E

C

S

a

b

D

RA

B

Tnc

f

l

0d

pilepsy Research (2011) 97, 236—242

j ourna l ho me pag e: www.elsev ier .com/ locate /ep i lepsyres

an seizure-alert dogs predict seizures?

tephen W. Browna, Laura H. Goldsteinb,∗

Peninsula College of Medicine & Dentistry, Chy Govenek, Threemilestone, Truro TR4 9LD, Cornwall, UKMRC Centre for Neurodegeneration Research, Department of Psychology, PO77, Institute of Psychiatry, King’s College London,e Crespigny Park, London SE5 8AF, UK

eceived 8 April 2011; received in revised form 11 October 2011; accepted 12 October 2011vailable online 1 November 2011

KEYWORDSEpilepsy;Seizure-alert dogs;Animal-assistedtherapy

Summary An index observation where a dog was trained to alert to, as well as respond to,human tonic—clonic seizures led to further research and refinement of training techniques.This was followed by anecdotal reports of pet dogs spontaneously anticipating human epilepticseizures. An industry has since developed training Seizure-Alert Dogs (SADs) to give humanswarnings of their seizures. In some cases this has been accompanied by a reduction in seizurefrequency. SADs may be trained along with the person with epilepsy, responding specifically tothat person’s seizures, or may be trained separately. Recent sceptical reports of non-epilepticseizures in some people with SADs have cast doubt on dogs’ ability to anticipate true epilepticseizures. This may reflect selection criteria for training programmes as well as training methodsused, but does not necessarily indicate that SADs might not be able to predict epileptic seizures.

Whether the seizures are epileptic or non-epileptic, it is speculated that SADs probably alert tosubtle pre-ictal human behaviour changes, but may also be sensitive to heart rate or olfactorycues. As yet, however, no rigorous data exist as to whether seizure prediction by SADS is betterthan chance, and what false positive and negative prediction rates might be.© 2011 Elsevier B.V. All rights reserved.

ac

ackground

he therapeutic use of companion animals in assisting diag-osis, treatment and management of various human healthonditions, and in supporting carers, has attracted a certain

∗ Corresponding author. Tel.: +44 020 7848 0218;ax: +44 020 7848 5506.

E-mail addresses: stephen.brown@cantab.net (S.W. Brown),aura.goldstein@kcl.ac.uk (L.H. Goldstein).

f‘captvct

920-1211/$ — see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.eplepsyres.2011.10.019

mount of interest, although rigorous clinical trial data isurrently lacking. The specific use of dogs in these situationsalls into two main categories, ‘‘response’’ and ‘‘alert’’.‘Response’’ dogs assist the human by behaving in a spe-ific and useful way when a particular event occurs, suchs drawing the hearing-impaired human’s attention to theresence of a visitor at the door. ‘‘Alert’’ dogs are saido anticipate certain types of impending events and pro-

ide a useful warning to the human. As far as epilepsy isoncerned dogs used in these two situations are referredo as Seizure Response Dogs (SRDs) who display specific

paAseMcmaatdbbbap(nbs1sf(o4aitfifif

aiawwfsT1tbnotrdwInufwaw

Can seizure-alert dogs predict seizures?

behaviours during or immediately after a human seizure,and Seizure Alert Dogs (SADs) who display specific behavioursprior to any appreciation of an impending seizure event bythe person with epilepsy or observer (Brown, 2005). Clini-cians specialising in epilepsy will be familiar with carers andrelatives of people with epilepsy occasionally making a claimthat they can detect changes in the person’s appearance,presentation or behaviour that are premonitory to seizureactivity even before the person is aware of the situation.Further discussion sometimes also reveals a claim that thefamily pet (usually, but not always a dog) may display spe-cific behaviours prior to the person having a seizure, whereno warning signs have been noticed by the person or theirhuman carers or family (Pinikahana and Dono, 2009). In suchcases the dog has spontaneously acquired the properties ofa SAD.

The UK experience with SADs began with a particularindex observation. A 28-year-old female with tonic—clonicseizures and a degree of physical disability asked for helpfrom a dog trainer in training her 18-month-old cross col-lie to assist with her post-ictal recovery. Such training wasrelatively straightforward, but on completion the trainerdecided to pursue the possibility of training the dog towarn the person that a seizure was imminent. Previouslythe dog had never shown any response to seizures. Afterthree months’ training, the dog (using a vocal signal) wasable to indicate that a seizure was imminent 20 min inadvance of a tonic—clonic seizure. Publicity from this caseled to many self-referrals to the trainer and provided casesfor a pilot study (Strong et al., 1999; Brown and Strong,2001).

