|Home | About | Journals | Submit | Contact Us | Français|
There have been few treatment trials for psychogenic nonepileptic seizures (PNES). Some psychotherapies have been shown to improve PNES and comorbid symptom outcomes. We evaluated a pharmacologic intervention to test the hypothesis that sertraline would reduce PNES.
We conducted a pilot, double-blind, randomized, placebo-controlled trial in an academic medical hospital with epilepsy center outpatients. Subjects aged 18 to 65 years diagnosed with video-EEG–confirmed PNES were treated with flexible-dose sertraline or placebo over 12 weeks. Seizure calendars and symptom scales were charted prospectively. Secondary outcome measures included psychiatric symptom scales and psychosocial variables.
Thirty-eight subjects enrolled, and 26 (68%) completed the trial. Thirty-three subjects with nonzero nonepileptic seizure rates at baseline were included in intent-to-treat analysis of the primary outcome. Subjects assigned to the sertraline arm experienced a 45% reduction in seizure rates from baseline to final visit (p = 0.03) vs an 8% increase in placebo (p = 0.78). Secondary outcome scales revealed no significant between-group differences in change scores from baseline to final visit, after adjustment for differences at baseline.
PNES were reduced in patients treated with a serotonin selective reuptake inhibitor, whereas those treated with placebo slightly increased. This study provides feasibility data for a larger-scale study.
This study provides Class II evidence that flexible-dose sertraline up to a maximum dose of 200 mg is associated with a nonsignificant reduction in PNES rate compared with a placebo control arm (risk ratio 0.51, 95% confidence interval 0.25–1.05, p = 0.29), adjusting for differences at baseline.
Reports of pharmacologic therapy for psychogenic nonepileptic seizures (PNES) were first published at the turn of the 20th century1 and have reappeared in later case reports2; however, no definitive randomized controlled trials (RCTs) have been completed to date.3–6 Some medically unexplained symptoms (presumed psychogenic) have been shown to be responsive to pharmacologic interventions.7–11 In addition to their established efficacy for treating depression and anxiety,12 serotonin selective reuptake inhibitors (SSRIs) have shown promise in trials for conversion or somatoform disorders11,13 and some personality disorders.14 These frequently occurring comorbidities in patients with PNES15 make SSRIs particularly attractive as a potential treatment for patients with PNES.
Acknowledging the heterogeneity in the PNES population and the observed benefit of SSRIs in other somatoform disorders, we proposed a pharmacologic treatment with therapeutic breadth that addresses both comorbidities and PNES directly. Based on the high frequency of Axis I and II serotonergic-mediated symptoms in PNES (i.e., depression,16 anxiety and impulsivity17), we initially hypothesized that treating the comorbidities in patients with PNES would reduce PNES. SSRIs are a reasonable choice to safely treat these conditions. Of the SSRIs, sertraline (Zoloft®; Pfizer, New York, NY) has the broadest US Food and Drug Administration indications and the fewest drug-drug interactions, a concern because many such patients with seizures also use antiepileptic drugs (AEDs). An open-label trial of flexible-dose sertraline in 8 patients with PNES for proof of concept was conducted preceding the current pilot RCT.13
The primary hypothesis of this pilot RCT was to assess the magnitude of seizure frequency reduction by treatment to inform a power analysis for a full-scale RCT. Secondary hypotheses were to identify potential predictors of treatment response.
We received approval from the Rhode Island Hospital (RIH) Institutional Review Board, received written informed consent from all patients participating in the study at enrollment, and provided the ClinicalTrials.gov identifier: NCT00159965.
Patients were referred to the RIH neuropsychiatry/behavioral neurology clinic between July 2002 and June 2008, after being diagnosed with PNES. PNES diagnosis was established by capturing at least 1 of the patient's typical PNES on video-EEG (vEEG). The standard 10–20 electrode system was used and was recorded by cable, 16-channel telemetry, combined EEG and video recording. EKG was monitored. Data were collected in a standard fashion that included interictal samples and all recorded episodes. The combined vEEG recordings were reviewed by a board-certified epileptologist (A.S.B.).
The diagnosis of PNES was defined as stereotypic, motor manifestations (including the initiation or cessation of motor activity/staring), with or without change in level of consciousness, on vEEG with no recognizable buildup of rhythmic epileptiform (ictal) activity immediately before, during, or after the event. Patients, and family members, if present, were given the diagnosis in the RIH comprehensive epilepsy center in a standard format explaining the differences between epilepsy and PNES and their divergent treatments. Patients who were potential study candidates underwent neuropsychiatric examination and clinical screening by a board-certified neurologist and psychiatrist (W.C.L.).
