Measuring Outcome in Psychogenic Nonepileptic Seizures: How Relevant Is Seizure Remission?

Authors


Address correspondence and reprint requests to Dr. M. Reuber at Academic Neurology Unit, University of Sheffield, Royal Hallamshire Hospital, Glossop Road, Sheffield, S10 2JF, U.K. E-mail: mreuber@doctors.org.uk

Abstract

Summary: Purpose: To examine whether seizure remission is a comprehensive marker of outcome in psychogenic nonepileptic seizures (PNESs).

Methods: A postal questionnaire was returned by 147 patients with PNESs a mean of 4.2 years after diagnosis (mean age at follow-up, 38.1 years). The proportion of patients who were “unproductive” (receiving health-related state benefits) at follow-up was determined, with a comparison of markers of ongoing psychopathology (Global Severity Index, anxiety and depression scores of the Symptom Checklist 90, Somatization Index DSM of the SOMS-2) in three outcome groups: group 1, continuing seizures; group 2, seizures stopped but patients “unproductive;” and group 3, seizures stopped, patients “productive.”

Results: Of the patients, 71.4% continued to have seizures, and 28.6% had achieved seizure remission; 60.0% of patients with continuing seizures and 42.7% of patients in remission were “unproductive” (difference, NS). Ongoing psychopathology was related to the factor “group membership” with higher values in groups 2 and 3 than in group 1 (GSI, p < 0.001; anxiety, p = 0.01; depression, p = 0.02; Somatization Index DSM, p < 0.001). Across all patients and in the subgroup with PNESs and additional epilepsy, differences were significant only between groups 2 and 3, not between groups 1 and 2. In patients with PNESs alone, differences were significant only between groups 1 and 2.

Conclusions: Seizure remission is not a comprehensive measure of good medical or psychosocial outcome in PNESs. Nearly half the patients who become seizure free remain unproductive. Many of these patients continue to report symptoms of psychopathology. Seizure control should not be the only focus of treatment in clinical practice or outcome observation in research studies.

Psychogenic nonepileptic seizures (PNESs) can be defined as a form of paroxysmal behavior that mimics epileptic seizures and is characterized by a sudden and time-limited disturbance of motor, sensory, autonomic, cognitive, and emotional functions, or a combination of these. PNESs are usually thought to represent an automatic expression of psychological distress rather than a wilfully simulated event, but establishing underlying intentionality is difficult (1–3).

Since the introduction of video-EEG into routine clinical practice, it has become clear that PNESs are more common than expected. Five percent of patients who carried a diagnosis of epilepsy in primary care (4), 10% referred to an epilepsy clinic with seizures refractory to antiepileptic drugs (AEDs) (5) and 30% of patients referred to an American epilepsy center for prolonged video-EEG monitoring had PNESs rather than epilepsy (6). What is more, patients with PNESs have a greater tendency than do those with epilepsy to develop seizure status, causing emergency admissions to hospital (7). One fourth of patients with PNESs seen at an epilepsy centre had been taken to an intensive care unit to control their seizures (8). This means that PNES patients are at high risk of iatrogenic harm or even death from inappropriate or inadvertent medication administration (9–12).

PNESs are costly to patients and to society at large. The mean proportion of patients reporting complete cessation of seizures in a range of outcome studies was 31%, and only 43% of patients were living independently at the time of follow-up (Table 1) (13–25). One large study showed that 69% of patients with PNESs were working at the time of initial manifestation of seizures, but only 20% were still working at the time the diagnosis was made (26). Another study revealed that the proportion of patients receiving disability benefits increased from 33% at the time of diagnosis to 60% after a mean follow-up interval of 17 months (15). The cost of the misdiagnosis of PNESs as epilepsy has been estimated as $100,000 per patient (26). The indirect costs through loss of employment of patients and carers are likely to be much greater than the direct costs of inappropriately prescribed AEDs or professional time (27).

Table 1. Overview of studies examining outcome of PNES in adults
StudyStudy groupOutcome

Number
Mean follow-up
after diagnosis (mo)

Concurrent epilepsy*

% PNES-free
% living
independently
  1. *+, Patients with concurrent epilepsy included; −, patients with concurrent epilepsy excluded; NR, not recorded.

