Assistance Dogs: What Studies Show for People and Dogs
Michael Sauerwein
Assistance dogs guide blind people, alert people with diabetes to changes in blood glucose, respond to seizures and live with veterans with posttraumatic stress disorder (PTSD) and children with autism. Many studies with waitlist comparisons and owner reports describe benefits, and a nonrandomized controlled trial found lower PTSD severity after three months with a service dog, including on a blinded clinician interview (Leighton, Rodriguez, Jensen, et al., 2024). At the same time, a randomized trial comparing PTSD service dogs with emotional support dogs found no significant difference on its primary outcomes, while measured accuracy of diabetes alert dogs depended strongly on the test procedure (Richerson et al., 2023; Los et al., 2017; Rooney et al., 2019).
This article reviews the evidence for PTSD service dogs, autism assistance dogs, diabetes and seizure alert dogs, and guide and mobility dogs, including what predicts training success and why dogs leave work early. It examines the little that is known about the welfare of the dogs themselves, the methodological weaknesses of the field, and what follows for trainers and programs.
1. Terms and Types of Evidence
1.1 Assistance Dog as an Umbrella Term
In this article, "assistance dog" is the umbrella term for dogs trained to support one person with a disability in daily life: guide, hearing and mobility dogs, medical alert and response dogs, for example for diabetes or epilepsy, psychiatric service dogs, for example for PTSD, and dogs placed with children with autism. The literature uses these terms inconsistently. One systematic review included guide, hearing, medical and mobility dogs but excluded psychiatric service dogs, emotional support animals and pets (Rodriguez et al., 2020), while a review on veterans spoke of psychiatric assistance dogs (Leighton et al., 2022).
Therapy, visiting and school dogs usually work with many people under a handler's guidance instead of living with one person with a disability (how school, therapy, visiting and assistance dogs differ). What that work means for the dog and what studies show in classrooms are reviewed separately (what animal-assisted work means for the dog; what studies show about school dogs). An assistance dog, by contrast, is with its partner most of the day, for years, and its work and home life are hard to separate.
1.2 "Service Animal": The US Legal Term
Much of the research comes from the United States, where "service dog" is the common term. Under the Americans with Disabilities Act (ADA), service animals are defined as dogs that are individually trained to do work or perform tasks for people with disabilities; dogs whose sole function is to provide comfort or emotional support do not qualify (U.S. Department of Justice, 2020). The legal situation in other countries is not addressed here.
The distinction also raises a research question: Do specifically trained assistance tasks add benefits beyond the presence of a dog? One randomized study compared PTSD service dogs with emotional support dogs and found no significant group difference on its primary outcomes (Richerson et al., 2023).
1.3 What Kind of Evidence Exists
The evidence base is young. According to a 2022 review, psychiatric assistance dogs for veterans made up more than 19% of assistance dog partnerships worldwide, and all peer-reviewed papers on them had appeared within the five years before that review (Leighton et al., 2022). Typical designs are waitlist comparisons, owner reports and records, small scent-detection studies and large records from guide dog organizations; randomized trials are almost absent.
2. PTSD Service Dogs
2.1 Observational and Waitlist Studies
A widely cited study compared 141 post-9/11 military members and veterans with PTSD: 75 who had received a service dog in addition to usual care and 66 on the waitlist with usual care alone (O'Haire & Rodriguez, 2018). Participants with a dog showed clinically significant reductions on the PTSD Checklist (PCL) from baseline, unlike those with usual care alone, and had lower depression, higher quality of life and better social functioning than the waitlist, with medium to large effect sizes. Average scores did not fall below the diagnostic cut-off, and the authors wrote that dogs "may confer clinically meaningful improvements" without being associated with a loss of diagnosis. Employment status did not differ, but employed participants with a dog had less health-related absenteeism. The study was not randomized, outcomes were self-reported, the waitlist group was self-selected and may have differed, for example in time since application, and the dog provider was also the recruitment source.
A small uncontrolled study followed two cohorts of veterans, seven and five people, through a three-week training program in which they were partnered with a psychiatric service dog (Kloep et al., 2017). PTSD and depression symptoms decreased significantly up to the six-month follow-up, with clinically significant and reliable change for most participants, along with less anger and better perceived social support and quality of life. The authors named the missing control group and the small sample as limitations. Because the dog came together with the program, the two effects cannot be separated.
Physiological measures complicate the picture. In 73 veterans, 45 with a service dog and 28 on the waitlist, the dog group had a higher cortisol awakening response, along with lower self-reported anxiety, anger, sleep disturbance, alcohol abuse and PTSD severity (Rodriguez et al., 2018). In this nonrandomized comparison, the cortisol response itself was not significantly associated with PTSD symptom severity, within or across groups. The authors suggested it "may indicate better health", but its clinical significance "cannot be confirmed" in a cross-sectional design. Cortisol was measured in the veterans, not in the dogs (how stress is measured and where the methods fail).
A Dutch cross-sectional study of 65 people compared PTSD with a service dog (20), with a companion dog (10), without a dog (12) and people without PTSD (23) (van Houtert et al., 2022). The service-dog group reported fewer symptoms and better sleep and wellbeing than the companion-dog group, and fewer symptoms than the group without a dog, but salivary cortisol did not differ between groups, and changes in cortisol and activity were not linked to wellbeing. With groups of 10 to 23 people, small differences are easily missed or overestimated.
2.2 A Nonrandomized Controlled Trial with Blinded Ratings
A prospective nonrandomized trial included 156 veterans and military members with PTSD, 81 receiving a service dog plus usual care and 75 on the waitlist with usual care, allocated by application date and all recruited from one accredited non-profit provider (Leighton, Rodriguez, Jensen, et al., 2024). Mean age was 37.6 years, and 75% were male. After three months, the service-dog group had lower self-reported severity on the PCL-5 (41.9 versus 51.7; difference −11.5, 95% CI −16.2 to −6.6) and lower severity in the CAPS-5 interview (30.2 versus 36.9; difference −7.0, 95% CI −10.8 to −4.5), conducted by clinicians blinded to trial topic, timing and allocation. Odds of depression (OR 0.45), anxiety (OR 0.25) and social isolation (OR 0.34) were also lower, and companionship was higher (OR 2.83), but participation in social activities was significantly lower (OR 0.24), which the authors linked to possible access denials and stigma in public. The authors concluded that the partnership "was associated with lower PTSD symptom severity and higher psychosocial functioning" and that service dogs "may be an effective complementary intervention".
