An exploratory randomized controlled trial of immediate versus delayed withdrawal of antiepileptic drugs in patients with psychogenic nonepileptic attacks (PNEAs)


  • Maria Oto,

    1. West of Scotland Regional Epilepsy Service, Institute of Neurological Sciences, Southern General Hospital, Glasgow, Scotland, United Kingdom
    Search for more papers by this author
  • Colin A. Espie,

    1. Department of Psychological Medicine, Faculty of medicine University of Glasgow, Institute of Neurological Sciences, Southern General Hospital, Glasgow, Scotland, United Kingdom
    Search for more papers by this author
  • Roderick Duncan

    1. West of Scotland Regional Epilepsy Service, Institute of Neurological Sciences, Southern General Hospital, Glasgow, Scotland, United Kingdom
    Search for more papers by this author

Address correspondence to Dr Maria Oto, Department of Neurology, Southern General Hospital, Glasgow G51 4TF Scotland, United Kingdom. E-mail:


Purpose: To determine whether withdrawal of antiepileptic drugs (AEDs) in patients with psychogenic nonepileptic attacks (PNEAs) improves outcome.

Methods: Randomized controlled trial of AED withdrawal in patients with PNEAs. Patients were randomized to immediate or delayed (9 months) withdrawal of AEDs. We recorded spell frequency, changes in work status, use of emergency medical services, and psychological status at baseline, 9 months, and 18 months.

Results: Of 193 patients screened, 38 fulfilled entry criteria, 13 declined participation, and 25 were randomized. Fourteen patients were randomized to immediate withdrawal (IW) and 11 patients to delayed withdrawal (DW). There was a significant reduction in spell frequency from baseline to 9 months in the IW group but not in the DW group (p = 0.028). There was a significantly greater reduction in use of rescue medication in the IW group compared to the DW group between baseline and 9 months (p = 0.002). Emergency health care utilization dropped to zero in both groups by the end of the study. Psychological measures reflecting internal locus of control increased significantly more in the IW group (p = 0.005).

Discussion: Stringent diagnostic criteria and an increasing tendency for patients to be referred before AED prescription limited the recruitment and the power of the study. Our data nonetheless provide evidence that some outcomes are improved by AED withdrawal in patients with PNEAs.

Psychogenic nonepileptic attacks (PNEAs) may be defined as paroxysmal events that resemble or may be mistaken for epileptic seizures, are not associated with any measurable alteration in brain electrical activity, and could plausibly be attributed to a psychological cause. PNEAs present to virtually all health care professionals, account for approximately 18% of patients presenting with spells, and represent a significant management problem for epilepsy specialists (Kotsopoulos et al., 2003). A variety of psychological interpretative paradigms have been proposed to explain PNEAs (Gates & Erdahl, 1993; Bowman & Markand, 1996; Krishnamoorthy et al., 2001).

A recent Cochrane review concluded that there was no reliable evidence to support the use of any specific treatment (Brooks et al., 2007). There is, however, increasing consensus that an initial clear and unambiguous communication of the diagnosis of PNEAs and the withdrawal of an erroneous diagnosis of epilepsy is a necessary part of management and may be the only required intervention for some patients (Aboukasm et al., 1998; Hall-Patch et al., 2010).

Up to 80% of patients with PNEAs are exposed to anti-epileptic drugs (AEDs), and up to 40% remain on AEDs after the diagnosis has been established (Benbadis, 1999; Reuber & Elger, 2003; O’Sullivan et al., 2007; Hall-Patch et al., 2010). Studies across a range of physical and psychological disorders have indicated that inconsistency between diagnosis and management undermines compliance and prejudices outcome (Barsky & Borus, 1999). Therefore, it is possible that failure to withdraw AEDs, by undermining the clarity of a communicated diagnosis of “no epilepsy,” may have a negative effect on outcome in PNEAs.

In a previous study we showed that AED withdrawal in the context of PNEAs can be carried out safely (Oto et al., 2005). In the present study we investigate the potential therapeutic effect of planned withdrawal of AEDs following the diagnosis of PNEAs, by carrying out an exploratory randomized, controlled trial comparing outcomes in patients who had immediate withdrawal of AEDs at diagnosis with those who had withdrawal delayed by 9 months.



