Diagnostic Validity of Electroencephalography ... - Wiley Online Library

J Vet Intern Med 2001;15:385–393

Diagnostic Validity of Electroencephalography in Equine Intracranial Disorders V.A. Lacombe, M. Podell, M. Furr, S.M. Reed, M.J. Oglesbee, K.W. Hinchcliff, and C.W. Kohn

Electroencephalography (EEG) is a valuable diagnostic test to identify functional disturbances in brain activity. The purpose of this study was to assess the validity of EEG as a diagnostic indicator of intracranial diseases in horses. The validity of EEG was estimated by comparing clinical, clinicopathologic, and histopathologic findings to EEG findings in 20 horses examined for seizures, collapse, or abnormal behavior between 1984 and 1997. A bipolar left-to-right, back-to-front montage and a bipolar circular montage were recorded from sedated (4) and anesthetized (16) horses. Visual and semiquantitative masked analysis of EEG recording 1st was validated on 10 horses presented for problems other than intracranial diseases. EEG pattern was normal in 7 of the 20 clinically affected horses. Abnormal EEG patterns included high-voltage slow waves and discrete paroxysmal activity with or without generalized activity in 13 horses. Histopathologic diagnoses in 10 horses included meningoencephalitis, neuronal necrosis, congenital anomalies, cerebral edema, and abscess. All of these horses had abnormal EEG patterns (sensitivity, 100%) with a positive neuroanatomic correlation in 7 animals. Localization of histopathologic and EEG abnormalities did not correlate in 15% of the horses (3/20). The cause of neurologic signs could not be explained at postmortem examination in 10 animals and the EEG pattern was normal in 7 of these horses (specificity, 70%). In conclusion, equine EEG was a sensitive tool in the diagnosis of intracranial disorders. Key words: Brain abscess; Collapse; Horses; Meningoencephalitis; Seizures.

E

lectroencephalography (EEG) is the graphic recording of the rhythmic bioelectrical activity arising predominantly from the cerebral cortex. A noninvasive method of determining abnormal cerebral electrical activity, EEG has been used extensively in humans and in small animals as a screening procedure in patients with intracranial neurologic disorders.1,2 EEG is most useful in the investigation and management of human patients with epilepsy, in particular in establishing probability of recurrence, need for treatment, and prognosis.1 Similar utilization as a diagnostic aid has been demonstrated in epileptic dogs,3 but information is lacking regarding the use of EEG in horses. EEG was 1st used in studying natural sleep and its disorders in large animals by means of electrodes surgically implanted in the skull.4–8 These electrocortigraphic recordings in healthy burros and horses had a high frequency (25– 40 Hz) with dominant voltage in the 50-mV range.9–11 More recently, EEG has been used to monitor electrical activity of the cerebral cortex during general anesthesia in horses and was found useful in evaluating central nervous effects of different anesthetic agents.12–18 Although little clinical evaluation has been performed in horses, normal findings From the Department of Veterinary Clinical Sciences (Lacombe, Podell, Reed, Hinchcliff, Kohn), and Department of Veterinary Biosciences, College of Veterinary Medicine (Oglesbee), and Department of Neuroscience, College of Medicine (Podell), The Ohio State University, Columbus, OH; and the Marion duPont Scott Equine Medical Center, Virginia-Maryland Regional College of Veterinary Medicine, Virginia Polytechnic Institute and State University, Blacksburg, VA (Furr). Sections of this study were presented previously as a poster at the 16th ACVIM meeting, San Diego, 1998. Reprint requests: V.A. Lacombe, DVM, Department of Veterinary Clinical Sciences, College of Veterinary Medicine, The Ohio State University, 601 Tharp Street, Columbus, OH 43210; e-mail: [email protected]. Submitted February 3, 2000; Revised November 20, 2000; Accepted February 9, 2001. Copyright q 2001 by the American College of Veterinary Internal Medicine 0891-6640/01/1504-0010/$3.00/0

for sedated and tranquilized newborn foals and adult horses have been reported.19 However, the establishment of normal values remains difficult because frequency and amplitudes are state-dependant (awake, drowsy, sleeping, sedated, or anesthetized), and chemical restraint influences EEG patterns. Horses given xylazine have generalized slow wave activity and those given acepromazine have fast activity with low to moderate voltage.19 Few studies have examined the role of EEG as a diagnostic tool for intracranial lesions in large animals. Two studies reported EEG patterns in cattle and sheep with cerebrocortical necrosis.20,21 In 1 study,22 EEG revealed seizure activity in 3 horses suffering from equine protozoal myeloencephalitis. The validity of EEG as a diagnostic tool in horses suffering from intracranial disorders has not been reported previously. The clinical problem is differentiation of underlying intracranial diseases in horses with paroxysmal onset of neurologic disturbances. The difficulty lies in the rapid return to normal activity without persistent neurologic deficits coupled with a small number of facilities capable of imaging the equine brain. The purpose of the study was to establish normal EEG parameters in horses and to assess the validity of EEG as a diagnostic indicator of intracranial diseases in horses. The specific aims of this project were 1st to validate the use of the semiquantitative analysis approach to EEG recordings in 10 normal anesthetized horses and to apply this scoring system to abnormal horses. Second, the diagnostic validity of EEG was evaluated by comparing clinical, clinicopathologic, and histopathologic findings to EEG findings in horses presented for suspected intracranial diseases.

