An umbrella review of the evidence for equine‐assisted interventions

Abstract

Objective

To establish the evidence base for animal‐assisted interventions, and specifically, hippotherapy and other interventions involving horses, in the context of improving biological, physical, psychological and social functioning, and outcomes in humans.

Method

A review of systematic reviews (an umbrella review) was carried out using a modified Joanna Briggs Institute methodology and following PRISMA guidelines. Inclusion criteria were any interventions that included a live horse, for humans of any age and for any therapeutic purpose. Service and assistance horses and people who were incarcerated were excluded. Systematic reviews with a clearly articulated and comprehensive search strategy, with evidence of critical appraisal, both qualitative and quantitative and mixed methods, and in English were considered.

Results

Of 19 systematic reviews considered eligible for inclusion, 13 were included in this umbrella review. These 13 systematic reviews covered 79 individual studies of equine‐assisted interventions, including primarily hippotherapy and therapeutic horse riding. Participants included children with cerebral palsy, children and adolescents with autism spectrum disorder, adults with multiple sclerosis, elderly people, adults post‐stroke, people with spinal cord injury, adults with serious mental illness, adults with balance problems, people with brain disorders, breast cancer survivors, and obese women.

Conclusions

The evidence for equine‐assisted interventions for adults and children across a range of conditions and presentations is equivocal. The current evidence base is marred by multiple methodological weaknesses and thus, therapeutic interventions that include a horse cannot be asserted as best practice at this time. Rigorous research is indicated to determine the utility of equine‐assisted interventions.

Key words:

• animal‐assisted interventions
• equine‐assisted interventions
• hippotherapy
• systematic reviews
• umbrella review

WHAT IS ALREADY KNOWN ABOUT THIS TOPIC

• Animal‐assisted interventions have grown in popularity over the last decade.

• There are numerous claims that animal‐assisted therapies are beneficial for human physical, psychological, and social health.

• Equine‐assisted interventions, which involve a horse, are one particular type of animal therapy that is growing in recognition.

WHAT THIS TOPIC ADDS

• There are a number of studies that evaluate equine‐assisted interventions for human health: all of them have methodological flaws.

• This is the first umbrella review on equine‐assisted interventions which demonstrates that these methodological flaws compromise the evidence.

• Based on the current evidence, equine‐assisted interventions cannot be advocated as best practice.

INTRODUCTION

The establishment of the term “human–animal bond” in the late 1970s signified the recognition of the importance of the potential therapeutic role animals may play in human health and well‐being (Hines, 2003). Since then interventions with animals in the context of health and social care have become common practice in many countries (Fine, 2015). The term “animal‐assisted interventions” (AAIs) is now an accepted industry term, and a focus of human–animal research in the field known as “anthrozoology,” with 2017 marking the 30th anniversary of “Anthrozoös,” the Journal of the International Society for Anthrozoology (http://www.isaz.net/isaz/anthrozoos/). AAIs are defined as: … goal‐oriented and structured interventions that intentionally incorporate animals in health, education, and human service for the purpose of therapeutic gains and improved health and wellness. Animal‐assisted therapy (AAT), animal‐assisted education (AAE) and animal‐assisted activities (AAAs) are all forms of animal‐assisted interventions. In all these interventions, the animal may be part of a volunteer therapy animal team working under the direction of a professional or an animal that belongs to the professional himself. (https://petpartners.org/learn/terminology/)

Due to increase in both the popularity and general awareness of AAIs, a number of studies have been conducted across a range of settings and populations and including a variety of different animals, including horses. In line with the biopsychosocial model (Engel, 1980), AAI's have been assessed in relation to biological, psychological, and social outcome variables (see e.g., Friedmann & Son, 2009). The majority of empirical studies are quantitative in nature although numerous qualitative studies have also been undertaken (see e.g., Shen, Xiong, Chou, & Hall, 2018). Subsequently, there has been some attempt at synthesising this evidence, with a number of systematic reviews conducted over the last 10–15-years. Results are mixed with most suggested limited beneficial effects as well as an acknowledgement in the literature of the methodological weaknesses common in this field of research (Chur‐Hansen, Stern, & Winefield, 2010; Herzog, 2015; Stern & Chur‐Hansen, 2013). Since there are a number of systematic reviews that have been conducted in the area that either focus on specific patient populations and settings or across them, an umbrella review is warranted to bring together and compare the results of these systematic reviews to determine whether AAIs which involve horses are effective across different populations and settings. To date, no such umbrella review for AAIs with a focus on horses exists. Two umbrella reviews have considered the best available intervention evidence for children with cerebral palsy in which hippotherapy was but one of a range of diverse interventions considered (Anttila, Suoranta, Malmivaara, Mäkelä, & Autti‐Rämö, 2008; Novak et al., 2013). Similarly, Abraha et al. (2017) conducted an umbrella review investigating non‐pharmacological interventions to treat behavioural disturbances in older patients with dementia which included one systematic review on AAIs, however, this included dogs, cats, and fish as well as robotic and toy animals, but not horses. This study aims to address a gap in the literature, by reviewing the reviews of equine‐assisted interventions (EAIs) to ascertain the levels of evidence for this AAI on the biological, psychological and social health and well‐being of children, adults and older people.

Equine‐assisted interventions

The popularity of and interest in EAIs and hippotherapy is perhaps best illustrated by an examination of the Table of Contents of a key text in the field, “Handbook on Animal‐Assisted Therapy. Theoretical Foundations and Guidelines for Practice”, edited by Fine, and now in its 4th edition (2015). In the 1st edition, published in 2000, “horses,” “hippotherapy,” “riding therapy,” and “vaulting” were covered in pages 93–94 and 217–218. By Edition 4, an entire 23‐page chapter was devoted to the role of the equine in AAT (Latella & Abrams, 2015).

A study of the literature shows diverse terminology for the use of horses interacting with humans for therapeutic purposes (Latella & Abrams, 2015). Granger and Kogan (2006) explained that horses in a therapeutic setting can be divided into the categories of hippotherapy, riding therapy, riding for rehabilitation and vaulting. Equine‐assisted psychotherapy is a further category of therapy, which by definition must include a credentialed mental health‐care professional and involve activities with a horse such as riding, grooming, or handling, or other related actions (Kruger & Serpell, 2006). Equine‐facilitated psychotherapy similarly involves a mental health practitioner and an equine specialist (Lee, Dakin, & McLure, 2016).