The blinded interview is a step beyond the earlier dependence on self-report. Still, allocation was not random, follow-up lasted three months, the authors described the other outcomes as "limited by self-reporting biases", and the waitlist group knew it would receive a dog later. Funders included NIH and pet-industry sources such as Merrick Pet Care and the PetCo Foundation, which according to the authors had no role in design or analysis.
2.3 Daily Life and Public Access
A nonrandomized study of 168 veterans used ecological momentary assessment: short questions twice daily for 14 days, before and three months after placement, corresponding to up to 9,408 scheduled prompts, not all of which were completed (Leighton, Rodriguez, Zhuang, et al., 2024). A service dog was associated with better social interactions, more positive and less negative affect, lower odds of panic attacks and more activities, but also with lower odds of being away from home (OR 0.77). The authors read this as possible support for public stigma as an obstacle and called for education on "service dog etiquette". They also suggested that trained tasks may matter more for social outcomes and the dog's presence more for emotional ones, an interpretation the design cannot test directly.
A preregistered integrative review of 43 US articles found consistent reports of stigma, discrimination and access denials for service dog partners, possibly more common with invisible disabilities such as diabetes or mental health conditions (Leighton et al., 2025). The authors attributed this in part to inadequate legal protection and to "service dog fraud". It contains no data on the dogs, although the dog is present at such confrontations at entrances.
2.4 The Randomized PTSD Service Dog Trial
The only randomized trial to compare PTSD service dogs directly with emotional support dogs was a multicenter study by the US Department of Veterans Affairs (Richerson et al., 2023). Veterans with PTSD were randomly assigned to one or the other and followed for 18 months; of 227 randomized, 181 received a dog. Mean age was 50.6 years, and 80% were male. Primary outcomes were functioning (WHODAS 2.0) and quality of life (VR-12); secondary outcomes included PTSD symptoms, suicidal ideation, depression, sleep, health-care use and costs, medication adherence and employment. There was no group difference on the primary outcomes. The service-dog group showed a 3.7-point greater reduction in PTSD symptoms on the secondary PCL-5 measure (p = 0.036), which the authors did not treat as an established clinically meaningful advantage, and 10 percentage points higher antidepressant adherence; health-care use and costs were not reduced. The authors concluded that both groups appeared to benefit and that the findings on PTSD symptoms and adherence "should be explored further".
Without a group lacking a dog, "both benefited" cannot be separated from time or expectations. Participants and local study staff were masked to dog assignment only during the pre-pairing observation period; after allocation and pairing, masking could not be maintained. The design paper describes prioritizing the safety of participants and dogs, choosing the comparison group partly to make enrollment feasible, using "pragmatic methods" to ensure healthy, well-trained dogs, and having "no blueprint" for a trial of this size (Saunders et al., 2017). It is nevertheless the strongest test so far of whether trained tasks add something beyond a dog's company, and on its primary outcomes, they did not.
2.5 What Reviews Conclude
A scoping review screened 126 publications and found only six empirical studies among the 19 it included; most of the literature concerned human-animal interaction in general, and the dog-specific ones assessed only veterans' self-perceived wellbeing (van Houtert et al., 2018). The authors saw "no undisputed empirical evidence" of effectiveness beyond positive self-reports and warned that missing knowledge on dog welfare "creates risks for both human and animal welfare". A later systematic review included 41 articles, 29 peer-reviewed papers and 12 dissertations, and its meta-analysis of nine articles found a large effect on PTSD severity (g = −1.129), with inconclusive effects on physical health and quality of life (Leighton et al., 2022). Its studies were largely nonrandomized, waitlist-controlled and self-reported, several by the review's own research group, so the pooled effect summarizes these weaknesses rather than overcoming them.
2.6 Dog and Team Factors
In 82 veterans and their PTSD service dogs, most dog characteristics did not predict outcomes (Jensen et al., 2022). Lower excitability of the dog was associated with less PTSD severity at follow-up and a closer relationship (how temperament and coping styles are measured). Better mental health was associated with less use of dominance-based training methods, lower perceived costs of the dog and a closer relationship (what aversive methods do in the dog). Veterans spent on average 82% of their time with the dog, and calming or comforting during anxiety was the most used task. More frequent use of a "make a friend" task was associated with worse mental health, and use of a "cover/watch back" task with more anxiety but less PTSD severity. These are correlations: a veteran who is already doing better may train differently, and a calm dog may simply be easier to live with.
In practice: Prepare handlers of psychiatric service dogs for reward-based everyday handling as carefully as for the dog's tasks, including what to do when the dog makes mistakes, and discuss high excitability honestly before placement instead of hoping the dog will settle. This is a practice recommendation, not a study result.
3. Autism Assistance Dogs
3.1 What Parents Report
At one accredited center in Ireland, 134 parents with an assistance dog and 87 on the waiting list were surveyed (Burgoyne et al., 2014). Because the age distributions differed, children over ten were excluded from the group comparisons, leaving 80 families with a dog and 84 on the waitlist. In this comparison, parents with a dog rated their children as safer from environmental dangers (p < 0.001), perceived the public as more respectful (p < 0.001) and felt more competent as parents (p = 0.023), but also described the commitment that caring for a dog requires. As this was a nonrandomized parent-report comparison, it cannot establish a causal effect of the dog.
3.2 A 2025 Review: Effects Similar to Companion Dogs
A mixed-methods systematic review pooled 14 studies with 1,106 people with autism, one randomized trial, three cohort, seven cross-sectional and three single-arm studies, and synthesized 13 studies qualitatively (Mulraney et al., 2025). There was some quantitative evidence of benefit, for example for adaptive functioning and family outcomes, but pooled effects were similar to those of companion dogs, and the comparisons with companion dogs were not statistically significant. All included studies carried a high risk of bias. Qualitatively, parents described benefits but also challenges they were unprepared for, such as time, costs and access denials in public. The authors found a "lack of compelling evidence" that assistance dogs confer unique benefits, called the evidence base "limited and of poor quality" and concluded that many potential benefits could be achieved by a companion dog. They wrote the review while employed by Australia's National Disability Insurance Agency, a funder of assistance animals, and declared this as a competing interest. This does not make autism assistance dogs useless; as in the PTSD trial, what training adds to a dog's presence remains open.
4. Diabetes Alert Dogs
4.1 Can Dogs Smell Low Blood Glucose?
In a laboratory study, six dogs aged one to ten years, trained for about six months with positive reinforcement, chose among seven cans containing one perspiration sample from hypoglycemia, two from normal glucose and four blanks (Hardin et al., 2015). All dogs performed above chance (14%), with sensitivity from 50.0% to 87.5% and specificity from 89.6% to 97.9%. The authors concluded that trained dogs can recognize such samples "in an in vitro setting using smell alone" (how the sense of smell shapes behavior). The task is possible in principle, but recognizing a sample in a lineup is not the same as noticing a change in a person at home or at night.
