SEARCH

SEARCH BY CITATION

Keywords:

  • Psychogenic nonepileptic seizures;
  • Time–frequency mapping;
  • EEG

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Summary: Purpose: Approximately 30% of patients admitted for video-EEG monitoring have psychogenic nonepileptic seizures (PNES). Differentiation of “convulsive” PNES from convulsive seizures can be difficult. The EEG often displays rhythmic movement artifact that may resemble seizure activity and confound the interpretation. We sought to determine whether time–frequency mapping of the rhythmic EEG artifact during “convulsive” PNES reveals a pattern that differs from that of epileptic seizures.

Methods: EEGs from 15 consecutive patients with “convulsive” PNESs were studied with time–frequency mapping by using NEUROSCAN and compared with 15 patients with convulsive epileptic seizures. Fast Fourier transforms (FFTs) were performed to determine the dominant frequency for 1- to 2-s windows every 2 s through the seizures.

Results: The dominant frequency remained stable within a narrow range for the duration of the PNES, whereas in the epileptic seizures, it evolved through a wide range. The coefficient of variation of the frequency during the seizures was considerably less for patients without epilepsy (median, 15.0%; range, 7.2–23.7% vs. median, 58.0%; range, 34.8–92.1%; p < 0.001). The median frequency did not differ significantly between groups (4.2 vs. 4.6 Hz; p = 0.290).

Conclusions: “Convulsive” PNES display a characteristic pattern on time–frequency mapping of the EEG artifact, with a stable, nonevolving frequency that is different from the evolving pattern seen during an epileptic seizure.

Individuals with psychogenic nonepileptic seizures (PNES) have recurrent episodes of altered movement, sensation, or experience that resemble epileptic seizures but are not associated with abnormal electrical activity in the brain (1). The etiology of PNES remains unclear; however, they are presumed to relate to underlying psychogenic disturbances, with multiple factors including personality traits playing a role in both etiology and prognosis (2).

PNES represent a common diagnostic and management problem, not just for the neurologist, but also for general practitioners, emergency departments, and other treating physicians. The estimated prevalence of PNES is between 2 to 33 per 100,000 individuals, making PNES one of the more common conversion disorders in the community (3). It has been reported that between 11 and 54% of patients admitted for inpatient video-EEG monitoring (VEM) are diagnosed as having PNES (1,4–7). A significant proportion of patients requiring intubation for status epilepticus are ultimately diagnosed with PNESs, resulting in potentially serious morbidity to the patient, and considerable cost to the community (8).

The differentiation between epileptic and nonepileptic seizures can be difficult. Seizure semiology is important in the diagnostic algorithm, and a plethora of features have been reported as being more likely associated with PNES. These have included a stable ictal heart rate (9), induction of the event with suggestion (10), eyes closed versus open, pelvic thrusting or “no-no” head shaking during the event (11), and even the presence of a teddy bear brought in during EEG monitoring (12). However, none of these features alone is diagnostic of PNES, as all can be seen on occasions during epileptic seizures. VEM is the investigation of choice in confirming the diagnosis of seizure disorders. However, even with EEG monitoring, the diagnosis may not be straightforward. The rhythmic movement artifact often obscures the EEG during a seizure, which can confound interpretation. Rhythmic movement artifact may also resemble spike–wave abnormality on the EEG, making the distinction between epileptic and nonepileptic seizure disorder difficult (13) (Fig. 1). Additionally, VEM monitoring is highly resource and labor intensive and therefore is relatively expensive and limited in availability. More definitive diagnostic methods for PNES are required, particularly those that may be applicable outside of the VEM setting.

image

Figure 1. Segment of an EEG during a psychogenic nonepileptic seizure, with rhythmic movement artifact resembling spike–wave discharges.