Krumholz 1983 (25) 3460+29NR
Lempert et al. 1990 (24) 5024+3428
Meierkord 1991 (23) 7060+4031
Kristensen 1992 (22) 2870+4545
Betts 1992 (21)12124+31NR
McDade 1992 (20) 1312+33NR
Walczak 1995 (19) 5116+35NR
Krahn 1997 (27) 71172455
Ettinger 1999 (18) 56185232
Kanner et al. 1999 (17) 4514+29NR
Quigg et al. 2002 (15) 3017+3340
Carton et al. 2003 (14) 85302870
Reuber et al. 2003 (13)16449+2944
Mean, M; Total, TT, 818M, 31.6N/AM, 31M, 43

The realization that PNESs are common, serious, and costly has stimulated increasing interest in effective treatments. Randomized controlled studies of different psychotherapeutic interventions are finally under way, and the results of a National Institutes of Health (NIH)-supported study of antidepressant therapy are awaited (28,29). Yet it remains unclear how the outcome of treatments for PNESs should be measured. Although some of the most recent treatment studies have included a range of outcome measures, the existing literature shows that the number of seizures or the proportion of patients becoming seizure free after intervention have been the most popular outcome measures (20,21,28,30–38). However, some evidence indicates that patients can remain disabled by psychosocial problems and dependent on state benefits although they have stopped having seizures. For instance, one study showed that although PNESs had fully remitted in 35% and decreased substantially in a further 41% of patients 16 months after diagnosis, occupational status had either not changed or deteriorated in 80% of patients (19).

In epilepsy, the complete remission of seizures is an important treatment goal for several reasons. These include the reduction of the risk of seizure-related injury and sudden unexpected death (39), reduction of stigma (40), and improvement of socioeconomic status and quality of life (41,42). However, it is not clear whether the achievement of seizure remission is a similarly important indicator of outcome in PNESs. To approach this question, we focused on measures of continuing psychopathology in patients with PNESs whose long-term outcome we described previously (13). In this study, we compare psychopathology in patients whose seizures were continuing with that in those whose seizures had stopped but also examine employment status as a cofactor.

PATIENTS AND METHODS

Patients

The database of the Department of Epileptology, University of Bonn, Germany, was used to identify all patients in whom a diagnosis of PNESs was newly established between April 1991 and April 2001 (n = 329). Patients were included only if the diagnosis was secure (if typical events had been recorded by using video-EEG, EEG, observation, and ictal examination, or by the provocation of a typical seizure by intravenous injection of 0.9% saline). Two hundred ten patients had PNES alone, and 119 had PNESs and concurrent epilepsy (PNES+E). The previous studies listed in Table 1 used different definitions of PNES+E. For the purpose of this study, an additional diagnosis of epilepsy was recorded not only if ictal EEG or video-EEG evidence was available (66 of 119), but also if an experienced epileptologist made this diagnosis based on seizure descriptions, medical history, interictal EEG recordings, imaging, and neuropsychological findings (53 of 119). Outcome was determined by a postal questionnaire that included the psychological self-report measures listed later. Patients were considered “unproductive” if they had retired on ill-health grounds or were unemployed. Patients were not considered “unproductive” if they were employed, self-employed, at school, at university, or if they were older than 60 years and receiving an old-age pension. Educational achievements were graded by using the categories special school, lower-tier (“Hauptschule”), intermediate-tier (“Realschule”), or top-tier secondary school (“Gymnasium”), and university course (“Studium”) and considered an ordinal variable.

Psychopathologic self-report measures

The Symptom Checklist-90 (SCL-90-R) is a 90-item, 2-point, self-report clinical rating scale widely used to measure current psychopathology (43). Apart from a general psychopathology score (global symptom index, GSI), the nine subscale scores include somatization, obsessional compulsion, interpersonal sensitivity, depression, anxiety, anger–hostility, phobic anxiety, paranoid ideation, and psychoticism. The reliability and validity of the German version of the SCL-90 is similar to that of the original version (44). For the purpose of this study, we focused on the GSI depression and anxiety scores.

The Screening for Somatoform Disorders-2 (SOMS) is a 68-item questionnaire measuring all physical symptoms relevant for a diagnosis of somatization disorder according to the DSM-IV and ICD-10, with a 2-point rating scale for each item. Patients are asked to report all symptoms experienced over the last 2 years for which they had consulted a doctor but for which no physical explanation had been found. Our analysis was based on the somatization index DSM-IV (SIDSM) (45).

