Article Text

Download PDFPDF

Do patients whose psychogenic non-epileptic seizures resolve, ‘replace’ them with other medically unexplained symptoms? Medically unexplained symptoms arising after a diagnosis of psychogenic non-epileptic seizures
  1. Paul S McKenzie,
  2. Maria Oto,
  3. Christopher D Graham,
  4. Roderick Duncan
  1. West of Scotland Regional Epilepsy Service, Institute of Neurological Sciences, Southern General Hospital, Glasgow, UK
  1. Correspondence to Dr R Duncan, Department of Neurology, Southern General Hospital, Glasgow G51 4TF, UK; r.duncan{at}


Background In clinical practice, it is sometimes observed that patients in whom psychogenic non-epileptic seizures (PNES) cease, develop another medically unexplained symptom (MUS).

Methods In order to determine how many patients develop new MUS post diagnosis and whether patients whose attacks cease are more likely to do so, new MUS were recorded 6–12 months after the diagnosis of PNES in 187 consecutive patients.

Results Compared with baseline, the overall proportion of patients with MUS increased slightly, from 70.1% to 76.5%, with 44/187 patients (23.5%) developing new MUS. There were no significant differences between attack free and non-attack free patients. Binary logistic regression analysis showed that predictors of new MUS diverged between attack free and non-attack free patients. Among patients continuing to have attacks, those with previous health related psychological trauma were 18.00 times more likely to develop new MUS (p<0.0005). In patients who became attack free, patients drawing disability benefits were 5.04 times more likely to have new MUS (p=0.011).

Conclusions The results suggest that almost 25% of patients develop new MUS following a diagnosis of PNES, although most of those have MUS pre-diagnosis. Patients with a history of health related psychological trauma whose attacks continue after diagnosis are at particularly high risk of developing new MUS. The data do not support the hypothesis that PNES that resolve are likely to be ‘replaced’ by other MUS.

Statistics from


Outcome in patients with psychogenic non-epileptic seizures (PNES) is variable but generally thought to be poor.1–4 Psychological intervention may not be successful,5 and in one large study only 28.8% of patients were attack free at 1–10 years after diagnosis.4 In contrast, some patients have early resolution of attacks with no intervention other than communication of the diagnosis.6–10 In one study,10 38% of patients were attack free at the 6–12 month follow-up. Outcomes other than freedom from attacks are less well studied. Healthcare utilisation may change favourably at diagnosis10 11 but there is evidence that social and employment outcome is poor, many patients remaining effectively disabled long term.3 4

In clinical practice, it is sometimes observed that patients in whom PNES resolve then develop another medically unexplained symptom (MUS). Patients with PNES have a high liability to other MUS in any event,12 13 and it may be that this simply continues post diagnosis. It is also possible that when PNES resolve, the putative psychological function served by them (eg, as a way of coping with or expressing distress) is lost, and there is a tendency to acquire new MUS to ‘replace’ that function.

In order to determine whether patients with PNES are at risk of developing new MUS post diagnosis and, if so, whether some patient groups are particularly at risk of doing so, we studied the post diagnosis incidence of new MUS in a cohort of patients with confirmed PNES.


Our cohort is that described in our previous outcome study10 consisting of 260 consecutive patients seen at our PNES clinic between March 1999 and August 2004. The diagnosis was confirmed using inpatient or outpatient14 video EEG (249 patients) or by ambulatory EEG recording (one patient) of attacks confirmed as typical by eyewitnesses. In the remaining 10 patients, events were either directly observed in the clinic or eyewitness descriptions were typical of PNES, and antiepileptic drugs successfully withdrawn with appropriate post withdrawal monitoring.15 Epilepsy was diagnosed by video EEG recording of 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 dataset was that in our previous outcome study of the same cohort.10 The variables acquired were: patient age at presentation and onset of PNES, delay to diagnosis, gender, presence of epilepsy or learning disability, psychiatric diagnoses, previous referral to secondary mental health services, history of self-harm, sexual abuse or other traumatic past experiences, MUS other than PNES, attack frequency and type, employment and benefits status, and whether on anticonvulsant or antidepressant drugs.