Hippotherapy is often utilised for physical therapy, for instance to improve balance, posture and gait. Such treatment is delivered by health‐care professionals including occupational, physical, or speech therapists (Latella & Abrams, 2015). Hippotherapy requires the horse to lead, while in riding therapy, the rider can be more active and take control, in order to improve flexibility or gross motor control, for example. In riding for rehabilitation, the rider similarly takes more control, but in addition, treatment aims to address psychological and social problems as well as physical ones. Vaulting is a variation of riding therapy, which requires the rider to perform gymnastic exercises on horseback, with an aim to improve muscle strength, mobility and self‐esteem as well as a number of other physical and psychological outcomes. Latella and Abrams (2015) describe a further category of equine therapy, “therapeutic carriage driving,” which allows people to connect with a horse or pony as an alternative to riding, which may be particularly suitable for children and people who require a wheelchair. Furthermore, not all interventions include an actual horse, with simulated horseback riding and robotic horses also incorporated into therapeutic interventions (Lee, So, Lee, & Moon, 2018).

Biopsychosocial applications

A number of studies have been conducted to assess the efficacy of hippotherapy and therapeutic horse riding on biological and physical parameters, for example, for improving muscle symmetry and tone for children with cerebral palsy (Sterba, 2007). At the other end of the lifespan, a systematic review focused on the benefits of hippotherapy and horse‐riding simulation for healthy older adults for outcomes including improved balance, mobility, gait, muscle strength, and physical fitness (Hilliere, Collado‐Mateo, Villafaina, Duque‐Fonseca, & Parraça, 2018). A recent Australian review and meta‐analysis considered hippotherapy for adults with acquired brain injury to improve motor function (Marquez, Weerasekara, & Chambers, 2018).

With a focus on psychological health Lee et al. (2016) note that the incorporation of horses to treat mental illness is an innovative and novel approach, with limited but promising evidence for the treatment of adolescents. Selby and Smith‐Osborne (2013) adopted a biopsychosocial approach in their review of the effectiveness of complementary and adjunct therapies and interventions involving equines for chronic illness and health challenges. Research on social functioning and hippotherapy tends to focus on children with autism spectrum disorder (ASD), for example, a recent systematic review considered the effectiveness of equine‐based therapy in the treatment of social and behavioural aspects of children with ASD (Wiese, Simpson, & Kumar, 2016).

Focus of this review

This umbrella review examines the use of AAI's involving horses, “equine‐assisted interventions” or EAIs, on the health and well‐being of children, adults and older people by synthesising the results of previous systematic reviews. The review questions are: “What is the effectiveness of EAIs on the health and well‐being outcomes of children, adults and older people?” and “What are the experiences of children, adults and older people receiving EAIs?” This umbrella review aims to establish the current evidence base, highlight strengths and weaknesses of the methods used for gathering data to evaluate claims regarding EAIs, and offer suggestions regarding future research and practice in the field.

METHODS

This review followed the Joanna Briggs Institute (JBI) methodology for conducting umbrella reviews (Aromataris et al., 2017), however to accommodate the diversity of interventions, outcomes and populations included in this review, a modified approach to how the data are presented has been utilised (i.e., a narrative synthesis is provided instead of the prescribed tabulated findings and summary of evidence stop‐light indicator system). The protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO) on July 19, 2018 (registration number CRD42018100796). It follows the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) guidance for the conduct and reporting of systematic reviews (Moher, Liberati, Tetzlaff, & Altman, 2009).

Inclusion criteria

Participants

This umbrella review considered any reviews that included people of all ages receiving AAIs which involved a live horse (i.e., EAIs). For the purpose of the review children were defined as those up to 17-years of age, adults aged 18–59-years, and older people as 60-years and over. Participants were receiving EAIs for any specific health condition or diagnosis (e.g., Dementia, ASD) or for no specific health condition or diagnosis (e.g., for loneliness, mood).

Intervention(s)

Reviews were considered that evaluated AAIs as defined above and including both AAT and AAA. For this review AAT was defined as: “a goal‐oriented, planned, structured and documented therapeutic intervention directed by health and human service providers as part of their profession.” (https://petpartners.org/learn/terminology/) AAA was defined as activities that “provide opportunities for motivational, educational, and/or recreational benefits to enhance quality of life. While more informal in nature, these activities are delivered by a specially trained professional, paraprofessional, and/or volunteer, in partnership with an animal that meets specific criteria for suitability.” (https://petpartners.org/learn/terminology/)

While the initial search included interventions involving any animal or combination of animals delivered by trainers and handlers, only horses are the subject of this review. There was no limit in frequency or duration of AAIs and they could be delivered in a group format or individually. Animal‐assisted education (as defined by Pet Partners, see https://petpartners.org/learn/terminology/) was not included. Interventions could be compared with other combinations of AAIs, alternative therapies (e.g., art therapy, robotic animals, simulators), or no treatment. Reviews focusing on people with service/assistance animals or people who were incarcerated were excluded.

Phenomena of interest

The phenomenon of interest was the experiences (both positive and negative) of those receiving AAIs involving horses for any purpose.

Outcomes

The outcomes of interest were any outcome related to health and well‐being. For the purpose of this review outcomes are grouped under the biopsychosocial model (Engel, 1980): biological (e.g., blood pressure, heart rate, mobility), psychological (e.g., depression, anxiety, quality of life, cognitive status), and social (e.g., social functioning, social interaction). Outcomes could be measured using any validated instrument, via observation or by self‐report, and measured during or immediately after the intervention or at a follow‐up period.

Context

There were no limits to the geographical or care setting of the intervention and could include (but were not limited to): acute care hospitals, residential care facilities, rehabilitation facilities, and schools and mental health facilities.

Study types

This umbrella review included systematic reviews with or without meta‐analysis. Systematic reviews that met the following criteria were considered for inclusion: containing a clearly articulated and comprehensive search strategy using multiple databases and having evidence of critical appraisal/assessment of risk of bias. Reviews could be quantitative or qualitative in nature or both (in the case of comprehensive or mixed methods reviews) as long as the quantitative or qualitative component could be extracted separately.

