Materials and Methods

Identification of research subjects—Data for the present study were collected as part of a larger research study in which owners and breeders of Poodles with seizures were recruited through national and regional Poodle clubs, dog-fancier magazines, and the Internet. Each owner was asked to provide a 4- or 5-generation pedigree for the affected dog and answer questions about the dog's seizures and medical history, with emphasis on any illness or injury that could have contributed to the seizures.

When > 1 affected Poodle in a bloodline was identified, the breeder was asked to provide litter records for as many litters as possible that were closely related to the affected Poodles, even if none of the dogs in those litters had reportedly had any seizures. A litter record included information on the parents of the litter, the numbers of males and females in the litter, the date the litter was born, and the contact information for owners of dogs in the litter. During this initial data collection period, several large families with multiple affected dogs were identified. However, there was only 1 family for which complete litter records were obtained and for which there was a clear line of descent. Specifically, in this family, all litters were found to be descended from the same maternal bloodline, although there were 4 relatively unrelated sires (dogs 1 to 4; Figure 1). The dams included the foundation bitch of this family (dog 5), a full sister of the foundation bitch (dog 6), and 3 daughters of the foundation bitch (dogs 7, 8, and 9). Parentage of selected dogs in the family was verified by means of microsatellite marker DNA testing performed by a commercial company.^a At the time litter records were obtained, 4 dogs from the family were reported to have had seizures. Among these 4 dogs were the foundation bitch (dog 5) and 1 of the daughters of the foundation bitch (dog 7), both of which had been bred multiple times prior to the onset of seizures.

View Full Size

Pedigree of a family of Standard Poodles with seizures attributed to PIE. White squares and circles represent unaffected male (squares) and female (circles) dogs, black squares and circles represent dogs with PIE, and gray squares and circles represent dogs for which seizure status was unknown. A slash indicates the dog was dead at the time of the study. Dotted lines connect female dogs from their birth litter to litters they produced. The 2 horizontal arrows at the top indicate that there were an unknown number of dogs in these litters. An asterisk indicates a dog that was reported to have had only 1 seizure episode.

Citation: Journal of the American Veterinary Medical Association 231, 10; 10.2460/javma.231.10.1520

• Download Figure
• Download figure as PowerPoint slide

For each of the 126 dogs in this family, owners were sent a letter from the breeder indicating that she was cooperating with epilepsy researchers because 2 of the dams in the family had had seizures after they were bred and encouraging the owners to participate in the study. Letters included a postcard to be returned to the first author (BGL) that identified the Poodle in question and asked the owner to respond to the following 3 questions: Has your dog ever had any seizures? Has your dog ever had any other episodes of unusual behavior or uncontrolled movements of its body? May we contact you for further information? Question 2 was asked to determine whether the dog might have had focal (partial) seizures that the owner did not identify as seizures.

For letters returned by the post office as undeliverable, the researchers attempted to track down the owner's new address through free Internet search engines. Owners who did not return the postcard were contacted by the first author by telephone to determine whether they had received the letter and were willing to participate; all owners who were contacted in this manner consented to participate.

Determination of seizure status—The first author contacted all owners who reported that their dogs had had seizures or any other episodes of unusual behavior or uncontrolled movements of its body. Owners were asked to provide open-ended descriptions of the dog's seizures and other unusual episodes. Standard follow-up questions were then asked to determine the first signs exhibited during each episode and to assess the dog's responsiveness during episodes. Additional follow-up questions were asked to clarify ambiguities in the owners' descriptions, such as whether there was abnormal motor activity (eg, shaking, stiffness, or difficulty standing or walking), urination, defecation, salivation, or repetitive stereotypical behaviors (eg, repeatedly licking mouth). If necessary, owners also were asked to clarify the episodic nature of abnormal behaviors and uncontrolled movements (ie, whether these behaviors were relatively continuous or intermittent with normal behavior between episodes) and to clarify whether the owner could cause any particular episode to stop. Owners' descriptions and answers to follow-up questions were used to determine whether episodes were seizures and, if so, what type of seizure. Owners also were asked how old the dog was at the time of the first episode, the number of episodes the dog had had during its lifetime, whether any antiseizure medications had been administered, whether the dog had any illnesses or injuries that preceded or were concurrent with the first episode, and whether the dog had any other health problems.