Download figure to PowerPoint

We have observed that the evolution of the frequency of the movements and rhythmic EEG artifact in PNES differs from that in convulsive epileptic seizures, and this may be a reliable and practical differentiating feature. However, no study has systematically investigated the EEG patterns and rhythmic movement artifact across PNES and epilepsy groups. We have hypothesized that the frequency remains stable over time in PNES, but evolves over time in convulsive epileptic seizures. The aim of this study was to determine whether time–frequency mapping of the rhythmic EEG artifact during “convulsive” PNES revealed a pattern that differed from that seen during epileptic seizures.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Patients

The EEGs of consecutive patients with “convulsive” PNES were analyzed from the database of the Royal Melbourne Hospital and Monash Medical Centre from January 2001 until July 2003. During this period, 360 patients underwent VEM, of whom 86 (24%) were diagnosed with PNES. Of these 86 patients, 59 (69%) patients had nonconvulsive PNES, and 27 (31%) patients had convulsive PNES. “Convulsive” PNES were defined as psychogenic motor seizures as per the classification described by Groppel et al. (14), characterized by the occurrence of simultaneous clonic and motor movements of the upper or lower limbs or both, recurrent head movements, pelvic thrusting movements, tonic posturing of the head, or a combination of these. The diagnosis of a nonepileptic seizure was made primarily on seizure semiology evident on the video but confirmed with history, EEG recording, neuropsychiatry, and epileptologist's opinion. Eligible patients for this study had “convulsive” PNES with movements during an event that produced rhythmic movement artifact on the accompanying EEG. These events were included if they lasted ≥30 s and were analyzed to a maximum time of 180 s. Up to five seizures for each patient were analyzed. Of the 27 patients with convulsive PNES, nine patients did not have rhythmic movement artifact on the EEG lasting ≥30 s, and three patients had data files that were not able to be analyzed because of technical problems with the data archiving. Thirty-two seizures from 15 patients with convulsive PNESs were analyzed for the study.

The 15 patients with PNES included in the study comprised six male and nine female patients, and the epileptic seizure group, four male and 11 female patients. The median age of the PNES patients was 39 years (range, 33–59 years), and of the epileptic seizure patients, 37 years (range, 14–56 years). Twelve of the 15 PNES patients had normal interictal EEGs, with the remaining three patients having mild right-sided focal slowing but no frank epileptiform activity. Fourteen of the 15 PNES patients had normal magnetic resonance imaging (MRI) scans of the brain, with one patient having an area of frontal lobe gliosis after craniotomy. The ictal EEG in these patients showed only PNES, with none having epileptic seizures captured. The patients with epilepsy had various seizure disorders, including symptomatic generalized epilepsy, temporal lobe epilepsy, and seizures secondary to focal cortical malformations. No patients with frontal lobe epilepsy, which can mimic PNES, were included. All had generalized convulsive seizures from which the time–frequency mapping was performed. The median length of the seizure was greater in the PNES than in the epileptic seizures (125 s vs. 74 s; p = 0.021). The clinical characteristics of the patients are outlined in Table 1.

Table 1. Clinical characteristics of patients
Clinical featuresPNESEpileptic seizures
  1. PNES, psychogenic nonepileptic seizure; MRI, magnetic resonance imaging; SWD, spike–wave discharge.

Number of patients1515
Median age in yr (range) 39 (33–59)37 (14–56)
Male/Female6 : 94 : 11
Interictal EEG
Normal (%)12 (80%)6 (40%)
Abnormal
 Focal epileptiform activity07 (46%)
 Generalized SWD01 (7%) 
 Generalized slowing01 (7%) 
 Focal slowing 3 (20%)0
Ictal EEG
 PNESs only 15 (100%)0
 Generalized onset seizure03 (20%)
 Focal onset, secondarily generalized seizure012 (80%) 
MRI
 Normal14 (93%)9 (60%)
 Mesial temporal sclerosis03 (20%)
 Focal cortical dysplasia01 (7%) 
 Other focal cortical abnormality1 (7%)2 (13%)