Original studies

A pilot study (Strong et al., 1999) was carried out in col-laboration with the UK charity (not-for-profit organisation)Support Dogs based in Sheffield, UK and involved 6 self-referred individuals aged 9—52 (3 male, 3 female) withtonic—clonic seizures. In two cases the possibility of addi-tional psychogenic non-epileptic seizures (PNES) had beenraised, although the clinicians involved had not made a def-inite diagnosis other than epilepsy. In all cases, SADs weresuccessfully trained to provide a useful warning to the per-son of an impending seizure.

The authors felt that in these cases (both in trained dogsand in those who learn spontaneously) the mechanism isprobably based on canine sensitivity to what are often sub-tle behavioural changes in humans. Important points in thetraining protocol used in these cases included introducingthe selected dog to the human and socialising them togetherunder supervision before training the dog to recognise andanticipate the person’s seizures using reward-based operantconditioning. No adverse effects on canine or human healthor safety were observed.

The purpose was to provide the person with epilepsy withtime to avoid coming to harm as a result of a seizure and toseek a safe environment where appropriate. In addition to

this, however, all participants reported a decrease in seizurefrequency. This was an unexpected finding and had not orig-inally been regarded as an outcome. A further study with10 people was therefore undertaken (Strong et al., 2002)

ebfr

237

artly to explore refinement of the training procedure, butlso to investigate the possible effect on seizure frequency.

strict protocol was employed with every effort madepecifically to exclude participants who might have, or hadver been suspected of having PNES (Strong et al., 2002).edication was not changed during the study and inclusionriteria also required at least four tonic—clonic seizures peronth. Tonic—clonic seizure frequency was monitored over

48 week period including 12 weeks baseline after entry, 12 week training period, and 24 weeks follow-up. Using araining paradigm similar to that described below in moreetail, it was possible to train dogs to show seizure-alertingehaviour. Because of the nature of the training (reward-ased operant conditioning) the specific behaviour exhibitedy the dogs involved gaining the person’s attention as if innticipation of a reward. This might include blocking theerson’s way, jumping up, making eye contact, vocalisingsometimes including but not confined to barking) in a man-er which the person recognised as the reward anticipatoryehaviour that the dog demonstrated in association with aeizure. Comparing baseline seizure frequency to the last2 weeks of follow-up, there was a 43% mean reduction ineizure frequency (p = 0.002), with a mean monthly seizurerequency at baseline of 13.8 (range 6.3—45.6) falling to 8.5range 2—30) during the last 12 weeks of follow-up. Ninef the 10 participants showed a 34% or greater reduction,/10 showed a 50% or greater reduction in seizure frequencynd only one showed no improvement. A significant dropn seizure frequency was seen during the first 4 weeks ofraining (p = 0.0078), a further drop occurred between therst and last 4 week period of training (p = 0.038) and thisnal improvement was maintained for the whole 24 weekollow-up.

At this stage, one obvious next step would be to design larger, randomised trial with a control group compris-ng individuals with untrained pet dogs. However, the sameuthors also became aware of potential hazards of workingith untrained dogs. They reported 36 cases of pet dogsho suffered significant adverse health effects resulting

rom spontaneously reacting to, or anticipating, epilepticeizures in their human owners (Strong and Brown, 2000).hese included three cases in which the dog died and2 cases in which the dog exhibited aggressive behaviourowards humans. They also noted that where dogs haveeen specially trained as SADs such adverse effects hadot been seen. They stressed the importance of selectionf suitable dogs and provision of appropriate structuredraining from a behavioural specialist. Otherwise the dog’sesponse may be unreliable, the health and welfare of theog may be compromised and the safety of the personith epilepsy and the general public may be jeopardised.

n view of these reports the authors concluded ‘‘we doot feel that future research in this area could justifiablyse a control group with untrained dogs. . . on animal wel-are grounds. This obviously raises methodological issueshich will need to be addressed in due course.’’ (Strongnd Brown, 2000). This contrasts with Kirton et al. (2004)ho, in their study of 62 families with both a child with

pilepsy and at least one pet dog, found no obvious draw-acks. It is possible, therefore, that there may be a potentialor using untrained pet dogs as a control group, but cur-ently it is unclear how to avoid the potential problems