For dogs with seizures, owners were contacted a second time after at least 2 years had passed since the first seizure and asked whether the dog had had any additional seizures during this period. To help rule out symptomatic epilepsy (ie, seizures due to an identifiable brain abnormality), owners were asked whether the dog had had any interictal neurologic abnormalities. Seizure descriptions were again solicited, but by a different interviewer than the first author, to increase the reliability of seizure descriptions and classification.

Because most of the affected dogs in this family had infrequent and relatively mild seizures, most did not undergo complete diagnostic testing when seizures began. Thus, the term PIE was used, rather than IE, when referring to affected dogs in the study because intracranial abnormalities could not be ruled out with confidence. Dogs were classified as having PIE if they had had at least 1 seizure without any evidence of an underlying cause. Specifically, the following conditions all had to be met: the owner's answers to health-related questions revealed no illnesses or events (eg, head trauma) that could plausibly account for the seizures, at least 1 year had passed since seizure onset during which no interictal neurologic abnormalities were observed, and the dog was between 6 months and 7.5 years old when seizures began. This age range was adopted on the basis of results of a study³ involving dogs with seizures that underwent extensive diagnostic testing. In that study, seizures were most likely to begin between 1 and 5 years of age in dogs with IE, but began as early as 6 months of age and as late as 7.5 years of age in some dogs.

Dogs that did not have any history of seizures or other episodes of unusual behavior or uncontrolled body movements were classified as unaffected. At the time analyses were conducted, all dogs classified as unaffected were at least 6 years old, decreasing the chance that dogs with PIE would be falsely classified as unaffected while maintaining a sufficient number of dogs for analyses. Seizure status for dogs that reportedly had not had any seizures during their lifetimes but that died before turning 6 years old was recorded as unknown. In addition, seizure status was recorded as unknown for dogs whose owners could not be contacted for the study.

Classification of seizures—Each dog's seizures were classified on the basis of a modified version of the International League Against Epilepsy classification system²⁹ used for seizures in humans. Modifications to the International League Against Epilepsy classification system were made to accommodate differences between human and canine seizures, such as the inability to evaluate a dog's consciousness with methods used for human patients.²⁵

For assessments of consciousness, dogs were considered to have lost consciousness if the dog was showing no responsiveness to the owner and the dog was not navigating the environment (eg, walking or running) in any way. Dogs were considered to have impaired consciousness if the dog was showing no responsiveness to the owner, but the dog was navigating the environment in some way (eg, the dog was wandering around). Dogs also were considered to have impaired consciousness if the dog's behavior was out of context (eg, cowering or hiding for no reason or aggression in an otherwise nonaggressive dog). Dogs were considered to have preserved consciousness if they were showing normal responsiveness to the owner and their behavior was not out of context.

To increase inter-rater reliability of seizure classifications, specific operational definitions were established for all seizure types. For a seizure to be classified as generalized (either primarily or secondarily generalized), movements had to have been bilateral, the entire body had to have been involved, and consciousness had to have been lost. A seizure was classified as focal (partial) if consciousness was not lost or movements were limited to a single part of the body (eg, head shaking only) or were unilateral. Focal seizures for which consciousness was preserved were classified as simple focal seizures; focal seizures for which consciousness was impaired were classified as complex focal seizures. For seizures that had a focal onset but secondarily generalized, owners were asked to rate the dog's responsiveness during the focal onset and during the worst part of the seizure.

Consistent with the human classification system, it was noted when seizures were accompanied by motor signs (eg, jerking, rigidity, or difficulty walking or standing), autonomic signs (eg, salivation, urination, defecation, vomiting, or panting), or psychic signs (eg, cowering or hiding or other signs of unwarranted fear, unprovoked aggression in a nonaggressive dog, or snapping at imaginary flies). However, somatosensory and special sensory symptoms (eg, tingling or numbness of a limb) were not recorded because these symptoms could not be reliably assessed in dogs. Automatisms (ie, organized behaviors that generally are purposeful but that have no meaning in the context of the seizure and are performed automatically, such as repeatedly licking the mouth and swallowing or legs paddling as if running or swimming) were recorded.

