Elsevier

The Lancet Neurology

Volume 16, Issue 7, July 2017, Pages 523-531
The Lancet Neurology

Articles
Individualised prediction model of seizure recurrence and long-term outcomes after withdrawal of antiepileptic drugs in seizure-free patients: a systematic review and individual participant data meta-analysis

https://doi.org/10.1016/S1474-4422(17)30114-XGet rights and content

Summary

Background

People with epilepsy who became seizure-free while taking antiepileptic drugs might consider discontinuing their medication, with the possibility of increased quality of life because of the elimination of adverse events. The risk with this action, however, is seizure recurrence. The objectives of our study were to identify predictors of seizure recurrence and long-term seizure outcomes and to produce nomograms for estimation of individualised outcomes.

Methods

We did a systematic review and meta-analysis, and identified eligible articles and candidate predictors, using PubMed and Embase databases with a last update on Nov 6, 2014. Eligible articles had to report on cohorts of patients with epilepsy who were seizure-free and had started withdrawal of antiepileptic drugs; articles also had to contain information regarding seizure recurrences during and after withdrawal. We excluded surgical cohorts, reports with fewer than 30 patients, and reports on acute symptomatic seizures because these topics were beyond the scope of our objective. Risk of bias was assessed using the Quality in Prognosis Studies system. Data analysis was based on individual participant data. Survival curves and proportional hazards were computed. The strongest predictors were selected with backward selection. Models were converted to nomograms and a web-based tool to determine individual risks.

Findings

We identified 45 studies with 7082 patients; ten studies (22%) with 1769 patients (25%) were included in the meta-analysis. Median follow-up was 5·3 years (IQR 3·0–10·0, maximum 23 years). Prospective and retrospective studies and randomised controlled trials were included, covering non-selected and selected populations of both children and adults. Relapse occurred in 812 (46%) of 1769 patients; 136 (9%) of 1455 for whom data were available had seizures in their last year of follow-up, suggesting enduring seizure control was not regained by this timepoint. Independent predictors of seizure recurrence were epilepsy duration before remission, seizure-free interval before antiepileptic drug withdrawal, age at onset of epilepsy, history of febrile seizures, number of seizures before remission, absence of a self-limiting epilepsy syndrome, developmental delay, and epileptiform abnormality on electroencephalogram (EEG) before withdrawal. Independent predictors of seizures in the last year of follow-up were epilepsy duration before remission, seizure-free interval before antiepileptic drug withdrawal, number of antiepileptic drugs before withdrawal, female sex, family history of epilepsy, number of seizures before remission, focal seizures, and epileptiform abnormality on EEG before withdrawal. Adjusted concordance statistics were 0·65 (95% CI 0·65–0·66) for predicting seizure recurrence and 0·71 (0·70–0·71) for predicting long-term seizure freedom. Validation was stable across the individual study populations.

Interpretation

We present evidence-based nomograms with robust performance across populations of children and adults. The nomograms facilitate prediction of outcomes following drug withdrawal for the individual patient, including both the risk of relapse and the chance of long-term freedom from seizures. The main limitations were the absence of a control group continuing antiepileptic drug treatment and a consistent definition of long-term seizure freedom.

Funding

Epilepsiefonds.

Introduction

Antiepileptic drugs suppress seizures in 65% to 85% of people with epilepsy.1 Because of the fear of seizure relapse many people with epilepsy continue antiepileptic drug treatment even when free from seizures and despite the side-effects of the drugs. Up to 88% of patients often have several adverse effects from antiepileptic drugs.2, 3 As a result, quality of life for seizure-free patients is significantly better when antiepileptic drugs are discontinued,4 provided they remain seizure free.

Results from a meta-analysis estimated that the cumulative seizure recurrence rate after antiepileptic drug withdrawal is around 34%.5 For those who have seizure recurrence, about 80% will be able to control their seizures by reinstating antiepileptic drug treatment.6 The remaining 20% will develop treatment-refractory epilepsy, although there is no convincing evidence that this refractoriness occurs as a consequence of antiepileptic drug withdrawal. Nonetheless, there is some debate around whether antiepileptic drug withdrawal is safe at all.7, 8

Research in context

Evidence before this study

We did a systematic review of the English-language scientific literature in PubMed and Embase published up to Nov 6, 2014, using the search terms “antiepileptic”, “withdrawal”, “recurrence”, and “seizure-free”, and their synonyms. Overall risk of bias for separate studies was low for study participation, study attrition, prognostic factor measurement, and outcome measurement. 25 variables were identified as significant predictors of seizure recurrence in at least one peer-reviewed article. However, differences in study design, population, and methods limited the possibility to determine which were the strongest predictors, and how to combine those predictors to identify risks for the individual patient.