Only reviews published in English were included. A lower date limit of 1990 was set as per the guidance provided by JBI since there were very few systematic reviews published prior to that time (Aromataris et al., 2017).

Search strategy

The search strategy aimed to locate both published and unpublished systematic reviews. The search strategy was developed in consultation with a librarian and consisted of a combination of keywords and index terms which were customised for each database/resource (Table 1). The databases searched were: The JBI Database of Systematic Reviews and Implementation Reviews; The Cochrane Database of Systematic Reviews; The Campbell Collaboration Online Library; MEDLINE (via OVID); CINAHL with full text (via EBSCOhost); PsycINFO (via OVID); EMBASE (via ELSEVIER); and Web of Science. The registers searched were: PROSPERO and the Centre for Reviews and Dissemination. The search for unpublished reviews included: Proquest Dissertations and Theses Global; MedNar; and Google Scholar. The search was conducted between July and August 2018.

Table 1. Search strategy. CINAHL: search ran August 31, 2018

Search ID	Search terms
S1	(MH “Pet Therapy”)
S2	TI Pet Therap* OR AB Pet Therap*
S3	TI Pet assisted OR AB Pet assisted
S4	TI Pet facilitated OR AB Pet facilitated
S5	(MH “Animal Assisted Therapy (Iowa NIC)”)
S6	TI animal assisted OR AB animal assisted
S7	(MH “Equine‐Assisted Therapy”)
S8	TI Equine Assisted OR AB Equine Assisted
S9	TI equine facilitated OR AB equine facilitated
S10	TI hippotherap* OR AB hippotherap*
S11	TI horseback n3 therap* OR AB horseback n3 therap*
S12	TI dog assisted OR AB dog assisted
S13	TI canine assisted OR AB canine assisted
S14	TI animal facilitated OR AB animal facilitated
S15	(MH “Systematic Review”)
S16	TI Systematic Review OR AB Systematic Review
S17	(MH “Meta Analysis”)
S18	TI Meta Analysis OR AB Meta Analysis
S19	TI metaanalysis OR AB metaanalysis
S20	TI meta synthesis OR AB meta synthesis
S21	(MH “Meta Synthesis”)
S22	TI metasynthesis OR AB metasynthesis
S23	TI qualitative n5 synthesis OR AB qualitative n5 synthesis
S24	S1 OR S2 OR S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14
S25	S15 OR S16 OR S17 OR S18 OR S19 OR S20 OR S21 OR S22 OR S23
S26	S24 AND S25
S27	S24 AND S25 Limiters—Published Date: 19900101‐20180731

Study selection

Following the search, all identified citations were collated and uploaded into Endnote X8 and duplicates were removed. Titles and abstracts were screened independently by the two authors for assessment against the inclusion criteria for the review. Reviews that met the inclusion criteria were retrieved in full and assessed in detail against the inclusion criteria. Full‐text articles that did not meet the inclusion criteria were excluded and reasons for exclusion are provided in Supporting Information Table S1. Any disagreements that arose between the reviewers were resolved through discussion.

Assessment of methodological quality

Selected reviews were critically appraised by the two authors for methodological quality using the standardised critical appraisal instrument from JBI (Aromataris et al., 2017). The decision was made a priori to only include articles that scored “yes” for the following questions: Q1 (Is the review question clearly and explicitly stated?), Q2 (Were the inclusion criteria appropriate for the review question?), Q4 (Were the sources and resources used to search for studies adequate?), Q5 (Were the criteria for appraising studies appropriate?), and Q8 (Were the methods used to combine studies appropriate?) (Stern & Chur‐Hansen, 2018). Any disagreements that arose were resolved through discussion.

Data extraction

Data were extracted from articles included in the review using the standardised data extraction tool from the JBI (Aromataris et al., 2017). The data extracted included specific details about the interventions, phenomena of interest, populations, study methods, and outcomes of significance to the review questions.

Data synthesis

For quantitative reviews extracted findings are presented narratively and in tabular format for each intervention, population, and outcome. Due to extensive variation across each of these elements, results are not presented using the traffic light approach as recommend by JBI and as outlined in the protocol (Stern & Chur‐Hansen, 2018).

Similarly, for qualitative systematic reviews the findings are presented narratively and in tabular format instead of the traffic light approach due to the inclusion of only one review containing qualitative evidence and the way in which the results were presented. Any overlap of original research studies in each of the included review is also presented.

RESULTS

Study inclusion

Due to a larger body of work being undertaken by the review team the search was broad and encompassed interventions involving any animal or combination of animals delivered by trainers and handlers. The search identified a total of 644 potentially relevant citations. Following the removal of duplicates (n = 158) 486 were assessed by title and abstract for relevance. There were 337 records excluded, of those 10 full‐text articles could not be located or obtained and 27 authors were contacted to determine whether their review had been completed and published; seven did not respond and the remaining authors advised that their review was either incomplete or at varying stages of the publication process. The full‐text of 149 articles was examined and 130 did not meet inclusion criteria leaving 19 articles that were appraised for methodological quality. The majority of articles excluded at full‐text did not meet our definition of a systematic review, largely because no critical appraisal took place but also due to the search only including one database, the article indicating the study was a primary study or a protocol or the description of the methods signifying the article was a literature review or discussion piece. The PRISMA flowchart provided in Figure 1 outlines the search and inclusion process.

short-legendFigure 1.

Methodological quality of included studies

Nineteen reviews were assessed by two independent reviewers for methodological quality. The total number of “yes” responses varied from 5 to 10 out of a possible 11, with 14 of the 19 articles scoring at least 8. Nearly all of the studies scored poorly in regard to publication bias not being assessed (Q9) while just under half reported on methods to minimise errors in data extraction (Q7). Under half of the reviews did not clearly mention if appraisal had been conducted by at least two reviewers independently (Q6), however, all but one outlined appropriate research recommendations (Q11). All articles appraised included adequate sources and resources for searching (Q4), however, only 75% of reviews had appropriate search strategies outlined (Q3). Recommendations for policy and/or practice were supported by data in just over three‐quarters of the reviews assessed.

Since five of the questions on the checklist were deemed essential for inclusion, six reviews were subsequently excluded, leaving a total of 13 reviews included in this umbrella review (see Table 2 for appraisal results).