4.2 Owner Records
In records kept by 17 clients with trained alert dogs, all reported positive effects such as fewer paramedic call-outs and more independence; eight of ten dogs with adequate records alerted more often when glucose was out of range, and seven of nine clients had a higher proportion of tests in range after obtaining the dog, while HbA1c fell only slightly and not significantly (Rooney et al., 2013). A larger analysis of routine owner records for 27 dogs from one UK charity, with almost 4,000 out-of-range episodes, found a median sensitivity of 70% (interquartile range 50–95%), 83% for hypoglycemia and 67% for hyperglycemia, and a median positive predictive value, the proportion of alerts made at out-of-range glucose, high or low, of 81% (Rooney et al., 2019). Four dogs had a positive predictive value of 100%. Performance varied with dog, partnership and household factors, for example whether the dog had previously been a pet and whether its partner was an adult or a child; the authors called it "better than previous studies suggest". Two authors were full-time employees of the training charity, Medical Detection Dogs, and the first author a part-time paid employee; the author who analyzed the data had no financial interest.
4.3 Comparisons with Continuous Glucose Monitoring
Independent measurement produced lower figures. In an exploratory study of eight people aged 4 to 48 years, alerts were compared with capillary blood glucose and with continuous glucose monitoring (CGM) to which participants were blinded (Los et al., 2017). Satisfaction with the dogs was high, and dogs alerted 3.2 times more often during hypoglycemia than during normal glucose, but sensitivity was 36%, the positive predictive value for hypoglycemia alerts was only 12%, and when both alerted, the CGM would usually have been first, by a median of 22 minutes. The authors concluded that a dog's alert alone is "unlikely to be helpful" to distinguish glucose states. One author reported ties to a CGM manufacturer.
In owner diaries from 18 owners, overall sensitivity was 57.0% and specificity 49.3%, with sensitivity for low glucose ranging from 33% to 100% between dogs (Gonder-Frederick, Grabman, Shepard, Tripathi, et al., 2017). With masked CGM in 14 owners, only 3 of 14 dogs performed better than chance while awake and 1 of 11 during sleep, and dogs responded to absolute glucose levels rather than rapid changes (Gonder-Frederick, Grabman, & Shepard, 2017). Waking sensitivity was 35.9% for low and 26.2% for high glucose. The authors pointed to a gap between owners' positive experiences and the accuracy data.
The UK charity also tried to remove the owner from the measurement: alerts of eight accredited dogs were recorded by CCTV and coded blind to flash glucose values (Wilson et al., 2019). Over 5 to 14 days, dogs alerted 2.80 times more often during hypoglycemia and 2.29 times more often during hyperglycemia than in range; hypoglycemia sensitivity ranged from 33.3% to 91.7%, and the mean positive predictive value was 69.7%. Performance was associated with owners' adherence to the training protocol. The first author was salaried by the charity to develop the paper.
4.4 Why the Figures Are Not Comparable
The often quoted 12% and 81% do not measure the same performance. The 12% is the positive predictive value of hypoglycemia alerts against blinded CGM (Los et al., 2017); the 81% is a median for alerts at out-of-range glucose, hyperglycemia included, from owner records of the training charity whose staff co-authored the study (Rooney et al., 2019). Data collection, glucose thresholds and conditions also differed. Definitions alone shift results: in the CCTV study, the positive predictive value was 69.7% or 50.4%, depending on how the relevant glucose range was defined (Wilson et al., 2019). In another study, most dogs did not exceed chance (Gonder-Frederick, Grabman, & Shepard, 2017). A narrative review listed sensitivities of 0.29 to 0.80, specificities of 0.49 to 0.96 and drawbacks such as costs, lack of provider certification, poorly harmonized training, variability between and within dogs, owners' ability to recognize alerts and lack of quality assessment (Lippi & Plebani, 2019). Some dogs alert reliably and others do not, so group averages say little about an individual dog.
In practice: Before placing an alert dog, agree with the family on how every alert will be recorded and checked against a glucose measurement, review these records together regularly, and present the dog as a supplement to established monitoring, not a replacement. This is a practice recommendation, not a study result.
5. Seizure Alert and Response Dogs
5.1 What the Field Studies Show
Response dogs act during a seizure, for example by staying with the person or fetching help; alert dogs are said to signal before one. A scoping review included only five of 28 peer-reviewed studies, with 9 to 22 participants each (Catala et al., 2018). Reported alert times ranged from 10 seconds to 5 hours and owner-reported accuracy from at least 70% to 85%, and alerts were mostly attention-getting behaviors by dogs of many breeds and sizes; most studies reported better quality of life and fewer seizures. The authors judged rigor "generally poor" and the data "too scarce and preliminary" for conclusions.
The prospective study among them followed ten consecutive referrals over 48 weeks with 12 weeks of baseline, 12 of training and 24 of follow-up, and found a 43% mean reduction in seizure frequency; nine of ten participants improved by at least 34%, and one did not improve (Strong et al., 2002). There was no control group, the dogs came from the authors' own service, and people referred at a time of frequent seizures may improve anyway as frequency returns toward its usual level. In a survey of 63 patients, 29 owned dogs, nine reported that the dog responded to seizures and three that it also alerted, with no association with the dog's age, sex or breed. Drawing also on interviews with 15 trainers, the authors concluded that success depends largely on the handler's awareness of the dog (Dalziel et al., 2003).
5.2 Is There a Seizure Odor?
In an odor discrimination study, five trained dogs chose among seven breath and sweat samples from five patients, collected during seizures, calm activity and sport; the dogs had never encountered these patients' odors and completed nine trials each (Catala et al., 2019). Three dogs reached 100% sensitivity and specificity, two 67% and 95%; overall sensitivity was 86.8% and specificity 98%, with chance at 14% and no learning effect across trials. The authors called this "first proof" of a seizure-associated odor, but the samples were taken during seizures, so anticipation was not tested, and funding included the assistance-dog association Handi'Chiens.