Materials and Methods Control Group Between 1998 and 1999, 10 horses were presented to the Marion duPont Scott Equine Medical Center for problems other than intracranial diseases and constituted a control group for this study. Complete physical and neurologic examinations were performed on these horses. EEG was performed under general anesthesia on these 10 horses to

386

Lacombe et al

help to establish normality of the EEG and to validate the use of the semiquantitative analysis approach to EEG recording. The horses were premedicated with xylazinea (0.6 mg/kg IV), induced with guaifenesinb (55 mg/kg, IV) and thiopental sodiumc (5.5 mg/kg IV), and then intubated endotracheally and maintained on isofluraned at 1.2 minimum alveolar concentration. Horses were allowed to rest for 20 minutes before the start of any recording. EEG recordings were performed for 15 minutes on 8-channel machines (Neurofaxe) with 2 different types of montages: a bipolar left-to-right, back-to-front montage and a bipolar circular montage. The bipolar left-to-right, back-to-front montage contained reference electrodes, which allowed determination of absolute voltage and polarity. In the first 4 channels, 1 of the 2 electrodes was placed over an electrically inactive site. However, the common reference site was never completely inactive and contributed to the output signal, and can become contaminated.23 The circular montage was a bipolar montage that linked a serial pair of electrodes in coronal lines. The advantage of bipolar recording included elimination of the possibility of a contaminated reference, easy localization by phase reversal of relatively discrete focal abnormalities and prevention of problems that can arise from unbalanced amplifier input with common reference. On the other hand, voltage and polarity determinations are always relative to electrode placement.23 The recording parameters were similar for all patients and included a recording speed of 30 mm/ s, a high linear frequency filter of 70, a sensitivity of 7 mV/cm, a time constant of 0.16 Hz, and the presence of a 60-cycle filter. When muscle artifacts were detected during the recording, the skin and muscle underlying the electrodes were desensitized with 2% mepivacainef by subdermal injection. A single visual and semiquantitative analysis was performed on all recordings by an observer (MP) blinded to other findings. A total score was assigned based on frequency and voltage values, the presence or absence of asymmetrical patterns, and paroxysmal activity. The development of this scoring system was based 1st on a determination of what is considered normal for anesthetized horses based on prior case material from The Ohio State University. Categories were chosen for the most abnormal activity (score of 2): slowest frequency, highest voltage, greatest degree of paroxysmal activity, and most asymmetry. By default, intermediate range activity between normal and most abnormal was then given the intermediate value (score of 1). These intermediate changes have been associated with abnormal neurologic conditions in other species.24 The frequency was measured by taking the average of a minimum of 20 random 10-second blocks of activity throughout the recording. Scores of 2, 0, or 1 were assigned if the frequency measurement was ,8 Hz, 8–13 Hz, or .13 Hz, respectively. Scores of 1, 0, or 2 were assigned if the voltage measurement was ,25 mV, 25–50 mV, or .50 mV, respectively. Scores of 0, 1, or 2 were assigned if low, moderate, or high asymmetry pattern was recorded, respectively. Scores of 0, 1, or 2 were assigned if the paroxysmal activity recorded was ,25%, 25–50%, or .50%, respectively. EEG with only a change in background activity related to the anesthetic drugs was interpreted as normal despite a score . 0. Our hypothesis was that EEG should be considered normal when a moderate voltage and frequency (8–13 Hz, alpha rhythm, and 25–50 mV, respectively) without any asymmetry or paroxysmal activity were observed with a score of 0. To test this hypothesis, chi-square analysis of observed scored EEG versus expected normal scored EEG from a normal horse population (score of 0) was performed. The null hypothesis was that no difference existed between observed and expected scores, thus further validating the scoring system. In foals #10 months, the normal patterns included high-voltage slow wave activity19 with a total score of 4. The EEG was considered abnormal when abnormal voltage and frequency associated with or without asymmetrical patterns and paroxysmal activity were recorded. Epileptiform paroxysmal activity was defined as spikes or sharp waves, isolated or followed by a wave, spike and wave, or polyspike and wave complexes.3 After the EEG recording, cerebrospinal fluid (CSF) was collected from each horse via the atlantooccipital space and submitted for cy-

tologic examination (including total protein, red blood cell count, white blood cell count, and white blood cell count differential). The horses then were euthanized with an overdose of barbiturate and postmortem examinations were completed. The brain and spinal cord were removed. After fixation in 10% buffered formalin, brains were sectioned in a standardized fashion, and routinely mounted for histopathogic evaluation. Twenty samples of brain were taken in a predetermined standard format (4 from the cortex, 1 from the hippocampus, 2 from the periventricular region, and 3 from brainstem, in duplicate). This protocol was evaluated and approved by the Virginia Tech Animal Care and Use Committee.