Time–frequency mapping of events

The time–frequency analysis was performed with the aid of the NEUROSCAN (Compumedics, Melbourne, Australia) EEG analysis software package. The identified events were “blocked” for analysis (Fig. 2), and then divided into 1- and 2-s time epochs. Fast Fourier transforms (FFTs) were performed for each epoch and plotted as frequency–amplitude spectra (Fig. 3). The frequency–amplitude spectra were then visually inspected to determine the dominant primary frequency of every 2-s epoch. This dominant frequency was then plotted against time for the duration of the seizure, up to a maximum of 180 s, to produce the time–frequency map (Fig. 4). The patterns of evolution of the time–frequency maps for the events from the 15 PNES patients were qualitatively compared with those from the 15 epileptic seizure patients, particularly with regard to the evolution of the frequency over time.

image

Figure 2. Ten-second EEG recordings for a patient having a psychogenic nonepileptic seizure (PNES): (A) the beginning, (B) 90 s into the event, and (C) at the end of the event (180 s). Also shown is the EEG from a typical convulsive seizure: (A) the beginning, (B) 30 s into the event, and (C) at the end of the event (60 s). Although the EEG recording for the PNES resembles a rhythmic spike-and-wave discharge, the frequency of the activity stays the same throughout the event (4 Hz). In contrast, during the convulsive seizure, the activity evolves through different frequencies.

Download figure to PowerPoint

image

Figure 3. Frequency-amplitude spectra from fast Fourier transforms (FFT) of 2-s time window during the psychogenic nonepileptic seizure (PNES) from Fig. 1: (A) the beginning, (B) 90 s into the event, and (C) at the end of the event (180 s). Also shown are FFT frequency spectra from 2-s epochs during a typical convulsive seizure: (A) the beginning, (B) 30 s into the event, and (C) at the end of the event (60 s). In the PNES, the dominant frequency remains stable at 4 Hz. for all time windows, whereas during the epileptic seizure, the frequency varies between the windows.

Download figure to PowerPoint

image

Figure 4. Time–frequency plot of the dominant frequency for the sequential time windows during the event for (A) the psychogenic nonepileptic seizure (PNES) from Fig. 1, showing a very stable frequency at 4 to 4.5 Hz throughout the seizure. In contrast, the frequency of the activity for the epileptic seizures (B, C) progressively evolves over a wide range of frequencies during the seizure. Two major patterns were seen in the epileptic seizures: (B) The frequency started low and then progressively increased before decreasing again toward the end of the seizure; (C) the frequency started high and progressively declined during the seizure.

Download figure to PowerPoint

Statistical methods

A quantitative statistical analysis of the variability over time of the dominant frequency between the groups was performed by comparing its coefficient of variation (CV) for the multiple time points throughout the event. The CV is calculated by

  • image

The CV provides a measure of the percentage variability of the frequency, normalized to the mean frequency. A seizure in which the frequency evolves over a wide range will have a high CV (Fig. 4B,C), whereas those in whom the frequency remains stable during the seizure will have a low CV (Fig. 4A). The median CV for the 15 PNES patients was compared with that for the 15 epileptic seizure patients. When the patients had more than one seizure, the average CV for each patient was used for the intergroup analysis. The average frequency and average seizure duration also were compared between the groups. The Mann–Whitney U test (two-tailed) was used to compare groups.

To estimate sensitivity and specificity, patients were categorized according to CV ≤23.7 versus CV >23.7. This value corresponded to the highest observed CV in PNES patients, although complete separation of CV values was seen for each group in the range 23.7 to 34.8, so any cut point within this range would have given identical results. Sensitivity and specificity are given with exact binomial 95% confidence intervals.

Statistical analysis was performed with Stata Version 7.0 (19).

