Article Text
PDF
Abstract
Background Psychogenic non-epileptic seizures (PNES) or attacks consist of paroxysmal behavioural changes that resemble an epileptic seizure but are not associated with electrophysiological epileptic changes. They are caused by a psychopathological process and are primarily diagnosed on history and video-EEG. Clinical presentation comprises a wide range of symptoms and signs, which are individually neither totally specific nor sensitive, making positive diagnosis of PNES difficult. Consequently, PNES are often misdiagnosed as epilepsy. The aim of this study was to identify homogeneous groups of PNES based on specific combinations of clinical signs with a view to improving timely diagnosis.
Methods The authors first retrospectively analysed 22 clinical signs of 145 PNES recorded by video-EEG in 52 patients and then conducted a multiple correspondence analysis and hierarchical cluster analysis.
Results Five clusters of signs were identified and named according to their main clinical features:
dystonic attack with primitive gestural activity (31.6%);
pauci-kinetic attack with preserved responsiveness (23.4%);
pseudosyncope (16.9%);
hyperkinetic prolonged attack with hyperventilation and auras (11.7%);
axial dystonic prolonged attack (16.4%).
When several attacks were recorded in the same patient, they were automatically classified in the same subtype in 61.5% of patients.
Conclusion This study proposes an objective clinical classification of PNES based on automatic clustering of clinical signs observed on video-EEG. It also suggests that PNES are stereotyped in the same patient. Application of these findings could help provide an objective diagnosis of patients with PNES.
- Psychogenic non-epileptic seizure
- video-EEG
- diagnosis
- classification
- clinical neurology
- EEG
- epilepsy
- neuropsychiatry
- paroxysmal disorder
Statistics from Altmetric.com
- Psychogenic non-epileptic seizure
- video-EEG
- diagnosis
- classification
- clinical neurology
- EEG
- epilepsy
- neuropsychiatry
- paroxysmal disorder
Background
Psychogenic non-epileptic seizures or attacks (PNES) consist of paroxysmal behavioural changes comprising various combinations of motor, sensory, emotional signs and changes in consciousness. They may be mistaken for epileptic seizures (ES) because of semiological similarities. But they are not associated with ictal cerebral electrical discharges.1
Their prevalence is around 1 to 33/100 000, and they account for 5–20% of patients thought to have epilepsy and 24–32.3% of patients in epilepsy centres.2–6 The rate of associated epilepsy ranges from 5% to 50%.3 PNES occur mainly in young adults (second and third decade) and in women (75%).7 8
The semiological analysis of PNES is often made in contrast with the clinical features of epileptic seizures.8 No isolated clinical characteristic is specific and sensitive.4 9 10 Clinical variability from one attack to another was thought to be an important distinguishing criterion. However, only a few studies have attempted to describe different types of PNES as a conjunction of clinical signs. Three of these base their description on a subjective classification1 11 12 and two on an objective statistical method of classification.13 14 Both the latter focused on a limited number of clinical items (seven and 12 items) and identified three classes: (1) major motor, (2) minor motor or trembling and (3) unresponsive or atonic psychogenic seizures.
Despite these efforts, PNES remain a major cause of misdiagnosis after a first paroxysmal clinical event15 and constitute a public-health issue.
Our purpose was therefore to achieve an objective classification of PNES based on the conjunction of clinical symptoms and thus to improve diagnostic accuracy.
Methods
Patients
We retrospectively reviewed video-EEG and medical records of all adult patients who underwent a prolonged video-EEG monitoring from January 2003 to June 2008 at the Clinical Neurophysiology unit of the University Hospital of Nancy, France. The records of 55 consecutive patients having PNES with or without associated epilepsy, with or without any treatment but with at least one PNES recorded on video-EEG were retained for analysis. PNES was diagnosed by consensus on video-EEG analysis and medical-history data by two epileptologists. The following definition of PNES was applied: paroxysmal episodes of motor, sensory, emotional signs or changes in consciousness, not associated with ictal cerebral electrical discharges, after having eliminated other pathologies (cardiological in particular). All recorded seizures of each patient were studied.
Patients had undergone video-EEG for one of the following reasons: diagnostic elucidation of refractory seizures, presurgical evaluation as part of an epilepsy surgery programme or suspected PNES.