Table 2. Results of the critical appraisal of eligible studies using the JBI checklist for systematic reviews and research syntheses

Review	Q1	Q2	Q3	Q4	Q5	Q6	Q7	Q8	Q9	Q10	Q11
Bremer, Crozier, and Lloyd (2016)	Y	Y	N	Y	Y	Y	Y	Y	N	Y	Y
Bronson, Brewerton, Ong, Palanca, and Sullivan (2010)	Y	Y	Y	Y	Y	Y	U	Y	N	N	Y
Dewar, Love, and Johnston (2014)	Y	Y	Y	Y	Y	U	U	Y	N	Y	Y
Francis (2007)	Y	Y	Y	Y	Y	Y	U	Y	N	U	Y
Hilliere et al. (2018)	N	U	Y	Y	Y	U	Y	U	N	Y	Y
Jormfeldt and Carlsson (2018)	Y	Y	Y	Y	Y	U	U	Y	N	Y	Y
Marquez et al. (2018)	Y	Y	Y	Y	Y	Y	Y	N	N	Y	Y
Martín‐Valero, Vega‐Ballón, and Perez‐Cabezas (2018)	Y	Y	U	Y	Y	U	U	Y	U	U	U
Rampling et al. (2016)	Y	Y	Y	Y	Y	N	Y	Y	N	Y	Y
Selby and Smith‐Osborne (2013)	Y	U	N	Y	Y	U	Y	N	Y	Y	Y
Snider, Korner‐Bitensky, Kammann, Warner, and Saleh (2007)	Y	Y	Y	Y	Y	Y	U	N	N	Y	Y
Srinivasan, Cavagnino, and Bhat (2018)	Y	Y	Y	Y	Y	Y	Y	Y	N	Y	Y
Sterba (2007)	Y	U	Y	Y	N	U	U	U	N	Y	Y
Stergiou et al. (2017)	Y	Y	Y	Y	Y	Y	Y	Y	N	Y	Y
Tseng, Chen, and Tam (2013)	Y	Y	Y	Y	Y	Y	U	Y	N	Y	Y
Whalen and Case‐Smith (2012)	Y	Y	U	Y	Y	Y	U	Y	N	Y	Y
White‐Lewis, Russell, Johnson, Cheng, and McClain (2017)	Y	Y	Y	Y	Y	Y	U	Y	N	Y	Y
Wonsetler and Bowden (2017)	Y	Y	Y	Y	Y	U	Y	Y	N	Y	Y
Zadnikar and Kastrin (2011)	Y	U	Y	Y	N	Y	Y	U	Y	N	Y

Note. Y: Yes; N: No; U: Unclear.

Q1. Is the review question clearly and explicitly stated?

Q2. Were the inclusion criteria appropriate for the review question?

Q3. Was the search strategy appropriate?

Q4. Were the sources and resources used to search for studies adequate?

Q5. Were the criteria for appraising studies appropriate?

Q6. Was critical appraisal conducted by two or more reviewers independently?

Q7. Were there methods to minimise errors in data extraction?

Q8. Were the methods used to combine studies appropriate?

Q9. Was the likelihood of publication bias assessed?

Q10. Were recommendations for policy and/or practice supported by the reported data?

Q11. Were the specific directives for new research appropriate?

Characteristics of included studies

Thirteen reviews comprising 128 studies relevant to this review were included in the umbrella review. Of those 128 studies, 79 were unique with 30 of these duplicated at least once across reviews. The most frequently included studies were Casady and Nichols‐Larsen (2004), Kwon et al. (2011), and MacKinnon, Noh, Laliberte, Allan, and Lariviere (1995), each appearing in four reviews. Nine studies were included in three reviews, 18 studies appeared in two reviews and 51 studies appeared only once. Please refer to Table S2 for the complete listing of duplicate studies. Studies included in the reviews were undertaken between the years of 1988 and 2016. One review (Martín‐Valero et al., 2018) also included two systematic reviews—one which is already included in this review (Tseng et al., 2013) and the other which was excluded following critical appraisal (Zadnikar & Kastrin, 2011) and as such is not reported on.

All reviews were quantitative in nature with Jormfeldt and Carlsson (2018) also including a qualitative component. The quantitative study designs varied from randomised controlled trials to single case designs while the qualitative studies consisted of a case study, an evaluation study, an exploratory study, and a longitudinal study.

The number of participants included in each review (that were relevant to this review) ranged from 36 (Bronson et al., 2010) to 601 (White‐Lewis et al., 2017), however, duplication of studies needs to be considered. Participants ranged in age from 2 to 85-years. Six studies focused exclusively on children/adolescents (mainly those with cerebral palsy), five were focused exclusively on adults and two reviews contained mixed populations. Most focused on a health condition/impairment such as cerebral palsy, serious mental illness (SMI), multiple sclerosis, ASD, stroke, and motor function impairment while two had no specific pre‐defined criteria.

Four of the included reviews were not specific to EAIs and examined broader interventions such as exercise, physical therapy and non‐pharmacological interventions of which EAIs were one of multiple interventions included. The way EAIs were described, the terminology used, and the person delivering them varied, however, most reviews distinguished between hippotherapy (considered an equine‐assisted therapy for this review) and therapeutic horseback riding (THR; considered an equine‐assisted activity). Most sessions were provided once or twice per week lasting between 30–60-min and lasted anywhere from 4 to 33-weeks.

Much like the interventions, the outcomes measured were diverse with many outcome tools/methods used on an individual basis. The most frequently measured outcome was the Gross Motor Function Measure (GMFM) which measures gross motor activities. Other outcomes were predominately biological, largely due to many of the reviews focusing on conditions such as cerebral palsy, multiple sclerosis, stroke and motor function impairment where movement and/or function are diminished. The reviews focusing on ASD tended to measure outcomes that can be considered as psychosocial in nature.

The majority of reviews presented results as a narrative summary with some providing effect sizes. Only Tseng et al. (2013) and Stergiou et al. (2017) undertook meta‐analysis. The review including qualitative studies described findings narratively and made no attempt to pool the results in any type of qualitative synthesis. Refer to Table 3 for a table outlining the characteristics of the included reviews.

