Antidepresivos para personas con epilepsia y depresión
Resumen
Antecedentes
Los trastornos depresivos son la comorbilidad psiquiátrica más común en las personas con epilepsia, y afectan a alrededor de un tercio, con un impacto negativo significativo sobre la calidad de vida. Existen la preocupación de que las personas puedan no estar recibiendo el tratamiento apropiado para la depresión debido a la incertidumbre con respecto a qué antidepresivo o clase funciona mejor y al riesgo percibido de exacerbación de las crisis. Esta revisión tuvo como objetivo considerar estos temas e informar la práctica clínica y la investigación futura.
Ésta es una versión actualizada de la revisión Cochrane original publicada en el número 12, 2014.
Objetivos
Evaluar la eficacia y la seguridad de los antidepresivos para el tratamiento de los síntomas depresivos y el efecto sobre la recurrencia de las crisis epilépticas en personas con epilepsia y depresión.
Métodos de búsqueda
Para esta actualización se hicieron búsquedas en CRS Web, MEDLINE, SCOPUS, PsycINFO y ClinicalTrials.gov (febrero de 2021). Se buscó en el Registro de ensayos clínicos de la Organización Mundial de la Salud en octubre de 2019, pero no fue posible realizar la actualización por la imposibilidad de acceso. No hubo restricciones de idioma.
Criterios de selección
Se incluyeron los ensayos controlados aleatorizados (ECA) y los estudios no aleatorizados de intervenciones (ENAI) prospectivos que investigaran a niños o adultos con epilepsia que fueran tratados con un antidepresivo en comparación con placebo, un antidepresivo comparativo, psicoterapia o ningún tratamiento, para los síntomas depresivos.
Obtención y análisis de los datos
Los desenlaces principales fueron los cambios en las puntuaciones de la depresión (proporción con una mejoría mayor al 50%, diferencia de medias y porcentaje que alcanzó la remisión completa) y el cambio en la frecuencia de las crisis epilépticas (diferencia de medias, porcentaje con recurrencia de las crisis o episodio de estado epiléptico). Los desenlaces secundarios incluyeron el número de participantes que se retiraron del estudio y las razones de los retiros, la calidad de vida, la función cognitiva y los eventos adversos.
Dos autores de la revisión extrajeron de forma independiente los datos de cada estudio incluido. Luego se cotejó la extracción de datos. El riesgo de sesgo se evaluó mediante la herramienta Cochrane para los ECA y la ROBINS‐I para los ENAI. Los desenlaces binarios se presentaron como razones de riesgos (RR) con intervalos de confianza (IC) del 95% o del 99% para eventos adversos concretos. Los desenlaces continuos se presentaron como diferencias de medias estandarizadas (DME) con IC del 95% y diferencias de medias (DM) con IC del 95%.
Resultados principales
En la revisión se incluyeron diez estudios (cuatro ECA y seis ENAI), con 626 participantes con epilepsia y depresión, que examinaron los efectos de los antidepresivos. Un ECA fue un estudio multicéntrico que comparó un antidepresivo con la terapia cognitivo‐conductual (TCC). Los otros tres ECA fueron estudios realizados en centro único que compararon un antidepresivo con un control activo, placebo o ningún tratamiento. Los ENAI informaron principalmente sobre desenlaces en participantes con epilepsia focal antes y después del tratamiento para la depresión con un inhibidor selectivo de la recaptación de serotonina (ISRS); un ENAI comparó los ISRS con la TCC.
Un ECA se consideró con riesgo bajo de sesgo, tres ECA con riesgo de sesgo poco claro y los seis ENAI con alto riesgo de sesgo. No fue posible realizar un metanálisis de los datos de los ECA debido a la heterogeneidad de las comparaciones terapéuticas. La certeza de la evidencia se consideró moderada a muy baja en todas las comparaciones porque los estudios individuales aportaron pocos datos de los desenlaces, y por el riesgo de sesgo, en especial de los ENAI, que no ajustaron por las variables de confusión.
La mejoría de más del 50% en los síntomas depresivos varió del 43% al 82% en los ECA, y del 24% al 97% en los ENAI, en función del antidepresivo administrado. La venlafaxina mejoró los síntomas depresivos en más del 50% en comparación con ningún tratamiento (diferencia de medias [DM] ‐7,59; intervalo de confianza [IC] del 95%: ‐11,52 a ‐3,66; un estudio, 64 participantes; evidencia de certeza baja); los resultados entre otras comparaciones no fueron concluyentes. Dos estudios que compararon los ISRS con la TCC informaron de resultados no concluyentes en cuanto al porcentaje de participantes que lograron la remisión completa de los síntomas depresivos.
Los datos sobre la frecuencia de las crisis epilépticas no indican un aumento en el riesgo de crisis con los antidepresivos en comparación con los tratamientos de control o con la situación inicial. Dos estudios midieron la calidad de vida; los antidepresivos no parecieron mejorar la calidad de vida respecto al control. Ningún estudio informó acerca de la función cognitiva.
Dos ECA y un ENAI proporcionaron datos comparativos sobre los eventos adversos; los antidepresivos no parecieron aumentar la intensidad ni el número de eventos adversos en comparación con los controles. Los ENAI informaron de mayores tasas de retiros debido a eventos adversos que debido a la falta de eficacia. Los eventos adversos informados para los antidepresivos incluyeron náuseas, mareos, sedación, cefalea, trastornos gastrointestinales, insomnio y disfunción sexual.
Conclusiones de los autores
La evidencia existente sobre la efectividad de los antidepresivos para el tratamiento de los síntomas depresivos asociados con la epilepsia continúa siendo muy limitada. Las tasas de respuesta a los antidepresivos fueron muy variables. Hay evidencia de certeza baja de un ECA pequeño (64 participantes) de que la venlafaxina podría mejorar los síntomas depresivos más que ningún tratamiento; esta evidencia se limita al tratamiento entre ocho y 16 semanas y no informa sobre los efectos a más largo plazo. Evidencia moderada a baja indica que no hay un aumento ni una exacerbación de las crisis epilépticas con los ISRS.
No existen datos comparativos para fundamentar la elección del fármaco antidepresivo o la clase de fármaco en cuanto a la eficacia o la seguridad para el tratamiento de las personas con epilepsia y depresión.
Para documentar mejor la política de tratamiento se necesitan ECA de antidepresivos que utilicen intervenciones con otras clases de tratamiento aparte de los ISRS, en muestras grandes de pacientes con epilepsia y depresión. Los estudios futuros deberían evaluar las intervenciones con duraciones más prolongadas del tratamiento para tener en cuenta el retraso en el inicio de la acción, la perdurabilidad de las respuestas al tratamiento y para proporcionar una mejor comprensión del impacto en el control de las crisis.
PICO
Resumen en términos sencillos
Antidepresivos para personas con epilepsia y depresión
Antecedentes
Los trastornos depresivos ocurren en aproximadamente un tercio de las personas con epilepsia, y a menudo requieren tratamiento antidepresivo. Sin embargo, a menudo la depresión no se trata en las personas con epilepsia, en parte debido al temor de que los antidepresivos puedan causar crisis epilépticas. Aunque existen diferentes clases de antidepresivos, todos procuran aumentar sustancias químicas neuronales clave en el cerebro y, de ese modo, aliviar los síntomas depresivos.
Características de los estudios
Se encontraron diez estudios que incluyeron a 626 pacientes con epilepsia y depresión tratados con un antidepresivo. Cuatro fueron ensayos controlados aleatorizados y seis fueron estudios de cohortes prospectivos no aleatorizados. Los estudios observaron el efecto de diferentes antidepresivos, principalmente una clase de antidepresivo denominado inhibidor selectivo de la recaptación de serotonina (ISRS). Un ensayo controlado aleatorizado y un estudio prospectivo también observaron el efecto de la terapia cognitivo‐conductual en la depresión.
Resultados
Al tener en cuenta toda la evidencia, la revisión halló que hay evidencia muy limitada de que los antidepresivos reducen los síntomas depresivos más que otros tratamientos, el placebo o ningún tratamiento en la epilepsia. Hubo información limitada acerca del efecto de los antidepresivos sobre el control de las crisis epilépticas; sin embargo, en los estudios que informaron este desenlace, no pareció haber un empeoramiento significativo de las crisis. La evidencia está actualizada hasta febrero de 2021.
Calidad los estudios
Se evaluaron los estudios con respecto al sesgo y la calidad. En general, la calidad de la evidencia se consideró moderada a baja para los ensayos clínicos y baja a muy baja para los estudios de cohortes prospectivos no aleatorizados. Se necesitan ensayos grandes y de alta calidad sobre los antidepresivos para examinar cómo se comparan las diferentes clases de antidepresivos y qué impacto pueden tener en el control de las crisis.
Authors' conclusions
Summary of findings
| Paroxetine compared to doxepin for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| doxepin | paroxetine | |||||
| > 50% reduction in depressive symptoms Follow‐up: 8 weeks | 706 per 1000 | 819 per 1000 | RR 1.16 | 67 | ⊕⊕⊕⊝ |
|
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks | NA | The mean HAMD depression score in the intervention groups was | NA | 67 | ⊕⊕⊕⊝ | |
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Withdrawals Follow‐up: 8 weeks | 88 per 1000 | 13 per 1000 | RR 0.15 | 67 | ⊕⊕⊕⊝ moderatea | doxepin: 3 withdrew paroxetine: 0 withdrew |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: 8 weeks | Reported adverse events: blurred vision, dizziness, dry mouth, sleep disorders, and urinary retention | Reported adverse events: blurred vision, dizziness, dry mouth, and sleep disorders | NA | 67 (1 RCT) | ⊕⊕⊕⊝ moderatea | There were no significant differences between treatment groups for any reported adverse events |
| *The basis for the assumed risk is the event rate in the doxepin group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded for imprecision, because only one small study contributed to the outcomes. | ||||||
| Amitriptyline compared to nomifensine for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| nomifensine | amitriptyline | |||||
| > 50% reduction in depressive symptoms Follow‐up: 12 weeks | 786 per 1000 | 432 per 1000 | RR 0.55 | 28 | ⊕⊕⊝⊝ |
|
| Mean depression scores Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Withdrawals Follow‐up: NA | ‐ | ‐ | ‐ | 0 | ‐ |
|
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the nomifensine group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded twice for imprecision, because only very small study contributed limited outcome data. | ||||||
| Venlafaxine compared to no treatment for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| no treatment | venlafaxine | |||||
| > 50% reduction in depressive symptoms Follow‐up: 8 weeks | 125 per 1000 | 406 per 1000 | RR 3.25 | 64 | ⊕⊕⊝⊝ lowa,b |
|
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks | NA | The mean HAMD depression score in the intervention group was | NA | 64 | ⊕⊕⊝⊝ lowa,b |
|
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Withdrawals Follow‐up: NA | ‐ | ‐ | ‐ | 0 | ‐ |
|
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the no treatment group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded for imprecision, because only one small study contributed to the outcomes. | ||||||
| Sertraline compared to cognitive behavioural therapy for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| CBT | sertraline | |||||
| > 50% reduction in depressive symptoms Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Mean depression scores (BDI scores; lower = better) Follow‐up: 16 weeks | NA | The mean BDI depression score in the intervention group was 0.50 lower (4.47 lower to 3.47 higher) | NA | 117 | ⊕⊕⊕⊝ moderatea | At 8 weeks: MD ‐2.50 (95% CI ‐6.28 to 1.28; 104 participants) |
| Seizure frequency Follow‐up: 16 weeks | NA | The mean frequency of GTCS per month in the intervention group was 0 lower (‐0.10 lower to 0.10 higher) The mean frequency of focal seizures with impaired awareness per month in the intervention group was 3.00 lower (7.81 lower to 1.81 higher) | NA | 96 with GTCS plus 75 with focal seizures (1 RCT) | ⊕⊕⊝⊝ lowb | At 8 weeks: GTCS per month: MD ‐0.10 (95% CI ‐0.26 to 0.06; 86 participants) focal seizures with impaired awareness per month: MD ‐2.60 (95% CI ‐6.52 to 1.32; 75 participants) |
| Withdrawals Follow‐up: 16 weeks | 176 per 1000 | 222 per 1000 (113 to 434 per 1000) | RR 1.26 (0.64 to 2.46) | 140 | ⊕⊕⊕⊝ moderatea | CBT: 6 withdrew, 6 lost to follow‐up sertraline: 7 withdrew, 9 lost to follow‐up |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Quality of life (QOLIE‐89 scale; lower = better) Follow‐up: 16 weeks | NA | The mean QOLIE‐89 score in the intervention group was 3.10 higher (3.41 lower to 9.61 higher) | NA | 118 (1 RCT) | ⊕⊕⊕⊝ moderatea | at 8 weeks: MD 6.10 (95% CI ‐0.28 to 12.48; 104 participants) |