Statistical analysis

We initially examined a range of demographic measures and the psychopathology scores in the whole patient group (PNES-only and PNES+E). To investigate the possible difference between these two patient groups, measures of psychopathology also were analyzed separately in the PNES-only and PNES+E subgroups. In view of the fact that Levene's test suggested inequality of variance of the biographical measures (age at onset, etc.) and the psychological scores, only nonparametric tests were used in this study. The χ2 or Mann–Whitney U test was used for comparisons between two groups. The significant effect of the factor “patient group” was demonstrated by the Kruskal–Wallis test, before individual groups were compared by using the Mann–Whitney U test (MWU) for post hoc analysis.

RESULTS

Patients

Of 329 patients with PNESs, 164 (49.8%) returned the postal questionnaire (98 PNES-only, 66 PNES+E), 83 (25.2%) could not be traced, six (1.8%) did not want to take part, and 76 (23.1%) failed to return the questionnaire. The response rate of patients who could be contacted was 66.7%. Responders did not differ significantly from nonresponders with regard to age at onset of PNES, age at diagnosis, latency from manifestation to diagnosis, PNES frequency, sex, length of follow-up, or presence of concurrent epilepsy. In those patients with an additional epileptic seizure disorder, no significant relation was found between response and age at onset of epilepsy or epileptic seizure frequency. Seventeen of the responders had to be excluded from this study because their responses were not sufficiently complete, leaving 86 patients with PNES-only and 61 with PNES+E.

Global outcome

The mean age at onset of PNESs was 21.3 years (SD, 14.5); the mean age at the time of diagnosis was 33.9 years (SD, 13.6); and the mean age at the time of the study was 38.1 years (SD, 13.9). Of the 147 patients included in this study, 105 (71.4%) continued to have seizures, and 42 (28.6%) had been free of seizures for ≥1 year. Sixty percent of the patients who continued to have seizures and 42.7% of patients who were seizure free were “unproductive.” The difference in the rate of utilization of state benefits in seizure-free patients and in those still having seizures was not significant. Of the 86 patients with PNES-only, 44.4% of those still having seizures and 26.1% of those who were seizure free took AEDs at the time of study (difference, NS).

Further to evaluate the relevance of seizure status as an indicator of outcome, patients were split into three groups. Group 1 was made up of patients continuing to have PNESs; group 2, of patients who were seizure free but remained “unproductive”; and group 3, of patients who were seizure free and “productive.” Group membership was significantly influenced by patients' educational achievements, their age at diagnosis and at the time of the study, but not by gender, age at onset, or markers of severity of the seizure disorder (see Table 2 for further details). The demographic details of patients in the three groups are not shown separately for patients with PNES-only and PNES+E because significant differences between these two patient categories were limited to the age of manifestation (PNES-only = 25.6 ± 14.2 years; PNES+ E = 15.3 ± 12.8 years; p < 0.001; MWU = 1,262.0) and age at the time of the study (PNES-only = 36.3 ± 15.5 years; PNES+E = 40.6 ± 10.8 years; p = 0.014; MWU = 1,999.0).

Table 2. Demographic and clinical features in the whole patient group (PNES and PNES+E, N = 147)


Feature
Seizures
continue
(n = 105)
PNES stopped,
unproductive
(n = 18)
PNES stopped,
productive
(n = 24)


Difference
  1. aKruskal–Wallis χ2.

  2. bSpecial school, 1; lower-tier secondary school, 2; intermediate-tier secondary school, 3; top-tier secondary school, 4; university education, 5.

Age at onset (yr, mean ± SD)22.7 ± 14.517.1 ± 14.718.3 ± 13.9NS
Time to diagnosis (yr, mean ± SD) 8.3 ± 10.38.3 ± 9.54.0 ± 4.1NS
Age at time of study (yr, mean ± SD)39.4 ± 14.138.8 ± 12.432.1 ± 12.7p = 0.039
(6.509)a
% Female81.072.079.2NS
% With additional epilepsy40.050.041.7NS
Educational achievement (mean category)b2.7 ± 1.02.3 ± 1.03.2 ± 1.1p = 0.020
(7.789)a
% Emergency hospital treatment72.466.762.5NS
% ICU treatment27.627.88.3NS