Inclusion criteria were age between 18 and 65 years and vEEG diagnosis of PNES. Patients had to have experienced at least 1 event in the month before enrolling. Patients with only subjective sensory seizures without apparent loss of consciousness or behavioral arrest were excluded. Patients with mixed epileptic seizures (ES) and PNES who could clearly distinguish between their events were included (n = 2). Other exclusion criteria included using monoamine oxidase inhibitors or pimozide within 30 days before study, receiving optimized sertraline currently (≥100 mg daily for 3 weeks), presence of current psychosis, suicidality, or DSM-IV substance dependence diagnosis, inability to complete written surveys, pending litigation, or disability application. Participants currently taking an antidepressant were allowed to enroll, but all medication dosages were held constant during the trial. Participants currently receiving psychotherapy were allowed to enroll; however, those beginning new therapy were excluded.
After enrolling, patients documented their pre-enrollment PNES frequency for 2 weeks before enrollment and rated their psychosocial functioning and symptoms. Baseline measures from the 38 patients in this study were used in a larger cross-sectional study of quality of life (QOL) in PNES.18 As part of the initial examination establishing PNES and the comorbid diagnoses, participants were also administered the Structured Clinical Interview for DSM-IV Axis I Disorders and Structured Interview for DSM-IV Personality Disorders by trained research interviewers. Additional historic and medical data were collected from chart review, patient query, and self-report surveys. Self-report and clinician symptoms scales were prospectively administered biweekly during the visits, with patients reporting symptoms for the 2 prior weeks.
Patients were treated in a double-blind, randomized, placebo-controlled trial. Patients were randomly assigned in blocks of 10, by a computer-generated schedule, in a 1:1 ratio to either the placebo or the sertraline group. Both patient and physician were blinded to treatment group. Allocation was concealed by having the RIH pharmacy generate and maintain the randomization schedule. Pharmacy prepared similar-appearing capsules of 25-, 50-, or 100-mg dosages of the medications. The blind was not broken until after the entire study was completed. Patients were followed up prospectively for 2 weeks without treatment to establish a baseline for measures. Beginning day 15, patients were started on either 25 mg sertraline or 25 mg placebo equivalent. The flexible-dosage design was that sertraline or placebo dose was increased in biweekly intervals to 50 mg, and then by 50-mg increments to a maximum of 200 mg daily, unless increase was limited by side effects. Subjects were seen in six 30-minute, biweekly sessions, according to a pharmacologic trial protocol19 by a clinician with more than 10 years of experience in neurologic and psychiatric pharmacotherapy (W.C.L.). The session consisted of delineating subjects' PNES frequency and side effects, and adjusting medication dose. Missed appointments were made up during the same or following week.
Beginning at enrollment, patients recorded their PNES prospectively using a daily seizure calendar, which was aggregated into biweekly intervals. Collateral information from family informants was encouraged because some patients with ES and PNES may be unaware of their events. Secondary outcome measures are listed in table 1. A trained rater, blinded to treatment group, assessed symptom and psychosocial functioning scales. The Oxford Handicap Scale and Clinician Global Improvement Scale were assessed by a clinician blinded to treatment group.
Data were analyzed using SAS for Windows 9.1.3 (SAS Institute, Inc., Cary, NC). Continuity-adjusted χ2 was used to compare treatment groups on baseline categorical variables. Between-group differences in continuous variables at baseline were evaluated using the Student t test. Seizure counts were modeled using Poisson regression. The Poisson is a distribution on the positive integers appropriate for describing seizure counts; its sole parameter describes both its central tendency and its variance. The Poisson distribution can be approximated by the normal distribution when the event rate or sample size is high; therefore, count data can be modeled adequately using normal linear regression. However, for small sample sizes, Poisson regression is preferred. Mean seizure rate at final as the outcome and mean seizure count at baseline as an offset were used to estimate the within-group change in relative frequency of seizures through the course of the study. The resulting rate ratio was used as a measure of the treatment effect across the 2 study conditions. Patients with zero seizures in the baseline period were removed from this analysis because percentage improvement from baseline to follow-up could not be calculated for such subjects. A scale parameter was also included to adjust standard errors for overdispersion (variance greater than the mean). Analysis of covariance in which change scores from baseline to final visit were compared across study arms, after adjustment for differences at baseline, was used to estimate within-group changes on continuous secondary outcomes. Under an intent-to-treat (ITT) approach, subjects with missing data at the end of the study had their retrospectively reported baseline values carried forward to visit 6.