| Adverse effects Follow‐up: 16 weeks | NA | The mean adverse event profile score in the intervention group was 2.10 lower (6.21 lower to 2.01 higher) | NA | 118 (1 RCT) | ⊕⊕⊕⊝ lowa,c | Sertraline resulted in more cases of tiredness than CBT (RR 3.54, 99% CI 1.40 to 8.96; 140 participants) Sertraline did not result in more cases of any other adverse effects than CBT. |
| *The basis for the assumed risk is the event rate in the CBT group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded once due to risk of bias: participants and personnel not blinded, and lack of blinding may have influenced outcome cCertainty of the evidence downgraded once due to imprecision: adverse event data not available for all participants who received an intervention | ||||||
| Citalopram (before and after treatment) for people with epilepsy and depression | ||||
|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of participants | Certainty of the evidence | Comments |
| Citalopram (before and after) | ||||
| > 50% reduction in depressive symptoms Follow‐up: 4 months | 11 out of 45 participants (24%) showed a 50% or more improvement in depression scores after treatment compared to baseline. | 45 (1 NRSI) | ⊕⊕⊝⊝ lowa | |
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks to 4 months | Improved depression scores were shown after citalopram compared to before (see comment) | 88 | ⊕⊕⊝⊝ | SMD in HAMD score was 1.17 (95% CI 0.96 to 1.38), indicating improved outcomes and a large treatment effect. |
| Seizure frequency Follow‐up: 8 weeks to 4 months | See comment | 88 | ⊕⊝⊝⊝ | Results were mixed between studies; due to very high heterogeneity (I² = 81%), we did not present the overall effect estimate. |
| Withdrawals Follow‐up: 8 weeks to 4 months | 6/45 participants (13%) withdrew from one study; 0/43 from the other study | 88 (2 NRSI) | ⊕⊕⊝⊝ lowa | |
| Cognitive functioning Follow‐up: NA | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: 8 weeks to 4 months | 22/45 participants (56%) experienced adverse events in one study; 5/43 (12%) in the other study | 88 (2 NRSI) | ⊕⊕⊝⊝ lowa | Specific adverse events reported: nausea, sexual dysfunction, headache, dizziness, drowsiness, and fatigue |
| CI: confidence interval; HAMD: Hamilton Rating Scale for Depression; NRSI: non‐randomised studies of interventions | ||||
| GRADE Working Group grades of evidence | ||||
| aCertainty of the evidence downgraded twice as studies were judged to be at serious risk of bias due to lack of blinding, which may have influenced participant‐recorded outcomes, and lack of adjustment for confounding variables. | ||||
| Selective serotonin reuptake inhibitorscompared to cognitive behavioural therapy for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| CBT | SSRIs | |||||
| > 50% reduction in depressive symptoms Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Mean depression scores (BDI scores; lower = better) Follow‐up: 12 weeks | NA | The mean BDI depression score in the intervention group was 4.90 lower (14.90 lower to 4.80 higher) | NA | 15 | ⊕⊝⊝⊝ very lowa,b | at 6 weeks: MD ‐2.60 (95% CI ‐11.58 to 6.38; 15 participants) |
| Seizure frequency Follow‐up: 12 weeks | NA | The mean frequency of seizures per month in the intervention group was 1.60 lower (5.63 lower to 2.43 higher) | NA | 15 (1 NRSI) | ⊕⊝⊝⊝ very lowa,b | |
| Withdrawals Follow‐up: 12 weeks | 286 per 1000 | 126 per 1000 (14 to 1000 per 1000) | RR 0.44 (0.05 to 3.85) | 15 (1 NRSI) | ⊕⊝⊝⊝ very lowa,b | CBT: 2 lost to follow‐up SSRI: 1 lost to follow‐up in the |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life ‐ QOLIE‐31 scale Follow‐up: 12 weeks | NA | The mean QOLIE‐31 score in the intervention group was 0.50 lower (19.67 lower to 18.67 higher) | NA | 15 (1 NRSI) | ⊕⊝⊝⊝1,2 very low |
|
| Adverse effects ‐ adverse event profile Follow‐up: 16 weeks | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the CBT group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1. Certainty of the evidence downgraded twice as the study was judged to be at serious risk of bias with regards to lack of blinding which may have influenced participant recorded outcomes and lack of adjustment for confounding variables. 2. Certainty of the evidence downgraded once due to imprecision: very small study of 15 participants, confidence intervals around effect estimates wide | ||||||
Background
This is an updated version of a Cochrane Review published in Issue 12, 2014 (Maguire 2014).
Description of the condition
Depressive disorders are the most common psychiatric comorbidity in people with epilepsy (Tellez‐Zenteno 2007), and they are the strongest predictor of poor quality of life (Boylan 2004). Symptoms of depression include low mood, tiredness, and apathy. Sleep and cognitive functioning may also be affected. Depressive disorders occur in approximately one‐third of the people with epilepsy (Baker 1996; Indaco 1992; Jacoby 1996; Mendez 1986). These disorders are broadly divided into unipolar (depression only) and bipolar disorders (depression associated with mania or hypomania; (APA 2000)). Depressive disorders in epilepsy may be mediated via the interplay of neurobiological, psychosocial, and iatrogenic factors (Lambert 1999). Depressive symptoms or episodes may occur inter‐ictally (i.e. they appear unrelated to seizures) or peri‐ictally (preceding, during, or following seizures). This is an important distinction, as a person may require modification of his or her antiepileptic drug regimen, commencement of antidepressant drug therapy, or both. In some people, the depressive symptoms may follow a significant period of seizure remission in previously uncontrolled epilepsy, thought to occur via neuro‐biochemical changes, and termed 'forced normalisation' (Trimble 1998). Studies examining clinical predictors of risk for depression in people with epilepsy have produced inconsistent results (Lin 2012). There is a perceived greater risk of depression in people with temporal lobe epilepsy, although elevated rates of depression have been found in generalised and extra‐temporal focal epilepsy (Adams 2008). Epilepsy‐related factors as predictors of risk for depression are inconsistent. Psychosocial factors, such as life stress, coping style, social support, perceived stigma, and personality are more consistent predictors of depression in people with epilepsy (Hermann 2000).
In 2008, the Food and Drug Administration issued a health alert about an increased risk of suicidal ideation in people taking antiepileptic drugs (Hesdorffer 2009). This alert was based on a meta‐analysis of approximately 28,000 participants who had participated in randomised controlled trials (RCTs) investigating 11 antiepileptic drugs. There were four completed suicides, all of whom had taken antiepileptic drugs, compared to no suicides in the placebo groups (odds ratio (OR) 1.8; 95% confidence interval (CI) 1.24 to 2.66). Since this alert, a number of observational studies have investigated the association, reporting conflicting results, and the International League Against Epilepsy (ILAE) Commission on Neuropsychobiology a has published a consensus statement on the risk of suicide with antiepileptic drugs (Mula 2013). Whilst the exact risk of suicide with antiepileptic drugs is unknown, depression, as a treatment emergent adverse effect, is associated with some antiepileptic drugs (GABAergic antiepileptic drugs: benzodiazepines, vigabatrin, gabapentin; and also topiramate, levetiracetam, and zonisamide; (Mula 2009)). Other antiepileptic drugs appear to have mood‐stabilising properties (valproic acid, lamotrigine, carbamazepine, oxcarbazepine), which may benefit people with epilepsy and depression. Enzyme‐inducing antiepileptic drugs (i.e. carbamazepine) may lower plasma levels of antidepressants, thus impacting on their effectiveness.
Case control studies have shown that participants with depression have a two‐ to seven‐fold higher risk of developing epilepsy, implying a bi‐directional relationship (Hesdorffer 2000; Hesdorffer 2006; Hesdorffer 2012). A number of factors may explain this, for example shared pathophysiology involving disturbances in several key neurotransmitter systems (Bagdy 2007), structural lesions (frontal lobe tumours), or a genetic susceptibility. However, there is also the possibility that the use of antidepressants may trigger seizures. This is a common concern for healthcare professionals, and may influence their decisions to start antidepressant treatment (Cotterman‐Hart 2010).
Description of the intervention
Antidepressants are a heterogeneous class of drugs that have been the mainstay of pharmacological treatment for depressive disorders. There are 10 classes of antidepressants used to treat depressive disorders, with 60% to 70% of depressive episodes responding to current treatment (Klerman 1990; Sackeim 2006). These are:
-
tricyclic antidepressants;
-
selective serotonin reuptake inhibitors;
-
serotonin‐norepinephrine reuptake inhibitors;
-
monoamine oxidase inhibitors;
-
serotonin/antagonist reuptake inhibitors (i.e. trazodone);
-
dopamine and norepinephrine reuptake inhibitors (i.e. bupropion);
-
a‐2 antagonists (i.e. mirtazapine);
-
norepinephrine reuptake inhibitors (i.e. reboxetine);
-
selective serotonin reuptake enhancers (i.e. tianeptine); and
-
serotonin 5HT₂C receptor antagonists (i.e. agomelatine).
These drugs work by targeting serotonergic, or noradrenergic, or dopaminergic neurotransmission, or a combination, with the aim of increasing their synaptic concentrations (Stahl 2000). Glutamate antagonists represent a novel class of drug currently being tested in refractory depression (Zarate 2006).
The risk of seizures with antidepressants was reported in early studies of the first generation antidepressants, notably tricyclic antidepressants (Preskorn 1992; Wroblewski 1990). Alper 2007 reviewed the incidence of seizures in 75,000 non‐epileptic participants in phase II and phase III trials of antidepressant treatment. They reported lower incidence rates of seizures in those randomised to an antidepressant versus placebo (standardised incidence ratio 0.48; 95% CI 0.36 to 0.61). Coupland 2011 examined 60,746 primary care participants aged 65 and over, treated for depression with antidepressants, between 1996 and 2007, and showed increased risks of epilepsy or seizures for selective serotonin reuptake inhibitors (hazard ratio (HR) 1.80; 95% CI 1.32 to 2.43), and other antidepressant classes (HR 2.20; 95% CI 1.46 to 3.30) versus tricyclic antidepressants. Venlafaxine was associated with the highest risk of seizures.
How the intervention might work
There appears to be a significant relationship between epilepsy and depression. From studies, it is emerging that they share common neurobiological substrates that involve hyperactivity of the hypothalamic pituitary adrenal axis, and the disturbance of different neurotransmitter systems, mainly serotonin and norepinephrine (Dell'osso 2013). The density of serotonin receptors is high in the mesial temporal and prefrontal areas (Gilliam 2005b). In critical brain regions, such as the limbic system and prefrontal areas, enforced serotonergic circuits seem to be responsible for increasing seizure threshold (Kondziella 2009).
Antidepressants of the selective serotonin reuptake inhibitor family have been reported to be safe in treating depression in people with epilepsy, and to possess antiepileptic properties in animal models of epilepsy (Hamid 2013). Based on clinical data, it has been suggested that selective serotonin reuptake inhibitors can decrease the seizure frequency in refractory epilepsy (Kondziella 2009). This is believed to be due to the increase in the concentration of serotonin. The study shows that the concentration of endogenous serotonin (5‐HT) and the activity of its receptor subtypes, 5–HT(1A), 5‐HT(2C), 5‐HT(3), and 5–HT(7), play a significant role in the pathogenesis of epilepsies (Bagdy 2007). Therefore, medications with serotonin agonist and antagonist properties can play a significant role in the pathogenesis of epilepsies.
Why it is important to do this review
Depression is common in people with epilepsy, and has a significant negative impact on quality of life (Gilliam 2005b; Kondziella 2009). There is concern that they may not be receiving appropriate treatment for their depression because of uncertainty around which antidepressant, or class, works best, and the perceived risk of exacerbating seizures. This review aims to address these issues, and to inform clinical practice and future research.
Objectives
To evaluate the efficacy and safety of antidepressants in treating depressive symptoms and the effect on seizure recurrence, in people with epilepsy and depression.
