) [27, 30, 32–34, 36, 37] segments, though participants with sacral injuries were also represented [37]. Where reported, participants represented ASIA Impairment Scale (AIS) grades between A-D [27, 29, 32–36], with some studies recruiting participants exclusively with incomplete injuries [29, 31, 35] and others solely recruiting participants with complete injuries [30, 36]. Time since injury was highly heterogeneous, with participants both in sub-acute (e.g., mean of 8.8 months) [35] and chronic stages (e.g., mean of 14.2 years) [32]. Studies were predominantly conducted in North America (n = 7) [29–34, 37], though studies from Switzerland (n = 2) [27, 36] and South Korea (n = 1) [35] also featured. The vast majority of studies were led by physiotherapists (n = 7) [27, 30, 32–36] and occupational therapists (n = 5) [30, 32–34, 37], in addition to certified instructors (n = 2) [29, 31]. Outcomes were heterogeneous in nature, including biomechanical and physiological markers such as pushrim and upper limb kinetics, muscular activity, and task completion [30, 32–34], spasticity and balance [27, 35, 36], gait [31], and cardiorespiratory values [32]. Self-reported outcomes included pain, affect, well-being, and motivation [29, 36, 37]. One biological indicator, cortisol, was also assessed [37].

Narrative synthesis

Findings were organized by intervention modality, leading to the creation of two categories: (i) canine-assisted treatment interventions (n = 5) [30, 32–34, 37] and (ii) equine-assisted treatment interventions (n = 5) [27, 29, 31, 35, 36]. This grouping foregrounded conceptual clarity and supported a structured mapping of the available, empirical evidence across diverse study designs and outcome measures. Tables 1 and 2 summarize the findings and statistics from the canine-assisted treatment and equine-assisted treatment interventions respectively.

Table 1.

Summary of canine-assisted treatment interventions and results.