Outcomes at 6 or 12 months based on face to face clinical follow-up were available in 187 patients. As in the baseline data,16 we pragmatically defined a MUS as one that was severe or persistent enough to result in referral to secondary care but which had no documented medical explanation. This was based on interview of the patient and examination of the secondary care case records at baseline and at follow-up. We included past and present MUS. For the purposes of the present study, we included symptoms that had been attributed to tension headache, irritable bowel syndrome, chronic fatigue and fibromyalgia. These data were for practical reasons gathered unblinded to seizure outcome. Health related psychological trauma was in the baseline dataset and formed part of our previous study of antecedent factors and their associations in the present population.17 Patients in this category would include, for example, those who had a documented previous life threatening illness or medical experience (eg, myocardial infarction or cancer) that they continued to regard as frightening or traumatic.

The study was carried out with the approval of the Research Ethics Committee of the Southern General Hospital, Glasgow.

Statistical analysis

Statistical analysis was carried out using SPSS v13.

The χ2 test was used to evaluate between groups differences.

Simultaneous binary logistic regression models were used to evaluate the ability of independent variables (those in the baseline dataset—see above) to predict outcomes (in this case, the occurrence or not of new MUS and new medical tests). Exploratory bivariate analysis was carried out for each one, to eliminate baseline variables unlikely to contribute to a significantly predictive model and to prevent baseline variables that covaried from being entered into the same model. Thus independent variables correlating with outcome variables at the 10% level or less (p<0.1) were considered for entry into the model. Where screening for colinearity identified two independent variables correlating at the 30% level (p<0.3) or less, the variable correlating less significantly with the dependent variable was eliminated. The remaining independent variables were entered into an initial model. Independent variables without significant predictive value at the 5% level (p<0.05) were then eliminated, and final analysis carried out. Where the number of predictors exceeded that allowed by the number of cases, the excess predictors were eliminated least significant first.


Of our whole cohort, 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 PNES 30.8 years (range 6–70, SD 14.2) and median diagnostic delay was 3.8 years (range 0.1–51.2, IQR 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). Twenty-six of 260 (10.0%) patients had additional epilepsy and 22 (8.5%) had a learning disability. Of the 187/260 patients (71.9%) who attended at least one follow-up appointment, 131/187 (70.1%) had MUS other than PNES at baseline. At the 6–12 month follow-up, 71/187 patients (38.0%) were attack free.

At the 6–12 month follow-up, the total number of patients with MUS had increased to 143/187 (76.5%) from 131/187 (70.1%). This change was not significant. New MUS were recorded in 44/187 (23.5%) patients (table 1). The proportion of patients who developed new MUS was slightly lower in attack free patients (14/71, 19.7% vs 30/116, 25.9%). This difference was not significant (p=0.336).

Table 1

Psychogenic non-epileptic seizure attacks post diagnosis and the occurrence of new medically unexplained symptoms: attack free patients versus non-attack free patients

Predictors of new MUS

Baseline variables predicting new MUS are shown in table 1.

Where the independent variable is categorical, binary logistic regression reports the factor by which the presence of a given independent variable increases the likelihood of the outcome. Where the independent variable is continuous, the model reports the factor by which each unit increase in the independent variable increases the likelihood of the outcome. This factor is expressed as an OR, and is conventionally quoted with its significance (p) and its 95% CL.

In the whole cohort (n=187), patients with a prior history of health related psychological trauma were 4.34 times (95% CI 1.56 to 12.06) more likely to have new MUS at follow-up (p=0.005). The model for the whole cohort accounted for 6.1% of the variance.

When the cohort was divided into patients who were attack free at follow-up and those who were not, this variable only had predictive value in patients who continued to have attacks. However, its predictive power increased impressively: patients still having attacks were 18.00 times (95% CI 3.62 to 89.61) more likely to develop new MUS if they had a history of health related psychological trauma (p<0.0005) and the binary logistic regression model in these patients explained 20.1% of the variance.

