Early outcomes and predictors in 260 patients with psychogenic nonepileptic attacks

The literature suggests that outcome in patients with psychogenic nonepileptic attacks (PNEA) is variable, but generally poor.^1–4 Some, though not all, studies suggest that patients with longer delay to diagnosis and poorer psychiatric status do worse.^3–8 In the largest published study, 28.8% of patients were spell-free at 1 to 10 years after diagnosis.⁴ In contrast, some patients have early resolution of spells with no intervention other than communication of the diagnosis.^6,9,10 The proportion in which this happens remains uncertain, and may depend on how the diagnosis is communicated.^11,12

Outcome studies have concentrated primarily on spell frequency, though some provide evidence that social and employment outcome is poor.^3,4 Health care utilization, which is high in this patient group, may change favorably at diagnosis and the available data suggest the possibility of an important effect.¹³ It is not clear to what degree this effect requires an improvement in spell frequency.

To try to resolve these issues, we have carried out a retrospective observational study of 6-and 12- month outcome in 260 consecutive patients with PNEA. We have analyzed attendance at follow-up, changes in spell frequency, use of emergency medical services, employment and benefits status, and have used binary logistic regression analysis to determine predictors of outcome.

METHODS

Between March 1999 and August 2004, 260 consecutive patients seen at our PNEA clinic were included in a retrospective outcome study. The clinic is based in a regional neurosciences unit, serving a population of 2.5 million, and takes referrals for known or suspected PNEA from all sources. A total of 196 patients (75.4%) were female. Mean age at first attendance was 37.8 years (range 13–87, SD 14.2), mean age at onset of PNEA 30.8 years (range 6–70, SD 14.2), and median diagnostic delay 3.8 years (range 0.1–51.2, interquartile range 11.0). (The data for diagnostic delay were positively skewed, so the median is quoted. For comparison with previous studies, the mean diagnostic delay was 7.0 years [SD 8.0].) Antecedent sexual abuse was reported by 86/260 (33.1%), 26/260 (10.0%) had additional epilepsy, and 22 (8.5%) had learning disability (LD).

The diagnosis was confirmed by inpatient or outpatient¹⁴ video EEG (249 patients), or by ambulatory EEG recording (1 patient) of spells confirmed as typical by eyewitnesses. In the remaining 10 patients, events were either directly observed in the clinic, or eyewitness descriptions were typical of PNEA, and antiepileptic drugs successfully withdrawn with appropriate postwithdrawal monitoring.¹⁵ Epilepsy was diagnosed by video EEG recording seizures, or by patient and eyewitness accounts assessed by epilepsy specialists and supported by interictal EEG. Information was acquired using a semi-structured interview with the patient and an eyewitness.

The baseline visit was that at which the patient was given a confirmed diagnosis. The baseline dataset was defined and collected as in our previous publication¹⁶: 1) demographic data; 2) presence of epilepsy; 3) history of PNEA; 4) PNEA clinical semiology; 5) history of sexual and physical 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; 6) history of other traumatic life events; 7) medical and drug history, including 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 negative investigation; 8) 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.

Follow-up at the PNEA clinic at 6 and 12 months postdiagnosis was arranged. Patients were encouraged to attend follow-up, and were offered alternative appointments when they failed to attend or canceled. The following data were recorded for comparison with baseline data, and as outcomes for logistic regression analysis: 1) whether patient attended; 2) PNEA frequency; 3) antiepileptic drugs (AED); 4) emergency room visits and emergency admissions to hospital for PNEA since last follow-up; 5) employment and social security payments.

RESULTS

Our patients were poor attenders, so we retrospectively accepted variance of 2 months either way in follow-up visit date. If the 2 follow-up dates were within 4 months of each other, only data from the second visit were entered. Of 260 patients, 187 patients (71.9%) attended at least 1 follow-up visit, while 105 (40.4%) attended 2.

DISCUSSION

In the present study, early outcome in terms of spell frequency was good in a substantial minority of patients, and the improvement occurred before psychotherapy was given. Use of emergency services dropped markedly immediately after diagnosis, irrespective of whether or not the patient became spell-free. Although there was a small improvement in the number of patients in employment, the number of patients drawing social security payments did not improve.

At the time of the study, waiting times for neuropsychology were up to 1 year. Therefore, our outcomes are unlikely to be related to psychological intervention, as no patient in the study had completed treatment by 1 year after baseline, and only a handful had their initial session. This is in keeping with previous data suggesting that a proportion of patients do well with communication of the diagnosis as the only intervention.^6,9,10 The outcomes shown in table 2 show, however, that not all patients who became spell-free did so immediately after baseline.

Factors determining short-term outcome may be different from those determining long-term outcome, and factors predicting good outcome with no psychological intervention may be different from those predicting good outcome after intervention. This, and methodologic differences, make direct comparison of outcome studies difficult. However, our figure of 61% of patients (including those spell-free) with more than 50% reduction in spell frequency (table 1) may be compared with a previous study³ finding that 57% of patients rated themselves markedly improved. Our spell-free rate of 38% (table 1) can be compared with a previous study² that found 25.4% of patients were spell-free with the caution that their follow-up period was 5 years. Similarly, 4 years after diagnosis, 28.8% of patients in the largest available outcome study⁴ were free of spells. In those of our 105 patients who had 2 follow-up visits, outcomes continued to improve between 6 and 12 months (table 2). This argues against early relapse, but the question of later relapse will have to await longer-term outcome figures from our cohort.