Author (Year)	Intervention	Results
Champagne et al. [32]	Mobility Assistance Dog facilitating wheelchair propulsion (630 m outdoor course)	• VO₂ ↓38% • VE ↓ 36% • VT ↓ 23% • RQ ↓ 9% • RR ↓ 21% • HR ↓ 10% • RPE ↓ 68% (4.92 → 1.58) • Time ↓32% (564 to 384 s). Recovery period also showed reductions in: • VO₂ (−28%) • VE (−28%) • VT (−25%) • RQ (−13%) • RR (−5%, ns) • (HR (−8%). • Large effect sizes (d = 0.93–2.00) for most measures, including oxygen consumption, ventilation, and time to complete the course, while heart rate yielded a moderate effect size (d = 0.71)
Gagnon et al. [30]	Wheelchair ramp ascent (8.5°) with vs. without Mobility Assistance Dog	• Handrim force ↓34.5% (83.5 to 54.7 N), max ↓46.1% (134.5 to 72.5 N). • Tangential force ↓26.3% (57 to 42 N), max ↓36.6% (103.5 to 65.6 N). • Efficiency ↑29.8% (47 to 61%). Joint kinetics: • Wrist ↓60.5% • Elbow ↓45% • Shoulder ↓36.9% Mean muscle activation reductions: • Biceps −38.5% • Triceps −59.5% • Deltoid −26.6% • Pectoralis −44.9%.
Martin-Lemoyne et al. [33]	Wheelchair ramp ascent (8.5°) with vs. without Mobility Assistance Dog	• Speed ↑38.3% (0.61 to 0.99 m/s, p = 0.037, d = 0.47) • Push phase time ↓45.4% (p = 0.013, d = 0.56). Forces: • Total, mean ↓37.3% (p = 0.007, d = 0.61) • Tangential, mean ↓34.4% (p = 0.013, d = 0.56). Shoulder moments: • Flexion ↓28.7% (p = 0.0.47, d = 0.45), • Adduction ↓41.7% (p = 0.017, d = 0.54), • Internal rotation ↓48.7% (p = 0.28, d = 0.50). Muscle activity: • Biceps mean ↓37.5% (p = 0.005, d = 0.56), • Anterior deltoid mean ↓14.7% (p = 0.013, d = 0.56), • Pectoralis mean ↓18% (p = 0.038, d = 0.49), • Triceps mean ↓ 31% (p = 0.139 ns). • Perceived effort ↓62.9% (VAS 3.4 to 1.3, p = 0.005). • Effect size: r = 0.45–0.63 (medium to large), favoring mobility assistance dog use across spatiotemporal, kinetic, joint moment, EMG, and perceived effort outcomes.
Martin-Lemoyne et al. [34]	Wheelchair propulsion (tiled vs. carpeted) with vs. without Mobility Assistance Dog	Time reduced: • Tiled ↓21.9% (p = 0.005, r = 0.62) • Carpet ↓30.3% (p = 0.01, r = 0.61). Muscle activity reductions: • Pectoralis (tiled mean ↓1.6%, tiled max ↓28.4%, tiled integrated ↓27.0%, carpet mean ↓43.3%, carpet max ↓37.8% carpet integrated ↓59.2%) • Anterior deltoid (tiled mean ↓18.9% ns, tiled max ↓8.8% ns, tiled integrated ↓54.8%, carpet mean ↓74.5%, carpet max ↓49.7%, integrated mean ↓92.4%) • Biceps (tiled mean ↓9.8% ns, tiled max ↓44.7%, tiled integrated ↓53.9%, carpet mean ↓9.8%, carpet max ↓44.7%, carpet integrated ↓57.6%) • Ticeps (tiled mean ↓14.2% ns, tiled max ↑4.4% ns, tiled integrated ↓45.7%, carpet mean ↓41.6%, carpet max ↓38.9%, carpet integrated ↓67.3%) • Effect size: r = 0.11–0.62 (small to large), with predominantly medium-to-large effects for task performance and shoulder muscle effort, favoring mobility assistance dog use, particularly on carpeted surfaces.
Thompkins et al. [37]	Canine-assisted therapy during OT (4×30 min sessions)	• PANAS ↓ significantly in AAT group (Session 1 mean −5.86, d = 0.89; Session 2 mean −2.29, d = 0.44) • Pain unpleasantness ↓ (6.0 → 4.93) vs. control ↑ (4.8 → 5.73), interaction p < 0.05. • Pain severity, positive affect, cortisol levels, BPI: ns. • Large effect size (d = 0.89) for the reduction of negative affect in session one; small effect size (d = 0.44) in session two.

↑ = increase(d).

↓ = decrease(d).

VO₂ = Oxygen consumption (measured in mL/kg/min).

VE = Ventilation (measured in L/min).

VT = Tidal volume (measured in L).

RQ = Respiratory quotient.

RR = Respiratory rate (measured in cycles/min).

HR = Heart rate (measured in beats per minute).

RPE = Rate of perceived exertion.

N = Newtons.

PANAS = Positive and Negative Affect Schedule (negative affect).

BPI = Brief Pain Inventory.

ns = not significant.

Table 2.

Summary of equine-assisted treatment interventions and results.

Author (Year)	Intervention	Results
Asselin et al. [29]	Therapeutic horseback riding (2 years)	• Improved core strength, posture, and balance • Decreased lower limb spasms • Regained awareness of legs • Increased confidence, trust, motivation for exercise, enjoyment of outdoors, • Social benefits
Choi et al. [35]	Hippotherapy (25–30 min sessions)	• No significant group changes in balance indices. Subgroup improvements: • C3/C4 injuries WDI ↑ (7.6 ± 2.9 to 8.9 ± 2.9); Brown-Sequard WDI ↑ (7.3 ± 3.3 to 9.2 ± 3.3).
Knight & Coffey [31]	Two 6-week horseback riding programmes (30–45 min sessions, weekly)	• Right hip angle at heel strike ↓ across sessions. • Stride length ↓ (0.91 to 0.83 m first session; 0.78 to 0.71 m second). Positive maintenance for hip/knee angles despite spasticity. Joint angles: • Left ankle +15° heel strike/toe-off (second session).
Lechner et al. [27]	Hippotherapy-K (25–30 min, avg. 11 sessions)	Ashworth scores: • 93% of sessions immediate ↓, 6% no change, 1% ↑. • Mean ASV significantly reduced (p < 0.001). • Greatest decreases in participants with baseline ASV > 38 (−10 points or more).
Lechner et al. [36]	Hippotherapy vs. Bobath roll, rocking seat, control (8 sessions, 4 weeks)	Ashworth scores: • Hippotherapy significantly ↓ vs. control (median −8.0 to +0.5). Self-reported pain: • Hippotherapy greater ↓ (VAS − 4.6 to −0.05 cm, p ≤ 0.05). Well-being: • Significant ↑ only with hippotherapy (p = 0.048). • No long-term effects ≥4 days post-intervention.