Study Group Between 1984 and 1997, EEG was performed on 113 horses examined at The Ohio State University Veterinary Teaching Hospital because of seizures, collapse, or abnormal behavior. The validity of EEG was estimated by comparing neurologic examination, findings from other diagnostic testing, and histopathologic characteristics with EEG findings. Horses only were included in this study when postmortem examinations were performed because postmortem examination was considered as the gold standard. Physical and neurologic evaluations, CBC, serum chemistry (including ionized calcium concentration), and EEG were performed. Of the 27 horses initially included in the study, EEG was performed on 16 horses under general anesthesia and on 11 standing horses, sedated with xylazinea (0.4 mg/kg IV), in a quiet and dark environment. For general anesthesia, 11 horses were premedicated with xylazinea (0.6 mg/kg IV), induced with guaifenesinb (55 mg/kg IV) and thiopental sodiumc (5.5 mg/kg IV) or thiamylal sodiumg (4 mg/kg IV), and then intubated endotracheally and maintained with halothane.h Anesthesia was induced and maintained under isofluraned in 2 foals. One foal was premedicated with xylazinea (1.1 mg/kg IV) and induced with ketaminei (2.2 mg/kg IV). The mean 6 SD time of the anesthesia was 63 6 6 minutes. The protocol of general anesthesia was not reported in the medical record for 2 horses. Three horses received diazepamj IV before or at the time of EEG recordings. Analog EEG recordings were performed on one of two 10-channel machines (Grass model 8k and Neurofaxe) with 2 different types of montages (a bipolar left-toright, back-to-front montage and a bipolar circular montage) by means of 9 platinum subdermal needle electrodes placed in the scalp on the frontal, occipital, and parietal areas (Fig 1). The recording parameters were similar for all EEG recordings and included a recording speed of 25 mm/s, a high linear frequency filter of 70, a sensitivity of 7 mV/ cm, and a time constant of 0.16 Hz. The same semiquantitative analysis as described above was performed on all of the recordings by the same observer blinded to other findings. Nine EEGs from the 27 horses presented for suspected intracranial disorders could not be interpreted because of severe artifacts on the EEG tracings of standing and sedated horses. As a result, 7 cases were removed from the study because no other interpretable EEG recordings were available. Among the remaining 20 patients, EEG recordings were performed on 4 standing horses, with repeated recordings on 3 occasions (2 standing and 1 under general anesthesia). For repeated EEG, only the EEG recording performed nearest to the postmortem examination was selected in this study. CSF was collected in 18 of 20 horses from the atlantooccipital or lumbosacral spaces and submitted for cytologic examination. For horses examined after 1994 (7), serum and CSF also were analyzed for the presence of antibodies against Sarcocystis neurona by means of Western blot analysis. Additional diagnostic tests in some horses included skull radiographs (7) and titer for eastern equine encephalitis (EEE) or western equine encephalitis (WEE) (2), as indicated by the history and clinical examination. Euthanasia was performed on all horses because of their lack of clinical improvement. In 3 horses, the cause of the death was not related to the primary complaint. These animals died naturally or were

Diagnostic Validity of EEG in Horses

387

euthanized because of radial fracture, respiratory disease, and colic, respectively. Postmortem examinations of the central nervous system were performed in accordance to a standard protocol, including gross and histopathologic examinations of the frontal, temporal-parietal and occipital cortex, cerebellum, thalamus-brainstem, and spinal cord. All postmortem examinations were reviewed retrospectively without prior knowledge of EEG interpretation (MJO). Tissue sampling for histologic examination was incomplete for 2 horses. A positive correlation between the EEG and postmortem findings was made when 1 of the differential diagnoses made from the EEG interpretation corresponded to the diagnosis based on histologic examination. An anatomic correlation between EEG and postmortem findings was made when abnormalities were localized to a common site. Findings also were considered to correlate when the postmortem findings indicated widespread disease that included regions to which EEG abnormalities were localized. Postmortem examination was considered as the gold standard for intracranial disorders and the prevalence or pretest probability in our population was determined. From the correlations previously described, sensitivity, specificity, and likelihood ratio were established for EEG as a diagnostic tool. The likelihood ratio for a positive EEG result was determined by the equation: sensitivity/(1 2 specificity). After the pretest probability and likelihood ratio were calculated, the posttest probability was estimated with a nomogram.25

Results History and Diagnostic Evaluation Control Group. Eight Thorougbreds, 1 Tennessee Walking Horse, and 1 Quarter Horse constituted this group. They were presented to the Marion duPont Scott Equine Medical Center for chronic lameness (6 horses), for suspected wobbler (1 horse), and for scoliosis (1 horse), and the remaining 2 horses were purchased and appeared healthy. Six were geldings, 3 were females, and 1 was an intact male. The age of the animals ranged from 2 to 29 years with a mean 6 SD and median age of 12.6 6 7.7 years and 12 years, respectively. Findings from all neurologic examinations were within normal limits. Values from the CSF analyses were within published normal limits in all horses. Study Group. In this study, the most frequent complaint by the owners in horses with intracranial disorders was observation of seizures (single isolated episode or recurrent events). Among the 20 animals included in this study, 14 (70%) had a history of seizures, 3 had a history of abnormal behavior, and the 3 remaining horses were presented for other neurologic problems (collapse or incoordination). Three horses were treated with antiepileptic drugs before presentation. Quarter Horse, Arabian, Saddlebred, and Appaloosa were represented more than other breeds in the present study (6, 3, 3, and 3 horses, respectively). The age of the animals ranged from 1 day to 17 years with a mean 6 SD and median age of 6.1 6 1.3 years and 6 years, respectively. No gender predisposition was apparent (8 females, 7 geldings, and 5 males). Initial neurologic examination did not disclose any abnormalities in 3 animals. In the remaining 17 horses, cranial nerve dysfunction including nystagmus, head tilt, decreased palpebral light reflex, strabismus, dysphagia, and decreased sensation of the face was identified in 7 horses. Gait abnormalities such as weakness, ataxia, or spasticity of the front or rear limbs were observed in 10 horses (weakness, ataxia, and spasticity ranging from 1 to 4, according to the De Lahunta classifi-