Psychopathology in different patient groups

The markers of ongoing psychopathology were first examined in all patients (Table 3a), then in the PNES-only (3b) and PNES+E subgroups (3c). In the whole patient group, the markers of continuing psychopathology examined in this study were significantly related to the factor “group membership.” The markers were most highly elevated in patients with continuing PNESs, less elevated in those whose PNESs had stopped but who were unproductive, and lower again in patients who were seizure free and productive. Group differences were most significant with regard to the GSI of the SCL-90 (Kruskal–Wallis χ2= 16.232; p < 0.001) and the Somatization Index DSM (SIDSM) of the SOMS-2 (Kruskal–Wallis χ2= 21.860; p < 0.001).

Table 3. (a) Measures of ongoing psychopathology in whole patient group (PNES only and PNES+E, n = 147)



Measure
Group 1
Seizures
continue
(n = 105)
Group 2
Seizures stopped,
unproductive
(n = 18)
Group 3
Seizures stopped,
productive
(n = 24)

Significance
of factor
“group”

Difference
Groups
1 vs. 2

Difference
Groups
2 vs. 3
  1. aKruskal–Wallis χ2.

  2. bMann–Whitney U.

Global severity index42.6 ± 20.637.8 ± 24.223.1 ± 20.1p < 0.001NSp = 0.044
(GSI, SCL-90) (16.232)a (137.0)b
Depression (SCL-90)6.8 ± 3.86.2 ± 3.63.7 ± 3.5p = 0.02
(12.823) 
NSp = 0.033
(132.5)b
Anxiety (SCL-90)5.3 ± 2.84.3 ± 3.42.8 ± 2.8p = 0.01
(14.745)a
NSNS
Somatization (SI-DSM, SOMS-2)10.1 ± 5.2 7.4 ± 4.34.3 ± 5.0p < 0.001
(21.860)a
NSp = 0.018
(123.5)b
 
(b) Measures of ongoing psychopathology in subgroup with PNES only (n = 86)



Measure
Group 1
PNESs
continue
(n = 63)
Group 2
PNESs stopped,
unproductive
(n = 9)
Group 3
PNESs stopped,
productive
(n = 14)

Significance
of factor
“group”

Difference
Groups
1 vs. 2

Difference
Groups
2 vs. 3
 
Global severity index45.5 ± 21.929.0 ± 26.524.8 ± 21.4p = 0.002p = 0.041NS
(GSI, SCL-90) (12.644)a(163.5)b 
Depression (SCL-90)7.2 ± 3.84.4 ± 3.84.1 ± 3.8p = 0.006
(10.072)a
p = 0.043
(165.0)b
NS
Anxiety (SCL-90)5.5 ± 3.03.0 ± 3.52.6 ± 2.7p = 0.002
(12.714)a
p = 0.025
(152.5)b
NS
Somatization (SI-DSM, SOMS-2)10.6 ± 5.5 7.8 ± 4.85.7 ± 5.1p = 0.014
 (8.596)a
NSNS
 
(c) Measures of ongoing psychopathology subgroup with PNES+E (n = 61)



Measure
Group 1
Seizures
continue
(n = 42)
Group 2
Seizures stopped,
unproductive
(n = 9)
Group 3
Seizures stopped,
productive
(n = 10)

Significance
of factor
“group”

Difference
Groups
1 vs. 2

Difference
Groups
2 vs. 3
 
Global severity index38.3 ± 17.946.3 ± 19.320.7 ± 18.9p = 0.019NSp = 0.020
(GSI, SCL-90) (7.912)a (23.5)b
Depression (SCL-90)6.2 ± 3.78.0 ± 2.43.2 ± 3.0p = 0.014
(8.537)a
NSp = 0.005
(10.5)b
Anxiety (SCL-90)5.1 ± 2.55.6 ± 2.83.0 ± 3.0NSNSNS
Somatization (SI-DSM, SOMS-2)9.4 ± 4.97.0 ± 3.92.2 ± 4.2p = 0.001
(14.024)a
NSp = 0.007
(12.5)b

In the whole patient group, the post hoc comparison of the mean scores of GSI, SIDSM, anxiety and depression subscales of the SCL-90 of patients with continuing seizures, and patients who were seizure free but unproductive revealed no significant differences. However, the comparison of patients who were seizure free and unproductive and those who were seizure free and productive showed significant differences in the GSI, SIDSM, and depression but not anxiety scores (Table 3a).