↑ = increase(d).

↓ = decrease(d).

WDI = Weight Distribution Index.

ASV = Ashworth value.

VAS = Visual Analog Scale.

Canine-assisted treatment interventions

Five studies examined canine-assisted treatment interventions [30, 32–34, 37]. In the case of a 26-year-old man sustaining a T7 SCI, ramp ascension with a mobility assistance dog (MAD), as opposed to without an MAD, reduced a number of pushrim and joint kinetics [30]. Increases in mechanical efficiency were also achieved in the MAD condition, while relative muscular utilization of the biceps, triceps, anterior deltoid, and pectoralis major were observed [30]. In ascending an 8.5° outdoor ramp, 10 participants with chronic SCI (C6–T11; AIS A, C, and D) registered faster speeds, reduced push phase time, and a prolonged recovery phase when aided by an MAD [33]. Significant reductions in upper limb effort were demonstrated, accompanied by decreases in all measured pushrim kinetics, and further reductions in anterior deltoid, biceps, triceps, and pectoralis major muscular activity [33]. Elsewhere, 13 chronic SCI participants (C5-T11; AIS A-D) demonstrated significant improvements in all measured cardiorespiratory outcomes, including rate of perceived exertion, oxygen consumption, and ventilation, while navigating a 630-meter outdoor course with the assistance of an MAD [32]. Significant reductions in all cardiorespiratory values were similarly observed in the recovery phase following MAD use, with all values, other than respiratory rate, demonstrating small to large effect sizes [32]. Participants with MAD assistance also took less time to complete the course [32]. Thirteen participants with SCI (C6-T11; AIS A, C, and D) attempted wheelchair propulsion on tiled and carpeted surfaces, with and without the assistance of an MAD [34]. In the MAD condition, the most pronounced mean reductions in electromyography (EMG) effort were observed on carpet, and were associated with the anterior deltoids, the pectoralis major, and the triceps. Effort in the biceps was also significantly reduced, though to a lesser extent [34]. During the tiled condition, all examined muscle groups demonstrated non-significant mean EMG reductions [34]. Mean time to complete the intervention was significantly reduced in both conditions when assisted by an MAD [34]. Finally, AATx combined with occupational therapy was shown to significantly decrease negative affect over two sessions for 31 participants with unspecified SCI characteristics, but no significant change in positive affect was evidenced [37]. Likewise, while pain unpleasantness significantly and consistently decreased across all four sessions, no significant differences were observed in pain severity or interference [37]. Compared to a standard treatment control group who demonstrated elevated cortisol stress levels, the MAD condition decreased stress levels, though findings were non-significant [37].