2

dbbd

S

Kttemtscct

S

Ddamatrtt

ddplww2ntldndidaM1dcdiatobcfuis

C

TobpCfsstitrmcsdbPgi

aPbttaHaahswtbpwnpsdtscnttNswttS

df

38

escribed by Strong and Brown (2000). It might be possi-le to design a randomised study with a waiting list control,ut that would fail to show any specific effects of trainedogs.

ome later studies

irton et al. (2008) described a retrospective study of araining program for SRDs with some similarities, both inraining technique and in outcome, to that used by Strongt al. (2002). Other than that, the subsequent literatureainly consists either of surveys of groups of families

o solicit observations of pet dogs’ natural responses toeizures, or of anecdotal retrospective reports of singleases or very small series, usually where the authors reportases involving PNES. There have unfortunately been no fur-her prospective series reporting on specific training.

urveys of groups of families

alziel et al. (2003) surveyed 63 patients of whom 29 hadogs. Nine of these dogs showed SRD behaviour and threelso had SAD features. Those patients with SADs seemedore likely to have complex partial seizures and auras

nd this raises the possibility that the ‘alert’ anticipa-ory behaviour of the dog might actually be a behaviouralesponse to an aura. Nevertheless the authors consideredhe findings to suggest that some dogs have an innate abilityo alert and/or respond to seizures.

Kirton et al. (2004) in a Refractory Epilepsy Clinic for chil-ren aged 1—18 investigated children who had lived with aog for at least one year and who had at least one seizureer month. Of 122 families who responded, 48 (39%) had ateast one pet dog, the total number of dogs being 62. No dogsere specifically trained. Sixty-four percent of the childrenith dogs were female with a median age of 10. A total of2 dogs from 20 families seemed to have learned sponta-eously to react to seizures in a stereotyped way and couldherefore be regarded as untrained SRDs. These tended to bearger animals (weighing >40 lb) compared to non-responsiveogs and were mostly cross breeds, although there waso difference in breed between SRDs and non-responsiveogs. The most common seizure response behaviour exhib-ted by SRDs was licking, especially of the face, which wasescribed in 13 of the 22 cases. Nine of the 22 SRDs werelso SADs. These tended to be larger animals, mainly female.edian anticipation time was described as 2.5 min (range0 s—5 h). Although these often very long prediction win-ows may tend to be undefined, and the authors have notonsidered the potential for correcting statistically for ran-om prediction with such intervals, Kirton et al. (2004) alsondicate that the reported accuracy of alerting was high with

median sensitivity estimate of 80%. They also reportedhat anticipatory behaviours were never demonstrated with-ut a seizure then occurring. However, response bias cannote ruled out without more rigorous methodology. Sixty per-ent of families thought this ability in their dog was present

rom the first seizure. SAD behaviours were described assually ‘‘protective’’. There was some evidence that qual-ty of life for the children was higher where there was aeizure-sensitive dog in the family.

itPs

S.W. Brown, L.H. Goldstein

ontroversies

here are unfortunately very few reports of EEG orther physiological monitoring during alert and responseehaviours. Ortiz and Liporace (2005) were able to study twoatients with North-American trained SADs in an Epilepsyare Unit undergoing video EEG telemetry. One patient hadour seizures while awake and four while sleeping. The SADlept through seven of these but did alert for the eightheizure. For other epileptic events the dog awoke duringhe seizures and then alerted family members by bark-ng/walking round the bed. The behaviour described washerefore of the SRD type even though the animal waseferred to as a SAD. The dog in this case did not seem toonitor the patient during the hours of normal sleep, in

ontrast to the behaviour described in UK-trained SADs. Theecond patient had a North-American trained SAD presenturing one of her seizures which alerted her seven minutesefore the seizure, but telemetry confirmed this to be aNES. Studies of this type are likely to be rare, however,iven the practical difficulties posed by having dogs presentn seizure monitoring units (Ortiz and Liporace, 2005).