Methods
Criteria for considering studies for this review
Types of studies
-
Randomised controlled trials (RCTs)
-
Prospective non‐randomised cohort controlled and uncontrolled studies (NRSI; with a control group including participants acting as their own control group (i.e. before‐after studies)).
-
We considered prospective non‐randomised cohort studies in this review because of the known delayed effect of antidepressants on depressive symptoms, which may not be effectively detected in short‐term randomised trials. Similarly, prospective non‐randomised studies are more likely to recruit populations of participants who better reflect clinical practice, since depression can affect any person with epilepsy.
-
Types of participants
We considered participants who satisfied all of the following criteria:
-
any age;
-
diagnosis of epilepsy (any type);
-
treated with antidepressants for co‐existing depression (including participants with major depressive disorder, adjustment disorder, and dysthymic disorder), based on standardised criteria, according to participant scores on validated tools, or both (e.g. Hamilton Rating Scale for Depression).
Types of interventions
-
Intervention group: participants who received an antidepressant drug in addition to an existing antiepileptic drug regimen
-
Control group(s): participants who received a placebo, comparative antidepressant, psychotherapy, or no treatment in addition to an existing antiepileptic drug regimen
Types of outcome measures
Primary outcomes
-
Depression scores
-
The proportion of participants with a greater than 50% improvement in depressive symptoms (defined as a 'response') compared to baseline
-
Mean difference in depression scores following treatment (compared to baseline or between‐group comparison)
-
The proportion of participants achieving complete remission of depressive symptoms
-
If the data allowed, we planned to analyse outcomes at ≤ 12 weeks (short‐term), 13 to 26 weeks (medium‐term), and ≥ 26 weeks (long‐term); however, we were unable to perform these analyses.
-
Change in seizure frequency
-
The mean difference in seizure frequency
-
The proportion of participants with a seizure recurrence
-
The proportion of participants with an episode of status epilepticus
-
Secondary outcomes
-
Withdrawals
-
For specific reasons
-
For any reasons
-
-
Global state
-
Clinically important change in global state (as defined by the individual studies)
-
Relapse (as defined by the individual studies)
-
-
Mental state
-
Clinically important change in general mental state score
-
General mental state score (average and end point)
-
Clinically important change in specific symptoms (sleep, anhedonia, suicidal ideas)
-
Specific symptom score (average and end point)
-
-
General functioning
-
Clinically important change in general functioning
-
General functioning score (average and end point)
-
-
Cognitive functioning
-
Clinically important change in overall cognitive functioning
-
Overall cognitive functioning score (end point and average)
-
Clinically important change in specific cognitive functioning (attention, concentration, memory, language, executive functioning)
-
Specific cognitive score (average and end point)
-
-
Quality of life
-
Clinically important change in quality of life
-
Any change in quality of life score (average and end point)
-
-
Behaviour
-
Clinically important change in general behaviour
-
Any important change in general behaviour (average and end point)
-
Clinically important change in specific aspects of behaviour
-
Any important change in specific aspects of behaviour score (average and end point)
-
-
Adverse effects
-
Death
-
Any non‐serious general adverse effects (e.g. gastrointestinal effects, anorexia, dizziness, dry mouth, insomnia, sexual dysfunction, hypotension)
-
Any serious, specific adverse effects (hypersensitivity reaction)
-
Any change in general adverse effect score (average and end point)
-
Clinically important change in specific adverse effects
-
Any change in specific adverse effects score (average and end point)
-
Search methods for identification of studies
Electronic searches
We ran searches for the original review in March 2013. We ran subsequent searches in May 2014, October 2016, July 2018, and October 2019. For the latest update, we searched the following databases.
-
Cochrane Central Register of Controlled Trials (CENTRAL), in CRS Web (searched 1 February 2021), using the search strategy shown in Appendix 1;
-
MEDLINE Ovid (1946 to 29 January 2021), using the search strategy shown in Appendix 2;
-
SCOPUS (1823 to 1 February 2021), using the search strategy shown in Appendix 3;
-
PsycINFO EBSCOhost (1887 to 1 February 2021), using the search strategy shown in Appendix 4;
-
ClinicalTrials.gov (searched 1 February 2021), using the search strategy shown in Appendix 5;
-
WHO International Clinical Trials Registry Platform (ICTRP; searched 22 October 2019), using the search strategy shown in Appendix 6. We were unable to update this search because the ICTRP website was inaccessible.
CRS Web includes randomised or quasi‐randomised, controlled trials from the Specialized Registers of Cochrane Review Groups, including Epilepsy, CENTRAL, PubMed, Embase, ClinicalTrials.gov, and ICTRP.
There were no language restrictions.
Searching other resources
We checked the reference lists of retrieved studies for additional reports of relevant studies.
We also contacted lead study authors for any relevant unpublished material.
We identified duplicate studies by screening reports according to title, authors’ names, location, and medical institute, omitting any duplicated studies.
We identified any grey literature studies published in the last five years by searching:
1. Zetoc database;
2. ISI Proceedings;
3. International Bureau for Epilepsy (IBE) congress proceedings database;
4. International League Against Epilepsy (ILAE) congress proceedings database;
5. Abstract books of symposia and congresses, meeting abstracts, and research reports
Data collection and analysis
Selection of studies
Two authors (MM, SN) independently assessed all citations generated from the searches for inclusion. Where disputes arose, we acquired the full report for more detailed scrutiny.
Data extraction and management
Two authors (MM, SJN) undertook separate data extraction for each included study. We then cross‐checked the data extraction. We extracted data using pre‐standardised data extraction forms. We discussed any disagreement, documented decisions, and if necessary, contacted trialists for clarification.
We extracted the following information from the included studies.
Methodological and trial design
-
Year of publication
-
Number of study centres
-
Language
-
Industry funding
-
Study design (RCT, prospective cohort study, retrospective cohort study)
-
Blinding
-
Type of control group (placebo, comparative antidepressant, no treatment)
-
Sample size
-
Follow‐up period
-
Class of antidepressant as intervention
-
Dose range of intervention
-
Inclusion and exclusion criteria
Participant demographic information
-
Age range
-
Number of male/female participants
-
Duration of epilepsy
-
Previous number of antiepileptic drugs
-
Epilepsy type (focal, generalised, unclassified)
-
Location of epilepsy (temporal, extra‐temporal)
-
Baseline mean depression score or severity
-
Baseline mean seizure frequency/month
Outcomes
-
The number of participants experiencing each outcome recorded per treatment group
-
Number of dropouts
Assessment of risk of bias in included studies
Two review authors (SJN, MJM) independently assessed the risk of bias for the included studies.
Due to the non‐randomised design of some studies, we assessed risk of bias for non‐randomised studies using the ROBINS‐I tool (Sterne 2016). This tool considers seven domains of bias: two domains of bias pre‐intervention (bias due to confounding and bias in selection of participants into the study); one domain of bias at intervention (bias in the measurement of interventions); and four domains of bias post‐intervention (bias due to departures from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported result). We planned to perform a separate 'Risk of bias' assessment for each outcome of interest in the study.
Important confounders of interest in this Cochrane Review included:
-
mean age;
-
epilepsy type (focal or generalised);
-
mean duration of epilepsy;
-
location of epilepsy;
-
mean baseline seizure frequency;
-
mean baseline depression score.
Each domain of bias contained signalling questions to facilitate judgements of risk of bias. The response options for the signalling questions were: yes; probably yes; probably no; no; and no information. We specified the signalling questions for each domain in Appendix 7.
The 'Risk of bias' judgement options for each domain were:
-
low risk of bias: the study is comparable to a well‐performed randomised trial with regard to this domain;
-
moderate risk of bias: the study is sound for a non‐randomised study with regard to this domain, but cannot be considered comparable to a well‐performed randomised trial;
-
serious risk of bias: the study has some important problems in this domain;
-
critical risk of bias: the study is too problematic in this domain to provide any useful evidence on the effects of the intervention;
-
no information on which to base a judgement about risk of bias for this domain.
We presented guidance for an overall risk of bias for a study, based on outcomes from 'Risk of bias' judgements of each domain, in Table 1.
| Risk of bias judgement | Criteria based on seven risk of bias domains |
|---|---|
| Low risk of bias: the study is comparable to a well‐performed randomised trial | The study is judged to be at low risk of bias for all domains |
| Moderate risk of bias: the study appears to provide sound evidence for a non‐randomised study but cannot be considered comparable to a well‐performed randomised trial | The study is judged to be at low or moderate risk of bias for all domains |
| Serious risk of bias: the study has some important problems | The study is judged to be at serious risk of bias in at least 1 domain, but not at critical risk of bias in any domain |
| Critical risk of bias: the study is too problematic to provide any useful evidence on the effects of intervention | The study is judged to be at critical risk of bias in at least 1 domain |
| No information on which to base a judgement about risk of bias | There is no clear indication that the study is at serious or critical risk of bias, and there is a lack of information in 1 or more key domains of bias (a judgement is required for this) |
For RCTs, we assessed all domains of the Cochrane tool for assessing risk of bias (Higgins 2011). We rated each of the following six domains as low, high, or unclear risk of bias: method of generating random sequence, allocation concealment, blinding methods, incomplete outcome data, selective outcome reporting, and other sources of bias.
The two review authors resolved any discrepancies in the 'Risk of bias' judgements by discussion.
Measures of treatment effect
For binary outcomes (50% or greater improvement in depressive symptoms, complete remission of depressive symptoms, and % treatment withdrawal), we presented results as risk ratios (RR) with 95% confidence interval (CI). To allow for multiple statistical testing, we presented RRs with 99% CIs for specific adverse events.
For continuous outcomes (mean change in depression score), we presented results as mean differences (MD) or standardised mean differences (SMD) with 95% CIs.
Unit of analysis issues
Studies using a variety of depression measures created issues when we wanted to combine results in a meta‐analysis. Where appropriate, we used the SMD to allow for these variances.
Dealing with missing data
We sought missing statistics from studies through contact with the study authors. We sought reasons for missing data to determine whether the data were missing at random or not. We found no data missing at random.
Assessment of heterogeneity
We assessed clinical heterogeneity by comparing the distribution of important participant factors between studies (age, epilepsy type, duration of epilepsy, baseline depression score, baseline seizure frequency) and trial factors (study design, type of control group, antidepressant drug class, type of depression disorder). We assessed statistical heterogeneity by using the I² statistic. We considered an I² value of 75% or more indicated considerable heterogeneity, 50% to 90% indicated substantial heterogeneity, and 30% to 60% indicated moderate heterogeneity. If the I² value was 75% or more, we made an a priori decision not to carry out meta‐analysis; instead, we used a narrative form for the review, and discussed all comparisons according to the findings presented within the studies. We planned meta‐regression techniques, where possible, to investigate possible sources of heterogeneity, however, we were unable to investigate this within this review.
Assessment of reporting biases
1. Protocol versus full study
We investigated outcome reporting bias using the ORBIT classification system, allocating studies a letter from A to I if we suspected the presence of selective outcome reporting bias (Kirkham 2010).
2. Funnel plot
Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results (Higgins 2020; Sterne 2000). Funnel plots can be used to investigate reporting biases, but are of limited power to detect small‐study effects. We did not use funnel plots for outcomes for which there were 10 or fewer studies, or when all studies were of similar sizes.
Data synthesis
We synthesised data using the RR, the MD, or the SMD, depending on the measures used in both the controlled and uncontrolled studies. We carried out a sensitivity analysis to check for differences between a random‐effects model and fixed‐effect model in influencing conclusions. If differences between the models existed, we intended to report outcomes based on the random‐effects model, which incorporates an assumption that the different studies are estimating different, yet related, intervention effects.
For controlled studies, we intended to carry out meta‐analysis using the Mantel‐Haenszel method for dichotomous outcomes, and the inverse variance method for continuous outcomes. For before‐after studies, we used the inverse variance methods for continuous outcomes in meta‐analysis.
We did not combine data for outcomes measured in both randomised and non‐randomised studies. We reported combined data on outcomes for randomised and non‐randomised studies separately.
We stratified each comparison by type of control group, study design, study characteristics, or a combination, to ensure appropriate combination of study data.
Subgroup analysis and investigation of heterogeneity
Where possible, we planned to stratify subgroup analysis by antidepressant drug class, epilepsy type, and age. For investigation of heterogeneity, please see Assessment of heterogeneity.