Unblinded randomized controlled trial comparing immediate versus delayed AED withdrawal, with repeated measures, followed by a replication phase.


The trial was carried out in the Glasgow PNEA clinic in the regional neurosciences center at the Southern General Hospital. Patients were referred from a wide variety of sources throughout the West of Scotland, including primary and secondary care (catchment population 2.4 million).

Patients referred to the clinic between April 2001 and January 2004 were screened, and those with a video-EEG (electroencephalography) confirmed diagnosis of PNEA without coexisting epilepsy taking at least one AED were invited to participate.

Exclusion criteria

History suggestive of past epilepsy, pregnancy, and inability to give informed consent.


The diagnosis of PNEA was communicated during a standardized semistructured interview, at the end of which informed consent was obtained. Patients were randomly allocated to immediate withdrawal (IW) of AEDs, or delayed withdrawal (DW).

Patients who were randomized to IW were withdrawn beginning at the time of the initial visit. Patients who were randomized to DW continued to take their medication for an additional 9 months. In both cases, AEDs were withdrawn sequentially according to our standard protocols (see Oto et al., 2005).

Assessment and data collection took place prior to randomization (baseline), and at 9 months and 18 months postrandomization. The period of 9 months was chosen to accommodate the time required to withdraw patients from variable numbers of AEDs.

Other than AED withdrawal, management was as usual, including referral to the team neuropsychologists.

Outcomes and data collection

The primary outcome measure was change in self-reported spell frequency from baseline to 9 months and at the end of replication phase at 18 months. We also recorded changes in illness health care utilization, work status, physical and psychological health status, and illness attribution from baseline to 9 and 18 months.

Patients completed the Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, 1983) the Side Effects and Life Satisfaction inventory (SEALS; Brown & Tomlinson, 1982), and Illness Perception Questionnaire (IPQ; Weinman et al., 1996).


Patient and doctor blinding were not desired, as we wished to measure the sum of the physical and psychological effects of AED withdrawal. The research assistant who administered rating scales at baseline was blind to participant allocation. Scale scores were made available to the clinician only at the end of the trial.


Patients were allocated to groups using a list of random numbers generated by Excel (Microsoft Corp., Redmond, WA, U.S.A.). Randomization was administered by a secretary with no knowledge of the trial.

Statistical methods

Analyses were carried out using SPSS version 14 (SPSS, Chicago, IL, U.S.A.). Groups were compared using Fisher’s exact test or Wilcoxon paired rank test. Analyses were on an “intention to treat” basis; therefore, the last observation carried forward method was employed to accommodate missing data points.


Sample characteristics at baseline

From April 2001 to January 2004 our service assessed 193 patients with PNEAs of which 38 (19.6%) of 193 fulfilled our entry criteria. We were able to randomize only 25 (12.9%) of 193 patients, as 13 patients (34.2% of those randomizable) declined participation.

Of the 155 80.3%) of 193 patients who did not meet our entry criteria, 79 (50.1%) of 155 were not on AEDs, 37 (23.9%) of /155 were awaiting video-EEG confirmation of the clinical diagnosis, and the remaining 21 (13.5%) of 155 were excluded for other reasons (i.e., pregnancy, past history of epilepsy, lack of capacity).

Comparison between randomized and nonrandomized groups in terms of gender, marital status, employment, antecedent trauma, age, and delay to diagnosis and spell type showed no significant differences.

One patient from the IW group dropped out at 4 months but was included in analysis using the last observation carried forward.


There were no significant differences at baseline between IW and DW groups in terms of demographic data, or any of the variables measured. All subjects were offered up to six sessions of cognitive behavioral therapy–based psychological intervention; both groups received a mean number of four sessions, with two patients in each group declining to see a psychologist.

Changes in spell frequency, health care utilization, employment, and benefits status are shown in Table 1.