Analysis of mode of inheritance—To determine the mode of inheritance of PIE, pedigree data were analyzed by means of segregation analysis. Six classes of genetic models were evaluated, and nested models were compared to determine their relative goodness of fit to the observed data. The 6 classes of models that were evaluated included environmental (ie, nongenetic), dominant, recessive, simple dominant, simple recessive, and complex. The distinction between dominant and recessive, on the one hand, and simple dominant and simple recessive, on the other, was that the dominant and recessive models allowed for both phenocopy (ie, dogs with the disease phenotype that did not have the disease genotype) and incomplete penetrance (ie, dogs with the disease genotype that did not have the disease phenotype), whereas the simple dominant and simple recessive models allowed only for incomplete penetrance. In this context, the existence of incomplete penetrance was considered an indication that there might be an additional gene or genes that modified the expression of the disease gene, or that there were environmental factors influencing the expression of the disease gene.

For each of the 6 classes of genetic models considered, the expectation-maximization algorithm,³⁰ where the E-step was facilitated with a Markov-chain Monte Carlo procedure, was implemented to obtain maximum likelihood estimates of penetrance values and gene frequency (Appendix). This was done to accommodate the complex familial relationships, including inbreeding, among dogs in the family. This procedure allowed us to determine the best model from each of the 6 classes of potential models, by identifying the model that best fit the observed data. Subsequently, the best models from each of the 6 classes of potential models were compared by means of likelihood ratio tests for nested models³¹ to determine the single model that best explained the observed data overall. For all tests, a value of P < 0.05 was considered significant.

All analyses were conducted twice. In the first set of analyses, all dogs with ≥ 1 episode were classified as affected. In the second set of analyses, dogs with only a single episode were classified as seizure status unknown, rather than as affected, to determine whether considering dogs with only a single episode as affected altered the results.

Results

Clinical characteristics—Of the 126 Standard Poodles in the family, there were 90 (71.4%) for which a phenotype determination (ie, affected vs unaffected) could be made. Thirty of the 90 (33%) had had ≥ 1 episode judged to be a seizure and met the criteria for being classified as having PIE (Figure 1). Of these 30, 4 (all males) had had only a single observed episode. The remaining 60 (67%) dogs were classified as unaffected. Owners of 27 dogs could not be located, and these dogs were classified as seizure status unknown. An additional 9 dogs were classified as seizure status unknown because they had not been observed to have any seizures, but died before they were 6 years old. Of the 30 dogs with PIE, 17 were male and 13 were female. The proportions of affected males and females were not significantly (P > 0.50) different.

Twenty-nine of the 30 dogs classified as having PIE were followed up for at least 2 years after seizure onset, and none had any interictal neurologic abnormalities during the follow-up period. The dog that was not followed up for at least 2 years after seizure onset was a male that died after being hit by a car approximately 1 year after the onset of seizures. This dog had begun having seizures when it was between 2.5 and 3 years old.

Seizures could be classified in 29 of the 30 dogs with PIE; in the remaining dog, seizures could not be classified because after the initial report of a seizure, the owner could not be contacted to obtain a detailed description. Of the 29 dogs in which seizures could be classified, 27 (93%) had evidence of focal onset seizures. Of the 2 owners who had never observed any signs of focal onset seizures, only 1 reported seeing the seizures from the beginning of the episodes. Of the 27 dogs with focal onset seizures, 9 (33%) had had at least 1 seizure that secondarily generalized. Thus, 11 of the 29 (38%) dogs had had at least 1 generalized seizure (either primary or secondarily generalized). In 6 of these 11 dogs, generalized seizures included tonic (stiffness and rigidity) and clonic (shaking,jerking, or paddling) signs, and in 4, generalized seizures included only clonic signs. For the remaining dog, the owner observed only tonic signs, but did not see the seizures from the beginning.

Of the 18 dogs that had focal seizures without secondary generalization, 6 had simple focal seizures (ie, consciousness was preserved) and 12 had complex focal seizures (ie, consciousness was impaired). Of the 9 dogs that had focal onset seizures that secondarily generalized, 8 had evidence of impaired consciousness during the focal onset (ie, complex focal seizures); consciousness could not be rated during the focal onset in the remaining dog.