Recordings
Recordings (8 h per day over 1 to 3 days) had been made with 23 EEG electrodes according to the 10/20 international system, at a 1 kHz sampling rate (SD64 Headbox, Micromed, Italy). Diurnal sleep and wakefulness EEGs were performed by requesting the patients to perform up to one hyperventilation per recording hour every day for 6 min with their eyes closed, under the supervision of a trained EEG technologist.16 Facial movements were analysed from a high-resolution zoom video. A specialised video-EEG nurse under medical supervision collected any subjective data observed during the seizures or immediately afterwards.
Data collection
Sociodemographical and medical history data were gathered from standardised medical files which had been specifically designed for the video-EEG monitoring unit prior to the study: sex, age at onset of symptoms, age at diagnosis, antiepileptic treatment, current and past psychiatric comorbidity, history of psychological traumatism, results of cerebral imaging and previous EEG, previous tongue biting, loss of urine, trauma during a seizure, stays in intensive care units.
For the purpose of the study, 22 clinical variables were collected from the video-EEG recordings and entered into a statistical database by the same observer (CH). These variables were chosen prior to the study, based on previously published studies and on our own experience.1 8 11–14 This choice also reflected a compromise between the accuracy of description and the complexity of the subsequent statistical analyses. We therefore prioritised the most objective signs such as motor manifestations.
The 22 clinical variables collected for each seizure were: duration of seizure (<1 min, 1–5 min, >5 min), prodromes (dizziness, feeling of strangeness, abdominal pain…), modality of onset and end (sudden or progressive), responsiveness (interactions with observer through verbal answers, execution of instructions), dystonic movements, tremor, myoclonus, focal signs (one limb, head, halfbody, two lower or two upper limbs), axial extension, axial immobility, closed eyes, side-to-side head shaking, one-sided rotation of the head, face asymmetry, movements of the mouth (chewing, protraction of the tongue), vocalisation, hyperventilation and vegetative signs (sweating, flushes, pallor), sensory manifestations (paraesthesias), hypermotor primitive gestural activity (looking oriented to a purpose, ie, hiding one's face, punching, grasping), variation of symptom intensity (eg, tremor fluctuating between high and low amplitude or frequency) and postictal state (defined as the abnormal condition occurring between the end of an attack and return to baseline condition).
Clinical data were systematically reviewed by an expert epileptologist (LM or JPV), not blinded to the findings of the first. If there was disagreement, the final interpretation was based on a consensus between the two observers.
Statistical analysis
The first part of the analysis was descriptive (mean and standard deviations for continuous variables, absolute and relative frequencies for categorical variables).
The second part involved determining a typology of PNES, using the multiple correspondence analysis and the hierarchical clustering analysis.
A multiple correspondence analysis was first performed to construct principal components optimally summarising the data using dichotomised variables (semiological parameters).17 This method generates a set of coordinate values summarising the seizure and the semiological categories, and thus allows associations between the seizure and the semiological parameter to be displayed graphically (CORRESP procedure, SAS). Each principal component was interpreted by examining the contribution of each semiological category to the variance of the category coordinates. The contribution of a category was considered significant when it was above the average contribution 1/p (where p is the total number of categories of the variables). Finally, graphical displays of the categories of semiological parameters were constructed using the principal components in a series of two-dimensional graphs, plotting one component against another within a set of axes (PLOT procedure).
An ascending hierarchical classification was then used to determine the number of clusters (CLUSTER procedure, Ward's minimum variance method) from these principal components (Everitt BS). The cubic clustering criterion was positive, which means that the seizures were distributed according to a continuum in the space constituted by the components used as axes systems. χ2tests were used to determine which semiological parameter contributed to the formation of cluster groups to interpret the different profiles of PNES.
All statistical analyses were performed with SAS v9.1.