In previous studies of longer-term outcome, the length of time a patient has to be free of spells to be classified as such is either not specified^2,4,6 or is 6 months.^3,8 As our focus was on early outcome, we elected to report spell frequency over a shorter period, enabling us to capture changes in spell frequency occurring later in our 2 6-month follow-up periods.

Our data cast doubt on the hypothesis that patients with a short latency to diagnosis are more likely to have a good early outcome.^3,6 Our patients with long delay to diagnosis were just as likely to have a good early outcome as those diagnosed quickly. Early age at onset and age at presentation have also been associated with good longer-term outcome.⁴ We had a large patient group with a broad range in these variables, so we should have detected even a relatively weak effect.

The phenomenon of aggravation of spells following diagnosis has attracted little attention in the literature. In the present study, 35/187 (18.7%) reported an increase of 50% or more in spell frequency over baseline. Our data did not suggest a reason for this, though patients who had MUS other than PNEA and those with LD were between 3 and 4 times less likely to report it.

Some of our predictors of outcome were expected. Our clinical experience is that patients drawing social security payments do badly. Poor psychiatric status,^3,9,17 poor personality and psychopathology inventory scores,⁴ and poor ability to form relationships¹⁸ have all been associated with poor long-term outcome. It is possible that our crude measure of psychological health, i.e., past diagnosis of anxiety or depression, covaries with these factors. We also hypothesized that patients with antecedent sexual abuse would do worse. This was in fact the case, but only so long as psychiatric status was not entered into the same logistic regression model: this suggests that while patients with antecedent sexual abuse indeed do worse, they do so because of the associated psychiatric status.

We did not expect women to do worse than men. Indeed, one previous study found the contrary.⁸ Men less often report antecedent sexual abuse,^19–21 which itself has been associated with poorer psychiatric status in patients with PNEA.²² However, the predictive effects of gender and psychiatric status were in this study independent of each other, and of sexual abuse.

While bullying predicted good early outcome, the numbers involved were small: 10 out of 12 patients who reported bullying became spell-free. The effect was not mediated by lack of other antecedent trauma, lack of psychiatric history, or any other factor that we recorded. This requires further study.

In the present study, reduction in use of emergency services was striking. A marked reduction is in keeping with one previous smaller (though more detailed) study.¹³ The fact that reduction in use of emergency services occurred in many patients whose spells persisted or even became more frequent argues that outcomes other than spell frequency need to be taken into account when assessing the effect of an intervention. Militating neither for nor against psychological intervention, our data provide compelling argument that resources given to achieving early diagnosis and communicating it are likely to be effective in reducing demand for emergency health care.

Our protocols¹⁵ call for sequential withdrawal of AED with appropriate clinical monitoring in patients who do not have evidence of underlying epilepsy. The process takes time, and in some patients had to be halted for reinvestigation when “new” spells arose. In a few cases, AED were restarted by other physicians. Hence a small number of patients who had no evidence of epilepsy remained on some AED treatment.

The fact that a minority of patients got back to work, while there was no change in social security payments, was expected from published data.^3,4,13 Our own data suggest that return to work is largely dependent on success in stopping spells. Those originally drawing social security payments continued to do so, spell-free or not.

Our dataset was part of the standard comprehensive assessment offered at our clinic, and as such was decided before patients were seen at baseline. However, poor patient attendance data precluded a prospective study, hence our retrospective design. We were able to obtain face-to-face follow-up in the majority of patients, but the fact that we have no data in 28.1% should be borne in mind when interpreting our results. Two of the larger published studies included only patients in whom follow-up was available,^2,6 one obtained follow-up information by telephone in 76% of 72 patients,³ and the largest⁴ obtained follow-up information on 49.8% of 329 patients, also by telephone. In the present study, those patients with follow-up and those without did not differ in respect of any of the variables acquired, with the exception that patients with LD were more likely to attend (presumably because they were often brought by carers).

Our logistic regression models explained a relatively small proportion of the variance in the dataset, so it is likely that predictors of outcome exist other than those we recorded. The present study, like others before it, is not population-based. Although our clinic operates within a neurosciences center that contains all the practicing neurologists within our population, it is highly unlikely that we saw all the PNEA cases in the population during the time of the study. Therefore, while our population is probably large enough such that major sampling error is unlikely for most variables, the possibility (indeed probability) of systematic sampling bias due to referral patterns and other effects remains. The acquisition of population-based samples of patients with PNEA would be required to overcome this.

AUTHOR CONTRIBUTIONS

Statistical analysis was conducted by Dr. R. Duncan.

DISCLOSURE

P. McKenzie, Dr. Oto, and Dr. Russell report no disclosures. Dr. Pelosi has received travel expenses and/or honoraria for lectures or educational activities not funded by industry; served on the international editorial advisory board of the British Medical Journal; receives honoraria from the University of Glasgow and the University of Edinburgh; and is a full-time employee of the National Health Service in Scotland. Dr. Duncan has received speaker honoraria and funding for travel from UCB.