Table 1.   Clinical and social outcome by groups from baseline to end of study
 Baseline9 Months18 Months
  1. IW, Immediate withdrawal group; DW, delayed withdrawal group.

  2. *p < 0.05.

Spells/month, median (range)20 (5–720)12 (6–120)2 (0–290)6 (0–100)1 (0–6)4 (0–32)
Attack free0/140/113/14 (21%)3/11 (27%)7/14 (50%)3/11 (27%)
Use of emergency medication6/14 (43%)4/11 (36%)0/14*4/11 (36%)0/140/11
Emergency services use for PNEAs3/14 (21%)5/11 (45%)1/14 (7%)3/11 (27%)0/140/11
Other use of emergency services10/14 (71%)8/11 (73%)4/14 (29%)4/11 (36%)0/140/11
Working2/14 (14%)0/112/14 (14%)0/114/14 (28%)*0/11
Receiving state benefits11/14 (79%)10/11 (91%)11/14 (79%)10/11 (91%)9/14 (64%)8/11 (73%)

The median number of spells per month declined throughout the study in both groups, from 20 (range 5–720) to 2 (0–290) to 1 (0–6) in the IW group, and from 12 (6–120) to 6 (0–100) to 4 (0–32) in the DW group (see Table 1). The reduction in spell frequency from baseline to 9 months was statistically significant (p = 0.028) for the IW group, but not for the DW group (p = 0.415). At the end of the controlled phase (9 months) a similar proportion of both groups achieved remission (IW 3 of 14, DW 3 of 11). By the end of the study, however, twice as many IW patients achieved remission (IW 7 of 14, DW 3 of 11, p = 0.173).

Use of emergency services for PNEA improved by 9 months with only 1 of 14 patients from the IW group compared with 3 of 11 from the delayed group using this resource (p = 0.183). Use of rescue medication dropped over the period of the study phase, and in this case the difference between IW and DW groups at 9 months (0 of 14 vs. 4 of 11) was significant (p = 0.026). Use of emergency health care services for reasons other than PNEA similarly dropped over the period of the study to no contacts by the end of the study in both groups. Few patients (IW 3 of 14, DW 2 of 11) presented with new symptoms arising since the start of the trial at 9 months.

By the end of the study two patients in the IW group got back to work compared with none in the DW group (p = 0.303), and there was a slight reduction in the number of patients claiming disability benefit payments (p = 0.300).

Changes in psychological rating scales are shown in Table 2.

Table 2.   Psychological scales scores: outcomes by groups from baseline to end of study
 ScaleBaseline9 Months18 Months
  1. IW, Immediate withdrawal group; DW, delayed withdrawal group.

  2. *p < 0.05.

 Anxiety mean (SD)10.9 (6.4)9.9 (3.4)8.7 (4.4)8.1 (4.3)10.0 (3.2)9.5 (3.4)
 Depression mean (SD)9.0 (5.5)6.2 (4.0)6.2 (4.8)7.5 (3.5)6.9 (3.5)7.0 (3.8)
 Number of symptoms8.3 (2.2)7.3 (2.5)7.4 (4.3)8.0 (2.4)6.6 (4.2)5.2 (3.5)
 Physical causes14.6 (5.6)16.3 (4.7)14.4 (4.5)14.7 (4.4)14.3 (4.9)14.4 (3.6)
 Psychological causes8.0 (3.4)8.0 (2.6)10.0 (2.4)9.0 (3.4)11.2 (1.8)8.3 (2.8)
 Control cure3.0 (0.4)3.0 (0.2)3.1 (0.7)3.2 (0.6)2.5 (0.8)2.7 (0.3)
 Consequence3.8 (0.7)3.3 (0.8)3.4 (6.5)3.4 (6.0)3.3 (0.7)3.2 (1.0)
 Time line3.1 (0.4)3.0 (0.5)2.7 (0.5)3.0 (0.3)3.4 (0.8)3.2 (0.7)
 Worry, mean (SD)8.0 (2.8)7.9 (4.9)7.7 (2.4)6.3 (3.5)8.0 (2.5)7.6 (2.6)
 Cognition, mean (SD)25.8 (12.6)24.9 (11)24.2 (12.1)31.2 (7.6)26.3 (12)26.8 (12.9)
 Temper, mean (SD)7.0 (3.7)6.0 (3.1)7.2 (3.0)6.4 (3.3)6.6 (3.5)6.2 (3.2)
 Tiredness, mean (SD)7.6 (3.9)9.2 (4.7)7.1 (4.1)*10.6 (3.8)9.2 (3.5)8.2 (4.7)
 Dysphoria, mean (SD)9.5 (5.2)10.9 (3.3)8.7 (4.7)9.4 (4.0)11.3 (3.8)11.4 (3.2)