Sensitivity analysis
We intended to carry out sensitivity analysis if peculiarities in study quality were found (Assessment of risk of bias in included studies). We planned to report the analysis for all studies, and then compare this to an analysis of only studies at low risk of bias. However, as we could not combine RCTs in meta‐analysis, and we judged all NRSIs to be at serious risk of bias, we did not perform any sensitivity analyses.
Summary of findings and assessment of the certainty of the evidence
We made an overall summary judgement of risk of bias for each study per outcome, followed by an overall judgement per outcome across studies. We had planned to incorporate the 'Risk of bias' judgements into the analysis using a sensitivity analysis, so that a secondary analysis of the data included only studies rated as low risk of bias. However, we were unable to do this due to the small amount of studies and lack of data. We presented both results in the Results section of the review. Where applicable, we created 'Summary of findings' tables for outcomes, and graded each outcome using the GRADE approach (Guyatt 2008). Outcomes reported in 'Summary of findings' tables include: depression scores, seizure frequency, withdrawals, cognitive functioning, quality of life, and adverse effects.
Results
Description of studies
Results of the search
We identified 18 potentially eligible studies. We made our final assessment of eligibility by checking the full text of the reports. Figure 1 outlines the flow diagram of search results, eligible records, and study exclusions.
Study flow diagram
We excluded six studies; three studies did not meet the inclusion criteria, and a further three studies met the inclusion criteria but did not report any results for any of the primary and secondary outcomes. We attempted to contact trial authors of these three studies, but received no response.
Two studies are currently ongoing (EUCTR2017‐000990‐35‐IT; EUCTR2018‐003464‐32‐HU).
Included studies
Of the ten remaining studies, four were randomised controlled trials and six non‐randomised prospective cohort studies examining the effect of antidepressant drugs.
We found four randomised trials of antidepressant versus cognitive behavioural therapy, active drug control, placebo, or no treatment, which reported on the primary efficacy outcome (Gilliam 2019; Li 2005; Robertson 1985; Zhu 2004). One was a multi‐centre study (Gilliam 2019), the rest were single centre. A total of 313 participants were randomised in these studies; 199 participants had focal epilepsy. The remaining six non‐randomised prospective cohort studies reported on a total of 313 participants treated with an antidepressant, reported on the primary efficacy outcome (Hovorka 2000; Kanner 2000; Kuhn 2003; Orjuela‐Rojas 2015; Specchio 2004; Thome‐Souza 2007). Two hundred and ninety seven participants had focal epilepsy; 290 were treated with a selective serotonin reuptake inhibitor, and seven with cognitive behavioural therapy.
Seven studies reported outcomes for adults participants only; three studies reported outcomes for adults and children (Kanner 2000; Thome‐Souza 2007; Zhu 2004).
Eight studies included participants with focal onset epilepsy; two studies included participants with generalised onset epilepsy (Li 2005; Zhu 2004). In all trials, there was a larger or equal number of female participants.
Six studies evaluated the efficacy of selective serotonin reuptake inhibitors (citalopram, sertraline, fluoxetine) versus no treatment, a tricyclic antidepressant (doxepin), a norepinephrine reuptake inhibitor (reboxetine), or an alpha‐2 antagonists (mirtazapine; (Hovorka 2000; Kanner 2000; Kuhn 2003; Li 2005; Specchio 2004; Thome‐Souza 2007)). One study evaluated a serotonin‐norepinephrine reuptake inhibitor (venlafaxine) versus no treatment (Zhu 2004); another, a tricyclic antidepressant (amitriptyline) versus dopamine and a norepinephrine reuptake inhibitor (nomifensine; (Robertson 1985)). Two studies evaluated the efficacy of selective serotonin reuptake inhibitors (sertraline, citalopram) versus cognitive behavioural therapy (Gilliam 2019; Orjuela‐Rojas 2015).
Five studies used the Hamilton Rating Scale for Depression (HAMD; (Hovorka 2000; Kuhn 2003; Li 2005; Robertson 1985; Zhu 2004)). Two studies used both the Mini International Neuropsychiatric Interview (MINI) and Beck Depression Inventory (BDI; (Gilliam 2019; Orjuela‐Rojas 2015)). One study used the Montgomery‐Åsberg Depression Rating Scale (MADRS; Specchio 2004)), one study used the Kiddie SADS depression score (Thome‐Souza 2007), and one study did not report the use of a specific depression rating scale (Kanner 2000).
Randomised Controlled Trials
Gilliam 2019 was a published multi‐centre, randomised controlled trial, conducted in the USA, with 140 participants. Participants had a mean age of 39.6 years, 77 participants were female, and 56% of participants had focal epilepsy. Participants were randomised to receive sertraline (50 mg to 200 mg/day) or cognitive behavioural therapy (CBT). Changes in the MINI score, CES‐D, Beck Depression Inventory (BDI), QOLIE‐89 score, and adverse event profile (AEP) scores were compared from baseline to 8 weeks and 16 weeks. Changes in monthly seizure rates were compared from a retrospective 3‐month baseline, to 8 weeks and 16 weeks. Recurrence of a GTCS during the study was compared in those subjects who had not had a GTCS in a retrospective six‐month period prior to study enrolment. Of the 140 participants, 42 did not complete treatment as assigned, 15 of whom were lost to follow‐up. All participants were included in the reported analysis.
Li 2005 was a published, single‐centre, randomised controlled trial, conducted in China, with 67 participants. Forty‐two participants had generalised onset epilepsy. The participants were aged between 14 and 62 years, and 35 participants were female. Thirty‐three participants were randomised to paroxetine, which was started at 10 mg/day and titrated up to 40 mg/day, depending on response. Thirty‐four participants were randomised to doxepin, which was started at 25 mg/day and titrated up, according to response (mean dose 100 mg/day). The Hamilton Rating Scale for Depression (HAMD) score was measured at eight weeks and compared to the baseline score. Seizure frequency was not assessed. Three participants in the doxepin treatment arm dropped out and were not included in the primary analysis.
Robertson 1985 was a published, single‐centre, randomised, placebo‐controlled trial, conducted in the UK, with 42 participants. The majority had focal onset epilepsy. The participants were aged between 18 and 60 years, and 26 were female. Participants were randomised to amitriptyline, nomifensine, or placebo. All treatment arms completed a six‐week phase, and then both active treatment arms continued the study for a further six weeks. At 12 weeks of treatment, Hamilton Rating Scale for Depression (HAMD) scores were compared to baseline. Three participants withdrew from the study. Twenty‐eight participants in the active treatment arms were included in the primary outcome analysis at 12 weeks.
Zhu 2004 was a published, single‐centre, randomised trial of venlafaxine versus no treatment, conducted in China, with 64 participants. The participants were aged between 7 and 60 years. Thirty‐two participants were randomised to venlafaxine 25 mg/day to 75 mg/day; 32 participants received no treatment. Depression scores were measured at eight weeks of treatment, using the Hamilton Rating Scale for Depression (HAMD), and compared to baseline. Seizure frequency was not reported. There were no dropouts, and all participants were included in the primary outcome analysis.
Non‐Randomised Prospective Cohort Studies
Hovorka 2000 was a published, single‐centre, prospective cohort study, conducted in the Czech Republic, with 43 participants. Two‐thirds of the participants had focal epilepsy. Participants were between the ages of 12 and 49 years, and 35 participants were female. . All participants received citalopram (mean daily dose 22.6 mg ± 8.3 mg) for eight weeks. At four and eight weeks, Hamilton Rating Scale for Depression (HAMD) depression scores and seizure frequencies were measured and compared to an unspecified baseline period. There were no treatment withdrawals; all 43 participants were included in the reported analysis.
Kanner 2000 was a published, single‐centre, prospective cohort study, conducted in the USA, with 100 participants. Participants were aged between 6 and 62 years, 95% had focal onset epilepsy, and 49 participants were female. All participants received sertraline (25 mg/day to 200 mg/day; mean dose of 108 mg/day ± 56.9 mg/day), and were followed up for 0.2 to 38 months. Monthly seizure frequencies were compared during the treatment period, and to a 3‐ and 12‐month retrospective baseline period. No changes in depression scores were reported. Of the 100 participants, 18 withdrew from the study. All participants were included in the primary efficacy analysis.
Kuhn 2003 was a published, single‐centre, prospective cohort study, conducted in Germany, with 75 participants. All had focal onset epilepsy (temporal lobe). The participants were aged between 19 and 68 years, and 45 participants were female. Twenty‐seven participants received mirtazepine (mean daily dose 32.2 mg), 33 participants received citalopram (mean daily dose 24.2 mg), and 15 participants received reboxetine (mean daily dose 6.9 mg). Changes in Hamilton Rating Scale for Depression (HAMD) depression scores and treatment responders were measured at four weeks and 20 to 30 weeks, and compared to baseline scores. Changes in seizure frequency were not measured. Forty‐two participants dropped out; eight dropped out between baseline and week four, 34 dropped out between week four and weeks 20 to 30. The last observation carried forward method was used, and all participants were included in the primary efficacy outcomes.
Orjuela‐Rojas 2015 was a published, single‐centre, prospective study, conducted in Mexico, with 15 participants. All participants had temporal lobe epilepsy, and 11 participants were female. Seven participants received 12 sessions of cognitive behavioural therapy, and 8 participants received an SSRI (either sertraline (200 mg/day to 400 mg/day or citalopram 20 mg/day) over a 12‐week period. The Beck Depression Inventory (BDI) score, HADS score, QOLIE‐31, MINI, and monthly seizure frequency were compared at 6 and 12 weeks to baseline scores. There were two dropouts in the CBT group, and one participant was lost to follow‐up in the SSRI group. All participants were included in the reported analyses.
Specchio 2004 was a published, multi‐centre, prospective cohort study, conducted in Italy, with 45 participants. Forty‐four participants had focal onset epilepsy. The participants had a mean age of 42.7 years, and 31 were female. All participants received citalopram for four months. Montgomery–Åsberg Depression Rating Scale (MADRS) depression scores and seizure frequency were measured at two and four months on citalopram, and compared to baseline measures. Six participants withdrew from the study and were omitted from the primary outcome analysis.
Thome‐Souza 2007 was a published, single‐centre, prospective cohort study, conducted in Brazil, with 36 participants with focal onset epilepsy. The participants were aged between six and 18 years, and 19 were female. Twenty‐eight participants received sertraline (50 mg/day to 200 mg/day), and eight participants received fluoxetine (20 mg/day to 80 mg/day) for 12 to 78 months. Change in Kiddie SADS score was measured during the treatment phase, and compared to a six‐month baseline score. Seizure exacerbation was also observed during the treatment phase. One participant dropped out of the study. All participants were included in the primary outcome analysis.
Excluded studies
We excluded six studies. One was a clinical trial comparing escitalopram and no treatment (NCT01244724). This trial was terminated early due to problems with recruitment, and there were no available published data. The second study was a very small pilot trial comparing escitalopram and referral to psychiatry (NCT03464383). Only three participants were recruited to each study arm, and limited outcome data were available. The third study reported on a small case series of participants with epilepsy, who were taking a combined tricyclic antidepressant and SSRI for depression, and did not fulfil the inclusion criteria (Blumer 1997).
A further three studies were excluded because they did not report any primary or secondary outcome data and trial authors could not be contacted (Harmant 1990; Machado 2010; NCT00595699).
See Characteristics of excluded studies for more details of the studies.
Risk of bias in included studies
We rated risk of bias across each domain for each study, and then made an overall judgement on risk of bias for each study, using the Cochrane 'Risk of bias' tool for the RCTs, and the ROBINS‐I tool for the NRSIs.