Table 3. Characteristics of the included reviews

Author (year)	Objectives of systematic review	Review inclusion criteria	Characteristics of included studies
Bremer et al. (2016)	To examine and synthesise the literature with regard to the impact of exercise interventions on the overall behaviour of individuals with ASD from birth to 16-years of age	Participants: individuals aged 0–16-years with a diagnosis of ASD, including diagnoses of PDD made under the Diagnostic and Statistical Manual of Mental Disorders Interventions: exercise interventions (physical activity that is planned, structured, repetitive, and purposeful) compared to no intervention or a different intervention Outcomes: measure of stereotypic behaviours; measure of cognition or attention; measure of social–emotional behaviours; measure of additional behavioural outcomes	Included studies: 13 Year range: 2009–2013 All included studies in review applicable to this review: no, 4 relevant Study designs: 3 controlled trials (1 randomised); 1 cohort Country/setting: not stated Number of participants (relevant to this review): 121
Bronson et al. (2010)	To examine the evidence for hippotherapy as an intervention to improve balance in persons with MS	Participants: people with MS Interventions: hippotherapy Outcomes: balance	Included studies: 3 Year range: 1998–2007 All included studies in review applicable to this review: yes Study designs: pilot case–control study; single study experimental design; pilot case‐series study Country: not stated Number of participants: 36
Dewar et al. (2014)	To evaluate the efficacy and effectiveness of exercise interventions that may improve postural control in children with CP	Participants: children diagnosed with CP, 0–18-years Interventions: land‐based exercise intervention that required active participation by the child Outcomes: postural control	Included studies: 25 primary studies and 3 SRs Year range: 2007–2013 All included studies in review applicable to this review: no, 3 primary studies and 2 SRs relevant, however, SRs not included in results/analysis Study designs: single subject studies; group design Country/setting: not stated Number of participants (relevant to this review): 80
Francis (2007)	To determine the effects of hippotherapy THR as alternative treatment modalities for children with CP, DD, and ND	Participants: children diagnosed with CP (hemiplegic, diplegic, and quadriplegic), DD or ND between 2 and 16-years Interventions: hippotherapy or THR Outcomes: posture; trunk postural reactions; gross motor function; muscle tone; gait speed; functional development; kinematics; functional performance	Included studies: 8 Year range: 1988–2004 All included studies in review applicable to this review: yes Study designs: 3 repeated measures within‐subject design; 1 case study; 1 repeated measures, 1 cohort study, 1 single case design and 1 case series Country/setting: not stated Number of participants: 85
Jormfeldt and Carlsson (2018)	To review studies concerning equine‐assisted interventions among adult individuals diagnosed with schizophrenia. (1) What types of equine‐assisted therapeutic interventions characteristics have been evaluated? (2) What types of study designs have been used to evaluate equine‐assisted therapeutic interventions? (3) Which outcomes, effects and possible benefits of equine assisted therapeutic interventions have been proven?	Participants: people aged 18–65-years with schizophrenia or schizophrenia‐like diagnoses Interventions: EAIs such as EAT, EAP, and THR Outcomes: none pre‐specified	Included studies: 6 Year range: 2000–2016 All included studies in review applicable to this review: yes Study designs: qualitative case study design; qualitative evaluation design; quantitative longitudinal design; qualitative exploratory design; longitudinal qualitative design; comparative quantitative design Country: Italy; USA; Canada; UK Number of participants: 137
Martín‐Valero et al. (2018)	To summarise the grades of recommendation regarding the benefits of hippotherapy in children with CP	Participants: infantile CP Interventions: therapy with a real horse, a simulator and comparative studies of the two Outcomes: gross motor function; stability, mobility, functionality and balance of the trunk, head or upper limbs; the electromyographic activity of the adductor muscle; sitting independently and self‐perception; psychological factors; walking speed or length of stride; evaluation of the quality of life; various daily life activities, for example, jumping, balance in jumping, resistance and ascending and descending stairs	Included studies: 18; 6 systematic reviews were located and appraised but not included in analysis Year range: 2004–2016 All included studies in review applicable to this review: no, 17 relevant Study designs: RCT; non‐RCT; case–control; quasi‐experimental; feasibility study; pilot study; prospective study; case report Country/setting: not stated Number of participants (relevant to this review): 558
Rampling et al. (2016)	To aggregate the evidence base for non‐pharmacological interventions in reducing violence among adults with SMI and PD, and to assess the efficacy of these interventions	Participants: adults (18 and over) with a primary diagnosis of SMI and/or PD with a history of violence or aggression (not sexual) Interventions: any form of specific non‐pharmacological intervention Outcomes: violence outcomes	Included studies: 23 Year range: 2003–2015 All included studies in review applicable to this review: no, 1 relevant Study designs: RCT Country/setting: US Number of participants (relevant to this review): 90
Srinivasan et al. (2018)	To examine the effects of equine therapy on specific domains including social, communication, behavioural, and sensorimotor skills as well as broader functional outcomes including overall adaptive functioning and quality of life	Participants: no predefined criteria Interventions: equine therapy Outcomes: social, communication, behavioural, and sensorimotor skills as well as broader functional outcomes including overall adaptive functioning and quality of life	Included studies: 15 Year range: 2009–2016 All included studies in review applicable to this review: no, 14 relevant Study designs: 3 RCTs; 4 controlled clinical trials; 1 crossover; 5 single group pre‐test post‐test studies/multiple time points; 2 single subject design Country/setting: US; Spain; Canada; UK; Italy; Slovakia Number of participants: 368
Stergiou et al. (2017)	To determine whether THR and hippotherapy improve balance, motor function, gait, muscle symmetry, pelvic movement, psychosocial parameters, & the patients' overall quality of life	Participants: children, adults & the elderly with diagnoses associated with impairments in motor function Interventions: hippotherapy or THR Outcomes: postural control, balance, gait, and spasticity as well as overall quality of life indicators	Included studies: 16 Year range: 1995–2015 All included studies in review applicable to this review: yes Study designs: studies with a control/comparison group or self‐controlled Country/setting: not stated Number of participants: 505
Tseng et al. (2013)	To evaluate the literature on the efficacy of equine‐assisted activities and therapies on gross motor outcomes representing the ICF component of body functions and activity in children with CP	Participants: children with spastic CP under 18-years Interventions: hippotherapy or THR Outcomes: postural control; gait; gross motor activity; muscle activity and tone	Included studies: 15 Year range: 1988–2011 All included studies in review applicable to this review: yes Study designs: RCTs; observational studies Country/setting: not stated Number of participants: 277
Whalen and Case‐Smith (2012)	To determine the efficacy of hippotherapy or THR on motor outcomes in children with CP	Participants: a diagnosis of CP (all types), under 18-years Interventions: hippotherapy or THR; a physical therapist, occupational therapist, or an accredited therapeutic riding instructor implemented the intervention Outcomes: gross motor outcomes	Included studies: 9 Year range: 1988–2009 All included studies in review applicable to this review: yes Study designs: 2 RCTs; 2 one‐group pre‐test–post‐test designs; 5 repeated‐measures within‐subject designs Country/setting: not stated Number of participants: 157
White‐Lewis et al. (2017)	What are outcomes of equine‐assisted therapy interventions studies in adults? What is the significance of these outcomes? What are various interventions, controls and comparisons that are identified? What is the quality, including internal/external validity, bias, power and reporting? What research designs are reported? What are study strengths and limitations? What theoretical/conceptual frameworks have been used to guide this research? What doses, frequency and duration of equine‐assisted interventions have been used?	Participants: age 16-years of age or older Interventions: intervention involving a living horse or horses Outcomes: physical symptoms	Included studies: 31 Year range: 1988–2015 All included studies in review applicable to this review: yes Study designs: 12 RCTs; 19 quasi‐experimental Country/setting: Korea (n = 8); US (n = 7); Brazil (n = 4); Germany (n = 3); Italy (n = 3); Switzerland (n = 2); Canada (n = 1); Portugal (n = 1); Spain; (n = 1); Sweden (n = 1) Number of participants: 601
Wonsetler and Bowden (2017)	To examine changes in spatiotemporal variables and asymmetry measures in intervention studies associated with significant changes in SSWS To differentiate between spatiotemporal raw variables and measures of asymmetry in their potential relationship with motor recovery	Participants: adult participants; >18-years of age; clinically diagnosed with stroke, regardless of time since diagnosis and lesion site Interventions: any clinical physical therapy intervention to effect gait Outcomes: spatiotemporal variables and asymmetry measures associated with SSWS	Included studies: 46 Year range: 1995–2015 All included studies in review applicable to this review: no, 1 relevant Study designs: quasi‐experimental Country/setting: not stated Number of participants (relevant to this review): 15