Fig 1. Drawing of the forehead of a horse illustrating electroencephalographic electrode positioning and the 2 montages used: bipolar leftto-right, back-to-front montage (1) and bipolar circular montage (2). O, occipital; P, parietal; F, frontal; C, central.

cation26). Eight horses had generalized seizures during hospitalization and before EEG recordings. Results of CSF cytology were within normal limits in 14 patients. In 2 of these horses, Western blot analysis of CSF identified antibodies for S neurona. Cytology of the CSF was abnormal in 4 horses, including increased protein concentration, increased white blood cell count, or increased red blood cell count. In a 10-year-old mixed-breed gelding presented for abnormal behavior, cytology of xantachromic CSF disclosed a mixed inflammation associated with the presence of Cryptococcus neoformans. Skull radiographs were performed in 7 horses, and findings included thickening of the proximal part of the stylohyoid bone (1), hemorrhage into the guttural pouch (1), and fracture of the zygomatic arch (1) in 3 horses. However, the clinical relationship between these findings and the occurrence of seizures is questionable. In a 10-month-old Quarter Horse colt, serum chemistry results included a persistent hypocalcemia (ionized calcium, 3.5 mg/dL; reference range, 6.0–6.5 mg/dL), which was believed to be the cause of the seizures. Titers for EEE or WEE were negative in both horses examined. Based on the diagnostic tests described above, clinical diagnoses in 7 horses included idiopathic epilepsy, cryptococcal meningitis, neoplasia, neonatal maladjustment syn-

388

Lacombe et al

Fig 2. A normal adult equine electroencephalographic recording made under general anesthesia with a bipolar left-to-right, back-to-front montage. A moderate voltage (8–13 Hz) and moderate frequency (25–50 mV), without any asymmetry or paroxysmal activity were present. L, left; R, right; O, occipital; F, frontal.

drome, metabolic disorder, and equine protozoal myeloencephalitis.

Pathology Control Group. For all horses, no abnormalities were detected on gross examination of the brain and spinal cord, and on histopathogic examination of the 20 brain samples. Study Group. Brain and spinal cord were normal in 10 horses. Histopathologic diagnoses in the other 10 horses included nonsuppurative meningoencephalitis, cerebrocortical selective neuronal necrosis (neonatal maladjustment syndrome), congenital anomalies (porencephaly), cryptococcal meningitis, cerebral edema, focal malacia (without inflammatory process), diffuse gliosis, and Streptococcus equi abscess (associated with suppurative meningitis).

Electroencephalography Control Group. For the 10 control horses, moderate voltage (8–13 Hz) and moderate frequency (25–50 mV), without any asymmetry or paroxysmal activity, were present in 5 animals with a total score of 0. A high voltage (.50 mV) was recorded in the remaining horses because anesthetic drugs were suspected by the blinded reader to have induced artifact by changing the background activity. The total score of these EEG recordings was 2. Chi-square analysis of each category (frequency, voltage, asymmetry, and paroxysmal activity) indicated that observed scored EEG recordings (0 or 2) versus expected normal scored EEG recordings (score 0) from the normal horse population were not statistically different (P 5 .155), thus validating the scoring system.

Therefore, this semiquantitative analysis was applied for the interpretation of the EEG recordings in horses presented for suspected intracranial disorders. Study Group. Among the 20 interpretable EEG recordings, 35% (7/20) were considered normal, all of which were recorded under general anesthesia. A moderate voltage (8– 13 Hz) and moderate frequency (25–50 mV), without any asymmetry or paroxysmal activity, were present in 2 horses with a total score of 0 in adult horses (Fig 2). Higher scores were recorded in 2 horses (score 2 and 3) because of induction of background artifact by anesthetic drugs. Highvoltage slow wave activity was recorded in 3 foals, with a total score of 4 (Tables 1, 2). Among the 20 interpretable EEG recordings, 65% (13/ 20) had abnormal patterns, including diffuse high-voltage slow waves and focal or multifocal discrete paroxysmal activity. The distribution between generalized and asymmetric activity in these recordings was 46% (6/13) and 54% (7/ 13), respectively. Generalized slowing of frequency (,8 Hz) and increase in amplitude (.50 mV) were recorded in 31% of EEG recordings (4/13). Paroxysmal activity was recorded in 46% of abnormal EEG recordings (6/13; score $1; Table 1). Focal abnormal electrical activity (ie, restricted to a localized area of the scalp) was more frequently recorded over the frontal cortex than over the parietal or occipital cortex (80, 13, and 7% abnormalities detected, respectively). The distribution of the abnormal focal activity between right and left forebrain was 9 and 6, respectively. Among these 13 abnormal recordings, 8 semiquantitative analyses were scored between 2 and 4, and 2 semiquantitative analyses were scored .4 (Table 2). Diffuse high-


Table 1. Summary of semiquantitative analysis of electroencephalographic (EEG) recordings in 20 horses. No. EEG Recordings Low