Results
Demographical and medical history data
Out of the 55 patients initially included, three were excluded for technical reasons (insufficient quality of recordings). The sex ratio was 0.27 (38 women and 14 men); mean age at diagnosis of PNES 34.9 years (12–68); and mean age at onset of PNES evaluated at 30 years. Twenty-four patients (46.2%) had a confirmed associated epilepsy (PNES+E), and 28 (53.8%) had PNES only (PNES), among whom 19 (67.8%) had been misdiagnosed as epileptic and were on long-term antiepileptic treatment (14 patients being on two or more antiepileptic drugs). Interictal and ictal EEGs and cerebral CT or MRI were available for all patients. Among the PNES patients, 17.8% had interictal EEG abnormalities (four focal and one generalised slow waves). None of the PNES+E patients had normal interictal EEG (12 patients had focal slow waves, 15 paroxysmal events and three generalised slow waves). Twenty-eight per cent of PNES and 66.6% of PNES+ES patients had abnormal cerebral imaging. Six PNES patients and nine PNES+E patients had a family history of epilepsy. Eleven patients had been hospitalised in an intensive care unit, among whom five (9.6% of the entire cohort) were for pseudo-status. Among the 29 patients with a documented history of psychological trauma, 11 (37.9%) had experienced physical violence and 10 (34.6%) sexual aggression.
In the PNES group, six patients had a history of seizure-related injury, six of urinary incontinence and five of nocturnal seizure. In the PNES+E group, six patients had a history of seizure-related injury, seven of urinary incontinence and eight of nocturnal seizure.
Cluster analysis
A total of 145 seizures were analysed from the 52 patients. The frequencies of individual symptoms in the whole series are presented in table 1. Some clinical items were frequent in all the categories but did not constitute discriminatory elements to classify attacks: side-to-side head shaking, moans and tears, normal postictal state and closed eyes.
Frequency of individual symptoms in the whole series
Other symptoms were most frequent and discriminatory of one or two categories. Their distribution in each class is detailed in table 2.
Main semiological characteristics of the five clinical subtypes of psychogenic non-epileptic seizures
Statistical analyses generated seven categories of attacks according to their main semiological features (figure 1). Two patients were finally excluded because of the high number of attacks they had (patient 43 had seven attacks, and patient 47 had 12) leading to a distinct category for each of these patients (class 6 and 7). These two categories were semiologically close to class 2, but we preferred to exclude both patients from our final analysis to avoid any bias in the description of the class 2 characteristics.
Graphical representation of psychogenic non-epileptic seizures according to axes 1 and 2. The axes 1 and 2 each represent specific items. Axis 1 for: duration<1 min, duration >5 min, axial extension, axial immobility, side-to-side head shaking, face asymmetry, focal signs (one limb or head), sudden onset or end, myoclonus, hyperextension of the upper limb, dystonia in the upper limb, myoclonus of the upper limb, fluctuation of the symptoms, vocalisation, immobility of the axis, flexion or extension of the axis, vegetative signs, variation of symptom intensity and postictal state. Axis 2 for: focal signs (halfbody), focal signs (head or limb), upper-limb myoclonus, closed eyes, vegetative signs, responsiveness, one-sided rotation of the head, upper-limb tremor and hyperkinesia. According to the factorial map of the variables, seven classes appear. Each attack is represented by a number and a colour (colour according to their membership of a category). The colour code was: ‘black’ for class 1, ‘green’ for class 2, ‘blue’ for class 3, ‘purple’ for class 4, ‘red’ for class 5, ‘orange’ for class 6, and ‘cyan’ for class 7 (see table 1).
The average number of attacks per patient was not significantly different between the five retained classes (ANOVA, p=0.99).
We named these five classes according to their main clinical features (table 2):
Dystonic attacks with primitive gestural activity (class 1, 31.6%). This was typically an attack lasting less than 5 min with tonic movements of the four limbs and primitive gestural activity (hiding the face, grasping, punching; 48.7% of the seizures) and an altered responsiveness. It could also include fine tremor (30.8%), vocalisation or wailing (33.3%).
Pauci-kinetic attacks with preserved responsiveness (class 2, 23.4%). These were seizures of variable duration with sudden onset and end and preserved responsiveness in most cases. They could be announced by variable and non-specific symptoms. There was no movement of the trunk. Movements were limited to a fine tremor usually in a limb or the head. Wailing was observed in 37.9% of the seizures.
Pseudosyncopes (class 3, 16.9%). This term was previously coined by Benbadis.18 This class was characterised by short and sudden loss of responsiveness without any vocalisation or movement of the trunk. Abnormal movements were limited to bilateral myoclonus (57%) or bilateral tremor (42%).
Hyperkinetic prolonged attacks (class 4, 11.7%) were typically characterised by a prolonged duration (>5 min), a progressive onset and end, and preceded by warning symptoms. They also comprised hyperventilation, partial loss of responsiveness and abnormal movements of fluctuating intensity and various type, involving the limbs and head but sparing the trunk and consisting mainly of tremor and tonic posturing.