There have been other cases observed where dogslert to PNES in circumstances where the person withNES believed themselves to have epilepsy and thereforeelieved the dog was a SRD. The subsequent discovery thathe seizures were non-epileptic has caused some workerso exhibit scepticism about the possible significance of SRDsnd the existence of true SADs. For example, Doherty andaltiner (2007) described a two year-old Blue Heeler wholerted and responded to seizures by laying on the patientcross her chest prior to and during seizures and fetchedelp from neighbours after seizures. The dog was alsoaid to predict the patient’s husband’s seizures, althoughhether or not these were epileptic is unclear. Video EEG

elemetry captured events including pelvic thrusting, headobbling movements, psychomotor unresponsiveness andost-event flaccidity and led to a diagnosis of PNES. The dogas not present during the monitoring, however and it isot stated whether the recorded events were typical of theatient’s seizures. Krauss et al. (2007) reported a series ofix out of nine patients in which ‘‘seizure response serviceogs’’ alerted to non-epileptic events. In one of these caseshe dog abruptly licked the patient’s face 30 s prior to hiseizure (observed during telemetry), while in a furtherase the dog was reported to lick the patient’s head andeck to alert him, although the dog was not present duringelemetry. Unfortunately the paper does not describe howhe dogs were trained and since some, and probably most,orth American SRDs/SADs are trained by using actors withimulated seizures rather than being socialised specificallyith the person with epilepsy, it should not be surprising

hat the dogs respond to non-epileptic seizures. Thereforehis does not provide any useful information in relation toRDs/SADs in epilepsy.

The behaviour of the dogs in Krauss et al.’s (2007) reportoes raise a question as to whether the dog’s response rein-orce PNES occurrence (Litt and Krieger, 2007). For example

f the dog licks the person’s face or neck prior to PNEShis might conceivably be an event that could trigger aNES. Indeed, similar issues arise when considering epilepticeizures which could possibly be conditioned in this way.

siOstpsabpaidDmw

dittcoptbin

£

T

Scfpauhdgdu

A

Amavttasso

Can seizure-alert dogs predict seizures?

Training protocol used by support dogs

This is the current method used by Support Dogs of SheffieldUK which is essentially that used in the two original studies(Strong et al., 1999, 2002). It differs from techniques usedin other countries in the emphasis on socialising and train-ing the individual dog with the person with epilepsy fromthe beginning. People with epilepsy often self-refer to thecharity.

Inclusion and exclusion criteria

Applications must be supported by the physician treat-ing the epilepsy and strict inclusion criteria are applied.There must be a confirmed diagnosis of epilepsy with atleast 10 seizures per month which may consist of gener-alised tonic—clonic seizures, atonic seizures or complexpartial seizures, not including seizures occurring in sleep.This is a stringent inclusion criterion, but allows adequateseizure occurrence to facilitate dog training. This raisesthe potential for improvement in seizure frequency to rep-resent ‘‘regression to the mean’’, which can only reallybe evaluated properly in a randomised trial. However aprevious reported study from this centre (Strong et al.,2002) from which data were presented earlier, used 4 ormore tonic—clonic seizures/month as part of their inclu-sion criteria, and inspection of the raw data provided inthe published paper indicates a similar pattern of resultsif those with 4—10 seizures per month at entry are con-sidered separately. Medical reports are obtained from theGeneral Practitioner and the hospital specialist treating theepilepsy. If the applicant has been regarded as having PNESit is not possible for them to enter the program. The appli-cant must be able to keep an accurate seizure diary, mustbe currently over 16 years of age, and there must be ahuman carer available. There must be adequate supportin the domestic situation to ensure that the dog’s needscan be met. There must be no changes to antiepilepticdrug therapy for six months prior to enrolment and through-out the assessment training period. Enrolment may not bepossible where there is already a pet dog in the family,and the person cannot have their own pet dog trained.The person must genuinely enjoy the close companionshipof a dog and want to form a working relationship and beable to care for dog adequately. Training currently may notbe possible if there is a vagus nerve stimulator device inplace.

Training program

If these criteria are met, both from the person’s point ofview and after obtaining medical reports, including specificexclusion of PNES, the applicant is offered an initial inter-view with training staff. At this, the person is assessed fortheir commitment and ability to care for the dog and moreinformation on seizure phenomenology and about the per-son is collected. If no contraindications are identified, the

applicant is offered a two day admission to assess how theyinteract with and react to dogs generally, and what typesof dog they may best bond with. If this initial admission issatisfactory, the person is placed on a waiting list until a

Tvpb

239

uitable dog for that person becomes available, the suitabil-ty being based on the results of the two-day assessment.nce a dog is identified, there is another two-day admis-ion with the selected dog where the trainer checks thathe person and the dog can bond with each other. If theotential to bond is satisfactorily demonstrated, the nexttep is a three-week admission to try and commence oper-nt conditioning in the context of increasing development ofonding. The dog is trained to make an extreme focus on theerson’s face, referred to as ‘‘look at me’’ training. When

seizure occurs, the dog is rewarded. The reward, admin-stered by the trainer is usually a food treat that varies withog, and this particular reward is only used for seizures.epending on how many seizures occur, the dog may oray not be acquiring a seizure alert pattern after threeeeks.

