Predictors of antecedent factors in psychogenic nonepileptic attacks

Psychogenic nonepileptic attacks (PNEA) are associated with sexual and physical abuse.^1–3 The association recurs through a range of observational studies but it remains uncertain whether or to what degree such antecedent trauma is causal.⁴ Antecedent psychological factors other than abuse may be involved in some patient groups,^5–7 as may nontraumatic factors such as learning disability (LD)⁸ and epilepsy.⁹

Bivariate analysis of observational data has detected associations between specific antecedent factors and some clinical variables, suggesting relationships that might inform further research into mechanisms, and indicating the possibility of clinically relevant subgroups related to LD, health-related psychological trauma, and sexual abuse.^5,8,10 Bivariate analysis is limited in respect of the number of analyses that can practically be carried out in a single dataset, in providing no indication of the strength of a relationship (only its significance), and in being unable to assess whether an association is direct, or is due to covariance with another variable. Multivariate analysis can address those issues, so we have used binary logistic regression analysis to assess relationships between commonly reported antecedent factors for PNEA and a range of clinical variables, as well as relationships among the antecedent factors themselves, in a large cohort of patients with confirmed PNEA.

METHODS

Between March 1999 and August 2004, 288 patients (age range 13–69 years) were seen at our PNEA clinic (West of Scotland regional epilepsy service, catchment population 2.7 million). The clinical diagnosis of PNEA was confirmed by inpatient or outpatient¹¹ video EEG (249 patients), or by ambulatory EEG recording (1 patient) of events confirmed as typical by eyewitnesses. In the remaining 38 patients, events were either directly observed in the clinic by one of our staff, or eyewitness descriptions were typical of PNEA, and antiepileptic drugs successfully withdrawn with appropriate post-withdrawal monitoring.¹² Epilepsy was diagnosed on the basis of video EEG recorded seizures or compatible patient and eyewitness accounts assessed by epilepsy specialists and supported by interictal EEG data. Throughout the period of the study information was acquired using a semistructured interview with the patient and an eyewitness.

Bivariate analysis of the present cohort with respect to LD has been previously reported,⁸ as has bivariate analysis of subsets of the present population with respect to age at onset of PNEA,⁵ and sexual abuse.¹⁰

The information extracted from the database for the present study included the following: demographic data; presence of concomitant epilepsy; history of and path to PNEA diagnosis; history of sexual abuse, pragmatically defined as any reported compelled act of a sexual nature or any witnessed sexual activity perceived as abusive by the patient and occurring at any stage in life preceding the onset of PNEA. In a number of patients this history was elicited from the patient or carers at a time subsequent to that first interview, and has been included in the data; history of other past life events that the patient or carers regarded as traumatic, including traumatic events related to medically explained health problems⁵; history of physical abuse, which was also pragmatically defined according to the patient’s account and perception of the nature of the experience; medically unexplained symptoms (MUS) other than PNEA, determined from patient records, pragmatically defined as any symptoms sufficiently problematic to have resulted in a referral to secondary care, with subsequent negative investigation; previous contact with secondary mental health services, past or present psychiatric diagnoses, and deliberate self-harm (including suicide attempt) as reported by the patient or carers, or documented in case records; and PNEA clinical semiology, situational associations, and triggers.

RESULTS

Of our 288 patients, 217 (75.3%) were women. Mean age at onset of PNEA was 30.6 ± 14.4 years, mean age at presentation was 37.7 ± 14.2 years, and median diagnostic delay was 3.5 (1.64–5.49) (median and interquartile range are quoted as the data distribution differed significantly from normal due to positive skew; for comparison purposes, the mean diagnostic delay was 7.50 ± 8.32 years).

DISCUSSION

This PNEA population in this study is similar to those previously published in that 24.3% were men, 11.1% had concomitant epilepsy, and 30.5% reported antecedent sexual abuse.^13–22

While our data provide an estimate of the relative proportion of patients with PNEA who report one or more of a group of traumatic antecedent factors, we are unable to say how this relates to a non-PNEA population. Our data do not therefore impact on the diagnosis itself, which rests on video EEG monitoring in the majority of cases. The acquisition of appropriate control data for some variables, particularly sexual abuse, is problematic, and we did not attempt it. Neither did we use a formal definition of psychological trauma. DSM-IV criteria for trauma²³ were specifically intended to aid the diagnosis of post-traumatic stress disorder. This was not our primary interest, so we based our data on what the patient and (particularly in the case of patients with LD) relatives reported as traumatic. Similarly, we took the approach to MUS that is common in recent publications,²⁴ in that our definition relied on symptoms, avoiding the use of diagnostic labels that might be regarded as controversial or of uncertain nature.

The question of the accuracy of reporting of traumatic issues in general and sexual abuse in particular remains controversial.⁵ We took the pragmatic step of basing our analysis on what the patient reported, and our data should be seen in this light. To minimize under-reporting, our time window for reporting traumatic factors was not restricted: antecedent traumatic events disclosed during treatment or subsequently have been included in our data (hence the small increases in numbers of patient with sexual and physical abuse in the present study compared to that reported in our previous study of the same population).⁸

Three quarters of our patients reported at least one event or factor in their life that they regarded as traumatic, the citation of any traumatic factor being predicted by later onset of PNEA. No single type of trauma explained this finding. Previous studies have cited sexual and physical abuse as antecedent traumatic factors, with some studies identifying health-related trauma, bullying, and bereavement, and others.^1–3,5–7 The previously known association between sexual and physical abuse aside,^1,10 there were no significant positive relationships between traumatic factors. In keeping with our previous study,⁵ health-related trauma and sexual abuse appeared to involve different patient subgroups, the one being a predictor of the absence of the other.