Equine-assisted treatment interventions

Five studies examined equine-assisted treatment interventions [27, 29, 31, 35, 36]. No significant changes in standing balance or stability were reported among 10 participants with cervical SCI (C3-C6; AIS D) following a hippotherapy intervention [35]. However, subgroup analyses showed significant improvements in weight distribution by participants with C3/C4 injuries [35]. Gait was assessed before and after two 6-week blocks of therapeutic horseback riding for a 33-year-old case participant with a C3-C4 SCI [31]. After the second block, the participant’s left ankle demonstrated an increased range of motion. While right hip angles decreased after the second block, range of motion was maintained during the riding intervention for most joints [31]. Nevertheless, stride length decreased from pre- to post-intervention in both 6-week blocks [31]. Thirty-two individuals with SCI (C4-T12; AIS A-D) participated in an average of 11 horse riding sessions, each lasting 25–30 min and joint-led by a trained equestrian and physiotherapist [27]. At group-level, pre-post spasticity reductions were significant, while at the individual level, Ashworth scores immediately reduced in 93% of all 351 sessions [27]. Individuals with severe pre-existing spasticity (Ashworth value > 38) demonstrated the most pronounced improvements [27]. A cross-over trial compared hippotherapy outcomes with two active interventions, (i) sitting astride a Bobath roll and (ii) sitting astride a rocking seat, in addition to a control condition [36]. Across 11 male participants with SCI (C7-T10; AIS A-B), only hippotherapy resulted in statistically significant reductions in spasticity outcomes [36]. Hippotherapy provided significantly greater reductions in self-rated spasticity than both active comparator conditions [36]. Only hippotherapy significantly improved well-being immediately post-session, though no effects on either spasticity or well-being persisting beyond a few days were observed for any condition [36]. A 44-year-old man sustaining C3-C4 SCI (AIS C) participated in a two-year therapeutic horseback riding programme [29]. Through verbal feedback, the participant reported a number of physical improvements, such as increased core strength, posture, and balance, reduced lower limb spasms, and greater awareness of his legs [29]. The participant also reported feeling motivated to resume exercise, more confident and trusting, and enriched by the social opportunities of competing in horse shows, interacting with fellow riders, and speaking at relevant conferences and events [29].

TIDieR assessment

All 10 studies provided clear and well-defined rationales based on neurophysiological or psychological recovery and contained highly detailed information on materials and procedural methods [27, 29–37]. All interventions were delivered face-to-face with the assistance of certified professionals (e.g., physiotherapists, occupational therapists, certified riding instructors). However, animal handling experience is detailed in only 6 studies [27, 29, 31, 34, 36, 37]. In all canine-assisted treatment interventions, dogs were trained either by the MIRA Foundation [30, 32–34] or Hand in Paw [37]. In only one equine-assisted treatment intervention were horses described as ‘specially trained’ [29]. Settings were divided between laboratory environments (for canine studies) [30, 32–34, 37] and community riding arenas (for equine studies) [27, 29, 31, 35, 36]. While some studies provided precise details regarding total number of sessions, frequency of sessions, intensity of sessions, etc. [36, 37], significant reporting gaps were observed in the remaining biomechanical studies [30, 33, 34], community-based case studies [29, 31], and clinical trials [27, 35], which compromises the repeatability of their protocols beyond the conditions of the immediate experimental task. Personalization was mentioned in half of the studies, often described as setting individualized goals [29, 37] or allowing for self-selected speeds during tasks [32–34]. Only one study explicitly addressed modifications, noting that handlers were prepared to stop sessions if dogs were stressed [37]. Information regarding fidelity and adherence represented the largest reporting gap. While technical data quality was often monitored (e.g., checking for EMG noise) [30, 33, 34], few studies described a formal plan to ensure the animal or handler adhered strictly to a treatment protocol, with one notable exception [37]. Full details of the TIDieR assessments are available as Supplementary Material 4.

Discussion

Summary of findings

Five studies reported canine-assisted treatment interventions [30, 32–34, 37], and five studies reported equine-assisted treatment interventions [27, 29, 31, 35, 36]. Muscular activity was measured and reduced in three of the five canine-assisted treatment interventions [30, 33, 34]. In particular, reductions in biceps, triceps, anterior deltoid, and pectoralis major activity were observed across all three studies [30, 33, 34]. Correspondingly, two studies observed reductions in pushrim kinetics, namely total and tangential force [30, 33]. Faster wheelchair propulsion and course completion speeds were also observed in three out of five canine-assisted treatment interventions [32–34]. Cardiovascular outcomes were also improved in the one canine-assisted treatment study in which they were measured [32]. While one canine-assisted treatment intervention reported mixed findings related to subjective assessments of pain and affect, measures of physiological stress (salivary cortisol) also yielded non-significant results [37]. Equine-assisted treatment was associated with reduced spasticity in three out of five studies [27, 29, 36], though some gains diminished over time [36]. Findings related to standing balance were mixed, with one study reporting perceived positive effects [29] and another quantitatively reporting no change [35]. Equally, equine-assisted treatment intervention resulted in both decreased stride length and increased range of motion [31]. Elsewhere, reports of improved weight distribution [35], social benefits [29], and motivational qualities [29] were associated with equine-assisted treatment.