Moderate

High

Total

Frequency Normal Abnormal

4 4

2 5

1 4

7 13

Voltage Normal Abnormal

0 2

4 6

3 5

7 13

Asymmetry Normal Abnormal

6 6

1 7

0 0

7 13

Paroxysmal activity Normal Abnormal

6 7

1 5

0 1

7 13

voltage slow wave and multifocal discrete paroxysmal discharge activity was found in 2 horses with cerebrocortical selective neuronal necrosis, the latter being consistent with neonatal maladjustment syndrome (Fig 3). EEG indicated a generalized continuous high-voltage slow activity in a Belgian colt suffering from an abscess in the cortex of the left cerebral hemisphere. In a 10-year-old mixed-breed gelding with cryptococcal meningitis, intermittent waxing and waning high-voltage activity superimposed by some sharp waves, recorded during EEG, was compatible with a neocortical inflammatory process. The EEG pattern of hypocalcemia-induced seizures included diffuse and symmetric slow wave activity associated with discrete epileptiform activities. Horses tranquilized with xylazine showed highvoltage slow wave activity in 7 instances. Other artifacts included muscle artifacts including eye and ear movements. A positive correlation between EEG findings and clinical diagnoses was established for 6 horses. A positive correlation between EEG findings and histopathologic findings was identified in 85% of the horses (17/20). EEG patterns were abnormal in all horses with postmortem abnormalities (sensitivity 100%) with a positive neuroanatomic correlation in 7 animals (70%; Table 3). In 10 animals, the cause of neurologic signs could not be explained at postmortem examination, and in 7 of these horses, the EEG pattern was normal (specificity 70% with 95% CI 42–98%). Abnormal EEG recordings were obtained from 3 horses with normal findings from neurologic examinations and from 8 horses with normal findings from CSF analysis (Table 4). Abnormal EEG recordings that correlated with histologic disease were obtained from 1 horse with normal findings from neurologic examination and from 7 horses with normal findings from CSF analysis. Despite moderate specificity, EEG had the highest sensitivity and specificity when compared to neurologic examination and other diagnostic tests (eg, CSF analysis) used in the workup of equine intracranial disorders (Table 5). The likelihood ratio for a positive EEG result was 3.3. Based on pretest probability of 50% in our population, we concluded that the posttest probability of EEG was 76%.

389

Table 2. Number of electroencephalographic recordings based on total score assigned during the semiquantitative analysis in 20 horses. Total score

0

1

2

3

4

5

6

7

8

Total

Normal Abnormal Total

2 0 2

0 3 3

1 4 5

1 1 2

3 3 6

0 1 1

0 0 0

0 1 1

0 0 0

7 13 20

Discussion These results demonstrate that EEG has excellent sensitivity with a good specificity (70%) in our population of horses with neurologic disease. Despite the retrospective nature of this study and lack of matched controls, EEG seems to be a useful diagnostic tool for diagnosis of equine intracranial diseases. This study demonstrated the utility of general anesthesia for EEG recordings in horses. EEG recording on anesthetized horses is preferable because the majority (69% or 9/ 13) of the tracings performed on standing horses with chemical restraint were not interpretable. However, consideration must be given to the degree of sedation and the type of drug used, both of which may interfere with EEG interpretation by changing background activity. In humans and small animals, awake recordings on patients are optimal because general chemical and gas anesthetics alter the normal EEG pattern by altering cortical activity.2 Therefore, a cautious interpretation of the EEG recording obtained under general anesthesia is crucial to avoid overinterpretation of drug-induced changes such as xylazine-induced high-voltage, slow wave activity. These findings are consistent with previous work on equine EEG.27 Xylazine-sedated horses displayed a hypersynchronous EEG pattern, with dominant activity being 1–3 Hz, 10–70 mV, and ketamine administration 3 minutes after xylazine injection caused an increase in high-voltage slow activity.13 Therefore, premedication with xylazine or induction with ketamine before EEG recording is not recommended. The depth of anesthesia also can interfere with the electrical background. EEG recordings made in ponies under halothane or isoflurane changed with the depth of anesthesia from high frequency and low amplitude to low frequency and high amplitude.12 Moreover, general anesthesia also has the potential to increase the threshold for seizures in patients during EEG. Antiepileptic properties of halothane and isoflurane have been demonstrated in humans, and ketamine hydrochloride has both proconvulsant and anticonvulsant properties.28 Changes during deep anesthesia induced by isoflurane included burst suppression, spikes, and isoelectric periods in ponies.12 Early detection and elimination of other types of artifacts (eg, muscle artifact including eye and ear movements) was critical for high-quality recording under general anesthesia. Collectively, our results demonstrate that use of a standardized anesthetic protocol enhances the quality of tracing with a decrease in the number of uninterpretable EEG recordings. The presence of a moderate frequency (8–13 Hz) and voltage (25–50 mV) has been established as normal for adult sedated horses in previous studies,19,27 and the normal

390

Lacombe et al

Fig 3. (Top) Electroencephalographic (EEG) recording made under general anesthesia from a 1-day-old filly that presented for lack of nursing and intermittent tremors. This recording revealed diffuse high-voltage slow waves and focal discrete paroxysmal discharge activity. Mutifocal paroxysmal activity was seen on the remaining EEG recording. (Bottom) Histopathologic examination (parietal cortex) showed cerebrocortical selective neuronal necrosis (arrows), compatible with neonatal maladjustment syndrome.


Table 3. Sensitivity and specificity of electroencephalography (EEG) as diagnostic tool in equine intracranial disorders in 20 horses.a Abnormal Postmortem Examination Findings

Normal Postmortem Examination Findings

Total

10 0 10

3 7 10

13 7 20

Abnormal EEG recording Normal EEG recording Total a

391

Table 5. Sensitivity and specificity of neurologic examination, cerebrospinal fluid (CSF) analysis, and electroencephalography (EEG) as diagnostic aids in equine intracranial disorders in 20 horses.a Diagnostic Tests

Sensitivity (%)

Specificity (%)

Neurologic exam CSF analysis EEG

90 (71–100) 33 (2–64) 100

20 (0–45) 67 (36–98) 70 (42–98)

a

Confidence intervals are given in parentheses.