Note. ASD: autism spectrum disorder; CP: cerebral palsy; DD: developmental delay; EAI: equine‐assisted intervention; EAP: equine‐assisted psychotherapy; EAT: equine‐assisted therapy; ICF: international classification of functioning, disability and health; MS: multiple sclerosis; ND: neurological disorders; PD: personality disorder; PDD: pervasive developmental disorder; RCT: randomised controlled trial; SMI: serious mental illness; SR: systematic review; SSWS: self‐selected walking speed; THR: therapeutic horseback riding.

Findings of the review

The findings of the review are presented narratively and all relevant data (as provided in the individual reviews) are outlined in Table S3. Findings are grouped by population and/or condition in order of the amount of reviews/trials located. The qualitative findings are presented after the quantitative findings.

Children with cerebral palsy

Six reviews representing 39 unique studies examined the use of hippotherapy and/or THR in children with cerebral palsy. Over one‐third of those primary studies were included in at least two reviews. All of the studies included in Dewar et al. (2014) were reported on in other reviews, and all but one study in both Whalen and Case‐Smith (2012) and Tseng et al. (2013) were included in other reviews. A wide variety of outcomes were measured with over 25 different tools used; many of which were only measured once: therefore, the most frequently reported outcomes and tools are summarised here. All reviews used the GMFM‐88 and/or the GMFM‐66 to measure gross motor activities. Results were mixed with some showing a significant improvement and others showing no difference. In those that showed improvement it was rare that this was maintained following conclusion of the intervention. Stergiou et al. (2017) conducted a meta‐analysis of four studies (three involving hippotherapy and one using THR) which demonstrated no significant difference (mean difference, MD: 1.61, 95% CI: −1.92 to 5.14, p = 0.37). This was also supported by Tseng et al. (2013) who pooled studies separately for the GMFM‐88 and GMFM‐66 and demonstrated no significant differences (noting that some of the same studies were included in analyses across reviews). The Pediatric Evaluation of Disability‐Inventory (PEDI) which measures functional skills and level of independence was evaluated in four reviews (Francis, 2007; Martín‐Valero et al., 2018; Stergiou et al., 2017; Whalen & Case‐Smith, 2012) and showed mixed results as did the Sitting Assessment Scale which measures postural control (three reviews: Dewar et al., 2014; Martín‐Valero et al., 2018; Tseng et al., 2013). Balance was measured using the Pediatric Balance Scale in three reviews (Martín‐Valero et al., 2018; Stergiou et al., 2017; Tseng et al., 2013) indicating significant improvements following the intervention. However, Stergiou et al. (2017) undertook a pooled analyses of two studies (both involving hippotherapy) which did not result in a significant difference between groups (MD: 3.21, 95% CI: ‑1.82 to 8.24, p = 0.21). A significant reduction in muscle activity post‐hippotherapy was demonstrated in two studies in Tseng et al. (2013) (pooled result, WMD: −32.20, 95% CI: −61.38 to −3.01) while stride length showed no significant improvement following hippotherapy (pooled result of two studies, Weighted Mean Difference (WMD): 3.37, 95% CI: ‑13.21 to 19.94).

For other outcomes measured please refer to Table S3.

Children and adolescents with autism spectrum disorder

Two reviews (Srinivasan et al., 2018; Bremer et al., 2016) made up of 14 unique studies included children and adolescents with ASD who received hippotherapy or THR as an intervention. The studies included in Bremer et al. (2016) were all included in Srinivasan et al. (2018) so the results of this review will be reported. Outcomes were grouped into domains with the most commonly measured outcomes being social communication, behavioural, and motor. Of the 11 studies measuring social communication outcomes, nine showed significant improvements, for behavioural outcomes, five of seven studies showed significant improvement and for motor outcomes, four of seven studies showed significant improvement. Three out of four studies measuring sensory outcomes showed significant results while for quality of life outcomes, two out of three studies demonstrated improvements. Physiological, functional participation, and executive functioning domains only comprised of one study each and all showed significant effects following the intervention. The authors of this review went on to calculate effect sizes across domains and from those in which confidence intervals did not include zero concluded that the intervention only produced positive effects in the behavioural domain.