Study findings in context

Our review of rehabilitative AATx interventions for individuals living with SCI only included canine-assisted and equine-assisted tasks. This may be explained by a number of factors. For instance, dogs are social animals that can read human emotional states [38], have been shown to reduce stress levels [39] and can easily be trained to perform specific tasks in many different therapeutic settings [40]. Equally, horses are sensitive to human body language and emotion [41], can support physical co-ordination and balance through riding [42], and, on account of their relative size, have been shown to engender self-confidence when a participant masters the interaction [43].

A recent review of AATx interventions in the wider neurological population [17] identified two trends. Firstly, different species were prioritized for different groups: while canine-assisted treatment predominated in dementia studies [13, 44–47], equine-assisted treatment predominated in post-stroke [10, 48–51] and multiple sclerosis groups [11, 52, 53]. As also reflected in our review, no animals besides dogs and horses were involved in AATx interventions in the wider neurological literature [17]. Secondly, and in conjunction, primary outcomes differed according to species and treatment group. Outcomes targeted using canine-assisted treatment for individuals with dementia were chiefly related to emotional and cognitive functioning, with a secondary focus on cortisol levels, balance, and motor activity [30, 32–34, 37]. Comparatively, the outcomes targeted using equine-assisted treatment in post-stroke and multiple sclerosis groups focused more directly on physical rehabilitation, in addition to fatigue, health perception, and mental health [27, 29, 31, 35, 36]. In the included SCI studies, equine-assisted treatment interventions were most commonly measured by physiological measures [27, 29, 31, 35, 36] as opposed to psychological or emotional [29, 36] outcomes. A similar pattern was observed in the included canine-assisted treatment interventions, with four studies reporting physiological outcomes [30, 32–34] and one reporting psychological or emotional outcomes [37].

Limitations

Firstly, inclusion was limited to studies published in English, Danish, Swedish, or Norwegian, corresponding to the languages that the author group could reliably interpret. Since four records were excluded due to language ineligibility, it remains unclear as to whether these records would have met the remaining eligibility criteria. Consequently, language-based eligibility may have limited the representation of studies from certain geographical regions. In a similar vein, the geographical scope of the included studies was comparatively homogenous, with most originating from North America (n = 7), thereby further restricting the generalizability of findings across different rehabilitation contexts and cultures. Secondly, while seven studies reported statistical significance (e.g., p-values) [27, 32–37], only four studies reported standardized effect size metrics (e.g., Cohen’s d, r) [32–34, 37]. Since the calculation of standardized effect sizes requires access to means and measures of variance or equivalent test statistics, which were not consistently reported across the remaining studies, and since p-values alone are insufficient for reliable effect size estimation [54], effect sizes could not be calculated across the included studies. While effect size synthesis is not required per the PRISMA-ScR reporting guidelines [18], we accept that in its absence, cross-study comparison is hindered. Additionally, it is acknowledged that the characteristics that distinguish AATx from AASP in the wider literature are subtle and that the two concepts can be conflated in common parlance. We attempted to overcome this issue by screening according to a pre-defined operational definition of AATx, informed by the most recent IAHAIO Position Paper, whereby the human-animal interaction served to achieve some specific, goal-directed intervention during rehabilitation, as opposed to being purely recreational or tied to well-being [1, 2]. This also meant that we included studies with MADs who were familiar to the participant, so long as they were involved in such an intervention. While we believe this created sufficient parameters to reliably perform our review, due to wider definitional ambiguity about what constitutes ‘rehabilitation’, there remains a risk that not all eligible studies were identified and included. Similarly, we acknowledge that AATx interventions require the guidance of a health, education, or human service professional. In line with our conservative approach to mapping the literature we included two studies led by certified trainers. While we considered their certification as evidence of relevant expertise, we recognize that a lack of specificity regarding the trainers’ educational background limits our ability to assess whether interventions adhered to standard practice. Equally, our search string compiles synonyms or terms broadly associated with animal-assisted treatment, canine-assisted treatment, and equine-assisted treatment. While we consider this approach appropriate given the preponderance of dogs and horses in AATx research, less frequently reported treatment-animal studies involving, for example, cats or birds, that were not explicitly defined as AATx or similar may not have been captured by the search. Lastly, the limited number of studies and heterogeneity in participant characteristics, outcomes, and intervention parameters precluded subgroup or stratified analyses.