In another study, sweat samples from 60 people in an epilepsy monitoring unit were tested by dogs of a service-dog organization (Maa et al., 2021). Across 680 observations, the observed discrimination of seizure and non-seizure samples yielded 65% sensitivity and 88% specificity; the often cited 93.7% (OR 14.89) is a model-derived estimate, not an observed accuracy rate. Of 19 non-epileptic events, 18 were not identified as seizures. With hourly sampling in a small subgroup, dogs also identified pre-seizure samples (model-based estimate 82.2%), but the confidence interval of that odds ratio included 1 (OR 4.60, 95% CI 0.98–21.69); the mean warning phase was 68.2 minutes. The samples were tested remotely after collection, so the results do not establish reliable live alerts in daily life.
5.3 Prediction Is Not Established
In a survey of 227 people with seizures, owners' reports of alerting behavior were associated with their own pre-seizure symptoms and with a stronger human-dog bond (Martos Martinez-Caja et al., 2019). This leaves open the possibility that dogs detect early changes in the person, but does not prospectively measure reliable warnings. Controlled scent studies showed that dogs could distinguish seizure from comparison samples (Catala et al., 2019), but did not establish positive-prediction and false-alarm rates for individual warnings in daily life. Reliable seizure prediction under everyday conditions therefore remains unproven.
6. Guide and Mobility Dogs
6.1 Outcomes for Users
In a telephone survey of more than 800 people with visual impairment, the most commonly cited benefits of a guide dog were independence, confidence, companionship, social interaction and mobility, and demographic and contextual factors shaped both reasons and perceived drawbacks (Whitmarsh, 2005). The author concluded that the most appropriate mobility aid for the individual is the hallmark of good rehabilitation. An Austrian survey of 36 blind people found that guide dog ownership was not associated with better quality of life, although medical costs were descriptively lower and owners believed dogs increase independence and health (Glenk et al., 2019). With 36 participants, the absence of a difference is not evidence of no effect. Across guide, hearing, medical and mobility dogs, 68% of 147 statistical comparisons were null, 30% positive and 2% negative; positive findings concerned psychological wellbeing, emotional functioning, self-esteem and vitality, but weak reporting and uncontrolled confounders prevented firm conclusions (Rodriguez et al., 2020). All 12 studies in a review on people with physical disabilities had design concerns such as small samples and no power calculations (Winkle et al., 2012).
6.2 Training Success and Its Predictors
At The Seeing Eye, the program one research group studied, about 70% of dogs entering training succeed, a figure for that program, not an international rate (Bray et al., 2017b). In 98 puppies followed from birth there, higher overall maternal behavior was associated with failure, while nursing styles that required more effort from the puppies were associated with success; these are associations, and genetic confounding prevents a causal interpretation. Failure was also associated with slow multistep problem-solving with perseveration and a short latency to vocalize at a novel object, interpreted as greater anxiety (how maternal care shapes puppy development).
With a 25-item cognitive test battery in 164 assistance dogs and a prospective sample of 180, predictions for assistance dogs in the top predicted quartile were on average 86% accurate, and successful dogs more often made eye contact with an experimenter when facing an unsolvable task (MacLean & Hare, 2018). In 1,401 juvenile guide dogs, seven questionnaire scales, among them Excitability, General Anxiety and Trainability, were associated with outcome, but thresholds identified only 8.4% of dogs later withdrawn for behavior and 8.5% of qualified dogs (Harvey et al., 2017). Such tools flag the extremes; they do not predict every dog (what puppy temperament tests can predict). A pilot study of 43 dogs found that laterality, reaction to an unfamiliar dog, latency to lie down and rest, coat color and breed predicted success, and that assessment at 14 months was more accurate than at six months (Batt et al., 2008). With many predictors and few dogs, these findings need replication. A review of selection tests and of improving performance through rearing, training and breeding noted that selection practices also affect working-dog welfare (Bray et al., 2021).
6.3 Puppy Raising and Maternal Style
Maternal style was described in 21 mothers and 138 puppies during the first three weeks as one stable component associated with breed, litter size, parity and maternal cortisol (Bray et al., 2017a). After the puppy leaves its mother, the raising environment matters. In 978 prospective guide dogs, more experienced raisers and another dog in the household were associated with better behavior scores, while being frightened by a person was associated with worse fear scores and, together with less experienced raisers and no other pets, with less successful training (Serpell & Duffy, 2016). In German Shepherds, aggression toward strangers increased from 6 to 12 months. The data are correlational and rated by the raisers (how behavior changes during adolescence).
In practice: Give volunteer puppy raisers more than a handbook: regular supervision by experienced staff, a plan for meeting people and dogs without frightening encounters, and quick support when a puppy shows fear. This is a practice recommendation, not a study result.
6.4 Withdrawal and Returns
One analysis of more than 7,500 dogs removed from service at Guide Dogs UK between 1994 and 2013 compared only dogs retired for age with dogs withdrawn for behavior; health-related withdrawals were excluded. Of these, 83% retired and 17% were withdrawn, mainly environmental anxiety, training problems and fear or aggression (Caron-Lormier, Harvey, et al., 2016b). The 17% is therefore not an overall rate of early withdrawal. Breed and sex affected the odds, fear and aggression tended to appear early, and training problems occurred at any age. A second analysis, of 7,686 dogs over 20 years, excluded behavioral withdrawals: 84% worked until retirement, and German Shepherds had shorter working lives; musculoskeletal disease, mostly arthritis, was the most common health reason for withdrawal, and skin conditions shortened working life the most (Caron-Lormier, England, et al., 2016). Behavioral withdrawals cost the workforce more working time than health-related ones, and trends over time reflected positive effects of the organization's interventions (Caron-Lormier, Harvey, et al., 2016a). Because the two analyses used different selections, their percentages cannot be combined into an overall withdrawal rate. Returns also occur after placement: in 118 pairings, 43 dogs were returned before retirement age, 26 for behavioral and 14 for physical reasons, mainly poor health, and three for handler-related reasons; second dogs were returned more often, and previously used guides could be matched again (Lloyd et al., 2016).
6.5 The End of a Partnership
In 36 New Zealand handlers with 77 ended pairings, most experienced reduced quality of life through loss of mobility, companionship, social contact and self-esteem; extreme grief was common, especially after the first dog, and more than a quarter reported negative effects on the relationship with the following dog (Lloyd et al., 2021). What retirement means for the dogs was not examined.