In the subgroup of patients with PNES+E, the significant interaction of the factor “group membership” with measures of psychopathology also was related to differences in productivity status rather than seizure status (Table 3c).

The post hoc analysis in the subgroup with PNES-only, however, showed that differences in the psychopathology scores were greater between patients who continued to have seizures and those who were seizure free than between unproductive and productive patients whose seizures had stopped (Table 3b).

DISCUSSION

We previously showed that the long-term outcome in patients with PNESs seen in an epilepsy center is poor. At a mean age of 38 years, and just over 4 years from the time of diagnosis, 71% of the patients continued to have seizures, and 51% were dependent on state benefits (13). In this study, we examined whether the achievement of complete cessation of seizures for >1 year was a meaningful, comprehensive marker of outcome.

In line with other outcome studies (see Table 1) (14,15,17–25,27), almost a third of patients with PNESs achieved seizure remission. Yet, even in this group, >40% were in receipt of health-related benefits at the time of follow-up and did not enter (or reenter) full-time employment or education. The observation that no significant difference occurred in the risk of continuing nonproductivity between patients whose seizures had stopped versus those whose seizures had not, does not mean that seizure status is of no importance to patients or benefit agencies. However, this observation does suggest that seizure control is not a major factor in the self-assessment and third-party assessment of disability in this patient group.

The comparison of measures of ongoing psychiatric symptoms, somatization, depression, and anxiety showed that mean scores tended to be lower in patients whose PNESs had stopped. However, across the whole patient group and in the subgroup with PNES+E, scores were significantly lower only in those patients who were seizure free and productive. Little difference was noted between patients whose seizures continued and patients whose attacks had stopped but who remained nonproductive. In the subgroup of patients with PNES-only (Table 3b), the achievement of seizure control may be a more meaningful marker of psychopathology, but even in this subgroup, nine (39.1%) of 23 patients remained unproductive, indicating that seizure control does not necessarily equate with good overall outcome.

Given that a significant interaction was found between age at manifestation and seizure and productivity outcome, it is possible that the differences between the PNES-only and PNES+E subgroups were related to the fact that the age at onset of PNES was lower in the PNES+E group. Conversely, real differences may exist in terms of psychopathology between patients with PNES-only and PNES+E (46), although such differences have not been found by all groups who have examined this issue (47).

The most important finding of this study, however, is that nearly half the patients whose PNESs had stopped remained disabled and continued to have symptoms of psychiatric disorders. This is in keeping with a study of 56 patients with PNESs that found evidence of depression in 52%, suicidal ideation in 39%, and a history of a recent suicide attempt in 20% of patients a mean of 18 months after diagnosis (48). Suicide attempts were as common in patients who continued to have seizures as they were in those whose seizures had stopped.

Evidence from studies in patients with epilepsy suggests that measures of psychopathology are correlated more closely with Quality of Life (QoL) than with physical markers of severity of the seizure disorder (49–52). One would therefore expect the findings of this study to mirror those of a previous investigation examining the relation between seizure frequency or remission and QoL in 30 patients with PNESs a mean of 17 months after diagnosis (15). Although patients who had achieved complete remission of seizures in this study showed greater improvements in QoL than did those with continuing seizures, the study also revealed a very substantial group of patients in whom seizures stopped but life did not improve. Only 50% of those with complete remission experienced a subjective improvement in QoL, and about half of those who achieved full seizure control had significant psychiatric symptoms or an unproductive occupational status. Furthermore, no significant correlation of QoL and seizure frequency overall was noted (15).