Ninety of the 128 patients who were assessed for eligibility were excluded. Seventy-three did not meet inclusion criteria, and 17 were eligible but declined participation or were geographically unable to participate and did not enroll. Reasons for exclusions included using an optimized SSRI (n = 29), inability to differentiate events (n = 11), infrequent events (n = 9), no vEEG (n = 8), age (n = 7), and other exclusions (n = 9) (figure).
Sociodemographic data, comorbidities, and clinical factors were similar in both placebo and sertraline groups. Patients in the placebo arm reported slightly higher unemployment, anxiety and Axis II diagnoses, AED use, and a family history of seizures. Patients in the treatment group reported slightly higher frequency of mood diagnoses, trauma history, prior treatment with antidepressants or with psychotherapy, and seizures; however, none of the differences between the 2 groups were significant (table 2). Similarly, baseline measures in the placebo group revealed higher baseline mean self-reported depression scores, impulsive symptoms, and overall symptoms, whereas the treatment group reported higher baseline mean trauma symptoms and dissociative symptoms; none of these differences were significant (table 3).
Of the 38 participants who enrolled, 26 subjects provided end-of-study data (68% study completion rate) (table 3). Sixteen of the 19 who received active drug tolerated a dose of at least 100 mg, with 10 patients tolerating up to the maximum of 200 mg daily. There were no severe adverse events. Thirty-three subjects were included in the ITT analyses for the primary outcome. Subjects excluded from the Poisson regression models consisted of 5 patients with a retrospectively reported biweekly baseline rate of zero seizures, because the outcome of interest (relative change in PNES rates from baseline to end of study) could not be calculated for such subjects.
There was no difference between treatment groups (risk ratio [RR] 0.51, 95% confidence interval [CI] 0.25–1.05, p = 0.29). However, further analyses were conducted using an overdispersed Poisson regression model to estimate relative change in biweekly seizure rates from baseline to study end, separately, by treatment group. These analyses indicated that patients in the sertraline arm manifested a 45% decline in biweekly seizure rates over the 12-week course of the intervention from 22.24 to 12.18 (ratio 0.55, 95% CI 0.32–0.93, p = 0.03). In contrast, control subjects experienced an 8% increase in biweekly seizure rates from 13.38 to 14.38 (ratio 1.08, 95% CI 0.65–1.77, p = 0.78). Using the ratio of these 2 ratios (RR) as a summary measure of treatment effectiveness, our study provides suggestive evidence that pharmacologic treatment reduced seizure rates in the sertraline arm relative to a placebo control arm, adjusting for differences in seizure rates at baseline. Table 4 presents the raw mean and median seizure counts for all visits before exit. Cross-sectional Poisson regressions revealed no between-condition differences significant at α = 0.05 at any of the intermediate time points.
Among subjects with nonzero retrospectively reported baseline seizure rates, 8 of 17 patients in the sertraline arm reported a 50% or greater reduction in the seizure frequency by their final session, compared with 3 of 16 patients in the placebo arm (ITT rates of 47.1% vs 18.8%, p = 0.18), resulting in a number needed to treat of 3.53. Of these responders, 6 patients in the sertraline group reported complete cessation, vs a single patient in the placebo group (ITT rates of 35.3% vs 6.3%, p = 0.08).
Among patients with nonzero baseline rates who provided information at study end, sertraline subjects with Axis II disorders (n = 5) reported higher baseline biweekly seizure rates than those without Axis II diagnoses (n = 9). In the placebo group, there were no significant differences in baseline seizure rates between patients with (n = 11) and without (n = 3) Axis II disorders.
Mean scores on secondary outcome scales assessing depression, anxiety, impulsivity, somatic symptoms, QOL scores, and psychosocial functioning did not reveal between-arm differences in change scores from baseline to final session, after adjustment for differences at baseline (all p >.05) (table 3).