7. Dog Welfare in Assistance Work
7.1 Cortisol in PTSD Service Dogs
In 19 service dogs for veterans with PTSD, saliva was collected on arrival, before and after training and after free play (van Houtert et al., 2021). Mean values ranged from 1.55 to 2.73 nmol/L, which the authors described as relatively low. Cortisol before and after training did not differ, and both were lower than on arrival and after free play. The authors concluded that the dogs "did not seem to experience training as stressful"; free play produced the highest values, a reminder that cortisol also rises with positive arousal (how acute and chronic stress differ in the body). Hair cortisol of 11 service dogs did not differ from that of 19 companion dogs (van Houtert et al., 2023); in this small sample, coat color, season, sex and mental-health diagnoses in the household showed no statistically significant association with hair cortisol. Both studies are small and cover single sessions or short periods, hair covers only about two months, and a null result does not show equivalence.
7.2 Autism Dogs in the Family
Interviews and home observations in families with autism service dogs identified stressors: lack of rest after working, unintentional maltreatment and prodding by children, unpredictable routines and too little recreation (Burrows et al., 2008). The dogs bonded primarily with the parents. The authors concluded that unmet physical, emotional and social needs can affect behavior, welfare, performance and parental satisfaction (why control and predictability matter for dogs).
In practice: Before placing a dog with a family, agree in writing on a resting place the child does not enter, daily free exercise and play, and one adult responsible for the dog, and check these arrangements at follow-up visits. This is a practice recommendation, not a study result.
7.3 Working Life and Retirement
Guide dog records measure whether a dog stays in service, not how it feels while working (Caron-Lormier, England, et al., 2016). None of the studies considered here systematically assessed dogs' welfare after retirement; Lloyd et al. (2021) examined the dogs' destinations and handlers' experiences when partnerships ended. Nor did any examine the welfare of diabetes or seizure alert dogs, although some are expected to alert at night (Gonder-Frederick, Grabman, & Shepard, 2017). For PTSD service dogs, the only physiological data come from two small studies from the same Dutch research group: one found no significant difference immediately before and after training, while the other found no significant difference in hair cortisol between service dogs and companion dogs. Validated measures of workload, rest at home or the strain of public access are lacking (how welfare science defines quality of life in dogs).
8. Methodological Weaknesses
8.1 Self-Report and Waitlists
Most human outcomes rest on self-report or parent report from people who applied for a dog. Waitlist designs compare people who have a dog with people still hoping for one (O'Haire & Rodriguez, 2018; Burgoyne et al., 2014); those still waiting know that they are scheduled to receive a dog (Leighton, Rodriguez, Jensen, et al., 2024).
8.2 Missing Randomization and Blinding
The randomized PTSD trial compared two kinds of dog, not dog versus no dog (Richerson et al., 2023). Participants can hardly remain blinded once they live with a dog, but assessors and coders can: blinded clinicians (Leighton, Rodriguez, Jensen, et al., 2024), blind-coded CCTV (Wilson et al., 2019) and masked CGM (Los et al., 2017). Where such measures were used, alert accuracy was lower or more variable than in owner records (how observer bias and blinding affect measurement).
8.3 Small Samples
Many studies are small: 12 veterans (Kloep et al., 2017), five or six dogs in scent studies (Hardin et al., 2015; Catala et al., 2019), eight owners (Los et al., 2017; Wilson et al., 2019). Such studies show what is possible, not how often it works in everyday settings.
8.4 Involvement of Training Organizations
Training organizations appear as employers of authors (Rooney et al., 2019; Wilson et al., 2019), funders (Catala et al., 2019), source of the dogs (Maa et al., 2021) or sole recruitment source (O'Haire & Rodriguez, 2018; Leighton, Rodriguez, Jensen, et al., 2024). Industry ties exist on the other side too (Los et al., 2017). This does not make results wrong, but it calls for disclosure and independent replication.
8.5 Concentration in Few Research Groups
Much of the PTSD evidence, from waitlist studies to the meta-analysis, comes from one research group (O'Haire & Rodriguez, 2018; Jensen et al., 2022; Leighton et al., 2022; Leighton, Rodriguez, Jensen, et al., 2024). Reviews remain cautious: one identified no undisputed empirical evidence for PTSD service dogs (van Houtert et al., 2018), another found a lack of compelling evidence for unique benefits of autism assistance dogs (Mulraney et al., 2025), a third noted the generally poor rigor of earlier seizure-alert studies (Catala et al., 2018), and a broad assistance-dog review found 68% of statistical comparisons without significant effects (Rodriguez et al., 2020).
9. Practical Consequences for Trainers and Programs
9.1 Selection and Development
The studies support assessing dogs repeatedly over time rather than with one test, since temperament scales, cognitive tests and later assessments predicted success to some degree (Harvey et al., 2017; MacLean & Hare, 2018; Batt et al., 2008). If about 30% of dogs do not succeed, as in one US guide dog program (Bray et al., 2017b), programs need a plan for dogs that do not qualify. Raising deserves as much attention as selection, because frightening experiences and inexperienced raisers were associated with worse outcomes and fear was a common reason for withdrawal (Serpell & Duffy, 2016; Caron-Lormier, Harvey, et al., 2016b) (why the sensitive period matters for later behavior).
9.2 Training and Placement
The evidence is consistent with reward-based everyday handling and careful matching: less use of dominance-based methods and lower dog excitability went with better outcomes in PTSD teams (Jensen et al., 2022), and alert performance with owners' adherence to the training protocol (Wilson et al., 2019). Training the person is part of training the team. Clients also deserve honest information: no significant primary-outcome difference from emotional support dogs in the randomized PTSD comparison (Richerson et al., 2023), no shown advantage over companion dogs in autism (Mulraney et al., 2025), widely varying diabetes alert accuracy (Los et al., 2017; Rooney et al., 2019) and no prospectively validated seizure prediction in daily life (Martos Martinez-Caja et al., 2019; Catala et al., 2019).
9.3 Monitoring the Dog
Because the dog's side is hardly measured, programs must monitor it themselves. The stressors found in autism families (Burrows et al., 2008) are a useful checklist for follow-up visits in any placement. Guide dog records point to musculoskeletal and skin conditions and to behavioral problems as early warning areas (Caron-Lormier, England, et al., 2016; Caron-Lormier, Harvey, et al., 2016b), and the end of a partnership belongs in the plan from the start (Lloyd et al., 2021).
In practice: Keep a record for each placed dog with follow-up dates, body condition, musculoskeletal or skin problems, behavior at home and in public, and the handler's account of rest and free time, and define in advance which changes lead to an earlier visit, less work or retirement. This is a practice recommendation, not a study result.