Axial dystonic prolonged attacks (class 5, 16.4%) were typically characterised by prolonged, violent and tonic axial manifestations. Opisthotonos was specific of this class. Head and limbs were involved by various abnormal movements often with tonic posturing and less frequently tremor, myoclonus and primitive gestural activity. Vocalisation and hyperventilation were also frequent.
Among the 39 patients with several recorded non-epileptic attacks, 24 (61.5%) had only one type of attack, and the 15 others had two to four different types. The most frequent associations were: classes 1–2–3 in three patients and classes 1–3 in three other patients.
Discussion
Our study has identified five classes of PNES leading to an objective classification based on the statistical analysis of conjunctions of clinical symptoms:
dystonic attacks with primitive gestural activity;
pauci-kinetic attacks with preserved responsiveness;
pseudosyncopes;
hyperkinetic prolonged attacks;
axial dystonic prolonged attacks.
Among the patients with several recorded attacks, 61.5% had only one type of attack, suggesting that most of the patients had a reproducible PNES semiology.
The demographic characteristics of our study were consistent with those of previous studies in terms of sex ratio (73% were women) and age at onset (second and third decade).2 4 7 Comorbidity with epilepsy was high, in the upper range of previous studies (5–50%) and reflected the PNES population of a French regional epilepsy centre.3 19 20 In our series, the time to diagnosis was slightly shorter than in previous studies of series from other western countries (4.9 vs 7–16 years).21 A shorter time to diagnosis was previously reported in a recent Chinese series and can probably be explained by differences in healthcare organisation.14
The high rate of inappropriate antiepileptic treatment in the PNES only patients (67.8%), reflecting misdiagnosis, is consistent with previous reports21 and is related to the semiological similarities with ES.22 The well-known clinical characteristics of psychogenic attacks such as closed eyes (present in 60.9% of the attacks), side-to-side head shaking (20.3%) and ictal pelvic thrusting (15.4%) were inconstantly found as in previous studies.9 23 Signs such as ictal pelvic thrusting, rocking of body, side-to-side head movements or rapid postictal recovery, which are classically considered as characteristic of PNES, can also be observed in frontal-lobe seizures.22 Clinical signs such as irregular, fine, low amplitude tremor (43%) and fluctuating intensity of symptoms (40.6%) have rarely been reported previously but were frequent findings in our series. These observations emphasise the need for a better clinical description based on the conjunction of ictal signs instead of a single sign.
Our results are similar to those of a recent study which defined six types of PNES: non-epileptic auras, rhythmic motor PNES, hypermotor, complex motor, dialeptic and mixed PNES.12 This study was based on 14 items only and consisted of subjective classification as the authors did not use an automatic classification method. It is interesting to note the numerous similarities between the two studies despite methodological differences: the ‘pauci-kinetic with preserved responsiveness’ (class 2) is close to the ‘rhythmic tremor’ type; the ‘pseudo-syncope’ (class 3) is close to the ‘dialeptic PNES,’ the ‘hyperkinetic prolonged attack’ (class 4) to the ‘hypermotor PNES’ and the ‘axial dystonic prolonged attack’ (class 5) to the ‘complex motor PNES.’
Furthermore, our study extends the results of two previous studies using automatic clustering analysis which focused on seven and 12 clinical items.13 14 They consistently found three types of PNES: psychogenic minor motor, psychogenic major motor and unresponsive seizures. We identified two types of attacks (class 4 and 5) that are very similar to the previously described major motor type which comprised hypermotor movements and axial tonic manifestations.12 13 It confirms the additional features of hyperventilation and vocalisation reported by An et al14 and further expands the description with the following features: long duration, gradual onset and end, fluctuation of intensity. We thus created two classes which are differentiated by pre-eminent axial and generalised tonic manifestations (class 5) and by pre-eminent varied movements, preceded by warning symptoms (class 4).
Likewise, the ‘pseudosyncope’ type identified in our series (class 3) is very similar to the ‘atonic seizure’ described by Groppel et al13 and the ‘unresponsive seizure’ described by An et al14 Eye closure and unresponsiveness were reported by An et al, while abrupt onset and end, and the axial immobility were described by Groppel et al We also observed myoclonic jerks in 52% of these attacks.