This admission is followed by a home placement for theog with the person. Video camera footage of daily activitiesn the home and continuance of the reward programme byhe carer are continued for about two months after whichhere is a further one week admission to the assessmententre to review progress, top-up any training and use thepportunity to review the video record to check that therogramme is being optimally applied. The expectation ofhe trainers is that the dog should be alerting to seizuresy then. Further follow-up depends on progress, but theres opportunity for further admission for top-up training ifecessary.

The program described above is currently costed at about10,000 (∼$15,000).

ypes of dogs

eizure Alert Dogs are drawn from three sources; dog res-ue centres, dogs otherwise specially donated to the charityrom various sources and ‘career change’ dogs who werereviously involved as disability assistance dogs but whore no longer required in that role. All dogs are individ-ally assessed by trainers for their ability to bond withumans and every effort is made to match up individualogs with suitable applicants. To date, no specific breed orender seems to be preferred; successful training is felt toepend on the person—dog bonding, which is very individ-al.

lert behaviour

lthough there are individual variations, in a typical pre-onitory alerting behaviour the dog will seek out the person

nd engage their attention by getting in their field ofision, staring at them and sometimes barking. In the ini-ial studies which concentrated on tonic—clonic seizures,he warning usually came 15—45 min before the observ-ble seizure onset (Strong et al., 1999, 2002). Trainerstill regard the mean anticipation period for tonic—cloniceizures to be about 30 min, but shorter times, in therder of 15 min are observed with complex partial seizures.

he action taken by the human after the dog alertsaries, and may for example include lying down andlacing a cushion under the head, or going to lie on aed.

2

S

Tawwptotgpdtottbil

S

N

IeKtTbnmfstpashspawa

aatcg

T

TvsChb

tbImbnsacar

ipacorcswtMbip

O

AisHirasnsnoMrottstcer

lliIe

40

ome questions answered

wo particular observations invite comment and occasion-lly lead to unhelpful speculation about the phenomena tohich dogs might be alerting. The first of these involves dogsho appear to alert from another room to that in which theerson is situated. The second refers to dogs that alert inhe night. Those of us with knowledge of the field have beenccasionally approached by journalists for opinions abouthe psychic abilities of dogs that these events seem to sug-est. Fortunately review of video evidence from the homelacement phase of training described above shows that aog situated in another room in the house typically entershe room and appears to check the human every 15 minr so, and therefore shows alert behaviour as a result ofheir regular checking and observation of the human, ratherhan entering the room because they are displaying alertehaviour. Likewise, video scrutiny shows that the sleep-ng pattern of SADs seems to include waking regularly andooking at the human (Strong, pers. com.).

ome unanswered questions

on-epileptic red herrings

f a dog can respond to or apparently predict PNES phenom-na (Ortiz and Liporace, 2005; Doherty and Haltiner, 2007;rauss et al., 2007) this does not necessarily mean thathe dog cannot or does not respond to epileptic seizures.he issue might be more clearly realised as being about theehaviour they are trained to detect rather than whether orot that behaviour is part of an epileptic seizure or a PNESanifestation, even though their semiology may clearly dif-

er (Mellers, 2005). Conditioning principles involved in eithereizure alerting or responding would apply equally to bothypes of phenomena. There is a question as to whether peo-le with abnormal illness behaviours are more likely to seeknimals for support, as suggested by Krauss et al. (2007),o that PNES may be over-represented in the population ofumans seizure support dogs of whatever type. It has beenuggested, however, that given the cost of training SADs,atients might be usefully screened to ensure that the dogsre being provided specifically for people with epilepsy forhom a life-saving role is particularly important (Dohertynd Haltiner, 2007).