The HADS mean scores for anxiety were at the lower end of mild (>10) in both groups at baseline and remained at the same level throughout the study. There were no significant between-group differences.

The SEALS questionnaire at the end of the controlled phase (9 months) showed a significantly higher score for tiredness in the DW group (10.6 vs. 7.1, p = 0.041). This difference disappeared at 18 months, and no other significant differences were found.

For the IPQ results, the number of reported symptoms declined in both groups throughout the study. For the domains that reflected an internal locus of control (stress, state of mind, own behavior), an increased score was recorded for both groups throughout the study, and this increase was statistically larger for the IW group by the end of the replication phase, with more patients attributing their spells to their mental state (p = 0.005).


This is the first randomized controlled trial aimed at determining the possible therapeutic effect of a scheduled AED withdrawal at the time of the diagnosis of PNEAs.

We detected no significant between-group differences that would support a therapeutic effect of AED withdrawal on our primary outcome measure, spell frequency. The fact that we achieved a significant drop in spell frequency in the IW group but not in the DW group, hints at an effect, however.

One obvious possible reason for the preceding result was our small sample size. The only available data on which to base a preliminary power calculation came from cohort studies that suggested that up to one third of the patients become attack free regardless of management (Iriarte et al., 2003). A power calculation based on an alpha of <0.05 suggested we would have required 87 patients per group to have 80% probability of detecting a medium-sized effect. We thought it unlikely that we would achieve this, but elected to proceed with an exploratory trial which, in the event, was indeed underpowered. The lack of significant differences between IW and DW groups in terms of our primary outcome variable should be viewed in this context.

Overall, however, outcomes in terms of spells were encouraging, with 50% in the IW group and 27% in the DW being spell free 18 months after diagnosis, with a nonsignificant difference in favor of the IW group. This compares favorably with series in the literature, including one published trial of cognitive behavioral therapy (Ettinger et al., 1999; Reuber et al., 2003a; Goldstein et al., 2004). Our patients received a great deal of attention during the trial (see below), which may have improved the overall outcome, particularly as our patient population often feels rejected and misunderstood by physicians (Mokleby et al., 2002).

Our secondary outcomes provide clearer indication of a positive effect of AED withdrawal. The use of emergency services for PNEAs dropped to zero in IW and DW groups, as did the use of rescue medication, but they did so earlier in the IW group, and in the case of rescue medication the difference between IW and DW groups was significant. This is consistent with our previous study, which showed a drop in use of emergency services at diagnosis that was largely independent of the result in terms of continuing to have spells (McKenzie et al., 2009), and with the study of Martin et al. (1998) who found that video-EEG monitoring and the subsequent delivery of diagnosis substantially reduced medical costs by 6 months and reduced the use of emergency rooms.

Reduction in demand for emergency health care suggests a change in attitude toward the spells themselves, possibly on the part of relatives and carers, if not patients. Health care contacts may act to perpetuate psychogenic disorders (Page & Wessely, 2003), and a reduction may, therefore, be beneficial in the long term as well as reducing the impact of the disorder in the short term. These issues require further study.

There exists a concern that patients who cease having PNEAs will go on to develop new psychogenic symptoms. Our data are reassuring in this regard, with only a small minority of patients developing new symptoms during the study, all of which resolved by study completion. This was also reflected in the IPQ, with a reduction of the number of symptoms reported throughout the study. It was also encouraging that use of emergency services for reasons other than PNEAs dropped, suggesting that patients did not simply replace PNEA with other emergency symptomatology.