Adults with multiple sclerosis

Three systematic reviews (Bronson et al., 2010; Stergiou et al., 2017; White‐Lewis et al., 2017) constituting eight unique studies examined the role of hippotherapy and/or THR in people with multiple sclerosis. The majority of outcomes measured related to balance and gait. All but one study that measured balance showed improvement following the intervention. This was also the same for gait. Other outcomes measured in individual studies that showed improvements included global severity, health status, depression, speed/stride, postural stability/control, spasticity, mobility, muscle tension, ground reaction force, and fatigue, while activities of daily living, pain, disability, somatisation, quality of life, sway, and function did not show improvements.

Elderly

Stergiou et al. (2017) and White‐Lewis et al. (2017) (constituting a total of nine unique studies) included elderly participants and looked at elements of mobility and balance. Pooled results of two studies (one hippotherapy and one THR) showed no significant improvement in balance between intervention and control groups (MD: 0.94, 95% CI: −0.04 to 1.92, p = 0.06). Results were mixed with the timed up and go test (two studies) and muscle strength (two studies) while improvements in step length and sway length, gait, brain activity, body composition, quality of life and hormone levels were noted (all measured in single studies).

Adults post‐stroke

Three systematic reviews (Stergiou et al., 2017; White‐Lewis et al., 2017; Wonsetler & Bowden, 2017) examined the effect of hippotherapy or THR on adults following stroke, however, this constituted only three unique studies. Motor impairment in the lower limbs (Stergiou et al., 2017; White‐Lewis et al., 2017), gait velocity (White‐Lewis et al., 2017), step length asymmetry (White‐Lewis et al., 2017; Wonsetler & Bowden, 2017), and quality of life (White‐Lewis et al., 2017) demonstrated statistically significant improvements while balance (White‐Lewis et al., 2017), ambulation (White‐Lewis et al., 2017), and cadence did not (White‐Lewis et al., 2017).

Spinal cord injuries

One review undertaken by White‐Lewis et al. (2017) included three studies of participants with spinal cord injuries. Out of the outcomes measured (spasticity, pain, and well‐being), only spasticity improved significantly following hippotherapy compared to the control. There was no difference in pain between the intervention and control groups and results were mixed for well‐being. One study included multiple morbidities and data regarding spinal cord injuries could not be extracted individually and thus is not reported here.

Adults with serious mental illness

Two reviews representing two studies focused on adults with SMI, one with schizophrenia/like diagnoses (Jormfeldt & Carlsson, 2018) and the other on people with SMI who had a history of aggression (Rampling et al., 2016). An improvement in negative psychiatric symptoms and a reduction in hospitalisations were noted in the first study using THR (statistical data not provided) while a statistically significant reduction in violent incidences and aggression was reported in the other study utilising hippotherapy.

Adults

White‐Lewis et al. (2017) included two studies measuring balance in adults and both reported improvements. The review also included two studies (published in the same article) looking at strength‐disabled adults and reported improvements in self‐concept and strength/coordination.

Brain disorders

Significant improvements post‐hippotherapy in balance, gait and walking were seen in White‐Lewis et al., 2017 (one study) in people with brain disorders. No changes in depression and functional ambulation were demonstrated.

Breast cancer survivors

The effect of THR on breast cancer survivors was measured in White‐Lewis et al. (2017) (one primary study). Significant improvement in cardiac fitness, body composition, strength, and quality of life were reported.

Obese women

White‐Lewis et al. (2017) also included a study looking at the weight and gait of obese women. Statistically significant improvement in BMI and gait were reported.

Other

White‐Lewis also included one study containing people with “physical/psychosocial‐intellectual disability,” however, details on the dosage of the intervention, measurement tools and outcome comparisons were not provided therefore results are not reported.

Qualitative findings

Adults with serious mental illness

The review by Jormfeldt and Carlsson (2018) included four qualitative studies. Only a narrative summary of each study was provided with no participant illustrations provided. Overall, all studies reported positive experiences of participants involved in the program with no negative experiences or issues mentioned. Positive elements reported related to the relationship between the horse and the participant; the experience being a “stepping stone” for participants; the acquirement of transferrable skills; and psychosocial benefits including improved confidence, self‐esteem, self‐concept, and self‐efficacy.

DISCUSSION

Based on the 79 unique studies that were considered in our umbrella review of 13 separate systematic reviews of the literature on EAIs, largely comprising of hippotherapy and THR, the conclusion from this present study is that the evidence that such interventions are beneficial across biological, social and psychological domains, across the lifespan, is equivocal. While some studies report improvements, and others do not, the current evidence base is flawed and compromised by serious methodological weaknesses in study design. At this point in time, therefore, while it may be that therapeutic interventions that include a horse have benefits, such interventions cannot be confidently advocated as best practice, based on current research evidence.

All of the review articles included in this study noted similar methodological challenges in their included articles, particularly around risk of bias (Wonsetler & Bowden, 2017). For example, Whalen and Case‐Smith (2012) noted that the research design of all studies in their review was a limitation, with the majority of studies using repeated measures, within‐subject or one group pre‐test–post‐test designs. Small sample sizes, no control group or comparison groups, no randomisation, and lack of detail about the intervention were further methodological weaknesses. Despite this, although, Whalen and Case‐Smith concluded that there is evidence that hippotherapy and THR may have positive effects on gross motor function in children with cerebral palsy. Bronson et al. (2010) identified small sample sizes, lack of standardised outcome measures, lack of randomisation, and case–control and case series designs as methodological challenges in their review, yet also concluded that hippotherapy has a positive effect on balance in persons with multiple sclerosis and enhances quality of life.