Added value of this study

This individual participant data meta-analysis of information from 1769 patients identified independent predictors of seizure relapse and eventual seizure freedom after antiepileptic drug withdrawal, and enabled the computation of individualised outcome risks. Our nomograms are validated across various populations and can be applied in all seizure-free patients, both children and adults, for whom antiepileptic drug withdrawal is being considered.

Implications of all the available evidence

The nomograms have the potential to improve patient consultations by providing evidence-based estimates of risk for antiepileptic drug withdrawal. Furthermore, future studies on prognostic factors for the outcome of antiepileptic drug withdrawal should correct for those identified as predictors in this paper.

The dilemma between overtreatment and side-effects of antiepileptic drugs on the one hand, and the risk of seizure recurrence on the other should be considered with every seizure-free patient. However, a robust tool to guide the decision to withdraw antiepileptic drugs is not available. 25 predictors of seizure outcome have been identified, but the published populations, methods, and results were too variable to distil a definitive set of independent predictors.5 Although many studies have focused on predictors of seizure recurrence, only a few have studied factors related to refractory epilepsy.6 A major limitation of prognostic meta-analyses that have used published aggregate data is that effect sizes associated with individual predictors cannot be produced because of different methods and reporting of the original studies. A method to overcome this issue is through a meta-analysis of individual participant data (IPD), in which the original data from previous studies are combined and more accurate, adjusted statistics can be computed for a large dataset.9

In this IPD meta-analysis we aimed to identify independent predictors of seizure recurrence and long-term seizure outcome, and ultimately provide an evidence-based tool using nomograms to predict the short-term and long-term seizure outcomes for individual seizure-free patients who face the decision of whether to withdraw antiepileptic drugs.

Section snippets

Search strategy and selection criteria

To select articles eligible for this study, we did a systematic search of PubMed and Embase on Nov 6, 2014 (with no date restrictions). Inclusion criteria were that they had to be original full-text articles reporting on a cohort of seizure-free patients who started antiepileptic drug withdrawal and containing information regarding seizure recurrences during and after antiepileptic drug withdrawal. We excluded surgical cohorts, reports with fewer than 30 patients, and reports on acute

Results

We identified 45 reports as eligible for inclusion; 33 authors were ultimately contacted and invited to collaborate, of whom ten agreed to participate and provide IPD (appendix p 1). 1771 (25%) of 7082 patients were included in the initial analysis. Many authors provided additional, unpublished details on the cohorts, such as longer follow-up durations. No important issues that could compromise the analysis were identified in checking IPD from contributing cohorts. The cohorts consisted of a

Discussion

This prognostic IPD meta-analysis of the risks of antiepileptic drug withdrawal in 1769 seizure-free people with epilepsy yielded clinically useful nomograms to predict individual seizure outcome. Relapse occurred in 812 (46%) patients, while only 136 (9%) of 1455 in the cohort of patients with available information had seizures in the last year of follow-up. The proportion of relapsing patients who did not regain freedom from seizures decreased with longer follow-up times. The strongest

References (34)

  • HJ Lamberink et al.

    Antiepileptic drug withdrawal in medically and surgically treated patients: a meta-analysis of seizure recurrence and systematic review of its predictors

    Epileptic Disord

    (2015)
  • D Schmidt et al.

    Uncontrolled epilepsy following discontinuation of antiepileptic drugs in seizure-free patients: a review of current clinical experience

    Acta Neurol Scand

    (2005)
  • D Schmidt

    AED discontinuation may be dangerous for seizure-free patients

    J Neural Transm

    (2011)
  • E Beghi

    AED discontinuation may not be dangerous in seizure-free patients

    J Neural Transm

    (2011)
  • RD Riley et al.

    Meta-analysis of individual participant data: rationale, conduct, and reporting

    BMJ

    (2010)
  • JA Hayden et al.

    Annals of internal medicine academia and clinic evaluation of the quality of prognosis studies in systematic reviews

    Ann Intern Med

    (2006)
  • TAMO Cardoso et al.

    Is low antiepileptic drug dose effective in long-term seizure-free patients?

    Arq Neuropsiquiatr

    (2003)
  • Cited by (191)

    View all citing articles on Scopus
    View full text