See Characteristics of included studies for review authors' judgements about each 'Risk of bias' item for each included RCT, and Table 2 for review authors' judgements about each 'Risk of bias' item for each included non‐randomised study.
| Domain and risk of bias judgement | Study | |||||
|---|---|---|---|---|---|---|
| Bias due to confounding | Serious | Moderate | Serious | Moderate | Serious | Serious |
| Bias in selection of participants into the study | Moderate | Low | Moderate | Low | Serious | Moderate |
| Bias in classification of interventions | Low | Low | Moderate | Low | Low | Low |
| Bias due to deviations from intended interventions | Low | Low | Moderate | Moderate | Moderate | Low |
| Bias due to missing data | Low | Low | Serious | Serious | Serious | Low |
| Bias in measurement of outcomes | Moderate | Serious | Moderate | Moderate | Moderate | Moderate |
| Bias in selection of the reported result | Low | Moderate | Low | Moderate | Low | Low |
| Overall judgement | Serious | Serious | Serious | Serious | Serious | Serious |
| Support for judgement | No adjustment for confounding; unclear if participants were recruited consecutively; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes | Some analyses to investigate prognostic variables, but not a complete analysis of confounders; unclear which variables were of interest in advance, and if other characteristics were tested and analysed; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes. The measure of depression was not an accurate or reliable measure | No adjustment for confounding; many discontinuations due to adverse events and non‐compliance, with outcome data analysed by last observation carried forward; unclear if participants were already receiving the intervention on entry into the study, and exactly how groups were assigned. Lack of blinding may have influenced some participant‐reported outcomes. | Some analyses to investigate prognostic variables, but not a complete analysis of confounders; unclear which variables were of interest in advance, and if other characteristics were tested and analysed; small groups and dropouts, with outcome data analysed by last observation carried forward, Lack of blinding may have influenced some participant‐reported outcomes. | No adjustment for confounding; outcome data included only for those who completed analysis (6 participants excluded); and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes | No adjustment for confounding; unclear if participants were recruited consecutively; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes. |
Overall, we rated one RCT at low risk of bias (Gilliam 2019), three RCTs at unclear risk of bias (Li 2005; Robertson 1985; Zhu 2004), and all six NRSIs at serious risk of bias.
Allocation
Generation of random sequence and allocation concealment (RCTs)
We rated three RCTs at low risk of bias for sequence generation, as they used adequate methods (Gilliam 2019; Li 2005; Robertson 1985). For allocation concealment, we rated Li 2005 at unclear risk of bias, Robertson 1985 at low risk of bias, and Gilliam 2019 at low risk of bias, as it was not possible to conceal allocation due to different treatment types (CBT versus sertraline). We rated the fourth RCT at unclear risk of bias for both sequence generation and allocation concealment (Zhu 2004). (See the 'Characteristics of included studies' for more detailed information on methodology).
Selection of participants into the study (NRSIs)
We judged two studies, which recruited consecutive participants over a specified time frame, to be at low risk of bias in the selection of participants into the study (Kanner 2000; Orjuela‐Rojas 2015); three studies to be at moderate risk of bias, as we were unclear if participants were consecutive, the time frame of recruitment was unclear, or we were unclear if participants were already taking an antidepressant at recruitment (Hovorka 2000; Kuhn 2003; Thome‐Souza 2007); and one study to be at serious risk of bias, as although consecutive participants were recruited, only those who completed the intervention were included in the study results (Specchio 2004).
Classification of interventions (NRSIs)
We judged five studies to be at low risk of bias in classification of interventions: four studies had a before‐after design, measuring outcomes before and after treatment with antidepressant medications (Hovorka 2000; Kanner 2000, Specchio 2004; Thome‐Souza 2007), and one study allocated participants to antidepressants or CBT according to the feasibility of participants travelling to attend CBT (Orjuela‐Rojas 2015). We judged one study, with three treatment groups receiving different antidepressant medications, to be at moderate risk of bias, as it was unclear exactly how participants were assigned to these groups, and if they were already taking the treatment at recruitment into the study (Kuhn 2003).
Deviations from intended interventions (NSRIs)
All studies aimed to assess the effect of starting and adhering to an intervention, and in all studies, all included participants received an intervention plus any anti‐seizure medications needed to maintain stability during the study. We judged three studies to be at low risk of bias, as all participants were included in the study and analysis (Hovorka 2000; Kanner 2000; Thome‐Souza 2007), and three studies to be at moderate risk of bias, since participants who discontinued the intervention or the study were not included in analysis (Kuhn 2003; Orjuela‐Rojas 2015; Specchio 2004).
Blinding
Blinding of participants, personnel, and outcome assessors (RCTs)
We rated two RCTs at low risk of bias, as study personnel, participants, and outcome assessors were blinded (Li 2005; Robertson 1985). One RCT did not report any clear methods of blinding, therefore, we rated this study at unclear risk of bias (Zhu 2004). One RCT reported an absence of blinding for study personnel or participants due to different treatment interventions (CBT versus sertraline), but did report blinding of study investigators. Therefore, we rated this RCT at low risk of bias (Gilliam 2019).
Measurement of outcomes (NSRIs)
By design, participants were not blinded in any of the six NRSIs, since they completed their own seizure diaries, in an un‐blinded manner. In one study, with three treatment groups receiving different antidepressant medications, a blinded outcome assessor assessed depression; in the other studies, depression was measured before and after the intervention in an un‐blinded manner. As lack of blinding may have influenced some participant‐reported or subjectively assessed outcomes, we judged five studies to be at moderate risk of bias due to measurement of outcomes (Hovorka 2000; Kuhn 2003; Orjuela‐Rojas 2015; Specchio 2004; Thome‐Souza 2007). We judged one study to be at serious risk of bias, because they used an insufficient measure of depression, by looking for 'complete resolution of identified target psychiatric symptoms' as a measure of response to treatment (Kanner 2000).
Incomplete outcome data
For the RCTs, two reported missing data and did not perform an intention‐to‐treat analysis, but they did report both numerator and denominator data (Li 2005; Robertson 1985). We rated these as unclear risk of bias. The third RCT did not report any missing data and carried out an intention‐to‐treat analysis, therefore, we rated this study at low risk of bias (Zhu 2004). The fourth RCT reported missing data, with 15 participants lost to follow‐up, but reported an intention‐to‐treat analysis. We rated this study at low risk of bias (Gilliam 2019).
We judged three NRSIs to be at low risk of bias, as they included all participants in the study; none of the participants discontinued the study (Hovorka 2000; Thome‐Souza 2007), or all participants were included in an intention‐to‐treat analysis (Kanner 2000). We judged three studies to be at serious risk of bias, since participants who discontinued the intervention or the study were not included in the analysis, or data were imputed using the simple method 'last observation carried forward', which assumes no change in outcomes, so this may have introduced bias into the results (Kuhn 2003; Orjuela‐Rojas 2015; Specchio 2004).
Selective reporting
We rated the four RCTs at low risk of bias, as they reported outcomes that were clearly stated in their methods section (Gilliam 2019; Li 2005; Robertson 1985; Zhu 2004).
We judged four NSRIs to be at low risk of bias in their selection of the reported results when summary statistics were presented for all outcomes defined in the methods, and no formal statistical analyses were conducted (Hovorka 2000; Thome‐Souza 2007), or all outcomes and analyses defined in the methods were reported in the results (Kuhn 2003; Specchio 2004). We judged two studies to be at moderate risk of bias, when some participant characteristics were considered in the analyses, but the methods contained no details about which participant characteristics were of interest, and how many characteristics were examined and tested (Kanner 2000; Orjuela‐Rojas 2015).
Other potential sources of bias
For one RCT, seizure data at baseline was collected retrospectively, so we judged this trial to be at high risk of recall bias (Gilliam 2019). For the other RCTs, it was unclear whether there were any other potential sources of bias, so we rated all the RCTs as unclear risk of bias for this domain.
Confounding variables (NSRIs)
We judged four studies to be at serious risk of bias due to confounding, as they conducted no adjustments for any confounding variables (Hovorka 2000; Kuhn 2003; Specchio 2004; Thome‐Souza 2007); and we judged two studies to be at moderate risk of bias due to confounding, as they conducted some analyses to investigate differences in participant subgroups, but they did not consider all the important, prespecified confounders for this review, and it was unclear if the analyses were conducted specifically to investigate confounding (Kanner 2000; Orjuela‐Rojas 2015; see Assessment of risk of bias in included studies).
Effects of interventions
See: Summary of findings 1 Paroxetine compared to doxepin for people with epilepsy and depression; Summary of findings 2 Amitriptyline compared to nomifensine for people with epilepsy and depression; Summary of findings 3 Venlafaxine compared to no treatment for people with epilepsy and depression; Summary of findings 4 Sertraline compared to cognitive behavioural therapy for people with epilepsy and depression; Summary of findings 5 Citalopram (before and after treatment) for people with epilepsy and depression; Summary of findings 6 Selective serotonin reuptake inhibitors compared to cognitive behavioural therapy for people with epilepsy and depression
We evaluated six comparisons in this review:
-
Paroxetine compared to doxepin for people with epilepsy and depression (summary of findings Table 1). We judged the certainty of evidence for this comparison to be moderate; evidence was downgraded due to imprecision, as only one small RCT contributed data.
-
Amitriptyline compared to nomifensine for people with epilepsy and depression (summary of findings Table 2). We judged the certainty of evidence for this comparison to be low; we downgraded the evidence due to serious imprecision, as only one very small RCT contributed data.
-
Venlafaxine compared to no treatment for people with epilepsy and depression (summary of findings Table 3). We judged the certainty of evidence for this comparison to be low; we downgraded the evidence due to risk of bias and imprecision, as only one small RCT contributed data.
-
Sertraline compared to cognitive behavioural therapy (CBT) for people with epilepsy and depression (summary of findings Table 4). We judged the certainty of evidence for this comparison to be moderate to low; we downgraded the evidence due to risk of bias (lack of blinding, and retrospective collection of seizure frequency data at baseline).
-
Citalopram (before and after) for people with epilepsy and depression (summary of findings Table 5). We judged the certainty of evidence for this comparison to be low to very low; we downgraded the evidence from two NRSIs due to serious risk of bias, and substantial statistical heterogeneity was present where data could be pooled.
-
SSRIs (sertraline or citalopram) compared to cognitive behavioural therapy (CBT) for people with epilepsy and depression (summary of findings Table 6). We judged the certainty of evidence for this comparison to be very low; we downgraded the evidence due to serious risk of bias and imprecision, as only one very small NRSI of 15 participants contributed evidence, without any adjustment for confounding variables
Primary outcomes
Depression scores
1. Proportion with a greater than 50% improvement
Three of four RCTs reported on the proportion with a 50% or more improvement in depression scores. The RCTs analysed different treatment comparisons, and we were unable to combine the data in meta‐analysis.
Li 2005 (N = 67) compared paroxetine (20 mg/day to 40 mg/day) to doxepin (mean dose 100 mg/day). There were 27/33 (82%) responders in the paroxetine group and 24/34 (71%) in the doxepin group. The risk ratio (RR) for the proportion with a 50% of more improvement in depression scores for paroxetine versus doxepin was 1.16 (95% confidence interval (CI) 0.88 to 1.52; P > 0.05; Analysis 1.1).
Robertson 1985 (N = 42) compared amitriptyline (75 mg/day), nomifensine (75 mg/day), and placebo. At 12 weeks, there were 6/14 (43%) responders in the amitriptyline group and 11/14 (79%) in the nomifensine group. The risk ratio for the proportion with a 50% or more improvement in depression scores for amitriptyline versus nomifensine was 0.55 (95% CI 0.28 to 1.06; P > 0.05; Analysis 2.1).
Zhu 2004 (N = 64) compared venlafaxine (25 mg/day to 75 mg/day) to no treatment. There were 22/32 (69%) responders in the venlafaxine group and 6/32 (19%) in the no treatment group. More participants had a 50% or more improvement in depression scores in the venlafaxine group than in the no treatment group (RR 3.25, 95% CI 1.19 to 8.90; P < 0.05; Analysis 3.1).
Four of the six NRSIs reported on the proportion of participants with a 50% or more improvement in depression scores. One study did not use a validated depression scale (Kanner 2000).
Hovorka 2000 observed 28/43 (65%) participants with a 50% or more improvement in depression scores following eight weeks of treatment with citalopram (mean dose 23 mg/day) compared to baseline.
Kuhn 2003 observed 17/27 participants (52%) in the mirtazepine group (mean dose 32 mg/day), 12/33 participants (36%) in the citalopram group (mean dose 24 mg/day), and 8/15 participants (53%) in the reboxetine group (mean dose 7 mg/day) with a 50% or more improvement in depression scores following 20 to 30 weeks of treatment, compared to baseline.
Specchio 2004 observed 11/45 participants (24%) with a 50% or more improvement in depression scores following four months of treatment with citalopram (20 mg/day), compared to baseline.
Thome‐Souza 2007 observed 35/36 participants (97%) with a 50% or more improvement in depression scores following one year of treatment with sertraline (mean dose 111 mg/day), or fluoxetine (mean 46 mg/day), compared to baseline.
2. Mean difference
Three of the four RCTS reported on the mean difference in depression scores. Two RCTs used the HAMD scale (Li 2005; Zhu 2004), and one used the BDI‐II (Gilliam 2019). The RCTs compared different treatment groups, and we were unable to combine the data in any meta‐analysis.
