Key messages
• Abnormal results for a test that records brain activity (an electroencephalogram or EEG) are probably associated with an increased risk of having another seizure.
• Other factors, such as abnormal brain scans, seizures during sleep (nocturnal seizures), Todd’s paresis (weakness after a seizure), or a family history of epilepsy, may also increase the risk, but the evidence is less certain.
• We need higher-quality studies to help doctors more accurately predict who will have further seizures after a first unprovoked seizure.
What are unprovoked seizures?
A seizure is a sudden surge of electrical activity in the brain. It can cause shaking, stiffening of the body, staring, confusion, or a loss of awareness. An unprovoked seizure is a seizure that occurs without an immediate or obvious cause, such as a fever, infection, or recent injury. It may be a sign of an underlying condition like epilepsy.
Why are unprovoked seizures a concern?
Many people will have a single seizure in their lifetime — around 1 in 20 by the age of 85. After this first seizure, doctors often cannot tell who will go on to have more. This uncertainty makes it difficult to make decisions about driving, working, or starting treatment.
Doctors need reliable ways to predict seizure risk after a first unprovoked seizure, to help guide diagnosis and treatment. This is especially important because a diagnosis of epilepsy can now be made after one seizure in some people, if the chance of more seizures is high.
What did we want to find out?
We wanted to find out if certain features — either of the person, their first seizure, or results from medical tests — could help predict whether they would have another seizure. We looked at:
• personal characteristics, such as age, sex, and family history of epilepsy;
• medical tests, including brain scans (magnetic resonance imaging (MRI) or computed tomography (CT)) or EEGs (brain wave tests);
• seizure features, including whether the seizure happened during sleep or lasted a long time.
What did we do?
We searched for good-quality studies that followed people after a first unprovoked seizure to see if they went on to have more. We included studies that followed people for at least six months and included at least 30 participants.
We then combined the results of the studies to find out how strong the evidence was for each possible risk factor.
What did we find?
We found 23 studies involving 5918 people. Some studies included adults, some children, and some both. The strength of the evidence for each risk factor varied.
Main results
• Abnormal electroencephalogram (EEG): people with abnormal EEG results probably have a higher risk of another seizure.
The following factors may increase the risk of seizure recurrence:
• abnormal brain scan (imaging);
• Todd’s paresis (temporary weakness after a seizure);
• family history of epilepsy;
• seizures during sleep (nocturnal seizures).
We are very uncertain if the following factors increase the risk of seizure recurrence:
• focal neurological signs (problems affecting a specific body part, caused by a problem in a specific area of the brain);
• febrile seizures (seizures that occur during a fever) in childhood;
• status epilepticus (when a seizure doesn’t stop on its own, or repeated seizures without regaining full consciousness between them);
• focal seizures (a seizure that starts in one part of the brain);
• male sex.
There is conflicting evidence about whether being younger than 16 is a risk factor: some studies suggest higher risk, others suggest lower risk. Overall, the evidence is uncertain.
What are the limitations of the evidence?
Many studies measured and reported results differently, making them hard to compare. Some were small or did not adjust for all important factors. Most were done in high-income countries, so the findings may not apply everywhere.
Larger, high-quality studies that measure and report on seizure risk factors in the same way would help doctors make better predictions.
How can patients and carers use this information?
Understanding which features may be linked to having more seizures can help patients and families have informed discussions with their doctors. This can support decisions around treatment, driving, school, work, and daily life.
How current is this review?
The review includes studies published up to December 2022.
Read the full abstract
Assessing the risk of seizure recurrence after a first unprovoked seizure remains a clinical challenge but is essential for counselling, especially given its impact on driving, employment, and treatment decisions. The International League Against Epilepsy now allows for an 'operational' diagnosis of epilepsy after a single unprovoked seizure, based on an individual’s recurrence risk. This shift highlights the need for more precise tools to predict seizure risk and guide accurate diagnosis and management.
Objectives
To identify which prognostic factors predict the risk of subsequent unprovoked seizures and the development of epilepsy at any time following a first unprovoked seizure, a cluster of seizures within 24 hours, or a first episode of status epilepticus – regardless of seizure type.
Search strategy
We searched the following databases between 12 and 15 December 2022, with no language restrictions: CENTRAL, MEDLINE, SCOPUS, ClinicalTrials.gov and the World Health Organization (WHO) International Clinical Trials Registry Platform.
Selection criteria
We included retrospective and prospective cohort studies and randomised controlled trials (RCTs) that assessed prognostic factors for seizure recurrence following a first unprovoked seizure. Eligible studies reported on at least one index prognostic factor (including demographic variables such as age and sex, clinical features such as seizure type, and investigation results such as electroencephalogram (EEG) and neuroimaging) and assessed seizure recurrence outcomes. Prognostication began at the time of the initial seizure, with outcomes assessed at a minimum of six months' follow-up.