One challenge of reviewing and combining the evidence in the area of EAIs is the diversity of methods of intervention (e.g., frequency, duration, content, and personnel involved), terminologies used, outcome measures and tools employed, as well as the presenting issue or condition for treatment. This dilemma has been noted by other reviewers (Lee et al., 2016). The interdisciplinarity of EAIs also poses a challenge, insofar as the approaches taken in the design of studies and the outcome measures of interest are not common across studies. This review highlights the focus on biological or physical aspects of health, such as balance and gait and a corresponding dearth of research into psychological health: for example, there were no studies identified on EAIs for depression. Anestis, Anestis, Zawilinski, Hopkins, and Lilienfeld (2014), in their review of equine‐related treatments for “mental disorders,” concluded that existing studies are so flawed, and the evidence base so weak, that these interventions should not be offered to the public. This strong stance is noteworthy in the context that many authors support the use of EAIs and argue for its benefits, even in light of the lack of evidence. Jormfeldt and Carlsson (2018) state that even though further research is necessary for a solid evidence base, people with severe mental illnesses such as schizophrenia may benefit from EAIs. However, Anestis et al. are not alone in their position that advocating for equine therapies is unwarranted. Tseng et al. (2013) similarly drew the conclusion, based on their meta‐analysis, that there is no evidence that THR or hippotherapy improves gross motor function for children with cerebral palsy.

The limited number of studies that exist is also noteworthy, given the seeming popularity of interventions that include horses. Stergiou et al. (2017), in their meta‐analysis, decried that they could only identify eight studies for inclusion. These eight studies, on hippotherapy and THR, included participants from children through to the elderly, and with a range of physical health problems. Consistent with other reviews, the need for further, better designed research was the conclusion made, though the authors concluded that these interventions appear to be viable options for balance, gait, and psychomotor disorders. The limited number of high‐quality studies on which to base systematic reviews and meta‐analyses and the heterogeneity of participants, compromise the evidence base for EAIs.

There are some limitations to this umbrella review. Manuscripts that were not in English that may have been eligible were excluded (Mansfeld, 2002; Moraes, da Silva, Copetti, Abreu, & de David, 2015; Prieto, Silva, Silva, Santos, & Gutierres Filho, 2018). The majority of eligible studies were quantitative. While this is not a limitation to our study, it is interesting that given the number of qualitative studies in the area of AAT and AAI, there was only one located involving EAIs that was subsequently included. Future research is indicated that is based on well‐designed qualitative methods.

Although the search of databases and resources considered unpublished and/or grey literature and the review subsequently included one thesis, publication bias is also acknowledged. During study screening a few theses were identified that may have met the inclusion criteria but could either not be located or due to budget constraints the full‐text was not obtained. Unpublished and/or grey literature plays an important role in reducing publication bias by producing a more balanced picture of available evidence and should thus be included in reviews wherever feasible.

As alluded to previously, the heterogeneity of the included articles in terms of populations, interventions and comparators, outcomes and study design precluded pooling of studies across the majority of reviews. The relatively small number of studies and the large amount of duplication across reviews suggests more primary studies may be needed in the area as well as indicating potentially unnecessary or redundant synthesis has been undertaken. While some independent replication can be worthwhile, it has also been criticised (see Konstantinos, Siontis, Hernandez‐Boussard, & Ioannidis, 2013) and deliberation on approaches to efficient synthesis of evidence and the reduction of “research waste” should be considered.

Although the assessment of methodological quality of the 13 systematic reviews that were included in this umbrella review indicated that the majority of included articles were of good quality and specific assessment criteria deemed essential were established a priori, some methodological limitations were noted. In particular, Q8 of the appraisal tool around methods to combine studies; although most (if not all) of the articles acknowledged the heterogeneity across studies many did not explicitly outline their approach to data synthesis and made no mention of the intention to statistically combine studies but simply provided a narrative summary. Important statistical data were not provided in some articles (in one review supplementary data was not accessible) and in one article systematic reviews were included and appraised but subsequently not included in the results. Additionally, for Q2 regarding appropriate inclusion criteria, many articles were broad in defining inclusion criteria in terms of outcomes, associated tools and interventions. Assessment of publication bias was rarely undertaken and some of the appraisal tools used were general and not specific to study design. The short‐term nature of outcome measurement was common with very few measuring long‐term follow‐up. In terms of data, some discrepancies in figures presented in the narrative and the tabular format were evident and in one article the authors reanalysed the primary data and reported variations in results between their calculations and that of the primary study authors. Finally, it was unclear why some studies calculated effect sizes but did not go on to pool studies. Based on these issues a modified approach to how to present data was warranted in this umbrella review; however, the authors acknowledge this as a deviation from the JBI methodology.

The need for well‐designed (and well‐reported), rigorous quantitative and qualitative methodologies to assess the impact of EAIs is clear. This call for better methods is not new. Indeed, researchers in the field of anthrozoology have been calling for higher level standards in research for a number of years (Chur‐Hansen et al., 2010). With regards to quantitative approaches, including experimental designs, Kazdin (2015) lists randomisation to conditions, clear inclusion and exclusion criteria for participants, strong control and comparison groups, the use of treatment manuals and protocols, assessment of treatment integrity, the use of multiple (reliable and valid) outcome measures, the evaluation of both clinical significance as well as statistically significant change, and follow‐up or longitudinal data collection. There are some promising signs that some progress is being made: for example, Srinivasan et al. (2018) in their systematic review of equine therapy for people with ASD were the first to report of the size of treatment effects by calculating effect size estimates and their 95% confidence intervals. Tseng et al.'s (2013) meta‐analysis is noteworthy, with a second meta‐analysis published in 2017 by Stergiou et al. (2017). Herzog (2015) has called for an AAT Clinical Trial Registry, which to date, does not exist. Qualitative research similarly needs to follow trustworthy and rigorous methods, with clear research questions articulated (Stern & Chur‐Hansen, 2013), and well‐designed methods of data collection, analysis and reporting adhered to (Braun & Clarke, 2006, 2013). One contribution of the current umbrella review is the evidence that standardised approaches to the implementation and evaluation of interventions is a fundamental requirement for this field to produce credible and convincing data to support the claim that EAIs are effective. Without this, health professionals cannot reasonably or ethically assert that interventions with horses are beneficial to humans. Similarly, the community cannot be confident that the treatments they are receiving have demonstrated efficacy. It is noteworthy that Serpell, McCune, Gee, and Griffin (2017) have highlighted public and media pressure to report positive findings related to AAIs. Consequently, findings that show no effects, or where human or animal welfare might be compromised, are likely underreported, in both the academic and wider community literature. As scientist‐practitioners, psychologists are well placed to make meaningful contributions to knowledge in this area, and we call upon scholars to conduct high quality research to provide an evidence base from which to advocate for EAIs, or not, from a position of theoretical and methodological strength.