Sensitivity: 100%; specificity: 70% (CI 42–98%).

pattern in neonatal sedated foals has been characterized by high voltage and slow wave activity.19 However, the establishment of normal values is difficult because frequency and amplitudes are state-dependant (awake, sedated, or anesthetized) and EEG is influenced by the method of restraint, technical aspects of recording and the subjectivity of interpretation. In this study, all normal EEG recordings had a total score of 0 except for normal foals (with a characteristics EEG pattern as described above), and for horses with drug-induced artifacts. Developmental changes in the EEG recordings of calves occurred during the first 10 weeks and were characterized by the appearance of an alpha rhythm.29 To the best of our knowledge, the age at which mature EEG develops in horses is unknown. The substantial increase between pretest and posttest probability (126%) confirmed the usefulness of EEG by adding significant sensitivity to the diagnostic testing, which is in accordance with findings in humans.1 Because EEG recordings obtained from 1 horse with normal findings from neurologic examination and from 7 horses with normal findings from CSF analysis indicated abnormal electrical patterns that correlated with histologic disease, EEG can be considered as useful test for early detection of functional intracranial disorders.20 Moreover, in comparison to the neurologic examination and other diagnostic testing, EEG was more helpful in making a diagnosis, as demonstrated in small animals.30 However, the practical applications of EEG are influenced by cost and limited availability of equipment and trained personnel. Therefore, EEG remains a tool of referral institutions. Because of moderate specific-

ity, EEG is a complementary test to, rather than an alternative to, neurologic examinations and other diagnostic tests, such as imaging studies and CSF analysis. Additionally, despite excellent sensitivity, we cannot exclude the possibility that we failed to detect seizures during EEG recordings, because of the effect of the drugs (as previously described) or because of intermittent episodes or the limited time of recording. However, frequent and prolonged sophisticated EEG monitoring, as routinely performed in humans, is not practical in large animals. Therefore, an animal suspected of having epilepsy should undergo a 2nd EEG examination if the 1st EEG recording is normal.31 The negative and positive predictive values are directly correlated to sensitivity of the test and prevalence of intracranial disorders, respectively (Fig 4).25 Thus, these parameters are less useful than sensitivity, specificity, and odds-ratio to assess the validity of a diagnostic test. The high number of abnormal EEG recordings (65% or 13/20) found in the study may be caused by the fact that EEG were performed in 14 horses (70%) that had already experienced several episodes of seizures before presentation, increasing the pretest probability of intracranial disease and the chance of characterizing the underlying cause of the seizure. Moreover, 40% of the horses developed seizures before EEG, and the proximity of seizures to the EEG examination increases the frequency of epileptiform activity present in the EEG recording.3 Despite a moderate specificity, a good neuroanatomic correlation between EEG and histopathologic findings was established in 70% of the cases (7/10), which makes EEG a potentially useful tool for

Table 4. Correlation between clinical, clinicopathologic, and histopathologic characteristics, and electroencephalographic (EEG) findings in 20 horses. No. of Horses Neurologic Examination Findings

EEG recording Normal Abnormal

Cerebrospinal fluid Analysis Findings

Normal

Abnormal

Normal

Abnormal

0 3

7 10

4 8

2 4

8 9

6 6

3 3

Postmortem examination findings Normal 2 Abnormal 1

Fig 4. Positive predictive value (PPV) of electroencephalography based on prevalence of intracranial disorders, with a given sensitivity of 100% and a specificity of 70%.

392

Lacombe et al

neuroanatomic localization of intracranial disorders in horses in a manner similar to that used in humans.1 However, EEG does not provide good neurolocalization in some cases described in the literature and in our study. For example, the EEG pattern is not pathognomonic for abscess localization32 and generalized high-voltage slow wave activity was recorded in most dogs with intracranial masses.33 Moreover, a focal slow frequency pattern has been interpreted as an expression of underlying structural brain damage and supports the diagnosis of epilepsy in humans and dogs, and similar findings occur with focal mass lesions.3 The high frequency of abnormal frontal rather than parietal and occipital electrical activity in our study may be explained by the anatomic variations between these areas (presence of a thin frontal bone).34 The EEG patterns described in this study are similar to those described previously in the veterinary literature. For example, the distribution between asymmetric and generalized activity in our recordings (54 and 46%, respectively) was almost similar to the distribution observed in epileptic dogs (41 and 59%, respectively).35 EEG recordings in sheep and cattle with experimental cerebrocortical necrosis included continuous slow wave and high-amplitude activity.20,21 Similar to our findings, the EEG pattern of dogs with intracranial masses, such as abscesses, included generalized slow wave activity and was not specific as to localization.33 The main difficulty encountered in this study included the inherent limitations of attempting to quantify analyses of equine EEG recordings because visual analysis is inherently biased by its subjective nature, regardless of whether the interpreter is blinded to the case or not. The semiquantitative method used in this study is a novel method of EEG analysis and used a scoring system to establish different categories from information obtained by visual analysis. Our goal was to provide consistency among EEG interpretations as well as some indication of the severity of changes. The latter then can be used to correlate clinical and histopathologic findings. A number of scoring systems have been employed in the evaluation of EEG activity in human medicine and research setting.36–38 The method of scoring EEG recordings by visual analysis has been employed in evaluating EEG recordings in humans before the advent of quantitative EEG.39 In particular, the degree of abnormal slowing of the EEG recording correlated quite well in human patients exhibiting bromism by means of a similar scoring system.39 However, this scoring system had not been validated in horses until the current study. Even if this approach gives consistency in the interpretation of EEG recordings, the total score of the semiquantitative analysis does not necessarily reflect normality. For instance, EEG recordings from 7 adult horses (5 from the control group and 2 from the clinical group) were interpreted as normal by the blinded reader despite a total score greater than 1 because of drug-induced background artifacts. In summary, the use of a semiquantitative analysis partially limits the variability inherent in subjective interpretation. Another limitation was variability caused by small sample size. The necessary matching of EEG and histopathologic findings limited the numbers of patients that could be included in our study. Ideally, a prospective study with matched con-