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

Time–frequency mapping

In the PNES, the dominant frequency remained stable within a narrow range for the duration of the event. From event onset, the frequency of the EEG artifact remained largely unchanged until the seizure end (e.g., Fig. 4A). In keeping with this observation, the CV of the mean frequency during a seizure was low, with a median of 15%, ranging from 7.2 to 23.7% (Fig. 5).

image

Figure 5. Box plot demonstrating the coefficient of variation (CV) of the frequency of the predominant rhythmic EEG activity during the seizure. The median for the psychogenic nonepileptic seizures was 15%, ranging from 7.2 to 23.7%, whereas the median for the epileptic seizures was 58%, ranging from 34.8 to 92.1%. The CV was significantly greater in the epileptic group (p < 0.001), with no overlap between the ranges in the groups, indicating that less variability was present in the frequency of the discharge during the seizures in these patients.

Download figure to PowerPoint

In contrast, during the epileptic seizures, the frequency of movement artifact evolved through a much wider range from seizure onset to offset. Two major types of patterns were seen. In the first (nine patients), the frequency started relatively low in the delta/theta range, progressively increased into the alpha/beta range, and then progressively decreased back into the delta range before terminating (Fig. 4B). In the second epileptic pattern, the frequency started fast in the alpha/beta range and then progressively decreased into the delta range by the end of the seizure (Fig. 4C). The CV of the dominant frequency during an epileptic seizure was high, with a median of 56%, ranging from 34.8 to 92.1%.

Table 2 compares the CV and the average frequency for the seizures in the 15 PNES and 15 epilepsy patients. The CV was greater in the epilepsy group (p < 0.001), with no overlap between the ranges in the groups (Fig. 5). CV >23.7 was 100% sensitive and 100% specific for distinguishing epileptic seizures from PNES (95% confidence interval for both sensitivity and specificity, 78.2–100%).

Table 2. Comparison of frequencies in groups of patients
Patient numberNonepileptic seizuresEpileptic seizures
Mean frequencyCV of frequencyMean frequencyCV of frequency
  1. CV, coefficient of variation; PNES, psychogenic nonepileptic seizure.

 15.9 7.58.654.3
 25.023.77.649.7
 34.011.58.271.3
 44.615.23.659.0
 55.4 8.54.692.1
 64.916.05.658.4
 74.215.03.834.8
 82.523.63.038.5
 94.5 7.43.043.2
104.217.47.847.5
114.121.13.967.2
124.2 7.24.362.5
134.611.34.949.6
143.522.84.660.4
153.017.64.968.3
Median for group4.215.24.658.4

The average dominant frequency of each seizure was similar between the PNES and the epilepsy groups (4.2 Hz; range, 2.5–5.8 Hz, vs. 4.6 Hz, range, 3.0–8.6; p = 0.290). In seven of the 15 PNES patients, brief pauses of movement lasting between 5 and 70 s during an event were noted, followed by resumption of movement at the same frequency.

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES

The distinction between epileptic and nonepileptic seizures is an important one with significant implications both for treatment and prognosis. The correct diagnosis allows potentially toxic antiepileptic drugs (AEDs) to be ceased and also saves hospital and medical resources being unnecessarily used in the mistaken belief that the patient is having uncontrolled epileptic seizures. Diagnosis of PNES results in a substantial reduction of direct medical costs (15). The correct diagnosis of PNES also creates the opportunity for referral for appropriate psychiatric assessment and treatment of psychological issues that may underlie the events. Long-term follow-up studies have shown that although the outcome of PNES is variable, poor prognostic features include older age at PNES onset and diagnosis, positive motor features, the presence of tongue biting and incontinence, and more extreme scores on measures of personality pathology in areas of emotional dysregulation, inhibitedness, and compulsivity (2).

Access to limited resources such as video-EEG monitoring is not uniform, and the clinical uncertainty of treating physicians often leads to delay in accurate diagnosis. This delay can further confound the management issues, as ≤80% of patients with PNES have been treated with AEDs before diagnosis, and ≤20% have a history of pseudostatus epilepticus (16).