Gilliam 2019 (N = 140) compared sertraline (50 mg/day to 200 mg/day) to cognitive behavioural therapy (CBT). The mean difference (MD) in depression scores for sertraline versus CBT was ‐2.50 (95% CI ‐6.28 to 1.28) at 8 weeks, and ‐0.50 (95% CI ‐4.47 to 3.47) at 16 weeks (Analysis 4.1).
Li 2005 (N = 67) compared paroxetine (20 mg/day to 40 mg/day) and doxepin (mean dose 100 mg/day). The mean difference in depression scores following treatment for paroxetine versus doxepin was 0.65 (95% CI ‐2.15 to 3.45; Analysis 1.2).
Zhu 2004 (N = 64) compared venlafaxine (25 mg to 75 mg/day) versus no treatment. Depression scores following treatment with venlafaxine were better than in the no treatment group (MD ‐7.59, 95% CI ‐11.52 to ‐3.66; Analysis 3.2).
Four out of the six NSRIs reported on the mean difference in depression scores; either the mean difference before and after citalopram treatment (Hovorka 2000; Specchio 2004), or the mean difference between treatment groups following treatment (Kuhn 2003; Orjuela‐Rojas 2015).
We were able to meta‐analyse data from two before‐after studies of citalopram. The average dose of citalopram was 22 mg/day in Hovorka 2000, and 20 mg/day in Specchio 2004. The standardised mean difference (SMD) in depression scores indicated a large effect, with an estimate of 1.17 (95% CI 0.96 to 1.38; 2 studies, 88 participants; Analysis 5.1). The I² was 53%, which may reflect differences between the treatment periods. Specchio 2004 reported outcomes following four months of treatment; Hovorka 2000 after two months of treatment. Due to the level of statistical heterogeneity, we conducted a random‐effects analysis, which showed similar results (SMD 1.17; 95% CI 0.86 to 1.47).
Kuhn 2003 reported mean depression scores before and after treatment with mirtazepine, citalopram, and reboxetine. From baseline to 20 to 30 weeks of treatment, the mean depression scores on the HAMD decreased with mirtazepine from 23 to 13.5, citalopram from 22.5 to 14, and reboxetine from 23 to 13.5.
Orjuela‐Rojas 2015 reported mean depression scores before and after treatment with an SSRI (sertraline or citalopram) versus CBT. The mean difference in depression scores for SSRIs versus CBT was ‐2.60 (95% CI ‐11.58 to 6.38) following six weeks of treatment; and ‐4.90 (95% CI ‐14.60 to 4.80; Analysis 6.1) following 12 weeks of treatment.
3. Remission
One of the four RCTs reported the proportion of participants who achieved a remission in depressive symptoms.
Gilliam 2019 reported 38/72 (53%) participants in the sertraline group and 41/68 (60%) in the CBT group who achieved a remission in depressive symptoms, measured on the MINI at 16 weeks of treatment (RR 0.88, 95% CI 0.65 to 1.17; 140 participants; Analysis 4.2).
One of the six NRSIs reported the proportion of participants who achieved a remission in depressive symptoms.
Orjuela‐Rojas 2015 observed 7/8 (87%) in the SSRI group and 4/7 (57%) in the CBT group who achieved a remission in depressive symptoms, measured on the BDI (BDI < 14) at 12 weeks of treatment (RR 1.53, 95% CI 0.77 to 3.06; 15 participants; Analysis 6.2).
Seizure frequency
1. Mean difference
One RCT reported mean difference in seizure frequency between baseline and end of treatment period (Gilliam 2019). One RCT reported no change in seizure frequency in either the paroxetine or the doxepin treatment groups (Li 2005).
Gilliam 2019 reported inconclusive results for a change in the average number of generalised tonic‐clonic seizure (GTCS) per month at eight weeks (MD ‐0.10, 95% CI ‐0.26 to 0.06), and 16 weeks (MD 0.00, 95% CI ‐0.10 to 0.10) of treatment with either sertraline or CBT; and for focal seizures with impaired awareness at eight weeks (MD ‐2.60, 95% CI ‐6.52 to 1.32), and 16 weeks (MD ‐3.00, 95% CI ‐7.81 to 1.81; Analysis 4.3) of treatment with either sertraline or CBT.
Four of the six NRSIs reported on changes in mean seizure frequency; either the mean difference before and after citalopram treatment (Hovorka 2000; Specchio 2004), or the mean difference between treatment groups following treatment (Kuhn 2003; Orjuela‐Rojas 2015).
We conducted a meta‐analysis with data from the two before‐after studies of citalopram (Hovorka 2000; Specchio 2004). Calculating SMD, and using both a fixed‐effect and random‐effects model, the I² was 81%. Therefore, we did not report the pooled effect estimate. Individually, Hovorka 2000 showed inconclusive results between groups (MD 0.03, 95% CI ‐0.23 to 0.30; 43 participants), while Specchio 2004 showed possible improvement after treatment (MD 0.50, 95% CI 0.21 to 0.78; 45 participants; Analysis 5.2). Possible reasons for a high I² value are differences in mean baseline seizure frequency, treatment duration, and mean age of participants.
Kanner 2000 reported no statistically significant difference in seizure frequency between baseline and post‐treatment for the 100 participants treated with sertraline (mean dose 108 mg/day).
The results for a change in seizure frequency between baseline and 12 weeks of treatment, between the eight participants receiving an SSRI (sertraline or citalopram) and the seven participants receiving CBT, were inconclusive (MD ‐1.60, 95% CI ‐5.63 to 2.43; 15 participants; Analysis 6.3; Orjuela‐Rojas 2015).
2. Seizure recurrence
One RCT reported on the recurrent of GTCS during the study in participants who had not experienced a GTCS in the six months prior to enrolment (Gilliam 2019). It reported a recurrence in 4/51 people in the sertraline group versus 4/53 in the CBT (RR 1.04, 95% CI 0.27 to 3.94; 104 participants; Analysis 4.4).
One NRSI reported that one participant experienced a recurrence of seizures (Specchio 2004).
3. Episode of status epilepticus
None of the 10 studies reported episodes of status epilepticus.
Secondary outcomes
Withdrawals for any reason
Three of the RCTs clearly reported the number of participants who withdrew for any reason (Gilliam 2019; Li 2005; Robertson 1985).
Gilliam 2019 reported 7/72 withdrawals in the sertraline group (three withdrew consent; four left the study due to worsening depression), and 6/68 in the CBT group (two withdrew consent; four left the study due to worsening depression). There were also 9/72 participants lost to follow‐up from the sertraline group, and 6/68 from the CBT group (RR 1.26, 95% CI 0.64 to 2.46; 140 participants; Analysis 4.5).
Li 2005 reported that 3/34 participants (9%) withdrew from the doxepin group versus 0/33 from the paroxetine group (RR 0.15, 95% CI 0.01 to 2.74; P = 0.20; 67 participants; Analysis 1.3). The specific reasons for withdrawal were not reported.
Robertson 1985 reported that 1/14 participants (7%) withdrew from the amitriptyline group, 1/14 from the nomifensine group, and 1/14 from the placebo group. The participant withdrew from the nomifensine group because of increased seizures; the other reasons were not reported.
All six prospective non‐randomised studies reported on the number of participants withdrawing from the study.
Hovorka 2000 reported no treatment withdrawals from the study.
Kanner 2000 reported that 18/100 participants (18%) withdrew due to adverse events from the sertraline.
Kuhn 2003 reported that 20/27 participants (74%) withdrew from the mirtazepine group, 16/33 participants (48%) from the citalopram group, and 6/15 participants (40%) from the reboxetine group withdrew after 20 to 30 weeks of treatment. Adverse events accounted for 8/20 withdrawals from the mirtazepine group, 6/16 from the citalopram group, and 3/6 from the reboxetine group. Three participants (two from mirtazepine and one from citalopram) withdrew due to inefficacy. The remaining 22 participants were lost to follow‐up.
Orjuela‐Rojas 2015 reported that 2/7 withdrew from the CBT group (1 due to health problems unrelated to epilepsy, and 1 due to severe psychosocial situation deemed unrelated to either epilepsy or depression). There were no withdrawals from the SSRI group, but one was lost to follow‐up due to a road traffic accident. There were inconclusive results between groups (RR 0.44, 95% CI 0.05 to 3.85; 15 participants; Analysis 6.4).
Specchio 2004 reported that 6/45 participants (13%) withdrew from the study (four because of adverse events from the citalopram, one due to poor compliance, and one due to concurrent illness).
Thome‐Souza 2007 reported that 1/36 participant, treated with sertraline (SSRI) withdrew because of an exacerbation of seizures.
Global State
None of the included studies reported on global state outcomes.
Mental State
None of the included studies reported on mental state outcomes.
General Functioning
None of the included studies reported on general functioning outcomes.
Cognitive Functioning
None of the included studies reported on cognitive functioning outcomes.
Quality of life
One of the four RCTs assessed quality of life, using the 89‐item Quality of Life in Epilepsy Inventory (QOLIE‐89; (Gilliam 2019)).
There were inconclusive results in improved quality of life between the sertraline and CBT groups at 8 weeks (MD 6.10, 95% CI ‐0.28 to 12.48; 104 participants), and at 16 weeks (MD 3.10, 95% CI ‐3.41 to 9.61; 118 participants Analysis 4.6).
One of the six NRSIs assessed quality of life with the QOLIE‐31 (Orjuela‐Rojas 2015).
There were inconclusive results between groups (MD ‐0.50, 95% CI ‐19.67 to 18.67; 15 participants; Analysis 6.5)
Behaviour
None of the included studies reported on behaviour outcomes.
Adverse effects
Two RCTs reported the number of participants who experienced specific side effects (Gilliam 2019; Li 2005).
Li 2005 compared paroxetine and doxepin.
The results were inconclusive between groups for blurred vision (RR 0.34, 99% CI 0.09 to 1.32; 67 participants), dizziness (RR 0.21, 99% CI 0.03 to 1.37; 67 participants), dry mouth (RR 0.26, 99% CI 0.06 to 1.20; 67 participants), sleep disorders (RR 0.32, 99% CI 0.08 to 1.20; 67 participants), and urinary retention (RR 0.34, 99% CI 0.01 to 21.99; 67 participants; Analysis 1.4).
Gilliam 2019 compared sertraline and CBT.
There were 10 serious adverse events in the sertraline group, three of which were possibly related to the study (one mania and psychosis, two worsening depression requiring hospitalisation). There were 12 serious adverse events in the CBT group, three of which were possibly related to the study (worsening depression and suicidal). One participant in the CBT group died of sudden unexpected death in epilepsy (SUDEP).
The difference in adverse events profile (AEP) scores was inconclusive between the groups at 8 weeks (MD ‐2.70, 95% CI ‐6.62 to 1.22; 106 participants), and 16 weeks (MD ‐2.10, 95% CI ‐6.21 to 2.01; 118 participants; Analysis 4.7).
There were more cases of tiredness reported by the participants on sertraline compared to those receiving CBT (RR 3.54, 99% CI 1.40 to 8.96; 140 participants; Analysis 4.8)
Results were inconclusive between the groups for the other five most commonly reported adverse events: headache (RR 1.48, 99% CI 0.69 to 3.19), insomnia (RR 2.16, 99% CI 0.73 to 6.38), shakiness (RR 12.28, 99% CI 0.88 to 171.59), nausea (RR 2.60, 99% CI 0.62 to 10.96), and diarrhoea (RR 19.85, 99% CI 0.49 to 805.53; 140 participants; Analysis 4.8).
Five of six NRSIs reported on adverse events. The studies analysed different treatment comparisons, so we were unable to combine the data in meta‐analysis.
Hovorka 2000 reported that 3/43 participants (7%) experienced nausea, and 2/43 (5%) experienced sexual dysfunction following eight weeks of treatment with citalopram.
Kanner 2000 reported that 9/100 participants (9%) experienced sedation, 7/100 (7%) experienced hypomanic symptoms, 1/100 (1%) experienced rheumatic pain, and 1/100 participants (1%) experienced myoclonus following an average of 10 months on sertraline.
Kuhn 2003 reported that 13/75 participants experienced side effects. The most common were; weight gain (5/75), sedation (2/75), and sexual dysfunction (2/75).
Specchio 2004 reported that 22/45 participants (56%) experienced side effects. The most common were; headache (15%), nausea (11%), dizziness (9%), drowsiness (7%), and fatigue (7%).