Our previous study identified significant differences in a range of simple indicators of psychological health between patients reporting sexual abuse and those not,¹⁰ as did exploratory bivariate analysis of the present dataset. However, the present logistic regression analysis showed that most of these indicators, anxiety and depression, past contact with psychiatric services, and a documented diagnosis of personality disorder were not independent predictors of sexual abuse, but that medically unexplained symptoms other than PNEA and self-harm were. A patient from our population with both factors was just over five times more likely to report sexual abuse than a patient with neither. The fact that this is consistent with data from populations of adult survivors of sexual abuse, in which both self-harm and MUS are more common than in control populations,^25–28 suggests that the predictive value we found of self-harm and MUS for sexual abuse may well be more widely generalizable. It might therefore have some clinical utility, particularly in the early stages of assessment of patients with possible or confirmed PNEA.

Despite the known association between sexual and physical abuse, it was striking that their predictors were quite different. This suggests that physical abuse may have a specific role in some patients, rather than just being something that happens to accompany sexual abuse. Our study did lump together physical abuse in childhood with that in marital and other contexts: separation of different histories of physical abuse in future studies might provide more useful information.

A history of bullying was significantly predicted by early age at onset of NES, consistent with the fact that in all but two cases the bullying had occurred at school. Our finding that a history of presentation to emergency departments predicted the absence of a history of bullying was unexpected: however, in schools in Scotland, it is common to contact parents when a child has a spell, rather than contacting an ambulance immediately. The delay may have been enough to allow recovery in a proportion of patients, avoiding a call to ambulance services.

Bereavement has previously been recorded as an antecedent factor, for PNEA and other functional symptoms.^6,7 It is of course a usual part of life: we recorded it as an antecedent factor only if the patient cited it in response to an open question about traumatic life events preceding the onset of PNEA. In some cases, patients gave no specific information that would distinguish the bereavement from an ordinary life experience. In others, subsequent psychological assessment did reveal harrowing circumstances and suggested long-lasting effects. Whether the spectrum of trauma-related bereavement in our patients differed from that which might be found in an appropriate control sample remains open to question, the more so as bereavement was not predicted by any other factor in our sample. This may also apply to the 8% of our patients who had PNEA beginning immediately or shortly after a traumatic assault or accident. Head injury has been recorded as a common antecedent factor for PNEA,²⁹ but was not recorded separately in the present study.

Our previous study⁵ found a high prevalence of health-related trauma in late onset patients. Consistent with this, multivariate analysis found that late age at onset was a strong predictor of health-related trauma: this is consistent with the hypothesis that in our previous study late age at onset acted to a degree as a surrogate to select patients with health-related trauma. There is evidence that adolescents and children with PNEA have differing ranges of antecedent factors.⁷ Therefore, our present relationships between bullying and health-related trauma and age at onset of PNEA may simply reflect population differences in prevalence of different types of traumatic event at different times of life.

As expected, our findings relating to LD are consistent with previous bivariate analysis of the same population,⁸ in finding that epilepsy, triggers for attacks, and previous episodes of pseudostatus predicted LD. Male gender survived in the model as a significant independent predictor of LD while low prevalence of sexual abuse was eliminated, indicating that the increased proportion of males found in this subpopulation is not simply an indirect effect of a reduced prevalence of sexual abuse.

The predictive factors we found, or in some cases clusters of predictive factors (see table 2), add to evidence for heterogeneity of PNEA populations, and appear to relate at least some of that heterogeneity to common antecedent factors. Patients with sexual abuse were best identified by a combination of female gender, physical abuse, MUS, and a history of self-harm. Patients with LD, in contrast, were best identified by a combination of male gender, epilepsy, spells that were triggered by social circumstance, and a history of pseudostatus epilepticus. The impression that these two clusters of predictors identify different subpopulations is reinforced by the fact that that sexual abuse predicted the absence of LD (albeit there was a caveat regarding reporting of sexual abuse in patients with LD; see above). Similarly, patients with health-related trauma were predicted mainly by increased age, and by the absence of sexual abuse, suggesting the existence of a further subpopulation.

The predictive power of our models was limited by small numbers in some cases, while in others a relatively small proportion of variance was explained. Furthermore, an observational study such as the present one cannot determine whether relationships are causal. Nonetheless, some of the predictors we identified suggest relationships that may be worth further research. It would be particularly useful to be able to distinguish effects that are specific to PNEA from population effects of sexual abuse and other common antecedent factors. This would require appropriate control populations, a challenging proposition in some cases. Finally, it would be useful to know to what extent our findings are specific to our sociodemographic and health care environments.

ACKNOWLEDGMENT

The authors thank Dr. David Young, Department of Statistics and Modeling Science, Strathclyde University, Glasgow, Scotland, and Dr. Jason Ellis, Department of Psychological Medicine, University of Glasgow, Scotland, for advice.