In this pilot RCT, we assessed SSRI treatment to reduce seizure frequency in PNES. The trial was not powered for establishing treatment efficacy; rather, it was conducted to establish an effect size for a pharmacologic intervention and to demonstrate feasibility of conducting a future multicenter RCT for PNES. Given the small, pilot nature of this trial, it is not surprising that no significant differences were found in seizure rates between sertraline and placebo groups. Nonetheless, patients in the sertraline arm manifested a significant 45% decline in biweekly seizure rates vs control subjects, who experienced a nonsignificant 8% increase, suggesting that subjects assigned to the sertraline arm received some benefit relative to placebo. This pilot trial can neither substantiate nor refute the utility of SSRI treatment in patients with PNES.
Analysis of secondary outcome measures including psychiatric symptoms, QOL, family functioning, and psychosocial functioning did not reveal significant differences between the treatment and placebo groups. This finding differs markedly from that in a trial of cognitive behavioral therapy for PNES conducted in parallel with this study that showed not only reduced PNES frequency, but also improved symptoms of depression and anxiety, QOL scales, and family functioning.20 The lack of significant improvement in secondary outcome measures in this pilot pharmacologic trial provides indirect evidence that the pharmacologic trial may not have been “contaminated” with psychotherapy, a potential concern for pharmacologic trials.19
The population enrolled in our trial was reflective of this disorder's complexity, as noted in clinical practice and reported in the literature. Sociodemographic characteristics of the 38 participants who enrolled and completed baseline measures are consistent with current PNES literature. Neurologically, a number of patients with PNES 1) have neuroimaging abnormalities of uncertain relevance, 2) have interictal EEG abnormalities despite no epileptiform abnormalities during their seizure, and 3) have abnormal neurologic examination results despite the absence of a “focal lesion” causing their events. Psychiatrically, patients with PNES are a heterogeneous population having at least 1 comorbid condition, including depression, anxiety, posttraumatic stress disorder (PTSD), or a personality disorder.15 In this study, most of the participants had more than 1 Axis I disorder accompanying their diagnosis of PNES. More than half had a mood, anxiety, or personality disorder. Given that SSRIs are the treatment of choice for the comorbidities, and to maximize generalizability to PNES populations seen in hospitals and clinics, we included patients with anxiety, mood, or personality disorder or a combination of the disorders. If current psychiatric clinical trial exclusion criteria were applied to “real-world” outpatients, the majority of patients seen in practice, up to 90%, would be excluded from RCTs, thus limiting generalizability.21,22 Future studies may benefit from stratifying groups on the presence of personality disorders.
None of the prior PNES treatment studies approached Class I evidence.3,4,23–25 This pilot study misses only 1 Class I criterion (68% enrollment, vs criterion d, at least 80% completion).26 Although results did not attain significance at the customary 5% level, the present study provides Class II evidence that flexible-dose sertraline up to a maximum dose of 200 mg is associated with a nonsignificant reduction in PNES rate compared with a placebo control arm (RR 0.51, 95% CI 0.25–1.05, p = 0.29), adjusting for differences at baseline. The study also provides preliminary evidence of a serotonergic-mediated intervention directly on PNES, because seizure reduction in the sertraline group was not accompanied by a mean reduction in symptoms in common comorbidities of depression or PTSD. Based on this study, we modified our initial hypothesis that treating comorbidities may reduce PNES. SSRIs may have a direct effect on PNES. In fact, other studies have recently reported a direct effect of SSRIs on somatoform disorders, independent of mood and anxiety symptoms.8,27 That the treatment group began showing seizure improvement at lower doses of the SSRI may indicate that somatoform disorders may have a lower serotonergic response threshold than mood and anxiety disorders. This response was not observed in the placebo group, arguing against an early placebo response in the treatment group.
Half of the patients had received an antidepressant before enrollment. Despite using antidepressants at some point in the past, they did not have symptomatic improvement in seizures during their pre-enrollment regimen, suggesting that optimizing the dose of antidepressant may be an important treatment component. Also, patients who do not respond to one SSRI may respond to another. The patients taking AEDs at baseline were prescribed the drug not only for seizures, once thought epilepsy, but also for other AED-responsive conditions, including migraine prophylaxis (n = 4), mood disorder (n = 3), pain (n = 4), and comorbid epilepsy (n = 2). One could argue that ongoing AED use can in some cases reflect lack of confidence in the diagnosis of PNES, and that may influence outcomes. However, the explanation given to the participants was clear that 1) AEDs do not treat PNES, and 2) if they continued on their AED for other indications, it was not being used for seizure reduction in their treatment. The patients understood the indication for their AEDs, and with this clarification, we hoped to mitigate any reduction in confidence in the PNES diagnosis.