Motor signs in the dogs with focal onset seizures (with or without generalization) included shaking, jerking, or shivering (n = 22); incoordination characterized by staggering or an inability to stand (15); stiffness or rigid-ity (12); and unusual movements or body positions such as a head tilt or awkward leg lifting (3). Autonomic signs included drooling (n = 11), panting (4), urination (1), and increased heart rate (1). Thirteen dogs had psychic signs (anxiety), and 3 had automatisms (2 with licking, lip smacking, or swallowing and 1 with circling). Autonomic signs in dogs with generalized seizures (primary or secondarily generalized) included drooling (n = 5), urination (3), defecation (3), panting (1), vomiting (1), and pupillary dilatation (1). Automatisms that were reported included licking, lip smacking, or swallowing (n = 2) and coordinated paddling as if swimming or running (4).

Median age at the onset of seizures was 3.7 years (range, 6 months to 7 years). Six dogs were > 5 years old at the onset of seizures, including 3 dogs in which seizures were first observed when dogs were between 5 and 6 years old, and 3 dogs in which seizures were first observed when dogs were between 6 and 7 years old. Of the 30 dogs with PIE, 4 were reported to have had only 1 observed episode, 3 were reported to have had 2 observed episodes, 10 were reported to have had 3 to 5 observed episodes, and 13 were reported to have had > 5 observed episodes. Only 4 dogs were receiving antiseizure medication, and all appeared to have improved with medication. Three dogs were treated with phenobarbital and 1 was treated with potassium bromide. Only 1 dog, a female, had seizures that progressed, becoming more frequent over the years. During the first 2.5 years following seizure onset, this dog had had only 3 episodes. However, during the next 2 years, the owner estimated that the dog had 20 episodes a year. The dog was never treated with antiseizure medications and was euthanized because of the frequent seizures.

Mode of inheritance—None of the conclusions changed when results of analyses of pedigree data in which all dogs with ≥ 1 episode were classified as affected were compared with results of analyses of pedigree data in which the 4 dogs with a single episode were classified as seizure status unknown. Therefore, results of only the first set of analyses are presented.

Parameter estimates were obtained for the best model from each of the 6 classes of potential models examined (Table 1). For the environmental model, no estimate was obtained for disease gene frequency because this model assumes that epilepsy is not inherited. In addition, for this model, penetrances were all equal, regardless of genotype.

Likelihood ratio tests were used to compare nested models for their ability to explain the observed data (Figure 2). For these analyses, the simple recessive mode of inheritance was considered a subclass of models nested within the recessive mode of inheritance, and the recessive mode of inheritance was considered a subclass of models nested within the complex mode of inheritance. Similarly, the simple dominant mode of inheritance was considered a subclass of models nest-in which they were nested. For example, for the recessive model, penetrances for the genotypes DD and Dd, which represented the phenocopy rate for this model (ie, in this model, if a dog with either of these genotypes were affected, it would be because of phenocopy), were set to be equal to each other and could vary from 0 to the penetrance for the dd genotype. In contrast, for the simple recessive model, the phenocopy rate was set to 0. That is, penetrances for the DD and Dd genotypes were not free to vary. Thus, if the model that was determined to be the best of all recessive models had a phenocopy rate of 0, this would provide evidence that the simple recessive model was a better model for the obtained data than the more general recessive model.

View Full Size

Parameter estimates for the best model from each of 6 classes of models for mode of inheritance of PIE in a family of Standard Poodles.

Citation: Journal of the American Veterinary Medical Association 231, 10; 10.2460/javma.231.10.1520

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Illustration of the nested relationships of the 6 potential models that were considered for mode of inheritance of PIE in the family of Standard Poodles in Figure 1. Arrows connect nested models, with the arrow pointing toward the larger of the 2 models (ie, the model with more free parameters). Numbers associated with each arrow correspond to the P value for the χ² statistic of the likelihood ratio test for whether the larger model explained the observed data significantly better than the smaller model. ENV = Environmental. SR = Simple recessive. REC = Recessive. DOM = Dominant. SD = Simple dominant. COMP = Complex.

Citation: Journal of the American Veterinary Medical Association 231, 10; 10.2460/javma.231.10.1520

• Download Figure
• Download figure as PowerPoint slide

The complex model had the fewest restrictions placed on its penetrance values, as possible penetrances included values obtained for all other models analyzed, and penetrance values could differ for each of the 3 genotypes. Thus, the complex model had the greatest opportunity to provide the best fit with the obtained data. If the complex model yielded parameters different from those obtained for all of the other models and fit the observed data best, this would have provided evidence that none of the other models examined adequately fit the data, and additional models would need to be considered, such as polygenic models. In contrast, if the best complex model yielded the same parameters (penetrances, gene frequency, and likelihood) as one of the other models, this would have indicated that the other model provided the best explanation of the data. This is because the parameter values of the other model were chosen as best, even though the complex model had all the other options from which it could choose.