Available case reports certainly suggest that PNES can bessociated with response behaviour in dogs. If there is suchn effect in PNES, however, it does not rule out effects inrue epilepsy and the thrust of at least some of the literatureould be described as unhelpful in assessing the effect inenuine epilepsy.

o what are the dogs alerting?

he consensus from those who spend time reviewing theideo evidence is that dogs probably alert to specific and

ubtle human behaviour (Strong, personal communication).ompanion dogs have evolved a close relationship withumans which is facilitated by close attention to humanehaviour and anticipation of aspects of human behaviour

bbaa

S.W. Brown, L.H. Goldstein

hat have consequences for the dog, such as going out oreing fed. Most people who have pet dogs will recognise this.t is perfectly possible that some relevant human behavioursight not be obvious to humans themselves but are notedy dogs because they signify anticipation of reward. It isot currently possible to state whether dogs are sensitive toubtle changes in human respiratory rate, or whether theircute hearing may enable sensitivity to human heart ratehanges, or whether some olfactory phenomenon may play

part. It is possible that more than one mechanism may beelevant.

The possibility has been raised that seizure activity man-festing as the behavioural component of a partial seizurerior to generalisation might account for at least some situ-tions. It could be speculated that odour or other autonomicharacteristics might be detected by a dog where humanbservers might lack sensitivity, e.g. changes in human heartate might be heard by a dog. Certainly changes in heart ratean precede the onset of a clinically observable seizure byeconds or minutes (Sevcencu and Struijk, 2010). Patientsith PNES are more likely than those with epileptic seizures

o report a range of autonomic symptoms (Goldstein andellers, 2006) including chest pain, palpitation, shortness ofreath and sweating just prior to or during their PNES. Theres clearly a need for more careful observation to identify thehenomena to which SADS are actually responding.

ther issues

part from the very few cases where dogs have been presentn clinical monitoring situations, all other accounts rely onelf-report, using diaries or questionnaires (Doherty andaltiner, 2007) which may raise legitimate issues of report-

ng bias in terms of whether all seizures are reported inelation to whether or not the dog alerts/responds. Inddition, people with SADs may be a highly selected orelf-selected sample (Kirton et al., 2008). This issue iso different from the consideration of potential bias inelf-reporting of seizures that might underlie whether orot adults appear to be able to predict their own seizureccurrence (Haut et al., 2007; Litt and Krieger, 2007).aiwald et al. (2011) studied a small number of patients who

eported being able to predict their seizures on the basisf prodromal symptoms. On the basis of prospective time-racked data using personal digital assistants they concludedhat such events did not provide a reliable and statisticallyignificant basis on which patients could necessarily predictheir own seizures. This argues strongly for more tightlyontrolled prospective studies of SADs’ alerting abilities,specially if such alerting abilities are thought to occur inesponse to pre-ictal changes in the person with epilepsy.

It is of course noteworthy that patients using SADs areikely to be those for whom conventional treatment has noted to satisfactory seizure control, and any improvementn seizures or associated quality of life may be beneficial.t may be the case that people with epilepsy may experi-nce reduced anxiety and increased independence if they

elieve their seizures may be pre-warned and that they cane alerted, for example, if outside their home. The ther-peutic effects of contact with dogs more generally hasroused interest in helping recovery from physical injury

uurfu

buaSibccscwsdc

tcobStsatoafctopt

G

Tr

1

2

3

Can seizure-alert dogs predict seizures?

(e.g. Hastings et al., 2008) and, for example, in improv-ing psychological function in elderly people with Alzheimer’sdisease (Mossello et al., 2011) and psychiatric disorders(Moretti et al., 2011).

It is interesting to consider the role that assistance dogsmay play in other disease areas. There are now diabetesassistance dogs who have allegedly been trained to alerttheir owners to hypoglycaemic episodes. It is suggested thatthey are responding to the subtle scent that hypoglycaemiacreates. There are no published studies, but the trainingorganisations suggest that their dogs are never 100% accu-rate in prediction, and quote a figure of 85—90% accuracy(Dogs4Diabetics; Alert Service Dogs). There are also CancerDetection Dogs that are claimed to be trained to detect thesmell of bladder cancer in urine. Otherwise most of the gen-eral literature on Assistance Dogs is mainly concerned withhelping people with mobility problems or sensory impair-ments (e.g. Sachs-Ericsson et al., 2002). As with SADs forpeople with epilepsy (Dalziel et al., 2003), the person withthe illness will have to learn to recognise the dog’s warningbehaviour, may not be able to take the dog everywhere withthem if buildings do not provide access to animals and willalso have to develop new relationships with replacementSADs when the dog reaches the end of its working life.