10. Summary at a Glance
For PTSD service dogs, waitlist studies report fewer symptoms and better functioning, and a nonrandomized trial found lower severity after three months, also on blinded clinician interviews (O'Haire & Rodriguez, 2018; Leighton, Rodriguez, Jensen, et al., 2024). The randomized trial comparing service dogs with emotional support dogs found no significant difference in functioning and quality of life on its primary outcomes (Richerson et al., 2023). For autism, parents report benefits for family life, but nonrandomized comparisons do not establish causality, and a 2025 review found pooled effects similar to those of companion dogs (Burgoyne et al., 2014; Mulraney et al., 2025).
Diabetes alert figures are not directly comparable: the positive predictive value of hypoglycemia alerts was 12% against blinded CGM, while owner records from a training charity whose staff co-authored the study showed a median of 81% for alerts at high or low glucose (Los et al., 2017; Rooney et al., 2019). Dogs can distinguish seizure odor samples, but prediction is not established (Catala et al., 2019). In guide dogs, maternal style, cognition and temperament are associated with training success to some degree, and in a UK analysis limited to age retirements and behavioral withdrawals, 17% were behavioral, not an overall withdrawal rate (Bray et al., 2017b; Harvey et al., 2017; Caron-Lormier, Harvey, et al., 2016b). Dog welfare data are limited to two small cortisol studies without significant differences on their main comparisons and one qualitative study (van Houtert et al., 2021, 2023; Burrows et al., 2008).
11. Limitations and Research Gaps
11.1 What Training Adds
Whether trained tasks add anything beyond a dog's presence is a central open question (Richerson et al., 2023; Mulraney et al., 2025). No study included here combined random assignment, a no-dog group and blinded outcome assessment.
11.2 Follow-Up and Self-Report
Controlled follow-up ranges from three months (Leighton, Rodriguez, Jensen, et al., 2024) to 18 months (Richerson et al., 2023). Most outcomes are self-reported, and physiological measures in people either did not differ significantly or are difficult to interpret (van Houtert et al., 2022; Rodriguez et al., 2018).
11.3 Alert Accuracy
Diabetes alert accuracy depends on the comparison standard and the provider (Lippi & Plebani, 2019). Seizure evidence rests on small odor studies and uncontrolled field reports (Catala et al., 2018, 2019; Maa et al., 2021; Strong et al., 2002); no study in this review tested live prediction against objective seizure monitoring.
11.4 Dog Welfare
Welfare data are limited to two small cortisol studies, one qualitative study and guide dog records of withdrawal (van Houtert et al., 2021, 2023; Burrows et al., 2008; Caron-Lormier, Harvey, et al., 2016a, 2016b). Workload, recovery, public access and retirement remain unmeasured.
11.5 Populations and Settings
Many studies come from the United States and the United Kingdom, and public-access data only from the United States (Leighton et al., 2025). Hearing and mobility dogs are covered mainly by reviews (Rodriguez et al., 2020; Winkle et al., 2012), and from German-speaking countries the only user study here is an Austrian survey of 36 people (Glenk et al., 2019).
12. Conclusion
The research on assistance dogs is more encouraging than skeptics assume and weaker than providers often claim. Users frequently report benefits, and a nonrandomized controlled trial with blinded ratings found lower PTSD severity with a service dog (Leighton, Rodriguez, Jensen, et al., 2024). However, a randomized PTSD trial showed no significant group difference on its primary outcomes, and independently measured diabetes alert accuracy was variable (Richerson et al., 2023; Los et al., 2017).
For trainers and programs, this means selecting and raising dogs carefully, training handlers as thoroughly as dogs, informing clients honestly and monitoring each dog's welfare over its working life. The dogs carry much of the load in these partnerships, and research has measured very little of it.
Key Insights (Takeaways)
The only randomized trial comparing PTSD service dogs with emotional support dogs found no difference between service dogs and emotional support dogs on functioning and quality of life, with a 3.7-point greater PTSD symptom reduction in the service-dog group (Richerson et al., 2023).
A nonrandomized trial with 156 veterans found lower PTSD severity after three months with a service dog, including on clinician interviews blinded to allocation (Leighton, Rodriguez, Jensen, et al., 2024).
In children with autism, parents with assistance dogs reported greater perceived safety and more positive experiences in public, but the nonrandomized design cannot isolate a causal effect of the dog, and a 2025 review found pooled effects similar to those of companion dogs (Burgoyne et al., 2014; Mulraney et al., 2025).
Diabetes alert figures measure different things: 12% for hypoglycemia alerts against blinded CGM, a median of 81% for alerts at high or low glucose in owner records, and 69.7% or 50.4% in one study depending on the glucose range used (Los et al., 2017; Rooney et al., 2019; Wilson et al., 2019).
Dogs distinguished seizure odor samples under controlled conditions, but this does not establish reliable prediction of individual seizures in everyday life (Catala et al., 2019).
In one US guide dog program, about 70% of dogs entering training succeed, and in UK records comparing only age retirements with behavioral withdrawals, 17% were behavioral, not an overall withdrawal rate (Bray et al., 2017b; Caron-Lormier, Harvey, et al., 2016b).
Welfare data on assistance dogs are scarce: two small cortisol studies in PTSD service dogs found no effect, and one qualitative study described lack of rest and prodding by children in autism families (van Houtert et al., 2021, 2023; Burrows et al., 2008).