Thome‐Souza 2007 reported that 1/36 participants (3%) experienced facial rash, and 1/36 (3%) experienced gastrointestinal disorders.
Discussion
Summary of main results
We identified four randomised controlled trials (RCTs; N = 313) of antidepressant treatment for people with epilepsy, three of which had small sample sizes (42 to 67 participants). One trial compared antidepressant medication with cognitive behavioural therapy (CBT). Venlafaxine improved depressive symptoms when compared to no treatment; amitriptyline was not better than nomifensine in improving depressive symptoms; depressive symptoms were improved with both sertraline or CBT, but one was not better than the other. The RCTs did not allow meaningful comparisons among the different classes of antidepressants. Therefore, we do not know which antidepressant or class of antidepressant is most effective.
Only one RCT reported on changes in seizure frequency. The results were inconclusive for a change in frequency of generalised tonic clonic seizures for sertraline or CBT. Only one RCT reported on adverse events. The results were inconclusive between groups for adverse events including suicidal risk except tiredness which occurred more often in the sertraline group. The top reported side effects were; sedation, headache, insomnia and gastrointestinal symptoms.
The six prospective cohort studies (N = 313) were of serious risk of bias examining small numbers of participants treated predominantly with selective serotonin reuptake inhibitors (SSRIs). One study compared SSRIs to CBT. The combined meta‐analysis of two studies examining citalopram at 20 mg/day showed that citalopram improved depressive symptoms. We could not combine the data on changes in seizure frequency due to marked heterogeneity. However, in the three studies reporting changes in seizure frequency, the results were inconclusive. Only one study reported on one participant on antidepressant treatment (citalopram) who experienced seizure recurrence. Whilst the data are of low certainty in terms of impact on seizures, there is broad agreement across the prospective cohort studies of limited or no impact on seizures with selective serotonin reuptake inhibitors (SSRIs).
Participants withdrawing from antidepressants were more likely to do so because of adverse events rather than lack of efficacy. Reported adverse events for SSRIs included nausea, dizziness, sedation, gastrointestinal disturbance, and sexual dysfunction. We have no reliable information on the comparative risk of adverse events with different classes of antidepressant treatment.
Overall completeness and applicability of evidence
This review has ascertained that there is very limited evidence of an effect of antidepressants on depressive symptoms in participants with epilepsy. Depressive symptoms were improved with both sertraline or CBT, but one was not better than the other. We do not have any reliable high certainty evidence to inform on the best choice of antidepressant drug or class of drug for treating depression with the lowest risk of seizure exacerbation.
Quality of the evidence
Overall, we rated one study at low risk of bias (Gilliam 2019), five prospective cohort studies at high risk of bias (Hovorka 2000; Kanner 2000; Kuhn 2003; Orjuela‐Rojas 2015; Thome‐Souza 2007) and four studies (three RCTs and one prospective cohort study) at unclear risk of bias (Li 2005; Robertson 1985; Specchio 2004; Zhu 2004).
The 'Summary of findings' tables for each comparison examined shows that the certainty of the evidence for the outcomes ranged from moderate certainty to low certainty. For comparisons where RCTs were available, we downgraded for imprecision as for these comparisons, only one small study contributed to the outcomes. For comparisons where only NRSIs were available, we rated the certainty of evidence as low to very low and evidence was downgraded due to risk of bias, imprecision and substantial heterogeneity for some outcomes.
Potential biases in the review process
It is possible that despite the exhaustive searches carried out in this review, we did not identify other sources of unpublished data. This can be more of an issue for reviews including non‐randomised study designs, such as this review.
Including non‐randomised evidence, which is inherently at risk of additional biases to RCT evidences, may have also introduced bias into the review. However, as stated in Types of studies, prospective non‐randomised studies may be more reflective of clinical practice in terms of the populations recruited, and because of the known delayed effect of antidepressants on depressive symptoms, which may not be effectively detected in short‐term randomised trials.
We carried out a detailed quality assessment of the non‐randomised studies included in this review, using an appropriate tool (ROBINS‐I), and interpreted evidence from the non‐randomised studies taking into account the likely biases present within these studies.
Agreements and disagreements with other studies or reviews
We are not aware of any other reviews on this topic.
Study flow diagram
Comparison 1: RCT: paroxetine versus doxepin, Outcome 1: > 50% reduction in depressive symptoms
Comparison 1: RCT: paroxetine versus doxepin, Outcome 2: Mean depression scores
Comparison 1: RCT: paroxetine versus doxepin, Outcome 3: Withdrawals (any reason)
Comparison 1: RCT: paroxetine versus doxepin, Outcome 4: Adverse effects
Comparison 2: RCT: amitriptyline versus nomifensine, Outcome 1: > 50% reduction in depressive symptoms
Comparison 3: RCT: venlafaxine versus no treatment controls, Outcome 1: > 50% reduction in depressive symptoms
Comparison 3: RCT: venlafaxine versus no treatment controls, Outcome 2: Mean depression scores ‐ HAMD
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 1: Mean depression scores (BDI)
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 2: Remission in depressive symptoms
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 3: Seizure frequency
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 4: Seizure recurrence
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 5: Withdrawals (any reason)
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 6: Quality of life (QOLIE‐89)
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 7: Adverse events profile
Comparison 4: RCT: sertraline versus cognitive behavioural therapy (CBT), Outcome 8: Adverse events
Comparison 5: NRSI: citalopram (before and after), Outcome 1: Mean depression scores HAMD‐21
Comparison 5: NRSI: citalopram (before and after), Outcome 2: Mean monthly seizure frequency
Comparison 6: NRSI: SSRIs (sertraline or citalopram) versus CBT, Outcome 1: Mean depression scores (BDI)
Comparison 6: NRSI: SSRIs (sertraline or citalopram) versus CBT, Outcome 2: Remission in depressive symptoms
Comparison 6: NRSI: SSRIs (sertraline or citalopram) versus CBT, Outcome 3: Seizure frequency per month at 12 weeks
Comparison 6: NRSI: SSRIs (sertraline or citalopram) versus CBT, Outcome 4: Withdrawals (any reason)
Comparison 6: NRSI: SSRIs (sertraline or citalopram) versus CBT, Outcome 5: Quality of life (QOLIE‐31 overall score)
| Paroxetine compared to doxepin for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| doxepin | paroxetine | |||||
| > 50% reduction in depressive symptoms Follow‐up: 8 weeks | 706 per 1000 | 819 per 1000 | RR 1.16 | 67 | ⊕⊕⊕⊝ |
|
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks | NA | The mean HAMD depression score in the intervention groups was | NA | 67 | ⊕⊕⊕⊝ | |
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Withdrawals Follow‐up: 8 weeks | 88 per 1000 | 13 per 1000 | RR 0.15 | 67 | ⊕⊕⊕⊝ moderatea | doxepin: 3 withdrew paroxetine: 0 withdrew |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: 8 weeks | Reported adverse events: blurred vision, dizziness, dry mouth, sleep disorders, and urinary retention | Reported adverse events: blurred vision, dizziness, dry mouth, and sleep disorders | NA | 67 (1 RCT) | ⊕⊕⊕⊝ moderatea | There were no significant differences between treatment groups for any reported adverse events |
| *The basis for the assumed risk is the event rate in the doxepin group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded for imprecision, because only one small study contributed to the outcomes. | ||||||
| Amitriptyline compared to nomifensine for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| nomifensine | amitriptyline | |||||
| > 50% reduction in depressive symptoms Follow‐up: 12 weeks | 786 per 1000 | 432 per 1000 | RR 0.55 | 28 | ⊕⊕⊝⊝ |
|
| Mean depression scores Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Withdrawals Follow‐up: NA | ‐ | ‐ | ‐ | 0 | ‐ |
|
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the nomifensine group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded twice for imprecision, because only very small study contributed limited outcome data. | ||||||
| Venlafaxine compared to no treatment for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| no treatment | venlafaxine | |||||
| > 50% reduction in depressive symptoms Follow‐up: 8 weeks | 125 per 1000 | 406 per 1000 | RR 3.25 | 64 | ⊕⊕⊝⊝ lowa,b |
|
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks | NA | The mean HAMD depression score in the intervention group was | NA | 64 | ⊕⊕⊝⊝ lowa,b |
|
| Seizure frequency Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Withdrawals Follow‐up: NA | ‐ | ‐ | ‐ | 0 | ‐ |
|
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the no treatment group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded for imprecision, because only one small study contributed to the outcomes. | ||||||
| Sertraline compared to cognitive behavioural therapy for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| CBT | sertraline | |||||
| > 50% reduction in depressive symptoms Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Mean depression scores (BDI scores; lower = better) Follow‐up: 16 weeks | NA | The mean BDI depression score in the intervention group was 0.50 lower (4.47 lower to 3.47 higher) | NA | 117 | ⊕⊕⊕⊝ moderatea | At 8 weeks: MD ‐2.50 (95% CI ‐6.28 to 1.28; 104 participants) |
| Seizure frequency Follow‐up: 16 weeks | NA | The mean frequency of GTCS per month in the intervention group was 0 lower (‐0.10 lower to 0.10 higher) The mean frequency of focal seizures with impaired awareness per month in the intervention group was 3.00 lower (7.81 lower to 1.81 higher) | NA | 96 with GTCS plus 75 with focal seizures (1 RCT) | ⊕⊕⊝⊝ lowb | At 8 weeks: GTCS per month: MD ‐0.10 (95% CI ‐0.26 to 0.06; 86 participants) focal seizures with impaired awareness per month: MD ‐2.60 (95% CI ‐6.52 to 1.32; 75 participants) |
| Withdrawals Follow‐up: 16 weeks | 176 per 1000 | 222 per 1000 (113 to 434 per 1000) | RR 1.26 (0.64 to 2.46) | 140 | ⊕⊕⊕⊝ moderatea | CBT: 6 withdrew, 6 lost to follow‐up sertraline: 7 withdrew, 9 lost to follow‐up |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Quality of life (QOLIE‐89 scale; lower = better) Follow‐up: 16 weeks | NA | The mean QOLIE‐89 score in the intervention group was 3.10 higher (3.41 lower to 9.61 higher) | NA | 118 (1 RCT) | ⊕⊕⊕⊝ moderatea | at 8 weeks: MD 6.10 (95% CI ‐0.28 to 12.48; 104 participants) |
| Adverse effects Follow‐up: 16 weeks | NA | The mean adverse event profile score in the intervention group was 2.10 lower (6.21 lower to 2.01 higher) | NA | 118 (1 RCT) | ⊕⊕⊕⊝ lowa,c | Sertraline resulted in more cases of tiredness than CBT (RR 3.54, 99% CI 1.40 to 8.96; 140 participants) Sertraline did not result in more cases of any other adverse effects than CBT. |
| *The basis for the assumed risk is the event rate in the CBT group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| aCertainty of the evidence downgraded once due to risk of bias: participants and personnel not blinded, and lack of blinding may have influenced outcome cCertainty of the evidence downgraded once due to imprecision: adverse event data not available for all participants who received an intervention | ||||||
| Citalopram (before and after treatment) for people with epilepsy and depression | ||||
|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of participants | Certainty of the evidence | Comments |
| Citalopram (before and after) | ||||
| > 50% reduction in depressive symptoms Follow‐up: 4 months | 11 out of 45 participants (24%) showed a 50% or more improvement in depression scores after treatment compared to baseline. | 45 (1 NRSI) | ⊕⊕⊝⊝ lowa | |
| Mean depression scores (HAMD scores; lower = better) Follow‐up: 8 weeks to 4 months | Improved depression scores were shown after citalopram compared to before (see comment) | 88 | ⊕⊕⊝⊝ | SMD in HAMD score was 1.17 (95% CI 0.96 to 1.38), indicating improved outcomes and a large treatment effect. |
| Seizure frequency Follow‐up: 8 weeks to 4 months | See comment | 88 | ⊕⊝⊝⊝ | Results were mixed between studies; due to very high heterogeneity (I² = 81%), we did not present the overall effect estimate. |