Summary and conclusions

Some domestic dogs seem to show identifiable behavioursin response to epileptic seizures in their significant humans.Of these, a proportion seem to be able to display premon-itory behaviour to seizures while the person with epilepsyand their carers are still unaware of the impending event,as described above. In at least some cases, untrained dogsmay react adversely to human seizures with negative con-sequences for both dog and human. It is not known howfrequent such adverse reactions are, or how to predict whichdogs will show adverse reactions.

Although there is a limited scientific literature on thistopic, it does seem that some dogs can be trained both torespond to and provide useful premonitory signals to humanseizures following specific individual training with the signif-icant human using a systematic operant conditioning basedapproach. Such cases may be associated with a decrease inreported seizure frequency. Therefore, the potential valueof dogs trained in this way lies firstly, in providing a usefulwarning of impending seizures so that a safe environmentcan be achieved and secondly, in providing an adjunct toother treatments such as antiepileptic drugs (AEDs).

If reliable warning of seizures can be given by a dog wherethe human previously could not self-predict seizures, it isfeasible that this could contribute to reduction in anxietyand help increase independence and quality of life. Takingpart in more activities in general may play a part in reduc-tion of seizure frequency. There is also the possibility thatthe alerting behaviour of the dog can increase the level ofhuman arousal and play a part in aborting a seizure, althoughwe currently lack data on the frequency of ‘false-positive’

alerts.

Some dogs are also able to show stereotyped responsepatterns to PNES. It is less clear whether in such casesreliable premonitory behaviour is also displayed. It is also

4

241

nclear whether the presence of a PNES alert dog has anyseful effect on PNES frequency, since most of the caseeports available refer to people with seizures severe andrequent enough to require intensive monitoring in specialistnits.

The observed reduction in seizure frequency in a num-er of cases where specially trained SADs are used makes itnlikely that the seizures are a consequence of the dog’slerting behaviour, especially since the training requiresADs to expect a reward when a seizure occurs. However,t is not clear to what trigger(s) SADs are responding. Subtleehavioural or mood changes in the human may be suffi-ient explanation in some cases, but the possibility that dogsan sense and respond to physiological changes in humans,uch as heart or respiratory rate or pheromone productionannot be excluded. Bearing in mind the variable length ofarning before seizures that is described, it would seem rea-

onable to allow for several mechanisms to be relevant, withifferent ones applying in different situations and possiblyorresponding to different seizure types.

The only systematic prospective studies using speciallyrained dogs have supported three hypotheses; that dogsan be trained as SADs, that this can be done safely with-ut hazard to dog or human, and that such training maye associated with reduction in reported seizure frequency.tarting with these observations there is therefore a needo design and carry out a larger scale controlled trial of thepecific training protocol described above. Careful consider-tion would have to be given to identifying the conditions forhe control group. There are no helpful systematic reportsf the outcome where dogs are trained by other means suchs with simulated seizures or with a person with epilepsy dif-erent to the intended significant human. Indeed, the onlyontribution these training methods have made to the scien-ific study of the subject may be the presumably inadvertentbservation that dogs may respond to PNES. Therefore it isrobably more appropriate to limit research to specificallyrained dogs, at least for the foreseeable future.

roup Discussion

he Group Discussion elaborated on a number of the issuesaised in the talks.

. The difficulties in undertaking a randomised controlledtrial were further highlighted but it was considered thata waiting list control design might be the best designoption.

. Although cats can alert people to their oncomingseizures, the typical response of a cat would be to runaway rather than stay with the person and this makesthem less suitable for the role of an alert animal.

. Once exposed to a SAD this may influence the abilityof the person with epilepsy to detect their oncomingseizures and may increase their epilepsy-related aware-ness and their ability to implement countermeasures.Thus the loss of a dog at the end of a study, for exam-

ple, might not necessarily mean that the person revertsto the pre-study status quo in terms of seizure control.