References
Batt, L. S., Batt, M. S., Baguley, J. A., & McGreevy, P. D. (2008). Factors associated with success in guide dog training. Journal of Veterinary Behavior, 3(4), 143–151. https://doi.org/10.1016/j.jveb.2008.04.003
Bray, E. E., Otto, C. M., Udell, M. A. R., Hall, N. J., Johnston, A. M., & MacLean, E. L. (2021). Enhancing the selection and performance of working dogs. Frontiers in Veterinary Science, 8, 644431. https://doi.org/10.3389/fvets.2021.644431
Bray, E. E., Sammel, M. D., Cheney, D. L., Serpell, J. A., & Seyfarth, R. M. (2017a). Characterizing early maternal style in a population of guide dogs. Frontiers in Psychology, 8, 175. https://doi.org/10.3389/fpsyg.2017.00175
Bray, E. E., Sammel, M. D., Cheney, D. L., Serpell, J. A., & Seyfarth, R. M. (2017b). Effects of maternal investment, temperament, and cognition on guide dog success. Proceedings of the National Academy of Sciences, 114(34), 9128–9133. https://doi.org/10.1073/pnas.1704303114
Burgoyne, L., Dowling, L., Fitzgerald, A., Connolly, M., Browne, J. P., & Perry, I. J. (2014). Parents' perspectives on the value of assistance dogs for children with autism spectrum disorder: A cross-sectional study. BMJ Open, 4(6), e004786. https://doi.org/10.1136/bmjopen-2014-004786
Burrows, K. E., Adams, C. L., & Millman, S. T. (2008). Factors affecting behavior and welfare of service dogs for children with autism spectrum disorder. Journal of Applied Animal Welfare Science, 11(1), 42–62. https://doi.org/10.1080/10888700701555550
Caron-Lormier, G., England, G. C. W., Green, M. J., & Asher, L. (2016). Using the incidence and impact of health conditions in guide dogs to investigate healthy ageing in working dogs. The Veterinary Journal, 207, 124–130. https://doi.org/10.1016/j.tvjl.2015.10.046
Caron-Lormier, G., Harvey, N. D., England, G. C. W., & Asher, L. (2016a). A new metric for quantifying the relative impact of risk factors on loss of working life illustrated in a population of working dogs. PLOS ONE, 11(11), e0165414. https://doi.org/10.1371/journal.pone.0165414
Caron-Lormier, G., Harvey, N. D., England, G. C. W., & Asher, L. (2016b). Using the incidence and impact of behavioural conditions in guide dogs to investigate patterns in undesirable behaviour in dogs. Scientific Reports, 6, 23860. https://doi.org/10.1038/srep23860
Catala, A., Cousillas, H., Hausberger, M., & Grandgeorge, M. (2018). Dog alerting and/or responding to epileptic seizures: A scoping review. PLOS ONE, 13(12), e0208280. https://doi.org/10.1371/journal.pone.0208280
Catala, A., Grandgeorge, M., Schaff, J.-L., Cousillas, H., Hausberger, M., & Cattet, J. (2019). Dogs demonstrate the existence of an epileptic seizure odour in humans. Scientific Reports, 9, 4103. https://doi.org/10.1038/s41598-019-40721-4
Dalziel, D. J., Uthman, B. M., McGorray, S. P., & Reep, R. L. (2003). Seizure-alert dogs: A review and preliminary study. Seizure, 12(2), 115–120. https://doi.org/10.1016/S105913110200225X
Glenk, L. M., Přibylová, L., Stetina, B. U., Demirel, S., & Weissenbacher, K. (2019). Perceptions on health benefits of guide dog ownership in an Austrian population of blind people with and without a guide dog. Animals, 9(7), 428. https://doi.org/10.3390/ani9070428
Gonder-Frederick, L. A., Grabman, J. H., & Shepard, J. A. (2017). Diabetes alert dogs (DADs): An assessment of accuracy and implications. Diabetes Research and Clinical Practice, 134, 121–130. https://doi.org/10.1016/j.diabres.2017.09.009
Gonder-Frederick, L. A., Grabman, J. H., Shepard, J. A., Tripathi, A. V., Ducar, D. M., & McElgunn, Z. R. (2017). Variability of diabetes alert dog accuracy in a real-world setting. Journal of Diabetes Science and Technology, 11(4), 714–719. https://doi.org/10.1177/1932296816685580
Hardin, D. S., Anderson, W., & Cattet, J. (2015). Dogs can be successfully trained to alert to hypoglycemia samples from patients with type 1 diabetes. Diabetes Therapy, 6(4), 509–517. https://doi.org/10.1007/s13300-015-0135-x
Harvey, N. D., Craigon, P. J., Blythe, S. A., England, G. C. W., & Asher, L. (2017). An evidence-based decision assistance model for predicting training outcome in juvenile guide dogs. PLOS ONE, 12(6), e0174261. https://doi.org/10.1371/journal.pone.0174261
Jensen, C. L., Rodriguez, K. E., MacLean, E. L., Abdul Wahab, A. H., Sabbaghi, A., & O'Haire, M. E. (2022). Characterizing veteran and PTSD service dog teams: Exploring potential mechanisms of symptom change and canine predictors of efficacy. PLOS ONE, 17(7), e0269186. https://doi.org/10.1371/journal.pone.0269186
Kloep, M. L., Hunter, R. H., & Kertz, S. J. (2017). Examining the effects of a novel training program and use of psychiatric service dogs for military-related PTSD and associated symptoms. American Journal of Orthopsychiatry, 87(4), 425–433. https://doi.org/10.1037/ort0000254
Leighton, S. C., Hofer, M. E., Miller, C. A., Mehl, M. R., Walker, T. D., MacLean, E. L., & O'Haire, M. E. (2025). Obstacles to inclusion and threats to civil rights: An integrative review of the social experiences of service dog partners in the United States. PLOS ONE, 20(3), e0313864. https://doi.org/10.1371/journal.pone.0313864
Leighton, S. C., Nieforth, L. O., & O'Haire, M. E. (2022). Assistance dogs for military veterans with PTSD: A systematic review, meta-analysis, and meta-synthesis. PLOS ONE, 17(9), e0274960. https://doi.org/10.1371/journal.pone.0274960
Leighton, S. C., Rodriguez, K. E., Jensen, C. L., MacLean, E. L., Davis, L. W., Ashbeck, E. L., Bedrick, E. J., & O'Haire, M. E. (2024). Service dogs for veterans and military members with posttraumatic stress disorder: A nonrandomized controlled trial. JAMA Network Open, 7(6), e2414686. https://doi.org/10.1001/jamanetworkopen.2024.14686
Leighton, S. C., Rodriguez, K. E., Zhuang, R., Jensen, C. L., Miller, E. A., Sabbaghi, A., & O'Haire, M. E. (2024). Psychiatric service dog placements are associated with better daily psychosocial functioning for military veterans with PTSD. Psychological Trauma: Theory, Research, Practice, and Policy, 16(Suppl. 3), S707–S717. https://doi.org/10.1037/tra0001543
Lippi, G., & Plebani, M. (2019). Diabetes alert dogs: A narrative critical overview. Clinical Chemistry and Laboratory Medicine, 57(4), 452–458. https://doi.org/10.1515/cclm-2018-0842