Analysis of the best models for each of the 6 potential classes of models revealed that the best recessive and complex models yielded virtually the same parameter values as did the best simple recessive model (Table 1). For example, for all 3 models, the estimated probability of being affected for a dog with 2 copies of the disease gene was at least 0.999, and the estimated probability of being affected for dogs with either of the other 2 genotypes was 0.

The recessive model could not be directly compared with the dominant model and the simple recessive model could not be directly compared with the simple dominant model because one was not nested within the other. However, the complex model could explain the observed data significantly (P < 0.001) better than the dominant model, which precluded consideration of the dominant model or, by extension, the simple dominant model, as the best model. Furthermore, because the recessive model fit the data significantly (P < 0.001) better than the environmental model, and the simple recessive model was virtually the same as the best recessive model, the observed data were best explained by a simple recessive inheritance model.

Follow-up information for unaffected dogs—After completion of the study, researchers attempted to obtain follow-up information for all dogs classified as unaffected until they were at least 7 years old. Of the 60 dogs classified as unaffected, 43 could be followed up until they were ≥ 8 years old, and 15 could be followed up until they were between 7 and 8 years old. Additional follow-up information was not available for the remaining 2 dogs.

During the follow-up period, 2 dogs initially classified as unaffected were subsequently identified as having PIE, with seizure onset between 6 and 7 years of age. One additional dog was reported to have had 2 seizures. However, this dog was not classified as having PIE because it was almost 9 years old at the onset of seizures, it had been identified as having lupus, and it had likely been exposed to a hallucinogen. All 3 dogs had focal onset seizures, and all 3 were from litters that had other dogs with seizures. Although the analyses were not repeated to include the 2 dogs that were identified as having PIE after completion of the study, the conclusion that the simple recessive model best explained the data would not have changed. This is because both dogs were from litters that had other dogs with seizures, the analyses provided strong support for the simple recessive model, and the conclusions remained unchanged when the analyses were conducted with and without the 4 dogs that had had only 1 seizure episode.

Discussion

Results of the segregation analyses in the present study strongly suggested that PIE in this family of Standard Poodles was inherited as a simple (ie, 1 gene locus) recessive autosomal trait with complete or almost complete penetrance and a gene frequency of 0.428. Our findings were similar to those of 2 previous studies,^7,10 both of which had data that were consistent with a simple recessive autosomal mode of inheritance. As was the case for the present study, these 2 previous studies included dogs that were highly related to each other, which likely decreased confounding associated with genetic heterogeneity. However, unlike these previous studies, the present study allowed us to conclude that a simple recessive autosomal mode of inheritance with complete or almost complete penetrance was the best model for explaining the observed data in this family.

Until researchers can map the causal genes or markers that are linked to the causal genes in dogs with IE, the true mode of inheritance will not be known with certainty. However, examination of the pedigree for the family used in the present study, in conjunction with results of segregation analyses, suggested that all dams in the family were affected or carriers, that sires 1 and 2 were carriers, and that sires 3 and 4 were not carriers. When sire 1 or 2 was bred to any of the dams, some affected offspring were always produced, whereas when sire 3 or 4 was bred to any of the dams, including any of the affected dams, affected offspring were not produced.

Importantly, we do not know whether the same mode of inheritance exists in other families of Standard Poodles with IE. On the basis of unpublished data, we believe that there are differences among Standard Poodle families in terms of the phenotypic expression of epilepsy. Anecdotally, we have found, for example, that in some families there are a relatively high number of cases with severe seizures that are refractory to medication. This contrasts with the relatively mild phenotype in the family analyzed in the present study. Thus, it is possible that in other Standard Poodle families, there may be different genetic mutations or different modes of inheritance.