The major limitation of this pilot study's conclusions is sample size. A full-scale RCT is needed to establish efficacy for a pharmacologic intervention for PNES. The future full-scale trial will need to adjust for potential dropouts, which largely occurred in this trial because of the patients' concern that they would receive the placebo, despite the equipoise that exists for PNES treatments. Although the evenly dispersed baseline slight differences among the 2 groups could have contributed to the apparent treatment difference, our analyses confirmed that there were no significant differences among any of the factors related to illness severity. One of the major strengths of this study was that all patients had vEEG-documented PNES. However, excluding patients who did not have vEEG may present a potential sampling bias.
The trial provides feasibility and patient tolerability for a pharmacologic intervention for PNES. The potential influence of patient characteristics was also highlighted in this pilot study. Future studies on response durability, documenting treatment effect duration, need to be conducted. A multicenter RCT is being designed to address the efficacy of treatments for PNES.
Statistical analyses were conducted by Drs. George D. Papandonatos and Jason T. Machan.
The authors thank Drs. Orrin Devinsky and Michael Trimble for initial conceptual guidance; Dr. Lawrence Hirsch, who acted as data safety monitor for this trial; and Joan Kelley for database management.
Dr. LaFrance serves on the editorial boards of Epilepsia and Epilepsy & Behavior; receives royalties from the publication of Gates and Rowan's Nonepileptic Seizures, 3rd ed. (Cambridge University Press, 2010); receives research support from the NIH (NINDS 5K23NS45902 [PI]), Rhode Island Hospital, the American Epilepsy Society, the Epilepsy Foundation, and the Siravo Foundation; and has acted a legal expert for Healthcare Litigation Support. Dr. Keitner reports no disclosures. Dr. Papandonatos serves on the editorial boards of the Journal of Consulting and Clinical Psychology and Health Psychology; has served as a statistical consultant for Weinstock & Barylick Associates; receives research from the NIH (R01AG016335 [Biostatistician], P50CA84719 [Biostatistician], R01DA019558 [Biostatistician], R01DA018079 [Biostatistician], R01HL064342 [Biostatistician], R01HL064342 [Biostatistician], R01MH079153 [Biostatistician], R01NR010559 [Biostatistician], R01AA016799 [Biostatistician],R21CA137211 [Biostatistician], R01CA132854 [Biostatistician], and U01CA150387-0 [Biostatistician]); and receives research support from the American Legacy Foundation and Miriam and Rhode Island Hospitals. Dr. Blum serves as Editor of BMC Neurology; receives royalties from the publication of The Clinical Neurophysiology Primer (Humana-Springer, 2007); serves as Medical Supervisor for DigiTrace/SleepMed Inc.; and receives/has received research support from UCB, Eisai Inc., and Abbott. Dr. Machan receives research support from the NIH (5P20 RR024484 [Biostatistician], 5R01CA123544 [Biostatistician], 5U19AI070202 [Biostatistician], NIAMS 1R01-AR056834 [Biostatistician], 1R01 AR056834–01S1 [Biostatistician], and 1R01AA017895-01A2 [Biostatistician]). Dr. Ryan receives royalties from the publication of Evaluating and Treating Families: The McMaster Approach (Routledge, 2005); and has received research support from the Firan Foundation. Dr. Miller receives research support from the NIH (NIMH R34 MH070743-01 [Coinvestigator], NIMH R34 MH078855 [PI], NIMH R01 MH071766 [Coinvestigator], NIAAA R01 AA015950 [PI], NIMH R34 MH073625 [PI], NIMH R34MH079108 [Coinvestigator], R01DA023072 [PI], NIDA R01DA023190 [Coinvestigator], NIMH R34MH083065-01 [Coinvestigator], NIMH U01MH088278 [Co-PI], and NIMH R34MH08221 [Coinvestigator]).
Address correspondence and reprint requests to Dr. W. Curt LaFrance, Jr., Rhode Island Hospital, 593 Eddy St., Potter 3, Providence, RI 02903 william_lafrance_jr/at/brown.edu
e-Pub ahead of print on August 25, 2010, at www.neurology.org.
Study funding: Supported by the NIH/NINDS 5K23 NS045902.
Disclosure: Author disclosures are provided at the end of the article.
Presented in part as a poster at the 2009 American Academy of Neurology, Seattle, WA.
Received November 20, 2009. Accepted in final form May 27, 2010.