Data collection and analysis
Two review authors screened titles and abstracts identified through searches and removed irrelevant articles. We extracted data using a data extraction form. We conducted separate meta-analyses to pool odds ratios (ORs), risk ratios (RRs), and hazard ratios (HRs) reported from univariable regression analyses in the included studies.
We conducted meta-analyses using a random-effects, generic inverse-variance model, which accounted for any between-study heterogeneity in the prognostic effect. We summarised the meta-analysis by the pooled estimate (the average prognostic factor effect), its 95% confidence interval (CI), the I² (heterogeneity) estimates, and a 95% prediction interval for the predictive effect in a single population.
Two review authors independently extracted data and assessed risk of bias using the QUality In Prognosis Studies (QUIPS) tool. We adapted the GRADE framework to assess the certainty of evidence for each prognostic factor–outcome association. We rated evidence certainty as high, moderate, low, or very low, based on study phase, internal validity, effect size and precision, heterogeneity, generalisability, and reporting bias.
Main results
We included 23 studies (5918 participants). Cohort sizes varied from 50 to 1885 participants (median 134).
Most studies were cohort designs (15 prospective, seven retrospective), with one RCT. Median follow-up was 35 months (range six to 283 months). Seven studies included participants on anti-seizure medication (ASM) after their first seizure; 16 did not. Eight studies were adult-only, 12 paediatric-only, and three included both age groups. Using the QUIPS tool, nine studies (39%) had a low risk of bias, 11 (48%) unclear, and three (13%) high.
Abnormal EEG (RR 1.90, 95% CI 1.60 to 2.25; P = 0.022, I² = 55.4%; 9 studies, 1904 participants; HR 1.45, 95% CI 1.17 to 1.79; P = 0.018, I² = 75%; 3 studies, 939 participants) probably increases the risk of seizure recurrence (moderate-certainty evidence for RR; low-certainty for HR). Low-certainty evidence suggests abnormal brain imaging (RR 2.19, 95% CI 1.74 to 2.76; P = 0.085, I² = 54.6%; 4 studies, 890 participants), nocturnal seizures (HR 1.41, 95% CI 1.13 to 1.75; P = 0.674, I² = 0%; 3 studies, 967 participants; RR 1.23, 95% CI 1.04 to 1.47; P = 0.017, I² = 70.7%; 4 studies, 1248 participants), family history of epilepsy (RR 1.47, 95% CI 1.16 to 1.85; P = 0.423, I² = 0%; 6 studies, 1290 participants), and Todd’s paresis (RR 1.48, 95% CI 1.02 to 2.13; P = 0.102, I² = 56.2%; 3 studies, 836 participants) may increase seizure recurrence, but findings remain uncertain due to limitations in study quality, consistency, and precision.
Very low-certainty evidence means we are uncertain about associations of febrile seizures (RR 1.02, 95% CI 0.82 to 1.28; P = 0.516, I² = 0%; 3 studies, 841 participants), focal neurological deficit (HR 1.21, 95% CI 0.92 to 1.60; P = 0.028, I² = 67%; 4 studies, 981 participants), status epilepticus (RR 1.05, 95% CI 0.81 to 1.36; P = 0.0507, I² = 56.4%; 5 studies, 1456 participants), male sex (RR 1.14, 95% CI 0.94 to 1.39; P = 0.190, I² = 39.8%; 3 studies, 738 participants), initial focal seizures (OR 1.19, 95% CI 0.77 to 1.85; P = 0.004, I² = 88.1%; 2 studies, 473 participants), and age under 16 years (OR 1.80, 95% CI 1.16 to 2.79; I² = 0%; 5 cohorts, 522 participants; low-certainty evidence but RR 0.69, 95% CI 0.47 to 1.01; I² = 37.9%; 3 studies, 480 participants; very low-certainty evidence) with seizure recurrence, due to significant inconsistency, imprecision, indirectness, and risk of bias across studies.
Authors' conclusions
We aimed to identify prognostic factors predicting seizure recurrence after a first unprovoked seizure. Considerable heterogeneity and inconsistency in how the included studies defined, measured, and reported prognostic factors significantly limited our analyses and led to downgrading of the evidence due to imprecision and methodological variability. These limitations highlight the need for standardisation. Future studies would benefit from adherence to an international core outcome set currently under development, which will standardise reporting and collection of prognostic factor data, enhancing comparability and reliability. Identifying high-risk cohorts remains critical for guiding clinical decisions, shaping healthcare policy, and enabling recruitment into trials of disease-modifying treatments.