The ‘pauci-kinetic attack’ type observed in our series (class 2) is very similar to the previously described ‘trembling’ or ‘minor motor seizures.’13 14 Our study further reports a preserved responsiveness, a gradual onset and end, an often focal localisation of tremor and possible associated sensory manifestations.
Finally, we propose a class (‘dystonic attacks with archaic gestural activity’) which represents 31.6% of recorded attacks and is characterised by archaic gestural activity and oro-alimentary movements, dystonic movements and less than 5 min duration. This class is clearly differentiated from the major motor type, since it comprises neither axial tonic manifestations nor hypermotor movements. The reason it has not been previously identified is possibly because variable archaic gestural activity and oro-alimentary movements were not explicitly assessed in the previous studies.13 14
Our study has some limitations. The rate of patients with intricate PNES and epilepsy was in the upper range of previous comparable studies (46% vs 5 to 50%), and reflected the recruitment of a tertiary centre with a potential bias towards difficult cases.
A second limitation is the potential variable number of recorded attacks per patient leading to a possible over-representation of one type of attack. However, we chose this statistical unit to address the issue of intrapatient clinical variability of PNES. Moreover, the average number of attacks per patient was not statistically different between classes. The third possible limitation is the retrospective nature of the study and the inherent risk that some data will be missing as was the case for ‘history of previous trauma.’ However, this was not the main focus of the current study. All the data on the medical history of PNES were systematically assessed through a standardised questionnaire specifically designed for our video-EEG monitoring unit prior to the study. The fourth limitation is the subjectivity involved in categorising clinical manifestations. To limit this subjectivity, all data were reviewed and categorised by consensus between two experienced examiners.
We chose diagnosis by consensus between the two examiners over a double-blinded interpretation by two independent raters because a previous study about inter-rater reliability for classifying PNES showed only a moderate κ of 0.57.24
The clinical classification proposed in this study should help to improve the clinical diagnosis of PNES and may also facilitate differential diagnosis of PNES subtypes against specific epileptic seizures. Indeed, it will allow the physician to focus on the signs that specifically distinguish each subtype of PNES from its closest epileptic or non-epileptic-non-psychogenic counterpart: for example, hyperkinetic prolonged attack (class 4) and axial major prolonged attacks (class 5) have similarities with tonic–clonic generalised seizures. The preceding subjective symptoms (class 4) could be confused with a primary partial onset preceding the secondary generalisation. However, the fluctuating intensity of signs, and the associated hyperventilation frequently observed in both classes could help to differentiate them from primary or secondary generalised tonic–clonic seizure. Our class 3 (pseudosyncope) PNES might be difficult to differentiate from syncope or secondary convulsive syncope. However, in this class, the frequent closed eyes (71%) could be highly discriminatory, whereas this is not the case when considering all types of PNES together.9 25 Our class 1 (dystonic with primitive gestural activity) could be confused with temporal or frontal-lobe epileptic seizures in terms of duration and the pre-eminence of primitive gestural activity or dystonic movements. In this class, a video-EEG recording might be required early in the diagnostic phase, since no clear clinical variable discriminatory from frontal or temporal lobe seizures emerged from our study. Finally, PNES with pure isolated aura may be difficult to differentiate from simple partial seizures with normal ictal EEG. In our study, PNES with auras were always accompanied by other signs. There were two clinical situations:
Pauci-kinetic attacks with preserved responsiveness; auras were mostly bilateral sensory manifestations (tingling). This topography and the presence of a fine and low amplitude tremor which could be clearly differentiated from tonic or clonic manifestations oriented the diagnosis towards PNES.
Hyperkinetic prolonged attacks with an initial aura, the subsequent clinical behaviour which comprised varied movements, hyperventilation, fluctuating intensity of symptoms and prolonged duration clearly oriented towards PNES. If such a sequence had been epileptic, a normal ictal and postictal EEG would have been very unlikely.
To conclude, our study shows that the clinical presentation of PNES is stereotypic and can be objectively classified. Such classification could provide the physician with clinical criteria to diagnose patients with PNES. Interobserver reliability, psychopathophysiological correlates and the diagnostic value of this classification need now to be evaluated prospectively.
Acknowledgments
We are grateful to M Debouverie, M Reuber and R Duncan for their careful review of the manuscript.