Lloyd, J., Budge, C., La Grow, S., & Stafford, K. (2016). An investigation of the complexities of successful and unsuccessful guide dog matching and partnerships. Frontiers in Veterinary Science, 3, 114. https://doi.org/10.3389/fvets.2016.00114
Lloyd, J., Budge, C., La Grow, S., & Stafford, K. (2021). The end of the partnership with a guide dog: Emotional responses, effects on quality of life and relationships with subsequent dogs. Frontiers in Veterinary Science, 8, 543463. https://doi.org/10.3389/fvets.2021.543463
Los, E. A., Ramsey, K. L., Guttmann-Bauman, I., & Ahmann, A. J. (2017). Reliability of trained dogs to alert to hypoglycemia in patients with type 1 diabetes. Journal of Diabetes Science and Technology, 11(3), 506–512. https://doi.org/10.1177/1932296816666537
Maa, E., Arnold, J., Ninedorf, K., & Olsen, H. (2021). Canine detection of volatile organic compounds unique to human epileptic seizure. Epilepsy & Behavior, 115, 107690. https://doi.org/10.1016/j.yebeh.2020.107690
MacLean, E. L., & Hare, B. (2018). Enhanced selection of assistance and explosive detection dogs using cognitive measures. Frontiers in Veterinary Science, 5, 236. https://doi.org/10.3389/fvets.2018.00236
Martos Martinez-Caja, A., De Herdt, V., Boon, P., Brandl, U., Cock, H., Parra, J., Perucca, E., Thadani, V., & Moons, C. P. H. (2019). Seizure-alerting behavior in dogs owned by people experiencing seizures. Epilepsy & Behavior, 94, 104–111. https://doi.org/10.1016/j.yebeh.2019.02.001
Mulraney, M., Rincones, O., Bryant, C., Saredakis, D., Ghezzi, E., & Lampit, A. (2025). A mixed methods systematic review of assistance dogs for people with autism. Neuroscience & Biobehavioral Reviews, 173, 106160. https://doi.org/10.1016/j.neubiorev.2025.106160
O'Haire, M. E., & Rodriguez, K. E. (2018). Preliminary efficacy of service dogs as a complementary treatment for posttraumatic stress disorder in military members and veterans. Journal of Consulting and Clinical Psychology, 86(2), 179–188. https://doi.org/10.1037/ccp0000267
Richerson, J. T., Wagner, T. H., Abrams, T., Skelton, K., Biswas, K., Illarmo, S., McSherry, F., Fallon, M. T., Frakt, A., Pizer, S., Magruder, K. M., Groer, S., Dorn, P. A., Huang, G. D., & Stock, E. M. (2023). Therapeutic and economic benefits of service dogs versus emotional support dogs for veterans with PTSD. Psychiatric Services, 74(8), 790–800. https://doi.org/10.1176/appi.ps.20220138
Rodriguez, K. E., Bryce, C. I., Granger, D. A., & O'Haire, M. E. (2018). The effect of a service dog on salivary cortisol awakening response in a military population with posttraumatic stress disorder (PTSD). Psychoneuroendocrinology, 98, 202–210. https://doi.org/10.1016/j.psyneuen.2018.04.026
Rodriguez, K. E., Greer, J., Yatcilla, J. K., Beck, A. M., & O'Haire, M. E. (2020). The effects of assistance dogs on psychosocial health and wellbeing: A systematic literature review. PLOS ONE, 15(12), e0243302. https://doi.org/10.1371/journal.pone.0243302
Rooney, N. J., Guest, C. M., Swanson, L. C. M., & Morant, S. V. (2019). How effective are trained dogs at alerting their owners to changes in blood glycaemic levels? Variations in performance of glycaemia alert dogs. PLOS ONE, 14(1), e0210092. https://doi.org/10.1371/journal.pone.0210092
Rooney, N. J., Morant, S., & Guest, C. (2013). Investigation into the value of trained glycaemia alert dogs to clients with type I diabetes. PLOS ONE, 8(8), e69921. https://doi.org/10.1371/journal.pone.0069921
Saunders, G. H., Biswas, K., Serpi, T., McGovern, S., Groer, S., Stock, E. M., Magruder, K. M., Storzbach, D., Skelton, K., Abrams, T., McCranie, M., Richerson, J., Dorn, P. A., Huang, G. D., & Fallon, M. T. (2017). Design and challenges for a randomized, multi-site clinical trial comparing the use of service dogs and emotional support dogs in veterans with post-traumatic stress disorder (PTSD). Contemporary Clinical Trials, 62, 105–113. https://doi.org/10.1016/j.cct.2017.08.017
Serpell, J. A., & Duffy, D. L. (2016). Aspects of juvenile and adolescent environment predict aggression and fear in 12-month-old guide dogs. Frontiers in Veterinary Science, 3, 49. https://doi.org/10.3389/fvets.2016.00049
Strong, V., Brown, S., Huyton, M., & Coyle, H. (2002). Effect of trained seizure alert dogs on frequency of tonic–clonic seizures. Seizure, 11(6), 402–405. https://doi.org/10.1053/seiz.2001.0656
U.S. Department of Justice, Civil Rights Division. (2020, February 28). ADA requirements: Service animals [Government document]. https://www.ada.gov/resources/service-animals-2010-requirements/
van Houtert, E. A. E., Endenburg, N., Rodenburg, T. B., & Vermetten, E. (2021). Do service dogs for veterans with PTSD mount a cortisol response in response to training? Animals, 11(3), 650. https://doi.org/10.3390/ani11030650
van Houtert, E. A. E., Endenburg, N., Vermetten, E., & Rodenburg, T. B. (2023). Hair cortisol in service dogs for veterans with post-traumatic stress disorder compared to companion dogs (Canis familiaris). Journal of Applied Animal Welfare Science, 26(4), 623–633. https://doi.org/10.1080/10888705.2022.2033119
van Houtert, E. A. E., Endenburg, N., Wijnker, J. J., Rodenburg, B., & Vermetten, E. (2018). The study of service dogs for veterans with Post-Traumatic Stress Disorder: A scoping literature review. European Journal of Psychotraumatology, 9(Suppl. 3), 1503523. https://doi.org/10.1080/20008198.2018.1503523
van Houtert, E. A. E., Rodenburg, T. B., Vermetten, E., & Endenburg, N. (2022). The impact of service dogs on military veterans and (ex) first aid responders with post-traumatic stress disorder. Frontiers in Psychiatry, 13, 834291. https://doi.org/10.3389/fpsyt.2022.834291
Whitmarsh, L. (2005). The benefits of guide dog ownership. Visual Impairment Research, 7(1), 27–42. https://doi.org/10.1080/13882350590956439
Wilson, C., Morant, S., Kane, S., Pesterfield, C., Guest, C., & Rooney, N. J. (2019). An owner-independent investigation of diabetes alert dog performance. Frontiers in Veterinary Science, 6, 91. https://doi.org/10.3389/fvets.2019.00091
Winkle, M., Crowe, T. K., & Hendrix, I. (2012). Service dogs and people with physical disabilities partnerships: A systematic review. Occupational Therapy International, 19(1), 54–66. https://doi.org/10.1002/oti.323