A recessive mode of inheritance is consistent with results of previous studies of IE in dogs^{6,7,8,9,10,17,18} and some ro-dents.^32,33,34 However, most types of IE in humans that have been successfully mapped have an autosomal dominant mode of inheritance.^27,28 One possible explanation for the discrepancy between humans and dogs is that with dogs, matings are arranged by breeders. Thus, breeders may try to avoid breeding dogs with any disease that is clearly inherited, even if the disease is relatively mild. With a dominant disorder, the role of genetics will be more obvious to breeders because the disorder is more likely to emerge in successive generations and dogs that are affected will be more likely to produce affected offspring. In other words, dominant diseases have an excellent chance of being eliminated by selective breeding. However, with a recessive disorder, the role of genetics will be less obvious to breeders because the disorder can skip generations and, even when it occurs, a smaller proportion of offspring are likely to be affected. Thus, diseases that have remained in the canine gene pool for many decades, such as epilepsy, will probably be recessive. However, this does not rule out the possibility that some of the molecular defects that cause IE (eg, mutations in ion channel genes) may be the same for dogs and humans.

It is noteworthy that classifying dogs with only a single observed seizure episode as affected, rather than as seizure status unknown, did not alter the conclusions of our analyses. Although dogs that have only 1 seizure are rarely considered to have epilepsy, the argument for classifying dogs with 1 seizure as affected in the initial analyses was that most dogs never have any seizures. Thus, the occurrence of even 1 unprovoked seizure suggested that the dog differed neurologically and, probably, genetically from dogs that have never had even 1 seizure. On the other hand, it is recognized that when only 1 episode is observed, particularly an episode that looks like a focal seizure, it is easy to misclassify a nonepileptic episode as a seizure. Nonetheless, the fact that the same conclusions emerged regardless of how dogs with only 1 episode were classified raises the possibility that dogs with only 1 episode may have inherited the underlying genetic defect. Consistent with this, Famula and Oberbauer¹⁹ found that having 2 parents with only 1 seizure per parent increased the chances that there would be affected offspring. From a broad sample of Belgian Tervuren, they found that the probability of producing seizure-free offspring when mating 2 parents, each of which had had only 1 seizure, was only 0.58. In contrast, when both parents were seizure-free, the probability of producing sei-zure-free offspring was 0.99.

When deciding on the minimum age for dogs without any history of seizures to be classified as unaffected, researchers must balance the need to avoid false-nega-tive results with the need to include a sufficient number of dogs in the analyses. The former consideration would argue to set the age high (eg, 7 years or older); the latter would argue to set the age low (eg, 5 years). The decision should be influenced by the upper limit for the age at which IE is likely to emerge in general, as well as by the upper limit for the age at which it emerges in the families being studied. In prior studies, the minimum age for dogs without any history of seizures to be classified as unaffected was 5 years,^6,10,19 4 years,¹¹ or 3 years.^8,9,15,18 For the Standard Poodle family that was analyzed in the present study, use of 6 years led to 2 false-negative classifications, although reclassifying these 2 dogs would not have altered the study's conclusions. However, it is possible that in studies in which a lower age cutoff (eg, 3 years) is used, a high penetrance might appear to be falsely low because of a high rate of false-negative classifications.

One striking finding of the present study was that nearly all affected dogs in this family (27/29 [93%]) had evidence of focal-onset (partial) seizures. This was consistent with findings of an earlier study²⁵ that examined seizure phenotype in Standard Poodles from several families, including 6 of the 41 affected Standard Poodles in the present study. In that study, 88% of the dogs for which seizures were observed from the beginning of the episodes had evidence of a focal onset during at least some of their seizure episodes. Because all dogs in the present study were from the same family, it was not surprising that we found an even higher percentage of dogs with the same seizure type.

As pointed out by Patterson et al,^10,17 it has often been argued that when dogs have focal-onset seizures, they are most likely to have symptomatic epilepsy. In contrast, findings of the present study and these previous studies suggest that in certain breeds, IE may be characterized by focal-onset seizures. In 1 study,¹⁰ for instance, 79% of Vizslas with IE had focal-onset seizures, and in another,¹⁷ 53% of English Springer Spaniels with IE had focal-onset seizures. Studies^27,28 of epilepsy in humans also demonstrate that focal-onset seizures can be due to inherited IE.