Given that the value of seizure frequency or control appears limited, what outcome measures would be more relevant? In view of the heterogeneity of PNES disorders and related interventions, we suggest the following:

  • 1Outcome measures must be chosen with the specific condition in mind. PNESs involve unpredictable, disabling attacks that often lead to dependence on carers and state benefits, which are associated with other psychiatric and medically unexplained physical symptoms and high health care utilization.
  • 2Measures must be suitable for particular patient groups; those suitable for children may not be applicable to men or women.
  • 3Outcome measures of PNESs treatment studies should include psychosocial domains such as self-efficacy, activities, and perceived support.
  • 4Measures should have face validity for the particular intervention: a score measuring change in illness representations may be a good choice for a brief intervention tackling patients' beliefs and assumptions about their condition, whereas such an intervention may not effect changes of social and work adjustment in the short term.
  • 5Both subjective measures and objective measures should be considered because the two reveal different aspects of HRQoL (15,53), and subjective measures cannot be relied on in isolation (54).
  • 6The applicability of outcome measures may depend on the health care system they are applied in. A person funding his or her own health care may be interested in significant subjective improvements, whereas public health purchasers may only be prepared to fund interventions that reduce health care costs. In countries where health care and social budgets are financed by tax income, interventions may be funded only if reductions in direct and indirect health care costs can be demonstrated.

Some of the more recent intervention studies have used a range of suitable measures in addition to seizure counts. For instance, a pilot study of a 12-session cognitive behavioural therapy programme demonstrated significant psychosocial improvements on the Work and Social Adjustment Scale (55), Hospital Anxiety and Depression Scale (56), and positive changes of beliefs and assumptions on the Illness Perceptions Questionnaire (57). Another small but innovative open-ended group psychotherapy program that failed to show appreciable effect on seizure control (31) demonstrated significant benefits in posttraumatic and dissociative symptoms and emotionally based coping by using the Coping Inventory for Stressful Situations (58), the Davidson Trauma Scale (59), the Curios Experiences Survey (60), and the State-Trait Anger Expression Inventory-2 (61). Our own experience suggests that a count of unexplained physical symptoms that have caused patients to consult a doctor over the previous 6 months is a measure that may be closely correlated with seizure severity and longer-term global outcome (45,62).

This study has several limitations. The analysis was based on postal returns of a self-report questionnaire. Because of this, it was not possible to distinguish clearly between patients with ongoing epileptic and nonepileptic seizures in the PNES+E group. Although the overall response rate was satisfactory, and no identifiable differences were seen between those PNES patients who did and those who did not return the questionnaire, it is possible that the patient sample described here is biased in some way. It should be pointed out that elevated scores in self-report measures do not mean that patients are necessarily symptomatic or would pass the threshold of a DSM-based psychiatric diagnosis. It may be relevant that significant differences were found between the three patient groups with regard to education and age, although we did not think that these differences would be relevant to the comparisons made here. Although the results in the whole group would not have been different, the results of the analysis in the PNES+E and PNES-only subgroups clearly depend on the definition of “additional epilepsy.” The PNES+E group would have been smaller and the PNES-only group bigger if we had included only patients with ictal EEG or video-EEG evidence of epilepsy or a specific interictal EEG focus. Finally, we did not have sufficient data to analyse the impact of seizure freedom on health care costs. However, no significant difference was found in the rate of patients still taking (inappropriate) AEDs. Yet the effect may have been much more significant had the more relevant indirect costs been taken into account

CONCLUSIONS

Traditional outcome measures of both epilepsy and PNESs have been based on the frequency of seizures. Although seizure control is important in both disorders, it is not the strongest predictor of QoL (49). In addition, distinct differences appear to exist between patients with epilepsy and PNESs with regard to the relevance of seizure remission. In epilepsy, seizure-free individuals can return to near-normal levels of functioning (63). In patients with PNESs, this does not appear to be the case. In fact, the long-term outcomes in PNESs appear to be worse than those in epilepsy but comparable to those with dissociative and somatization presentations (64,65).

Seizure remission cannot be considered a comprehensive measure of medical or psychosocial outcome in PNESs. Nearly half the patients who become seizure free remain unproductive. Many of these patients continue to have symptoms of psychopathology including other somatoform, depressive, and anxiety disorders. This suggests that seizure counts should be complemented by other outcome measures in trials of medical and nonpharmacologic interventions in PNESs. It also implies that seizure control should not be the only focus of treatment programmes. Further work is required to elucidate differences in psychopathology between patients with PNES-only and those with additional epilepsy. We also must discover not only how to improve seizure control but also how to improve QoL in patients with PNESs.

Acknowledgments

Acknowledgment:  This study was supported by the St James's Trust for Nervous System Diseases, Special Trustees of the Leeds General Infirmary, Leeds, England, and the Verein für die Förderung der Epilepsieforschung e.V., Bonn, Germany.

Ancillary