trols (by age and breed) should be performed to confirm the diagnostic utility of EEG in horses. Horses with clinical neurologic disease that lacked postmortem lesions were identified. Such findings are consistent with functional disorders (eg, idiopathic epilepsy), but we cannot exclude the possibility that our tissue sampling procedure failed to detect focal histologic lesions. No correlation was assigned between EEG and postmortem examination findings when the clinical and histopathologic significance of a lesion was uncertain; but we could not exclude the possibility that such lesions were the cause of the intracranial disorders. For example, in 1 horse, left cerebral paroxysmal activity was recorded on the EEG recording and at postmortem examination a left frontal cortex ependymal-lined cavitation was detected, but no correlation was assigned. Cholesterol granuloma was reported in 2 horses, but it was not considered as a cause of seizures. Because of the difficulty in imaging the equine brain, EEG can be considered as the most important diagnostic examination in horses suffering from intracranial disorders, as in small animal and human epileptology.31,40 This tool may allow the clinician to construct a list of differential diagnoses that helps identify the cause of the seizures in the light of the clinical and clinicopathologic findings, and to plan the therapy. Serial recording also may be helpful as a prognostic indicator for following the response to therapy and disease progression.31 Comparison of EEG and advanced imaging diagnostic techniques such as computed tomography or magnetic resonance imaging for horses with intracranial disorders needs to be made in the future. In conclusion, diagnosis of intracranial disease remains a challenge in equine medicine. However, EEG provides a sensitive and noninvasive method of clinical evaluation of the central nervous system that may enhance the clinician’s understanding of the functional and anatomic sequelae of the underlying disease process in horses.

Footnotes Rompun, Miles, Shawnee, KS Guailaxin, A-H Robins, Richmont, VA c Pentothal, Abott/Ceva Laboratories, North Chicago, IL d Aerrane, Anaquest Inc, Liberty Corner, NJ e Neurofax, Nihon-Kodon, Irvine, CA f Carbocaine, Abbott, North Chicago, IL g Bio-Tal, Bio-Ceutic, St Joseph, MO h Halothane, Fort Dodge Laboratories, Fort Dodge, IA i Ketaset, Aveco company, Fort Dodge, IA j Valium, Roche, Exton, PA k Grass model 8, Grass, Amherst, MA a

b

Acknowledgments We would like to acknowledge Jenny Bolan for her technical assistance and the Virginia Tech Clinical Research Program for partial funding. This study was performed at the Department of Veterinary Clinical Sciences, College of Veterinary Medicine, The Ohio State University, Columbus, OH; and at the Marion duPont Scott Equine Medical Center, Virginia-Maryland Regional College of Veterinary


Medicine, Virginia Polytechnic Institute and State University, Blacksburg, VA.

References 1. Aminoff MJ. Electroencephalography: General principles and clinical applications. In: Aminoff MJ, ed. Electrodiagnosis in Clinical Neurology, 2nd ed. New York, NY: Churchill Livingstone; 1986:21– 75. 2. Redding RW. Electroencephalography. In: Oliver JE, Horlein BF, Mayhew IG, ed. Veterinary Neurology. Philadelphia, PA: WB Saunders; 1987:111–145. 3. Berendt M, Hgoenhaven H, Flagstad A, et al. Electroencephalography in dogs with epilepsy: Similarities between human and canine findings. Acta Neurol Scand 1999;99:276–283. 4. Ruckebush Y, Barbey P, Guillemot P. Les e´tats de sommeil chez le cheval. C R Se´ances Soc Biol 1970;164:638–665. 5. Ruckebush Y, Gaujoux M, Eghbali B. Sleep cycles and kinesis in the foetal lamb. Electroencephalogr Clin Neurophysiol 1977;42: 226–237. 6. Ruckebush Y. Sleep deprivation in cattle. Brain Res 1974;78: 495–499. 7. Ruckebush Y. Development of sleep and wakefulness in the foetal lamb. Electroencephalogr Clin Neurophysiol 1972;32:119–128. 8. Ruckebush Y. The relevance of drowsiness in the circadian cycle of farm animals. Anim Behav 1972;20:637–643. 9. Usenick MW, Kitchell RL, Herschler RG, et al. A surgical technique for permanent implantation of electrocorticographic electrodes in the burro and pigs. Am J Vet Res 1962;23:70–73. 10. Stromberg MW, Kitchell RL, Unsenik FA, et al. Electrocorticographic patterns in normal burros and pigs. Am J Vet Res 1962;23: 737–743. 11. Grabow JD, Anslow RO, Spalatin J. Electroencephalographic recordings with multicontact depth probes in a horse. Am J Vet Res 1969;30:239–243. 12. Auer JA, Amend JF, Garner HE, et al. Electroencephalographic responses during volatile anesthesia in domestic ponies: A comparative study of isoflurane, enflurane, methoxyflurane, and halothane. J Equine Med Surg 1979;3:130–134. 13. Purohit RC, Mysinger PW, Redding RW. Effects of xylazine and ketamine hydrochloride on the electroencephalogram and the electrocardiogram in the horse. Am J Vet Res 1981;42:615–619. 14. Ekstro¨m PM, Short CE, Geimer TR. Electroencephalography of detomidine-ketamine-halothane and detomidine-ketamine-isoflurane anesthetized horses during orthopedic surgery. Vet Surg 1993;22:414– 418. 15. Johnson CB, Young SS, Taylor PM. Analysis of the frequency spectrum of the equine electroencephalogram during halothane anaesthesia. Res Vet Sci 1994;56:373–378. 16. Miller SM, Short CE, Ekstro¨m PM. Quantitative electroencephalographic evaluation to determine the quantity of analgesia during anesthesia of horses for arthroscopic surgery. Am J Vet Res 1995;56: 374–379. 17. Otto KA, Voigt S, Piepenbrock S, et al. Differences in quantitated electroencephalographic variables during surgical stimulation of horses anesthetized with isoflurane. Vet Surg 1996;25:249–255. 18. Johnson CB, Taylor PM. Comparison of the effects of halothane, isoflurane and methoxyflurane on the electroencephalogram of the horse. Br J Anaesth 1998;81:748–753. 19. Mysinger PW, Redding RW, Vaughan JT, et al. Electroenceph-