Another common notion is that when dogs begin having seizures after 5 years of age, the seizures are likely due to factors other than IE. In contrast, 3 of the dogs with PIE in the present study were between 5 and 6 years old at the onset of seizures, and 5 (including the 2 reported to develop seizures after the conclusion of the study) were between 6 and 7 years old. Thus, the onset of seizures was after 5 years of age in 8 of the 32 dogs (25%) with PIE in this Standard Poodle family. While it is possible that the seizures in these dogs were due to factors other than IE, the fact that none of these dogs had evidence of neurologic deterioration over time suggests that the diagnosis of IE was correct and that in dogs with IE, seizures can begin after 5 years of age. This is consistent with the findings of Patterson et al,¹⁷ who found that 20% of English Springer Spaniels had an onset of seizures between 5 and 6 years of age, with the oldest age of onset being 7 years. Similarly, Podell et al³ identified some dogs with confirmed IE that were as old as 7.5 years at the time of seizure onset.

An important limitation of the present study was that relatively few dogs underwent sufficient diagnostic testing at the time of the first seizure to confidently rule out causal factors other than IE. While it would have been desirable to have more diagnostic information on all dogs, we would argue that what was gained in terms of including the entire family far outweighed the drawbacks of including dogs without thorough diagnostic testing. Additionally, there were multiple years of follow-up information for affected dogs that revealed no neurologic deterioration. Although this does not rule out all causes of seizures other than IE, it does rule out progressive disorders that can cause seizures. Additionally, finding a clear mode of inheritance was consistent with the notion that seizures were due to IE.

Because owners were informed that 2 dams in this family had had seizures, some owners could have been influenced to report unusual episodes as seizures, even though they were not. In addition, because the researchers knew that this was an affected family, they could have been biased toward classifying ambiguous unusual episodes as seizures when, in fact, they were not. In clinical practice and research studies of epilepsy in humans, electroencephalography is used to help clarify whether ambiguous episodes are in fact seizures.³⁵ Although a few stud-ies^36,37,38,39 have compared electroencephalographic findings in dogs with and without seizures, electroencephalography is used infrequently in dogs. This appears to be due, in part, to the fact that no reliable standards have been established for recording techniques (eg, placement and number of electrodes) or for what constitutes normal electroencephalographic activity in dogs.³⁷ Additionally, dogs must be anesthetized or sedated to perform electroencephalography, and there is little research on the effects of anesthetic and sedative agents on electroencephalographic findings.³⁷ The response to antiseizure medications can be used to determine whether ambiguous episodes are in fact seizures. However, few dogs in the present study were treated with antiseizure medications, although those that were treated were reported to have improved.

Although bias cannot be completely ruled out in the present study, steps were taken to avoid classifying nonepileptic events (eg, dreaming or anxiety) as seizures. First, open-ended descriptions of the various episodes provided by the owners were followed up with specific, standardized questions to clarify the nature of those episodes. Second, to increase the reliability of the information provided by the owners, those with dogs classified as affected were interviewed by a second individual 2 years after the initial interview. Third, any bias that may have existed on the part of the researchers would have led them to classify ambiguous episodes as seizures, regardless of who the dog's sire was. That is, researchers had no a priori expectation that use of certain sires (sires 1 and 2) would produce affected offspring in each litter, but that use of other sires (sires 3 and 4) would never produce affected offspring. Further, during follow-up of unaffected dogs conducted after completion of the study, interviewers did not know whether they were inquiring about a dog whose sire was a putative carrier (sires 1 and 2) or presumed to be clear (sires 3 or 4). Thus, the interviewers could not have influenced the owners to respond in a way that would confirm the pattern of inheritance found in the study. As indicated, these interviewers identified 2 dogs with PIE that were offspring of sires 1 and 2, but no dogs with seizures or even ambiguous episodes that were offspring of sires 3 and 4. Thus, while one cannot completely rule out a false-positive classification for any particular dog, it is not plausible that bias led to the overall pattern of results in the present study.

In addition to helping clarify whether an ambiguous episode is a seizure, electroencephalography is also used in humans to help clarify whether a seizure is focal or generalized. Thus, electroencephalography could have strengthened the conclusion that affected dogs in this family had almost exclusively focal-onset seizures. On the other hand, specific operational definitions were provided for all seizure types to increase the reliability of seizure classifications. Additionally, the focal-onset seizure phenotypes described in the present study were consistent with focal-onset seizure phenotypes described in other studies^10,23,24,39 of epilepsy in dogs.