In patients who did become attack free, those who drew disability benefits were 5.04 times (95% CI 1.46 to 17.45) more likely to develop new MUS (p=0.011). The model explained 14.7% of the variance.

Being attack free did not predict the development of new MUS, either in the whole group or in either subgroup. There was no significant predictive effect of follow-up interval (ie, patients whose last follow-up was at 12 months were no more likely to have new MUS than those whose follow-up was at 6 months).


The characteristics of this patient population have been published,10 16 and are similar to other published PNES populations. Our dataset was determined with the intention of detecting MUS arising after a new diagnosis of PNES. It was not possible within the limits of the study to reassess each symptom (and indeed as neurologists we would not necessarily have been competent to do so). Therefore, we used the practical definition detailed in the methods section. We had no follow-up data in 28.1% of patients. However, those with follow-up and those without did not differ in terms of any of the baseline variables acquired, with the exception that patients with a learning disability were more likely to attend. Our sample is not population based. In the context of the present study, it is sufficiently large that major sampling error is unlikely for the variables of interest. However, systematic sampling bias due to referral patterns and other effects cannot be excluded.

Just under a quarter (23.5%) of our patients had new MUS. We do not have data relating specifically to the 6–12 months prior to PNES diagnosis, so we are unable to say whether or not the rate of development of new MUS changed, but given the short time frame of the study and the predominantly young age of the patients, we considered that value to be high. However, the overall increase in the proportion of patients with MUS (70.1% to 76.5%) was modest, as many patients who developed new MUS post diagnosis (37/41, 84.1%) already had MUS pre-diagnosis. Those who continued to have attacks were just as likely to have new MUS as patients who were attack free. Therefore, our data would support the hypothesis that patients with PNES simply have a continuing liability to develop MUS, rather than the hypothesis that they ‘replace’ PNES with other MUS. Indeed, we found a non-significant trend for patients who continued to have PNES to be more likely to develop new MUS than those who became attack free (25.9% vs 19.7%), which might be considered more consistent with a ‘continuing liability to MUS’ hypothesis, and would be in keeping with evidence suggesting that patients with PNES who somatise more do worse.18 Consistent with the lack of between group differences, our binary logistic regression models found that freedom from attacks or otherwise did not predict the development of new MUS. These data do not rule out a ‘replacement’ effect entirely: it remains possible that ‘replacement’ happens beyond the relatively short follow-up period of this study, a question requiring a long term study with the appropriate dataset. It also remains possible that ‘replacement’ operates in a specific subgroup of patients, although our logistic regression analysis failed to detect such an effect, within the limits of the variables we recorded and the numbers of patients reporting each. We did not, for example, record change in severity of existing MUS: in some patients, pre-existing MUS may have become more or less severe post PNES diagnosis.

Intuitively, it seemed to us possible that there would be differences between patients who became attack free early and those who did not. Our data provide some support for this hypothesis, in that we found an intriguing divergence between predictors of new MUS between attack free and non-attack free patients. The effect was unambiguous: not only did the ORs and significance values become more impressive in the split data, but the models explained markedly increased proportions of the variance. The most impressive predictor of new MUS we found was a patient report at diagnosis of previous health related traumatic experiences. This suggests that these patients, in whom health anxiety is important and in whom PNES are often of later onset,16 are highly susceptible to developing MUS. The predictive effect of previous health related traumatic experiences on the development of new MUS was specifically and strongly linked to the continuance of attacks, with an OR of more than 18.00: this provides strong evidence against a ‘replacement’ effect in this particular group.

Drawing disability benefits at baseline was found in our previous study to be associated with reduced likelihood of becoming attack free.10 In the present study, we found that drawing disability benefits predicted the development of new MUS in patients who had become attack free; 61% of patients were drawing disability benefits at follow-up, and this suggests a further aspect of the link with poor outcome, a link that has been found in other MUS.19


The authors thank Teng Cheng Khoo for help in proof reading this paper.



  • Competing interests None.

  • Ethics approval This study was conducted with the approval of the Southern General Hospital Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.