The concept of evolution of frequency of the rhythmic activity during an epileptic seizure is well described (17,18). Epileptic seizure activity often evolves through many frequencies, from delta-range right through to beta-range frequencies over the course of a single seizure and is evident both from the ictal EEG and direct observation of the patient. This pattern has been demonstrated from scalp electrodes as well as intracranial monitoring and is a useful tool to analyze and categorize seizure disorders further. Two patterns of evolution of frequency of the EEG activity on time–frequency mapping were demonstrated during convulsive epileptic seizures in this study. In the most common pattern, the frequency started low in the delta range, steadily increased into the alpha/beta ranges, and then gradually decreased again back into the delta range before ceasing (Fig. 4B). In the second pattern, the seizure started with a fast frequency discharge in the alpha/beta range, which then gradually slowed into the delta range by the end of the seizure (Fig. 4C).

This study demonstrates that unlike epileptic seizures, convulsive PNES have a stable frequency of rhythmic movements, which produces a stable rhythmic artifact on the EEG. Analysis of this pattern reveals that patients with PNES have a frequency of movement that remains relatively constant throughout the seizure, with little variation or evolution. Our study demonstrated that this feature is 100% sensitive and specific for PNES, although this finding may possibly be related to the sample size. We are currently validating the diagnostic accuracy of time–frequency mapping with a larger prospective study. Whereas a potential confounding variable in this study is the use of the EEG in the diagnosis of the seizure disorder, the diagnosis of PNES was made predominantly by the seizure semiology on video and confirmed by history, epileptologist, and neuropsychiatric opinion. Although the EEG was analyzed for concomitant epileptiform activity, it was not paramount in the diagnostic algorithm, and patients were not diagnosed with PNES based on the EEG. Another characteristic finding among PNES was the presence of brief pauses in rhythmic movement during a seizure, lasting between 5 and 70 s. These pauses were followed by resumption of movement at the same frequency: the “on-off-on” pattern. No such pauses were observed during epileptic events. The explanation for this pattern during PNES may be that an individual has a set fundamental frequency at which subconscious or conscious rhythmic movement preferentially occurs. Therefore when the person resumes movements after a break, the frequency returns to this same frequency. In contrast, during epileptic seizures, the limb movements are driven by the frequency of the evolving seizure discharge, which overrides the normal motor control mechanisms.

Time–frequency mapping can reliably distinguish epileptic from nonepileptic seizures and remove any uncertainty with respect to EEG artifact or epileptiform activity. To our knowledge, this is the first study using time–frequency mapping of the EEG to document this stable frequency of movement in PNES. This pattern is unlike that seen in any convulsive epileptic seizure and may be a useful tool, both electrographically and at the bedside, to distinguish between epileptic and nonepileptic events. Clinically, this characteristic of PNES may be particularly useful in situations in which high-quality EEG services are not immediately available. A careful examination of the frequency of the limb movements, noting nonevolution, and in particular, if the “on-off-on” pattern is present, may allow correct diagnosis of PNES to be made before any potentially dangerous interventions being instituted. If necessary, the clinical observation could be confirmed by examining the frequency of the movement artifact on the rhythm strip of an ECG recording.

In this study we found complete separation of CV values for each group in the range of 23.7 to 34.8. We used the lower of these two values as the test cut-point simply for convenience. For the purpose of validation, we propose that initially a cut-point CV of 30 be used (i.e., approximate mid-point between 23.7 and 34.8). If subsequent studies show some overlap of CV values for each group, the cut-point can be adjusted to maximize sensitivity or specificity or both.

Accurate diagnosis of PNES is essential for both the long-term management of the patient and harm minimization. Unlike other clinical features of convulsive PNES, this study demonstrated that time–frequency mapping of the rhythmic movements has the potential to be used as a reliable diagnostic tool in the accurate classification of seizure disorders. A future application of these findings may be the use of a device to monitor the frequency of movements during events that will allow the outpatient diagnosis of PNES, avoiding the need for expensive inpatient VEM.

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. REFERENCES