Occupational status is poor in patients with PNEAs, and existing evidence suggests that it remains so even in patients whose spells resolve (Reuber & Elger, 2003). A small minority of patients may cease to be dependent (McKenzie et al., 2009), and our relatively short-term data are consistent with this.

All of our patients were on AEDs known to have mood-stabilizing effects (Ettinger & Argoff, 2007). One possible outcome of withdrawal would, therefore, be exacerbation of anxiety and depression. The HADS scores, however, remained stable throughout the study, with no differences between groups. As might be expected, SEALS scores did suggest that the IW group was significantly less tired than the DW group at 9 months, but provided no other evidence of a benefit in “well-being” as a result of AED withdrawal.

One of the theories behind the study was that removing medication would enhance the message that PNEAs are not a result of a neurologic condition and change patients’ illness attribution. Consistent with this, patients’ perceptions that their attacks were due to stress or mental state increased steadily in the IW group, becoming significantly higher at the end of the study reflecting a greater internal locus of control. For the rest of the IPQ results it is difficult to draw conclusions because of the potential for multiple testing effects. In any event, we found no evidence of deterioration.

Trials in this area are known to be difficult (LaFrance et al., 2007), and this trial posed a number of practical problems. Our final recruitment was smaller than we anticipated. This was mainly because of the increasing (and otherwise welcome) tendency for neurologists in the region to refer before prescribing AEDs, but partly because of our stringent diagnostic criteria (both for PNEAs and for excluding coexisting epilepsy). Because of difficulties recording spells, poor patient motivation, and other factors, the diagnostic process may be prolonged in some patients with PNEAs, and we underestimated the effect of this on recruitment. One third of patients declined participation.

The ethical implications of leaving subjects with PNEA on AEDs for 9 months was an issue discussed extensively by the research team and the research ethics committee. We considered that, although it is intuitive that patients with PNEA (only) should not be prescribed AEDs, in practice a large proportion of patients are left on medication (Ettinger et al., 1999; Reuber et al., 2003; O’Sullivan et al., 2007, Hall-Patch et al., 2010), to the point that it could be said to constitute “treatment as usual” for many physicians. Furthermore, there is no published evidence that withdrawing AEDs improves outcome, or that maintaining them worsens it. The trial was approved by our research ethics committee on this basis. Ethical constraints may also have had a negative impact on the results of the trial. In our daily practice, we positively emphasize the importance of withdrawing AEDs as a crucial step to a complete recovery. In the context of informed consent for the trial, it was not possible to be so one-sided, thus potentially diluting the psychological impact of withdrawal.

Our drop-out rate was low, but the numbers conceal the great effort it took to achieve this. Patients with PNEAs are poor attenders, and repeated reminders and much flexibility on the part of the research fellow were required to ensure that data were acquired.

Anticipating some of the above difficulties, we chose to carry out a pilot study. To expose the potential difficulties of a definitive exercise, we elected to stick as rigorously as possible to standard trial methodology. Unsurprisingly, we concluded that a definitive trial would have to sacrifice rigor for power and practicality in order to be successful.

This trial has not definitively answered the question of whether AED withdrawal has in itself a therapeutic effect on PNEAs. However, as an exploratory exercise, the results have provided a suggestion that there may be a positive effect, and have provided clearer evidence of a positive effect on use of emergency services and rescue medication, as well as on locus of control, as reflected by a shift in illness attribution. As such, our results may encourage the practice of AED withdrawal in appropriate patients.

Finally, and as a by-product of this work, we hope to inform other researches considering a trial in this area on the challenges likely to be encountered in the PNEA population.


We confirm that we have read the Journal’s position on issues involved ion ethical publication and affirm that this report is consistent with those guidelines. Dr. Oto has received support from UCB Pharma for educational meetings. Professor Espie has no disclosures to make. Dr. Duncan has received support from UCB Pharma and Eisai for educational meetings.