Sudden unexpected death in epilepsy (SUDEP) accounts for approximately 2% of deaths in population-based cohorts of epilepsy, and up to 25% of deaths in cohorts of more severe epilepsy. When it occurs, SUDEP usually follows a generalised tonic-clonic seizure. Unresponsiveness, apnoea, and cardiac arrest occur in SUDEP, rather than the typical gradual recovery. The great majority of tonic-clonic seizures occur without difficulty and how the rare seizure associated with SUDEP differs from others is unknown. Three mechanisms have been proposed for SUDEP: cardiac arrhythmia, neurogenic pulmonary oedema, and postictal suppression of brainstem respiratory centres leading to central apnoea. Recent studies have found that the incidence of SUDEP increases with the severity of epilepsy in the population studied. The duration of epilepsy, number of tonic-clonic seizures, mental retardation, and simultaneous treatment with more than two antiepileptic drugs are independent risk factors for SUDEP. Some studies have reported that carbamazepine use, carbamazepine toxicity, and frequent, rapid changes in carbamazepine levels, may be associated with SUDEP. Other evidence indicates that carbamazepine could potentially increase the risk for SUDEP by causing arrhythmia or by altering cardiac autonomic function. However, this evidence is tenuous and most studies have not found an association between the use of carbamazepine or any other individual antiepileptic drug and SUDEP. There is little information regarding antiepileptic drugs other than phenytoin and carbamazepine. The incidence of SUDEP with gabapentin, tiagabine, and lamotrigine clinical development programmes is in the range found in other populations with refractory epilepsy. This suggests that these individual antiepileptic drugs are no more likely to cause SUDEP than antiepileptic drugs in general. Best current evidence indicates that the risk of SUDEP can be decreased by aggressive treatment of tonic-clonic seizures with as few antiepileptic drugs as necessary to achieve complete control. At present there is no strong reason to avoid any particular antiepileptic drug. Further studies are needed to elucidate the potential role of individual antiepileptic drugs in SUDEP and establish clinical relevance, if any. These studies may be challenging to conduct and interpret because SUDEP is relatively uncommon and large numbers will be necessary to narrow confidence intervals to determine the clinical relevance. Also adjustments will be needed to account for the potent risks associated with other independent factors.