Implications on future research and clinical practice

While a risk of bias analysis was not performed in the current review, the overall methodological quality of included studies was noticeably limited. Only one RCT [37] was included, and three of the 10 included studies (30%) were case reports [29–31], which consistently rank low within reputable evidence hierarchies [55, 56]. The inherent lack of randomization in quasi-experimental studies [32–34] and the lack of control groups in pre-post studies [27, 35] similarly limits our ability to draw firm causal conclusions. While over half of the included studies evaluated a comparator condition [30, 32–34, 36, 37], only one study compared AATx with standard treatment [37], leaving the extent to which AATx provides additional benefits beyond standard practice unclear. Future research should therefore include standard-care control arms to determine whether AATx provides benefits beyond standard practice. While the effects of AATx were generally positive in the included studies, many reported benefits were measured exclusively over the short-term. This risks the reporting of transient or short-lived effects rather than sustained improvement. For instance, the effects of equine-assisted treatment on spasticity and well-being were shown to deteriorate over time [36], while in a canine-assisted treatment study, improvements in negative affect were demonstrated in the first two of four sessions [37]. Additionally, many of the included studies examined physiological outcomes only during the assisted task itself [30, 32–34]. Future research should therefore consider time-series designs with periodical, repeated measurements when evaluating the effects of AATx interventions among participants with SCI. Equally, important aspects of AATx in SCI rehabilitation remain insufficiently studied, such as dose-response relationships (e.g., session frequency, duration, and intensity), differential effects across levels of neurological injury and completeness, and the ethical considerations of involving animals in structured interventions. To address these shortcomings, high-quality RCTs and/or single-case experimental design studies with the ability to report intervention dose parameters, stratified or subgroup analyses based on injury characteristics, and explicit monitoring and reporting of animal welfare outcomes are necessary for the successful and safe implementation of AATx within SCI clinical practice.

In subscribing to our pre-specified operational definition of AATx [3], studies describing the provision of a companion or support animal, without the animal acting as a partner in a goal-directed therapeutic intervention, were excluded from our review. However, it is important to acknowledge that companion animals can provide inherent psychological and emotional benefits outside of formal therapeutic settings [57, 58]. Greater research is needed to understand the experience of pet ownership in an SCI population [59], which can in turn inform structured AATx interventions. Lastly, research into animal-assisted interventions has primarily focused on human health outcomes, with relatively few publications addressing the welfare of the treatment animals involved. This represents a significant gap in the field, which is only beginning to recognize the critical role that animal welfare plays in ensuring the quality and integrity of the entire intervention process [60, 61].

Author contributions

SDW co-wrote the protocol, performed the literature searches, screened studies for eligibility, synthesized the data, co-wrote the introduction and methods, and is the main author of the results and discussion. AOA co-wrote the protocol, screened studies for eligibility, extracted studies, and provided continuous feedback on the manuscript. AOE co-wrote the introduction and methods, checked duplicates, conducted the additional searches, extracted studies, and provided continuous feedback on the manuscript. CG provided continuous feedback on the manuscript. SLR designed the overall project, secured funding, co-wrote the protocol, screened studies for eligibility, and provided continuous feedback on the manuscript.

Funding

These materials have received financial support from The Danish Victims Fund. The execution, content, and results of the materials are the sole responsibility of the authors. The analysis and viewpoints that have been made evident form the materials belong to the authors and do not necessarily reflect the views of The Council of The Danish Victims Fund.

Data availability

The data analyzed during this study can be found within the published article and its supplementary material.

Competing interests

The authors disclose no conflicts of interest related to this work.

Ethical approval

Ethical approval was not required for the production of this scoping review, which was conducted without participants or the retrieval of sensitive, personal data.

Supplementary information

The online version contains supplementary material available at 10.1038/s41393-026-01223-z.