. The influence of SADs in reducing anxiety and improv-ing quality of life of the person with epilepsy may be

2

R

A

B

B

C

D

D

D

G

H

H

K

K

K

L

M

M

M

M

O

P

S

S

S

S

42

similar to that seen in terms of the benefits of pets topeople with a range of psychiatric disorders, althoughthe mechanisms here may additionally include the ben-efit of providing the person with confidence in that theirseizures can be pre-warned thereby making them morewilling to engage in activities out of the home.

eferences

lert Service Dogs. http://alertservicedogs.com/faqs.html(accessed 10.03.11).

rown, S.W., Strong, V., 2001. The use of seizure alert dogs. Seizure10, 39—41.

rown, S.W., 2005. Pet Therapy. In: Devinsky, O., Schachter, S.,Pacia, S. (Eds.), Complementary and Alternative Therapies forEpilepsy. Demos, New York, pp. 311—317.

ancer Detection Dogs. http://cbd.cancerdogs.co.uk/home.php(accessed 10.03.11).

alziel, D.J., Uthman, B.M., Mcgorray, S.P., Reep, R.L., 2003.Seizure-alert dogs: a review and preliminary study. Seizure 12,115—120.

ogs4Diabetics. http://www.dogs4diabetics.com/ (accessed10.03.11).

oherty, M.J., Haltiner, A.M., 2007. Wag the dog: skepticism onseizure alert canines. Neurology 68, 309.

oldstein, L.H., Mellers, J.D.C., 2006. Ictal symptoms of anxiety,avoidance behaviour and dissociation in patients with dissocia-tive seizures. J. Neurol. Neurosurg. Psychiatry 77, 616—621.

astings, T., Burris, A., Hunt, J., Purdue, G., Arnoldo, B., 2008. Pettherapy: a healing solution. J. Burn Care Res. 29, 874—876.

aut, S.R., Hall, C.B., LeValley, A.J., Lipton, R.B., 2007. Canpatients with epilepsy predict their seizures? Neurology 68,

262—266.

irton, A., Wirrell, E., Zhang, J., Hamiwka, L., 2004. Seizure-alerting and -response behaviours in dogs living with epilepticchildren. Neurology 62, 2303—2305.

S

S.W. Brown, L.H. Goldstein

irton, A., Winter, A., Wirrell, E., Snead, O.C., 2008. Seizureresponse dogs: evaluation of a formal training program. EpilepsyBehav. 13, 499—504.

rauss, G.L., Choi, J.S., Lesser, R.P., 2007. Pseudoseizure dogs.Neurology 68, 308—309.

itt, B., Krieger, A., 2007. Editorial, Of seizure prediction, statistics,and dogs. A cautionary tail. Neurology 68, 250—251.

aiwald, T., Blumberg, J., Timmeer, J., Schulze-Bonhage, A., 2011.Are prodromes preictal events? A prospective PDA study. EpilepsyBehav. 21, 184—188.

ellers, J.D.C., 2005. The approach to patients with ‘non-epilepticseizures’ Postgrad. Med. J. 81, 498—504.

oretti, F., De Ronchi, D., Bernabel, V., Marchetti, L., Ferrari, B.,Forlani, C., Negrett, F., Sacchetti, C., Atti, A.R., 2011. Pet ther-apy in elderly patients with mental illness. Psychogeriatrics 11,125—129.

ossello, E., Ridolfi, A., Mello, A.M., Lorensini, G., Mugnai, F., Pic-cini, C., Barone, D., Peruzzi, A., Masotti, G., Marchionni, N.,2011. Animal-assisted activity and emotional status of patientswith Alzheimer’s disease in day care. Int. Psychogeriatr. 23,899—905.

rtiz, R., Liporace, J., 2005. Seizure-alert dogs: observations froman inpatient video/EEG unit. Epilepsy Behav. 6, 620—622.

inikahana, J., Dono, J., 2009. The lived experience of initial symp-toms of and factors triggering epileptic seizures. Epilepsy Behav.15, 513—520.

achs-Ericsson, N., Hansen, N.K., Fitzgerald, S., 2002. Benefits ofassistance dogs: a review. Rehabil. Psychol. 47, 251—277.

evcencu, C., Struijk, J.J., 2010. Autonomic alterations and cardiacchanges in epilepsy. Epilepsia 51, 725—737.

trong, V., Brown, S.W., Walker, R., 1999. Seizure alert dogs — factor fiction? Seizure 8, 62—65.

trong, V., Brown, S., 2000. Should people with epilepsy have

untrained dogs as pets? Seizure 9, 427—430.

trong, V., Brown, S., Huyton, M., 2002. Effect of trained SeizureAlert Dogs on frequency of tonic—clonic seizures. Seizure 11,402—405.