| Withdrawals Follow‐up: 8 weeks to 4 months | 6/45 participants (13%) withdrew from one study; 0/43 from the other study | 88 (2 NRSI) | ⊕⊕⊝⊝ lowa | |
| Cognitive functioning Follow‐up: NA | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life Follow‐up: NA | ‐ | 0 (0 studies) | ‐ |
|
| Adverse effects Follow‐up: 8 weeks to 4 months | 22/45 participants (56%) experienced adverse events in one study; 5/43 (12%) in the other study | 88 (2 NRSI) | ⊕⊕⊝⊝ lowa | Specific adverse events reported: nausea, sexual dysfunction, headache, dizziness, drowsiness, and fatigue |
| CI: confidence interval; HAMD: Hamilton Rating Scale for Depression; NRSI: non‐randomised studies of interventions | ||||
| GRADE Working Group grades of evidence | ||||
| aCertainty of the evidence downgraded twice as studies were judged to be at serious risk of bias due to lack of blinding, which may have influenced participant‐recorded outcomes, and lack of adjustment for confounding variables. | ||||
| Selective serotonin reuptake inhibitorscompared to cognitive behavioural therapy for people with epilepsy and depression | ||||||
|---|---|---|---|---|---|---|
| Patient or population: people with epilepsy and depression | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | No of participants | Certainty of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| CBT | SSRIs | |||||
| > 50% reduction in depressive symptoms Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ | |
| Mean depression scores (BDI scores; lower = better) Follow‐up: 12 weeks | NA | The mean BDI depression score in the intervention group was 4.90 lower (14.90 lower to 4.80 higher) | NA | 15 | ⊕⊝⊝⊝ very lowa,b | at 6 weeks: MD ‐2.60 (95% CI ‐11.58 to 6.38; 15 participants) |
| Seizure frequency Follow‐up: 12 weeks | NA | The mean frequency of seizures per month in the intervention group was 1.60 lower (5.63 lower to 2.43 higher) | NA | 15 (1 NRSI) | ⊕⊝⊝⊝ very lowa,b | |
| Withdrawals Follow‐up: 12 weeks | 286 per 1000 | 126 per 1000 (14 to 1000 per 1000) | RR 0.44 (0.05 to 3.85) | 15 (1 NRSI) | ⊕⊝⊝⊝ very lowa,b | CBT: 2 lost to follow‐up SSRI: 1 lost to follow‐up in the |
| Cognitive functioning Follow‐up: NA | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| Quality of life ‐ QOLIE‐31 scale Follow‐up: 12 weeks | NA | The mean QOLIE‐31 score in the intervention group was 0.50 lower (19.67 lower to 18.67 higher) | NA | 15 (1 NRSI) | ⊕⊝⊝⊝1,2 very low |
|
| Adverse effects ‐ adverse event profile Follow‐up: 16 weeks | ‐ | ‐ | ‐ | 0 (0 studies) | ‐ |
|
| *The basis for the assumed risk is the event rate in the CBT group. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| 1. Certainty of the evidence downgraded twice as the study was judged to be at serious risk of bias with regards to lack of blinding which may have influenced participant recorded outcomes and lack of adjustment for confounding variables. 2. Certainty of the evidence downgraded once due to imprecision: very small study of 15 participants, confidence intervals around effect estimates wide | ||||||
| Risk of bias judgement | Criteria based on seven risk of bias domains |
|---|---|
| Low risk of bias: the study is comparable to a well‐performed randomised trial | The study is judged to be at low risk of bias for all domains |
| Moderate risk of bias: the study appears to provide sound evidence for a non‐randomised study but cannot be considered comparable to a well‐performed randomised trial | The study is judged to be at low or moderate risk of bias for all domains |
| Serious risk of bias: the study has some important problems | The study is judged to be at serious risk of bias in at least 1 domain, but not at critical risk of bias in any domain |
| Critical risk of bias: the study is too problematic to provide any useful evidence on the effects of intervention | The study is judged to be at critical risk of bias in at least 1 domain |
| No information on which to base a judgement about risk of bias | There is no clear indication that the study is at serious or critical risk of bias, and there is a lack of information in 1 or more key domains of bias (a judgement is required for this) |
| Domain and risk of bias judgement | Study | |||||
|---|---|---|---|---|---|---|
| Bias due to confounding | Serious | Moderate | Serious | Moderate | Serious | Serious |
| Bias in selection of participants into the study | Moderate | Low | Moderate | Low | Serious | Moderate |
| Bias in classification of interventions | Low | Low | Moderate | Low | Low | Low |
| Bias due to deviations from intended interventions | Low | Low | Moderate | Moderate | Moderate | Low |
| Bias due to missing data | Low | Low | Serious | Serious | Serious | Low |
| Bias in measurement of outcomes | Moderate | Serious | Moderate | Moderate | Moderate | Moderate |
| Bias in selection of the reported result | Low | Moderate | Low | Moderate | Low | Low |
| Overall judgement | Serious | Serious | Serious | Serious | Serious | Serious |
| Support for judgement | No adjustment for confounding; unclear if participants were recruited consecutively; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes | Some analyses to investigate prognostic variables, but not a complete analysis of confounders; unclear which variables were of interest in advance, and if other characteristics were tested and analysed; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes. The measure of depression was not an accurate or reliable measure | No adjustment for confounding; many discontinuations due to adverse events and non‐compliance, with outcome data analysed by last observation carried forward; unclear if participants were already receiving the intervention on entry into the study, and exactly how groups were assigned. Lack of blinding may have influenced some participant‐reported outcomes. | Some analyses to investigate prognostic variables, but not a complete analysis of confounders; unclear which variables were of interest in advance, and if other characteristics were tested and analysed; small groups and dropouts, with outcome data analysed by last observation carried forward, Lack of blinding may have influenced some participant‐reported outcomes. | No adjustment for confounding; outcome data included only for those who completed analysis (6 participants excluded); and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes | No adjustment for confounding; unclear if participants were recruited consecutively; and by design, blinding was not possible, which may have influenced subjectively‐assessed outcomes. |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1.1 > 50% reduction in depressive symptoms Show forest plot | 1 | 67 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.16 [0.88, 1.52] |
| 1.2 Mean depression scores Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 1.2.1 HAMD scores | 1 | 67 | Mean Difference (IV, Fixed, 95% CI) | 0.65 [‐2.15, 3.45] |
| 1.3 Withdrawals (any reason) Show forest plot | 1 | 67 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.15 [0.01, 2.74] |
| 1.4 Adverse effects Show forest plot | 1 | Risk Ratio (M‐H, Fixed, 99% CI) | Subtotals only | |
| 1.4.1 Blurred vision | 1 | 67 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.34 [0.09, 1.32] |
| 1.4.2 Dizziness | 1 | 67 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.21 [0.03, 1.37] |
| 1.4.3 Dry mouth | 1 | 67 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.26 [0.06, 1.20] |
| 1.4.4 Sleep disorders | 1 | 67 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.32 [0.08, 1.20] |
| 1.4.5 Urinary retention | 1 | 67 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.34 [0.01, 21.99] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 2.1 > 50% reduction in depressive symptoms Show forest plot | 1 | 28 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.55 [0.28, 1.06] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 3.1 > 50% reduction in depressive symptoms Show forest plot | 1 | 64 | Risk Ratio (M‐H, Fixed, 95% CI) | 3.25 [1.19, 8.90] |
| 3.2 Mean depression scores ‐ HAMD Show forest plot | 1 | 64 | Mean Difference (IV, Fixed, 95% CI) | ‐7.59 [‐11.52, ‐3.66] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 4.1 Mean depression scores (BDI) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 4.1.1 BDI‐II scores at 8 weeks | 1 | 104 | Mean Difference (IV, Fixed, 95% CI) | ‐2.50 [‐6.28, 1.28] |
| 4.1.2 BDI‐II scores at 16 weeks | 1 | 117 | Mean Difference (IV, Fixed, 95% CI) | ‐0.50 [‐4.47, 3.47] |
| 4.2 Remission in depressive symptoms Show forest plot | 1 | 140 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.88 [0.65, 1.17] |
| 4.3 Seizure frequency Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 4.3.1 Generalised tonic‐clonic seizures per month at 8 weeks | 1 | 86 | Mean Difference (IV, Fixed, 95% CI) | ‐0.10 [‐0.26, 0.06] |
| 4.3.2 Generalised tonic‐clonic seizures per month at 16 weeks | 1 | 96 | Mean Difference (IV, Fixed, 95% CI) | 0.00 [‐0.10, 0.10] |
| 4.3.3 Focal seizures with impaired awareness per month at 8 weeks | 1 | 75 | Mean Difference (IV, Fixed, 95% CI) | ‐2.60 [‐6.52, 1.32] |
| 4.3.4 Focal seizures with impaired awareness per month at 16 weeks | 1 | 73 | Mean Difference (IV, Fixed, 95% CI) | ‐3.00 [‐7.81, 1.81] |
| 4.4 Seizure recurrence Show forest plot | 1 | 104 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.04 [0.27, 3.94] |
| 4.5 Withdrawals (any reason) Show forest plot | 1 | 140 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.26 [0.64, 2.46] |
| 4.6 Quality of life (QOLIE‐89) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 4.6.1 QOLIE‐89 score at 8 weeks | 1 | 104 | Mean Difference (IV, Fixed, 95% CI) | 6.10 [‐0.28, 12.48] |
| 4.6.2 QOLIE‐89 score at 16 weeks | 1 | 118 | Mean Difference (IV, Fixed, 95% CI) | 3.10 [‐3.41, 9.61] |
| 4.7 Adverse events profile Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 4.7.1 Adverse event profile at 8 weeks | 1 | 106 | Mean Difference (IV, Fixed, 95% CI) | ‐2.70 [‐6.62, 1.22] |
| 4.7.2 Adverse event profile at 16 weeks | 1 | 118 | Mean Difference (IV, Fixed, 95% CI) | ‐2.10 [‐6.21, 2.01] |
| 4.8 Adverse events Show forest plot | 1 | Risk Ratio (M‐H, Fixed, 99% CI) | Subtotals only | |
| 4.8.1 Anxiety | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 8.51 [0.19, 386.16] |
| 4.8.2 Chest pain | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.47 [0.05, 4.21] |
| 4.8.3 Cold | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.63 [0.13, 3.13] |
| 4.8.4 Diarrhoea | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 19.85 [0.49, 805.53] |
| 4.8.5 Dizziness | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 1.51 [0.37, 6.14] |
| 4.8.6 Dry mouth | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 3.78 [0.22, 65.10] |
| 4.8.7 Headache | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 1.48 [0.69, 3.19] |
| 4.8.8 Insomnia | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 2.16 [0.73, 6.38] |
| 4.8.9 Memory difficulty | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 0.71 [0.10, 4.82] |
| 4.8.10 Muscle strain or pain | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 1.21 [0.36, 4.12] |
| 4.8.11 Nausea | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 2.60 [0.62, 10.96] |
| 4.8.12 Rash | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 4.72 [0.29, 76.72] |
| 4.8.13 Sexual dysfunction | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 8.51 [0.19, 386.16] |
| 4.8.14 Shakiness | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 12.28 [0.88, 171.59] |
| 4.8.15 Tiredness | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 3.54 [1.40, 8.96] |
| 4.8.16 Unsteadiness | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 1.57 [0.25, 9.81] |
| 4.8.17 Worsening depression | 1 | 140 | Risk Ratio (M‐H, Fixed, 99% CI) | 1.13 [0.25, 5.07] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 5.1 Mean depression scores HAMD‐21 Show forest plot | 2 | 176 | Std. Mean Difference (IV, Fixed, 95% CI) | 1.17 [0.96, 1.38] |
| 5.2 Mean monthly seizure frequency Show forest plot | 2 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 6.1 Mean depression scores (BDI) Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 6.1.1 BDI at 6 weeks | 1 | 15 | Mean Difference (IV, Fixed, 95% CI) | ‐2.60 [‐11.58, 6.38] |
| 6.1.2 BDI at 12 weeks | 1 | 15 | Mean Difference (IV, Fixed, 95% CI) | ‐4.90 [‐14.60, 4.80] |
| 6.2 Remission in depressive symptoms Show forest plot | 1 | 15 | Risk Ratio (M‐H, Fixed, 95% CI) | 1.53 [0.77, 3.06] |
| 6.3 Seizure frequency per month at 12 weeks Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | ‐1.60 [‐5.63, 2.43] | |
| 6.4 Withdrawals (any reason) Show forest plot | 1 | 15 | Risk Ratio (M‐H, Fixed, 95% CI) | 0.44 [0.05, 3.85] |
| 6.5 Quality of life (QOLIE‐31 overall score) Show forest plot | 1 | 15 | Mean Difference (IV, Fixed, 95% CI) | ‐0.50 [‐19.67, 18.67] |