393

alographic patterns of clinically normal, sedated, and tranquilized newborn foals and adult horses. Am J Vet Res 1985;46:36–41. 20. Suzuki M, Sitizyo K, Takeuchi T, et al. Electroencephalogram of Japanese black calves affected with cerebrocortical necrosis. Jpn J Vet Sci 1990;52:1077–1087. 21. Itabisashi T, Horino R, Hirano K, et al. Electroencephalographic observation on sheep and cattle with experimental cerebrocortical necrosis. Jpn J Vet Sci 1990;52:551–558. 22. Dunigan CE, Oglesbee MJ, Podell M, et al. Seizures activity associated with equine protozoal myeloencephalitis. Prog Vet Neurol 1995;6:50–54. 23. Sharbrough FW. Electrical fields and recording techniques. In: Daly DD, Pedley TA, ed. Current Practice of Clinical Electroencephalography. New York, NY: Raven Press; 1990:29–49. 24. Zifkin BG, Cracco RQ. An orderly approach to the abnormal EEG. In: Daly DD, Pedley TA, ed. Current Practice of Clinical Electroencephalography. New York, NY: Raven Press; 1990:253–267. 25. Sackett DL, Haynes RB, Guyatt GH, et al. Clinical Epidemiology: A Basic Science for Clinical Medicine, 2nd ed. Boston, MA: Little Brown; 1991:89–139. 26. De Lahunta A. Veterinary Neuroanatomy and Clinical Neurology, 2nd ed. Philadelphia, PA: WB Saunders; 1983:218–219. 27. Andrews FM, Fenner WR. Indication and use of electrodiagnostic aids in neurologic disease. Vet Clin North Am Equine Pract 1987;3:293–322. 28. Engel J, Pedley TA. Epilepsy. Philadelphia, PA: Lippicott-Raven; 1998:1978–1984. 29. Takeuchi T, Sitizyo K, Harade E. Analysis of the electroencephalogram in growing calves by use of power spectrum and cross correlation. Am J Vet Res 1998;59:777–781. 30. Knecht CD, Sorjonen DC, Simpson ST. Ancillary tests in the diagnosis of seizures. J Anim Hosp Assoc 1984;20:455. 31. Klemm WR. Electroencephalography in the diagnosis of epilepsy. Probl Vet Med 1989;1:535–557. 32. Strain GM, Claxton MS, Turnquist SE, et al. Evoked potential and electroencephalographic assessment of central blindness due to brain abscesses in a steer. Cornell Vet 1987;77:374–382. 33. Steiss JE, Cox NR, Knecht CD. Electroencephalographic and histopathologic correlations in eight dogs with intracranial mass lesions. Am J Vet Res 1990;51:1286–1291. 34. Shorvon SD. The routine EEG. In Dam M, Gram L, eds. Comprehensive Epileptology. New York, NY: Raven Press; 1990:321–338. 35. Jaggy A, Bernardini M. Idiopathic epilepsy in 125 dogs: A long-term study. Clinical and electroencephalographic findings. J Small Anim Pract 1998;39:23–29. 36. Balzamo E, VanBeers P, Lagarde D. Scoring of sleep and wakefulness by behavioral analysis from video recordings in rhesus monkeys: Comparison with conventional EEG analysis. EEG Clin Neurophysiol 1998;106:206–212. 37. MacLean AW, Lue F, Moldofksy H. The reliability of visual scoring of alpha EEG activity during sleep. Sleep 1995;18:565– 569. 38. Drane DB, Martin WB, Viglione SS. The application of pattern recognition techniques to the scoring of EEG sleep patterns. Electroencephalogr Clin Neurophysiol 1971;30:94–95. 39. Greeblatt M, Levin S, Schegloff B. Electroencephalographic findings in cases of bromide intoxication. Arch Neurol Psychiatry 1945;53:431–436. 40. Engel J. Seizures and Epilepsy. Philadelphia, PA: Davis Company; 1989:312.